Grammaticalization and Parametric Variation

Grammaticalization and Parametric Variation

This page intentionally left blank

Grammaticalization and Parametric Variation Edited by M O N T S E R R AT BATL LO R I , M A R I A- L LU ¨I SA H E R NA N Z , C A R M E P I C A L LO, A N D F R A N C E S C ROC A

1

3

Great Clarendon Street, Oxford ox2 6dp Oxford University Press is a department of the University of Oxford. It furthers the University’s objective of excellence in research, scholarship, and education by publishing worldwide in Oxford New York Auckland Cape Town Dar es Salaam Hong Kong Karachi Kuala Lumpur Madrid Melbourne Mexico City Nairobi New Delhi Shanghai Taipei Toronto With oYces in Argentina Austria Brazil Chile Czech Republic France Greece Guatemala Hungary Italy Japan Poland Portugal Singapore South Korea Switzerland Thailand Turkey Ukraine Vietnam Oxford is a registered trade mark of Oxford University Press in the UK and in certain other countries Published in the United States by Oxford University Press Inc., New York ß Editorial matter and organization Montserrat Batllori, Maria-Lluı¨sa Hernanz, Carme Picallo, and Francesc Roca 2005 ß The chapters their several authors 2005 The moral rights of the author have been asserted Database right Oxford University Press (maker) First published 2005 All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, without the prior permission in writing of Oxford University Press, or as expressly permitted by law, or under terms agreed with the appropriate reprographics rights organization. Enquiries concerning reproduction outside the scope of the above should be sent to the Rights Department, Oxford University Press, at the address above You must not circulate this book in any other binding or cover and you must impose the same condition on any acquirer British Library Cataloguing in Publication Data Data available Library of Congress Cataloging in Publication Data Data available Typeset by SPI Publisher Services, Pondicherry, India Printed in Great Britain on acid-free paper by Biddles Ltd., King’s Lynn ISBN 0-19-927212-3 978-0-19-927212-9 1 3 5 7 9 10 8 6 4 2

Preface The following chapters are a selection of the papers presented at the Seventh Diachronic Generative Syntax Conference (DIGS 7), held at the University of Girona (UdG), 27–9 June 2002. The organizers of the event gratefully acknowledge the signiWcant Wnancial support of the following institutions for their generous donations or conference grants: Facultat de Lletres de la UdG, Departament de Filologia i FilosoWa de la UdG, Fundacio´ Privada. Girona: Universitat i Futur, Ministerio de Ciencia y Tecnologı´a: Direccio´n General de Investigacio´n (for the BFF2001-4713-E), and above all, Generalitat de Catalunya: DURSI—Direccio´ General de Recerca (for the conference grants corresponding to 2001ACES00022 and 2002ACES00077). Many thanks to all who attended the meeting for their keen participation; thanks are also due to the following colleagues for their valuable practical assistance with refereeing: Artemis Alexiadou, Paola Beninca`, Ignacio Bosque, Josep-Maria Brucart, Josep-Maria Fontana, Eric Haeberli, Liliane Haegeman, Ans van Kemenade, Anthony Kroch, David Lightfoot, Giuseppe Longobardi, Christiane Marchello-Nizia, Ana-Maria Martins, Susan Pintzuk, Marı´a-Luisa Rivero, Ian Roberts, Beatrice Santorini, Ann Taylor, Arhonto Terzi, Juan Uriagereka, Wim van der WurV, Sten Vikner, Nigel Vincent, Marie-The´re`se Vinet, and Anthony Warner. Likewise, we would like to show our gratitude to the four anonymous reviewers selected by OUP for their useful remarks and suggestions and, particularly, to John Davey, our editor at Oxford University Press, for his help and advice. Finally, we thank the contributors for their diligence and dedication during the course of editing and publication. Montserrat Batllori, Maria-Lluı¨sa Hernanz, Carme Picallo, and Francesc Roca Girona and Bellaterra, June 2005

This page intentionally left blank

Contents Notes on contributors Abbreviations 1 Grammaticalization and Parametric Change Montserrat Batllori, Maria-Lluı¨sa Hernanz, Carme Picallo, and Francesc Roca

Part I. Grammaticalization 2 Grammaticalization, Reanalysis, and CP Layering Nicola Munaro

ix xiv 1

27 29

3 Verb Movement and Clitics: Variation and Change in Portuguese Acrisio Pires

48

4 Changes in Clausal Organization and the Position of Clitics in Old French Marie Labelle and Paul Hirschbu¨hler

60

5 Exceptional Case Marking in Brazilian Portuguese Heloisa Maria Moreira Lima Salles

72

6 Reanalysis and Conservancy of Structure in Chinese John Whitman and Waltraud Paul

82

7 DeWnite Reference in Old and Modern French: The Rise and Fall of DP Paul Boucher 8 The Reanalysis of the French Prepositional System: A Case of Grammaticalization in Competing Grammars Mireille Tremblay, Fernande Dupuis, and Monique Dufresne 9 Accusative Alternation in Old and Modern Romance Anna Bartra Kaufmann

95

109 124

Part II. Parametric Variation

147

10 Parametric Comparison and Language Taxonomy Cristina Guardiano and Giuseppe Longobardi

149

viii

Contents

11 Clitic Placement, VP-Ellipsis, and Scrambling in Romance Ana Maria Martins 12 Microvariation in French Negation Markers: A Historical Perspective France Martineau and Marie-The´re`se Vinet 13 Sı` and e as CP Expletives in Old Italian Cecilia Poletto 14 Split wh-Constructions in Classical and Modern Greek: A Diachronic Perspective Eric Mathieu and Ioanna Sitaridou 15 The Syntax of Objects in Old English Susan Pintzuk 16 Clause Type Asymmetries in Old English and the Syntax of Verb Movement Eric Haeberli References Index

175

194 206

236 251

267 284 309

Notes on contributors Anna Bartra Kaufmann is professor in Catalan Philology at the Universitat Auto`noma de Barcelona. Her teaching and research Welds are syntax, lexicon, applied linguistics, and language acquisition and change. She has worked on Catalan weak pronouns and case, and on lexical and functional categories. She has taken part in several European and Spanish research projects on theoretical syntax and on second language acquisition. She has collaborated in the Grama`tica del Catala` Contemporani. She has written several grammar books for secondary school teaching. Montserrat Batllori is professor of Spanish grammar (diachrony) at the University of Girona. She received her Ph.D. in Spanish Philology at the Universitat Auto`noma de Barcelona in 1996. She has published articles on Old Spanish syntax, Old Spanish derivational morphology, sound change— phonetics and phonology—and lexicography. She directs a funded research project on the computerization of Corominas and Pascual’s etymological dictionary. Paul Boucher was born in the United States, where he did a BA in French language and linguistics, before moving to France where he did his doctoral studies in linguistics. He is full professor of linguistics in the UFR de Langues at the Universite´ de Nantes, where he teaches generative syntax, diachronic studies in French and English, and language acquisition. He recently published Many Morphologies (Cascadilla Press) and is preparing a volume entitled Principles of English Syntax for publication in France. Monique Dufresne is an assistant professor in French linguistics at Queen’s University specializing in theoretical and historical linguistics . She completed her Ph.D. at the Universite´ du Que´bec a` Montreal in 1993. Her research interests include syntactic and morphological change in Medieval French, and she has published several papers on clitics, particles, and aspect. She is also an associate researcher at the Centre d’analyse de textes par ordinateur at UQAM. Fernande Dupuis is lecturer in Linguistics at the University of Quebec in Montreal. Her research focuses on the historical development of the morphosyntax of French, with special emphasis on the evolution of word order, aspectual systems, and the use of computer-assisted text analysis in corpus linguistics.

x

Notes on contributors

Cristina Guardiano is a researcher in linguistics at the Universita` di Modena e Reggio Emilia. Having graduated from the University of Pisa in 1999 (studies in classical languages, with special reference to Indoeuropean dialects and historical-comparative linguistics), she got her Ph.D. in linguistics at the University of Pisa in 2003 with a dissertation on the ‘Structure and History of the Nominal Phrase in Ancient Greek: Parametric Hypotheses’. Her current research interests are diachronic and comparative linguistics, parametric and historical syntax, and Greek language. Eric Haeberli is associate professor of English linguistics at the University of Geneva. His main research interests are in the areas of diachronic syntax (in particular the history of English syntax), comparative Germanic syntax, and syntactic theory. His publications include Features, Categories and the Syntax of A-Positions: Cross-Linguistic Variation in the Germanic Languages (Kluwer, 2002) and articles on the syntax of Old and Middle English. Maria-Llu
sa Hernanz is professor of Spanish grammar at the Universitat Auto`noma de Barcelona. She obtained her Ph.D. in Romance Philology at the Universitat de Barcelona in 1978. Her main Weld of research is Spanish syntax. She has published several articles in Nueva revista de Wlologı´a hispa´nica and Verba, among other journals. She has also published two books and has contributed to the Grama´tica descriptiva de la lengua espan˜ola, edited by I. Bosque and V. Demonte (1999). Paul Hirschbhler is professor of linguistics at the Universite´ d’Ottawa. He earned his Ph.D. at the University of Massachusetts in Amherst (1978). His interests include French synchronic and diachronic syntax, comparative syntax, and cross-linguistic diVerences in the expression of arguments. He is currently working on locative alternation from a comparative and diachronic perspective. Marie Labelle is professor of linguistics at the Universite´ du Que´bec a` Montre´al. She received her Ph.D. at the Universite´ d’Ottawa in 1989. Her research interests include French syntax (synchronic and diachronic), psycholinguistics (particularly Wrst and second language acquisition by children), the semantics of tense and aspect, and the mapping of semantic arguments to syntax. Giuseppe Longobardi is currently professor of linguistics at the University of Trieste. He has also served as Visiting Professor at Vienna, Harvard, University of Southern California, University of California at Los Angeles, and as Directeur de Recherche E´tranger at the CNRS in Paris. His research

Notes on contributors

xi

interests include theoretical, comparative, and historical syntax and the study of the syntax/semantics interface. France Martineau is full professor of linguistics at the University of Ottawa, co-supervisor of the Laboratoire de Franc¸ais Ancien (LFA), and supervisor of a project on the evolution of Quebec French and microvariation. She has published papers on historical linguistics and Old and Preclassical French syntax. Her current areas of interest are negation, clitics, complementation, and valency patterns. She has co-edited two books, one on epistolarity and migrant discourse and the other on Old French corpora and the web. She is in charge of a large corpus of informal letters (more than 2 million words) from Quebec, Acadie, and the north-west of France, from the seventeenth to the nineteenth century. Ana Maria Martins is professor of linguistics at the University of Lisbon. Her main area of teaching and research is historical linguistics, in particular historical syntax. She has worked and published on diVerent topics in Romance diachronic syntax: clitic placement, clitic climbing, word order, negation and negative words, and the Portuguese inXected inWnitive. Eric Mathieu is assistant professor at the University of Ottawa. His research focuses on the syntax–semantics interface. He has published in Lingua, ‘The Mapping of Form and Interpretation: The Case of Optional WH Movement in French’ (2004). Joint works include a series of papers on Greek discontinuous elements, with Ioanna Sitaridou, and on bare nominals, with Sophie Heyd (University of Nancy). The results of his ongoing collaboration with Alastair Butler (University of Singapore) have culminated in the publication of a book entitled The Syntax and Semantics of Split Constructions: A Comparative Study (Palgrave-Macmillan, 2004). Nicola Munaro is a researcher in linguistics at the University of Venice. His research interests concern primarily the cross-linguistic variation in the leftperiphery of the functional structure of the sentence, with particular reference to the northern Italian dialects. Some of the issues addressed by his previous production are related to the distribution, the internal structure, and the interpretation of wh-phrases, both from a synchronic and from a diachronic point of view; more recently, his scientiWc activity has been devoted to the study of various syntactic strategies of clausal typing. Waltraud Paul is a researcher at the French Centre National de la Recherche ScientiWque (CNRS) and aYliated to the Centre de Recherches Linguistiques sur l’Asie Orientale (CRLAO), in Paris. Her main interest is the syntax of

xii

Notes on contributors

modern Mandarin, with occasional excursions into the diachronic syntax of Chinese. She has published articles on diVerent topics in Chinese syntax such as verb gapping, complex sentences, left-periphery in CP and IP, and adjectival modiWcation. Carme Picallo is professor in Catalan Philology at the Universitat Auto`noma de Barcelona. She obtained her Ph.D. in linguistics at the City University of New York in 1985. Her research interests are in the Welds of theoretical syntax and morphology. She has published the results of her research in Linguistic Inquiry, Natural Language and Linguistic Theory, Journal of Linguistics, and Syntax, among other journals. Susan Pintzuk is professor of English Language and Linguistics at the University of York. Her main research interests are syntactic variation and change, particularly in the history of English; quantitative models of syntactic change; and the construction of annotated corpora. Recent publications include Diachronic Syntax: Models and Mechanisms, Oxford University Press, 2001 (edited with G. Tsoulas and A. Warner); Phrase Structures in Competition: Variation and Change in Old English Word Order, Garland, 1999; and articles on Old English syntax. She is preparing a book, Old English Verbal Syntax, to appear in the Mouton de Gruyter series Topics in English Linguistics. Acrisio Pires is assistant professor at the Department of Linguistics, University of Michigan, Ann Arbor. He received his Ph.D. at the University of Maryland, College Park. His research focuses on syntactic theory, Minimalism, comparative syntax, and models of language acquisition and change. His work includes the investigation of defective clausal domains (especially gerunds and infinitives) and its implications for syntactic theory; Portuguese syntax; and the evaluation of models of syntactic change. He has been involved in various joint projects focusing on syntactic theory; the methodology of syntactic reconstruction; and the investigation of tense, agreement, and movement in Berber, Korean, and Spanish. Cecilia Poletto is currently associate professor in Linguistics at the University of Venice and affiliated to the Institute of Cognitive Science and Technologies of the CNR. She is mainly interested in the syntax of Romance languages and dialects and is collaborating to the ASIS project (Syntactic Atlas of Northern Italian dialects). She has written a number of articles and two books on Italian dialects both in a syncronic and diachronic perspective on topics including subject clitics, subject positions, the left periphery of the clause, verb second, wh-constructions, sentential particles, and aspectual auxiliaries.

Notes on contributors

xiii

Francesc Roca is professor of Spanish grammar at the Universitat de Girona. He obtained his Ph.D. in Spanish philology at the Universitat Auto`noma de Barcelona in 1997. He has published articles on Spanish syntax and morphology, and on Basque morphology and phonology. His current work focuses on the syntax, semantics, and morphology of nominal structures in Spanish. His general research interests are in the Welds of theoretical syntax and comparative linguistics. Heloisa Maria Moreira Lima Salles is adjoined professor at the Department of Linguistics, University of Brası´lia. Since her Ph.D. dissertation, she has been working in a comparative perspective with Brazilian Portuguese, European Portuguese, and English, focusing on prepositions and complementation. She is presently engaged in the study of (clausal) complementation, investigating the hypothesis of the equivalence between clausal and nominalized complements, by looking at the complementation system of (Brazilian) Indian languages from the Tupı´-Guaranı´ family, in particular the Guaja´ language, as compared to other linguistic families. Ioanna Sitaridou is lecturer in Romance Philology at the University of Cambridge. Her main areas of research are synchronic and diachronic syntax, in particular acquisition, language contact, dialectal variation, and language change. Her doctoral dissertation is entitled ‘The Synchrony and Diachrony of Romance InWnitives with Nominative Subjects’ (2002). During her postdoctoral research at the University of Hamburg she worked on the licensing of null subjects in Old French dialects, and the development of overt subjects in the history of French. Mireille Tremblay received her Ph.D. in linguistics from McGill University in 1991. She is now associate professor of French Linguistics and Coordinator of the Canadian Studies Program at Queen’s University in Kingston, Canada. Her main areas of research and publication are morphology, syntax, and the historical development of the French language. Marie-ThØrŁse Vinet is professor of linguistics at the Universite´ de Sherbrooke. Her previous works include D’un franc¸ais a` l’autre: la syntaxe de la microvariation (Fides, 2001) and La Variation en grammaire universelle (Presses de l’Universite´ de Montre´al, 1989), with co-author Yves Roberge. Her work mainly discusses microcomparative syntax in French and has revealed many novel facts in the grammar of various dialects of French. John Whitman is professor in the Department of Linguistics at Cornell University. He teaches historical linguistics, Japanese and Korean linguistics, and syntax. His publications include (with Koichi Takezawa) Nichieigo hikaku sensyo, vol. ix: Kaku to gojun to toˆgo koˆzoˆ (Kenkyuˆsha, 1998).

Abbreviations A A’-position acc Agr AgrO AgrOP AgrP AgrS AgrSP AP A-position Ar Asp / asp AspP Aux / aux Aux-S inversion BD Blg BN BNP BP Cat C cl ClGr CLLD ColBP CoP CP CSP CVC D D dat DeclP dem det

adjective non-argument position accusative agreement agreement object agreement object phrase agreement phrase agreement subject agreement subject phrase adjective phrase argument position standard Arabic aspect aspect phrase auxiliary auxiliary–subject–inversion bare determiners Modern Bulgarian bare noun bare noun phrase Brazilian Portuguese Catalan complementizer classiWer Classical Greek clitic left dislocation contemporary colloquial Brazilian Portuguese coordination phrase complementizer phrase chronological separation principle consonant–vowel–consonant determiner Modern German dative declarative phrase demonstrative determiner

Abbreviations DP E E ECM E-language EME EP EPP F fem FI Fin FinP FP Fr fut Got Grk HT I I-Wnal I-initial I-language IMP ind inf INFL INFL-Wnal INFL-initial IP It K KP Lat LD LF List Interpr masc MC MF MG MO MoodP

determiner phrase event Modern English exceptional case marking external language Early Middle English European Portuguese Extended Projection Principle (feature) focus feminine principle of full interpretation Wniteness Wniteness phrase focus phrase French future Gothic Modern Greek hanging topic inXection inXection in Wnal position inXection in initial position internal language imperative indicative inWnitive inXection inXection in Wnal position inXection in initial position inXection phrase Italian kase kase phrase Latin left dislocated (position) logical form list interpretation (position) masculine Modern Catalan Modern French Modern Greek Modern Occitan mood phrase

xv

xvi MP MS N neg NegP NIP nom NP NTGr Num NumP obl OC OE OF OI OO OP OS OV P p.c. perf PF pl/pl PLD PossP PP Ptg Q QF QP R refl RQC Rr S SC Sc. Sett. SCL/scl SCr sg/sg SP

Abbreviations Modern Portuguese Modern Spanish noun negation negative phrase new information phrase nominative noun phrase New Testament Koine´ or New Testament Greek number number phrase oblique Old Catalan Old English Old French Old Italian Old Occitan Old Portuguese Old Spanish object–verb (word order) preposition personal communication perfective aspect phonetic form plural primary linguistic data possessive phrase preposition phrase Portuguese quantiWcation, quantiWer Quebec French quantiWer phrase referent reXexive Restricted QuantiWcation Constraint Rhaetoromance South small clause scene setting (position) subject clitic Serbo-Croat singular sigma phrase

Abbreviations SIP Sp Spec sub subj SVO T TM TP UG v V V1 V2 V3 V4 VO VOS vP VP VS inversion VSO W wh-

split IP parameter Spanish speciWer subordinator subjunctive subject–verb–object (word order) tense Tobler-MussaWa tense phrase Universal Grammar verb (functional head) verb verb Wrst verb second verb third verb fourth verb–object (word order) verb–object–subject (word order) verb phrase (functional projection) verb phrase verb–subject inversion verb–subject–object (word order) West interrogative

xvii

This page intentionally left blank

1 Grammaticalization and Parametric Variation M O N TS E R R AT B AT L LO R I, M A R I A- L LUI¨ S A H E R NA N Z , CARME PICALLO, AND FRANCE SC RO CA

1.1 Introduction The title of this volume, Grammaticalization and Parametric Variation, is intended to capture the continuation of a tradition in the Weld of historical linguistics as well as the theoretical assumptions that have guided the study of grammatical variation and change in recent decades under the generative framework. As is well known, the term grammaticalization was used in Meillet (1912) to refer to processes of apparent category shifting and semantic weakening that some lexical items may undergo in language evolution and change. Elements that had entered into thematic relations partially lose features and gradually become functional-like elements such as clitics or inXectional aYxes.1 The phenomenon, however, has been diVerently understood and assessed, depending on the view adopted with respect to grammar and language. Under a functionalist perspective, where language is viewed in terms of its use, grammaticalization is considered a semantically pragmatic driven process (see Hopper and Traugott 1993). Linguistic change is, then, conceived as being mainly determined by non-linguistic cognitive faculties, and it is suggested that the properties of grammatical constructs can be inXuenced by social factors. These correlate with morphosyntactic and (sometimes) phonological changes in the particular grammars. Under this line of thought, phenomena related to language evolution and grammaticalization are understood as somewhat external to the linguistic endowment of the mind-brain.

1 Lehmann (1995, 2002) and Moreno Cabrera (1998) oVer a full description concerning the role of grammaticalization within language evolution.

M. Batllori, M.-L. Hernanz, C. Picallo, F. Roca

2

From the point of view of generative grammar, the grammatical properties of the utterances provided by historical material serve to characterize the system internalized by the speaker. The perspective that generative studies oVer on this issue is that of assuming that universal grammatical principles can be expressed within certain limited options for variation. The language faculty is viewed as a (relatively) autonomous system, which can be examined independently of other cognitive capacities. The nature of the cognitive linguistic capacity does not lie in its use, although the language capacity is, of course, what underlies communication and the expression of thought. The main objections that functionalists put down to this formal approach is that it only deals with reanalysis (a particular aspect of grammaticalization) and that it does not provide an explanation for graduality, the claim being that the generative perspective can only capture abrupt changes. Generative work has, however, faced some of these objections. Whitman (2000) proposes a minimalist typology of syntactic change where reanalysis (i.e. signiWcant structure changes) is claimed to be, in fact, only one of the mechanisms involved. The emphasis is placed on how and why grammars (i.e. these internal and autonomous systems) can change in the process of acquisition by new generations of individuals. Grammaticalization is not treated as a linguistic phenomenon related to language-external factors but as a regular case of parameter change (see Roberts and Roussou 2003). This theoretical approach is assumed in many of the chapters contained in this volume. The opposition between graduality and abruptness has been solved by appealing to the concepts of I(nternal)-language, where change can only be abrupt, and E(xternal)-language, where change follows a gradually changing path (see, for instance, Andersen 1973 and Lightfoot 1979, 1991, 1999, 2002a, 2002b).2 The double-base hypothesis proposed in Kroch (1989, 1994, 2000) oVers an account for variation at the same period of time and within individual speakers, both diachronically and synchronically. The hypothesis proposes that speakers may acquire more than one grammar. These grammars, in competition, can give rise to cases of internal diglossia explaining variation within the same period and texts. Following a similar line, Lightfoot (2002a) distinguishes grammatical change from grammar change. The Wrst is a consequence of contingent factors of language use, whereas the second is directly concerned with the basis of language acquisition. Thus, it can be said that grammatical changes are related to E-language and grammar changes 2

The concepts of I-language versus E-language were introduced in Chomsky (1986). The Wrst refers to a component of the mind-brain of the native speaker, whereas the latter understands language as independent of the internal properties of individuals.

Grammaticalization and Parametric Variation

3

aVect I-language. Both aspects are taken to be two sides of the phenomenon of language evolution. The questions that both language acquisition and language change from the internalist perspective pose as well as the theoretical and methodological assumptions that linguists adopt to face them, will brieXy be explored in this introduction. Section 1.2 presents some basic concepts that are directly relevant for setting the theoretical frame of the papers included in the volume. Our aim is brieXy to discuss some conceptual issues to a reader who may not be particularly familiar either with the Principles and Parameters framework, or with some recent developments of it. Section 1.3 is devoted to characterize the properties of the so-called functional categories, given the central role that these syntactic entities have in this theoretical framework, as well as their relation to grammaticalization processes. In section 1.4, the notion of grammaticalization is considered in light of the issues introduced in the preceding discussion. Finally, an overview of the chapters contained in this volume is given in section 1.5. To facilitate the presentation, this last section has been divided in two parts: in the Wrst one, we outline the studies dealing with grammaticalization phenomena; in the second section, the studies devoted speciWcally to several cases of parametric variation induced by a variety of grammatical factors are overviewed.

1.2 The Theoretical Framework Almost since its inception, generative grammar has aimed to accomplish the maximum theoretical restrictiveness in order to obtain a principled and explanatory account for the diversity of linguistic phenomena. As research was advancing, this goal was constantly being challenged by the need to introduce too many descriptive devices to account accurately for the available data. An approach to resolve the challenge to explanatory adequacy was oVered by the Principles and Parameters hypothesis. 1.2.1 The theory of Principles and Parameters

This theory, Wrst formulated in the late 1970s, puts forward the proposal of conceiving Universal Grammar (UG) as a system of principles and parameters. A principle is a condition or a restriction that all grammars obey without exception, independently of typological variation. A parameter is the set of limited options in which a principle of grammar can be expressed. For example, a principle of grammar requires that there must be a one-to-one correspondence between the thematic roles encoded in the lexical entry of a predicate and the arguments related to it (AGENT, THEME,

M. Batllori, M.-L. Hernanz, C. Picallo, F. Roca

4

EXPERIENCER, etc.). A variety of factors allow, however, a variation in the linear sequence in which arguments and predicate can appear within a sentence. In some languages, like the Catalan example (1a) or its English equivalent, the object with the role of THEME follows the verb; whereas in others, like in the Basque (1b), the object precedes it: (1) a. La dona ha llegit el llibre the woman has read the book b. Emakumeak liburua irakurri du woman-the book-the read has The parameter can be expressed in terms of the linearization of verb-object structures. As the glosses show, a similar parametric variation can also be observed in the above examples with respect to the order in which Determiner and Noun appear. Another much discussed cross-linguistic diVerence that was set in terms of parameter variability is that some languages, but not others, allow subjects of tensed sentences to be phonologically null. The phenomenon was known as the null subject (pro-drop) parameter. It is exempliWed by the contrast between the Spanish sentence (2a) and its French equivalent (2b): (2)

a. Ha leı´do estos libros has read these books b. Il/elle a lu ces livres (cf. *A lu ces livres) he/she has read these books

The formulation of this parameter triggered a lot of theoretical interest, since it was claimed to correlate with a number of other syntactic phenomena such as: the impossibility of having phonologically realized expletive subjects of the types it or there, the possibility of having postverbal subjects, and the apparent violation of conditions related to the distribution of certain empty categories. The latter phenomenon was discussed under the observation that a subject in many null subject languages could be extracted from a position immediately following a conjunction (i.e. the that-trace Wlter).3 In Romance languages, the correlation between the null subject property and the possibility of having clitic climbing out of an inWnitive clause has also been much discussed (see Rizzi 1982, Burzio 1986, or Kayne 1989). The correlations suggested that a single abstract grammatical property could account for a 3

See, among many others, Perlmutter (1971), Taraldsen (1978), Chomsky (1981), Rizzi (1982, 1986). The grammaticality of sentences like quien crees que salio´? ‘lit.: who do you think that left’ in Spanish constitute apparent violations of the that-trace Wlter.

Grammaticalization and Parametric Variation

5

complex array of phenomena that had previously been considered unrelated (see, for extensive discussion, Jaeggli and SaWr 1989, and all references cited there). The study of this parametric diVerence propitiated, moreover, a lot of research focused on the study and characterization of the so-called empty categories (i.e. phonetically null but syntactically present entities). The Principles and Parameters model was initially designed to explain synchronic diVerences among languages, but it proved to be a powerful theoretical tool to account for diVerent stages of language acquisition. Empty categories posed the question of how the child can interpret them, and deduce their syntactic presence, if they are absent from the speech signal. Empty categories have also been brought to the fore in the context of diachronic change. Thus, unlike Modern French, twelfth- and thirteenthcentury Old French could have null subjects in certain environments. This was also the case of existential sentences in Early Middle English and Old English (see Adams 1987 for French and Williams 2000 for English). This fact triggers the obvious question of which factors could be at work to explain the loss of the pro-drop property. The last twenty years has seen substantial developments in the Principles and Parameters model. The emphasis has gradually been shifting from the typology and characterization of empty categories and the study of movement operations with the abstract conditions applying to them, to a more parsimonious view of the syntactic objects that constitute the building blocks of sentential structures. 1.2.2 The Minimalist Programme

Earlier formulations of the Principles and Parameters framework assumed that syntactic movement can be applied freely, provided that principles of grammar are satisWed in a particular derivation. As research was advancing, it was observed that strong economy conditions appear to be at work to restrict the application of grammatical operations. The locality conditions on movement were observed to obey a ‘least eVort’ requirement as well as being a ‘last resort’ option, applying only in those cases where no other grammatical mechanism is possible. ‘Least eVort’ and ‘last resort’ are at work in the contrast observed between the construction (3a) and their ungrammatical counterparts (3b) and (3c), which show that the argument a Xy cannot be raised to the subject position of the inWnitive clause or cross-over expletive there: (3) a. There seems to be a Xy in the soup b. *There seems a Xy to be in the soup c. *A Xy seems there to be in the soup

6

M. Batllori, M.-L. Hernanz, C. Picallo, F. Roca

The more recent development of the Principles and Parameters model (the Minimalist Programme; see Chomsky 1995 and subsequent work) emphasizes the pivotal role of economy conditions in evaluating all the possible derivations of a given construction. Basically, economy consists of saving grammatical and computational resources as much as possible. When an expletive like there is selected to appear in a construction, it must be included in the computation (i.e. united or merged with another syntactic object) as early as possible, since the Merge operation is argued to be less costly than Move. Conditions of economy on cyclic operations account for the ungrammaticality of (3b) and (3c). In (3a) the selected expletive is Wrst merged in the lower cycle (the subject position of the inWnitive subordinate clause) and then moved to the main subject position, since Merge has priority over Move. In (3b) this computational priority (Merge over Move) has not been satisWed. In (3c) the argument has violated locality, crossing over the expletive that is in argument position: (4) a. Therei seems [ti to be a Xy in the soup] b. *There seems [a Xyj to be tj in the soup] c. *A Xyj seems [there to be tj in the soup]

(cf. (3a)) (cf. (3b)) (cf. (3c))

In addition to these constraints, a principle on grammatical constructs (Full Interpretation - FI) states that every element at the interface levels PF and LF (Phonetic Form and Logical Form, respectively) must receive an appropriate interpretation. The principle requires eliminating all the syntactic material that could have intervened in a syntactic computation but has no eVect on the ‘borders’ of the linguistic system, the LF or the PF levels that connect with the other non-linguistic systems. An element that has no eVect at LF is, for example, an expletive element like there in the example (3a) above, since it occupies a syntactic position (that of formal subject) but has no thematic role, because it is not an argument of predication. There is phonetically realized but it is non-interpretable at LF. Its Catalan, Italian, or Spanish counterparts (known as expletive pro) are also syntactically active, but phonologically null. Thus, in addition to its noninterpretability at LF, a null expletive subject is of course non-interpretable at PF.4 As said, LF and PF are the levels of representation that interact with the other systems of the mind or the body. The thesis that the Minimalist Programme puts forth is that grammar applies speciWc mechanisms 4 The Spanish equivalent to (3a) is parece haber una mosca en la sopa ‘lit.: seems to have a Xy in the soup’.

Grammaticalization and Parametric Variation

7

(i.e. computational operations) in order to be able coherently to link to the conceptual/intentional and the sensory/motor systems. In other words, the language faculty has to adapt in the best possible way to the conditions and restrictions imposed by external factors, in order to be usable for human interaction. This fact raises the question of the degree of ‘optimality’ that the linguistic system must meet. It allows one to consider the hypothesis of whether or not grammatical operations are, perhaps, the best possible adaptation to the restrictions imposed by the systems external to the linguistic cognitive module. Under this perspective, theoretical linguistics intends to characterize the minimal speciWcations that must be satisWed to Wt the language faculty with non-linguistic systems. Making explicit those speciWcations would presumably improve the explanatory capacity of the model and account for the fact that computational complexity must also be minimal for the grammar to be learnable. 1.2.3 Properties of features

Lexical items consist of bundles of features that are chosen, for each language, from a universal pool of features. These grammatical units combine to form ‘major’ categories such as nouns and verbs—as well as adjectives in many languages—together with the functional elements that constitute their extensions (see section 1.3). The Minimalist Programme framework has focused closely on the properties of features and explicitly adopts very strict notions of economy to drive the relations between grammatical objects in terms of them. The characterization of feature properties (i.e. whether they are interpretable or not) together with strict economy have allowed further uniWcation of all instances of movement (no longer an optional operation as in the early PP model) by assuming that Move only applies for the purposes of checking (and deleting/erasing) the non-interpretable features of syntactic objects. Thus, the checking procedure constitutes the motor of movement. One of the non-interpretable features that may trigger movement is the so-called EPP feature, which is present in at least one of the functional projections that constitute InXection (i.e. the functional extensions of VP).5 The EPP feature can either be checked and erased by raising an argument from a thematic position, as in the Spanish sentence (5a), or by merging with INFL an expletive element (overt or null, depending on the 5 The label EPP to refer to a feature is probably a misnomer, but it is being used for convenience. The acronym stands for Extended Projection Principle. It was initially formulated as a wellformedness condition that requires clauses to have a formal subject, either an argument or an expletive (see Chomsky 1981 and Rothstein 1983).

8

M. Batllori, M.-L. Hernanz, C. Picallo, F. Roca

language) that was previously selected from the lexicon to form a derivation, as in (5b):6 (5)

a. [IP Tu padrei [EPP][VP ha llegado ti ]] your father has arrived b. [IP pro EPP][VP ha llegado tu padre]] has arrived your father

Formal speciWcation of person (gender) and number on the verbs in tensed sentences (in agreement with the argumental subject) are also considered non-interpretable. These morphemes do not add any semantic information that is not already provided by the interpretable person (gender) and number features of the DP with which the verb agrees. Such features of the verb are also checked and erased by mechanisms taking place at the syntactic component. Although it is not LF interpretable, verbal inXection is spelled at PF in the Romance languages. The procedure is represented in the following Spanish sentence:

(6)

Han have[3PL]

florecido bloomed

muchos rosales many rosebushes-3PL

AGREEMENT Non-interpretable features in arguments are structural case and grammatical gender speciWcation in determiner phrases (DPs). The following examples in Catalan show that any combination of structural case (nominative or accusative) and gender (masculine or feminine) in DPs can be associated with any thematic role, a fact that suggests that case and gender are not present at the LF component and are erased at some point in the derivation: (7)

a. La cadira esta` trencada the chair-nom.fem is broken-fem b. Ha arreglat el sofa` have-3sg Wxed the sofa-acc.masc ‘S/he has Wxed the sofa.’ c. Aquest quadre ha guanyat el primer premi this painting-nom.masc have-3sg won the Wrst prizeacc.masc

6 The constructions in (5) exemplify sentences with an unaccusative verb, a predicate that cannot assign accusative case to its internal argument. The subject of these verbs is syntactically akin to the direct object of transitive verbs (see, among many others, Burzio 1986 and references cited therein). The VP structure of a sentence with an unaccusative verb is represented in (i):

(i)

[IP . . . [VP Vunaccusative DP]]

Grammaticalization and Parametric Variation

9

DiVerences within grammars (i.e. diVerent parameter settings) are conceived as stemming from the properties of the features that constitute the lexical items, the building blocks of phrases. Some grammatical changes are thought to be encoded in the properties of the formal features of the functional categories (see Borer 1984), to which we now turn.

1.3 Functional Categories Functional categories are grammatical objects that can express morphosyntactic or semantic information. They constitute the syntactic extension of the thematically related major categories (nouns, verbs, adjectives, and some prepositions) and serve to connect in various ways the syntactic relations that take place within a sentential structure. Functional categories consist of bunches of features, some of which are non-interpretable (like the EPP feature alluded to above, for example). Some other features are interpretable and, although they do not encode thematic relations, they may contribute to the interpretation of a sentence in several aspects. The articulated representation of a sentence contains three types of structural layers, VP, IP, and CP. The VP level encodes thematic information and is the locus where the assignment of thematic roles takes place. The inXectional level (IP) contains the functional elements corresponding to the inXectional morphological markers related to V. The complementizer level (CP) encodes the illocutionary force of the sentence. Like IP, the CP layer hosts a variety of categories, ranging from topics to operator-like elements, as we will see. The hierarchical organization of these three basic layers is represented in (8): (8) [CP . . . [IP . . . [VP . . . ]]] The elements contained in the CP and IP layers are not theta related but play a crucial role as they interact with lexical categories in a double way: they can bear morphological information related to lexical elements and they provide landing sites for movement rules. The representation (8) above is a rough schematized approach that cannot capture the vast array of syntactic phenomena taking place in the sentential structure. A more detailed look at functional elements follows in the next subsections. 1.3.1 The functional projections of nouns and the IP-VP layers

We will Wrst brieXy characterize the types of minimal functional categories that can be assumed to constitute the immediate extensions of thematically related categories such as nouns and verbs. The determiner (D) constitutes

M. Batllori, M.-L. Hernanz, C. Picallo, F. Roca

10

the extended projection of NPs. With the label D we can refer to deWnite and indeWnite articles, demonstratives, and quantiWers. This category can confer on a nominal construction the status of argument and can contribute to its interpretation by expressing some properties like (non-)speciWcity. Its semantic contribution can also be that of restricting the range of referents denoted by the nominal with which it occurs. Determiners can be overt or non-overt.7 The distribution of empty or phonologically realized determiners depends on several syntactic factors and on the types of nouns with which they co-occur. Examples of overt determiners are shown in (9a). Null determiners are realized in the bare NP object constructions like (9b): (9) a. I have seen [DP the=a=every[NP unicorn]] b. I want [DP D[NP water=Xowers]] Let us mention in this context, without elaborating the discussion in any way, that pronouns have been considered an expression of the category D in many languages. This fact is particularly clear when considering the morphology of articles and pronouns in the Western Romance Languages, which all involve the reanalysis and several degrees of grammaticalization of the Latin demonstratives ille–illa–illud (see Batllori and Roca 2000, and all references cited there). InXection (INFL) is also a cover term that serves a diversity of functional categories above VP. Pollock’s (1989) analysis of the position of the verb in English and in French led to the conclusion that an articulated structure of INFL had to be postulated. It has been proposed that INFL can split into Agreement nodes (AgrP), hosting non-interpretable agreement features related to the subject and the object (person, gender, and number) with an intervening Tense Phrase (TP), which expresses the eventive structure of the sentence. Pollock’s inXuential proposal has given rise to further developments in the functional domain immediately dominating the VP. The category Aspect (Asp) has been proposed to express semantic notions such as ongoing or completed states of aVairs, whereas verb (v) encodes the notion of causality. The hierarchy of these INFL projections is represented in (10): (10)

[CP C[AgrP AgrS[TP T[AspP Asp[AgrP AgrO[(subject)v[V(object)]]]]]]]

Other functional projections have been postulated besides the ones just mentioned. The expression of (negative) polarity is instantiated by means of a separate node NegP (see Pollock 1989 and Zanuttini 1997) or SP (Laka 1990). 7

See, however, the alternative put forward by Boucher’s chapter in this volume.

Grammaticalization and Parametric Variation

11

In addition, it has been suggested that a MoodP should also be postulated to encode mood (Rivero and Terzi 1995, Pollock 1997), and that diVerent classes of adverbs Wll the speciWer positions of distinct maximal projections, as proposed in Cinque (1999) (see Haegeman 1997a for a more detailed account of these issues). Several chapters in this volume discuss particular properties of the above mentioned functional categories for a variety of diachronic stages of several languages. For reasons of space, our purpose here is that of oVering only a very general background for their proposals. The next section is devoted to discussing the general properties and the distribution of the functional projections above INFL. 1.3.2 The CP layer

Rizzi (1997) proposes that the complementizer layer must be decomposed the same way as the IP layer, in order to capture a variety of phenomena in the left periphery of the sentence. He argues that the complementizer system, which minimally consists of a speciWcation of force besides a speciWcation of (non-) Wniteness for IP, may also consist of a Topic and a Focus, Weld, expressing the topic-comment and focus-presupposition articulation, respectively. The CP node splits into four separate projections, as represented in (11): (11) [ForceP F [TopicP Top [FocusP Foc [FinitenessP Fin . . . ]]]] The CP system may be seen as the interface between two kinds of information, one facing ‘outside’ the sentence and the other oriented to the lower sentential structure. The external boundary of the CP Weld instantiates the illocutionary value or the force of the sentence (declarative, interrogative, exclamative, relative, etc.). Thus, English complementizers like that and whether clearly determine the type of clause (declarative and interrogative, respectively) that they introduce, as well as its articulation to a higher clause. The downward boundary encodes the properties of the embedded IP, namely its Wniteness. This analysis captures the traditional assumption that complementizers are sensitive to certain characteristics of the verb, a fact that has usually been formalized by postulating agreement between C and I: in English, that selects Wnite clauses, whereas for takes inWnitival clauses. Between these two boundaries, Rizzi proposes that two nodes related to the informational structure may project Topic and Focus, as shown in (11).8 8

In the absence of Topic and Focus, the force-Wniteness system can be expressed on a single head. When the topic-focus system is activated, Force and Finiteness must be represented as two distinct projections. See Rizzi (1997, 2001) for further discussion.

12

M. Batllori, M.-L. Hernanz, C. Picallo, F. Roca

As opposed to the force-Wniteness system, Topic and Focus are independent from selectional constraints. The Topic projection (TopP) is activated to host topicalized constituents, expressing old information. The Focus projection (FocP) is the locus of focalized elements, which introduce new information. They are clearly distinguished in Italian and in other Romance languages. Topicalization is expressed by the construction usually labelled Clitic Left Dislocation (CLLD), as in (12a). Focalization is obtained by preposing the focal constituent, which receives contrastive intonation, as in (12b). Rizzi (1997) assumes that topicalized and focalized elements are located in the speciWer position of the functional projections TopP and FocP, respectively. The following Italian examples are from Rizzi (1997: 286): (12) a. Il tuo libro, lo ho letto the your book it have-1sg read ‘Your book, I have read it.’ b. il tuo libro ho letto (non il suo) the your book have-1sg read not the his ‘your book I have read (not his).’ The proposal sketched in (11) accounts for a vast array of empirical data, both diachronic and synchronic, whose analysis suggests that the CP layer is not as simple as previously assumed. For example, it is widely known that dialectal, colloquial, and standard varieties in a number of languages provide evidence showing that the speciWer and the CP head can be simultaneously realized, which appears to be violating the so-called ‘Doubly-Wlled Comp’ Wlter. The constructions include both a wh-phrase and a subordination particle, the former targeting the Spec,CP position, and the latter located in the head of CP, as in the following examples: (13) Ou` que tu vas? where that you go ‘Where are you going?’ (Radford 1988: 501) ! que es Julia! (14) Que´ guapa what beautiful that is Julia ‘How beautiful Julia is!’

(Colloquial French)

(Standard Spanish)9

9 In contrast with Colloquial French (among other languages), interrogatives with ‘DoublyWlled-COMP’ are banned in Spanish:

(i)

¿Que´ (*que) ha dicho Julia? what that have-3sg said Julia ‘What has Julia said?’

Grammaticalization and Parametric Variation

13

It has also been noted by many authors that the classical analysis for the CP level as consisting of a single functional projection is inadequate to account for some cases of preposing. For example, English negative inversion shows movement of both the preposed negative constituent and the auxiliary, as illustrated in (15): (15)

Lee said that at no time would she agree to visit Robin

Haegeman (2000: 47) notes that in (15) two constituents appear between the complementizer that and the speciWer of IP: a maximal projection (the preposed negative constituent at no time) and a head (the auxiliary would). Assuming that complementizer that is base generated in Force0 , it is clear that the landing site for the moved elements in (15) must be a lower projection, i.e. Focus Phrase. More precisely, the preposed negative constituent targets the Spec,FocP position, and the auxiliary moves to Foc0 : (16) Lee said[ForceP that [FocP at no time [Foc would[IP she agree to visit Robin]]]] Finally, a theory involving a single CP projection cannot account for the distribution of a variety of elements that have usually been located in the left periphery of the sentence: (i) interrogative (and exclamative) phrases, (ii) relative pronouns, (iii) topicalized elements, (iv) focalized elements, etc. As is well known, most of these elements have been argued to occupy the speciWer position of a functional projection. Furthermore, there is evidence that many of them obey strict ordering constraints, rather than being in complementary distribution. In Italian, and in other Romance Languages, CLLD normally precedes focalized elements, as in (17).10 Analogously, interrogative phrases must follow topics in main questions, as shown by the contrasts in (18):11 10

The example (17a) shows that CLLD may involve various constituents, a possibility that is banned for focused elements (see Rizzi 1997: 290). On the basis of this asymmetry and further distributional evidence, the representation in (11) may be reformulated as in (i): (i) ForceP (TopicP) (FocusP) (TopicP) FinP IP See Rizzi (1997: 289 V.) for a detailed discussion of some diVerences between Topic and Focus. 11 The ill-formedness of the examples in (17b) and (18b) seems to suggest that both focused elements and interrogative phrases behave alike, due to their quantiWcational nature. Additional support for this claim comes from the fact that they are mutually incompatible: (i) a. *A chi il premio nobel dovrebbero dare? to whom the prize Nobel should-3pl give b. *il premio nobel a chi dovrebbero dare? the prize Nobel to whom should-3pl give (Rizzi 1997: 298)

14

M. Batllori, M.-L. Hernanz, C. Picallo, F. Roca

(17) a. A Giorgio, un posticino, il dottor piva puo` Top, Foc ‘To Giorgio, a job, doctor piva can trovarglielo! Wnd it to him!’ b. *il dottor piva, a Giorgio, un posticino puo` trovarglielo! *Foc, Top (Beninca` 2001: 46) (18)

a. Il premio Nobel, a chi lo daranno? ‘The Nobel prize, to whom will they give it?’ b. *A chi, il premio Nobel, lo daranno? (Rizzi 1997: 289)

Top, Int *Int, Top

With respect to the distribution of relative operators and topics, the former have to precede the latter in Italian, as well as in other Romance languages, as shown in (19) below. A similar situation arises with the Wnite complementizers (that in English, che in Italian, que in Spanish and Catalan) encoding the force value of a sentence, as in (20). The examples are from Rizzi (1997: 289): (19)

(20)

a. Un uomo, a cui, il premio Nobel, lo daranno senz’ ‘A man, to whom, the Nobel prize, they will give it altro undoubtedly.’ b. *Un uomo, il premio Nobel, a cui lo daranno senz’altro a. Credo che il tuo libro, loro lo apprezzerebbero molto ‘I believe that your book, they would appreciate it a lot’ b. *Credo, il tuo libro, che loro lo apprezzerebbero molto (Rizzi 1997: 288)

Rel, Top

*Top, Rel che, Top *Top, che

The data provide empirical evidence that both relative operators and the Wnite complementizer occupy a higher position in Force, while elements located in the Topic/Focus Weld appear in a lower position. Rizzi’s (1997) seminal work has triggered a vast amount of research. Some of the chapters in this volume rely on Rizzi’s split-CP proposal and suggest further extensions of it (see the chapters by Martineau and Vinet, Poletto, Labelle and Hirschbu¨hler, and Munaro in this volume). It is worth emphasizing that a number of issues in this domain have received a lot of attention

Assuming that interrogative phrases target Spec,Focus in main questions, the ungrammaticality of the examples in (i) can be traced back to the assumption that the interrogative phrase a chi and the focused constituent il premio Nobel compete for the same position and hence cannot co-occur.

Grammaticalization and Parametric Variation

15

since the late 1990s. They are concerned with the nature of topics,12 the status of interrogative complementizers,13 and the functional representation of the sentential force,14 among others. Some questions remain open and diVerent reWnements or modiWcations of the structure represented in (11) have been proposed in recent literature.

1.4 Syntactic Change and the Minimalist Programme The gradual shift from the properties of categories to the properties of features from the early Principles and Parameters model to the Minimalist Programme framework that we have been discussing has aVected the characterization of what may constitute a possible parameter. Although in the earlier stages of the model it was already understood that each and every observed diVerence between one language and another could not be stated in terms of a parametric change, there was no formal characterization of the notion of parameter. A parameter was conceived as an independent property of a language that could trigger a cascade of consequences. The properties with respect to which languages vary were supposed to organize themselves in clusters, allowing researchers to propose a typology of languages with respect to the co-occurrences of phenomena associated with a given parameter. Within the set of general assumptions that we have summarized in the previous sections, parameters cannot be conceived the same way in the Minimalist Programme model. In this framework, language change has to be accounted for (at least in part) by taking into consideration the properties of individual features. Moreover, how non-interpretable units locally aVect, or are aVected by, the feature content of other syntactic objects through a checking procedure must be explained. Diachronic change can then be viewed as small diVerences in how features, with given values, relate to each other. Assuming economy guidelines and focusing the domain of inquiry on smaller and less observable grammatical entities require more Wne-grained analyses of syntactic constructs. 12

See Beninca` (2001), who argues that, in addition to TopP, a higher position (DiscourseP) for ‘hanging topics’ (HT) must be postulated. See also Rizzi (2002) for a detailed discussion of interpretive and distributional diVerences between topics and preposed adverbs. 13 As for interrogative complementizers, Rizzi (2001) suggests that in Italian and other Romance languages, interrogative se ‘if ’ is located in a lower position (IntP) than declarative che ‘that’. 14 This issue is addressed in Zanuttini and Portner (2002). These authors claim that the label illocutionary force, as it is based on the speaker’s intention, must be kept separate from the notion of clausal type, which refers to syntactic information such as the fact that a sentence is a declarative, a question, an exclamative, a relative, etc. This gives rise to the question of whether sentence typing and the illocutionary value of an utterance can be instantiated by diVerent speciWc functional projections in the CP domain.

16

M. Batllori, M.-L. Hernanz, C. Picallo, F. Roca

1.4.1 Learnability and variation within the generative framework

Within the theoretical viewpoint we have just sketched, the always puzzling question of why grammars should keep evolving instead of remaining invariant takes a renewed interest and can be considered from diVerent perspectives. Change occurs even in the absence of social or historical events able to interfere abruptly in the normal transmission of input data. Bearing in mind that change takes place as each new language learner constructs the I-grammar, the process could be sketched as follows (from Andersen 1973):

(21) Grammar I

output I

Grammar II

(I-Language)

output II

(E-Language)

Given the relevance attributed to the language environment, linguists have focused on the nature of the input in the acquisition process. Two diVerent hypotheses have been built to explain the process of learnability: the ‘overall input matching’ hypothesis and the ‘partial matching’ hypothesis. The Wrst one is highly UG driven and assumes that the learner seeks to match the input in full and all the potential values of UG for each parameter. Then, the diachronic change is triggered when the input is regarded as inconclusive by the learner (see Clark and Roberts 1993). This amounts to saying that change in grammars is endogenous. The ‘partial matching’ hypothesis considers that only a partial structural analysis of sentences is relevant for the setting of parameters. Lightfoot (1991, 1999) has pointed out that there is a tension between the facts (i.e. language evolution) and models of acquisition/change that assume that language learners analyse sets of sentences as input, to further proceed to select the grammars successfully matching such input. These models predict that grammars should be stable under normal conditions and do not successfully explain diachronic language change. Lightfoot suggests that the only two aspects that learners pay attention to are root clauses and, following the proposals put forth by Dresher (1999), claims that the trigger for variability across time cannot be sets of sentences. The triggers are partially analysed structures that have undergone slight changes and which act as cues that the learner derives from the input. A working methodology along these lines is being applied to the study of dialectal/linguistic variation (see the introductory remarks in Kayne 2000a). This issue is speciWcally addressed in Guardiano and Longobardi’s chapter in the present volume.

Grammaticalization and Parametric Variation

17

1.4.2 Grammaticalization and parametric change

The theoretical constraints we have been discussing in previous sections delimit the set of hypotheses one can reasonably formulate within the Minimalist Programme framework. The cascade of consequences that a single modiWcation in the properties of features can trigger can be more diYcult to assess, and even more reduced in scope than previously thought. Given that the framework allows us to formulate questions concerning the most basic items of syntactic variation, the notion of possible parameter and our understanding of issues concerning acquisition/learnability and evolution of the syntax of languages are being continuously re-evaluated. Roberts and Roussou (1999, 2002, 2003) concentrate on several questions raised by syntactic change in the context of the Minimalist Programme. They suggest that postulating the proliferation of features that have no syntacticexternal existence (i.e. non-interpretable features) should be avoided, since it can lead to redundancies and circularity in the system and can aVect the accuracy of the characterization of grammatical processes. These authors attribute variation among languages to the way functional categories are phonetically realized cross-linguistically; that is, languages vary with respect to the visible exponents of functional categories. Their work is specially relevant because it simpliWes the speciWcation of features, since, according to them, cross-linguistic variation results from diVerences regarding the features that have PF interpretation. This should reduce parametric variation to the presence at PF of a feature F* (a strong feature because it is phonologically realized). They describe the lexicon as containing the elements listed in (22): (22)

a. SpeciWed lexical categories [
N,
V], with their corresponding PF and LF properties. b. Substantive universal sets of features of functional heads. c. The diacritic *, assigned to features of functional heads in diVerent ways depending on the language.

Roberts and Roussou consider that PF realization is achieved either by movement (Move), by lexical insertion (Merge), or by lexical insertion and movement (Merge and Move), the favoured procedure being the most economical one (Merge preferred to Move).15 Their proposal is innovative 15 They introduce a notion of markedness into the parametric system so as to solve the conXict between a descriptively adequate analysis of grammaticalization and an explanatory adequate account of syntactic change as parametric change. The markedness of parameter values is deWned in terms of a simplicity metric which enables the language learner to consider Merge less marked than Move (see Roberts and Roussou 2003: 201).

18

M. Batllori, M.-L. Hernanz, C. Picallo, F. Roca

because it eliminates the circularity caused by having two coexisting primitives: feature checking and movement. Moreover, the model they put forth oVers a comprehensive approach to grammaticalization, which is seen as a regular case of parameter change involving structural simpliWcation (i.e. loss of movement) and a reanalysis of a lexical head as a functional head.16 Grammatizalization is deWned as the consequence of an upward reanalysis to avoid feature syncretism and to favour structural simpliWcation. Reanalysis generates a categorial change in a subset of linguistic items that share certain properties and undergo semantic bleaching and phonological reduction. The procedure gives rise to a new exponent for a higher functional head. According to them, grammaticalization is a natural form of endogenous change and is consistent with the fact that it is the most common parameter setting mechanism. In view of this approach, parametric change will take place whenever language learners converge on a type of parameter setting which diVers from the one adopted by adult grammars. Hence the phenomenon of grammaticalization highlights the relationship between parameter setting and syntactic change.

1.5 An Overview of this Volume 1.5.1 Grammaticalization

This part contains eight chapters devoted to the study of a series of linguistic facts concerned with grammaticalization processes. Munaro (Chapter 2) focuses on the grammaticalization of cosa/cossa in Italian. Originally, this lexical item was a noun, but it gradually became a wh-element that ended up in sentence initial position (as the speciWer of a functional projection within the CP domain) where wh-phrases usually appear. He proposes that cossa was a head N that moved up to a head Q within the extended projection of NP and adds that the change that took place would consist of the reanalysis of cossa as the head Q (i.e. cossa is directly merged as the head Q). This grammatical change can be seen as an instance of ‘F*Move > F*Merge’ consistent with Roberts and Roussou’s view of grammaticalization and syntactic change. In addition, the analysis and the development put forward take into account the connection with the wh-phrase che cossa, the uses of interrogative che and cosa/cossa in several dialects and in successive stages of Italian, and the properties of the functional elements that 16 They argue that only a change from FMove to FMerge gives rise to grammaticalization and that not all cases of loss of movement are instances of grammaticalization. See Roberts and Roussou (2003: 208) for a discussion related to loss of movement and downward reanalysis in English verbs, which cannot be considered a case of grammaticalization.

Grammaticalization and Parametric Variation

19

head the clause structure. In this sense, it is also noticed that in present stages of Italian, the use of interrogative cossa is linked to pragmatic factors (like speaker’s attitude). This suggests that cossa evolved from a nominal use to a ‘wh-interrogative’ element, and from this to a ‘wh-pragmatic’ element. Pires (Chapter 3) studies several patterns of verb movement and clitic placement in Portuguese (especially in Brazilian Portuguese), and how verb movement is crucial for the Wnal placement of clitics. He claims that verb raising to a functional head F higher than IP (Uriagereka 1995’s analysis) was lost in Brazilian Portuguese at the beginning of the twentieth century. This change, which can be formalized by considering that the head F became inactive in Brazilian Portuguese, is supported by the loss of the possibility of ‘verb-subject’ inversion (V-to-C movement) in cases where it had been possible during the eighteenth and nineteenth centuries, and by the increment of proclisis. The triggers for the acquisition of a grammar with no verb movement to F/C would be the fact that a high percentage of declarative sentences up to the eighteenth century displayed the order S-V because the subject also moved to a position higher than IP, and the fact that clitics appear optionally as enclitics or proclitics to the main verb. With these sentences on the language learner would not be provided with strong enough evidence and the sequence could be interpreted without movement of the verb. In other words, primary linguistic data do not contain clear evidence for verb movement to F (or to a functional head higher that IP) and this absence triggers the acquisition of grammars without verbal movement. This approach relies on the view of UG, language acquisition, and parameters in terms of cues put forward by Lightfoot (1991, 1999) and leaves aside the role that Roberts and Roussou (1999) attribute to markedness as a general tendency to structural simpliWcation, but it seems consistent with Roberts and Roussou’s (2003) claim that marked parametric values need robust cues. Labelle and Hirschbu¨hler (Chapter 4) examine the position of the verb and pronominal clitics from Early Old French to twelfth-century Old French and propose that the change follows from the grammaticalization of illocutionary force features within the CP layer of the clause. According to them, in the early stages of the language the relevant feature that triggers syntactic order in the clause is [þD] (discursive), but in later stages the features that acquire an active syntactic role are [þDECL] (for declarative sentences), [þWH] (in interrogative yes/no sentences), and [þIMP] (in imperative sentences). These features are checked, respectively, at T, in Fin, which is reanalysed as Focus, and in the highest functional position, which probably corresponds to Rizzi (1997)’s Force. The empirical facts discussed cover V2 phenomena and the

20

M. Batllori, M.-L. Hernanz, C. Picallo, F. Roca

Tobler-MussaWa generalization concerning the position of clitics. The V1 or V2 orders in Early Old French are attributed to the need to satisfy a Topic/ Focus feature in the highest functional projection. Checking of this feature can be achieved through V movement to its head, resulting in a V1 clause with a postverbal clitic, if it is the case, or through an XP moved to or merged in its speciWer, yielding a V2 clause with, if it is the case, a preverbal clitic. From a strictly theoretical point of view, it is important to note that the grammaticalization process put forward diVers from Roberts and Roussou’s (2003) approach in the fact that the grammaticalized element is not any lexical element but a (illocutionary) feature associated to a speciWc functional head. Salles (Chapter 5) studies completive subordination patterns in Brazilian Portuguese compared with European Portuguese and with Archaic and Classical Portuguese. The general issue addressed is the way in which modality is expressed in Brazilian Portuguese. This leads to the study of three particular points that are specially taken into account: the loss of indicative/subjunctive morphology in verbs, the loss of ‘Aux-subject’ inversion conWgurations, and the reanalysis of para ‘for’ as a complementizer. Salles considers that modality can be realized through merging of a lexical item in C or through movement to C. The Wrst option corresponds to inWnitive constructions introduced by the preposition para, which is reanalysed as a functional head C that encodes irrealis modality. Then, inWnitival clauses with para would have substituted subjunctive subordinate clauses as a consequence of the fact that in Brazilian Portuguese verbs do not express morphologically the distinction subjunctive/indicative. The second option is found in constructions with ‘Aux-subject’ inversion, which involve T-to-C movement and are in complementary distribution with ‘para þ inWnitive’ clauses. The approach outlined constitutes another piece of evidence for the idea that a change in the feature speciWcation (and overt realization) of a functional head is the trigger for the syntactic change (or for syntactic diVerences between two grammatical systems). The proposal is also in accordance with the idea that merge of any particle is preferred to movement of any element (cf. Roberts and Roussou 2003). The grammaticalization of para as C instead of T-to-C movement would exemplify it. Whitman and Paul (Chapter 6) oVer an analysis for the evolution of the particle ba in Chinese and argue against the idea that Chinese languages have moved from a basic order OV to VO and, then, back to OV. The authors show that the evolution of the particle ba (originally a verb meaning ‘take, seize’) can not be seen as an instance of relabelling of this verbal form as a preposition, a kind of diachronic change widely attested in Chinese, and that, instead, it should be considered a case of grammaticalization into a functional head. The change can be described as the step from a lexical V ba that can

Grammaticalization and Parametric Variation

21

assign a thematic role to its object, to a functional element ba directly merged above VP and below a higher vP. The relation between this functional head and a lexical head V can be seen as the typical head movement process in which V moves to a higher functional head. Clearly, this approach preserves the order OV within VP in every stage of the language, and it is fully consistent with the claims that syntactic change preserves structural relations (Whitman 2000) and that movement ‘up-to-the-tree’ is related to grammaticalization and to the possibility of having direct merge of a grammaticalized element. Boucher (Chapter 7) addresses the problem of the emergence of D heads in Romance and the interpretation and syntactic distribution of bare noun phrases. One of the main claims of the chapter is that there is no need to postulate the existence of a DP when no overt determiner is present. This statement applies to languages that lack overt determiners, like Turkish, Russian, Classical Latin, or Old German, and to bare NP constructions in languages with overt determiners like English or Romance languages. In order to provide an explanation for the existence of referential bare NPs and for the presence of these NPs in argument positions, Boucher modiWes DelWtto and Schroten’s (1991) Restricted QuantiWcation Condition and confers a central role to number and case morphology. In this sense, it is considered that a strong speciWcation of the head Num would do the work usually attributed to empty Ds. With respect to the expression of deWniteness, it is proposed that it follows from the interaction of three functional projections: KP, NumP, and DP. Under this view, the appearance of overt determiners is a consequence of the erosion of case and number morphology and consists of the grammaticalization of functional heads that lose their original values: the loss of the deictic feature in the case of the deWnite article, and the loss of the strong quantiWcational feature in the case of the indeWnite. Tremblay, Dupuis, and Dufresne (Chapter 8) compare Old and Modern French, explore the changes in transitivity that took place in certain prepositions, and analyse the split of the older prepositional system into two subsystems. They propose that both the changes in transitivity and the splitting of the system follow from one single diVerence between the two stages of the language: the loss of verbal prepositional particles. The Modern French distinction between simple prepositions (dans), which appear with overt objects, and morphologically complex prepositions (dedans), which take an implicit object, did not exist in Old French. The main diVerence between them at this period was that only simple prepositions were used as intransitive particles attached to verbs (issir hors ‘to go out/to exit’). The loss of the productive particle verbal system (at the end of Middle French) led to no diVerence at all between the two types of prepositions and, as a consequence,

22

M. Batllori, M.-L. Hernanz, C. Picallo, F. Roca

complex prepositions were reanalysed as non-derived prepositions and a system with competing lexical doublets was created. Later on, these doublets were reanalysed as two subsystems of prepositions diVering in transitivity, in accordance with a general ban on doublets (Kroch 1994). The authors also point out that the grammaticalization process aVected the ability of complex prepositions to assign case and that it involved structural simpliWcation (Roberts and Roussou 2003). Bartra (Chapter 9) investigates the evolution of the argument structure of (psychological) transfer verbs from Latin to Romance. She proposes that the original Latin structure, a construction labelled as accusative alternation in parallelism with the locative alternation or the dative alternation, consists of a small clause selected by the verb and headed by a Wnal coincidence abstract preposition that relates two NP arguments. The Latin abstract preposition would be replaced by a lexical preposition in Romance as a consequence of the loss of case noun morphology and the extension of the use of prepositions. This change would yield in Old Romance two alternating structures that semantically diVer in the SOURCE or GOAL meaning of the preposition. This semantic diVerence becomes crucial to the development of the argument structure in Modern Romance. If the preposition means SOURCE and the aspectual value of the predicate is [þtelic], the preposition gets grammaticalized as a verbal preWx and the simple transitive structure is obtained. However, if the meaning of the preposition is GOAL, the two arguments must be expressed overtly and the result is a ditransitive structure. This work takes into account general aspects of the evolution from Latin to Romance, such as the loss of the case system and the increasing use of the preposition system, and oVers a comparative picture of some Romance languages (Catalan, Spanish, French, Occitan) and several diachronic stages with respect to verbal argumental structure. 1.5.2 Parametric variation

The seven chapters of this section focus on many aspects of parametric variation including references not only to variation and phylogenetic relatedness among languages, but also to microvariation within the same language. Guardiano and Longobardi (Chapter 10) provide the reader with a research model that applies historical considerations—such as phylogenetic taxonomy, grammatical typology, and population genetics—to abstract cognitive structures characteristic of parametric linguistics (parameters and parameter values). This model allows us to look into the identiWcation of phylogenetic relatedness among languages and the deWnition of language families in greater

Grammaticalization and Parametric Variation

23

depth than ever before. After a systematic investigation of the facts, they supply data concerning the phylogenetic relation and the chance probability of diVerent relationships among 105 pairs of Indoeuropean languages. A crucial step in this research strategy is the choice of the parameters to be used in the comparisons. The authors consider that the best option is trying to be exhaustive in a limited subdomain well deWned within syntactic theory itself (for example, the Parameter Grid for nominal syntax). From this standpoint they can undertake a qualitative evaluation of the diVerent classes of parameters according to diVerent degrees of historical stability: Deep or Genetic Parameters (which are most often stable through time and deeply rooted in a language family); Areal Parameters (whose values tend to be borrowed from surrounding languages); and Shallow Parameters (the ones readiest to historical resetting). Martins (Chapter 11) puts forward an analysis that derives synchronic and diachronic variation with respect to clitic placement, VP ellipsis, and scrambling in Romance from the features of the functional heads S and AgrS (that is, the interplay between the ‘strength’ property of S and the EPP of AgrS). She shows that minimal answers to yes/no questions consisting of a bare verb can be attested in languages where the polarity encoding functional head S has strong features, as is the case of Latin, Portuguese, Galician, and several Old Romance languages. Verbal answers to yes/no questions are taken as a property of languages where strong S induces merging of this head with the V head. Enclisis is seen as the outcome of the merger between V and S in declarative aYrmative main clauses, where C and Neg do not project, and VP ellipsis is licensed by the Agree relation between strong S and V. The author also studies the correlations between clitic placement, IP scrambling, and interpolation structures in Old Romance. She gives a minimalist analysis according to which the nature of the EPP feature (which can attract one or more features), the subsequent possibility of projecting multiple speciWers, and the strong speciWcation of S become crucial to capturing the diVerences among old and modern stages of several languages and between a language like Modern Spanish, which does not allow enclisis with Wnite verbs, and Modern Portuguese, where this kind of enclisis is possible. Martineau and Vinet (Chapter 12) argue that, from a diachronic perspective, the realization of French negation has never been a homogeneous phenomenon and it can be studied as a case of microvariation. Along these lines, the presence or the absence of the particle ne follows from semantic and syntactic factors. From this standpoint, they undertake the discussion of certain facts connected to negation markers and to the left periphery of the sentence. They notice that interrogative clauses seem to display alternation between the presence of ne and its absence. In

24

M. Batllori, M.-L. Hernanz, C. Picallo, F. Roca

direct interrogatives and, more rarely, in indirect interrogatives of Old and Middle French and also Classical French, pas (or point) is commonly used with a presuppositional reading without ne. Furthermore, the comparison with Quebec French provides clear evidence for the claim that interrogative questions and presuppositional interrogatives must be overtly checked in a special position of the left periphery. The presence of two diVerent patterns in the use of either pas alone or ne . . . pas in yes/no interrogatives with discursive eVects is attributed to the existence of microvariation systems within yes/no questions. The regularity observed suggests that the source of this microvariation is due to the semantic features involved and to a type of clause bearing discursive eVects, such as presuppositional question forms and expressive structures. Poletto (Chapter 13) explores the process of grammaticalization of two elements related to CP in Old Italian: the adverb sı`, which is absent in Modern Italian, and the conjunction e. These elements are analysed, respectively, as an expletive merged in the speciWer position of Focus and as a topic marker. She oVers a detailed comparison between Old Italian and Modern Rhaetoromance concerning the syntactic behaviour of left periphery elements and examines several V2, V3, and V4 structures. This leads to one of the most relevant points in the chapter: the correlation established between the V2 phenomenon present in several languages and the existence of a fully articulated CP structure along the lines of Rizzi (1997). Such a correlation gives an answer to the disappearance of the expletive particle sı` once Italian turned into a non-V2 language, and serves to illustrate how reanalysis of an element as a functional marker (Focus or Topic) follows from the fact that it meets the appropriate structural conditions. Matthieu and Sitaridou (Chapter 14) oVer a comparison between Classical Greek, where split wh-constructions are generally allowed, and Modern Greek, where wh-splitting is impossible with the exception of genitive interrogative elements. The parametric variation between the two stages of the language is attributed to the bundle of f-features present in the lexical item to be extracted and to its ‘determiner’ or ‘non-determiner’ status. The authors point out that in Modern Greek elements that show rich agreement and are not determiners (adjectives, quantiWers, and demonstratives in their analysis), can be extracted without need to pied-pipe the nominal. The crucial change from Classical to Modern Greek in this respect involved reanalysis of the whelement as a determiner (seen as relabelling from modiWer to determiner in terms of Whitman 2000), a process that is also consistent with the general view that determiners evolve from adjective-like elements. This approach is in

Grammaticalization and Parametric Variation

25

accordance with Minimalist Programme claims according to which parametric variation can be reduced to the feature speciWcation of lexical items. The availability of split constructions is also related to the possibility of having null head modiWers (Greek and Romance languages versus English) and to the co-occurrence of demonstratives with the deWnite article. This enriches the comparison between Old and Modern Greek with parallel constructions in Romance languages and in English. Pintzuk (Chapter 15) addresses the question whether Old English should be considered as a head initial or a head Wnal language. She studies several contexts where there is clear variation in the linear order of verbs and their objects, paying special attention to the type of clause (main, conjoined, subordinate), the positive, negative, or quantiWed nature of the object, and Wnite or non-Wnite verbs. The conclusion is that the only way to account for all the variation exhibited concerning word order is to consider that there is grammatical competition between head Wnal and head initial structures both in the IP and in the VP. Consequently, this work is claimed to constitute strong evidence against the idea that Old English can be analysed as a head initial language (with subsequent leftward movement of several constituents), as proposed by Roberts (1997) and Nunes (2002). Haeberli (Chapter 16) oVers an account for Old English clause type asymmetries that relies on the presence of a feature subject to parametric variation in Fin. The author provides us with a quantitative description of the asymmetries concerning the position of Wnite verbs in main clauses (generally V2) and in subordinate and conjoined clauses (generally verb Wnal). He argues that in Old English main clauses the Wnite verb moves higher than in subordinate clauses. The trigger for this movement is the presence of a feature in Fin that has to be checked by V (through movement to Agr, in terms of Bobaljik and Thra´insson’s 1998 analysis). In subordinate clauses, Fin has a diVerent status because it hosts the complementizer. Hence, the verb does not have to check any feature in Fin and moves only to T. The frequent verb Wnal word orders in subordinate clauses result from the frequent use of head Wnal TP (cf. Pintzuk’s 1999 double-based hypothesis according to which TP can be head initial or head Wnal). Furthermore, Old English conjoined main clauses usually exhibit subordinate clause word order because they are conjoined at the AgrP level and the Wnite verb moves to Agr in the Wrst conjunct while it remains in T in the second conjunct. The fact that verb Wnal is not as frequent in conjoined main clauses as in subordinate clauses is explained in terms of variation concerning the categorial status of the conjoined constituents (AgrP/CP variation). This proposal has consequences with respect to the loss of verb movement in English.

This page intentionally left blank

Part I Grammaticalization

This page intentionally left blank

2 Grammaticalization, Reanalysis, and CP Layering NICOLA MUNARO

2.1 Introduction The aim of this chapter is to show that grammaticalization phenomena can function as the trigger of a process of reanalysis leading to a change in a given derivational path. The relevant empirical evidence is provided by the development of the distributional properties of the two wh-items che and cossa in some northern Veneto dialects, insofar as it is detectable diachronically from the sixteenth century up to the twentieth century. The wh-item cossa, which had initially an exclusively nominal use, has gradually taken up, in the course of the last centuries, the function of interrogative operator; this process is strictly connected to the change in the distributional properties of the whitem che: while in the first stages che occupied the initial position of a main wh-question, it appears in sentence internal position in modern dialects. The chapter is organized as follows: in Section 2.2 I summarize the diachronic development of the two wh-items che and cossa; in Section 2.3 I discuss some issues concerning the wh-items parche´ and parcossa; finally, Section 2.4 contains a few concluding remarks.

2.2 The Diachronic Development of Che and Cossa 2.2.1

The sixteenth century

During the whole of the sixteenth century, che functions as a wh-item appearing in sentence initial position, while cossa has an exclusively nominal use, as exemplified in (1) and (2) respectively: This chapter summarizes and updates issues addressed in previous work of mine (Munaro 1997, 1998, 2001). I thank P. Beninca`, G. Giusti, H. Obenauer, and anonymous reviewers for their comments on various aspects of the analysis, as well as the audience at DIGS VII (Girona, 27–29 June 2002) where this work was presented. The usual disclaimers apply.

30

N. Munaro

(1) a. Che a`-tu cata`? what have-scl found? ‘What have you found?’ b. Mo che faron-e? E de che viveron-e? now what do-fut-scl? And of what live-fut-scl? ‘What shall we do now? And of what shall we live?’ (Cavassico, Rime, first half of the sixteenth century) (2) a. . . . se ogni cossa e` per al miei! . . . if every thing is for the best! ‘. . . if everything is at its best!’ b. Al me recres a dirte cotal cossa scl me regrets to tell-you such thing ‘I regret to tell you such a thing’ c. Fazon un puoc de festa de tal cossa make a bit of feast of such thing ‘Let’s make merry for such a thing’ (ibid.) The complex wh-phrase che cossa resulting from the combination of the two elements was used mainly in predicative contexts:1 1 This resembles what happened in the early stages of standard Italian, namely in thirteenth-century Florentine, where the compound form che cosa was found exclusively in embedded contexts, while in main clauses only the form che was attested:

(i) a. Di’, che vai tu cercando? ‘Say, what are you looking for?’ (B. Latini, Il Tesoretto, 1274: 2473) b. Che fecero i tre cavalieri? ‘What did the three knights do?’ (Novellino, 1281–1300, 41: 15) c. Che ve ne pare? ‘What do you think about that?’ (Novellino, 1281–1300, 86: 10) d. Che e` cio`, segnore, che mi parli con tanta oscuritade? What is this, sir, that you speak to me with so much obscurity? ‘How come you speak . . .’ (Dante, Vita Nuova, 1292–3, 12: 16) (ii) a. . . . si conviene che’l parliere [ . . . ] diffinisca e faccia conto in brevi parole che cosa e` sacrilegio . . . ‘. . . it is convenient that the speaker [ . . . ] defines and reports concisely what is sacrilege . . .’ (B. Latini, La Rettorica, 1260–1: 96, 21–23) b. Tullio dimostra in questa picciola parte del testo che cosa e` appellata preghiera in questa arte ‘Tully demonstrates in this small part of the text what is called prayer in this art’ (B. Latini, La Rettorica, 1260–1: 112: 4–5) c. Vedi tu, figliuolo, che cosa il parente al parente e l’amico all’amico e` tenuto di fare . . . ? ‘Can you see, lad, what the relative for the relative and the friend for the friend is bound to do . . . ?’ (B. Giamboni, Il trattato di virtu` e di vizi, 1292, xiv. 4)

Grammaticalization and CP Layering

31

(3) a. Che cossa sea a servir a donzelle per amor, a me cost what thing be-subj to serve to maidens for love, at my cost [ . . . ] e´ impara` . . . [ . . . ] have learned . . . ‘I have learned at my expense [ . . . ] what it means to serve maidens for love . . .’ b. Per la me fe`, no i sa che cossa e` amar . . . for the my faith, not scl know what thing is to love . . . ‘Indeed, they don’t know what it means to love . . .’ (ibid.) Based on the possibility for che to be preceded by the definite article in contemporary standard Italian, in (4), in Munaro (1997) I analysed che as a quantificational element heading a QP located in the specifier of one of the functional projections of the extended nominal phrase: (4)

a. Gianni non e` venuto, il che mi ha stupito ‘John hasn’t come, which has struck me’ b. Quella persona ha un che di familiare ‘That person has something familiar’ c. Quel film non e` un gran che ‘That film is nothing special’

(5) a. [DP [QP[Q8 che]] [D8 ] . . . [FP tQP [F8] [NP [N8]]]] b. [DP [D8 ] . . . [FP [F8] [NP [N8cossa]]]] c. [DP [QP[Q8 che]] [D8 ] . . . [FP tQP [F8] [NP [N8 cossa]]]] Hence, the wh-items che, cossa, and che cossa in the examples (1)–(3) would be assigned the structural representation reported in (5a–c).2 2.2.2

The eighteenth century

Two centuries later, namely in the eighteenth century, che is still attested as a wh-item in sentence initial position, both in monoclausal and in biclausal structures:3 2 Following Kayne (1994), I assume that—at least in this initial stage—the wh-item che raises to Spec,DP, that is, to the highest functional specifier position inside the extended projection containing it. 3 Interestingly, at this stage the following case of doubling of the wh-item, which appears both in initial position and in sentence internal position, is also attested:

(i) Che ole´-u che e`ppia metu` che? what want-scl that have-subj put what? ‘What do you want me to have put ¼ I haven’t put anything’ (Villabruna, Fioretta, first half of the eighteenth century)

32

N. Munaro

(6) a. Che me dize-u? what me tell-scl? ‘What do you tell me?’ (Villabruna, La grandonazza, 1748) b. E tu che ghe respondee-tu? and you what him answered-scl? ‘And what did you answer to him?’ (Villabruna, Le passion de la Senta, 1755) c. Che ole´-u che seppia, mı`? what want-scl that know-subj, I? ‘I do not know anything’ (Villabruna, Fioretta, first half of the eighteenth century) The evolution from the first to the second stage is essentially triggered by the fact that the generic noun cossa undergoes a process of semantic bleaching, as a consequence of which the original lexical meaning is lost but can provide the variable restriction of the new wh-operator cossa, which raises to some specifier position of the left periphery.4 So, besides the nominal use of cossa, in (7), its use as a wh-item starts being attested, as in (8)—occasionally associated to a widening of its semantic value, as in (9): (7)

a. Tre cosse al n’ ha insegna` . . . three things scl-us-has taught . . . ‘Three things he taught us . . .’ (Coraulo, El filo`, 1780) b. Par altro in trope cosse . . . for else in too many things . . . ‘Moreover in too many things . . .’ (ibid.)

(8) a. Cossa mai dize-u? what ever say-scl? ‘What the hell are you saying?’ (Villabruna, Antonio, 1755)

4 Hopper and Traugott (1993) emphasize the gradualness of the loss of meaning induced by the process of grammaticalization and the property of persistence—the effect that the bleached form still reveals some traces of its original lexical meaning. They also point out that grammaticalization involves loss of morphosyntactic properties that would identify that form as a member of the original grammatical category (typically noun or verb). Indeed, when used as a wh-item, cossa does not inflect for number and cannot be preceded by any modifier, as witnessed by the contrast between (7) and (8)–(9).

Grammaticalization and CP Layering

33

b. Cossa dis-lo quel slibraz? what says-scl that damned book? ‘What does that damned book say?’ (Villabruna, Fioretta, first half of the eighteenth century) (9)

a. Cossa ave-u nome? what have-scl name? ‘What’s your name?’ (Villabruna, Fioretta, first half of the eighteenth century) b. Cossa ole´-u a tor qua` entre la me tosa? what want-scl to take here inside the my girl? ‘How much do you want to host my daughter here?’ (ibid.)

Roberts and Roussou (1999) propose that grammaticalization involves the diachronic development of lexical material into functional material;5 according to them, when an indefinite element is systematically found in a wh-dependency, it is reanalysed by the language learner as associated with the wh-feature of the dependency, that is, as a wh-word; this instantiates a case of F*move (where F¼wh), where the wh-indefinite moves and the highest position spelled out in the wh-dependency identifies the wh-feature of the dependency. A similar process could be at work in the cases we are considering here; more specifically, we can surmise that, once it acquires a wh-feature, cossa must raise by head movement to a DP external Q8 position in order to check it: (10) [QP [Q8 cossax ][DP [D8 tx ] . . . [FP[F8 tx ]...[NP [N8 tx ]]]] The checking of the wh-feature in Q8 could trigger in turn the raising of the whole QP containing cossa to a sentence initial operator position.6 2.2.3.

1800–1880

In the course of the nineteenth century cossa drops almost completely the nominal use and gradually substitutes che as a wh-item: 5 They discuss the development of wh-phrases from Classical Greek to Modern Greek: in Classical Greek wh-questions were formed by moving an indefinite to clause-initial position—like in (ia), while in Modern Greek indefinites have been reanalysed into a distinct class of wh-words by being associated with a wh-dependency—as in (ib):

(i) a. Tı` gar an tis kai poioi, o Sokrates? what prt.prt. one and do-3sg, voc. Socrates ‘What could one do, Socrates?’ b. Ti tha apandouses? what will answered-2sg ‘What would you answer?’ (La. 184d; Plato, Republic) 6

On the process of noun raising inside the DP projection and on the idea that a DP can be selected by a Q8 head, see Longobardi (1994) and Giusti (1993), respectively.

34 (11)

N. Munaro . . . la-me-ha comanda` tante cosse, ma faro` tut . . . scl-me-has ordered many things, but do-fut all ‘. . . she has ordered me many things, but I will do everything’ (Zuccagni-Orlandini, Raccolta di dialetti italiani, 1864)

(12) a. De tant sal che vos-to fa? of so much salt what want-scl do? ‘What do you want to do with so much salt?’ (Follador, Dove vasto, first half of the nineteenth century) b. Martin me, che fe-u? Martin my, what do-scl? ‘My dear Martin, what are you doing?’ (Pagani, Sonetti, around 1850) c. Che ole´-u che i se`pe i siori? what want-scl that scl-know-subj the rich ‘The rich do not know anything’ (Nazari, Parallelo fra il dialetto . . . , 1873) (13) a. . . . cossa me vien-tu a dir de la campagna? . . . what me come-scl to tell about the country? ‘. . . what news do you bring me about the country?’ (Zuppani, Raccolta di poesie, first half of the nineteenth century) b. Cossa sara`-lo mai lassu` tel Ziel? what be-fut-scl ever over there in the Sky? ‘What will there ever be over there in the sky?’ (Pagani, Sonetti, around 1850) c. E cossa te ha-lo conta`? and what you has-scl told? ‘And what has he told you?’ (Zuccagni-Orlandini, Raccolta di dialetti italiani, 1864) Interestingly, in the second half of the eighteenth century the existence of a doubling structure with sentence initial cossa and sentence internal che is attested: (14)

Notevole e` anche certa formula interrogativa che si usa nel dialetto. Finisce colla voce che, all’opposto precisamente di quello che richiede la lingua, nella quale col che si principia: Cosa vutu che? Cosa atu che?

Grammaticalization and CP Layering

35

‘Remarkable is an interrogative formula used in the dialect. It ends with the form che, precisely the opposite of what the (standard) language requires, in which che introduces the sentence: Cosa want-scl che? ¼ ‘What do you want?’ Cosa have-scl che? ¼ ‘What do you have?’ (Nazari, Parallelo fra il dialetto bellunese rustico e la lingua italiana, 1873: 63) This structure witnesses that the landing site of cossa is different, and higher, than the position hosting che, with the inflected verb raising to an intermediate position between the two wh-elements. I propose that che and cossa, the two subconstituents of the complex wh-phrase che cossa (surfacing again after three centuries), are attracted in two successive derivational steps to two different specifiers of the CP layer, as represented in (15a)—cf. Munaro (1998): (15)

a.

CP

DP cossa

C
C[F2] fatu

CP

QP che

C
C˚[F1] tfatu

b.

......

CP

DP cossa

C
C[F3]

CP C
C[F2] fatu

CP C


QP che C[F1] tfatu

......

So, structures with sentence initial cossa are probably analysed as involving the raising of cossa to a higher target landing site (possibly through an intermediate specifier position) to achieve checking of some feature F2

36

N. Munaro

which che cannot (or does not have to) check. We will take into account (15b) in section 2.2.5 below. 2.2.4

1880–1930

Between the nineteenth and the twentieth century the use of cossa as a whitem spreads further, while only some residual cases of sentence initial che are attested: (16)

a. . . . cosa deventara-tu ti, amor mio? . . . what become-fut-scl you, love my? ‘. . . what are you going to become, my darling?’ (Dalle Mule, Poesie, end of the nineteenth century) b. . . . cossa fara-lo mai ‘n zitadin . . . ?! . . . what do-fut-scl ever a citizen . . . ?! ‘. . . what shall a citizen ever do . . . ?!’ (Segato, ‘Una novella svizzera’, 1902) c. Cosa ghe a`-tu dat do? what him have-scl given down? ‘What have you given down to him?’ (Battisti, Testi dialettali italiani, 1914)

(17) a. Ste pore bestie a noi che fa-le mai? these poor beasts to us what do-scl ever? ‘How can these poor beasts ever harm us?’ (Lazzaris, Le stagioni dell’anno, 1931) b. Ma che vole-o, bisogna compatı` . . . but what want-scl, needs to pity . . . ‘But, you know, it’s necessary to pity . . .’ (ibid.) Moreover, a new structure with che in sentence internal position and no wh-operator in sentence initial position is also occasionally attested: (18)

a. . . . ghe pare-lo che? . . . you seems-scl what? ‘. . . what do you think?’ (Chiarelli, Poesie, beginning of the twentieth century) b. Vole-u che, pare? want-scl what, father? ‘What do you want, father?’ (Battisti, Testi dialettali italiani, 1914)

Grammaticalization and CP Layering

37

c. Fa-tu che lasu`? do-scl what, over there? ‘What are you doing, over there?’ (ibid.) After cossa has paved the way in activating the relevant projection, the checking of [F2] can plausibly be performed either by raising cossa to the relevant specifier, or by raising the inflected verb to the corresponding head position, as in (15a).7 2.2.5

From 1930 to the present

In the second half of the twentieth century, sentence initial che almost disappears, while sentence internal che spreads considerably and prevails over sentence initial cossa (which takes over a pragmatically marked function, as we will see in Section 2.3): (19) Che fon-e poš ade`ss de tuta š sta carne? what do-scl then now of all this meat? ‘What shall we do with all this meat?’ (Luciani, Poesie, second half of the twentieth century) (20) Cossa ze´rche-tu, bocia . . . what look-for-scl, boy . . . ‘What are you looking for, boy . . .’ (A filo` ko l ‘zˇenpedon’, 1985) In this fifth stage both the inflected verb and the past participle can precede the wh-item che, as shown by (21c): (21)

a. . . . ghe pare-lo che? . . . to-you seems-scl what ‘. . . what do you think?’ (Chiarelli, An ciapin, first half of the twentieth century) b. Neodo, u-tu mai che? nephew, want-scl ever what? ‘Nephew, what on earth do you want?!’ (Pison, Poesie, second half of the twentieth century)

7 As for the apparently optional location of che with respect to the inflected verb—compare (17) and (18)—we can hypothesize that at this stage two grammars coexist among speakers: as argued by Kroch (1989), learners may have two or more grammars when their triggering experience leads to incompatible analyses. Moreover, as pointed out by Lightfoot (1999), this kind of ‘internalized diglossia’ represents an interesting approach to solving the learnability problem concerning the acquisition of optional grammatical operations; alternations are diachronically unstable in language and represent a transition whereby one of the grammars is driven into disuse.

38

N. Munaro c. Ma a`-lo fat che de mal? but has-scl done what of bad? ‘But what harm has he done?’ (Studi bellunesi, from Luke 22: 39, 1981)

By looking at the input sequence described in (14), some learners may have parsed it not as (15a) but as (15b), which functions as Lightfoot’s (1999) cue, the abstract structure realized only in certain individual grammars which determines the parametric variation between grammars.8 The derivational procedure of this structure is plausibly reanalysed as in (15b), with cossa moving to its final landing site to perform overt checking of the feature [F3] associated to a higher functional projection (possibly connected with the nominal origin of cossa) while [F2] can be checked by (remnant) IP raising to the specifier of the relevant projection. This analysis is in line with Munaro, Poletto, and Pollock’s (2001) analysis, according to which che in main wh-questions raises to the specifier position of a N(ew)I(nformation)P, a low projection of the CP layer (hosting our [F1]); this wh-movement is followed by raising of the remnant IP to the specifier position of the higher projection GroundP (hosting our [F2]); finally, the cluster formed by the inflected verb and the subject clitic pronoun (merged in the low head TopP8) raises to the head of (Interrogative)ForceP (hosting our [F3]). This type of subject clitic, by its intrinsic morphological properties, can check the relevant feature, so further wh-movement of che is dispensed with: (22) [IntForceP [IntForce [
a-lo]x [GroundP [IP t
a magn
a ty ]z tx ][NIP [che]y tx [TopP pro[ToP 8tx ] tz ]]]] This derivation produces the deceptive appearance that in Bellunese there is no raising of the wh-element che to a functional specifier position of the left periphery. Such a parsing might eventually be extended to other bare whitems, including the wh-adverbials, ande´ and come´, which nowadays invariably appear sentence internally: (23)

8

a. Sie´-o š nda`di ande´? are-scl gone where? ‘Where have you gone?’ b. Se a`-lo comporta` come´? himself has-scl behaved how? ‘How did he behave?’

This also supports Lightfoot’s (1979) view that language learners converge on a grammatical system differing in at least one parameter value from the system of the speakers whose linguistic behaviour provides their input.

Grammaticalization and CP Layering

39

(Apparent) wh-in situ structures may eventually provide language learners with a factual basis for the application of a process of radical reanalysis, whereby the wh-item is processed as not having undergone overt syntactic movement to a functional specifier of the left periphery.9 9

In other words, they may be led to the following parsing of sentence internal wh-items:

(i) [CP Opy [C fax -tu][IP pro[I tx ][VP [V tx ][DP [QP chey ]]]]] The same diachronic pattern is displayed in these dialects by the wh-item chi in cleft and predicative interrogative structures. Until the nineteenth century chi appears sentence initially: (ii) a. Chi e´-lo quel doen, che gnen in qua? who is-scl that young, that comes in here ‘Who is that young man that is coming here?’ b. . . . chi e´-lo che fa el disnar . . . ? . . . who is-scl that makes the dinner . . . ‘. . . who prepares the dinner . . . ?’ (Villabruna, Fioretta, first half of the eighteenth century) c. Chielo, che no cognos la grega fede? whois-scl, that not knows the Greek faith ‘Who doesn’t know the Greek faith?’ (Coraulo, La Gerosalem liberada, 1782) (iii) a. . . . chi sara`-lo mai chel? . . . who be-fut-scl ever that ‘. . . who may be that person?’ (Anonymous author of the nineteenth century) b. Chi e´-lo che se ricorda pı` che mi sie stata? who is-scl that himself remembers more that I be been ‘Who remembers anymore that I have existed?’ (Nazari, Parallelo fra il dialetto . . . , 1873) At the beginning of the twentieth century a doubling structure is attested, and only later, in the second half of the century, some postcopular occurrences are attested, with both chi and che: (iv) a. Chi sara`-lo che la fala? who be-fut-scl that it fails ‘Who will fail it?’ che . . . ? b. Chi e´-lo chi che dit aree who is-scl who that said have-cond that . . . ‘Who would ever have said that . . . ?’ (Zanella, Poesie in dialetto rustico, 1901) (v) a. Chi e´-lo? Son mı`! E’-lo chi sto mı`? who is-scl? Am I! Is-scl who this I ‘Who is it? It’s me! Who is this me?’ (Luciani, I griff, second half of the twentieth century) b. Sara`-lo mai chi ’sta sort de strambel che . . . ? be-fut-scl ever who this kind of madcap that . . . ‘Who will ever be the madcap that . . . ?’ (Chiarelli, Finestra verta, 1955) c. E-lo chi de voialtri che . . . ? is-scl who of you that . . . ‘Who is among you that . . . ?’ (Studi bellunesi in onore del prof. Giovan Battista Pellegrini, 1981)

40

N. Munaro

2.3 Parche´ and Parcossa The diachronic pattern described in Section 2.2 has interesting implications for the distribution of the wh-items parche´ and parcossa in Veneto dialects. 2.3.1

The history of parche´ and parcossa

As discussed above, in the course of the eighteenth century cossa takes over the wh-operator function; at the same time, the wh-item parcossa, resulting from the merge of the preposition par with cossa, starts being attested for a while as an alternative to the form parche´: (24) a. Parche´ no vole´-u che vegna? why not want-scl that come-subj ‘Why don’t you want me to come?’ (Villabruna, Fioretta, first half of the eighteenth century) b. . . . parche´ po’ me domande-u? . . . why then me ask-scl ‘Then why do you ask me?’ (ibid.)

(vi) a. . . . e´-lo che da varda`? . . . is-scl what to look ‘. . . why are you looking?’ (De Matte`, Poesie, second half of the twentieth century) b. E-lo che mo, e´-lo che da zigar? is-scl what now, is-scl what to shout ‘Why are you now shouting like this?’ (Pellegrini, Poesie, second half of the twentieth century) This distribution is reflected in the synchronic variation attested in the dialects spoken in eastern Lombardy, as witnessed by the following examples: (vii) a. Ch’e`-l chi che maja le patate? wh is-scl who that eats the potatoes ‘Who eats the potatoes?’ b. E-l chi che porta al pa? is-scl who that brings the bread ‘Who brings the bread?’ c. Chi che l’ a` roba`? who that cl-has stolen ‘Who has stolen it?’ d. Chi ga maja`t la me turta? who has eaten the my cake ‘Who ate my cake?’ For an updated comparative analysis of cleft wh-structures in northern Veneto and eastern Lombardy, the reader is referred to Munaro and Pollock (2005).

Grammaticalization and CP Layering

41

c. E parche´ ade`s ste´-u la` senza far gnent? and why now stay-scl there without do nothing ‘And why do you stay there now without doing anything?’ (Coraulo, La Gerosalem liberada, 1782) (25) a. Parcossa scampe-lo, co son gnest mi? why flees-scl when am come I ‘Why is he fleeing, now that I have come?’ (Villabruna, Fioretta, first half of the eighteenth century) b. . . . parcossa mai [ . . . ] no e´-la usada ai nostri dı`? . . . why ever[ . . . ] not is-scl used at our days ‘. . . why the hell isn’t it used nowadays?’ (Anonymous, La polenta, eighteenth century) Both forms are attested in the nineteenth century, but parcossa is used in co-occurrence with negation, that is, in contexts which do not express a real request for information but rather the speaker’s attitude concerning the event referred to: (26) a. . . . parche´ rimprovera-u sta femena? . . . why scold-scl this woman ‘. . . why are you scolding this woman?’ (De Loda, from St Matthew, second half of the nineteenth century) b. Parche´ i Siri [ . . . ] t’ a`-li liga`? why the Siri [ . . . ] you have-scl bound ‘Why did the Siri bind you?’ (Pagani, Il libro di Giuditta, 1865) lassa` i toi? Parche´ t’ a`-lo piaze´st c. Parche´ a`-tu why have-scl left the your? why you has-scl liked vegner da noi? come to us ‘Why did you leave yours? Why did you feel like coming to us?’ (ibid.) (27) a. E par cossa no t’ a`-tu tolt la ombrela? and why not you have-scl taken the umbrella ‘And why didn’t you take the umbrella?’ (Zuccagni-Orlandini, Raccolta di dialetti italiani, 1864) b. Parcossa mo no sta`-tu tei to grun de ne`ole . . . ? why then not stay-scl in-the your lumps of clouds . . . ‘Why don’t you stay then in your lumps of clouds . . . ?’ (Nazari, Parallello fra il dialetto bellunese rustico, 1873) In the course of the twentieth century the form parche´ is by far the most frequently used one—see (28); the form parcossa tends to disappear, and the

42

N. Munaro

only attested occurrence, given in (29), still expresses the pragmatically marked reading described above: (28) a. Ma parche´ a`-lo na panza cossı` sgionfa? but why has-scl a belly so full ‘But why does he have such a full belly?’ (Battisti, Testi dialettali italiani, 1914) b. Parche´ po dormı`-o? why then sleep-scl ‘Then why are you sleeping?’ (Studi bellunesi, from Luke 22: 39, 1981) c. Parche´ le to man no tegne-le le me´e? why the your hands not hold-scl the mine ‘Why don’t your hands hold mine?’ (Dal Molin, Misteri, 1985) (29) Parcossa alora cruziarse par an vestito? why then worry for a dress ‘Why then worry over a dress?’ (Studi bellunesi, from Matthew 6: 19–34, 1981) Unlike other adverbial wh-items, parche´ and parcossa always occur sentence initially nowadays; this distributional asymmetry is predicted by the proposed analysis following Rizzi’s (2001) proposal that the (landing) site of perche´ in standard Italian is located higher in the CP field than the one hosting other wh-items.10 He proposes that, at least when construed locally, perche´ can be 10 This suggestion is supported by various arguments: a nominal subject can intervene only between perche´ and the inflected verb, in (i); perche´, unlike other wh-items, is compatible with a focalized constituent, in (ii) and (iii); in embedded wh-questions a focalized constituent must follow perche´ and precede other wh-items, in (iv): (i) a. Perche´ Gianni e` venuto? ‘Why did John come?’ b. *Dove Gianni e` andato? ‘Where did John go?’ c. *Cosa Gianni ha comprato? ‘What did John buy?’ (ii) a. Perche´ questo avremmo dovuto dirgli, e non qualcos’altro? b. *questo perche´ avremmo dovuto dirgli, e non qualcos’altro? ‘Why this / this why should we have told him, and not something else?’ (iii) a. *a gianni che cosa hanno detto (non a Piero)? b. *Che cosa a gianni hanno detto (non a Piero)? ‘to john what / what to john have they said (not to Peter)?’ (iv) a. Mi domando a gianni che cosa abbiano detto (non a Piero) ‘I wonder to john what they have said (not to Peter)’ b. Mi domando perche´ questo avremmo dovuto dirgli, non qualcos’altro ‘I wonder why this we should have said to him, not something else’

Grammaticalization and CP Layering

43

first merged in the specifier of Int(errogative)P, a relatively high projection of the CP layer. Under this view, it follows straightforwardly that the structural configuration underlying the reanalysed sentence internal occurrence of other bare wh-items does not affect parche´, which occupies a position higher than the one targeted by the IP raising process described in Section 2.2.5. 2.3.2

The interdependence of cossa and parcossa

2.3.2.1 Cossa in pseudo-questions In northern Veneto interrogative structures containing cossa in sentence initial position tend to be restricted nowadays to pragmatically marked contexts not implying a real request for information. The speaker rather expresses his personal attitude concerning the event, giving vent to a feeling which ranges from mild surprise to explicit disapproval: (30)

a. Cossa a`-lo magna` (che)?! what has-cl eaten (what) ‘What (the hell) has he eaten?!’ b. Cossa zı`ghe-tu (che)?! what shout-scl (what) ‘Why (the hell) are you shouting?!’ c. Cossa compre-tu n’altro giornal ((par far) che)?! what buy-scl another newspaper ((to do) what) ‘Why (the hell) are you buying another newspaper?!’

The particular semantic properties expressed by cossa in these cases are attributed by Munaro and Obenauer (2002) to the overt movement of the whitem to a specific structural position at the left edge of the sentence. Following the well-established tendency to associate each interpretively relevant feature to a specific head (projection) in the functional structure of the sentence, we proposed that, given the peculiar interpretive implications that are associated with the structures examined, the position occupied by cossa in this kind of sentence cannot be the same as that of wh-phrases in ordinary wh-questions. More precisely, we suggested that in pseudo-questions cossa occupies the specifier of a functional projection of the CP layer called Att(itude)-P which encodes the speaker’s attitude: (31)

[AttP cossa[IntForceP [IntForce
a-lo][GroundP [IP t
a magn
a (tche )][NIP (che) ...tIP ]]]]

Assuming that the marked value of the head Att8 corresponds to the speaker’s negative attitude towards the event, this reading—the marked option—must be activated by filling the corresponding specifier with cossa,

44

N. Munaro

leading to the expression of the speaker’s personal attitude towards the propositional content expressed by the sentence. 2.3.2.2 Parche´ and Parcossa in central Veneto In central Veneto dialects the wh-item why can be rendered by two different forms, parche´ and parcossa. Parche´ can be used in neutral questions and is incompatible with subject clitic inversion: a. Parche´ te scrivi na letera? why scl-write a letter b. *Parche´ scrivi-to na letera? why write-scl a letter ‘Why are you writing a letter?’ (33) a. Parche´ no te ve anca ti? why not scl-go also you b. *Parche´ no ve-to anca ti? why not go-scl also you ‘Why don’t you go as well?’ (32)

When parcossa is used, a cleft structure is needed in order to ask a neutral question: (34) Parcossa ze che no te ve anca ti? why is that not scl-go also you ‘Why don’t you go as well?’ If no cleft structure is used, parcossa is compatible both with proclitic and with enclitic subject pronouns; in the former case, the sentence tends to be interpreted as an echo question: (35)

a. Parcossa no te ve anca ti? why not scl-go also you ‘Why you don’t go as well?’ vien anca b. Parcossa el why scl-comes also ‘Why is he coming as well?’

lu? he

Questions introduced by parcossa with subject clitic inversion are used instead in pragmatically marked contexts in order to express the speaker’s evaluation of the event—cf. Zanuttini (1997) on this kind of structure: (36) a. Parcossa no ve-to anca ti? why not go-scl also you ‘Why (the hell) don’t you go too?’

Grammaticalization and CP Layering

45

b. Parcossa vien-lo anca lu? why comes-scl also he ‘Why (the hell) is he coming too?’ c. Parcossa (no) ghe-to prenota` do posti? why not have-scl booked two seats ‘Why (the hell) did(n’t) you book two seats?’ This use is likely to be related to the availability of the same kind of interpretation in questions with cossa, both in central and in northern Veneto. 2.3.2.3 The disappearance of parcossa in northern Veneto Parcossa is attested in northern Veneto from the eighteenth century, after cossa starts being used as a wh-operator; however, it is no more attested in present times, unlike in central Veneto. Its disappearance cannot be due to cossa’s occurrence in pseudo-questions, which is witnessed by the fact that this peculiar pragmatic use of cossa is already attested in the eighteenth century: (37) Cossa son-e stata qua` a zanzar, mo, mi? what am-scl been here to chatter now I ‘Why then have I been chattering until now?’ (Villabruna, Fioretta, first half of the eighteenth century) Moreover, the same use of cossa is attested in central Veneto, where parcossa has survived. An alternative hypothesis is that the disappearance of parcossa is due to the availability of the alternative form parche´, but, again, the two forms are attested in central Veneto. Indeed, on closer inspection, the decline of parcossa coincides with the spreading of the sentence internal wh-strategy, between the nineteenth and the twentieth century, which favoured the development of the three following structures, specialized for specific pragmatic functions: (38) a. Magne-tu che? eat-scl what ‘What are you eating?’ b. Cossa magne-tu (che)?! what eat-scl (what) ‘What the hell are you eating?!’ c. Cossa magne-tu (a ‘ste ore)?! what eat-scl (at these hours) ‘Why the hell are you eating (at this time)?!’

46

N. Munaro

So, while (38a) is a neutral wh-question requiring an answer, (38b) conveys the speaker’s surprise/reproach about what the person is eating (with cossa functioning as object of the verb), and (38c) can express the speaker’s annoyance at the mere fact that the person is eating (with cossa interpreted not as object but rather as causal adjunct). As Munaro and Obenauer (2002) point out, the division of labour splits standard interrogative argumental values expressed by che from non-standard values expressed by cossa. Given the autonomous development of three sequences pragmatically specialized, no possible ambiguity arises, and the resort to the form parcossa to express the interpretation conveyed by (38c) is dispensed with. As a consequence, parcossa, mainly used in pseudo-questions, disappears. In central Veneto, however, a string like (39a), containing a transitive verb, is potentially ambiguous between the three interpretations, so in one of the three readings it is roughly synonymous with (39b): (39) a. Cossa magni-to a ‘ste ore?! what eat-scl at these hours ‘What/why (the hell) do you eat at this time?!’ b. Parcossa magni-to a ‘ste ore?! why eat-scl at these hours ‘Why (the hell) do you eat at this time?!’ Hence, the form parcossa has presumably survived in these varieties in order to express unambiguously the non-argumental reading of cossa in structures like (39a). This hypothesis is confirmed by the fact that parcossa, unlike parche´, when associated with subject clitic inversion, systematically triggers the pseudo-question reading.

2.4 Conclusion In this case study I have analysed an instance of grammaticalization which affects the diachronic development of the distributional properties of some wh-phrases. I have thereby tried to show that grammaticalization phenomena can function as the (albeit indirect) trigger of a process of reanalysis leading to different parsing strategies, both in the derivation and in the syntactic representation of main wh-questions. More generally, if the diachronic path sketched here is correct, it provides support for the view that syntactic change is ultimately due to a parametric transformation in the lexical properties of individual functional heads,

Grammaticalization and CP Layering

47

intended as the necessity of the PF-realization of a given functional feature F, along the lines of Roberts and Roussou (1999). Primary Sources Battisti, C. (1914). Testi dialettali italiani in trascrizione fonetica. Halle an der Saale: Max Niemeyer Verlag. Chiarelli, I. (1955). Finestra verta. Bagnacavallo: Bottega della stampa ‘Zattoni’. Cian, V., and C. Salvioni (eds.) (1894). Le rime di Bartolomeo Cavassico notaio bellunese della prima meta` del secolo XVI. Bologna: Romagnoli Dall’Acqua. Circolo Dialettale Bellunese ‘Al zˇenpedon’ (1981). Studi bellunesi in onore del prof. Giovan Battista Pellegrini. Belluno: Tipografia Piave. —— (1985). A filo` ko l ‘zˇenpedon’—Antologia 1985. Belluno: Tipografia Piave. Coraulo, G. (1782). La Gerosalem Liberada del Tasso. Belluno: Tipografia Tissi. —— (1837). El filo`, ossia la veglia villereccia. MS Belluno. Lazzaris, L. (1931). Le stagioni dell’anno. Agordo: Tipografia Lise. Nazari, G. (1873). Parallelo fra il dialetto bellunese rustico e la lingua italiana. Belluno: Tipografia Tissi. Pagani, L. (1865). Il libro di Giuditta. Belluno: Tipografia Deliberali. Sacco, S. (ed.) (1975). Antologia dialettale della provincia di Belluno dalle origini ai giorni nostri. Belluno: Tipografia Piave. Segato, P. (1902). ‘Una novella svizzera tradotta in vernacolo feltrino’, in Antologia Veneta III: 129–49. Villabruna, V. (1748). La grandonazza e sparpositada gaodenzia de Barba Ettor dalla Valbruna par la nozza De Tarvis-Bellata. MS Feltre. —— (1755). Spatafio in lengua rustega feltrina . . . —Le passion de la Senta—Antonio e Barba Salvestro Bonisada. MS Feltre. —— (first half of the eighteenth century). Fioretta. MS Feltre. Zanella, V. (1901). Poesie in dialetto rustico feltrese del contadino Vetor Zanela. Feltre: Tipografia Zanussi and Curtolo. Zuccagni-Orlandini, A. (1864). Raccolta di dialetti italiani con illustrazioni etnologiche. Firenze: Tipografia Tofani. Zuppani, R. (1889). Raccolta di poesie. Belluno: Tipografia Cavessago.

3 Verb Movement and Clitics: Variation and Change in Portuguese ACR I SI O PI RE S

3.1 Introduction This chapter attempts to explain certain robust aspects of (diachronic) syntactic variation in Portuguese, focusing speciWcally on Brazilian Portuguese (BP), especially regarding verb movement and clitic placement.1 Analysing these phenomena is especially relevant for the evaluation of theories of diachronic and parametric syntactic variation in Romance and other languages. The kinds of phenomena discussed here are important for diVerent theories of grammatical reanalysis and change (cf. Hopper and Traugott 1993). The approach I adopt crucially hinges on the argument that the acquisition of innovating grammars is primarily the result of the interaction between properties of Universal Grammar (UG), language acquisition, and parsing/ interpretation of the primary linguistic data (PLD) by the learners (Lightfoot 1991, 1999; Hale 1998). However, the analysis proposed here does not necessarily appeal to markedness or complexity mechanisms that have been assumed in approaches to explain syntactic (diachronic) change—cf. Roberts and Roussou (1999). Notice, also, that the term ‘syntactic change’ itself is taken to be simply a shortcut to refer to the acquisition of innovative grammars by new generations of individuals, due to failure of the learners (children, in the usual case)

Thanks to David Lightfoot, Juan Uriagereka, and Norbert Hornstein for earlier discussion of some of the material discussed here. I especially thank David Lightfoot for reading a version of this chapter at the 7th Diachronic Generative Syntax Conference held in Girona, Spain (27–29 June 2002). 1

See Kayne (1991) for clitic movement. For related issues in European Portuguese (EP), Martins (1994a), Raposo (2000) and other papers in Costa (2000), Uriagereka (1995a, 1995b) and references therein. For Brazilian Portuguese (BP), see Galves (2001) and references therein.

Verb Movement and Clitics

49

to infer from the PLD grammatical settings that match in diVerent respects the grammars that were source of the PLD. The chapter is organized as follows. First, I summarize a possible analysis of the properties that distinguish diVerent Romance languages regarding the placement of clitics. Second, I argue that BP lost V movement to a position higher than the inXectional head, at least.2 I show that the loss of V movement in BP as I propose had diVerent consequences in BP, especially in the placement of clitics. In Section 3.6, I address the question of possible triggers for the kinds of change that took place in BP. I end the chapter by discussing other aspects of the change in clitic placement that took place in BP.

3.2 Three Parametric Settings in Romance Uriagereka (1995b) argues that French, Spanish, and Galician or European Portuguese (EP) are paradigmatic examples of three diachronic and parametric settings for clitic placement in Romance. He illustrates this in terms of clitic placement, suggesting that BP and French (Fr.) may behave alike: (1) a. < clitic, V > French, (spoken) Brazilian Portuguese (?) b. < clitic, . . . V > [þtense] Castilian Spanish, Aragonese, Catalan . . . c. < ø, V, clitic >; < Governor, clitic, V > EP, Galician, Leonese, most archaic dialects (Uriagereka 1995b : 157 (1)) One of the relevant contrasts is shown with inWnitivals in (2) and (3). French displays proclisis (2), whereas EP reverts from enclisis to proclisis in the presence of elements such as negation (3b) or a preposition (3c) introducing the embedded clause. (2) a. J’ai essaye´ [de l ’acheter] I have tried of it buy ‘I tried to buy it.’ b. J’ai essaye´ [de ne pas l ’acheter] I have tried of not it buy ‘I tried not to buy it.’

(Fr.)

(Fr.)

2 I have adopted a non-split INFL approach here, given that the approach and phenomena I present do not require a distinction between Agr and T.

50

A. Pires

c. [Sans l’acheter], on (ne) peut pas vivre. without it buy one not can not live ‘Without buying it, one cannot live.’ (3) a. Eu tentei [compra´-lo] I tried buy-it ‘I tried to buy it.’ b. Eu tentei [na˜o o comprar] I tried not it buy ‘I tried not to buy it.’ c. [Sem o comprar] na˜o se pode viver. without it buy not one can live ‘Without buying it, one cannot live.’

(Fr.)

(EP)

(EP)

(EP)

Uriagereka refers to an observation made by Wanner (1987), who argued that the increase in the frequency of attachment of clitics to verbs in Vernacular Latin was caused by the leftward appearance of the verb in certain emphatic clause types: imperatives, quotatives, presentatives, clauses with verb focus, and emotively marked utterances. Uriagereka (1995a: 163) then assumes that in the vernacular Latin cases and in the EP cases of V-CL such as (3a) the verb moves to a functional projection (FP) to the left of the clitic,3 given that FP has a strong F head that attracts the verb. F stands for a category that encodes point of view, thus it is used to express matters of topic, focus, emphasis, contrast, and so on. Uriagereka argues that the archaic Romance verb has a strong F feature in its morphology, which is checked against an active/strong F position (for a recent approach to active features such as the drive for movement, consider for example the probe-goal analysis in Chomsky 2000, 2001). According to Uriagereka, the feature checking requirements of the F head may be satisWed if: (4) a. an element moves to Spec,FP b. an element incorporates to F c. F cliticizes upwards. (4a, b) are represented by movement of a maximal (XP) or of a minimal projection (a head). (4b) is what accounts for the order V-CL in EP—see (3a). (4c) should account for the pattern XP-CL-V of (3b) and (3c), but the details of this alternative are much less clear,4 although they are not relevant for the 3 Uriagereka argues that the clitic left-adjoins to F and then V adjoins to F, yielding the order V-CL-(F). 4 For example, if cliticization is a phonological (PF) operation, it can satisfy the feature checking requirements of F only if this feature checking is required only at PF. Still, important ordering considerations arise.

Verb Movement and Clitics

51

analysis of the BP phenomena discussed below. For Uriagereka, F in EP and Galician is not only active/strong (i.e. attracting a V to adjoin to it) but also behaves as a clitic, which can relate to a host V to its left after V raising or by hopping onto the next head up as in (3b, c). Uriagereka argues that French has no syntactically overt F, given that V movement does not take place as it does in EP above. However, he also states that the licensing of clitics (especially third person clitics, for him) occurs by head movement to the head of FP. Given Uriagereka’s argument that French has no syntactically overt F, the claim that third person clitics left-adjoin to F needs to be revised, due to the existence of a full clitic paradigm in French, which includes third person clitics. If third person clitics can only be licensed in F, as proposed by Uriagereka, F should also be expected to be available in French, contrary to what (2a) and similar French cases show. To avoid arguing that certain syntactic projections are not realized in certain languages (F in this case), one alternative is to argue that F is not active/strong in French, so it cannot trigger V adjunction to its head, but it can still license clitics. Another alternative, which I assume here, is to take the clitic to adjoin to a functional projection (for example IP) lower than FP. This can explain the diVerent patterns in conjunction with the argument that V-toF takes place as in the case of EP and Galician. In one of the earliest approaches to movement, Kayne (1991) argues that clitics invariably leftadjoin to a functional head (IP), and the order V-CL in inWnitives results from the V having moved leftward past IP (see also Cardinaletti and Roberts 1991 for related issues; and Martins 1994a, Barbosa 2000, Raposo 2000 for diVerent analyses of the EP facts). However, what will be crucial here is the idea that there may be V movement (which may end up not being triggered in the acquisition of a new grammar) to a position higher than the landing site of clitics (presumably higher than IP), but not as high as C0 (given, for example, that Modern EP and late eighteenth-century BP show V movement but are not V2 or V-in-C languages). The argument that F is active/strong in EP and Galician and triggers V movement above INFL provides an explanation for why EP and Galician clitics cannot occur in initial position— contrary to BP, Modern Spanish, and the other languages in (1b).

3.3 Loss of V Movement in Brazilian Portuguese Turning to the current analysis, a striking change that took place in BP was the loss of the possibility of verb-subject (VS) inversion in the cases in which it was possible in the eighteenth and nineteenth centuries. First, Morais (1993: 284–5) shows that in the eighteenth century VS order was still fairly common

52

A. Pires

in declaratives in BP (38 per cent), although SV orders corresponded to between 62 and 72 per cent of her diVerent sets of data. By the late nineteenth century VS order decreased to about 10 per cent, suggesting that V movement to a position to the left of Spec,IP was no longer productive. The loss of VS order in questions may represent the last eVect of a loss of V movement to C. Old and Modern EP and also eighteenth-century BP (5a, 5b) usually displayed VS order (100 per cent in eighteenth-century direct questions), and a very low rate of SV order in wh-questions, whereas the SV order took over by the early twentieth century in BP. Contemporary Colloquial BP (ColBP) no longer allows inversion in wh-questions (5b, 5c). Furthermore, VS order in yes/no questions (which also decreased in eighteenthcentury EP) has now disappeared from BP (Duarte 1992; Rossi 1993). The loss of inversion as in (5) indicates clearly the loss of V-to-C. (5)

a. Que tem Deus de ver comigo? what has God of see with.me ‘What does God have to do with me?’ (Gil Vicente, sixteenth-century EP) b. O que faz a Maria? (EP, nineteenth-century BP, and *ColBP) the what does the Maria ‘What does Maria do?’ c. O que (que) a Maria faz? (ColBP) the what that the Maria does ‘What (does) Maria do?’

Furthermore, in EP and Galician and also in nineteenth-century BP overt VS inversion is possible in inWnitival complements (Longa 1994; Raposo 1987), whereas current BP rejects inversion in these cases (6). (6) a. Eu lamento terem os deputados trabalhado pouco. I regret have.inf.3pp the representatives worked little ‘I regret that the representatives worked little.’ (EP, Raposo 1987: (6), nineteenth-century BP, and *ColBP). b. Eu lamento os deputados terem trabalhado pouco. I regret the representatives have.inf.3pp worked little ‘I regret that the representatives worked little.’ (EP and BP) Based on such facts and on the placement of clitics discussed in the next section, I argue that V raising to F (or to any position above IP) was lost by the early twentieth century in BP, distinguishing it from EP and Galician. Provided that head movement proceeds locally, in V-to-C movement V would

Verb Movement and Clitics

53

actually have to move Wrst to F, and then to C. Arguably, the loss of V-to-C was one of the crucial triggers for the (subsequent) loss of V-to-F. If indeed it was the case that V-to-F was possible earlier and it was absent from the grammars of BP speakers after the late nineteenth century, the prediction is that such an innovation should have further eVects in these grammars. This is borne out regarding the placement of clitics in BP, as discussed in the next section.

3.4 Absence of Clitic Enclisis in Brazilian Portuguese A crucial fact in ColBP is the overwhelming occurrence of proclisis in clauses without auxiliaries, as observed since Teyssier (1976)—see (7): (7) Ela me / te viu. she cl.1ps cl.2ps saw ‘She saw me/you.’

(ColBP)

I argue that the occurrence of proclisis across the board in ColBP follows straightforwardly from a general loss of V raising to F in BP (which eliminated the source of enclisis in diVerent contexts) in the grammar of new generations. First, Cyrino (1993: 167–8) shows a systematic loss of clitic enclisis to inWnitives and to gerunds in BP. A rate of 86–100 per cent in 1850 reduces Wrst to 25–56 per cent by 1900 and then to zero by the 1950s. There was also a complete loss of clitics in enclisis to aYrmative imperatives (Cyrino 1993: 168),5 and clitics now only occur in proclisis to these constructions, contrary to what one Wnds in EP. These facts provide a clear indication that V raising to F (or to a higher position) in BP was in general lost by the mid-twentieth century. Clauses that have only an inXected main verb show a similar pattern (cf. Pagotto 1993), supporting the argument that V-to-F was lost by the midtwentieth century in BP.

3.5 Further Consequences of the Weakening of F in Brazilian Portuguese The argument for diVerent consequences of the loss/inactivation of F(P) in BP can be extended, connecting various properties found in EP and Galician and/ or in archaic languages (1c), but which are not found in BP (see Uriagereka 1995b). Such properties include not only V raising to F and the mixed pattern of proclisis and enclisis in (1c) discussed above, but also the ones in (8): 5

Data from Cyrino (1993: tables 2 to 4). The data are all from plays written by Brazilians.

54

A. Pires

(8) a. b. c. d. e.

interpolation of elements between clitics and V; overt F(ocus) elements; ‘recomplementation’; ‘sandwiched’ dislocations; overt expletives.

Uriagereka (1995b) argues that these and other properties hinge on the fact that F is strong/active. Hence, if this connection does hold, the argument that F became inactive/weak in BP provides a direct explanation for the loss of these properties as well.6

3.6 Triggers for the Loss of V Movement One crucial remaining question is how changes in the PLD would have triggered a reanalysis of the underlying structure of the clause, such that F became inactive as a host for X(P) movement, in the innovating grammars of nineteenth-century BP speakers. Two sets of possible triggers had to do with diVerent possibilities in S-V orders and in the placement of clitics. The Wrst set of triggers may be at the source of the loss of both V-to-C and V-to-F in BP. As early as the eighteenth century, when V-to-C was apparently still productive in BP, up to 83 per cent of the declarative sentences, in data analysed by Morais (1993: 284–5), already displayed SV order, which was open either to an analysis with V in a position higher than I (with S also occurring above IP) or to an analysis in which there was no V movement to a position higher than IP. It is likely that this high incidence of structures that were amenable to an analysis in which there was no V movement to a position higher than IP was a main source of the loss of V movement to F and to C in BP. Another possible source for the loss of V-to-F in BP may have been the variation in the position of clitics in complex clauses in early eighteenth century BP (Cyrino 1993: 168): (9) a. Maria Maria b. Maria Maria

queria falar-lhe wanted talk-him queria-lhe falar wanted-him talk

6 Uriagereka also links the possibility of inXected inWnitives to the existence of F(P). Although changes in the status of F(P) in BP might help explain the fact that inXected inWnitives have disappeared from colloquial BP, Pires (2001, 2002) argues that the loss of inXected inWnitives in ColBP follows more directly from a general weakening of the overt inXectional system of BP.

Verb Movement and Clitics

55

c. Maria lhe queria falar. Maria him wanted talk ‘Maria wanted to talk to him.’ The orders in (9b, c) did not require an analysis in which the lexical V had to move to a position higher than IP, since the lexical verb occurred to the right of the clitic. The order in (9b) later became the only possibility in BP, with the diVerence that the clitic stopped being enclitic to the auxiliary verb (see Section 3.7). Although the occurrence of the auxiliary verb to the left of the clitic in (9b) might still be analysed as requiring movement of the auxiliary to a position higher than IP, (9b) and (9c) represent further evidence that would have led to the reanalysis of the PLD by new generations, which would then acquire grammars in which there was no V movement to F or to C. Although more detailed evidence may be necessary regarding these possible sources of reanalysis in new grammars of BP speakers as early as the eighteenth century, the crucial point here is that new generations needed suYcient, unambiguous cues or triggers to set up individual grammars with V movement to F. Arguably, such evidence in the PLD was no longer enough or not adequately unambiguous to trigger the need for V-to-F, in the grammar of more and more speakers from the eighteenth to the twentieth century. Ambiguity in the PLD regarding SV orders and variation in the placement of clitics in complex clauses are factors that could have led to the absence of V-to-F movement in new grammars. One important point of the argument presented here is that the source of the reanalysis should have hinged simply on ambiguities in the ways the PLD could be interpreted by new generations, triggering grammars that did not require V movement to a position higher than IP. In order to acquire grammars that require V movement, new generations needed clear evidence from the PLD showing that the verb had to be analysed as being related to two diVerent positions, its base position (necessary, for example, to satisfy thematic requirements) and a dislocated position (which would be due to the need for the dislocated verb to satisfy certain features). Since the evidence for dislocation presumably became less robust, new generations were no longer able to identify such evidence in the PLD. In this respect, the source of the reanalysis depends entirely on properties of the PLD in interaction with general properties of the faculty of language (or Universal Grammar), and there is no need for to claim that the reanalysis is forced by some general tendency towards structural simpliWcation (see for example Roberts and Roussou 1999) imposed on learners, so that they choose a simpler grammar, in this case a grammar without movement. This kind of grammar may be

56

A. Pires

triggered simply because the PLD does not provide unambiguous cues or triggers for the need to assume that a certain category has to undergo movement from its base position.

3.7 Clitic Placement in Complex Verb Constructions I return here to clitic placement in complex verb constructions (Aux-V) as in (9). Proclisis to the auxiliary, (9c) and (10a), which was the predominant pattern until the mid-nineteenth century, becomes in general ungrammatical in BP by the twentieth century. Proclisis to the main verb in such cases (10b) is an innovation Wrst detected in the early nineteenth century (7.7 per cent in Cyrino 1993: 169) and becomes the only possible pattern by the late twentieth century (100 per cent by the mid-twentieth century, in Cyrino’s data). The rise of (10b) seems to have been triggered by a reanalysis of PLD such as (9b), which had the same order, but corresponded to enclisis to the auxiliary verb (Aux-CL V; Nunes 1993 proposes an analysis of the phonological innovation that took place here). (10)

a. <X CL (Aux) V> Maria me / te queria convidar. (early nineteenth-century BP and *ColBP) Maria cl.1p cl.2p wanted invite.inf ‘Maria wanted to invite me/you.’ b. <X (Aux) CL V> Maria queria me / te convidar. (twentieth-century BP and ColBP). Maria wanted cl.1p cl.2p invite.inf ‘Maria wanted to invite me/you.’

Considering Cyrino’s (1993: 169) data for the occurrence of a clitic in a clause that had an aVective element in position X, there was a drastic shift from the order in (10a)—90 per cent in the mid-nineteenth century—to the order in (10b)—100 per cent by the early twentieth century.7 Although (10a) shows that the clitic moved to a position to the left of INFL until the early nineteenth century in BP, there is no clear evidence that it moved up to C (in a V2-like pattern that was possible until the eighteenth century). Evidence for clitic movement up to C earlier, up to the eighteenth century, comes from the fact that clitics could occur above negation 7 Pagotto (1993: 192) cuts related data by centuries, so his analysis is less Wne-grained, although the same general shift can be seen clearly from the nineteenth to the twentieth century.

Verb Movement and Clitics

57

(11a), a possibility that was entirely lost by the early nineteenth century (see Cyrino 1997 and Pagotto 1993: 194). By the mid-eighteenth century (11b) became the predominant order (90 per cent in the late nineteenth century, in Cyrino 1993: 169)8 (11)

a. <X-CL NEG (Aux)-V> b. c.

(12)

a. Maria na˜o me / te

(until mid-eighteenth century) (late eighteenth century, early nineteenth century; example 12a) (late nineteenth century on; example 12b) tinha convidado.

Maria not cl.1p cl.2p had invited. ‘Maria hadn’t invited me/you.’ b. Maria na˜o tinha me / te convidado. Maria not had cl.1p cl.2p invited. ‘Maria hadn’t invited me/you.’

(early nineteenth century BP and *ColBP)

(ColBP)

Turning back to the innovative pattern in (10b) and (11c) [Aux-CL-V], I argue that the clitic did not behave as an aYx that required the auxiliary verb to move to its left. The loss of V-to-F movement argued for before must rule out the possibility that the new order (11c) was triggered by movement of the inXected auxiliary verb to the left of the clitic (i.e. to F), contrary to what was the case in earlier periods, when even main V-clitic order still occurred due to V-to-F. First, that is because by the time (10b)–(11c) arises (nineteenth century), other eVects associated with V movement above INFL had been lost or were being lost (for example, V-CL in simple clauses, subject-verb inversion and other properties given in (8)), and it would be incoherent to claim that V-to-F was triggered only in this innovative pattern. Thus, the new order [Neg Aux CL V] (11c), an instance of the (10b) order that became predominant from the nineteenth century onwards, did not involve movement of the auxiliary to F, but must be the result of further innovations in the grammar of BP speakers. One hypothesis is that besides loss of V movement to F, clitic raising to a position above (or in adjunction to) INFL was also lost (diVerent from French with auxiliaries avoir ‘have’ and eˆtre ‘be’, where the order is CL-Aux V).9 This is directly supported by the fact that the order in (11b)—[Negation CL-Aux V], an instance of (10a)—disappeared in favour of the order in 8 I do not address here the eighteenth-century factors that would have led to this speciWc innovation. 9 Under the alternative view that the clitic adjoins to F (Uriagereka 1995b), the loss of clitic movement should also be dependent on the inactivation of F.

58

A. Pires

(11c)—[Negation Aux CL V], an instance of (10b). Given that the clitic can be licensed by elements such as negation and prepositions in language with an active F as EP (3b) and (3c), the loss of V-to-F alone in the new grammars should not have aVected cases like (10a) and (11b), given that these do not require V-to-F even in grammars where V-to-F is possible, such as EP and Galician grammars. However, (10a) and (11b) still disappeared completely by the early twentieth century, and gave rise to (10b) and (11c) only. I argue that this was the result of a loss of clitic movement to the left of INFL, and not the result of a loss of V movement, which does not even play a role here.10 Considering these further facts, one possibility is that clitics stopped adjoining to the left of INFL and started adjoining to a position to the right of the auxiliary but to the left of simple verbs in BP (a possibility that indicates that there may be no overt movement of V to INFL in BP; it also requires inXected auxiliaries and main verbs to be distinct in terms of their overt position—see Lasnik 1999 for a similar distinction in English and French). In sum, other factors besides the loss of V-to-F should prevent the clitic from moving to a position higher than the auxiliary.

3.8 Complete Loss of Third Person Accusative Clitics This section brieXy discusses the situation of third person accusative clitics and why they were not considered in the discussion above. There are two important factors about the behaviour of third person accusative clitics in BP that distinguish them from Wrst person and second person clitics such as in (10), regarding the phenomena discussed here. First, the order in (10b) [X Aux CL V] did not extend to third person accusative clitics. They either continued to be realized to the right of the inWnitive verb (9a), or remained to the left of the inXected verb—in perfective tenses (13):11 (13) Maria o tinha visto. Maria cl.3p.acc had seen. ‘Maria had seen him.’ Second, the patterns of change applying to these clitics indicate that their occurrence decreased much faster than Wrst person and second person clitics. In Cyrino’s data (1993: table 10), they reduce from 48 per cent of occurrences 10 Lightfoot and Rodrigues (2002) put forward a detailed analysis of the loss of clitic climbing in BP along similar lines. 11 See Pagotto (1993). Galves (2001) makes interesting claims about the complete absence of third person accusative clitics from the native grammars of ColBP, given that they are allowed only on a restricted basis in more formal registers.

Verb Movement and Clitics

59

in the 1890s to 4 per cent in the 1970s (whereas null objects increase in frequency). Cyrino (1997) explores the possibility that the loss of third person accusative clitics was triggered by the rise of null objects in the grammar of new generations. However, other factors are necessary to explain why the pattern in (10b) did not extend to third person accusative clitics. Again, the loss of V-to-F should not be the reason for this or for the subsequent loss of third person accusative clitics, contrary to what would be predicted by Uriagereka (1995b). This is indirectly supported by the possibility that French has no active F but still displays such clitics, as discussed in section 3.2.

3.9 Conclusion This chapter has discussed innovations that took place in the grammars of BP speakers mostly between the mid-nineteenth and early twentieth centuries, especially in the position of verbs and clitic placement. It has been shown how they follow from the loss of both verb and clitic movement. It is expected that the core of these innovations in the grammars of new generations resulted from reanalysis of primary linguistic data (PLD) that triggered the acquisition of new grammars without V movement above INFL. It is expected that this reanalysis was possible only because of variation in the PLD that prevented new generations from correctly matching the properties of their own grammars to the ones that yielded the PLD they were exposed to. Possible sources for the trigger of this innovation in the PLD involved an increase in SV orders in declarative clauses and variations in clitic placement. Under this proposal, a loss of movement results simply from the absence in the triggering experience of clear evidence that a movement operation is required in the grammar.

4 Changes in Clausal Organization and the Position of Clitics in Old French ¨ HLER MARIE LABELLE AND PAUL H IRSCHBU

4.1 Introduction From the perspective of language change, grammaticalization is generally viewed as the process whereby ‘a lexical item or construction in certain uses takes on grammatical characteristics’ (Hopper and Traugott 1993/2003: 2). Roberts and Roussou (1999) account for this type of change as resulting from the reanalysis of lexical heads into functional heads. The diachronic changes we discuss in this chapter do not fall under this deWnition, but we view them as exemplifying a type of grammaticalization whereby illocutionary features come to be associated with distinct functional heads. We analyse the changes in the clausal organization of Old French as following from the fact that the Topic/Focus functional head common to all clause types of the Wrst stage gives way to a system with a number of separate illocutionary heads. We argue that the weakening of the Tobler–MussaWa (TM) constraint excluding object clitic pronouns from initial position in main clauses in Old French (OF) results from a gradual replacement of a common representation for V1 initial clauses by a new system where (1) satisfaction of a discourse-related [Top]/[Foc] feature by V is minimized, and (2) there is a reanalysis of the CP layer, with grammaticalization of illocutionary type features.

This work has been made possible thanks to a grant from the Social Sciences and Humanities Research Council of Canada (410–2001–0456). We thank Ana Maria Martins and Itziar Laka for their comments and suggestions on a previous version of this work.

Clausal Organization and Clitics in Old French

61

4.2 Old French as a V2 Language Old French is considered a V2 language (see for example Adams 1987, Roberts 1993a), with the verb generally following a topic or focus element in main clauses. We account for that characteristic by adopting a split-CP approach (Rizzi 1997), and assume that the C system is composed of two functional projections, ZP and FinP, as shown in (1). Z bears a discourse-related topic or focus feature, which we label [þD] (Discourse) for simplicity. Fin is the interface of the C system carrying a [
tense] feature, as in Rizzi. TP is the highest projection below the CP layer. (1 )

ZP

(XP)

Z


Z

FinP

Fin

TP

subject pronoun .........

Early Old French is a strong Top/Foc language in the sense that every main clause starts with a topic or focus element. We attribute this characteristic to the fact that the highest head in main clauses is Z. The [þD] feature of Z must be checked by a topic or focus element. We assume that a Z head with a [þD] feature selects a Finite head with a V feature forcing movement of V to Fin. V2 clauses result from an XP[þD] merging in or moving to Spec,ZP and V moving to Fin (to the left of pronominal subjects in Spec,TP);1 V1 clauses result from a V[þD] moving to the Z head. Crucially, this holds for all illocutionary types: declaratives, questions, and imperatives. If distinct [þIMP], [þWH], [þDECL] features are present for interpretive reasons, they have no visible eVect on the overt syntax of V in early Old French.

4.3 V1 Clauses and Tobler–MussaWa EVects The so-called ‘Tobler–MussaWa law’ corresponds to the observation made by Tobler (1875) and MussaWa (1886) that, in Medieval Romance languages, 1 We follow Vance (1997) who argues that the position of pronominal subjects is never lower than Spec,TP, while nominal subjects can occupy lower positions.

62

M. Labelle and P. Hirschbu¨hler

object clitics appear postverbally only when being preverbal would place them in clause initial position. Thus, in V2 clauses, clitics precede the verb, while in V1 clauses they follow it. The Tobler–MussaWa constraint is illustrated below with declaratives (2), questions (3), and imperatives (4): (2)

(3)

a. Allez en est en un verger suz l’ umbre. went gen is in an orchard under the shade ‘He went in an orchard under the shade.’ (Roland, 11) b. Vait s’en li pople. Goes refl gen the people ‘The people goes away.’ (Alexis, cxxi. 1) a. Sire, purquei m’ as dece´u¨e´? Sire why me-acc have deceived ‘Sire, why did you deceive me?’ (QLR, in Kok 1985: 78) b. Amerai le je s’ il ne m’ aime? will-love him-acc I if he neg me-acc love ‘Shall I love him if he does not love me?’ (Clige`s, 493)

(4) a. Car m’ eslisez un barun de ma marche. thus me-dat chose a baron from my march ‘Thus chose for me a baron from my march.’ (Roland, 275) senz dute les prendras, sis b. Pursiu les, chase them-acc without doubt them you-will-catch adv-them ociras. you-will-kill ‘Chase them, without doubt you will catch them, and kill them.’ (QLR, 58,8 in Kok 1985: 84) While many recent analyses of South Slavic languages attribute the postverbal realization of clitics in V1 clauses to their prosodic requirements (Franks 1999, Bosˇkovic´ 2001, Pancheva forthcoming, and references therein), in Old French, prosodic requirements play no role in the Tobler–MussaWa eVects displayed by object clitics. Thus, in Old French, object clitics can be Wrst in the intonational domain of which they are part, as shown in (5), which contrasts with the Serbo-Croatian examples in (6)—Franks (1999):2 2

Some examples, such as (i) below, may look as though the clitic les is postverbal because of a prosodic constraint. However, in this clause, li duze per is a left dislocated element, taken up by les. Just

Clausal Organization and Clitics in Old French

63

(5) Jo, qui voldreie parler a tei, le recevera´i. I who would-like talk to you him-acc will-receive ‘I, who would like to talk to you, will receive him.’ (QLR, in Kok 1985: 173) (6) a. *Ja, tvoja mama, sam ti obec´ala igracˇku. I your mother aux you-dat promised toy ‘I, your mother, promised you a toy.’ b. Ja, tvoja mama obec´ala sam ti igracˇku. I your mother promised aux you-dat toy ‘I, your mother, promised you a toy.’ Moreover, contrary to Wackernagel-type clitics, the object clitics of Old French were always adjacent to the verb. This distinguishes them from subject pronouns, always adjacent to Fin—cf. (1): (7) se il vers terre de Israel s’en vunt if they towards land of Israel refl gen go ‘If they go towards the land of Israel’ (QLR i. 13) Old French object clitics were generally proclitic and could even lean on a following word in the V-CL order: (8) Dunez m’ un feu. give me a Wef (Roland, 866) They were also less tightly connected to V than they are in contemporary French. In particular, they could have scope over V-headed constituents, as in (9), and license parasitic gaps, in (10). (9)

Qui la gart en pes et meintieigne. who her-acc keep in peace and maintain ‘Who keeps her and maintains her in peace.’ (Clige`s, 428)

as fronted adverbial clauses always behave as clause external in the initial stage, as discussed below, left dislocated elements are external to the minimal clause (Beninca` 1995). Similarly, the fact that the verb is clause initial is reXected by the postverbal clitic. (i) Li duze per, pur c¸o qu’il l’aiment tant, j DesW les ci, sire, vostre veiant. ‘The twelve Peers, because they love him soj (I) defy them, sire, in your presence.’ (Roland, 325–6)

64 (10)

M. Labelle and P. Hirschbu¨hler Ja la voloient en feu metre j Por rostir et por greı¨llier. already her-acc wanted in Wre put j for roast and for grill ‘They wanted to put her in the Wre to roast her and grill her.’ (Clige`s, 5936–7)

These facts suggest that object clitics left-adjoin to the highest head with a V feature (Fin or T according to the syntactic context) or that they are in the speciWer or multiple speciWers of that projection. For expository purposes we adopt the Wrst view, as the discussion of the scope of clitics over coordination or parasitic gaps falls outside the goals of this chapter. Sentence (11) is thus given the analysis in (12). (11)

[FinP li prierent [TP li meillor Sarrazin . . . ]]] [ZP Tant so-much him begged the best Sarrazins ‘The best Sarrazins begged him so much . . .’ (Roland, 451)

(12)

ZP tant [+D]

Z’ Z

Fin P Fin

li

TP

Fin prierent li meillor . . .

To account for the Tobler–MussaWa eVect, we assume that V, bearing a [þD] feature, moves from Fin to Z, to the left of the clitic. This is similar in spirit to analyses proposed by Beninca` (1995) and by Cardinaletti and Roberts (1991). Thus (3b) is analysed as in (13): (13)

ZP Z ameraii

FinP Fin

le

TP

taimerai je . . . . . . .

Notice that V and CL are in the C-system, to the left of subject pronouns in Spec,TP. Again, distinct illocutionary-type features do not play a role in the movement of VWn to Z.

Clausal Organization and Clitics in Old French

65

In negative main clauses, the verb is preceded by object clitics. (14) a. De cest services ne vos membre il gaires of this service neg you-refl remember expl-subj much ‘You don’t remember much of this service.’ (Charroi, 180) b. N’ i perdrat Carles, li reis ki France tient, neg loc will-lose Carles the king who France holds ‘Carles, the king who holds France, will not lose.’ (Roland, 755) In V2 negative clauses (14a), the initial XP occupies Spec,ZP, and the verb occupies Fin, as it precedes subject pronouns. Therefore object clitics are preverbal. As the verb also precedes subject pronouns and follows object clitics in V1 questions (15), we assume that it also occupies Fin in these clauses. We take the position that the Wnite verb also occupies Fin in V1 negative declaratives (14b), even though the absence of postverbal pronominal subjects in V1 declaratives makes it diYcult to show independently that the verb is within the CP layer in these clauses. (15)

Nel veez vos donques? neg-him see you thus ‘Don’t you see him?’ (Lancelot, 1701)

The philological tradition attributes the preverbal position of object clitics in such V1 negative clauses to the presence of clause-initial n(e). Within the present framework, the syntactic position of ne is not clear. One could assume that ne occupies Z (which would then host some polarity feature) or that it occupies a negative head above Fin, preventing V from moving higher (to Z) in V1 constructions. For concreteness, we will adopt the Wrst view.

4.4 Changes in the Position of Clitics 4.4.1 Adverbial clauses

In early OF, clitics were postverbal following an initial adverbial clause (16). We argue that the adverbial clause is ZP-external and that the main clause is structurally a V1 declarative: the verb is under Z and the clitic is adjoined to Fin.

66 (16)

M. Labelle and P. Hirschbu¨hler [Cum il vint a u´nes loges a´ pasturs en cel chemin] truvad i les freres Achazı´e, as he came to one hut to shepherd in this road found loc the brothers Achazie ‘As he came to . . . , (he) found there the Achazie brothers.’ (QLR iv. 195–6)

Around 1170, object clitics start being found in preverbal position after adverbial clauses (17). This change occurs in declaratives and yes–no questions, but not in imperatives. (17) [Quant nos lor donons nos aumosnes] lor faisons bien as when we them-dat give our alms them-dat make good to-the cors. body ‘When we give them alms, we do good to their body.’ (Sully, in Kok 1985: 91) Since the change occurred at a time when V1 declaratives were disappearing (Ska˚rup 1975), we interpret these concomitant changes as indicating the emergence of a new grammar where, in declaratives, focusing the V (resulting in narrative inversion) is excluded. This can be expressed by saying that, in declaratives, V may no longer carry a [þD] feature. As a result, V stays under Fin and some other element must check the [þD] feature of Z. This leads to the reanalysis of the adverbial clause in examples like (17) as being merged in Spec,ZP. The incorporation of adverbial clauses in Spec,ZP is, however, not the preferred option, the prevailing strategy favouring adjunction of the adverbial clause to a V2 main clause. The emergence at the same time of postverbal subject pronouns in declaratives introduced by an adverbial clause (18) is thus accounted for, as, in declaratives, postverbal subjects occur only in V2 clauses. (18)

[Ainz qu’ eu¨ssiez . . . ] j I avroit il feru M. cops d’ before that would-have loc would-have he dealt 1000 blows of espee sword ‘Before you would have . . . j he would have dealt you a thousand blows with his sword.’ (La Prise d’Orange, in Ska˚rup 1975: 279)

In yes–no questions starting with an adverbial clause, an identical change aVects the position of object clitic pronouns (19). This conWrms that adverbial clauses may Wll Spec,ZP.

Clausal Organization and Clitics in Old French

67

(19) se je descent a pie´ j me porrai je en vos Werj De . . . ? if I go-down on foot me-dat can-fut I in you trust to ‘. . . if I go on foot, can I trust you to . . . ?’ (Perceval, in Ska˚rup 1975: 357) However, the V1 construction survives in yes–no questions, and in that case, the word order continues to be V-CL (20). This shows that in interrogatives, V could still carry a [þD] feature, and raise to Z. (20) Conois la tu? know her-acc you ‘Do you know her?’ (Queste del Saint Graal, 112. 17, in Kok 1985: 82) Adverbial clauses never Wll Spec,ZP in imperatives, and V always moves to Z in these clauses. Thus this Wrst change results from the fact that the [þD] feature on tensed verbs is now excluded from declaratives. This is the Wrst sign of a breakdown of the uniform system characterizing early Old French. While illocutionary features did not seem to play a syntactic role in the Wrst stage, the diVerent behaviour of declaratives on the one side and questions and imperatives on the other creates a situation where such features are more likely to come to play a syntactic role. Verbs having a [þWH] or a [þIMP] feature raise to Z in V1 clauses, but not verbs having a [þDECL] feature. Notice that the change does not concern V2 clauses. Despite this Wrst change, there is no change in clause structure, i.e. Z continues to carry a [þD] feature, forcing the CP layer to be projected. 4.4.2 Coordination

In early OF, object clitics are postverbal after a coordinator joining main clauses, which indicates that, in main clauses, coordinators selected full ZPs: (21)

e´ mist la al lit David ( . . . ) e´ de dras bien and put her-acc in-the bed David and of sheets well la cuvrid . . . her-acc covered ‘. . . and placed it on David’s bed ( . . . ) and covered it well with sheets . . .’ (QLR i. 39)

Clitics start being preverbal after et ‘and’ and ou ‘or’ in declaratives, yes–no questions, and imperatives (22), around the time they start being preverbal following an adverbial clause.

68

M. Labelle and P. Hirschbu¨hler

(22) Leve´s sus et me preste´s trois pains get-up and me give three breads ‘Get up and give me three loaves of bread.’ (Sully, in Kok 1985: 93) Following insights by Beninca` (1995) and Kiparsky (1995), we suggest that et and ou now take FinP as a second conjunct when this conjunct is a V1 clause. This results in the clitic-Verb order. It is not clear why this change occurred. We can speculate that, since sequences of the type et V, generally with a null subject, are extremely frequent in declaratives, the ban against V[þD] in declaratives results in a reanalysis of such sequences as having V under Fin. This reanalysis can be extended to the other types of clauses because, in such coordinations, the second conjunct inherits the illocutionary force from the Wrst. Interestingly, following mais ‘but’, the order remains Verb-clitic. For example, in Chrestien de Troyes’s Chevalier de la Charrette, all examples with et have preverbal clitics (23a), but mais has postverbal clitics (23b): (23)

a. Car te teis / et me croi thus you-refl hush and me-acc believe ‘Shut up and believe me!’ (Charrette, 5030–1) b. Nel me celez, je sui des voz,j Mes dites not-it-acc me-dat hide I am of-the yours but say le seu¨remant it-acc securely ‘Do not hide it from me, I am one of yours, but speak openly.’ (Charrette, 5066–7)

We argue that mais continues to coordinate ZPs because it requires each conjunct to express its own illocutionary force. Observe the following contrasts from contemporary French, where the two coordinated clauses have a distinct illocutionary force: (24) a. [decl Pierre est doue´], mais/ *et/ *ou [qu a-t-il fait les Pierre is gifted but and or has-he made the sacriWces requis?] sacriWces required ‘[decl Pierre is gifted], but/*and/*or [qu has he made the necessary sacriWces?]’

Clausal Organization and Clitics in Old French

69

b. [imp Suis-moi si tu veux], mais/ *et/ *ou [qu sais-tu follow-me if you want but and or know-you dans quoi tu t’ embarques?] in what you refl are getting ‘[imp Follow me if you want], but/*and/*or [qu do you know what you are getting into?]’ c. [decl Pierre n’ est pas rapide], mais/ *et/ *ou Pierre neg is not quick but and or [imp fais-lui conWance!] do-him trust ‘[decl Peter is not quick],but/*and/or [imp trust him!]’ 4.4.3 Object clitics in absolute clause initial position

About thirty years after the preceding change, object clitics start being found in absolute initial position of declaratives and yes–no questions, but not imperatives. (25) a. Ou est sa feme? La nos mosterez vos? where is his wife her-acc us-dat show-fut you-nom ‘Where is his wife? Will you show her to us?’ (Orange, 528) b. Sire, l’ ai tant quis que je l’ ai Sire him-acc have so much searched that I him-acc have trove´ en vostre cort. found in your court ‘Sire, I have looked for him so much that I have found him in your court.’ (Perceval, in Ska˚rup 1975: 354) The fact that, here, clitics precede the verb indicates that there is no movement of V to Z in V1 declaratives and questions. Moreover, the fact that no XP precedes the clitics indicates that Z is not projected in these clauses. While early Old French was a strong Topic/Focus language where ZP was always projected, this is not the case anymore. In V2 clauses, the verb precedes pronominal subjects (26), as before. Given that these clauses have an XP checking the [þD] feature of Z, it appears that when ZP is projected, FinP must also be projected, which forces V-movement to Fin. (26) et pui i fu ele portee and then loc was she carried ‘and then she was carried there.’ (Clari, 67, 77, in Kok 1985: 75)

70

M. Labelle and P. Hirschbu¨hler

In V1 questions (25a), the verb in clause initial position precedes the subject pronoun as it did in previous stages, indicating that V is within the CP layer. But here, the position that the verb occupies is not selected by Z with a [þD] feature. We suggest that, at this stage, this position comes to be analysed as a head position bearing a [þWH] feature. In V1 positive declaratives, pronominal subjects do not occur. It is thus not clear whether V is within the CP layer or in T in sentences like (25b). It could be that the verb is under Fin. But, in the absence of Z selecting a Fin head with a V feature, it is also possible that the CP layer is simply no longer projected in declaratives. Adams (1987) claimed that null subjects were always postverbal in Old French, but Hirschbu¨hler and Junker (1988) showed that preverbal null subjects could be found in embedded clauses, opening up the possibility that they could also be preverbal in V1 declaratives. Also, Vance (1997) argues that SVO clauses were TPs at that time. We will thus adopt the view that, in the new V1 declaratives of type (25b), the verb remains within TP. In V1 imperatives, just as in previous stages, the verb precedes the clitics, and it must therefore occupy the highest position in the CP layer. To sum up, we suggest that the grammatical system of that period is the following. V2 clauses have, at least superWcially, the same grammar as in early Old French: the verb occupies a position above TP, probably Fin, and there is a higher projection above FinP, ZP, whose speciWer hosts an XP. This ZP is, however, no longer obligatorily projected. In V1 clauses, the uniform CP layer syntax for all clause types typical of early Old French has been replaced by a system where imperatives, questions, and declaratives have a distinct syntax. We observe the emergence of a typological change where distinct illocutionary force features are associated with distinct verb positions: . In topicless declaratives,V remains within TP. . In yes–no questions, V is higher than TP, but lower than Z, as it follows object clitics. . In imperatives, V occupies the highest position of the CP layer, to the left of object clitics. Thus, while Z and Fin remain in V2 clauses, in V1 clauses we see the emergence of specialized functional heads a` la Rizzi.

4.5 Conclusion In this chapter, we have looked at changes which are manifested most clearly in the respective position of the clitics and the verb in main clauses. We have

Clausal Organization and Clitics in Old French

71

adopted an analysis that derives these changes not from changes in the syntax of clitics, but from a reorganization of the CP layer. We have suggested that there was a reanalysis of the uniform CP layer, with grammaticalization of illocutionary type features. While features like [þDECL], [þWH], [þIMP] did not play a syntactic role in the earliest stage, they came to be crucially involved in the syntax of the clause. We suggest that in the last stage discussed, these features were checked in diVerent heads : [þDECL] is checked with T, allowing V to remain within TP in the absence of a motivation for raising higher; [þWH] is checked above TP, suggesting that speakers have reanalysed the Fin position in V1 interrogatives as Rizzi’s Focus projection (Rizzi 1997); [þIMP] is checked in the highest position, which is perhaps not Z in this case, but a distinct head (like Rizzi’s Force) bearing a [þIMP] feature. Primary Sources La Chanson de Roland, ed. G. Moignet (1970). Paris: Bordas. [Abbr. Roland] Le Charroi de Nıˆmes, ed. R. van Deyck (1970). Saint-Aquilin-de-Pacy: Malier. [Abbr. Charroi] ChrØtien de Troyes. Chevalier de la Charrette, ed. M. Roques (1958). Paris: Champion. [Abbr. Charrette] —— Clige`s, ed. C. Me´la and O. Collet (1994). Paris: Le livre de Poche (edition from MS BN fr12560). —— Lancelot, ed. K. Uitti and A. Foulet (1989). Paris: Bordas (Classiques Garnier). La Prise d’Orange: chanson de geste de la Wn du XIIe sie`cle, ed. C. Re´gnier (1970). Paris: Klincksieck. [Abbr. Orange] Li Quatre Livre des reis, ed. E. R. Curtius (1911). Dresden-Halle: Niemeyer. [Abbr. QLR] La Vie de Saint Alexis, ed. C. Storey (1968). Geneva: Droz. [Abbr. Alexis]

5 Exceptional Case Marking in Brazilian Portuguese HELOISA MARIA MOREIRA LIMA SALLES

5.1 Introduction A well-known fact about Brazilian Portuguese (BP) is the occurrence of an inWnitive clause introduced by the preposition para ‘for’ (pra/pro, in its reduced form) with an embedded oblique subject (cf. (1)): (1) Maria disse pra mim / pro Pedro estudar. Maria said for me for.the Pedro study-INF ‘Maria told me / Pedro to study.’ Several studies have related the emergence of these constructions to the loss of the inXection on the inWnitive in BP (cf. Lightfoot 1991; Botelho-Pereira and Roncaratti 1993; Pires 2001). In previous work (cf. Salles 1999, 2003), I have challenged this approach, pointing out that inXected inWnitives are still found in BP and proposing that the para-INF construction evolved as an alternative to subjunctive constructions.1 In particular, it was shown that given the neutral character of the matrix predicate dizer ‘say’ as to (lexical) modality marking (cf. Botelho-Pereira 1974) and given the (tendency to the) neutralization of the morphological distinction between indicative and subjunctive in BP, Wnite (subjunctive) clauses are excluded as complements of the (matrix) predicate dizer ‘say’ in the directive reading, hence inWnitive clauses become obligatory.2 The epistemic reading in turn remains encoded by the Wnite indicative clause—compare (1) and (2): I would like to dedicate this chapter to the linguist Adriana Maria Soares Viana (in memoriam). 1 Los (1999) independently posits a similar correlation for the emergence of ECM constructions in (Middle) English. The idea of linking ECM and subjunctive constructions was put foward in Emonds (1976) and also in Kempchinsky (1986). 2 Rocha’s (1997) study, which is formulated within the variationist framework and looks at data collected from informants of diVerent social classes, shows that in the contexts where the subjunctive is

Exceptional Case Marking

73

(2) Maria disse que eu estudei/ Pedro estudou. Maria said that I studied/ Pedro studied It was further shown that the emergence of (1) interacts with another feature of BP, as opposed to EP, namely the fact that datives are licensed by the preposition para (cf. (3) and (4)): (3) Maria disse pro Pedro que Gabriel chegou. Maria said to Pedro that Gabriel arrived (4)

Maria entregou a carta pra mim / pro Gabriel. Maria gave the letter to me to Gabriel

Interestingly, EP also displays a (directive) construction with para introducing an inWnitive clause (cf. Mateus et al. 2003). However, the EP construction crucially diVers from that of BP as the embedded subject is either a control position or a (non-overt) nominative one, the latter licensed by the inXected inWnitive (cf. (5)). The inWnitive construction in (5) alternates with the Wnite subjunctive construction in (6)—the latter not found in BP, as already noted: (5) Maria disse-me para Ø estudar/ disse a`s Maria said-to.me for study-inf-1s said to.the para Ø estudar(em). for study-inf-3pl (6) Maria disse-me que Ø estudasse/ disse a`s Maria said-to.me that study-subj-1s said to.the estudassem. study-subj-3pl

crianc¸as children

crianc¸as que Ø children that

According to Salles (2003), in (1) para is merged as a C head encoding irrealis modality. Following Roberts and Roussou’s (1999, 2003) theory of syntactic change, I will presently propose that the emergence of (1) in BP illustrates a grammaticalization process, by which para is reanalysed as the functional head C introducing the embedded inWnitive clause. In connection with the facts illustrated from (2) to (6), this process points to a development that recalls the rise of E(xceptional) C(ase) M(arking) constructions in English. As is well known, the English counterpart of (1) from BP (she said for / wants (for) me to study) is an innovation of (late) M(iddle) E(nglish) (see Fischer 1988, Lightfoot 1991 for a generative approach) and as pointed out in Los (1999: 287), ‘there is solid evidence of to-inWnitival ECMs at an earlier period expected, the subjunctive alternates with the indicative in BP. I take this result to indicate that the morphological contrast between the indicative and the subjunctive collapsed in BP.

H. M. M. L. Salles

74

after the ditransitive verbs, which points to a reanalysis of the to-inWnitival object control construction with which these verbs occurred in OE’—a fact that Wnds a correlate in (5) from EP, also an object control construction with a ditransitive verb. In spite of the aYliation to the Indoeuropean stock, it is a remarkable fact that BP and English, which belong to diVerent linguistic families, share various features in their development (cf. Roberts 1993b; Salles 1997). Assuming BP to have evolved from Classical Portuguese (the variety that was brought to Brazil during the colonization period (the sixteenth to nineteenth centuries), these facts show that languages vary as to whether Wnite or non-Wnite clauses are found as complements, optionality being one of the patterns (as in EP). Drawing on Salles’s (2003) analysis, I shall present additional evidence supporting the idea that the development of the complementation system in BP is not dependent on the loss of inXection on the inWnitive, taking into consideration the fact that inXection on the inWnitive is not a stable feature in EP as it alternates with non-inXection in a number of cases. I will leave aside the control construction in (5) from EP, as the discussion of its properties is a topic on its own, taking us too far aWeld (but see Pires 2001 for an approach including BP). The study will be presented as follows: in Section 5.2, I will discuss the syntax of (inXected) inWnitives in BP as opposed to EP. In Section 5.3, I shall propose an analysis for the contrast in the complementation system of BP and EP within Roberts and Roussou’s (2003) theory of language change, pointing out the structural steps of reanalysis in BP. In Section 5.4, I shall present the Wnal remarks.

5.2 Aspects of the Complementation System in EP and BP 5.2.1 InXected inWnitives in EP

A well-known feature of European Portuguese grammar is the presence of inXected inWnitive clauses with nominative subjects.3 However, these constructions cannot occur as root sentences (cf. (7)): *Os meninos saı´rem. the boys leave-inf-3pl

(7)

In fact, they are found in embedded contexts as adjunct clauses (cf. (8)), preverbal subject clauses (cf. (9)), and completive clauses (the latter to be detailed below): 3

The exposition on the distribution of inXected inWnitives in EP follows Raposo (1987).

Exceptional Case Marking

75

(8) Maria comprou canetas para as crianc¸as desenharem. Maria bought pens for the children draw-inf-3pl ´ (9) E difı´cil as crianc¸as desenharem (com essas canetas). is diYcult the children draw-inf-3pl with these pens Completive (inXected) inWnitive clauses are found in the complement position of epistemic and factive predicates. With epistemic predicates, auxiliary-subject inversion is obligatory, as illustrated in (10), whereas with factives both the VS and SV word orders are allowed (cf. (11)): (10) Maria acredita terem as crianc¸as estudado. Maria believes have-inf-3pl the children studied ‘Maria believes them to have studied.’ (11) a. Maria Maria b. Maria Maria

lamenta regrets lamenta regrets

terem as crianc¸as chorado. have-inf-3pl the children cried as crianc¸as terem chorado. the children have-inf-3pl cried

InXected inWnitives are also found in causative and sensitive constructions, as illustrated in (12a) and (12b), respectively:4 (12)

a. Maria Maria b. Maria Maria

mandou made viu as saw the

as crianc¸as estudarem. the children study-inf-3pl crianc¸as estudarem. children study-inf-3pl

Given the nominal character of inXected inWnitives, Raposo (1987) proposes that their distribution is determined by a requirement on case assignment by a governing head (P in adjunct clauses, INFL in preverbal subject clauses, and V in completives), variation in word order depending on whether CP or IP is selected. Madeira (1995) in turn proposes that all inXected inWnitive clauses are CPs and that the embedded subject is licensed by a 4 I shall not go into the rather complex discussion of causative (and sensitive) constructions. Apart from (12), EP displays the so-called Romance causative construction with a non-inXected inWnitive (cf. (i)), diVering from other Romance languages in further allowing a Wnite clause, with the subjunctive (cf. (ii)):

(i) Maria mandou/fez estudar a lic¸a˜o ao menino. Maria made study-inf the lesson to.the boy ‘Maria made the boy study the lesson.’ (ii) Maria mandou que o menino estudasse a lic¸a˜o. Maria made that the boy study-subj the lesson In BP both the Romance and the Wnite causative are absent (the latter due to the absence of subjunctive morphology), hence only the (inXected) inWnitive construction is found (see Andrade 2002 on BP causatives within a minimalist framework).

76

H. M. M. L. Salles

dependency between the C and T heads. The C-T dependency is realized under T-to-C movement whenever C does not have semantic content, that is, in epistemic contexts. In the remaining contexts, C has semantic content and T-to-C movement is not obligatory. Formulated within a minimalist framework, Galves’s (2003) analysis postulates that cross-linguistic variation is found with respect to the positions in which Agr is generated in the clause structure. Given that inXected inWnitives are independent of Tense, their existence in a particular language may be explained by the possibility of generating Agr in C (cf. also Lobato 1988), the nominative case being assigned in Spec,CP. Obligatory auxiliary inversion with epistemic predicates is accounted for under the assumption that they select for an abstract Tense, as proposed in Raposo (1985, cited in Galves 2003), which blocks both the generation of Agr in C and the rise of the subject to Spec,CP, further requiring movement of the auxiliary to C as a condition for legitimating Agr of the inXected inWnitive. Pires (2001) takes Galves’s (2003) analysis to be essentially correct. He points out that this analysis could be given through Uriagereka’s (1995b) approach to Galician inXected inWnitives, which postulates a functional head F in the clause structure that expresses point of view. The subject of the inWnitive would be licensed above the T head, in the same fashion as in Galves’s system. I shall not go into the matter of discussing these approaches. However, in the spirit of Madeira’s (1995) analysis, I suggest that obligatory Aux-subject inversion with epistemic predicates is due to a requirement on lexicalization of C, which is realized under T-to-C movement in EP (to be developed below, following Roberts and Roussou’s (2003) theory of language change). 5.2.2 (InXected) inWnitives in BP

A fundamental fact concerning the syntax of inXected inWnitives in BP is that auxiliary-subject inversion is absent (cf. Silva 1996). Hence (inXected) inWnitive clauses with overt subjects (cf. (10)) are not found with epistemic predicates in BP, the Wnite clause with the indicative (cf. (2)) being the only option for encoding epistemic modality, also available in EP. Accordingly, in BP factive constructions with Aux-subject inversion (cf. (11a)) are not found either, although ones with the SV order remain (cf. (11b)). As already noted, another feature of BP, but not EP, is the existence of a para-INF clause in the complement position of the (matrix) predicate dizer ‘say’. The relevant example is quoted in (1), and repeated as (13):

Exceptional Case Marking (13)

77

Maria disse pra mim / pro Pedro estudar.

Interestingly, in north-eastern dialects of BP, the para-INF construction is also found with volitional predicates (pointed out in Salles 2003), as illustrated in (14):5 (14) Maria quer para mim/ pro Pedro estudar. Maria wants for me for-the Pedro study-inf The para-INF construction is also found as pre- and postverbal subject inWnitives, as illustrated in (15) and (16): (15)

Pra mim/ pro Pedro for me for.the Pedro (16) E´ difı´cil pra mim / is diYcult for me

ler o livro e´ difı´cil. read-inf the book is diYcult pro Pedro estudar. for.the Pedro study-inf

Examples from real speech of popular BP are given in (17) and (18).6 Minha ma˜e esperou pra mim nascer. my mother waited for me be.born-inf (18) Para voceˆ ir a um lugar e na˜o Wcar tranqu¨ilo na˜o for you go-inf to a place and not be-inf relaxed not adianta ( . . . ) is.worth (17)

As can be inferred from the gloss, (17) parallels with English, while (18) allows for a correlation with Chomsky’s example, given in (19): (19) a. It’s pleasant for the rich for the poor to do the hard work b. It is pleasant [BENEF for the rich] [SUBJ for the poor] to do the hard work

5.3 An Account for the Contrast Between the Complementation System of BP and EP In the present section the patterns of subordination found in BP and EP are discussed. Following minimalist assumptions (cf. Chomsky 1995, 2001), I will 5

Other dialects of BP use Wnite indicative clauses in this context (cf. (i)), the irrealis modality being encoded in the lexical properties of the matrix predicate—a possibility that is not available for predicates of saying, due to their neutral character as to (lexical) modality marking: (i) Maria quer que Pedro estuda. Maria wants that Pedro study-ind 6 Real speech data collected in Lemle and Naro (1977), cited in Botelho-Pereira and Roncaratti (1993).

78

H. M. M. L. Salles

assume, with Salles (2003), that merge of the complementizer que ‘that’ as the C head encodes Wniteness but is neutral as to modality marking. Hence irrealis reading is obtained under matching/agreeing of the (embedded) C with the subjunctive on the verb (cf. (6)), an operation requiring temporal dependency (and some sort of discontinuous morphology). Merge of the preposition para ‘for’ as the C head (cf. (1), (13), and (14)) encodes both clause-type (namely, non-Wniteness) and irrealis modality (see Roussou 2000, for an analysis proposing modality marking on C). Drawing on Roberts and Roussou’s (2003) theory of language change, according to which languages diVer on whether the grammatical features of a given functional head have an overt exponent or not, I propose that para is a particle that lexicalizes grammatical/ case features of C in BP, but not in EP. Moreover, merge of the dedicated particle para as C is in complementary distribution with movement of the auxiliary to C, a mechanism encoding opposite (semantic) features, namely irrealis and epistemic modality, respectively. This further illustrates Roberts and Roussou’s (2003) proposal that lexicalization of a functional head is obtained under either merge or movement. At this point, a distinction between Standard and Popular BP (along social classes, education level, etc.) should be made. In particular, a crucial diVerence is the absence of third person inXection on the inWnitive in the popular variety, a phenomenon also aVecting Wnite verbs and recurrently discussed in the literature (cf. Duarte 1993, 2000, among many others). I propose that this fact interacts with changes in the complementation system, although it does not cause them. Precisely, I take the development of para-inWnitive in BP to create a situation by which either an inXected or a non-inXected inWnitive is found. Notice that this possibility should also account for optionality in EP. Traditional grammarians Wrst noticed optional inXection on the inWnitive. The examples go back to Archaic and Classical EP, as illustrated in (20) as opposed to (21), and in (22) as opposed to (23) (observation and examples from Said Ali 1921/2001: 252): (20)

a. Vy estes portugueeses asi revolver a lide e ferir ta˜ estranhamente I.saw these portuguese . . . revolve-inf . . . and injure-inf . . . (Nunes, Crestomatia Arcaica, 48) b. Vera´ brac¸os e pernas hir andando you.will.see arms and legs go-inf walking (Camo˜es, Os Lusı´adas, 5. 35)

Exceptional Case Marking

79

(21)

Vimos as Ursas, a pesar de Juno, banharem-se nas a´guas de Neptuno we. saw the Ursas in spite of Juno bathe-inf-pl in.the waters of Neptune (ibid.)

(22)

Quem te deu o direito de apagar no sangue who to.you gave the right to erase-inf . . . . . . (Herculano, Eurico, 183)

(23) Quem te deu, pois, o direito de correres a who to.you gave then the right to run.away -inf-2sg from morte certa? death (ibid.) Similar facts are quoted in Mateus et al. (2003: 648) with respect to modern EP. In fact, a context par excellence of optionality is perception (cf. (20) and (21)) and causative constructions (cf. (24)). As proposed in various analyses (cf. Kayne 2000b; Andrade 2002; Mateus et al. 2003), the non-inXected inWnitive clause involves exceptional case marking on the DP os alunos by the matrix verb, whereas nominative case is assigned by the inXected clause. (24)

O professor mandou os alunos entrar(em). the teacher made the students enter-inf-3pl

Optionality in (22) and (23) in turn seems to involve the alternation between control (non-inXected) inWnitives and inXected inWnitives, as in (5), repeated as (25): (25) Maria disse a`s crianc¸as para Ø estudar(em). Maria said to.the children for study-inf-3pl Recall that the construction with directive predicates of dezir ‘say’ as in (25) allows for another pattern, namely the already mentioned para-INF with an embedded overt subject (cf. (1), repeated as (26a)). In present terms, this construction should allow for an alternation between an ECM and an inXected inWnitive construction, depending on the dialect, which is borne out by the facts (cf. (26b)): (26) a. Maria disse pra mim estudar. Maria said for me study-inf b. Maria disse pras crianc¸as estudar(em). Maria said for the children study-inf-3pl Looking at (25) and (26), it is possible to point out another feature of BP as opposed to EP, namely the syntax of (dative) pronouns. As already mentioned, dative complements in BP are licensed by the preposition para

80

H. M. M. L. Salles

(cf. Gomes 2003; Berlinck 2000; Ramos 2002; Salles 1997; Andrade 2002; Salles and Scherre 2003; among others). Moreover, as pointed out in Bispo (forthcoming), pronominalization in BP is realized as a PP (¼ pra ele), as opposed to EP, in which (dative) clitics are found. The fact that datives in BP are realized as PP (not as a dative clitic) should then create the structural condition for reanalysis, as illustrated in (27a) as opposed to (27b): (27) a. Maria [disse [pro Pedro] [pra (ele) sair]] b. Maria [disse [pro Pedro sair]] Finally, the emergence of para-INF construction encoding irrealis modality in BP, as opposed to the auxiliary inversion subject construction encoding epistemic modality from EP, can be analysed as a process leading to reduction in structural complexity. As proposed in Roberts and Roussou’s (2003) theory of syntactic change, the simplicity metric establishes a correlation between structural complexity and (formal) feature syncretism. Accordingly, merge of a dedicated particle as C should be less syncretic than movement of the auxiliary to C.

5.4 Conclusion In this study, the (completive) subordination patterns of BP, as opposed to EP, were examined in a diachronic perspective. It was shown that BP developed a pattern in which the directive predicate dizer ‘say’ only takes para-INF clauses as its complements, whereas declarative and epistemic predicates only take Wnite clauses (with overt subjects) in the complement position—the latter fact further relating to the rigidiWcation of word order. In EP epistemic predicates take either indicative clauses or (inXected) inWnitives as their complements, whereas directive predicates take either a (control) inWnitive structure or a Wnite (subjunctive) clause. It was argued that the emergence of the novel directive construction in BP is independent of the presence or absence of inXection on the inWnitive. Instead, it is crucially determined by the erosion of the indicative and subjunctive morphological distinction (excluding Wnite (subjunctive) clauses in the complement position of directive predicates), and by the loss of auxiliary-subject inversion (excluding (inXected) inWnitive clauses in the complement position of epistemic predicates). The optionality of inXection on the inWnitive found in a number of constructions in EP was taken to indicate that the emergence of the novel para-INF (with an overt subject) in BP does not exclude the presence of inXected inWnitives in the grammatical system, the relevant conWguration that

Exceptional Case Marking

81

allows for both a nominative and an ECM subject. This is an interesting result, given the fact that a distinction between Standard and Popular BP along social classes, education level, etc. should be taken into consideration: while the absence of third person inXection is a feature of the popular variety, the complementation patterns described for BP apply to both dialects. In the present analysis, Roberts and Roussou’s (1999, 2003) theory of syntactic change, and in particular their approach to grammaticalization, was applied to the analysis of BP and EP data. It has been shown that the para-INF construction from BP emerges under reanalysis of para as a C head, a process that involves reduction in complexity, as compared to Aux-subject inversion, in which C is realized under movement.

6 Reanalysis and Conservancy of Structure in Chinese J O H N W H I T M A N A N D WA LTR AUD PAUL

6.1 Introduction The relegation of syntactic variation to lexical features under the Minimalist Programme has led to a growing awareness that syntactic change should be characterized in the same way: by changes in the discrete features of individual lexical items. This insight has given rise to the following distinctively minimalist typology of syntactic change: (1) a. Loss or gain of a feature driving movement (Roberts 1993c, 1997). b. Grammaticalization as shift ‘up the tree’ to a functional category (Roberts and Roussou 1999). c. Reanalysis as relabelling: lexical items change categorial or projection ([
max,
min]) features under preservation of hierarchical (c-command) relations (Whitman 2000). The objective of this chapter is to apply this typology to several well-known examples of syntactic change in Chinese. We show that earlier analyses exaggerate the scope of syntactic change in the long-documented history of Chinese languages. The changes that are in fact attested can be characterized as featural change, without rearrangement of hierarchical structure.

6.2 Received Views about Structural Change in Chinese A persistent misconception dating to the 1970s holds that Chinese languages have undergone a change in basic word order from OV to VO back to OV again (Li and Thompson 1974). Li and Thompson’s claim that Chinese was SOV before the tenth century bc has been decisively rejected by numerous specialists (Djamouri 1988; Shen 1992; Peyraube 1996), on the basis of extensive studies of the earliest attested texts (fourteenth to eleventh century bc).

Reanalysis and Conservancy of Structure

83

These studies clearly show that VO represents the main word order (94 per cent of 26,000 utterance tokens in the Shang bone inscription corpus studied by Djamouri 1988). Nevertheless, the original Li and Thompson view continues to inXuence even sophisticated general theorists (see, for example, Newmeyer 1998: 242). From a standpoint where all types of syntactic change result from discrete changes in the features of individual lexical items, such radical shifts in overall word order are unexpected. This chapter shows that they do not occur in Chinese. SigniWcant syntactic changes instead turn out to be cases of reanalysis which obey the conservancy of structure restriction proposed in (1c): they change categorial features and eliminate or ‘prune’ unmotivated structure, but preserve hierarchical c-command relations. 6.2.1 The great Mandarin ba hoax

Li and Thompson (1974) claim that modern varieties of Chinese are in the process of shifting to OV order. Their analysis of the preverbal object pattern with the ‘object marker’ ba (2) requires a change in the category of ba and a fundamental change in clause structure.1 (2) Ta ba Lisi paoqi -le. 3sg ba Lisi abandon -perf ‘She abandoned Lisi.’

(Modern Mandarin)2

The generally accepted idea is that ba (etymologically a verb meaning ‘take, seize’)3 originally occurred in a serial verb construction of the object-sharing type (3). 1 It is important to point out that the ba construction is not comparable to the obligatory object shift in e.g. Scandinavian languages which is contingent on verb raising (cf. Holmberg 1986; Ferguson 1996; among others), because deWnite DPs, proper names, and pronouns may likewise occupy the canonical postverbal object position (i). Furthermore, ba appears to the right of negation (ii):

(i) Ta paoqi -le Lisi / wo-de pengyou / wo. 3sg abandon -perf Lisi / 1sg-sub friend / 1sg ‘She abandoned Lisi / my friend / me.’ (ii) Ta mei ba Lisi / wo -de pengyou / wo paoqi. 3sg neg ba Lisi / 1sg-sub friend / 1sg abandon ‘She has not abandoned Lisi / my friend / me.’ For an extensive discussion of the ba construction in Modern Mandarin, cf. Li (2001). 2 The following abbreviations are used in glossing examples: CL classiWer; NEG negation; PERF perfective aspect; PL plural (for example, 3PL ¼ third person plural); SG singular; SUB subordinator. 3 Ba is one of a pair of original verbs including jiang ‘take’ and chi ‘hold’ that underwent parallel developments; specialists in Chinese historical syntax generally treat them together (Wang Li 1958/ 1988: ch. 47; Zhu Minche 1957; Peyraube 1996: 168). While ba survives in the object marking function in Modern Mandarin, jiang is its counterpart in more formal registers of Minnanhua (Taiwanese) and Cantonese.

84 (3)

J. Whitman and W. Paul Sunzi jiang yi -ya si yong. Sunzi take one -duck privately use ‘Sunzi grabbed a duck and used it himself.’ (Zhang Zhuo, Chao ye qian zai, eighth century; from Zhu Minche 1957: 18)

Li and Thompson claim that the shared object was reanalysed as the object of the second verb and ba was reanalysed as a preposition, giving the structure in (4) for (2). (4) [[VP [PP ba Lisi] [VP paoqi -le]] ba Lisi abandon-perf To grasp the scope of this alleged reanalysis, we must Wrst provide a structure for the source serial construction in (3). It is widely agreed (cf. Zhu 1957: 24) that at the onset of the grammaticalization of ba and jiang in the sixth to eighth centuries, they appeared in two distinct constructions, the object sharing construction in (3) and the instrumental construction in (5), which does not involve object sharing: (5)

Qing jiang yu -zhang qiao hua -pian. lightly take jade-stick tap Xower -petal ‘She lightly tapped on the Xower petals with a stick of jade.’ (Zhang Hu, Gong zi xing, ninth century; from Wang Li 1958/1988: ch. 47, p. 539)

Traditional analyses have not posited a structural distinction between these two constructions. Instead, they typically posit a coordinate structure for both (Peyraube 1985: 208; Cui 1984) or an adjunction structure for both (Zhu 1957; Wang Li 1958/1988: ch. 47). But two diVerent structures are necessary to explain the presence of object sharing in (3) and its absence in (5). A straightforward way of doing this is to adopt a VP complementation analysis (Larson 1991; Collins 1993, 1997) for the object sharing construction in (3): [VP2 pro yong]]]]] (6) [v 0 ba [VP1 yi -ya [v10 tba [VP2 si take one -duck privately use In (6) ba (still with its lexical meaning ‘hold, take’) takes the second VP headed by yong ‘use’ as its complement. The shared object ya ‘duck’ is merged in the speciWer of the VP headed by ba and controls pro in the complement VP. Ba moves to v (and possibly higher; cf. Paul 2002), deriving the surface order.

Reanalysis and Conservancy of Structure

85

The source of the instrumental construction will be an adjunction structure as in (7). In this structure, there is no control relation between the object of the adjoined VP1 and the object of VP2 . (7) [qing[VP2 [VP1 jiang yu -zhang ] [VP2 qiao hua -pian]]] [ ¼ (5)] lightly take jade-stick tap Xower-petal All researchers agree, following Zhu (1957), that the source for the modern ba construction is the object sharing pattern in (3) (the instrumental pattern with ba is lost in Modern Mandarin). The reanalysis alleged by Li and Thompson posits a radical restructuring if the source structure is taken to be the VP complementation structure (8a) we have argued for in (6), or the symmetric coordinate VP structure (8b) assumed by Peyraube (1985): (8) a. [vP ba [VP1 NPobj [V10 tba [VP2 pro V2 ]]]] > [[VP [PP ba NPobj ][VP V ]] > [[VP [PP ba NPobj ][VP V]] b. [VP1 ba NPobj ] [VP2 V2 pro] In both (8a) and (8b) VP1, headed by ba, must be rebracketed as an adjunct of VP2, whose head then becomes the main verb. Such a restructuring would radically alter the hierarchical relations of the source structure.4 The radical restructuring in (8) is a by-product of the traditional analysis of modern ba as a preposition (Li and Liu 1955; Chao 1968; Huang 1982; Peyraube 1985, 1996; Li 1990). However, this analysis is rejected in recent work. There is a widespread consensus that ba is best viewed as the head of a functional projection above VP (Sybesma 1992, 1999; Zou 1993; Li 2001; Paul 2002).5 On this view, ba continues to take the projection of the verb to its right as its complement. The correctness of this analysis is clearly shown by the possibility of conjoining two occurrences of preverbal object plus VP under ba. (9)

4

Mama ba [[di ca -le you ca ] [zhuozi ma -le you ma]]. mom ba Xoor scrub -perf again scrub table wipe -perf again wipe

If both the original ba serial construction and the modern ba construction are analysed as adjunction structures, then no modiWcation of hierarchical structure, only relabelling from VP to P, is involved in the development of the modern structure. However, this analysis raises other questions: how object sharing is accounted for in the original serial structure, and how movement is possible in the modern ba construction, since an object moved into an adjunct headed by ba would not c-command its trace. 5 The analysis of ba as a non-prepositional head goes back to Hashimoto (1971), who analyses it as a verb. Hashimoto proposes a ternary branching structure where ba takes both an NP and a clause as its complements. Ross (1991), Chen Xilong (1993), and Bender (2000) basically follow Hashimoto’s analysis. Crucially, under this analysis, the NP following ba is not contained in the complement VP of ba; accordingly the coordination data in (9)–(10) cannot be accounted for.

J. Whitman and W. Paul

86

‘Mom again and again scrubbed the Xoor and again and again wiped the table.’ (Wu Meng 1982: 434)6 (10)

Wo ba [[Zhangsan jieshao gei Lisi], [Wangwu jieshao gei Laoli]].7 1sg ba Zhangsan introduce to Lisi Wangwu introduce to Laoli ‘I introduced Zhangsan to Lisi, and Wangwu to Laoli.’

This fact would be diYcult to explain if ba and the immediately following NP formed a constituent, as the prepositional adjunct analysis of ba in (4) holds. The behaviour of ba in (9)–(10) contrasts sharply with (11), where a true adjunct PP is involved: (11)

Wo [dui Wangwu] hen you yijian, [*(dui) Laoli] ye hen I towards Wangwu very have prejudice towards Laoli also very you yijian. have prejudice ‘I am very prejudiced against Wangwu, and also against Laoli.’

(11) is totally ungrammatical without the second occurrence of the preposition dui ‘towards’. A further argument against the prepositional analysis of ba is its inability to appear as a modiWer of a relational DP, in contrast with prepositions such as dui ‘towards’:8 6 Wu Meng (1982) explicitly cites (9) as problematic for the alleged prepositional status of ba; as he observes, no other preposition takes discontinuous complements. Wu attempts to account for (9) by appealing to the verbal origin of ba, an insight which is basically correct. 7 We owe (10) and the argument based on it to Thomas Ernst, who attributes it to Audrey Li. The acceptability of (10) in itself is, however, not conclusive, because the example can also be parsed as containing two conjoined clauses, the second of which is a topic-comment structure:

(i) [Wo ba Zhangsan jieshao gei Lisi], [[topic Wangwu]i [pro jieshao ti gei Laoli]]. 1sg ba Zhangsan introduce to Lisi Wangwu introduce to Laoli Adding an adverbial phrase like zui hao makes the parsing of the second conjunct as a topic-comment structure impossible: (ii) Ta zui hao [ba [[Zhangsan jieshao gei Lisi], [Wangwu jieshao gei Laoli]]]. 3sg most good ba Zhangsan introduce to Lisi Wangwu introduce to Laoli ‘He’d better introduce Zhangsan to Lisi and Wangwu to Laoli.’ These examples conWrm that ba is a higher head which can take two conjoined VPs as its complement. Naturally, it is also possible to conjoin two projections headed by ba: (iii) Ta zui hao [ba [Zhangsan jieshao gei Lisi]], [ba [Wangwu jieshao gei Laoli]]. 3sg most good ba Zhangsan introduce to Lisi ba Wangwu introduce to Laoli ‘He’d better introduce Zhangsan to Lisi and Wangwu to Laoli.’ 8

Prepositions are not allowed as modiWers of non-relational nouns:

ta] de hua (i) *[PP dui towards 3sg sub word ‘the words for him’

Reanalysis and Conservancy of Structure (12)

87

a. [Zhangsan [PP dui /*ba zhei-jian shi] de anpai ] bu Zhangsan towards / ba this-cl matter sub arrangement neg tuodang. suitable ‘Zhangsan’s arrangement of this matter is not suitable.’ b. Zhangsan [dui zhei-jian shi ] ba xijie dou anpai -hao -le. Zhangsan towards this-cl matter ba detail all arrange -good -perf ‘Zhangsan with respect to that matter arranged all the details.’ (Fu Jingqi, p.c.)

As (12) illustrates, ba can introduce the object DP of the verb anpai in a verbal projection (12b), but not in its nominal counterpart (12a). The unacceptability of (12a) also shows that ba as a higher verbal head has to be distinguished from lexical verbs, since verbs such as the relative of ba (descendent from the same root), meaning ‘guard’, can head relative clauses as nominal modiWers: (13)

men ] de nei -ge ren] shui jiao le. [DP [IP Ba guard door sub that -cl person sleep sleep part ‘The person who guards the door has fallen asleep.’

Last, but not least, since ba and the following NP do not form a constituent, they cannot be topicalized to the left of the subject as PPs can: (14) a. Gei Mali, wo (gei Mali) zuo -le hunduntang, gei Amei, for Mary 1sg for Mary make -perf wonton.soup for Amei wo (gei Amei) zuo -le chaomian. 1sg for Amei make -perf fried.noodles ‘For Mary, I made wonton soup, for Amy, fried noodles.’ b. (*Ba shu), ni keyi ba shu fang zai zhuozi -shang (*ba dayi), ba book 2sg can ba book put at table -top ba coat ni keyi ba dayi fang zai chuang -shang. 2sg can ba coat put at bed -top ‘The books, you can put on the table, the coat, you can put on the bed.’ (Paul 2002: 164)

(ii) [IP wo[PP dui ta] shuo de hua] 1sg towards 3sg speak sub word ‘the words I spoke to him’

88

J. Whitman and W. Paul

(15) Youju, cong zher, ni wang nan qu. post.oYce from here 2sg toward south go ‘The post oYce, from here, you go south.’ (Lu¨ et al. 1980/1995: 130) The analysis of the ba construction in Modern Mandarin that we propose shares the basic assumptions of previous analyses of ba as a higher head. Ba does not assign a thematic role to the preverbal object, and the surface position of the object is derived by movement.9 (16)

vP NP

[=(2)] v


v ba

baP Lisi

ba
tba

vP v

VP

paoqi -le

tLisi

The surface order of constituents is derived by movement of the object to the speciWer of ba and movement of ba to v.10 A concrete argument for the movement analysis is the observation that VP-level adverbs such as manner adverbs and frequentatives like you ‘again’—cf. (9) above—may occur between the preverbal object and the verb: (17) Ta (henxinde) ba Zhangsan (henxinde) paoqi -le. 3sg cruelly ba Zhangsan cruelly abandon -perf ‘She heartlessly abandoned Zhangsan.’ (slightly changed example from Tang 1990: 145) This would be diYcult to explain if the object had not moved into a position above VP. The same fact obtains already in late medieval ba sentences such as (18): (18)

Du ba Liangzhou fan ji pai. alone ba Liangzhou melody several.times play ‘Alone, I’ll play the Liangzhou melody several times.’ (Gu Kuang shi, eighth century; from Zhu Minche 1957: 28)

9 Carstens (2002) proposes a movement analysis for similar serial constructions in Yoruba and other Niger-Congo languages. 10 (16) is diVerent from Li’s (2001: 40–1, example 92) analysis where ba as the head of BaP stays in situ and takes a vP complement, with the direct object in its speciWer; on this view, adverbs to the right of ba, as in (17), must be adjoined to v’, a rather implausible assumption.

Reanalysis and Conservancy of Structure

89

At the same time, ba must be lower than the position occupied by modal and tense auxiliaries, as evidenced by the position of bare NP adverbs (cf. Li 1985: 373): (19) Ta zuotian ba Zhangsan (* zuotian ) paoqi -le. 3sg yesterday ba Zhangsan yesterday abandon -perf ‘She abandoned Zhangsan yesterday.’ (20) Ta (mingtian) yao (mingtian) qu Beijing. 3sg tomorrow want tomorrow go Beijing ‘He wants to go to Beijing tomorrow.’ While the auxiliary can occur to the left and the right of mingtian ‘tomorrow’, ba is restricted to the right of zuotian ‘yesterday’, which indicates that its position is below that of modal and tense auxiliaries. In (16) we locate ba in vP. Above we have shown that the Modern Mandarin ba construction retains the same basic hierarchical relations that its ‘ancestor’ structure had in Middle Chinese. Two changes have, however, taken place in the development of the modern structure. First, ba no longer assigns a thematic role to a complement NP; instead, the NP in the speciWer of ba, Lisi in (16), moves to that position. Second, the adverb placement data in (17) and (18) show that ba originates in a functional head position higher than lexical verbs. Reanalysis of ba thus exempliWes the Wrst of two types of featural change introduced in Section 6.1: reanalysis moves ba ‘up the tree’ to the position of a functional head (Roberts and Roussou’s characterization of grammaticalization). Crucially for the argument of this chapter, reanalysis of ba obeys the basic conservancy of structure constraint for feature-based reanalysis in Whitman (2000): c-command relations between ba and other components of the source structure are retained in the output of the change.

6.3 V > P Reanalyses A well-known pattern of categorial relabelling is change from V to P. In Chinese as well, this type of change is richly attested. Mandarin cong ‘from’, for example, functions as the head of a preverbal PP: (21)

Wo gang [PP cong nongcun ] hui -lai. 1sg just from village return -come ‘I have just come back from the village.’ (Lu¨ et al. 1980/1995: 130)

90

J. Whitman and W. Paul

The basic conWguration of (21), where cong heads an adjunct projection preceding the main verb, dates back at least to the Warring States period, as shown by (22) from the Zuozhuan.11 Cong at this time also occurs as a main verb meaning ‘follow, pursue’: (22) [Cong tai -shang] tan ren. from platform -upon shoot people ‘He shot people from up on the platform.’ (Zuozhuan: Xuangong 2, Wfth to third century bc) (23)

Xia, zhuhou zhi daifu [cong Jin hou ] fa Qin. summer feudal.lord sub high.oYcial follow Jin duke attack Qin ‘In summer, the high oYcials of the feudal lords, following the duke of Jin, attacked Qin.’ (Zuozhuan: Xianggong 14, Wfth to third century bc)

(23) illustrates the etymological source of cong. In Shang inscriptions (fourteenth to eleventh century bc), cong occurs only as a verb meaning ‘follow, pursue’ (R. Djamouri, p.c.). Note that it is diYcult to specify exactly when relabelling of cong from V to P took place (cf. Ohta 1958). The interpretation of the adjunct headed by cong in (21) shows semantic ‘bleaching’, familiar from the grammaticalization literature, but independent syntactic tests are required to demonstrate an actual change in category. We will cite here one piece of evidence for its prepositional status (see McCawley 1992 for additional arguments), viz. the unacceptability of [cong NP] as a nominal modiWer (recall that only relational nouns allow modiWcation by a PP in modern Mandarin—cf. (12) above): (24) a. *[DP [PP cong Beijing] de xuesheng ] from Beijing sub student ‘the student from Beijing’ b. [DP [IP [PP cong Beijing ] lai ] de xuesheng] from Beijing come sub student ‘the student who comes from Beijing’ Thus at some point (probably well) prior to Modern Mandarin, we must posit the change in categorial feature in (25): (25)

[VP cong NP] > [PP cong NP] [..v..] [..p..]

11 According to Ohta (1958/1981: 252), adjunct use of cong dates back to Old Chinese, primarily in the sense of ‘from (location)’ rather than ‘from (time)’.

Reanalysis and Conservancy of Structure

91

The relabelling in (25) relates (21) and the source pattern in (23), where cong occurs as the verbal head of an adjunct preceding the main verb. Hierarchical relations in the overall clause are retained. Cases of this sort exemplify ‘pure’ relabelling (1c), where change in categorial feature appears to be unaccompanied by any ‘shift up the tree’ in the sense of Roberts and Roussou (1999). Relabelling in (25) maintains the hierarchical relations among constituents retained after the change, but it is likely that reanalyses of this pattern involve some elimination of structure. To take a concrete example, given that congphrases like (23) are transitive, they must be associated with an external argument position, as in (26): (26) Xia, zhuhou zhi daifu [vP [vP PRO[VP cong Jin hou]] summer feudal.lord sub high.oYcal follow Jin duke Qin]]. [vP fa attack Qin ‘In summer, the high oYcials of the feudal lords, following the duke of Jin, attacked Qin.’ In (26) the external argument is represented as a PRO bound by the matrix subject. In current terms, the projection immediately dominating PRO is a higher category, such as vP. PPs headed by cong, in contrast, give no evidence of being associated with a subject position; in an example like (27) below, as in (15), there is no obvious controller for such a position. (27)

Qian pai, cong zuo-qi di si ren jiu shi ta. front row from left-onward number 4 person then be 3sg ‘Front row, the fourth person from the left is him.’ (Lu¨ et al. 1980/1995: 130)

If it is the case that PPs are not associated with a subject position, we need a mechanism to bring about the requisite ‘shrinkage’ in VP > PP reanalyses. Let us assume that the crucial step of reanalysis is relabelling. Once the head is relabelled with the categorial feature [p], selectional properties ensure that it may no longer be selected by v. This in turn ensures elimination of the external argument position, as shown schematically in (28): (28)

cong NP cong [..v..]

cong

cong NP

>

cong [..p..]

NP

92

J. Whitman and W. Paul

We dub such reduction of structure ‘pruning’, applying Ross’s (1967) term for deletion of excrescent structure to a diachronic change. Pruning eliminates structure, but does not change the c-command relations of the material that is retained, thus it satisWes the conservancy of structure constraint (1c). In the grammaticalization literature, relabellings of this type have typically been discussed completely apart from the syntactic environment where they occur. In the framework adopted here, however, it is predicted that only verbs in certain positions may undergo relabelling to P. Consider the extended VP structure in (29): (29)

[vP [XP ::VP1 ::][vP v[VP2 NP[V0 V2 [VP3 NP[V0 V3 Complement]]]]]]

In (29), VP1 (perhaps contained in a larger projection: vP, TP, or CP) is a vP-level adjunct. We have seen that V > P reanalysis of cong ‘from’ (< ‘follow’) conserves hierarchical relations in this context.12 Other instances of V > P reanalysis in Chinese of which we are aware—gen ‘with’ (< ‘accompany’), dui ‘towards’ (< ‘face, be opposite’), gei ‘to, for’ (< ‘give’)—are restricted to verbs which in the history of Chinese occur in this preverbal position.13 To recapitulate, V > P reanalysis occurs in Chinese in precisely the environment where relabelling preserves overall hierarchical structure. Returning to (29), VP2 represents the ‘main’ VP in the extended VP structure. Under the conservancy of structure constraint, the verb in this position is predicted not to be able to undergo V > P reanalysis. The resultant structure would be ill-formed: prepositions cannot support Tense, nor are they selected by v. It might appear that the non-existence of main verb > P reanalyses is simply taken for granted in the grammaticalization literature. But as we saw in Section 6.2, if the prepositional analysis of Mandarin ba generally assumed in this literature were correct, this would be precisely a case of a main verb that undergoes reanalysis as a preposition—like V2 in (29), i.e. the initial verb in an object sharing serial construction. In fact the basic development represented by ba is not an isolated one within the history of Chinese. The evolution of the passive marker bei is

12

Note that in these and many other cases (homophonous) verbs and prepositions coexist in Modern Mandarin (the same holds for the functional head ba (3rd tone) and the verb ba (3rd tone); cf. (13) above). This considerably weakens the common idea that the loss of the ‘original’ lexical item is decisive for grammaticalization to take place (see Longobardi 2001b) and that ‘functional elements (tend to) fall below the prosodic threshold for wordhood’ (Roberts and Roussou 1999: 1025). 13 For reasons of space, we cannot discuss cases attested in Chinese of V > P reanalysis in the complement position, i.e. VP3 in (29). We argue that reanalysis in this position as well observes the conservancy of structure constraint (1c).

Reanalysis and Conservancy of Structure

93

structurally parallel. Bei Wrst occurs as the main verb, V2 in (29), in a construction meaning ‘NP[VP2 suff ers V3 ]’: (30)

Cuo zu yi bei lu. Cuo Wnally therefore suVer slaughter ‘Therefore, Cuo Wnally suVered being slaughtered.’ (Shiji: Cu li lie zhuan, Wrst century bc; from Peyraube 1996: 176)

Peyraube (1996: 176) shows that bei Wrst occurs in this pattern without an overt agent associated with the complement VP; it is only towards the end of early medieval Chinese (second to sixth century) that the pattern ‘bei agent verb’ is attested: (31)

Liangzi bei Su Jun hai. Liangzi bei Su Jun kill ‘Liangzi was killed by Su Jun.’ (Shi shuo xin yu: Fang Zheng, Wfth century; from Peyraube 1996: 176)

For Modern Mandarin, it has been widely claimed that—unlike the immediately preverbal bei, cf. (30)—the bei followed by an agent NP has been reanalysed as a PP (see Bennett 1981; McCawley 1992: 225; Peyraube 1996: 177; Ma Beijia 2003).14 If correct, this analysis would posit exactly the kind of main verb > P reanalysis that we have been arguing does not occur. And once again, there is strong evidence that the PP analysis of ‘bei NP’ is wrong. Hashimoto (1988: 2.1.1) uses the same type of coordination argument that we presented for ba in (9)–(10) to show that bei is a higher head (for Hashimoto, a verb taking a clausal complement, with the agent NP as its subject or topic, respectively): (32) Ta bei [[qinren huaiyi] [wairen zhize]]. 3sg bei relatives suspect stranger criticize ‘He is not only criticized by strangers, but suspected even by his own people.’ (Hashimoto 1988: 332 (9)) Like ba in (9)–(10), bei takes a coordinated VP complement in (32); bei and the immediately following agent NP do not form a constituent.15 (32)

14 The status assigned to the verb-adjacent bei diVers (but is never that of a preposition). Peyraube (1996) considers it an auxiliary; for McCawley (1992), bei plus verb is a compound. 15 Also cf. Huang (1999), who analyses the bei in the agent passive as an experiential verb taking a clausal complement and the bei in the agentless passive as an experiential auxiliary with a VP complement.

94

J. Whitman and W. Paul

neatly contrasts with (33), where the preposition is required in both VP conjuncts: (33)

Ta[[[PP zai xuexiao] xuexi] [[PP (zai) jiali] xiuxi]]. 3sg at school study at home rest ‘He studies at school and rests at home.’ (Hashimoto 1988: 332 (10))

In the case of both ba and bei, original main verbs undergo substantial ‘bleaching’. Nevertheless, their relative structural position remains unchanged, as predicted by (1c). In sum, conservancy of structure predicts that verbs in the VP1 (adjunct) position may undergo relabelling to P, as is richly attested in Chinese. Verbs in VP2 (main verb) position may not, and such cases are not attested in Chinese.

6.4 Conclusion Specialists in Chinese historical syntax have been aware for some time that the radical word order shifts posited for this group of languages by researchers in the 1970s are simply not supported by the historical data. In this chapter we show that the changes in the extended VP syntax that are attested maintain the basic hierarchical structure of much earlier stages, and fall into just a few basic patterns: V > P reanalyses in adjunct position, and categorial shift of ‘main’ verbal heads to functional status.

7 DeWnite Reference in Old and Modern French: The Rise and Fall of DP PAUL B O UCHER

7.1 Introduction This chapter addresses the question of the emergence and evolution of the deWnite article in French. This problem is linked to the expression of ‘deWniteness’ as well as to the nature and evolution of ‘functional’ categories in a grammar. We will be interested in two parallel evolutions: the loss of bare NPs1 with speciWc reference in French and the grammaticalization of demonstrative pronouns, which could be adjoined to bare NPs in Late Latin, GalloRomance, and Early Old French, were reanalysed at some point as functional D heads, and have evolved in Modern French toward the status of ‘left-edge inXections’ (Miller 1991). Though not strictly simultaneous, these two processes turn out to be interdependent. As French nouns lost overt case and number morphology, they also lost the ability to be used alone as referential arguments of a verb. Likewise, as pronouns lost substantive content, they were reanalysed as functional heads and eventually replaced number and case aYxation altogether. The grammaticalization of D-words will thus be analysed here as an evolution from XP to Xo status, due to the erosion of interpretable semantic features on the root node. In a similar fashion, the loss of referential bare NPs will be analysed as the loss of uninterpretable grammatical features on the complement node of N.2 DeWnite reference will 1 In what follows the term ‘bare NP’ (BNP) will refer to referential NPs not accompanied by a determiner; ‘bare N’ (BN) will refer to predicative, i.e. non-referential noun heads, as in ‘They elected John president’, or ‘Nous allons faire aVaire’. BNPs are usually counted as plural or mass singular nouns, more rarely count singular. The term ‘bare plurals’ (BP) is often used in the literature (DelWtto and Schroten 1991, for example) to refer to BNPs. 2 The terms ‘root node’ and ‘complement node’ are used in the sense of Kerstens (1993) and will be deWned below.

96

P. Boucher

be analysed in terms of a modiWed version of DelWtto and Schroten’s (1991) Restricted QuantiWcation Condition, which posits that for existential reference to occur, two ‘logically distinct’ positions must be Wlled, respectively providing the ‘quantiWer’ and the ‘domain of quantiWcation’. In this framework, the expression of deWniteness can be formally related to the interaction between D and case and number morphology on N. This interaction will be shown to account for diVerences between Old and Modern French, as well as between Romance and Germanic and between French and the other Romance languages. The chapter will be organized as follows. In Section 7.2 I examine the problem of bare NPs and bare (pronominal) determiners (BD) in Romance and Germanic and discuss the Restricted QuantiWcation Condition, as well as the representation of morphological complexity developed in Kerstens (1993). In Section 7.3 I look at data on NPs in Old French between the ninth and fourteenth centuries. In Section 7.4 I will return to the reanalysis of D-words in OF and to the problem of case projections in Romance and Germanic.

7.2 The Expression of DeWniteness 7.2.1 The DP hypothesis

It is often assumed that referential NPs are complements of a D head within a DP projection (Abney 1987). The D node has come to be associated with the notion of (in)deWnite reference, though in many cases this seems to be little more than a stipulation. Chomsky (1995: 240) draws an explicit parallel between the three functional heads T, D, and C, opposing them to Agr: ‘T, D and C have semantic properties; Agr does not. ( . . . ) D may be the locus of what is loosely called ‘‘referentiality’’. . .’. It is not clear, however, just how being the ‘locus of referentiality’ should be represented formally, other than through the adjunction of some lexical item—a noun or a determiner—to a D-morpheme, along the lines of adjunction of a verbal head to a Tense morpheme under T or to a Q morpheme under C. This is essentially the line of reasoning adopted by Longobardi (1994), for instance, who further argues that a nominal expression can only attain argument status if it is dominated by some (overt or covert) D. However, there are many languages without determiner systems at all and yet these languages manage to compute deWniteness with little or no ambiguity. Turkish, for instance, has no article system and though it does possess a singular indeWnite article, bir, Knittel (2002) argues convincingly that this does not dominate NP and so cannot be considered to be the head of a DP projection. Nonetheless, Turkish can express four levels of deWniteness through combinations of number aYxation,

The Rise and Fall of DP

97

modiWcation and quantiWcation, verb agreement, and case marking. Likewise, Russian uses a combination of word order, prosody, case oppositions, aspect oppositions, and verb agreement to mark deWniteness in the absence of a determiner system (Paykin and Peteghem 2002). Many other cases come easily to mind: Chinese, Finnish, Classical Latin, Old Germanic, and so on (see Philippi 1997 and especially Lyons 1999 for discussion). I will thus adopt the position here that if a language has no determiners then it has no DP projection. 7.2.2 Number and case

Recent work on referential BNPs (DelWtto and Schroten 1991; Schroten 1991, 2001; Dobrovie-Sorin 2001; among others) has called the DP hypothesis into question. Dobrovie-Sorin (2001: 208) for instance has proposed that Bare Nouns do not necessarily have extended DP projections with a covert determiner adjoined to D, and that argument (versus predicate) status may be derived directly from the semantic nature of the N head. DelWtto and Schroten (1991: 156–7) also argue against solutions using covert determiners and propose instead a Restricted QuantiWcation Constraint (1) to account for BNPs in Germanic languages, adding the proviso in (2). (1) The Restricted QuantiWcation Constraint (RQC): the existential interpretation of DPs is only available if two ‘logically’ diVerent elements are present, Wlling distinct syntactic positions and respectively providing the domain of quantiWcation and the ‘quantifying in’ operator. (2) This universal condition on restricted quantiWcation can be satisWed by BNPs only if the ‘number aYx’ which is attached to the noun at S-structure is permitted to raise to D (i.e. to the operator position) at Logical Form (LF). In other words, BNPs may have an existential interpretation only if they are inXected for number and if the nature of the number aYx and the nominal root is such that the number aYx can ‘excorporate’ and raising can take place at LF. For this to occur, the number aYx must Wrst be strong. Secondly, the nominal must be ‘free’ (i.e. not bound) in order to avoid a violation of Rizzi’s Restricted Minimality condition. I will adopt the spirit of the RQC here, but will introduce two modiWcations: (a) ‘excorporation of the number aYx at (LF)’ will be replaced by ‘raising in the syntactic component for checking purposes’ (3); (b) ‘strong number aYxation’ will be considered a necessary but not suYcient condition for using BNPs in argument positions. Crucially, I will distinguish between BNPs in subject as opposed to object position (4).

98

P. Boucher

(3)

The RQC can be satisWed by BNPs only if NP is dominated by an extended number projection and if number aYxation is strong. (4) BNPs may be used in subject position only in languages where NumP is dominated by K(ase)P.

The reasoning behind (3) and (4) goes as follows. First of all, in languages without determiner systems the role of both number and case morphology is crucial. In Turkish (Knittel 2002) for instance, BNs can be used in both object and subject position. These have neither number nor case morphology and cannot be modiWed or quantiWed. Logically enough, they can only refer to kinds, never to individuals. Existential NPs (i.e. BNPs in our sense) must be dominated at least by NumP; speciWc, deWnite NPs must also be dominated by K(ase)P. Secondly, to return to the languages under investigation here, Modern French (ModF), which disallows BNPs in all argument positions, has lost both case and number morphology. Other Modern Romance languages, like Spanish or Italian, allow BNPs in object but not subject position. Contrary to French, they have never had case aYxation but, conversely, have maintained strong number aYxation. Old French (OF) and Germanic allow BNPs in both positions and had or still have both case and number morphology. The following examples illustrate the paradigm. (1) (2) (3) (4) (5) (6) (7)

Students occupied the building *Etudiants ont occupe´ le baˆtiment (French) *Estudiantes han ocupado el ediWcio (Spanish) Yesterday I saw students *Hier j’ai vu e´tudiants (French) Ayer encontre´ estudiantes (Spanish) Galois sont tuit par nature plus fol que bestes an pasture Gauls are all by nature more crazy than beasts in pasture ‘Gauls are all naturally crazier than beasts let out to pasture.’ (Foulet 1928) alumer Wrent (8) Mirre e timonie i Myrrh and incense there made-they burn ‘They burnt myrrh and incense.’ (Epstein 1995) (9) Hom qui traı¨st altre, nen est dreiz qu’il s’en vant man who betrays others not-of-it is right that-he self-of-it boasts ‘A man who betrays his fellows should not boast of it.’ (Foulet 1928)

The Rise and Fall of DP

99

The conclusion I draw from this is that only languages with case projections allow NP to attain a suYcient degree of speciWcity to be used in subject position. In the framework of Diesing’s (1992) Mapping Hypothesis, subject BNPs will be analysed here as ‘strong’ or ‘presuppositional’ quantiWers, which always map onto the ‘restrictive clause’, or IP. Object BNPs will be analysed as ‘weak’ or ‘cardinality’ quantiWers, which must remain under VP, where they function like variables inside the ‘nuclear scope’. In the next subsection, I will describe how case and number correlate with deWniteness to assure this. 7.2.3 Lexical and functional categories

Following Abney (1987: 309), I will consider that lexical items, that is, members of the categories N, V, A, P, are Wrst order predicates having scope respectively over objects, situations, attributes, and locations, the union of which corresponds to the universal set of individuals. Each lexical head therefore projects a ‘pure’, semantic projection and assigns one or more theta roles. All nouns predicate a quality, for instance the quality of being a table or a man, of some referent, which we will call R (Higginbotham 1986) and place in Spec,NP: (10) [NP R[N0 N . . . I will also assume that ‘semantic’ functional heads, in the sense of Chomsky (1995: 240), that is T, C, D, and so on, which I will group together under the heading ‘operators’, must bind an argument position in the lexical projection they c-command. For instance, D binds R(eferent) in the NP projection, T binds E(vent) in the VP projection, and so on. The classic means of binding R is to extend the NP projection to DP with a lexical determiner in D, which c-commands R. But in the case of BNPs, other extended functional projections may do so in the absence of DP. Let’s Wrst reduce DelWtto and Schroten’s RQC to its essential claim, which is that for there to be reference, two distinct positions must be Wlled: there must be a ‘quantiWer’ and a ‘domain of quantiWcation’. If there is no overt determiner at S-structure, then, they argue, there must be covert movement of the NumþN unit at LF, in order to satisfy the RQC. I claim that it is not necessary to resort to excorporation or covert movement at LF. It suYces to say that any referential NP must be dominated by at least NumP, which I will include in the closed set of ‘operator’ heads. In languages like English, with strong number morphology, the noun raises to Numo in the syntax. In this case a chain is formed by N-raising: the raised noun binds its trace/copy, which can now function as a variable since N has

100

P. Boucher

attained ‘operator’ status by adjoining to Num, and this licenses R. R is in a Spec-Head relationship with No and the semantic features of the noun must be checked against those of R for ‘predication’, which I am assimilating here to a form of u-role assignment, to ensue. The RQC will be satisWed since, by raising to Numo, a ‘semantic’ functional head, thus an operator in the sense deWned above, N can function as the ‘quantiWer’ and, by being licensed, R can constitute the ‘domain of quantiWcation’. (11)

[NUMP [NUM0 [Ni -Num][NP R[N0 ti ]]]]

In languages like French with weak number morphology, the noun cannot raise to Numo, no chain can be formed, so no operator is present, R cannot be licensed, and the RQC cannot be satisWed. An overt determiner must be present in Do in this case for the No position to be bound and R to be licensed. This gives us straightforwardly the diVerence between English (and other Germanic languages), where BNPs may be used in both subject and object positions, and Modern French, where BNPs may not be used in either position. But what about the other Romance languages where BNPs may be used in object but not subject position? This can be related to the fact that Germanic has retained a case projection, while the Romance languages other than French have not, as we will see below. For a BNP to be used in the subject position, I argue that the noun must adjoin not only to Numo, but also to Ko . Adjunction of N to Num allows NP to function like a ‘weak’ determiner in Diesing’s sense, that is, like a cardinality predicate, which can bind an indeWnite or non-speciWc variable. This forces it to remain in object position, i.e. under VP. In order to raise out of VP to Spec,IP and become a subject, N must be attached to Ko (if no determiner is present) in a language which overtly expresses ‘subject case’. This is, I assume, because subject case contains a ‘topicalization’ feature, and by adjoining to a K with this feature, N becomes a ‘presuppositional’ or ‘strong’ determiner, again in Diesing’s sense. This, as will be demonstrated below, is exactly the situation in Germanic and Old French. This analysis has several important formal consequences: (a) it eliminates the need for covert movement at LF, excorporation, empty D heads, and other such formal paraphernalia; (b) it allows QuantiWer Raising (QR) to be calculated directly in the syntactic component; (c) it relates the diVerences noted above between Old and Modern French, French and other Romance languages, Romance and Germanic, straightforwardly to the presence or absence of (strong) number and case morphology. This allows direct mapping from morphology to syntax to semantics.

The Rise and Fall of DP

101

7.2.4 From morphology to semantics

I will posit that referential BNPs and BDs (pronominal determiners) have in common a certain degree of morphological complexity. Following Kerstens (1993), I will represent this as a branching structure—[x0 [x y ]]—with interpretable semantic features located on the root node and uninterpretable grammatical features on the complement node. (12)

Det: [D0 [D
Deix,
Def [agr
Nom,
Fem,
Plur]]]

(13)

N: [N0 [N
Anim,
Count [agr
Nom,
Fem,
Plur]]]

The presence of semantic features on the root node of D or N identiWes them as predicates, thus allowing them to project argument structures—‘NP’ or ‘DP3’3—and to assign theta roles. The presence of phi features under the ‘agr’ node forces them to project extended agreement projections: Num (ber)P, K(ase)P, or Gen(der)P. Let’s assume that this process takes place in two steps: argument structure, i.e. a ‘pure’ projection, Wrst, then agreement structure, i.e. an ‘extended’ projection. If the grammatical features are attached to morphophonological features, then raising of the lexical head will take place, as described above, and the NP or DP can be used in its ‘bare’ form4 in an argument position. If the features are weak, then raising cannot take place and BNPs are not allowed. The grammaticalization process for French NPs can thus be shown to depend on the interaction between these two phenomena: as the phi features on French nouns became weak, BNPs were Wrst restricted to object positions, then became impossible in all A-positions. At the same time, the increasing adjunction of demonstrative pronouns to NP lead to the loss of semantic features on their root nodes, which in turn lead to their reanalysis as functional heads.

7.3 The Data 7.3.1 The ‘Old Stage’ of the language: Les Serments de Strasbourg, c.842

The earliest written text in French gives us a glimpse of what Gallo-Romance might have looked like. The text is divided into two symmetrical parts, one written in what is called the lingua rustica vel romana, or Vernacular Ro3

By ‘DP’ here I mean the projection of a pronominal D head. It is now clear that the term ‘bare’ is misleading, since it confuses two distinct types of NPs: true bare, that is predicative, nominal constructions, which can only refer to types; and NPs dominated by some extended functional projection other than DP, PossP, or QP—NumP or KP—which refer to individuals. 4

102

P. Boucher

mance, the other in Vernacular Germanic.5 It is uncertain to what extent the Romance half is representative of actual spoken Gallo-Romance, since it depends heavily on standard Latin oath-taking formulae of the time (see Table 7.1). The fact that Yvon (1946–7) describes the language of the Serments as corresponding to ‘l’e´tat ancien de la langue’6 may simply reXect this situation. Out of a total of twenty NPs, there are three demonstratives, eight possessives, three quantiWers, and nine BNPs. There are no articles, either deWnite or indeWnite. Semantically, the BNPs seem to fall into three groups:7 (i) DeWnite speciWc reference: Deo amur, christian poblo; (ii) IndeWnite speciWc reference: aiudha, damno, savir, podir; (iii) Generic reference: om, dreit. (14)

Deo amur / Christian poblo / om / dreit / aiudha / savir / damno / sagrement, etc. ‘God’s love / (the) Christian people / man / law / aid / knowledge / damnation / salvation, etc.’

Following Yvon (1946–7) and de Murc¸in (1815), and based on this text, I will therefore claim that Gallo-Romance NPs worked basically like Latin ones: demonstrative, possessive, or quantiWcational pronouns could be adjoined to NP for speciWc purposes; BNPs could refer both generically and speciWcally; there was no paradigm of deWnite and indeWnite articles. 7.3.2 A transitional stage: La Cantile`ne de Sainte Eulalie, c.878

Despite the short period of time separating the two texts (less than forty years), we can speak of a second, transitional stage as concerns the Cantile`ne. This is suggested by the presence of deWnite articles in thirteen out of thirtynine NPs (15) alongside nineteen BNPs (16). (15) li Deo inimi / les mals conseilliers / lo chief / enl fou / li rex pagiens ‘the God enemies / the bad advisers / the head / in-the Wre / the king (of the) pagans.’

5 See de Murc¸in (1815) for the complete text as well as a detailed discussion of the Modern French translation and examples of equivalent Latin texts of the same period. 6 Yvon (1946–7: 290): ‘This stage, which corresponds closely to that of Latin, can be called the old form of the language’. 7 I am using the terms ‘deWnite’, ‘indeWnite’, ‘speciWc’, and ‘generic’ in a preformal sense. This could be deWned along the lines of what is proposed in Carlier and Goyens (1998): speciWc deWnite reference refers to an entity identiWable through its relation to some anaphoric or deictic framework; speciWc indeWnite reference refers to an entity not identiWable through such a relationship; generic reference refers to entities identiWable in any possible time–space framework.

The Rise and Fall of DP

103

Table 7.1. BNPs in the Serments de Strasbourg and their (vulgar) Latin equivalents Serments

Latin equivalents

pro Deo amur savir et podir cum om son fradra

pro amore Dei sapere et potere cum homo suo fratri

(16) pulcella / corps / diaule / Deo / rex / pagiens / Wgure / colombe ‘(the) girl / (her) body / (the) Devil / God / (the) king / (the) pagans / (the) shape / (a) dove.’ All occurrences of deWnite articles in the text correspond to speciWc, deWnite individuals either known to the hearers (lo nom chrestiien, lo Deo menestier), previously mentioned in the text (les mals conseillers, li rex pagiens, la polle, la domnizelle), or implied by the context (les empedementz, enl fou). Two cases correspond to possessives (lo chieef, la mort). There are no examples of indeWnite or generic reference with a deWnite article. Alongside three cases where the bare nouns can be assimilated to proper names (deo, diaule, ciel), we Wnd BNPs being used as attributes (buona pucella, rex), materials (or, argent), or complements of prepositions (par manatce, a (grand) honestet, in Wgure, de colombe). Three BNPs refer to an unspeciWed number of individuals (paramenz, colpes) or all the members of a group (pagiens). Contrary to Modern French, we note the obligatory link at this stage between the deWnite article and a speciWc anaphoric or deictic context. 7.3.3 The Classic Old French period: Les Lais de Marie de France, c.1160

All texts from this period show a sharp increase in the number of deWnite articles used and the emergence of indeWnite articles, both for the singular (un, une) and the plural (un(e)s, des). Nearly 40 per cent of the NPs in the Lais, for instance, contain a deWnite article. The Prologue of the Lais yields the following: out of a total of thirty-eight NPs, there is one demonstrative, three possessive pronouns, nine quantiWers, eight deWnite articles (17), four indeWnite articles (18), and thirteen BNPs (19). (17)

als anciens / es livres / la letre / le surplus / le philesophe / le tens / l’onur etc. ‘to the elders / in the books / the letter / the surplus / philosophers / time / honor’ (18) uns granz biens / des lais / des aventures ‘(plur indef) great goods / (plur indef) tales / (plur indef) adventures’

104 (19)

P. Boucher escience/eloquence/custume/grevose oevre/grant dolu/ remembrance ‘science / eloquence / tradition / great works / great pain / memories’

DeWnite articles are now used when referring to all members of a group: als anciens, le philesophe, les contes, li Breton, les Lais; they may have generic reference in the sense that the anaphoric frame may be a type, or not be unique: la letre, l’onur, le surplus, le tens, le mestier, etc. We note the appearance of an indeWnite plural determiner, des, alongside the singular un: des lais, des aventures, un paı¨s, une aventure. There is now a restriction of the semantic extension of bare NPs, and deWnite speciWc reference is now impossible. For Carlier and Goyens (1998), li (le) at this stage expresses the fact that a given individual is identiWable relative to some framework. In the Cantile`ne, this framework was necessarily the known world of the hearers, the text itself, or its immediate context. Now we see the possibility of it being a typical situation. The next step in the process will be to extend the deWnite article to generic NPs, which is exactly what happens in Middle and Modern French. The link to a speciWc context identifying a known individual, or group of individuals, is lost. 7.3.4 The beginnings of the Modern period: Les Chroniques de Jean Froissart, c.1370

Out of a total of seventy-three NPs in a one-page extract, there are only fifteen BNPs, as opposed to twenty-five nouns preceded by deWnite articles and eighteen preceded by indeWnite articles, demonstratives, possessives, or quantiWers. The singular mass noun is now found preceded by the singular partitive article du. BNPs are no longer possible in subject position, but only in syntactically governed positions, as complements of V or P. BNPs are now restricted to generic or non-speciWc indeWnite meaning; (20)–(21). (20)

comment il ordonna en Artois ( . . . ) capitaines et gens d’armes pour les tenir ‘how he ordered in Artois ( . . . ) captains and soldiers to hold them’.

(21) et visoit ( . . . ) nuit et jour comment il pourroit porter contraire et dommage aux Anglais ‘and brooded night and day how he could cause harm and damage to the English’. To sum up this overview: (i) BNPs were the norm in Gallo-Romance, as they were in classic and vulgar Latin. Though demonstrative, possessive, or quantiWcational pronouns could be adjoined to NP for speciWc purposes, BNPs could refer both generically and speciWcally and there was no paradigm of deWnite and indeWnite articles, hence no DP projection.

The Rise and Fall of DP

105

(ii) A functional deWnite article emerged very early in the OF period (ninth century). From the very Wrst written texts the deWnite article had clitic status and could be both enclitic on prepositions and proclitic on nouns or adjectives. (iii) The opposition between general notions and unspeciWed individuals on the one hand and speciWc, deWnite notions or individuals on the other was formally marked in classic OF (eleventh to thirteenth centuries). Use of the deWnite article necessarily entailed a speciWc anaphoric or deictic context. IndeWnite articles began to appear. Genericity was still marked by BNPs in most cases, but could be marked by indeWnite articles. (iv) By the end of the Old French period a full set of indeWnite and partitive articles had emerged. DeWnite articles accounted for over 30 per cent of all NPs and could be used to express abstract notions and generic reference. It can be argued that French has now reached a stage where the deWnite article no longer signiWes deWniteness, but simply the unmarked case for nominal projections, since it can express either [þdeWnite] or [þgeneric], that is [deWnite]. This resembles the situation observed by Paykin and Peteghem (2002) for Russian, where deWniteness is never explicitly marked. Only indeWniteness can be marked in some contexts by word order, prosody, or genitive case. DeWniteness therefore represents the default interpretation. In a sense, the process which Lyons (1999) has termed the ‘deWniteness cycle’ has reached the point in French where the deWnite article plays much the same role as the case forms of Latin. In the Wnal section I will look brieXy at pronominal determiners and return to the question of case projections.

7.4 DP, KP, and NUMP 7.4.1 Reanalysis of pronominal D-words

As stated in the introduction, the analysis presented here formally links the expression of deWniteness to an interaction between three functional projections: KP, NumP, and DP. The stages through which the functional D-words emerged in French support this analysis. As seen above, deWnite atonal articles had already emerged by the ninth century and existed alongside overt case and number morphology for centuries. BNPs with speciWc indeWnite reference in both subject and object positions also lasted throughout the OF period. By the twelfth century, the numeral un ‘one’ began to be reanalysed as an indeWnite article. It could be used in both the singular (un, une) and the plural (unes) and could be used pronominally (les uns). By the fourteenth century, the enclitic forms des (deþles), du (deþun) were being

106

P. Boucher

used as partitives. This increasing use of D-words to mark deWniteness paralleled the loss of Wrst case morphology, then number morphology. Once case could no longer be used to mark an NP as ‘presuppositional’, therefore ‘deWnite’, the deWnite article was needed. Once number morphology had weakened suYciently, it had to be reinforced, then replaced by a paradigm of explicit indeWnite articles. In the framework developed here, this can be represented as the simultaneous loss of interpretable features on the root node of the D-word and of uninterpretable features on the complement node of the N-word. For deWnite articles, this meant the loss of the deixis feature. For the numeral, it meant the loss of a strong quantiWcational feature. Loss of the semantic feature meant that the head no longer projected an ‘argument projection’, i.e. was no longer interpreted as a predicate and was reanalysed as a functional head. This is precisely what we mean by ‘the emergence of DP’. Loss of the agreement features on the noun meant that N could no longer raise to Num (or K) and so bare nominal reference became impossible. The fact that in certain contexts deWnite articles could be used pronominally with the value of a demonstrative supports this analysis, since it implies that the deixis feature could be reactivated. Such contexts were: the ‘juxtaposition genitive’, vocatives, and certain temporal constructions (Jensen 1990). (22) il abati ton cheval et le Perceval ensemble he killed your horse and the Perceval together ‘. . . the one of Perceval (i.e. Perceval’s horse)’ (Foulet 1977: 53) 7.4.2 Case again

In Section 7.2, I claimed that Germanic languages have maintained a case projection, while Romance languages have not. Let me now justify that claim. Modern German determiners, nouns, and pronouns carry overt case morphology and adjectives have a ‘strong’ declension in the absence of an overt determiner, as opposed to a ‘weak’ declension in the presence of D. This can be taken to mean that either the determiner or the adjective must raise to K8, since the same aYxes appear on D, when it is present, as on A when D is absent. This trait can be claimed to be one of the fundamental characteristics of the Germanic languages within the Indoeuropean family, along with the two-tense system wherein preterites are marked with a dental suYx, and so on (Pyles and Algeo 1993: 85). While Modern Dutch and English have lost case morphology on determiners and nouns in general, they have retained it on pronouns and on possessive NPs. Moreover, Dutch adjectives have retained a trace of the

The Rise and Fall of DP

107

strong/weak declension opposition, in the form of an -e aYx which appears in the absence of an overt determiner. (23) a. Goed*(e) boeken zijn ‘Good books are b. Het goed (*e) boek ‘This good book

duur expensive.’ is duur is expensive.’

English also had an identical -e aYx which continued to appear on weak adjectives8 until the Early Modern period. I will therefore assume that KP is still an active projection in Germanic and that either the determiner or the Wrst adjective raises overtly to K, as in Modern German and perhaps Dutch and Middle English as well, or covertly at LF, as in Modern English. This explains the diVerence between Germanic and Romance in general. If Germanic nouns raise Wrst to Num8 then to K8, this would explain their ability to be used in subject as well as object position. This analysis can be extended to Old French, which patterned like Germanic. In the transition from Latin, contrary to the other Romance languages, with the exception of Old Occitan and Old Romanian (Marchello-Nizia 1995: 66), Old French retained some of the old Latin case system, in the form of a nominative or ‘subject’ case, le cas sujet, and an accusative or ‘object’ case, le cas re´gime. The OF agreement system was somewhat sporadic and confusing, however, since the same suYx, -s, was used to mark both case and number on masculine nouns, but only number on feminine nouns. Thus, while the case forms allowed considerable freedom in the use of BNPs in Gallo-Romance and Old French, as case morphology eroded and Wnally disappeared, the determiner gradually took over as the sole indicator of deWniteness. When, around the seventeenth century, number morphology also disappeared for all practical purposes, French found itself in the situation we see today: all referential NPs must contain an overt determiner. As for the other Romance languages, we can tentatively conjecture here that the absence of case morphology from the very beginning leads to more limited distribution of BNPs. On the other hand, the fact that Italian and Spanish never reached the stage that Modern French has now reached can certainly be attributed to the fact that they have maintained strong number aYxes. This guarantees weak indeWnite 8 The situation is not quite the same in each language. The rules in Dutch seem to be: adjectives are inXected if they are not preceded by a determiner and if they precede a ‘het’ word, that is, a noun which takes a deWnite determiner with a Wnal consonant -t (Kerstens 1993: 77–8); while in Middle English the rule seems to be: the adjective is inXected if it is in the plural, is monosyllabic, and/or is not preceded by a determiner. This added complexity seems to be due to the fact that the Wnal -e in Middle English could be the result of erosion of the weak adjectival endings or of the genitive plural forms (Pyles and Algeo 1993: 152).

108

P. Boucher

reference, at least, and therefore allows them to be used in object position, i.e. within VP. They cannot, however, raise out of VP to be used as subjects, since they do not have the additional KP projection that I claim is necessary to have BNP subjects.

7.5 Conclusion In the course of this chapter I have argued that representing deWniteness solely in terms of the presence of an (overt or covert) functional determiner is unsatisfactory, since many languages do not have a determiner category and yet are perfectly able to express [
deWnite] reference. Moreover, the diachronic study of French reveals the emergence of a functional determiner at some point in the transition from Latin to Old French. The analysis of diachronic data has therefore forced a change in perspective concerning DP, which I have argued emerged in French in response to two parallel evolutions: the loss of substantive content on an adjoined XP determiner word and the erosion of agreement morphology. The conditions allowing (or forcing) the emergence of a functional determiner category thus shed light on the role of determiners and the interdependence of three diVerent functional heads: D(eterminer), K(ase), and Num(ber). Finally, I have proposed an original analysis of the diVerence between bare NPs in Germanic and Romance and redeWned DelWtto and Schroten’s Restricted QuantiWcation Condition, in terms which are at once simpler and more compatible with current syntactic theory. My analysis oVers the possibility of direct mapping from morphology to syntax to semantics and allows us to calculate straightforwardly the parameters of diachronic change.

8 The Reanalysis of the French Prepositional System: A Case of Grammaticalization in Competing Grammars M I R E I L L E T R E M B L AY, F E R NA N D E D U P U I S , AND MONIQUE DUFRESNE

8.1 Introduction In this chapter, we consider the evolution of French prepositions and particles. Our comparative analysis of Old and Modern French shows a change in transitivity at the end of the Middle French period, followed by a split of the prepositional system into the two subsystems which characterize Modern French. We argue that both changes can be related to the lexicalization of verbal and prepositional preWxes. This study is based on the Base de franc¸ais me´die´val constituted by C. Marchello-Nizia (for Old French), and the corpus Lemieux (for sixteenth-century French). The Rabelais corpus comes from the CERHAC, Universite´ de Clermont-Ferrand.

8.2 The Data 8.2.1 Modern French

As is well known, Standard French has a dual prepositional system: simple prepositions such as sur, sous, dans, etc. and morphologically complex prepositions—also called orphan prepositions—such as dessus, dessous, dedans, etc. The two subsystems are (mostly) in complementary distribution: simple prepositions are usually found with overt objects (1a), while complex prepositions are used with implicit objects (1b). We wish to thank the participants to the 7th Diachronic International Generative Syntax (University of Girona, June 2002) for comments and suggestions. The research for this chapter was partially supported by the Social Science and Humanities Research Council of Canada (grants 410-96-1445 and 214-97-0016) and by Queen’s University’s Advisory Research Committee.

110

M. Tremblay, F. Dupuis, M. Dufresne

(1) a. Je marche sur la pelouse ‘I walk on the grass.’ b. Je marche dessus ‘I walk on it.’ According to Zribi-Hertz (1983), complex prepositions such as dessus, dedans, etc. are not truly intransitive, but take a null object pronoun pro. (2)

La pelousei , j’aime marcher dessus proi the grass I like walk-in on ‘The grass, I like walking on it’

The possibility of complex prepositions to license a null object is often linked or even attributed to their use as nominals, as in (3). (3)

a. Le ‘The b. Le ‘The

dessus top dedans inside

de of de of

la the la the

table table’ boıˆte box’

8.2.2 Old French

Interestingly, in Old French, both systems could be used either as prepositions or orphan prepositions. In (4a), a simple preposition (sus) is used with an overt object (la tombe), while in (4b), the object of sus is implicit. This is an instance where Modern French would use dessus. (4) a. Et fetes escrivre sus la tombe: Ci gist Gaheriet et and make write on the grave here lies Gaheriet and Gauvains que Lancelos ocist . . . Gauvain whom Lancelot killed . . . (Artu, 220) b. Lors pensent des plaies Lancelot et metent sus ce qu’il then they-dress wounds Lancelot and they-put on what they cuident . . . consider . . . ‘Then, they dress lancelot’s wounds and they put on them what they consider . . .’ (Artu, 203) The complex preposition desus can be used in exactly the same contexts. In (5a), desus appears with an overt object (desus la tombe), and in (5b), the object of desus is implicit.

Reanalysis of the French Prepositional System (5) a. Desus la tombe qui on the grave which letres qui disoient . . . letters which said (Artu, 251) b. . . . apre`s Wrent metre . . . after they-made put (Artu, 262)

111

estoit merveilleuse et riche avoit was marvellous and rich were

desus letres qui disoient: Ci gist . . . on letters which said here lies . . .

Moreover, many Old French prepositions could also be used as intransitive particles (Buridant 2000; Dufresne, Dupuis, and Tremblay 2003). (6)

aler avant aler ariere lever sus avaler/descendre jus monter contremont issir hors

‘to go forward’ ‘to go back’ ‘to stand up’ ‘to go down’ ‘to climb up’ ‘to exit/to go out’

In some cases, the particles could form a complex predicate with the verb (7), while in others they remained syntactically and semantically independent (8). (7) a. Lors li cort sus li rois . . . then him-dat runs sus the king-nom ‘Then the king attacks him.’ (Artu, 152) b. . . . que che ne fu se menchoingne non que on that this neg was either lie not which 3psg le avoit mis sus him-acc had put on ‘. . . that it was either a lie which he had been accused of.’ (Clari, 20) (8) a. . . . qu’ il est menterres de metre tel chose avant that it is deceptive to put such thing forward (Artu, 37) vont arriere a la nef b. Et cil and others go back to the ship (Partonopeu, 5901; cited in Buridant 2000: 441)

112

M. Tremblay, F. Dupuis, M. Dufresne

8.2.3 The issues

Two major changes took place between Old and Modern French. The Wrst change occurred at the end of the Middle French period, when French lost its productive particle system. Although residual particles are still found in Renaissance French (see section 8.4.2), such particle uses are very limited and restricted to lexicalized expressions such as courir sus. Modern French does not allow particle constructions and examples such as (9) are no longer grammatical. (9) *Nous allons arrie`re a` Paris ‘We are going back to Paris.’ The second change took place one or two centuries later. The prepositional system was split into the two subsystems that characterize Standard French. Was the split of the prepositions system triggered by the loss of particles, a subsequent reanalysis of complex prepositions as non-derived lexical items, or the existence of two systems of competing prepositional doublets? In the next section, we attribute the loss of particles to the lexicalization of aspectual preWxes, and consider the impact of such a change on the French particle system. The creation of a system of lexical doublets will be considered in section 8.4.

8.3 The Lexicalization of Aspect/Directionality in the History of French Tremblay, Dupuis, and Dufresne (2003) link the loss of particles to a change in the lexicalization of direction/aspect of verbs triggered by the loss of productivity of preverbs. Unlike Modern French verbs,1 Old French verbal predicates were not inherently directional and particles were required to lexicalize direction. For example, while in Modern French, an underived verb like entrer ‘to enter’ can appear with or without a prepositional complement (10), in Old French the verb entrer, being unspeciWed for direc-

1 According to Klipple (1997), diVerences in the lexicalization of the [þdirectional/aspectual] node (Di Sciullo 1996) account for language variation. For example, while in English direction/aspect is encoded in a particle (preposition), it is encoded in the verb in Modern French. This contrast is illustrated below.

(i) Jean est entre´ (dans la salle) ‘John came in (the room)’ (ii) Edwidge a monte´ l’ escalier ‘Edwidge went up the stairs’

Reanalysis of the French Prepositional System

113

tionality, required the presence of either a particle (11) or a prepositional complement (12) to lexicalize the [þdirectional/aspectual] node. (10) Jean est entre´ (dans la salle) John aux entered in the room) ‘John entered (the room).’ (11)

. . . et quant il vindrent a la porte, si la trouverent ouverte; and when they came to the door and it they-found open si entrerent enz et voient le leu si bel . . . and they-entered in and they-see the place so beautiful ‘. . . when they came to the door, they found it open and they went in and saw the place so beautiful.’ (Artu, 57)

(12)

. . . et quant il entrerent en la grant sale . . . and when they entered in the large room ‘. . . when they entered the large room . . .’ (Artu, 57)

Aspect was also marked by way of prepositional preWxes such as par-, con-, and re- in (13), (see Dufresne, Dupuis, and Tremblay 2003 for a discussion of the contrastive properties of preverbs and particles). (13)

pardire ‘to say completely’ paremplir ‘to Wll completely’ concueillir ‘to pick here and there’ confermer ‘to make strong’ refuir ‘to Xee as fast as possible’ regarder ‘to examine carefully’ (Dictionnaire de l’ancien franc¸ais, Larousse)

The ability to create new verbs by adding prepositional preWxes drastically decreased during the Middle Ages (Dufresne, Dupuis, and Longtin 2001; Hamacher 2002). Following the loss of productivity of preverbs, many preWxed verbs were reanalysed as non-derived lexical items. For example, verbs like arriver ‘to arrive’ or de´fouler ‘to unwind’ are no longer perceived as morphologically complex (aþriver, de´þfouler). This change aVected not only the once morphologically complex verbs, but also all verbs with the relevant semantic feature [DIRECTION] such as entrer, as both types of verbs were no longer morphologically distinct. As a result, following the change in the lexicalization pattern of direction, the verb entrer became þdirectional, and the presence of the prepositional complement (PP or particle) became

114

M. Tremblay, F. Dupuis, M. Dufresne

optional. Thus, under our analysis, the disappearance of (directional) particles was triggered by the relexicalization of directionality as a verbal property. In the next section, we argue that the loss of particles created a system of lexical doublets which is responsible for the split of the prepositional system.

8.4 The Reanalysis of the Preposition System 8.4.1 The lexicalization of complex prepositions

In order fully to understand the subsequent split of the prepositional system, we need to reconsider the categorial and morphosyntactic status of complex prepositions in OF. First, in OF, particles do not appear to be categorially distinct from prepositions as all particles could also function as prepositions, as illustrated in (14). (14)

avant ariere sus jus contremont hors

‘before, in front of ’ ‘behind’ ‘on’ ‘outside, down from’ ‘against’ ‘outside’

For example, particles/prepositions such as sus or avant ‘forward’ could be used either transitively (15), or intransitively (16). (15) a. Sus sa poitrine tenoit ses mains croisant on his chest kept his hands crossed ‘He was keeping his hands crossed on his chest.’ (Aliscans, 827; cited in Buridant 2000: §388) b. De deduit d’oiseax et de bois jne savoit nus hom avant lui ‘. . . No man knew how to be delighted so much before him.’ (Dole, 3) (16)

a. Si corent sus au chevalier 3ppl run sus towards-the knight ‘they attack the knight.’ (Artu, 174, 35) b. Lors saut avant GirXez et dist a la reı¨ne: . . . ´ then jumps forward GirXe and says to the queen ‘Then GirXe´ comes forward and says to the queen: . . .’ (Artu, 319)

Reanalysis of the French Prepositional System

115

Second, as mentioned earlier, not all prepositions could be used as particles: while simple prepositions such as sus, avant, fors/hors, ens could be used as either prepositions or particles, their preWxed counterpart desus, dedenz, defors/dehors could only be used as prepositions (from Tremblay, Dupuis, and Dufresne 2004). The very fact that complex prepositions were never used as particles indicates that, in Old French, such prepositions had probably not yet acquired a word-like status. Although they were probably no longer two distinct syntactic heads, each heading its own maximal projection [PP de[PP sus]] (Vincent 1999), they had not yet become the kind of syntactic atoms which can form complex predicates with verbs. In fact, in Old French, there was still morphological and syntactic evidence to maintain distinct representations for simple and complex prepositions. First, simple and complex prepositions had distinct distributions (only simple prepositions could be used as particles). Second, as argued above, preWxation was still a very productive morphological process. PreWxes were found not only on verbs but also on prepositions. For example, we still Wnd examples of prepositions used with a double de- preWx with no apparent change in meaning. (17)

Dedevant lui ad une perre byse in.front him cop-3ps a stone brown. ‘There was a brown stone in front of him.’ (La Chanson de Roland, 173)

Following the loss of particle constructions, however, the complex grammatical structure of complex prepositions became opaque, First, the disappearance of particles meant that there was no longer a distributional diVerence between the two types of prepositions. Second, preWxation on prepositions became non-productive, a change to be linked to the loss of productivity of preverbs. In Middle French, complex forms such as dedesoz or dedesus are no longer found. As there is no longer morphosyntactic evidence available to the language learner to posit a structural diVerence between simple and complex prepositions, complex prepositions are reanalysed as non-derived prepositions, thereby creating a whole system of competing lexical doublets. Modern French data seem to support the doublet hypothesis, as complex prepositions which are no longer members of a true doublet such as derrie`re, avant, devant, avec, apre`s, and depuis still allow both an overt or an implicit object.

116 (18)

M. Tremblay, F. Dupuis, M. Dufresne Elle est alle´e derrie`re (la maison) she aux gone behind the house ‘She went behind (the house).’

(19)

Elle est partie avant (la Wn du cours) she aux left before the end of-the class ‘She left before (the end of the class).’ (20) Il a place´ ses livres devant (son cafe´) he aux placed his books before his coVee ‘He placed his books in front (of his coVee).’ (21) Il n’a pas relu sa the`se depuis (la soutenance) he neg-aux neg reread his thesis since the defence ‘He has not reread his thesis since (the defence).’

Such data indicate that only prepositions that were part of a true lexical doublet underwent the reanalysis. 8.4.2 Lexical doublets in Renaissance French

Under the proposed analysis, the split of the prepositional system is the result of a reanalysis that took place during Classical French. We thus expect the similarity in distribution between simple and complex prepositions to increase slowly prior to the seventeenth century and reach a stage where they became lexical doublets. Renaissance French is particularly revealing in this respect as it both just follows the loss of particles, and precedes the split of the prepositional system. A comparison of simple preposition sus versus complex preposition dessus during the sixteenth century shows that the distribution of sus and dessus became increasingly similar when compared to their use during the thirteenth century (Table 8.1). In Old French, true prepositional uses of sus represent 40 per cent of all uses (68 per cent if we exclude uses as particles), while prepositional uses of desus represent 41 per cent of all uses. Moreover, while Table 8.1 Uses of sus/desus in thirteenth-century prose corpus (Old French) (%)

Preposition Particle Orphan preposition Other uses (locutions)

sus (200 occ.)

desus (96 occ.)

40 41 4.5 14

41 0 10 49

Source: Tremblay, Dupuis, and Dufresne (2004).

Reanalysis of the French Prepositional System

117

both sus and desus are used in locutions, this use accounts for 49 per cent of all uses of desus, but only 14 per cent of all uses of sus (23.7 per cent if we exclude uses as particles). Interestingly, while both prepositions can form a prepositional phrase with the preposition par (22), only desus occurs in a DP structure (23). (22)

a. Vait s’en Yseut, Yvains l’en meine tot droit aval par sus goes cl-cl Yseut Yvain her take all straight down towards l’ araine. the shore ‘Yseut goes, Yvain takes her straight down to the shore.’ (Beroul, 138) b. ( . . . ) fetes un char jtormoier par desus mon cors. make a cart pass over my body ‘( . . . ) make a cart pass over my body.’ (Dole, 148) (23) . . . et s’il avient par mescheance que Mordre´s viengne au desus del roi Artu et qu’il veinque ceste bataille, . . . ‘. . . and if it happens unfortunately that Mordre´ defeat King Arthur and that he win this battle, . . .’ (Artu, 219)

As expected under the reanalysis hypothesis, during the sixteenth century, the similarity in distribution between sus and dessus appears to have increased (Table 8.2). First, while dessus is still more broadly used in prepositional phrases, the diVerence in proportion between sus (5.5 per cent) and dessus (27.8 per cent) has decreased. Second, sus is almost never used as a particle. In Table 8.3, the high percentage of particle uses of sus is attributable to a single author: de la Vigne. In de la Vigne, sus is only used as a particle (prepositional uses being restricted to either dessus or sur). Finally, both sus and dessus are now equally used as prepositions, and compete with preposition sur (sur is Table 8.2 Uses of sus/dessus in sixteenth-century prose corpus (Pre-Classical French) (%)

Preposition Particle Orphan preposition Nominalization Other uses (phrases)

sus (36 occ.)

dessus (37 occ.)

66,7 25 0 0 5.5

61 2.8 5.6 2.8 27.8

118

M. Tremblay, F. Dupuis, M. Dufresne

seldom found in Old French but starts competing with sus and desus at the end of the Middle Ages). As shown in Table 8.3, the use of sur, unlike that of sus and dessus, is restricted to prepositional uses. To summarize, we have seen that the lexicalization of verbal preWxes in Middle French led to the disappearance of particles, and the creation of a system of increasingly competing lexical doublets during the Renaissance. During the Classical French period (seventeenth century), these doublets were reanalysed into the Modern French subsystems of transitive (sur, sous, dans, etc.) vs. orphan prepositions (dessus, dessous, dedans, etc.). While the proposed analysis explains why such reanalysis took place, it leaves open the question of why it is complex prepositions which were reanalysed as orphan prepositions (and not simple prepositions). In the next section, we argue that the answer to this question lies in their ability to appear in a nominal context.

8.5 Implicit Objects 8.5.1 Pro-drop in French

It is a well-documented fact that Old French was a pro-drop language. Interestingly, the language allowed not only null subjects (Adams 1987; Dupuis 1989; etc.) as illustrated in (24), but also null objects, either as object of V (25), or as object of P (26). (24) Lors dist a Perceval que il essait a l’ espee then he-said to Perceval that he tries at the sword ‘Then he told Perceval to try the sword.’ (Queste, 6. 23; cited in Vance 1997: 200) (25)

Un consoil sot li nains du roi, Ne sot que il a secret knew the dwarf of.the king neg knew only he ‘The dwarf knew a secret about the king, and he was the only one to know it’ (Beroul, 1307–8; cited in Buridant 2000: §365) Table 8.3 Prepositional uses of sur/sus/dessus in sixteenthcentury prose corpus (Pre-Classical French) (%)

sur sus dessus

Preposition (147 occ.)

Other uses (17 occ.)

68.7 16.3 15 100

0 70.6 29.4 100

Reanalysis of the French Prepositional System

119

(26) . . . et Wst prendre une biere chevaleresce et Wst metre enz and made take a coYn horse-adj and made put inside le cors monseigneur Gauvain the copse Messire Gauvain ‘. . . and he made someone take a coYn pulled by two horses and made someone put Messire Gauvain’s corpse inside.’ (Artu, 211) During the Middle Ages, French underwent a major morphological reduction. During this period Old French lost its subject-verb agreement, object participle agreement, pro-drop, morphological cases, and aspectual markers. While the loss of strong agreement on the verb meant that pro could no longer be licensed in subject position, the loss of morphological case resulted in the disappearance of object drop constructions: pro could no longer be licensed in case-governed positions (object of V and P). This change explains why simple prepositions such as sus, hors, enz, etc. are no longer used as orphan prepositions. As showed in (27) and (28), in Modern French, null objects are no longer allowed in verbal and most prepositional contexts.2 nain de´tenait un secret au sujet du roi. Il e´tait le dwarf knew a secret about the king. He was the only a` *(le) connaıˆtre. to know (it).’

(27)

*Le ‘The seul one

(28)

*Le gazon, j’ aime marcher sur. ‘The grass, I like to walk on.’

This analysis does not however explain why pro was not also barred from appearing after complex prepositions (dessus, dessous, dedans, etc.). As we saw earlier, in Renaissance French, complex prepositions could still be used with an overt object (29), or an implicit object (30). (29)

. . . en frappant torche lorgne, dessus le geant ‘. . . striking repeatedly on the geant’ (Rabelais; cited in Dictionnaire du moyen franc¸ais, 200; Rabelais, Cinquiesme Livre, 107)

(30)

. . . n’ entendre les mouvemens que faict la langue musculeuse, neg hear-inf the moves that makes the tongue muscular lorsque le boire dessus coule pour descendre en l’ estomac. when the drink on runs-down to go-down in the stomach 2

But see Cummings and Roberge (2003) and Larjavaara (2000) for a diVerent view.

120

M. Tremblay, F. Dupuis, M. Dufresne ‘. . . to hear the movements of the muscular tongue when drinks run down on it to go down the stomach.’ (Rabelais, Cinquiesme Livre, 177)

In the next section, we attribute the use of complex prepositions with covert objects to the fact that complex prepositions could also function in a nominal context. 8.5.2 Object drop and prepositions

According to Roberge (1998), the diVerence in distribution between the two prepositional systems of Modern French can be explained if one assumes that in Modern French, pro cannot appear in case-marked positions, and that in Modern French, prepositions belong to one of two paradigms: case-assigning (a`, de, pour, sur, dans, avec, apre`s) versus non-case-assigning (dessus, dedans, derrie`re, avec, apre`s) prepositions. The fact that denominal prepositional phrases do not require the presence of the genitive marker de when used with an implicit object constitutes a strong argument in favour of this proposal. (31) Je me suis assise a` coˆte´ de Paul. I refl aux sitted to side of Paul’ ‘I sat next to Paul.’ (32) Je me suis assise a` coˆte´ pro / * a` coˆte´ de pro I refl aux sitted to side ‘I sat next to him/her.’ (33)

a. Paul s’ est assis en face de Marie. Paul refl aux sitted in front of Marie ‘Paul sat in front of Marie.’ b. Paul s’ est assis en face pro / *en face de pro Paul refl aux sitted in front ‘Paul sat in front on him/her.’

Assuming Roberge’s analysis, we attribute the ability for complex prepositions to appear in the context of an implicit object to the fact that such complex prepositions are unspeciWed with respect to case. A detailed look at the distribution of prepositional phrases in the history of French provides evidence in favour of this hypothesis. First, nominalized complex prepositions are found in the language as early as the twelfth century.

Reanalysis of the French Prepositional System (34)

121

Sire, qu’est ce que vos avez fet qui estiez au desus ‘Lord, what you have done who were above your deadly enemy . . .’ de vostre mortel ennemi . . . (Artu, 202)

However, this should not be taken to indicate that complex prepositions were derived from nouns as, in Old French, such nominalized forms are never referential. Nouns derived from complex prepositions are not used referentially until Middle French (the end of the Middle Ages). (35) et quant ilz furent en and when they were in grande et toute tapissee, large and all covered d’ or . . . of gold (Jehan de Paris, 68)

la the le the

premiere salle, que moult estoit Wrst room which much was dessus et les coustez, d’ ung drap top and the sides of a drape

Thus, although nominal uses of complex prepositions go back to at least Old French, such nominal uses were generally outnumbered by prepositional uses. If a conversion process was ever involved, it appears to have gone from P to N and not from N to P. Under the proposed analysis, the reanalysis of complex prepositions as orphan prepositions follows from changes in the distribution of pro. In Old French, pro could appear in case-marked position. After the Middle Ages, pro can no longer appear as an object of V and as an object of a case-assigning P, and can only occur as an object of N or complex prepositions. In Old and Middle French, the use of complex prepositions in a nominal context indicates that such prepositions were unspeciWed with respect to case. A review of the use of the prepositions sus and dessus in Rabelais’s work reveals surprising distributional asymmetries between sus and dessus (Table 8.4). First, as expected, true prepositional uses of sus now represent 95 per cent of all uses, and no particle uses are found, while prepositional uses of desus Table 8.4 Uses of sur/sus/dessus in Rabelais (%)

Preposition Particle Nominalization Orphan preposition Other uses (locutions)

sus (537 occ.)

dessus (127 occ.)

sur (134 occ.)

95 0 0 0.4 4.6 100

17 0 5 16 62 100

98.5 0 0 0 1.5 100

122

M. Tremblay, F. Dupuis, M. Dufresne

decrease to 17 per cent of all uses. Second, we observe that dessus is now used as a true nominal (i.e. referential nominal) in 5 per cent of the occurrences. In comparison, sus is never used referentially. Another related fact is the increase of prepositional phrases with dessus. As shown in Table 8.5, a more detailed look at the distribution of dessus shows that 47 per cent of the uses of dessus occur in a DP structure (5 per cent of the total uses of dessus are referential nominals, and 42 per cent are prepositional phrases with a DP structure). Table 8.6 shows that a majority of such prepositional phrases occur in their Modern French uses (66 per cent involve a true DP structure and close to 20 per cent involve an implicit object). However, a surprising 15 per cent are used transitively, as illustrated in (36). Such examples provide additional evidence that in Renaissance French, complex prepositions were unspeciWed with respect to case. (36) ( . . . ) je feusse desja au dessus la sphere de la lune, . . . I was already above the sphere of the moon . . . ‘( . . . ) I was already above the sphere of the moon.’ (Rabelais, Pantagruel, 108) After the Renaissance, however, complex prepositions were reanalysed as strictly non-case assigning, which accounts for their distribution in Modern French. In dialects such as Canadian French, where complex prepositions can still be used with an overt object, no reanalysis took place and complex prepositions are still unspeciWed with respect to their case-assigning properties.

Table 8.5 Uses of dessus in Rabelais (%) dessus NP

dessus pro

au dessus (de)

nominalized

par dessus

dessus dessoubz

(cy) dessus.

la dessus

33

42

5

5.5

2.4

8.7

2.4

Table 8.6

Uses of au dessus in Rabelais (%)

au dessus (53 occ.) 15.1 18.9 66

au dessus NP au dessus pro au dessus de NP

? 1.6

Reanalysis of the French Prepositional System

123

8.6 Conclusions Our analysis is consistent with some of the most recent theoretical claims concerning language change. First, it shows how a series of apparently unrelated changes such as the loss of particles and the split of the prepositional system may be attributed to a single (although major) change in the morphological system (the loss of aspectual preWxes). Second, it supports Roberts and Roussou’s 1999 claim that grammaticalization involves structural simpliWcation. We have argued that the grammaticalization of the complex prepositions (and in particular the emergence of nominalized prepositions) is attributable to the lexicalization of the complex prepositions. Finally, it provides evidence that change is often triggered by the general ban on doublets (Kroch 1994). However, unlike reanalysis which seems to take place over a very short period of time, the elimination of doublets appears to be a much longer process. Primary Sources Beroul. Tristan, ed. L. M. Defourques (1947). Paris: Champion (CFMA 12). [Abbr. Beroul] La Mort le Roi Artu, ed. J. Frappier (1954). Gene`ve: Droz (TLF.). [Abbr. Artu] Larousse (1992). Dictionnaire de l’ancien franc¸ais. Larousse (1992). Dictionnaire du moyen franc¸ais. Le Roman de Jehan de Paris, ed. E. Wickersheimer (1923). Paris: Champion (Librairie Ancienne). [Abbr. Jehan de Paris] Renart, J. Le Roman de la Rose ou de Guillaume de Dole, ed. F. Lecoy (1962). Paris: Champion (CFMA 91). [Abbr. Dole] Robert of Clari. La Conqueˆte de Constantinople, ed. P. Lauer. Paris: Champion (CFMA 40). [Abbr. Clari]

9 Accusative Alternation in Old and Modern Romance A N NA B A RT R A KAU F M A N N

9.1 Introduction In traditional philological works,1 it has been asserted that several verbs underwent a change in the case of the (human) argument selected. Verbs like the Catalan perdonar ‘forgive’, pregar ‘ask, demand’, and others to be presented later on, together with their Spanish, French, and Italian equivalents, evolved from an Accusative to a Dative argument, whereas other verbs, like Cat. ajudar ‘help’, amenac¸ar ‘threaten’, and their Romance equivalents evolved from Dative to Accusative case. In (1)–(2) some cases are presented of the old and modern uses of the Wrst group of verbs, and in (3)–(4) of the second group of verbs: I thank Josep Maria Brucart and Gemma Rigau for their ideas, proposals, and stimulating brainstorming as we began to work together on this topic. I thank also Joan Torruella and Manuel Pe´rez Saldanya for oVering me several data from their databases (AITCM). Thanks also to Jaume Mateu for exciting discussions about Lexical Structure, to Luigi Rizzi for some comments on a previous version of this chapter, and to the audiences in DIGS VII, Workshop on Spanish Syntax at Universita` di Ca’Foscari in Venice, Linguistics Seminar at the Centro Linguistico (Universita` di Siena), XIII Colloquium on Generative Grammar at Ciudad Real (Spain), and I Seminari de Sintaxi Histo`rica at Universitat d’Alacant at La Nucia, especially Joan Rafael Ramos, for their observations. Usual disclaimers apply. This work has beneWted from Grant MCYT BFF2000-0403-C02 of the Ministerio de Ciencia y Tecnologı´a of Spain, DGR 1999-SGR-0013 of the Generalitat de Catalunya. 1 See, for instance, in the Catalan philological tradition, Par (1923: §430): ‘Es combinacio´ abundantı´ssima en Metge y escarsı´ssima avuy, la dels re`gims directe y relatiu, en qual cas lo primer era precisament lo de persona y lo segon lo de cosa, y may lo contrari. Modernament, substituhim en general aytal combinacio´, fent directe lo re`gim de cosa y posant lo de persona en re`gim indirecte. Aytal trasllat es particularment remarcable en los verbs qui signiWquen ‘‘pregar’’, ‘‘demanar’’, raho´ per la qual los donarem acoblats. [In Metge‘s works a lot of verbs that in today’s language are constructed with an Accusative designating a Person and a Dative argument designating a Matter or an Object were used with a direct or Accusative complement designating the person and a ‘‘relative’’ (i.e. Genitive or Ablative) complement designating the Matter or Object.] ( . . . ) A la llista de verbs ab re`gim relatiu

Accusative Alternations (1) a. Tu perdonist tots aquells you forgave all those (people)Acc (Tirant, DCVB) b. Tan fort lo pregave e li clamave merce` so hard himAcc he-begged and himDat claimed mercy ‘He/she begged him and claimed for mercy so hard.’ (Homilies, DECLC) c. De avarı´cia les has notades of greediness themAcc you-have noted ‘You have noted their greediness.’ (Metge, Somni; Par 1923) (2) a. Li perdonara`s totes les ofenses the oVences himDat you-will-forgive all ‘You will forgive him all the oVences.’ b. Li va pregar una almoina himDat he-prayed a mercy ‘He asked him/her for mercy.’ c. Tothom li nota l’ avarı´cia everybody him=herDat notes the greediness ‘Everybody notes her/his greediness.’ (3) a. Veent que en altra manera no li podia ajudar seeing that in another way not himDat could help ‘Seeing that he could’nt help him another way.’ (Metge, Somni, IV; DCVB)

125 (OC)

(OC)

(OC)

(MC)

(MC)

(MC)

(OC)

(§429) [i.e. conferir, conWar, curar, disposar, disputar, duptar, War, freturar, jugar, mentir, parlotejar, parlar, usar, profetar, tractar, respondre] hem d’afegir donchs, los segu¨ents, qui en lo text metgench porten noresmenys lo re`gim directe. [The verbs that take an Accusative human complement in Metge are, besides the list of verbs in italics presented, following Par: absoldre ‘‘acquit, absolve’’, certiWcar ‘‘certify’’, consellar ‘‘advise’’, escusar ‘‘excuse, exempt’’, privar ‘‘deprive’’, relevar ‘‘relieve’’, notar ‘‘notice, perceive, warn’’.]’ Par notes also that some verbs that in Latin took an inanimate DO and a person dative, were used by B. Metge (XIV cent.) with the human argument as the DO and a prepositional complement denoting the inanimate object. These were the verbs amonestar ‘warn, advise’, pregar ‘ask for’, demanar ‘ask for, demand’, enquerre ‘ask for, enquire’, suplicar ‘beg for’, avisar ‘warn, inform’, informar ‘inform’. Nevertheless, Par points out that some verbs exhibit the two constructions in Metge. See also Brunot and Bruneau (1949: §476, p. 319): ‘un grand nombre de verbes ont change´ de construction [a great number of verbs have changed their construction]’, among which they cite the direct constructions that have become indirect and vice versa. And ibid. (§570, p. 422): ‘Notons qu’en ancien franc¸ais un nom de personne se construit directement apre`s un verbe; on dit ‘‘demander quelque chose mon pe`re’’. L’emploi du cas-re´gime suYt a` marquer le rapport [Notice that in Old French a noun designating a person is constructed without a P, directly after the verbs; for instance, demander quelque chose mon pe`re ‘‘ask something to my father’’. The relation becomes suYciently clear by means of the ‘‘case regime’’]’.

126

A. Bartra Kaufmann b. Un rey infael lurs menassa` a mort a king inWdel themDat menaced to death ‘An inWdel king menaced them to death.’ (Llull, Felix, Pt. VIII, ch. 38)

(4) a. Has d’ajudar-lo you-have of help-himAcc ‘You have to help him.’ b. Va amenac¸ar-lo menaced-himAcc ‘He/she menaced him.’

(OC)

(MC)

(MC)

We will concentrate on the behaviour of the Wrst group of verbs (1)–(2), which have a meaning of (psychological) transfer.2 By taking as empirical starting point the changes in the argument structure of these verbs, we will argue that the apparent structural diVerences between two successive stages of a language are not to be viewed exclusively as a change in the argument structure, a very diYcult move to be argued for. What the changes reXect is a much more complex phenomenon. Two alternating structures will be shown to coexist at successive historical periods.3 The interaction of a bunch of features and processes is responsible for the fact that one of the two structures is abandoned in favour of the other one. We will mainly concentrate on Catalan and Spanish, but data will be also presented from other Romance languages. The alternation will be called accusative alternation, a label that makes clear the parallelism with other well-known alternations, such as the locative alternation (load hay on the truck / load the truck with hay) and the dative alternation (give John a book / give a book to John). As said before, several works have argued that the dominating pattern in modern languages is (5a), where in the old ones it was (5b): (5)

a. Datperson —Accthing b. Accperson —Oblthing

(5a) is the classical ditransitive structure, with an Accusative complement denoting an inanimate theme object and a GOAL or EXPERIENCER Dative denoting a human complement. In (5b) the human complement is an Accusative object and the inanimate theme is an Oblique or Ablative 2 For a general view of accusative-dative alternations in Catalan, from a Cognitive Grammar perspective, see Ramos (forthcoming a, forthcoming b). 3 Some of these verbs presented alternating structures in the old languages and still maintain both of them nowadays, whereas other predicates preferred one structure in the old stages and the other one in the modern language.

Accusative Alternations

127

complement introduced by the P de ‘of ’. For ease of exposition we will call the construction in (5a) Dative construction and the one in (5b) Accusative construction. The name accusative alternation receives its justiWcation from the fact that the verb can be combined with two diVerent types of Accusative arguments (presumably a THEME in (5a), and a GOAL or EXPERIENCER in (5b)). We will argue that none of these constructions can be derived from the other, diachronically or synchronically, but rather correspond to two independent lexical structures, associated with two slightly diVerent semantic interpretations.

9.2 The Two Alternating Structures 9.2.1 Verbs with the two structures

In this section we show that the two structures were already present in Old Romance. Some examples of the Accusative construction are shown in (6) to (9). (6)

4

a. un cavaller prega` lo rey que perdona`s al Wll de la a knight asked the kingAcc that he-forgave to-the son of the dona (OC) woman (Llull, DECLC) b. Honoratz lo preguava (OO) honoured himAcc he-prayed (Jensen 1994) c. Uno cavaliere pregava un giorno una donna d’ amore (OI) a knight asked one day a womanAcc of love (Poletto 2002 [my italics])4 d. Y en Wn, rogadas a detenerse, se sentaron (OS) and Wnally prayed-themFemPl to stop refl.sat down (Cuervo) e. Laissie´ me ester, ne m’ en proie´s ja mais! (OF) anymore Let me be not meAcc of-it ask ‘Leave me in peace, don’t ask me anymore’ (R. Violette, Tobler and Lommatzsch 1966)

As pointed by Jensen (1994: 190), pregar in the courtois sense of ‘court, woo’ is transitive. From our point of view, this fact should prove further that the semantic specialization of these verbs stems from their syntactic construction.

128

A. Bartra Kaufmann f. ainz pri Dieu que la mort me doinst therefore I-ask GodAcc that the death me he-gives (La Chastelaine de Vergi, Foulet 1974)

(OF)

(7)

a. Sson malvats; aquests aytals no sseran perdonats pel they-are evils these ones not will-be forgiven by-the savi hom . . . (OC) wise man . . . (A. March, AITCM) b. ny los perdones a varon ny a muger (OS) nor themAcc you-forgive a man nor a woman (Faz. Ultramar, Riiho 1988) c. vos, mon Wlh, los avetz totses perdonatz (OO) forgiven you my son themAcc you-have all (Jensen 1994) (8) Rey, mon senyor, suplicau De´u Wnida Sia per ell la King my lord beg GodAcc Wnished will-be by him the cosa mal seguida (OC) thing bad pursued (J. B. Masdovelles, apud AITCM) (9) Blanquerna aconsela` lo pastor con degue´s aconsolar Blanquerna advised the shepherdAcc how he-should comfort sa muller (OC) his wife (Llull, Blanquerna, apud DECLC)

The examples in (10)–(13) correspond to the Dative construction: (10) a. . . . e lo prom li prega` que li mostra`s and the man himDat asked that himDat he-showed com ho faı¨a (OC) how it he-did (Eiximenis, DECLC) b. Ella ruegue a Cristo por los pueblos errados (OS) she should-pray to Christ for the people mistaken (Berceo, Cuervo) c. prec li que tenha char s’ amor (OO) should-pray him=herDat that he/she-should-value his/her love (Jensen 1994)

Accusative Alternations d. Et qui nul bien t’ ore ne prie and who any good youDat demands nor asks (Tobler and Lommatzsch 1966)5 (11) a. Que perdona`s alcun maleWci a aquels que les that he.would.forgive some curse to those that them haurien acceptades would.have accepted (Jaume I, Furs, AITCM) b. La prison li pardone de gre´ et volontiers the prison himDat forgives willingly and intentionally (Tobler and Lommatzsch 1966) c. Ja Dieu ne leur puist pardonner already God not themDat can forgive (Tobler and Lommatzsch 1966) d. Lo mal qu’ el fan perdona lor the evil that himAcc they-made forgives themDat ‘He forgives them the evil that they caused to him’ (Jensen 1994) (12) A Tu, mon De´u, tostemps suplicare´ to you my God always I-will-beg (J. B. Masdovelles, apud AITCM) (13) Per ac¸o` lo cellereer consella` a l ’ abat e al for this the cellar monk advised to the abbot and to-the que destrysen . . . that they-destroyed . . . (Llull, Blanquerna, apud DECLC)

129 (OF)

(OC)

(OF)

(OF)

(OO)

(OC)

prior priest (OC)

5

In Old French the verb prier ‘pray’ presented also a third type of complementation: a human Dative complement and a complement of matter introduced by de ‘of’: (i) de tel chose li pre¨a of that thing himDat prayed (apud Tobler and Lommatzsch 1966) (ii) Si li prı¨ent de retorner back If himDat prayed of come (apud Tobler and Lommatzsch 1966)

(OF)

(OF)

130

A. Bartra Kaufmann

In Modern Catalan, these verbs are constructed with a human Accusative argument when there is only one argument. If two arguments are present, the human argument takes preferably a Dative form: (14)

a. No el preguis not himAcc you-pray b. La perdonare´ herAcc I-will-forgive

(15) a. Li prega que l’ ajudi him=herDat he/she-prays that him=herAcc he/she-helps b. No li va perdonar la traı¨cio´ not him=herDat he/she-forgave the betrayal

(MC) (MC) (MC) (MC)

9.2.2 Verbs that in old language preferred the Accusative construction

Other semantically and syntactically related verbs show a much clearer preference for one or another structure at each historical period. For example, Cat. absoldre ‘acquit, absolve’ takes, in a great majority of cases, a human Accusative and a de matter complement: (16)

a. Io A. Br. juge [ . . . ] absolvu la ecclesia ja dita de I A. Br. judge acquit the church already mentioned of Organya` de totes despeses e . . . Organya` from all charges and (OC) (Coromines 1971: 20, n. 3) b. Que e´l los absuelva de todos los pecados (OS) the sins that he themAcc absolves of all (Berceo, Cuervo) c. De tot ice t’ asolon (OF) of all that youAcc absolve (apud Tobler and Lommatzsch 1966) d. E v’ asolvo ben de questo pla (OI) plan and youAcc I-absolve good of this (Colussi 1987)

The Dative construction is also found, but it is much more rare: (17) Vos absol lo dan youDat I-absolve the damagesAcc (J. B. de Masdovelles, AITCM)

(OC)

Accusative Alternations

131

As for privar ‘deprive’, or notar ‘notice’, no examples of the Dative construction were found in OC and OS: (18)

a. ( . . . ) punia.l e privava.l d’ aquell he-punished-him and he-deprived-himAcc of that oWci (OC) profession (Llull, Blanquerna, DECLC) b. Los amichs no.l ploren [ . . . ] per c¸o com lo veen the friends not-him cry for this how himAcc they-see destituı¨t e privat dels be´ns temporals (OC) lacking and deprived of-the goods material ‘The friends do not cry for him, because they see him deprived from material goods.’ (Metge, Somni; DCVB)

(19) Y porque no parezca que los nota de poco and in order that not it-seeming that themAcc he-notes of less sabios (OS) wise . . . (Leo´n, Expos. De Job, apud Cuervo)6 Summing up, it can be asserted that in Old Romance both structures already existed, that the Accusative construction was clearly dominant over the Dative construction, and that some verbs, such as privar or notar, lacked the Dative construction. Therefore, there is a synchronic and a diachronic problem to be faced. On the one hand, the simultaneous existence of two structures for one single verb has to be explained. On the other, the fact that one structure becomes less and less used, to the extent that it appears to get lost, merits an explanation. Let us face Wrst the synchronic problem. As has been extensively argued by several people concerning other parallel alternations, such as locative alternation and dative alternation, there arise severe theoretical and empirical diYculties to derive one structure from the other.7 Baker’s Uniformity of Theta Assignment Hypothesis (1988) is incompatible with such a derivation 6

Example of a late stage: Siglo de oro, sixteenth to seventeenth centuries. There are, nevertheless, some predicates speciWc as for the [
animate] feature of the DO. In Modern Spanish acusar ‘accuse’ takes a human DO, and achacar ‘blame’ an inanimate DO: 7

(i) a. Juan acuso´ a su madre de sus fracasos Juan accused to his mother for his failures ‘Juan accused his mother for his failures’

132

A. Bartra Kaufmann

and would rather require the two structures to have diVerent theta roles.8 Furthermore, no restricted principles can be established to derive one structure from the other. From a diachronic point of view, the question of why some verbs that in Old Romance presented the two structures maintain them now, whereas others had only (6a) in the old language and (6b) in the modern language, suggests that a bunch of minimal grammatical features interrelate with a diVerence in weight. We argue that the two coexisting structures are two diVerent possibilities provided by Universal Grammar for the double Accusative structures in Latin to evolve when overt case markings get lost. Furthermore, the two structures can coexist because there is a (minimal) diVerence in meaning between them. Grammaticalization processes aVecting some Romance prepositions and verbal aYxes contribute to the assimilation of the basic meaning of the verbs to prototypical frames. In the next section, we will present the double Accusative structures in Latin and will try to provide an analysis of them.

9.3 The Double Accusative in Latin Three groups of verbs that take two Accusatives in Latin have been distinguished in the literature.9 I. The Wrst group contains verbs that take an Accusative of matter and a predicate, as in (20): (20)

malitiam sapientiam judicant ‘They regard malice as wisdom.’ (Cicero, OV. 2.10; apud Pinkster 1990)

b. *La the (ii) a. Juan ‘John b. La ‘The

policı´a de carretera acusa al mal traYc police accuses to-the bad achaca sus fracasos a su madre blames his failures on his mother’ policı´a achaca los accidentes al mal police attribute the accidents to the

tiempo de los accidentes weather for the accidents

tiempo bad weather’

8 Cf. for Locative Alternation: Demonte (1991: ch. 1), Goldberg (1995), Rappaport and Levin (1998), and Mateu (2002) and critical references within. For double object constructions vs. prepositional dative constructions, see, among many others, Larson (1988), JackendoV (1990), Goldberg (1995), Harley (2002). For an alternative view, based on his structure preserving movement theory, see Emonds (1993). 9 Ku¨hner and Stegman (1914), Pinkster (1990), Valentı´ Fiol (1960), among many others. As for Pinkster’s examples, the English glosses are his.

Accusative Alternations

133

II. The second group contains verbs that mean ‘to teach’, ‘to ask’, ‘to demand’, ‘to inform’, ‘to hide’: (21)

Te hoc beneWcium rogo ‘I ask you this favor.’ (Cicero, Att.14.13A.3; apud Pinkster 1990: §5.2.2)

(22)

Racilius . . . me primum sententiam rogavit ‘Racilius Wrst asked my opinion.’ (Cicero, Q.fr. 2.1.3)

(23) Non te celavi sermonem T. Ampii ‘I have not kept the conversation with T.A. from you.’ (Cicero, Fam. 2.16.3) A great majority of the Romance verbs that we study are semantically, syntactically, and etymologically related with the verbs of group II. III. The third group is made up by ‘compound verbs with a prepositional preWx. The preWx is related to a preposition which governs the accusative’ (Pinkster 1990: §5.2.2): (24)

equitum magnam partem Xumen traiecit ‘He brought a large part of the cavalry across the river.’ (Caesar Civ. 1.55.1)

In the passive, the constituent connected with the prepositional preWx remains unchanged: (25) ne maior multitudo Germanorum Rhenum traducatur ‘Lest a greater number of Germans be brought across the Rhine.’ (Caesar Gal. 1.31.16; apud Pinkster 1990: 46) Pinkster (1990) also notes the existence of three-place verbs with two constructions. The example he gives from such verbs is donare. Besides the usual construction with Accusative and a Dative with the semantic function of Recipient (verbs of transfer) or Addressee (verbs of communication), these verbs also have an Ablative þ Accusative construction: (26) anulus aureus quo tu istum in contione donavisti ‘The gold ring with which you endowed him during the meeting.’ (Cicero, Ver. 3.185; apud Pinkster 1990: §5.2.4) It would be desirable that one and the same structure could be postulated for all the types of Latin double Accusatives. Pinkster (1990) and Woolford (1997), among many others, notice that in group II verbs, the Person

134

A. Bartra Kaufmann

Accusative is the one that gets promoted into subject position in Latin Passive sentences. This fact suggests that if we Wgure out the double Accusative construction in a Larsonian structure, the Person Accusative would occupy the innermost position in a structure like (27): (27) a. [VP [NP a letter][V0 [V send][PP to Mary]]] b. [VP [Spec V’ [V0 [V send][VP [NPi Mary [V0 [V [V t][NPi e]][NP a letter]]]]]] 10 Nevertheless, we will not adopt a pure Larsonian analysis. We will keep Larson’s basic conceptual idea that in the predicate [send a letter to Mary] the NPs/DPs a letter and Mary stand in the relation of a subject and an object. But we will reinterpret it in a lexical structure framework that relies more heavily on a compositional analysis of the lexical and semantic properties of predicates. Following basically an idea put forward by Kayne (1994) and further developed by Hale and Keyser (1993, 1997, 1998), but also by Mateu (2002) and Harley (2002) among others, we hypothesize that in all the cases of Latin Double Object constructions, the two accusatives form a small clause. The diVerence lies in the element that articulates the small clause, the head of the predicative structure. Unlike Larson, we suppose that the preposition to of send a letter to Mary is not a case marking one (cf. Larson (1988: 352)), but rather it is the element responsible for the semantic relationship between a letter and Mary. In sentences like (20), the semantic value of the predicate is to equate one Accusative with the other. The predicate Accusative is related to the matter Accusative Object by means of a central coincidence empty preposition, which undertakes the job of relating the Object, a FIGURE, with the predicate, acting as a GROUND.11 We can represent this preposition as an abstract as. This analysis receives further support from an identical construction that Modern Romance languages exhibit in which the P is overt. Some of them are presented in (28): (28) a. El considero (com) el meu millor amic the my best friend himAcc I-consider as ‘I consider him my best friend.’ b. El tinc per honest himAcc I-have for honest ‘I take him for honest.’

10 11

For indirect passives, see Larson (1988: 363). In the sense of Talmy (1985), Mateu and Rigau (2002), Mateu (2002).

(MC)

(MC)

Accusative Alternations c. Enguany la tinc de professora this present year herAcc I-have of teacher ‘This year she is my teacher.’

135 (MC)

If we represent the empty preposition as ‘as’, we obtain the structure in (29):12 (29)

[V0 [judicant [PP malitiam [P0 [‘as’ [ sapientiam]]]]]]

At overt syntax, the verb judicare being a transitive verb, the subject of the small clause can check its accusative case in the usual way at Spec,vP and the predicate of the small clause gets its features copied in the usual way by means of the functional projections associated to the small clause. This analysis can be extended to the sentences of the type in (21)–(23), in which the two Accusatives also form a small clause headed by an empty P. In this case, we represent the meaning of the preposition by means of ab: (30)

[VP rogo [PP [DP te][P0 [P ‘ab’][DP hoc beneWcium]]]]

The transitive VP is selected by vP. Both DPs agree with v, their probe, in the two speciWers of vP.13 The structure of the third type of Latin verbs with two accusative complements is presumably more complicated, but also more interesting. These are preWxed verbs. As pointed out by Pinkster, the preWx selects the morphological case of the second argument of the verb. Also, this preWx determines the semantic relation between the two arguments in the small clause selected by the verb. We can express this semantic relation by taking the preWx to be the head of the small clause. This preWx incorporates into the verbal root. The only diVerence between this third class of verbs and the other two reduces to the fact that the preposition is overt instead of covert and that this P incorporates onto the verbal root. The core part of the structure is presented in (31):

12 It could also be argued, following Kayne (2001), that the preposition only ‘looks’ internal to the small clause, but it is in fact the complementizer of the small clause or, in the spirit of Kayne (2002), that the preposition merges to the VP in order to mark that its complement is a complex one. 13 At least for Modern Romance, the only necessary agreement between the subject and the predicate of the small clause is the one of case. Number and gender are lexically determined:

(i) Considera a sus hijas el ba´culo de su vejez (MS) he/she-considers to his/her daughtersfem:pl the supportmasc:sg: of his/her old years ‘He/she considers his/her daughters as his/her comfort in old age’ (ii) Considera el comiat la millor solucio´ (MC) solutionfem he/she-considers the farewellmasc the best ‘He/she considers that the farewell is the best solution’

136

A. Bartra Kaufmann

(31) [VP [V0 [Vj trai [V ducerej ]] [VP [V0 [V [tj ][PP [DP equitum magnam partem] [P0 [P ti ] [DP Xumen]]]]]]] In (31), ti is the copy of the preWx incorporated to the governing verb, and tj is the copy of the verb plus preWx raised to an empty CAUSE verbal head. Summing up, Latin double Accusative structures are instances of small clauses articulated by an empty P or an overt preWx incorporated onto the verb. In the next section, we will consider the way in which these structures evolved in Old Romance languages.

9.4 Two Structures at all Historical Periods In this section we will argue that in the cases where the two structures coexist at the same time,14 they correspond to two diVerent lexical argument structures of the verb.15 We will provide several theoretical and empirical arguments supporting this hypothesis. From a diachronic point of view, the structure missing in modern languages did cease to be used in favour of the dominating alternate one. The reasons for this move are to be found partly outside the structures under analysis, in the general process of grammaticalization of the prepositions that headed the small clauses. In this way two cycles of evolution are found in these structures. First, in Late Latin, prepositions conveyed semantic relations that were not clearly expressed by the morphological case of the arguments, and these structures were inherited in Old Romance. Then, a process of grammaticalization of these prepositions led to a reorganization of the lexical structures and to a change of the way arguments were licensed in syntax. Yet, in order not to derive one structure from the other, another scenario of discussion could be opened. It could be argued that the verbs under discussion take only one single argument and that the dative or oblique complement is an adjunct. This assumption would mean that a verb like perdonar ‘forgive’ has only one argument, a direct object referring either to a human or an inanimate entity: (32)

14

a. perdonar els pecadors (dels seus pecats) forgive the sinners of-the their sins

(MC)

This view is also supported by some philologists. See, for instance, Rohlfs (1969: §640): ‘In non pochi casi si puo` osservare oscillazione tra dativo e accusativo. Accanto all’odierno gli domandava, la lingua antica presenta lo domandava ( . . . ) Accanto a servire un signore, si dice servire a Dio [In many cases can be seen oscillation between dative and accusative. Besides the modern gli domandava, the ancient language showed lo domandava ( . . . ) Besides servire un signore, it was said servire a Dio]’ 15 Even a third structure will be identiWed. See section 9.6.

Accusative Alternations b. perdonar els pecats (a algu´) forgive the sins to somebody

137 (MC)

As for (32a), the oblique complement dels seus pecats would be interpreted as an ‘about’ complement, like the one in (33a). The dative complement in (32b) would be analysed as a possessive dative, parallel to that found in sentences like (33b): (33) a. parlar de talk about b. embenar la bandage the

polı´tica politics cama a algu´ leg to somebody

(MC) (MC)

It has to be said that the construction with only one argument—without a prepositional complement—has existed at all periods. However, in such a case, the predicate receives a holistic and telic interpretation in which the intended missing adjunct is interpreted as a universal quantiWer. In section 9.6 we will consider the structure of the verbs when constructed with a single human Direct Object. On the other hand, we consider the construction with a single non-human complement as a case of contextual ellipsis. We have already presented theoretical arguments against the derivation of one structure from the other one. In this section, we concentrate on several empirical arguments showing that there is a slight but real diVerence in meaning when constructed with an animate object or with an inanimate object, and that the construction of this meaning needs both arguments. 9.4.1 The holistic interpretation

In a way rather similar to the locative alternation, the construction with the person accusative has a holistic meaning, whereas the one with a matter accusative does not necessarily have a holistic meaning: (34)

a. el va perdonar del tot ‘He/she forgave him completely.’ b. ??/ *va perdonar els pecats del tot s/he-forgave the sins completely ` c. ??/*et perdono, pero no del tot you I-forgive but not completely d. els pecats es perdonen, pero` no del tot ‘the sins are forgiven, but not completely.’

(MC) (MC) (MC) (MC)

138

A. Bartra Kaufmann

9.4.2 Achievements vs. accomplishments

Closely related to the property presented above, the two alternating versions of the verbs studied behave diVerently with reference to the classical tests used to discriminate between achievements and accomplishments. For instance, diVerent degrees of grammaticality are obtained as durative or punctual adverbs are added: (35)

a. Va pregar he/she-prayed b. Va pregar he/she-prayed

De´u {durant una hora / God during one hour cleme`ncia {??durant una mercy during one

??a les cinc} (MC) at Wve hora / a les cinc} (MC) hour at Wve

The facts in 9.4.1 and 9.4.2 show that diVerent types of events are implied with an animate or inanimate complement. 9.4.3 Absolute constructions

Absolute constructions are more acceptable with human complements. (36) a. Absolt el absolved the confessionary confessionary b. ??Absolts els absolved the (37)

penitent, el mosse`n va abandonar el penitent the priest abandoned the

pecats, el penitent se’n va sins the penitent got away a. Perdonat el culpable, el ca`stig no forgiven the guilty man the punishment not b. ??Perdonades les culpes, el reu se sentı´ forgiven the faults the culprit felt

(MC) anar

(MC)

tenia sentit (MC) had sense alleugerit (MC) relieved

(38) a. Aconsellada pels seus pares, la noia va trencar amb advised by her parents the girl broke it oV with xicot Wance´ b. ??Aconsellat el divorci, l’ advocat se sentı´ deprimit having advised the divorce the lawyer felt depressed

el her (MC) (MC)

9.4.4 Passive

In today’s Catalan, each verb prefers a diVerent type of subject in the passive sentence, according to the construction preferred. Therefore, for the verb perdonar inanimate subjects are better in passive sentences, whereas with a verb like absoldre animate passive subjects are preferred:

Accusative Alternations (39)

a. Aquestes ‘These b. ??Aquests ‘These

coses things espies spies

no are no are

es never es never

perdonen mai forgiven.’ perdonen mai forgiven.’

139 (MC) (MC)

(40) a. En Riera no va ser absolt del robatori ‘Riera was not absolved of the robbery.’ b. ??Els pecats seran absolts nome´s a qui se’n penedeixi ‘The sins will only be forgiven to the one who will repent.’

(MC) (MC)

9.4.5 Pronominalization

The behaviour of the prepositional complement with respect to pronominalization shows its argumental status. Compare our de complements in (41) with (42), where adjunct de-phrases are intended to pronominalize: (41) a. D’aquestes faltes, ja n’ han absolt of these faults already of-it they-have absolved cops en Martines (MC) times the Martines b. D’aquestes ofenses, nome´s en podria perdonar Of these oVences only of-it I-could forgive molt estimat very beloved c. Tots te .n pregam (we) all you of-it we-pray (Canc¸oner, AICTM) d. Suplicau-ne la reyal majestat . . . pray-of-it the Royal Majesty (Canc¸oner, AICTM) (42)

a. *De of b. *De on

dos two

algu´ somebody (MC)

broma, en parla sempre joking of-it he/she-speaks always puntetes, en ballen les ballarines tiptoe of-it they-dance the dancers

(MC) (MC)

9.4.6 Weak islands

In the same line, we Wnd that the oblique complements of the perdonar-type verb scan cross weak islands:16

16

Thanks to Luigi Rizzi for suggesting the arguments in 9.4.5 and 9.4.6.

140

A. Bartra Kaufmann

(43) a. De quin deute lamentes que hagin perdonat en Gabriel? (MC) ‘From which debt do you regret that they have absolved Gabriel?’ b. De que` lamentes que hagin absolt per segona vegada ‘From what do you regret that they have absolved Martines en Martines? (MC) for the second time?’ (44)

a. *Amb quin argumenti lamentes with which argument you-regret en Gabriel Øi ? the Gabriel b. *Per por de queei lamentes que by fear of what you-regret that els estudiants Øi ? the students

que hagin perdonat that they-have absolved (MC) no hagin perdonat not they-have absolved (MC)

Summing up, the number of arguments presented show that there is a slight but real diVerence in meaning between the two structures, and that the oblique complement behaves like an argument when present in the structure. In next section, we present our proposal with reference to the argument structure of abstract transfer verbs.

9.5 The Lexical Argument Structure It is a classical line of investigation within diachrony and Universal Grammar to establish a causal connection between the loss of morphological case markings (in Latin) and the emergence of a more Wxed word order, together with the development of prepositions as markers of syntatic functions (van Kemenade and Vincent 1997, Lightfoot 1999, Pintzuk, Tsoulas, and Warner 2000, among many others). It has also been noted that the rise and generalization of prepositions in Latin precedes the fall of case. In eVect, several philologists (e.g. Perrot 1966) have observed that the Latin case system did not express the variety of semantic relations. In a number of relations, the preposition expressed the semantic relation conveyed by the argument, whereas the case was determined by a grammatical servitude. Folgar (1993) notes that in late Latin the P ad was used in combination with Accusative case to express the Indirect Object.17

17

It has to be said, nevertheless, that several hypotheses have been made in order to explain the coexistence of prepositions and case in Latin. See Sua´rez Martı´nez (1999) for a summary of these positions.

Accusative Alternations

141

We propose that Old Romance abstract transfer verbs exhibited the same structure as their Latin double Accusative antecedents, i.e. that the semantic relation between the two arguments is conveyed by means of a preposition. In Romance the preposition is overt, whereas in Latin it was covert. Following Hale and Keyser (1997, 1998) or Harley (2002), we assert that the two constructions under analysis diVer in the choice of the preposition that relates the two arguments, as we saw also for the Latin counterparts.18 Let us now examine each one of the structures. 9.5.1 The accusative structure

The relevant part of a sentence like (16a), repeated here as (45), would have the lexical structure represented in (46), where V2, incorporated in a causative abstract node V1, is the verb absolvu. (45)

absolvu [la ecclesia (ja dita de Organya`)] [de totes despeses] I-acquit the church . . . . . . . . . from all charges 0

(46) [VP-1 la ecclesiai [V-1 absolvu [VP-2 [DP-1 ti ] [V0 [V-2 tv] [PP ti [P de] [DP-2 totes despeses]]]]] The P is a Wnal or terminal coincidence one, which expresses a source relation between its complement and its speciWer. Once the relevant functional categories are projected at the syntax, DP1 in (46) is interpreted as the DO by checking the uninterpretable features of v at Spec,vP. As for DP2, there is no adequate functional element with an uninterpretable feature that attracts it before spell-out. Therefore, NP2 remains in its position. 9.5.2 The dative structure

In this case, the preposition which articulates the two arguments is a terminal coincidence preposition, clearly related to the ‘transfer’ meaning of the verb. The structure is represented in (47): (47)

18

0

[VP-1 lo dan [V-1 absoldre] [VP-2 [DP-1 ti ] [V0 [V-2 tv] [PP ti [P a ] [DP-2 algu´ ]]]]]

See, for instance, Hale and Keyser (1997): ‘we assume that the relation between the to-dative and the double object construction consists in the choice of preposition. The central coincidence preposition of the double object construction is generally non-overt in English, though a few verbs of the language do select an overt preposition, typically with, for this function (and many languages use an oblique case here; see, e.g. Dixon, 1971, for Dyrbal; and see Bittner, 1993, for Inuit)’ (Hale and Keyser 1997: 37).

142

A. Bartra Kaufmann

This structure is not simply the ‘mirror image’ of (46) with another P. The grammaticalization processes in which the two prepositions were involved determined the further reinterpretation of the structures that gave rise to their modern use. In Old Romance, the two arguments were licensed in syntax in the usual way, DP-1 at Spec,vP and DP-2 in situ by the P. Our next step is to investigate the reasons why Old Romance languages preferred to use the accusative construction whereas the modern languages prefer dative or ditransitive constructions.

9.6 Diachronic Evolution As said before, the preposition a ‘to’ was already adopted in Latin and afterwards in Romance to express the GOAL or BENEFACTIVE relation with transfer predicates. The verbs we analyse are (physical or abstract) transfer verbs. Furthermore, in many cases, there exists a possessive relation between the complement of matter and the complement of person. Take for instance the paternoster prayer in an already old version: (48)

perdoneu-nosi els nostresi pecats ‘forgive us our sins.’

If we analyse the possessive relation as in Kayne (1994) and subsequent work, we can suppose that the P introduces a constituent in which there is a possessive relation encoded, as in (49). (49)

els[D=PP pecatsi [a[IP en Pau[I ti ]]]

This, let’s say clausal, structure could also be on the basis of (47). It is unnecessary to point now to the amount of cases in Romance and other languages where there is an alternation between a Dative complement and an adnominal genitive possessive (perdonar els pecats a en Pau ‘forgive the sins to Paul’ / perdonar els pecats d’en Pau ‘forgive Paul’s sins’). Therefore, there is a conspiracy of factors that led to making this construction very productive. In any case, the ditransitive structure has been maintained until nowadays with the same meaning and structure. The strong hypothesis of the Syntactic Invariance (see Longobardi 2001b) can be maintained. Even if we suppose that the Dative preposition got grammaticalized, no additional assumptions have to be made. If so, the grammaticalized P would be a functional projection over the PP, and the DP would have to adjoin to its speciWer in order to get its features checked.

Accusative Alternations

143

As for the evolution of the Accusative construction, another set of factors conspired in order to make this structure the less used. We list the main factors in (50): (50)

a. Arguments with the theta role of SOURCE are much more easily eliminated from a sentence than GOAL arguments. b. The motion-from-a-point meaning is not transparent, nor evident, in the verbs under analysis. c. In a high number of cases, the de complement was clausal. Finite clausal complements delete the preposition in nearly all Romance languages. d. The preposition de underwent a clear grammaticalization process in several diVerent constructions.

In (51), we exemplify the facts listed in (50): (51)

a. Sempre surten tard ‘They always come out late.’ b. El perdonen de tot ‘They forgive him of everything.’ c. . . . prega` lo rey que perdona`s . . . [¼ (6a)] prayed the king that he-would-forgive d. la Wlla del rei j intento de veure’l j the daughter of-the king I-intend of see him

The facts mentioned above interacted to mask the semantic source relation encoded by the small clause and the functional relations of the arguments. Moreover, the fact that the DO is an animate DP permits also an interpretation of the verb as a telic verb of change of state. We assume, following Tenny (1987, 1989, 1992), that telic predicates have an argument that ‘measures out’ the event, bringing about the aspectual property of delimitedness. Usually, the accusative object is the argument that delimits the event denoted by the verb. We suggest that verbs such as perdonar, absoldre, etc. were interpreted as delimited events with an animate object. This fact, together with the grammaticalization of the preposition de, led to an interpretation of the verb as a change of state one. Therefore, the PP de totes despeses ceased to be an argument and acquired the status of an adjunct (perhaps an L-marked adjunct in the sense of Chomsky 1986, an element that, whenever present, completes the semantic content of the VP). We summarize the facts involved in the evolution of the structure in (52): (52) a. The verb selects a PP with a Wnal coincidence P. b. The meaning of the verb has no more transparent ‘source’ meaning.

144

A. Bartra Kaufmann c. The preposition gets grammaticalized. d. The speciWer of the PP is the argument that ‘measures out’ the event.19 e. The complement of the P is no more necessary for the nuclear predication to be complete, and becomes an adjunct.

As a result of this complex process, two competing structures arise: the SOURCE constructions with two arguments and the telic change of state construction with only one argument. Given the fact that the grammar has the dative structure to express the transfer meaning when two arguments are present, the SOURCE construction gets (almost) abandoned. Therefore, when used with an animate object, verbs such as perdonar, pregar, aconsellar, suplicar, absoldre, etc. have to be interpreted as causative change of state denominal verbs in the sense of Hale and Keyser. They have a paraphrase like ‘cause to become with N’. In this case, the direct object had to be interpreted as Na , i.e. the speciWer of V2 in the structure (53): (53)

[V* [V1 ] [VP [NP [Na alg u ] [V0 [V2 ] [PP [P0 amb [NP [Nb perdo ]]]]]]]]

An argument favouring this view comes from a correspondent synonymous construction with a light verb and a noun morphologically related to the verb. See the pairs in (54)–(57): (54) aconsellar algu´ / donar (un) consell a algu´ ‘to advise somebody / to give advice to somebody’ (55) pregar (a) algu´ / fer un prec a algu´ ‘to ask somebody for / to make a request to somebody’ (56) perdonar algu´ / donar el perdo´ a algu´ ‘to forgive somebody / to give forgiveness to somebody’ (57) absoldre algu´ / donar l’absolucio´ a algu´ ‘to acquit somebody / to give absolution to somebody’ 9.6.1 PreWxed verbs

In the case of verbs such as absoldre ‘absolve’ or deslliurar ‘release, exempt’, there exists a relation between the components of the complex structure of the verb and the lexical structure selected. The semantic relation between the verb 19 See also Ackerman (1992) for further support to the idea that telicity is involved in this process. In his study of Hungarian locative alternation constructions, he shows that the GOAL argument must bear Accusative marking and function as a DO when it co-occurs with complex predicates containing either a perfective or a resultative aYx.

Accusative Alternations

145

and the prepositional small clause that it selects is much more transparent. In fact, the verbal preWx is the element governing the choice of the P selected; it determines the formal, categorical, and semantic nature of the arguments selected by the verb. The preWx denotes a source relation, which is implemented by the head of the PP.20 Mateu (2001a) argues that in German, a preverb such as ver- in verga¨rtnern is the head of an independent projection to which the verbal element is conXated. Adapting this conceptual view to our cases, it can be proposed that the preWx not only selected the P, but rather it was originally itself the head of the PP selected by a verb (also derived by incorporation of N into empty light verb). Our structure, reinterpreted in the spirit of Mateu’s, is that in (58): (58)

[VP [V
[Vabi [V solvu]] [pp [DP ecclesia] [P
abi [DP totes despeses]]]]

Meanwhile, the semantic composition of the verb abþsolvere ‘to release from’ began to lose transparency. Consequently, the P head of the phrase governed by the V also lost part of its semantic content and underwent a grammaticalization process. Summing up, we have argued that the semantic relation between the two arguments is expressed in Latin by means of an empty preposition and in Romance by means of a lexical preposition. If the preposition is a Wnal coincidence one, meaning SOURCE, the person argument occupies the speciWer (of the light verb) position and the preposition gets grammaticalized. This fact, together with the aspectual properties of the predicate, which gets telic closure with the Wrst argument, gives rise to the simple transitive structure. If the preposition means GOAL, the two arguments are needed in the nuclear predication and we obtain the ditransitive structure. Primary Sources Arxiu informatitzat de textos catalans medievals. Bellaterra: Seminari de Filologia i Informa`tica de la UAB. [Abbr. AITCM] Coromines, J. (1980–91). Diccionari etimolo`gic i complementari de la llengua catalana. Barcelona: Curial. [Abbr. DECLC] Alcover, A. M., and Moll, F. B. (1930–61). Diccionari catala` -valencia` -balear. Palma de Mallorca. (http://dcvb.iecat.net)[Abbr. DCVB] Cuervo, R. J. (1951–94). Diccionario de construccio´n y re´gimen de la lengua castellana. Bogota´: Instituto Caro y Cuervo. [Abbr. Cuervo]

20

Remember also group III of Latin double accusative verbs (trans-ire).

This page intentionally left blank

Part II Parametric Variation

This page intentionally left blank

10 Parametric Comparison and Language Taxonomy CRISTINA GUARDIANO AND GIUSEPPE LO N GOB A R DI

10.1 Introduction It is still a widely held opinion that the contribution of formal syntactic theories to modern linguistics is essentially focused on synchronic generalizations and that it does not extend to the most classical problems of the historical paradigm. Such a belief is beginning to fade away as a consequence of increased work in diachronic syntax (e.g. Roberts 1993a, the contributions contained in van van Kemenade and Vincent 1997, Lightfoot 1999, and Longobardi 2001b, among others). In this chapter we will argue that the generative theory of syntactic parameters might interestingly renovate another stronghold of the historical study of language, phylogenetic taxonomy, and undermine the Humboldtian received distinction between the latter and grammatical typology.

10.2 The Challenge of Historical Adequacy Longobardi (2003b) tried to relate the synchronic and the historical study of formal grammar within a uniWed framework made available by the rise of ‘parametric linguistics’. To clarify the goals of this enterprise, Wve levels of adequacy were deWned, two more than in Chomsky’s (1964) classical approach, each deWned by a general question on language: We are indebted to Lucio Crisma, Paola Crisma, Chiara Gianollo, and Francesco Guardiano for inspiring discussions and advice on several mathematical and empirical issues. We are also especially grateful for help with data and hypotheses to Manuela Ambar for Portuguese, Roumyana Slabakova for Bulgarian, and Abdelkader Fassi Fehri for Arabic. The two authors worked jointly on the whole content of this article; however, as far as legal requirements are concerned, C. Guardiano takes oYcial responsibility for sections 10.6, 10.7, 10.8, and 10.9, and G. Longobardi for the remaining parts.

C. Guardiano and G. Longobardi

150

(1) 1. What are the recorded samples of human linguistic behaviour? 2. What are the actual human languages? 3. What are the biologically possible human languages? 4. Why do we have precisely these actual languages? 5. Why do we have precisely these biologically possible languages? 1. Chomsky’s observational adequacy (philology, corpus linguistics) 2. Chomsky’s descriptive adequacy (descriptive synchronic linguistics) 3. Chomsky’s explanatory adequacy (general linguistic theory) 4. Actual historical adequacy (historical linguistics, cultural history) 5. Evolutionary adequacy (phylogenesis of the language faculty, natural history) The level of adequacy relevant for our present purposes is the fourth one: that concerned with the cultural history of language. Attaining such historical adequacy with the tools of formal syntax was suggested to be representable, again, as a stepwise process, involving Wve degrees:1 (2)

(I) (II) (III) (IV) (V)

mechanical description of closed corpora description of grammatical changes as parameter resetting explanatory restrictions on grammatical changes comparison and reconstruction of grammatical patterns evaluation of genetic kinship

Although each level seems to enjoy some logical or practical priority over the successive ones, it is not at all impossible to grasp Xashes of understanding of higher levels while investigation is still ongoing or just beginning at lower levels. Indeed, pushing the inquiry to its extreme goals may a fortiori, in the case of successful results, grant hope for the achievement of historical adequacy at lower levels, ultimately for the success of the whole enterprise of studying history through sophisticated cognitive models. It is also for this reason that the present research aims directly at degree (V) adequacy.

10.3 Background Assumptions: Lexical Comparison and Language Taxonomy 10.3.1 The comparative method

The identiWcation of phylogenetic relatedness among languages and, as a consequence, the deWnition of language families, has so far been conducted 1

Longobardi (2001a, 2003b).

Parametric Comparison and Language Taxonomy

151

by comparing lexical items (i.e. sets of words and morphemes—whether roots, aYxes, or inXections—saliently characterized by Saussurean arbitrariness); very rarely was the task performed by comparing rules of grammar, such as the syntactic ones, which appear less arbitrarily variable across languages. With the development of the classical comparative method in the second half of the nineteenth century, the early taxonomic conclusions reached on the grounds of pretheoretical resemblance of lexical items in form and meaning have been supported by the identiWcation of exact and regular sound correspondences, adduced as a decisive proof of phylogenetic kinship. In the actual practice of phylogenetic taxonomy, this method was very successful as a proof for the most obvious language families (say Indoeuropean, Semitic, Austronesian, Bantu, among others), but it turned out not to work equally well for grouping such families into higher taxa, which are likely to have existed further back in the prehistory (long-distance comparison), nor for identifying many of the lower taxa (cf. the numerous unsolved problems of subgrouping within Indoeuropean or Romance). 10.3.2 Mass comparison

The most daring attempt to overcome the limits of long-range comparisons is Greenberg’s mass or multilateral comparison (Greenberg 1987, 2000). The basic idea of the method elaborated by Greenberg is that the absence of precision in comparing diVerent entities, namely the lack of exact sound correspondences, can be compensated for by comparing sets of lexical elements (on the grounds of mere phonetic and semantic resemblance) not just in pairs of languages, but across a larger number of languages at the same time: the increasing amount of compared languages would lower the probability of chance similarities. Epistemologically, Greenberg’s method does not provide so sharp a demarcation criterion as the one founded on precise sound laws, even though, of course, this does not automatically imply that it could not be used in agreement with sound scientiWc standards. The problem to be addressed, however, concerns the probative value of the proposed rapprochements since the comparanda are seldom obvious entities, owing to the inescapable vagueness of ‘phonetic similarity’ and ‘semantic similarity’. The discussion below will serve as a useful introduction to the relevance of probabilistic methods and standards in taxonomic linguistics.

152

C. Guardiano and G. Longobardi

10.3.3 Some preliminary evaluation of long-range comparison

Let us Wrst sketch some statistical considerations. A good deal of the longrange etymologies proposed in Ruhlen (1994), for example, connect (parts of) lexical items of the CVC form. According to Ruhlen, in a core phonological system containing seven consonants (three stops, say p, t, k, two nasals, say m, n, one liquid, and one sibilant) and three vowels, so that the CVC template can be instantiated by 7  3  7 ¼ 147 distinct forms, the probability of the same meaning to be encoded by chance in the same CVC form in two languages is 1/147. The probability of chance similarity decreases with further comparisons, thus it would be (1=147)2 ¼ 1=21,609 in three languages, (1=147)3 ¼ 1=3, 176,523 in four languages, and so on. The fact that, once confronted with actual practice, this optimistic perspective largely fades away (cf. Guardiano and Longobardi forthcoming) and that, in the view of many a scholar, hardly a single etymology proposed with the methods of mass comparison is by itself beyond doubt and able to speak for kinship should not concern us directly here. The relevant point is the fact that probabilistic values such as 1/147 and, a fortiori, its powers are already likely to be sometimes regarded as potentially signiWcant in the debates over long-range taxonomies. 10.3.4 Population genetics

Because of its shortcomings, Greenberg’s method needs at least to be complemented by other sources of evidence. Greenberg himself appealed to independent disciplines to further prove his points; one of these disciplines is population genetics (Cavalli Sforza, Menozzi, and Piazza 1994; Cavalli Sforza 1996). Theories of population genetics measure the distance among human populations with respect to a set of known genetic polymorphisms (e. g. blood groups) and the resulting groupings may coincide or not with the proposed linguistic taxonomies: in this case no question arises about the comparability of the basic entities, of course, since they are drawn from a universal and Wnite list of biological options. Let us then sum up the relevant features of the three comparative approaches discussed so far: 1. As we stressed before, within certain limited time bounds and in particularly favoured cases, the classical linguistic comparative method, traditionally based on the comparison of the lexicon, is able to provide some sharp taxonomic conclusions, immune from the need for serious probabilistic evaluation, since they are based on few patently highly improbable phenomena, like agreements in irregular morphology (cf. above) and recurrent

Parametric Comparison and Language Taxonomy

153

(‘regular’) sound correspondences. These properties largely solve the problem of the choice of the comparanda. 2. Beyond such cases, the only method so far devised, namely Greenberg’s mass comparison, still based on the lexicon, suggests taxonomic conclusions which are less rigorous and less widely acceptable because the very choice of the compared entities is rather unsafe. Obvious probabilistic questions, often of unmanageable complexity,2 arise and receive controversial answers. 3. A third comparative practice stems from population genetics: in this case no question arises about the comparability of the basic entities, of course, since they are drawn from a universal and Wnite list of biological options: a blood group must be compared to another blood group, obviously, not to other sorts of genetic polymorphisms. The only question may concern the statistical and empirical signiWcance of the similarities so discovered. It is also for this reason that such disciplines have been considered so useful to complement the methods of historical linguistics in the task of genetic taxonomy.

10.4 Background Assumptions: Parametric Variation A crucial step forward toward the theoretical and historical study of syntactic diversity is represented by the rise of parametric linguistics (cf. Longobardi 2001a, 2003b for an up-to-date presentation of the research programme). Most parametric analyses have reduced single instances of abstract morphosyntactic contrasts among languages to a model of grammatical variation exhibiting the following characteristics, the Wrst two of which are virtually analytic in the deWnition of parameter: (3) a. discreteness (i.e. the values of each parameter form no continuum and in the ideal case just amount to two); b. Wniteness (there is a Wnite number of parameters, hence of possible languages);3 c. limitedness with respect to the diVerences appearing on the surface (i.e. several superWcial diVerences turn out to cluster together at the appropriate level of analysis). Parametric theories suggest that a huge number of core properties of grammatical variation can be ultimately reduced to the setting of a Wnite and universal set of well-deWned biological options: in this sense 2

Cf. the discussion in some of the papers collected in Joseph and Salmons (1998). On mass comparison more generally, also cf. MatisoV (1990). 3 With the exception of the lexicon, of course.

154

C. Guardiano and G. Longobardi

parameters could be termed ‘syntactic polymorphisms’, in the optimal case ‘dimorphisms’. In the next sections we will see how parametric theories might contribute to the task of phylogenetic classiWcation.

10.5 Lexical and Syntactic Comparison A priori, the taxonomic impact of syntactic (parametric) comparison and that of traditional lexical comparison, whether performed by means of classical or long-range methods, could stand in a number of possible diVerent relations. To be concrete, let us consider the following statement of the question: (4) Given three or more languages, what is the relation between their phylogenetic tree based on syntactic comparison and the one arrived at on the grounds of lexical comparison? As a Wrst approximation, one may regard the possible empirical answer as a binary choice: (5)

a. Syntax and lexicon provide comparable taxonomic results. b. Syntax provides taxonomic results radically at odds with those of the lexicon.

We take (5a) to mean that at least signiWcant subportions of the two trees are overlapping. Now, if this is the case, three possibilities logically arise: (6) a. Syntax provides weaker insights, that is: the same taxonomic results, but only up to a shallower level of chronological depth (going back to the past, the threshold of chance is reached ‘earlier’).4 b. Syntax and lexicon provide exactly the same tree. c. Syntax provides stronger insights, that is to say: more ancient taxa can be identiWed (the threshold of chance is reached ‘later’, i.e. further back in the past). In this chapter we will begin an empirical evaluation of such possible answers to question (4). Whatever the Wnal result, it is clear that this kind of research can Wnally raise and shed light on the general question of the relative distribution of syntax and lexicon in time and space; therefore, its conclusions will be relevant for any universal theory of diachronic transmission and, ultimately, of language acquisition. Before addressing the question 4 Even in this case a method of parametric comparison might be useful for deciding about controversial cases of dialectal classiWcation.

Parametric Comparison and Language Taxonomy

155

empirically, it will be necessary to qualify better the nature and reliability of our theoretical tools.

10.6 Theoretical Premisses of Parametric Comparison 10.6.1 Mathematical signiWcance

In principle, the probative signiWcance of parametric syntax for the purposes of historical relatedness is at least no smaller than that of long-range lexical comparisons, indeed higher. It is true that the probability of two languages coinciding in the value of a speciWc binary parameter (with both values unmarked, that is: two a priori equiprobable events) is 1/2, thus slightly (though not incomparably) higher than the average 1/5 established as the correspondent value for a lot of long-range lexical comparisons (see Guardiano and Longobardi forthcoming for a more detailed analysis). Of course, as such, 1/2 has no probative value at all. However, exactly as for lexical comparison, it is easy to increase the number of comparanda: 30 binary independent parameters generate 230 languages ¼ 1,073,741,824. Now the probability for two languages to coincide in the values of 30 chosen parameters ¼ 1=230 , of three languages ¼ (1=230 )2 , that is to say: less than one in a billion of billions. But even choosing just 10 such parameters, the probability that their values coincide in three languages is less than one in a million. Of course, in the real world, one must be able to calculate the probabilistic signiWcance of partial agreements. Sticking to the usual assumption of binary parameters, the formula to compute the probability of partial agreement is the following: suppose n ¼ the total number of parameters in the sample and h ¼ the number of disagreeing (or agreeing, for that matter) parameters, then the probability of such an event will equal nno (n)h , in 2n i:e: h! h So, for example, in the case of 16 agreements and 4 divergences out of 20 independent parameters (or of 4 agreements and 16 divergences, which probabilistically is the same, but much less relevant in our domain; cf. below), the probability of this event will be computed in the following way: (20)4 20  19  18  17 116,280 ¼ ¼ ¼ 4,845 4! 432 24

156

C. Guardiano and G. Longobardi

In turn, this means a probability of 4,845 4,845 1 ¼ ¼ , 20 2 1,048,576 216:42 . . . still largely suYcient to draw conclusions. But now the enormous advantage of parametric over lexical comparison is the same as in the case of genetic polymorphisms: parametric theories provide precise comparanda virtually by deWnition, drawn from a universal list of Wnite biological (though culturally set) options (i.e. parameter values). Hence, parametric comparison does not work with the notion of resemblance and its ineliminable probabilistic uncertainties, but with identity: as in the case of genetic polymorphisms, the only question concerns a notion which is relatively easy to calculate, namely the probative value of parametric identities against chance. 10.6.2 The Inertial Theory

In order for a method of parametric phylogenetic comparison to be successful, it must be the case that parameter values exhibit a signiWcant degree of persistence through time in the history of E-languages (in Chomsky’s 1986 terms). Notice, in this context, that, according to the Inertial Theory of diachronic syntax, a conjecture proposed in Longobardi (2001b) on the grounds of Keenan’s (1994) crucial conceptual insights, parametric syntax should be among the most conservative aspects of a language, thus the most apt to preserve traces of original identity; this is true at least with respect to those parameters which are not primarily set (hence diachronically resettable) on the basis of phonological (segmental or prosodic) evidence. The Inertial Theory assumes that while semantic and phonological features can be sources of primitive changes (i.e. spontaneous deviations from the target language in acquisition), formal features (roughly in Chomsky’s 1995 terms) are ‘deeper’ and their change is in fact a direct or indirect consequence of some phonetic or semantic innovation (as is traditionally well known), or of some external inXuence on the grammar (interference, heterogeneous primary corpora).

10.7 Empirical signiWcance of parametric comparison 10.7.1 Comparing trees A straightforward method to begin to investigate question (4) of section 10.5 above is measuring the parametric distance of three or more languages against the lexical one (i.e. comparing the two types of phylogenetic trees) in cases whose actual history we already know

Parametric Comparison and Language Taxonomy

157

independently (cf. the discussion in a similar vein in McMahon and McMahon 2003). 10.7.2 The choice of parameters

A crucial step in this research strategy is the choice of the parameters to be used in the comparisons. The only viable approach, in our view, is trying to be exhaustive relatively to a limited subdomain, intrinsically (that is, neither randomly nor according to external criteria, but on the basis of its own independence of other modules) well deWned within syntactic theory itself (and suYciently vast hopefully to be representative). This strategy is the historical counterpart of the general typological approach proposed in Longobardi (2001a, 2003b) under the label ‘Modularized Global Parametrization’. For example, for the past Wfteen years a number of scholars have studied the parametric variation of the internal structure of nominal phrases in several languages: their work has discovered an amount of syntactic polymorphy, which may lead us eventually to expect a global relevance of no less than forty binary parameters (and, perhaps, no more than Wfty, though this projection is more impressionistic, pending further typological evidence). Some of the main dimensions of parametric variation have been discussed in Longobardi (2000). In order to devise a signiWcant empirical probe, we have drawn a very preliminary list of thirty binary parameters, reported in the following grid along with their presumable values in Wfteen ancient and contemporary languages. Such parameters have been tentatively proposed on the grounds of literature on and data from not only the Wfteen languages of the grid, but also independent ones outside the present sample, for example, Scandinavian languages, Hebrew, Celtic languages, Romanian, along with a few observations about Finno-Ugric. The Wfteen varieties considered in Table 10.1 are: Modern Italian, Spanish, French, and European Portuguese (It, Sp, Fr, Ptg, respectively), Latin (Lat), Classical Greek (ClGr), New Testament Koine´ (NTGr), Modern Greek (Grk), Gothic (Got), Old English (OE), Modern English and German (E, D), Modern Bulgarian and Serbo-Croat (Blg, SCr), and Standard Arabic (Ar). Some provisos are in order: the formulation of the parameters is sometimes tentative and unstable, as is obvious in a rapidly growing domain of inquiry, and parameters may diVer in terms of accuracy of description, depth of identiWed consequences, etc. Hopefully, this should not aVect most of the Wnal conclusions too severely, owing to the number of distinctions considered and to the relative sharpness of some of the mathematical results achieved.

Table 10.1.

1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17.

Parameter grid for nominal syntax

+ grammaticalized number + gramm. Count (‘null article’) + gramm. def in DP + gramm. Anaphoricity in D 3 + strong ref/def in D þ3 + number inX. On N (BNs) + ambiguous singulars + def checking APs + Art þ PP þarta + def spread on APs þ3 or þ8 + def. Spread on Mod þ3 or þ8 + enclitic def/deictic + inversely ordered As + N over external argument + N over GenO þ14 + N over low As þ15, 13 + N over M2 As þ16

It

Sp

Fr

Ptg

Lat

ClGr

NTGr

Grk

Got

OE

E

D

Blg

SCr

Ar

þ þ þ 0þ þ þ        þ þ þ þ

þ þ þ 0þ þ þ   þ     þ þ þ þ

þ þ þ 0þ þ e        þ þ þ þ

þ þ þ 0þ þ? þ   þ     þ þ þ þ

þ    0 þ   0 0 0   þ  0 0

þ  þ 0þ þ þ   þ  þ   þ  0 0

þ  þ 0þ þ þ   þ  þ   þ þ  0

þ þ þ 0þ þ þ     þ   þ þ  0

þ   þ 0 þ  þ 0þ     þ þ ? 0?

þ  þ 0þ ?e þ   0þ     þ þ  0

þ þ þ 0þ  þ         0 0 0

þ þ þ 0þ  þ   0þ     þ þ  0

þ  þ 0þ þ þ  þ 0  þ þ  þ þ  0

þ    0 þ  þ 0   þ  þ þ  0

þ  þ 0þ þ þ    þ   þ þ þ 0 0

Table 10.1.

18. 19. 20. 21. 22. 23. 24. 25. 26. 27. 28. 29. 30. a

(Continued)

+ N over high As þ17 + free APs in Mod + free Gen (non-agr.) + structural Gen þ20 + prep. Gen (vs. inX.Gen) þ21 + GenO 20 or (þ21, þ22) + GenS 20 or (þ21, þ22) + possessive pronouns + oblig. poss. Pronouns þ25 + def checking poss. þ25, þ3 + contextually inXected As + Consistency principle 13 + def. checking dem. þ3

It

Sp

Fr

Ptg

Lat

ClGr

NTGr

Grk

Got

OE

E

D

Blg

SCr

Ar

 þ þ  0þ 0 0e þ þ   þ þ

 þ þ  0þ 0 0e þ þ þ  þ þ

 þ þ  0þ 0 0 þ þ þ  þ þ

 þ þ  0þ 0 0e þ þ   þ þ

0 þ þ þ  0þ 0þ þ þ 0  ? 0

0 þ þ þ  0þ 0þ þ    ? 

0 þ  0þ 0 þ  þ    ? 

0 þ  0þ 0 þ   0 0  þ 

0 þ  0þ 0 þ þ þ ? 0 þ ? 0

0   0þ 0 þ þ þ ? þ þ þ þ

0  þ þ þ þe þ þ  þ  þ þ

0  þ þ þ þ þ þ  þ þ þ þ

0  þ  0þ 0 0 þ þ?    þ

0   0þ 0 þ  þ þ? 0   0

0  þ þ þ  þ  0 0  0 

þart must be conventionally understood as relevant only for languages having a lexical non-enclitic deWnite article and in which the latter is distinct from demonstrative pronouns, which is not the case, for example, in Modern German.

160

C. Guardiano and G. Longobardi

The setting of values is occasionally tentative, too, even more so for the non-contemporary languages of the sample, of course, for which we had to base ourselves on the analysis of particular corpora.5 Also, we are perfectly aware that very few of the labels used to deWne the thirty parameters are selfunderstandable to the reader, but the relevant explanations would necessarily exceed the limits of the present article.6 In Table 10.1, e marks the existence of limited lexical exceptions to the value set; 0 marks the irrelevance of a certain parameter in the language and results from the two conceivable types of partial interactions among parameters (Longobardi 2003b): 0 iV, as a consequence of other settings, no trigger for that parameter is ever manifested in the weakly generated language, 0þ or 0 iV, as a consequence of other settings, the surface manifestations are necessarily as if that parameter were set on one or the other of the two values (the diVerence is not always straightforward or relevant). The dependence of a certain parameter upon one of the values of another parameter is signalled by the numbers in the label of parameters 4, 5, 10, etc.: for example, þ3 or þ8 in 10 means that setting 10 is only relevant in languages already marked þ in 3 or 8, otherwise the value of 10 will be 0. In doubtful cases ‘þ?’ or ‘?’ were used; in a few cases it has so far been impossible safely to set the parameter for ancient languages, and a bare symbol ‘?’ was indicated in the relevant cell and counted as a ‘0’ for the purposes of the successive calculations. A few more clariWcations about some of the parameters assumed here can be found in Longobardi (2000) and references cited. 10.7.3 CoeYcients

The distance between any two languages can be expressed by an ordered pair of positive integers (or zero) , where n is the number of identities and m the number of diVerences. Call such pairs coeYcients. The relative distance among three or more languages can be computed, in principle, by comparing 5

We are indebted to the work of a number of colleagues for analyses and suggestions concerning the ancient and medieval languages: for Latin, Petronius’ Cena Trimalchionis, as analysed in great detail by Polo (2003) and by Chiara Gianollo in speciWc work in progress, was mainly used as a corpus, along with some materials from the Wrst century bc (Gianollo, forthcoming); by Classical Greek we refer to Attic prose of the fourth century, investigated in various respects by Guardiano (2003) on a corpus composed of three works of Plato: Apology of Socrates (399–390 bc), Symposium (385–380 bc), Cratylus (385–367 bc). Our notion of Old English is based on a corpus of prose around 1000 ad extensively studied by Paola Crisma (cf. Crisma 2000, forthcoming). Gothic nominal parameters were very tentatively set on the grounds of our own preliminary examination of WulWla’s text of the Gospels. 6 The thirty parameters used here are actually taken from a wider reference list which is now being constructed at the University of Trieste within the framework of Modularized Global Parametrization and will hopefully appear, along with commentary, explanation, and a list of their possible triggers in forthcoming work of ours with the collaboration of other researchers (Guardiano, Gianollo, and Longotardi forthcoming.)

Parametric Comparison and Language Taxonomy

161

the coeYcients of each subpair. The algebraic sum of n and m will be called the algebraic coeYcient.

10.8 A First Analysis of the Results 10.8.1 Goals of comparison

To begin with, it is necessary to clarify our goals. As often stressed, for example, by Greenberg, in phylogenetic linguistics, abstracting away from monogenist positions, two goals must be a priori distinguished: a stronger one, namely deciding about absolute relatedness between languages, and a weaker, purely taxonomic, one—i.e. establishing a relative degree of relatedness within sets of three or more languages. Pursuing the stronger goal with the devices of parametric syntactic comparison from the very beginning is too ambitious an objective. It is true that probabilistic considerations (cf. below) already suggest that the similarity exhibited, say, by Italian or Spanish with European Portuguese (coeYcient <25.1>), can hardly be due to chance (the probability is less than 1/2,500,000; cf. Table 10.2 below); but a priori we cannot always tell what is the border between chance and historically signiWcant similarity: some empirical considerations are needed. Thus, the second goal, the taxonomic one, will be primarily pursued here. 10.8.2 Anti-Babelic principle

If parametric comparison is historically signiWcant, we must expect the following empirical tendencies: given coeYcients of the assumed form , pairs of languages known to be closely related should clearly exhibit n > m, other pairs should tend toward n ¼ m, and n <m should tend toward non-existence. The last point, implicit in all work on historical linguistics, deserves a more formal statement as the Anti-Babelic principle (with obvious reference to Genesis 11, 7–9): (7) Anti-Babelic principle: similarities among languages can be due either to historical cause (relatedness) or to chance; diVerences can only be due to chance (no one ever confused languages on purpose). The Anti-Babelic principle is an obvious and reasonable postulate of this domain.7 Its main consequence for our model is that negative algebraic 7 Although something like the Anti-Babelic principle is implicitly assumed by all serious historical linguistics, it is clear that it may only have measurable eVects in domains of discrete entities drawn from a universal list, such as parametric syntax, hardly in a complex and heterogeneous Weld like the lexicon.

162

C. Guardiano and G. Longobardi

coeYcients must at least in part be due to chance; hence, in principle, when they exist, they must necessarily represent failures on the part of parametric comparison fully to capture historical generalizations—i.e. to distinguish them from products of chance. Hence, a system of parametric comparison producing too huge a number of negative coeYcients is unlikely correctly to represent historical reality and points to some ineYcient quantitative and qualitative choice of the sample of parameters. An a priori very encouraging result of the speciWc system presented above is that among the 105 pairs of languages considered here, none displays a negative coeYcient. As a tentative projection, this result already leads us to expect that the number of historically wrong (severely under- or overestimated) taxonomic relations provided by this parametric sample will be limited, probably in the order of few units.

10.9 Phylogenetic Grouping: Principles and Procedures 10.9.1 Probability

In Table 10.2 the chance probability of the various relationships between each pair of languages is computed. As can be seen, the table is divided into three subtables, of diVerent statistical signiWcance. The interpretation of these tables and the ratio for such partition will be the object of the following sections. 10.9.2 Statistical procedures

Some procedures for the partition and the interpretation of Table 10.2 are of a purely statistical nature, and, like the Anti-Babelic principle, can be devised independently of any empirical investigation of the results provided by the method on an initial sample. The introduction of a statistical parameter such as standard deviation seems to be relevant for our purposes, because its application allows us to identify by means of a statistic procedure (independent of any empirical considerations on speciWc language pairs) those pairs in which the intervention of chance for the distribution of similarities and diVerences can be considered null (irrelevant). After being applied to each single pair of our model, the calculations based on standard deviation allow us to identify a number of pairs which have a non-random percentage of at least 99 per cent: thus we propose to insert a threshold in Table 10.2, which separates those pairs that have essentially null possibilities of being due to chance (from <25.1> to <17.4>) from all the others. A further, simple, mathematical criterion that suggests establishing a threshold precisely at that point of Table 10.2 is that the probability of <17.4> is 1/350, still well below the a priori chance probability of Wnding one out of our 105 pairs with that coeYcient, while the next coeYcient <16.5>

Parametric Comparison and Language Taxonomy Table 10.2. Probability of chance results for the 105 pairs PAIR

PROBABILITY

COEFFICIENT

It-Ptg Ptg-Sp It-Sp It-Fr Sp-Fr D-E OE-D Fr-Ptg ClGr-NTGr NTGr-Grk Lat-ClGr OE-Got SCr-Blg OE-E

1/2,581,110

25.1

1/206,489

24.2

1/60,787 1/33,157 1/25,811 1/9,986 1/2,724 1/964 1/541

22.2 21.2 23.3 19.2 19.3 15.2 16.3

1/350

17.4

It-Grk Sp-D SC-Got

1/103 1/87 1/68

16.5 17.6 15.5

1/56

16.6

1/45

14.5

1/39

15.6

1/36

12.4

1/34

16.7

1/27

14.6

It-E Sp-E Fr-E Ptg-ClGr Ptg-NTGr ClGr-Grk NTGr-Got Got-D Sp-Grk Sp-OE Fr-Grk Blg-NTGr Ptg-Grk It-Lat Sp-Lat Ptg-Lat It-D It-Blg Fr-D Ptg-Blg Ptg-D NTGr-SCr SCr-OE

Continues

163

164

C. Guardiano and G. Longobardi Table 10.2. PAIR It-ClGr It-NTGr Sp-ClGr Sp-NTGr Ptg-E NTGr-OE NTGr-Lat It-OE Fr-OE Ptg-OE ClGr-Blg Sp-Blg Fr-Lat Lat-SCr It-Ar Grk-OE Grk-D E-Ar D-Ar Lat-Got It-Got Sp-Got Ptg-Got Blg-Got Grk-Got ClGr-Ar Grk-E Grk-Ar OE-Blg NTGr-D Fr-Blg Lat-Grk Lat-OE Lat-E Lat-Blg Lat-Ar ClGr-Got Sp-Ar Fr-Ar Ptg-Ar Grk-SC Grk-Blg OE-Ar

(Continued) PROBABILITY

COEFFICIENT

1/25 1/25

15.7 15.7

1/24

11.4

1/18

14.7

1/17

15.8

1/15 1/14

11.5 13.7

1/13

14.8

1/11

10.5

1/11

11.6

1/10

12.7

1/10

13.8

1/10

14.9

1/8

9.5

1/8

10.6

1/8

12.8

Parametric Comparison and Language Taxonomy Table 10.2.

165

(Continued)

PAIR Fr-ClGr Fr-NTGr ClGr-E ClGr-D Fr-Got Lat-D Got-E It-SC Sp-SC Ptg-SC Blg-Ar ClGr-OE NTGr-Ar D-SC D-Blg NTGr-E Fr-SC E-SC

PROBABILITY

COEFFICIENT

1/8

13.9

1/7

10.7

1/7

11.8

1/7 1/7 1/6

12.9 13.10 12.10

1/6

10.9

E-Blg ClGr-SC

1/6 1/5

11.10 9.8

Got-Ar SC-Ar

1/5

9.9

Note: Strong relations, in the sense deWned in section 10.9.3, are in bold.

(probability 1/103) already falls beyond the 1/105 border. The empirical examination of the data will directly support this conjecture.8 10.9.3 Empirical procedures

10.9.3.1 Strong relations The hypotheses about the threshold of reliability of the various coeYcients for the sake of phylogenetic taxonomy put forth in the previous section receive straightforward conWrmation when applied to the actual data. In order to see that, let us deWne the concept of strong relation: (8) A phylogenetic relation between a pair of languages in our sample is strong if and only if one of the languages derives from the other or both derive from a common ancestor within a time span of at most 2,000 years. 8

It is relevant to point out that these statistical computations do not provide us with a method to compare pairs of coeYcients that have diVerent sums (e.g. 20.1 vs. 16.2): they only contribute insights into the identiWcation of the probability of chance results for each single pair (in principle, one might compare only those pairs of coeYcients which have the same sum; that is, 20.1 vs. 19.2 and so on).

166

C. Guardiano and G. Longobardi

Given such a deWnition, it is easy to check that all the fourteen relations above the threshold identiWed by standard deviation (<17.4>) are strong in this sense, with the probable exception of that between Classical Greek and Latin, for which you can also see below. This observation provides further evidence, this time of a directly empirical sort, for the threshold identiWed above on statistical grounds. Furthermore, only six other clearly strong relations are not contained within the threshold (i.e. in the Wrst subtable). The missing relations are those between Latin and the Romance languages and, perhaps, those between Gothic and the two modern West Germanic varieties, though the latter must be located right at the edge of the two-millennia border (supposing common proto-Germanic to be still spoken around the last centuries before Christ). Especially the relations between Latin and French (<11.5>) and Gothic and English (<10.7>) seem to be severely underestimated by the system, as was noticed before, unless this fact alludes to some relevant peculiarity of the development of French and English within their respective families. It is, however, remarkable that the relation between Old English and Gothic, which instantiates a chronological depth of slightly more than one millennium, does fall within the first threshold (<17.4>). 10.9.3.2 DeWnition and distribution of mistakes The empirical relevance of the standard deviation threshold is also conWrmed by inspecting the position of factual mistakes. The concept of mistake will be a relative one, that is, it will apply not to pairs but rather to sets of at least three languages: (9) Factual mistake: two pairs of languages in the model form a mistake iV they both contain the same language X and two other languages Yand Z, respectively, and X turns out to be closer to Y than to Z (or vice versa), while the opposite is true according to independently well-established taxonomic results. Computing the relative distance among the languages of our sample reveals only three major classes of factual mistakes: (10)

a. Latin is not the language closest to the Romance ones b. Arabic (the only non-IE language) is not the language with the lowest algebraic coeYcient with respect to all the others in the lot c. Gothic is not the language closest to the other Germanic ones (all from the Western branch), except for OE

None of these sources of mistakes is instantiated by relations found in the Wrst subtable. The highest mistake is in fact represented by the closer relation

Parametric Comparison and Language Taxonomy

167

exhibited by Italian with Greek than with Latin, and it is found no higher than the second subtable, indeed it exactly marks the high border of the latter, converging again with the statistical evidence towards identifying the same threshold. On empirical grounds as well, then, one can safely assume that all relations within the threshold of the Wrst subtable must be phylogenetically signiWcant and that they do not yield any serious factual mistakes. 10.9.3.3 Chronological separation: the syntax of the ancients and of the modern Further inspection of Table 10.2 suggests some peculiar aggregation of the weaknesses of the system around the relations concerning ancient languages: Wrst, the three underestimated strong relations involve Latin and Gothic; second, if we abstract from pairs containing an ancient language, the highest factual mistake occurs at <13.7> (1/14), that is, with Arabic and Italian closer than Italian and Serbo-Croat (<11.8>). Accordingly, we can tentatively establish a second threshold at about 1/15 (so deWning the low edge of the second subtable), one within which no factual mistake arises if we abstract from relations involving ancient languages, and try to make initial taxonomic hypotheses keeping modern and ancient languages apart. This way the only factual mistakes we are left with are very low in the scale of improbability, thus of potential signiWcance, and all concern the position of one language, namely Arabic. Let us call this methodological resolution Chronological Separation Principle (CSP). The CSP is an operationally reasonable postulate, since in most real world cases where we ignore the actual phylogenetic relationships of a set of languages, at least we have fairly good information about the dates of attestation of the diVerent varieties, so that the CSP can be applied. Pushed to its extreme consequences, the CSP suggests that one hypothesizes the phylogenetic relations for contemporary languages (or at least for languages of approximately the same period) before trying to guess the diachronic relations among languages of sharply diVerent ages. Conceptually, the ground for the CSP is not yet clear: trivially, to a very little extent it may be related to the less precise setting of parameters imposed on the linguist by his working on closed corpora; but a good deal of the phenomenon might be due to our present incapability of understanding mechanisms of parameter resetting, in particular of parallel developments in closely or loosely related languages as consequences of their areal convergence on ‘fashionable’ settings during speciWc time periods. For instance, there may exist, so to speak, parametric patterns characterizing a ‘syntax of the ancients’ and one ‘of the modern’ in the history of Europe.

168

C. Guardiano and G. Longobardi

10.9.4 Tree hypotheses: modern languages

Following the CSP, we will Wrst hypothesize distinct (i.e. non-connected) trees for the modern and ancient languages (just keeping OE, that is, the only medieval language, ambiguously present in both sets). Sticking to the conclusion that the relations in the Wrst subtable of Table 10.2 must all be highly signiWcant, it becomes compulsory to acknowledge the existence of Wve distinct units among modern languages: one comprising Italian, French, Spanish, and Portuguese, one comprising English, German, and Old English, one with Bulgarian and Serbo-Croat, and then two language isolates, represented by Greek and Arabic. Let us suppose that all these units instantiate true families. In order to support the Wrst impressionistic guess about such grouping, it is possible to apply to these hypotheses two criteria which represent a kind of a posteriori proof of taxonomic correctness. Both criteria, which aim to measure the degree of internal cohesion of a proposed language family, share the basic idea of mass comparison; that is, they apply to the consideration of more than two languages at the same time. 10.9.5 Tree hypotheses: ancient languages

In agreement with the CSP, we may now consider the four ancient languages and the medieval one in abstraction from the modern ones: the relationships between Classical and New Testament Greek, as well as that between OE and Gothic and that between Latin and Classical Greek, turn out to fall quite saliently within the threshold of standard deviation. Thus, it is straightforward to assign Classical and New Testament Greek to a single close unit or lineage, and then to hypothesize a clear, though looser, relation between the other two pairs. Factually, this hypothesis is largely correct: perhaps it overestimates the relation between Latin and Classical Greek, which is not known to be genetically closer than that between, say, the Germanic and the Slavic family; however, it must be recalled that Latin and Classical Greek are the most anciently attested languages in the sample, so that the chronological depth of their separation is certainly very reduced in comparison to that of the modern IE subfamilies. It is suggestive, in this sense, that the relations between Romance languages and Modern Greek (<16.5> with Italian, <15.6> with the others) might still be slightly signiWcant, but fall short of the threshold, perhaps witnessing the progressive separation of the two lines. The second point about Latin and Classical Greek is that a possible slight overstatement of their relation can also be due to the centuries of contact between the two languages, beginning well before the literary age.

Parametric Comparison and Language Taxonomy

169

Anyway, all in all, the separate analysis of the Wve non-modern languages leads quite uncontroversially to the following hypotheses: there would be two salient families, one represented by OE and Gothic and another by Latin and Classical Greek, the latter language obviously quite closely related to New Testament Greek. The further binary relations are probably insuYcient to prove a particularly close connection between the two families, the highest one, that between Gothic and NT Greek, being at <14.5> (for this pair, some interference eVects can hardly be excluded, of course). 10.9.6 Tree hypotheses: diachrony and beyond

At this point, we are left with Wve ‘modern’ units and two ‘ancient’ ones: (W.) Germanic, (W.) Romance, (S.) Slavic, Greek, and Arabic, and then OE-Gothic and Latin-(Classical and NT) Greek, respectively. Now, we would like to build such units into a more comprehensive tree, Wnally giving up the CSP. In two cases, this is, again, straightforward: the relation of NT Greek to Modern Greek falls well within the threshold of Table 10.2 and the same is true of the relations between OE and Gothic on one side and OE and German and English on the other (recall that, owing to its intermediate chronology, OE was ambiguously compared from the very beginning against both ancient and modern languages). In this sense OE and NT Greek function as excellent bridges of other, apparently looser, relations within their families, which can be viewed once more as a typical application of the mass comparison criterion. In fact, while the relations between Classical and Modern Greek and that between German and Gothic (both <14.5>) are at least not discouraging for hypotheses of relatedness, that between English and Gothic (<10.7>) would be by itself completely unrecoverable. The situation might anyway suggest that Gothic must be regarded as a collateral cognate with respect to the taxonomic unit clearly formed by the other three Germanic languages, a factually correct result. Establishing descendence for the other ancient language of the sample—i.e. Latin—is much more diYcult, since it only appears once (with Classical Greek) in the high subtable of Table 10.2. The only way to proceed is then to look at the languages closest to Latin immediately after Classical Greek: Italian, Spanish, and Portuguese (three Romance languages); after the intervention of another ancient language, NT Greek, immediately comes French. Given that the unity of these four languages is beyond doubt, it might be signiWcant that all of them, despite their minor diVerences, display some agreement with Latin. On these grounds it is almost compelling to suppose that, if a descendence must be attributed to Latin within the range of the Wve

170

C. Guardiano and G. Longobardi

modern units arrived at above, this must be the Romance family. The relation (call it the Italic taxon) remains anyway hypothetical: it is natural to speculate whether it might become evident if we included some later and Vulgar variety of Latin to bridge the gap, along the model of what we have seen in the case of Classical, NT, and Modern Greek. The Wnal question arising now is that of the deeper relations among the four diachronically complex units to which we have so far reduced the original Wfteen varieties: the Greek-Italic unit, subdivided into the Greek family (three of our languages) and the hypothetical Latin-Romance unit (Wve languages), the Germanic family (four languages), the Slavic family (two languages), and Arabic. The taxonomic validity of none of the possible subsets formed out of this set of four members can be supposed on the sole observation of the high subtable of Table 10.2, because no inter-unit relation appears there. In particular, it is immediately clear that Arabic has a slightly diVerent status from the other languages, since it is the only one that does not occur at all in the high subtable of Table 10.2 and, consequently, could not be reduced into any family. In fact, it appears for the Wrst time rather low in Table 10.2, namely at <13.7>. It is reasonable, then, in order to make further taxonomic hypotheses, to try to examine pairs of languages at least down to this coeYcient, so as to be able to include all languages in the tree. This is a tentative methodological decision, of course; the other strategy could be that of concluding that the chance probability of <13.7> (1/14) is anyway too great to consider such evidence seriously, which would probably amount to already classing Arabic as extraneous to any unit formed by the extant languages. This conclusion would be factually correct, of course, in this case, but before accepting it, we would like to try to subject it to deeper scrutiny. At the same time, it is anyway necessary to ask questions about the mutual relationships of the other three units identiWed, a non-trivial problem, as noticed. For these reasons, it was decided indeed to speculate about evidence contained even below the high subtable of D, that is, between the Wrst threshold and the 13.7 threshold mentioned above.9 Clearly, since this

9 The computation of chance percentages (through the standard deviation) in the pairs of our sample shows that the probability of non-chance results for each of the pairs located between the Wrst threshold (immediately below 17.4) and the 13.7 threshold is comprised between 90 per cent and 60 per cent, and that the percentage for each of the pairs located in the third part of Table 10.2 (from 10.5 below) of being non-random is no more than 30 per cent, a result which can be considered statistically irrelevant.

Parametric Comparison and Language Taxonomy

171

evidence is probabilistically much less safe, here it will be more important to resort to mass criteria, i.e. to evaluate many relations at the same time. What we should compare is, thus, the closeness of the relations among the four taxa arrived at. Notice, for example, that the fact that Arabic does not appear above the standard deviation threshold, while the other languages do, per se means nothing about the possibility that all the others form a valid taxon together. What we are looking for is whether there is any of the possible proper subsets of the four-taxa set in which the relations are signiWcantly tighter than in all the other possible subsets. Arabic is not only the language with the lowest Wrst occurrence in Table 10.2, but also with the lowest occurrences altogether: actually the only two relations with zero algebraic coeYcient (<9.9>) concern Arabic. Also, and more relevantly, Arabic seems to be slightly aside from all the other languages, since each of the latter appears at least four times in the intermediate subtable of 10.2, where Arabic does not yet occur. Within this space, the relations among the established taxa are distributed as follows: the Greek-Italic unit bears fourteen binary relations to the Germanic family, seven to the Slavic family, and zero to Arabic, as noticed; four pairs relate Germanic to Slavic, none relates either of these two families to Arabic, of course. If we restrict our attention just to the ten modern languages, for which we have identiWed Wve units (Romance, Germanic, Slavic, Greek, Arabic), the space in question of Table 10.2 includes the following amount of inter-unit (i.e. external to their own unit) binary relations for each language: (11) Italian: 4 (Grk, Blg, E, D) Spanish: 4 (D, E, Grk, Blg) French: 4 (E, D, Grk, Blg) Portuguese: 4 (Blg, Grk, D, E) German: 4 (Sp, It, Fr, Ptg) English: 4 (It, Sp, Fr, Ptg) Greek: 4 (It, Sp, Fr, Ptg) Bulgarian: 3 (It, Ptg, Sp) Serbo-Croat: 0 Arabic: 0 Here Serbo-Croat might seem as isolated as Arabic, were it not for the accidental fact that we included a closely related, bridging, language into the sample, Bulgarian. But, if the inter-unit relations of these ten modern languages are also measured against the older languages, even Serbo-Croat appears in four pairs, while Arabic continues to appear in none.

C. Guardiano and G. Longobardi

172

10.10 Cognitive History 10.10.1 Some correct conclusions to be drawn from the model

The conclusions arrived at above by comparing abstract cognitive structures like parameters can be considered realistic to a large extent, at least in the sense that they closely resemble those which could be reached on purely lexical/phonological grounds (i.e. especially in the absence of extralinguistic historical evidence). Our method seems, as a matter of fact, to capture the following historically correct insights: (12) a. b. c. d. e.

The (Western) Romance unity The West Germanic unity The unity of Bulgarian and Serbo-Croat The slightly peculiar position of French within W. Romance The diachronic relation between Classical Greek and Modern Greek through NT Koine´ f. The non-obvious subdivision of West Germanic g. The Germanic unity

The attribution of slightly particular developments to French and English is based on their relations with the cognate languages: French is the Romance language whose aYliation with Latin is most unrecoverable by our method and, furthermore, is minimally more distant from the rest of Western Romance, which, signiWcantly, constitutes what Renzi (1985) on traditional comparative grounds termed Romania continua; English turns out to be quite distant from Gothic and, within West Germanic, though being very close to German, resembles OE less than German does, raising an unclear taxonomic situation, which forces us tentatively to represent West Germanic by means of a Xat ternary subtree. The apparently puzzling result that OE resembles German more than Modern English is, however, not too far from what we would probably derive through lexical comparison alone, that is, without the support of the geography of documentation and the knowledge of extralinguistic history (for example, of the West Germanic and Norman conquests of the British Isles). Three additional correct conclusions are likely to be derivable through the method of parametric comparison, though a little less neatly: (13) a. The diachronic relation between Latin and Romance b. The distance of Arabic from the other languages of the sample c. The progressive separation of some languages and subfamilies through time

Parametric Comparison and Language Taxonomy

173

The point in (13c) is well represented at least by the fact that Romance languages are more distant from Modern Greek than Latin was from Classical Greek;10 in other cases it may perhaps be obscured by chance or secondary, though still historically relevant, facts, like some modern convergence of Romance and Germanic in Western Europe, likely to be detectable also through lexical borrowings. The present version of the method of parametric comparison is not yet able to distinguish satisfactorily between certain eVects of kinship and convergence. Anyway, all in all, the method, as applied to the sample of languages and parameters in question here, seems deWnitely to provide question (4) above with the intermediate answer (5a), suggesting essentially the same phylogenetic tree as lexical comparison up to a depth of 2,000–500 years and apparently grasping some correct insights about chronologically much deeper relations (since it is likely to suggest the unity of all the languages of the sample except Arabic against the latter). On these grounds, there seems to be no reason to exclude the possibility that answer (6b)—or even (6c)—is the eventually correct one to question (4). 10.10.2 Insights expected from a qualitative analysis of parameters

Let us now brieXy review a few problems arising from this Wrst application of parametric comparison and calling for future amendments of the method; let us begin with the following couple of observations: (14) a. Certain parameters may understate identities and diVerences because they trigger a high number of zeros in the other parameters. b. Many apparently independent parallel developments may be due to similarity of input (primary data) persisting once a neutralizing setting of the protolanguage has been changed. The shifted value of parameter 3 of Table 10.1 from Latin to Romance, for example, implied the deneutralization of the value of all the other parameters depending on 3, augmenting the potential for comparison and, in principle, for identities among the Romance languages, but not between the latter and Latin. Something analogous is likely to occur in the parameters aVecting genitive case. This may be a reason for the diYculty encountered in promptly recovering the diachronic relation between Latin and Romance. On the other side, the identical settings of all the Romance languages for the parameters 10 Grouping the Classical languages together as a valid taxon excluding Slavic and Germanic is perhaps questionable but hardly an obvious mistake. Rather, it would be interesting to see whether the unity of the Classical languages would persist after subjecting to parametric comparison ancient Armenian, Indo-Iranian, and Celtic.

174

C. Guardiano and G. Longobardi

made suddenly relevant by the shift in 3 could depend on some common property of the triggers (primary data for parameter setting, in Clark and Roberts’s 1993 sense) already present, though irrelevant, in Latin. This may objectively understate the similarity between Latin and Romance. Of course, one way of reducing the negative eVects on taxonomic conclusions is quantitatively enlarging the set of parameters. The existence of such problems suggests, however, that one should also perform some qualitative evaluation of the diVerent parameters. The most natural assumption in this domain is that parameters have diVerent degrees of historical stability, classing them as: 9 (15) a. Deep (or Genetic) = parameters b. Areal ; c. Shallow Genetic parameters would be those most often stable through time and deeply rooted in a language family, areal parameters those whose values tend to be borrowed from surrounding languages, and shallow parameters the ones readiest to historical resetting (e.g. because mainly triggered by phonological or semantic interface evidence, in the Inertial framework). Various other amendments of the method proposed here are conceivable and being currently explored, promising to enhance our general understanding of diachronic syntax. However, one point needs to be stressed: given that the taxonomic results achieved thus far are largely correct, it would be probabilistically astonishing if this were due to chance and if welcome improvements of either parameter theory or the method of comparison disconWrmed such results. The evidence collected begins to be suYcient to question seriously the validity of one of the longest enduring assumptions of linguistic science: the bicentennial belief that grammatical typology may hardly serve the purposes and often cross-cuts the results of historical comparison. Such a belief fades away under a systematic and mathematically accurate investigation of the facts. This approach was crucially made possible by the application of historical considerations to abstract cognitive structures, that is to say, parameters and parameter values, perhaps more remote from common sense than surface patterns and items, but, as is often the case in other mature Welds, regarded by scientists as even more realistic entities of the world and more reliable objects of investigation.

11 Clitic Placement, VP-Ellipsis, and Scrambling in Romance ANA M ARIA MARTINS

11.1 Introduction My goal in this chapter will be to account for variation and change with respect to clitic placement, VP-ellipsis, and scrambling in non-V1 Wnite clauses in Romance. The analysis I will put forward derives synchronic and diachronic variation with respect to clitic placement, VP-ellipsis, and scrambling in Romance from the variable featural make-up of the functional heads S and AgrS, namely from the interplay between the ‘strength’ property of S and the EPP property of AgrS. I depart from previous diachronic analyses of clitic placement in Romance in essentially two ways: (i) I take Romance clitics to be heads from Old to Modern Romance (against Rivero 1986, 1991, 1997; Fontana 1993; 1997; Barbosa 1993; 1996; Halpern and Fontana 1994; Halpern 1995);1 (ii) I contend that V2, V3, etc. Wnite clauses with enclisis are not always ‘disguised’ V1 clauses (departing from Salvi 1990, 1991, 1993, 1997; and Beninca` 1995; among others); therefore I do not see variation between proclisis and enclisis in non-V1 Wnite clauses in Old Romance as an illusion created by the optional character of Left Dislocation.2 The research reported here was partially funded by the Collaborative Center on Multilingualism - 538 Sonderforschungsbereich Mehrsprachigkeit (University of Hamburg) and by FCT - Fundac¸a˜o para a Cieˆncia e a Tecnologia through the European Community programme POSI. I wish to acknowledge the Collaborative Center on Multilingualism for the very stimulating and friendly research environment that I was oVered during my stay in the Center in October 2001. FCT provided the Wnancial support that enabled me to attend DIGS VII, where this paper was presented. 1 For evidence in favour of the claim that Romance clitics are heads throughout the history of the Romance languages, see Martins (2003a) and Fischer (2002). 2 DiVerent authors addressing linguistic change in Romance have proposed that clitic pronouns evolved from second position clitics to verbal clitics, correspondingly changing their categorial status (from XP to X0 ) and their syntactic distribution. As clitics in Old Romance can surface in a third or more rightward positions, it has been claimed that leftward sentential adjuncts or left dislocated phrases are freely allowed in certain Old Romance languages; these constituents would be irrelevant for

176

A. M. Martins

The chapter is organized in Wve Sections. In Section 11.2 it is shown that in Romance there is a correlation which holds both synchronically and diachronically between the availability of VP-ellipsis and enclisis in Wnite clauses. It is proposed that VP-ellipsis and enclisis emerge in the languages where the polarity encoding functional head Sigma (S) has strong features, being not allowed in the languages where S is a weak functional head. In Section 11.3 the connection between clitic placement and Old Romance IP-scrambling is elucidated. Because clitics are minimal/maximal entities (Chomsky 1994, 1995), clitic movement may target either a head or a Spec position. If the later option is taken, clitics move into the same domain as non-clitic scrambled objects—that is Spec of AgrS. In some Old Romance languages AgrS allows multiple Spec positions. In these languages interpolation (i.e. nonadjacency between the clitic and the verb) is derived with the clitic placed in the outer Spec, AgrSP while scrambled objects or the subject occupy ‘embedded’ Spec positions in the AgrS domain. In Section 11.4 a concise description of the diVerent patterns of clitic placement in Old and Modern Romance is presented and a uniWed account of synchronic and diachronic variation is proposed. Section 11.5 summarizes and concludes the chapter.

11.2 The Correlation between Enclisis and VP-Ellipsis 11.2.1 The Modern Romance split: strong S vs. weak S

In Romance, enclisis in tensed clauses correlates with the possibility of VPellipsis and concomitantly with a pattern of minimal answer to a yes/no question consisting of a bare verb. The Romance languages split in two groups with respect to such correlation. The relevant contrast is exempliWed with data taken from Portuguese—cf. (1) below—and from Spanish—cf. (2) below. Galician groups with Portuguese; Catalan, French, Italian, and Romanian group with Spanish (see Martins 2003b for further exempliWcation). Portuguese: (1) a. Tu deste-lhe o livro? you gave him/her the book ‘Did you give him the book?’

(Enclisis OK)

the computation of the ‘second position’, being treated for this matter as sentence external. This type of analysis (that sees V2 sentences as ‘disguised’ V1 sentences) makes use across the board of the so-called Tobler-MussaWa law which bans clitics from the sentence initial position. Under this kind of approach it is predicted that in Clitic Left Dislocation constructions, if only the left dislocated constituent precedes the verb, clitics would obligatorily be enclitic on the verb in Old Romance (see Salvi 1991; Beninca` 1995; Barbosa 2000), the verb counting as the Wrst sentence internal constituent. This prediction however is contradicted by the empirical data, as shown in Martins (2003a).

Clitic Placement, VP-Ellipsis, and Scrambling b. *Tu lhe deste o livro? you him/her gave the book ‘Did you give him the book?’ c. Dei. gave ‘Yes, I did’ d. Sim, dei. yes gave ‘Yes, I did’ e. Sim, dei-lho. yes, gave-him-it ‘Yes, I did’

177

(*Proclisis)

(VP-ellipsis OK)

(id.)

Spanish: (2)

a. *Tu´ distele el libro? you gave-him/her the book ‘Did you give him the book?’ b. Tu´ le diste el libro? you him/her gave the book ‘Did you give him the book?’ c. *Di. gave ‘Yes, I did’ d. *Sı´, di. yes gave ‘Yes, I did’ e. Sı´, se lo di. yes him it gave ‘Yes, I did’

(*Enclisis)

(Proclisis OK)

(*VP-ellipsis)

(id.)

In order to account for the correlation illustrated by the data displayed above, I propose that enclisis and VP-ellipsis emerge in languages where the functional category S—which encodes polarity values, i.e. aYrmation, negation, modality (Martins 2000b)—bears strong features (Portuguese and Galician), whereas such phenomena are absent from languages where S is weak (Spanish, Catalan, French, Italian, and Romanian)—cf. Martins (1994b).3 Departing from Chomsky (2000), I take the ‘strength’ property of a functional category to be relevant in two ways: (i) strong functional 3 The S head (proposed by Laka 1990) parallels the Pol head of Zanuttini (1994, 1997). DiVerently from Laka (1990), I do not take aYrmative S to (necessarily) express emphatic aYrmation.

178

A. M. Martins

categories can license null constituents; (ii) strong functional heads are necessarily part of spelled-out morphological words in the Morphology component of grammar (cf. Chomsky 1994, 1995, 2000; Halle and Marantz 1993; Harley and Noyer 1999; Embick and Noyer 2001).4 As for strong S, this requirement is satisWed by merging S with a head related to S by bearing polarity features. I take the heads which encode polarity values and therefore can agree and merge with S to be C, Neg, and V. The merging operation may take place in the syntactic component or post-syntactically, in the Morphology component, as will be clariWed in Section 11.4 below. I further assume that main clauses are usually IPs, not CPs, following ideas implemented by Bosˇkovic´ (1996, 1997) and Thra´insson (1996), among others, on the nonuniversality of clause structure. In declarative aYrmative main clauses, C and Neg being not projected, strong S merges with V.5 Enclisis is the outcome of this merger between V and S, as will be shown in Section 11.4 below. VP-ellipsis, on the other hand, is licensed by the Agree relation (cf. Chomsky 2000) between strong S and V, independently of the fact that S merges with V, C, or Neg. Laka (1990) proposes that in Romance the lexical items that show up in minimal unmarked answers to yes/no questions incorporate in S. Under the analysis that I am proposing here, this is so independently of the strong or weak character of S. However, verbal answers to yes/no questions can only be a property of languages where the strong nature of S induces merging of this head with the V head; otherwise only an aYrmative/negative word will be able to give lexical content to the S head. Thus in Spanish, in contrast to Portuguese, only Sı´ ‘yes’ (not *Di ‘gave’) is a good aYrmative answer to the question Le diste el libro? (‘Did you give him the book?’). 11.2.2 The strong nature of Sigma in Old Romance

It is a well-known fact that verbal answers to yes/no questions are pervasively attested in Latin—see example (3) below. So Latin licensed VP-ellipsis (as well 4

Costa and Martins (2003) propose the following generalization: Strong F licensing A strong functional head is licensed iV it is lexicalized, (i) under syntactic merger (ii) under head or XP movement (iii) under morphological merger.

5 In yes/no questions S is similarly licensed by V. Even if yes/no questions are analysed as CPs containing a null operator, merging between S and C would not satisfy the licensing condition on strong functional categories (see footnote 4 above), because in yes/no questions neither C nor Spec of CP contain lexical material.

Clitic Placement, VP-Ellipsis, and Scrambling

179

as other kinds of null constituents).6 It follows from the hypothesis formulated in Section 11.2.1 above that S was a strong functional head in Latin. (3) Clodius insidias fecit Miloni? — Fecit. Clodius-nom plots made Miloni-dat? — made ‘Did Clodius plot against Milo? — Yes, he did.’ (Cicero; Pinkster 1990: 191) Hence Portuguese and Galician display a conservative feature which is absent from the other contemporary Romance languages. We may ask however when the split between the two identiWed groups of Romance languages arose. The empirical evidence oVered by Old Spanish, Old French, Old Catalan, and Old Occitan texts suggests that S was a strong functional category across the Old Romance languages. Sentences (4)–(5) from Old Spanish, (6)–(7) from Old French, (8)–(9) from Old Catalan, and (10)–(11) from Old Occitan are relevant examples. Further illustration of VP-ellipsis in Old Romance can be found in Keniston (1937), Foulet (1928), Moignet (1973), Blasco Ferrer (1984), and Jensen (1994). Spanish: (4)

¿I traedes uostros escriptos? — Rei, si traemos and bring your books? — king aff bring ‘And, do you bring your books? — Yes, King, we do.’ (Auto de los Reyes Magos; GiVord and Hodcroft 1966: 42)

(5) – Pues ¿quien esta arriba? ‘Who is upstairs?’ – ¿Quie´reslo saber? — Quiero. – want-it to know? — want ‘– Do you want to know it? — Yes, I do.’ (La Celestina. Cejador y Frauca 1913: 62) French: (6) — Dame, je crois bien qu il est vostre Wlz, — lady, I believe well that he is your son, ‘— My dear lady, I am sure he is your son, me´s il n’ est pas Wlz le roi. but he neg is not son the king but he is not the king’s son’ 6 See the arguments given in Lobeck (1999) and Martins (1994b) against an analysis of VP-ellipsis as a phonological process of deletion.

A. M. Martins

180

— Si est, dit la reyne aff is said the queen ‘— Yes, he is, said the queen.’ (Les Sept Sages de Rome; The Hamburg corpus of Old French—Collaborative Center on Multilingualism, University of Hamburg) (7) Phelippe dist . . . : ‘Sire, parle´ a moy a une part, vous et vos quatre compaignons’. Phelippe said: ‘Sir, speak to me to the side (aside), you and your four companions’ Et il respondirent et crierent tous cinc: and they answered and shouted all Wve: ‘Si m’aı¨t Des, ne ferons’ ‘if me helps God, neg will-do’ ‘We will not do it, so God help us.’ (Philippe de Novare; Foulet 1928: 237) Catalan: havie-hi altre cavaller, menys de vo´s’? ‘Sı´ havia’. dix Curial and was-there another knight, other than you? aff was. said Curial ‘And was there any knight besides you? Yes, there was – said Curial.’ (Curial e Gu¨elfa. Blasco Ferrer 1984: 181)

(8) ‘E

(9) ‘Senyor cavaller, prec-vos que em digats si partits Mister knight, (I-)beg-you that me-dat (you-)tell if (you-)left d’aqueix monastir qui esta´ aquı´ prop’. ‘Sı´ fac’. from that monastery that is here nearby. aff did ‘Sir, would you please let me know whether you are coming from that nearby monastery? Yes, I am.’ (Curial e Gu¨elfa; Blasco Ferrer 1984: 181) Occitan: (10)

et as per so to cor dolen? Si ai and (you-)have for that your heart painful? aff have ‘And is this why you feel sick at heart? Yes, it is.’ (G. de Bornelh; Jensen 1994: 282. Quoted in Fischer 2002: 171)

(11)

non saps? — Si fas not know aff do ‘Don’t you know it? — Yes, I do.’ (Flamenca; Jensen 1994: 282)

Clitic Placement, VP-Ellipsis, and Scrambling

181

While in Old Spanish diVerent verbs can be found in sentences displaying VP-ellipsis, in Old French, Old Catalan, and Old Occitan the attested manifestations of VP-ellipsis include mostly the verbs eˆtre/e´sser ‘to be’, avoir/haver ‘to have’, and faire/fer ‘to do’. We should not conclude from this scenario however that only auxiliary verbs would license VP-ellipsis in these languages. In fact the umbrella verb faire/fer does not appear to be an auxiliary since faire/fer is also used as an umbrella substitute in contexts where an auxiliary would be excluded; this is shown by example (12) below, from Old French. (12)

Je vous aim plus que vos ne facie´s mi I you-acc love more than you-nom neg do me ‘I am fonder of you than you are fond of me’ (Aucassin et Nicolet. Foulet 1928: 236)

On the other hand, while verbal answers with diVerent main verbs are well attested in Old Spanish, a use of hacer ‘to do’ similar to the umbrella use of faire/fer is found in Spanish, as noted by Keniston (1937: 593)—‘The verb of the question may be replaced by a form of the vicarious verb hacer, usually without an object pronoun’. I believe that it is the paucity of the available data with respect to VP-ellipsis in Old Romance that creates the illusion of a contrast between Spanish, on the one side, and French, Catalan, and Occitan on the other. In fact, although infrequent, verbal replies with main verbs (beyond ˆetre, avoir, faire)7 are found in Old French, as pointed out by Moignet (1973), and exempliWed by (13) below. (13) —Por quoi le demandez vos? fet li roi. for what-reason it ask you-nom said the king ‘Why are you asking that? – said the king’ —Por ce, fet messire Gauvains, que je ne cuit pas for the-reason, said sir Gauvains that I neg think not que vous le sachiez. that you it would-know ‘Because I believe that you do not know about it’ —Si sai bien, fet li roi, mes vos ne le savez pas —aff know well said the king but you neg it know not ‘—Yes, I certainly do, said the king, but you do not’ (Mort Artu; Moignet 1973: 288) 7

Note that eˆtre ‘to be’ and avoir ‘to have’ are not always auxiliaries. See the Old French example below:

(i) Et por ce as tu cuidie´ que je fusse semblables a toi . . . mes non sui and for that have you believed that I would-be similar to you . . . but not am ‘And so you believed me to be like you . . . but I am not’ (La Queste del Saint Graal; Foulet 1928: 235)

A. M. Martins

182

That the availability of VP-ellipsis is not lexically constrained in certain Old Romance languages appears to be also conWrmed by example (11) above, from Old Occitan, here repeated as (14): (14)

...

non saps? — Si fas not know? aff do ‘Don’t you know it? — Yes, I do.’ (Flamenca; Jensen 1994: 282)

In (14) VP-ellipsis shows up in the answer with faire, but it seems to be also displayed in the question with the main verb saber ‘to know’ (although we lack the ellipsis antecedent). From the data presented above, I draw the conclusion that Old Romance was uniform with respect to the nature and the licensing of S (although I am aware of the fact that the empirical evidence supporting this conclusion needs to be strengthened). We thus expect that the correlation between VP-ellipsis and enclisis identiWed in Section 11.2.1 above on the basis of Modern Romance is conWrmed by the Old Romance data. As will be shown in Section 11.4. below, this is in fact the case.

11.3 Clitic Placement and Scrambling Old Portuguese is an SVO language. Nevertheless the OV order is derived when the object is left dislocated, focused (identiWcational focus), or scrambled into the IP domain. The third type of movement is not permitted in Modern Portuguese and is optional in Old Portuguese as shown by the pair of sentences in (15) below. (15) a. sse pela ue˜tujra uos algue˜ a dita vy˜a enbargar if by chance you-dat someone the mentioned vineyard blocks ‘and if by chance someone blocks the vineyard from you’ (thirteenth-century legal document; Martins 2001: 377) b. sse pela ue˜tujra uos algue˜ enbargar a dita vy˜a if by chance you-dat someone blocks the mentioned vineyard ‘and if by chance someone blocks the vineyard from you’ (thirteenth-century legal document; Martins 2001: 376)

The OV order derived from IP-scrambling typically emerges in subordinate clauses, presumably instantiating an extraction out of focus strategy (cf. Costa 1998; Martins 2002). Object scrambling is found both in sentences with overt and null subjects—see sentences (15) and (16), respectively—and is not restricted to one constituent per clause. So sentences (17) and (18) below are

Clitic Placement, VP-Ellipsis, and Scrambling

183

examples of multiple object scrambling. Scrambling aVects DPs, PPs, APs, AdvPs, and reduced clauses (both inWnitival or participial). (16) como lhe entregou as terras que lhe de dar avia how him-dat returned the lands that him-dat of give-infin had ‘how he returned him the lands that he ought to give him’ (Ferna˜o Lopes — Wfteenth-century prose. Costa 2001: 184) (17)

de que˜ lhe ssobre elle embargo poser from whoever him-dat over it obstruction puts ‘[protecting him] from whoever tries to block it [the land] from him’ (sixteenth-century legal document; Martins 2001: 300)

(18)

herdade que de dar teudo ll’ era property that of give-infin obliged him-dat was ‘property that he was bound to give him’ (Cantigas de Santa Maria. Costa 2001: 181)

Martins (2002) proposes that object scrambling in Old Portuguese is movement to Spec of AgrS. In Old Portuguese AgrS has an uninterpretable selectional feature (i.e. an EPP feature) with an ‘Attract-all-F’ property, in the sense of Bosˇkovic´ (1999). Owing to its EPP property, AgrS allows multiple Specs—cf. Chomsky (1994, 1995), Grewendorf and Sabel (1999). Because object IP-scrambling is movement to Spec,AgrSP, the order OV is not derived when the verb moves to S; this usually happens in main clauses (in contrast to subordinate clauses). The analysis put forward to account for the Old Portuguese relevant facts extends to Old Spanish where similar data are found. The connection between object scrambling and clitic placement becomes clear when we couple the proposed analysis of IP-scrambling with a few assumptions with respect to clitics: (i) Under a bare phrase structure perspective, clitics qualify both as minimal and as maximal syntactic categories (Chomsky 1994, 1995). Thus clitics may move into positions unWt for regular heads, namely Spec positions. (ii) Clitics enter the domain of AgrS (see Cardinaletti and Roberts 1991 and Martins 2000a) whenever AgrS is projected in the Syntax as an independent inXectional head. Otherwise clitics left adjoin to the higher inXectional head where the verb moves (Kayne 1991). Putting together assumptions (i) and (ii) we derive the fact that clitics and full scrambled objects may target the same structural position, that is, Spec,AgrSP. (iii) A clitic occurs at the structural edge of the category (X0 or XP) that contains it, that is, the clitic is dominated by the highest segment of that

184

A. M. Martins

category. This is a reformulated version of the Edge Principle proposed by Raposo (2000).8 Given the Edge Principle, whenever verb movement and clitic movement target the same functional head, proclisis is derived. This is the case when both the verb and the clitic left adjoin to AgrS. Note on the other hand that when the verb moves to S through AgrS, if the clitic undergoes head movement it must enter the domain of CH(ain)AgrS , after the head [AgrS V ,T,AgrS] moves to S. Otherwise the clitic would not end up at the outer layer of the head S in accordance with the Edge Principle. See Raposo (2000: 291) for details. Still according to the Edge Principle, if the clitic moves (as a maximal category) to Spec,AgrSP in a conWguration of multiple Specs, it will attach to the outer Spec. So the clitic will precede scrambled objects and an overt subject, ending up non-adjacent to the verb if V is incorporated in AgrS. If V is incorporated in S, enclisis with adjacency will be derived as S immediately dominates AgrSP. We have now set all the necessary ingredients to understand the contours of synchronic and diachronic variation in Old Romance clitic placement in Wnite clauses. Joining together the variability of the targets for clitic movement (either a head or a Spec) and for verb movement (either S or AgrS), we will be able straightforwardly to derive the diVerent patterns of clitic placement that will be identiWed in the next section.

11.4 Syntactic Variation and Change: From Old to Modern Romance 11.4.1 Portuguese and Spanish

Old Portuguese (OP) and Old Spanish (OS) display variation between enclisis and proclisis in the kind of clauses where enclisis is obligatory in Modern Portuguese, that is, in aYrmative main clauses (without ‘proclisis triggers’),9 as exempliWed by (19) below.

8 Cf. Raposo (2000: 291): ‘clitics . . . always attach to the outer layer of any given functional category that contains them . . . clitics cannot appear embedded in a functional category that contains them. This is explicitly formulated in the following principle, which I take to be a primitive of the morphological component of UG: The Edge Principle A clitic occurs at the structural edge of a head H that contains it.’ 9 Proclisis is derived in wh-interrogatives and in main clauses introduced by focused constituents, certain quantiWers, or certain adverbs. I take all these constituents to belong to the CP domain (within a split CP perspective).

Clitic Placement, VP-Ellipsis, and Scrambling

185

(19) a. e agora llaman le Barcilona and now they-call it Barcilona ‘Nowadays, it is called Barcilona’ (Spanish. Primera Cronica General de Espan˜a; Mene´ndez Pidal 1978: 10b) b. e oy.en.dia le llaman Tarrac¸ona and nowadays it they-call Tarrac¸ona ‘Nowadays it is called Tarrac¸ona (Spanish. Primera Cronica General de Espan˜a; Mene´ndez Pidal 1978: 10b) In subordinate tensed clauses clitics are generally preverbal,10 but may occur either adjacent or non-adjacent to the verb—the phenomenon of non-adjacency is known as interpolation in the Romance philological literature. Observe the pair of sentences in (20) below: sentence (20a) displays adjacency between clitic and verb; sentence (20b), on the other hand, exempliWes interpolation. (20) a. como nesta carta se co˜the˜ how in-this letter se-[passive clitic] contains ‘how it is stated in this letter’ (sixteenth-century Portuguese legal document; Martins 2001: 554) b. como se nesta carta contem how se-[passive clitic] in-this letter contains ‘how it is stated in this letter’ (sixteenth-century Portuguese legal document; Martins 2001: 318) As sentence (20b) also shows, and is further illustrated by sentence (21) below, OP and OS allow IP-scrambling of non-clitic objects, therefore deriving the order OV. (21)

Et de mas mando al omne que vos esta mj carta mostrara and of more I-order to-the man who you this my letter would-show que enplase that he- cites a todos los que lo contra ella Wzjeren to all those that it against her [the letter]would-do ‘Moreover, I order the man showing you this letter of mine to call before a court of law all those acting against it’ (Wfteenth-century Spanish legal document; Mene´ndez Pidal 1919: 309 (text 234). Quoted in Rivero 1991: 244)

10 In all Romance languages some infrequent cases of enclisis in subordinate clauses can be found. This shows that embedded C (which generally encodes dependency relations between main and embedded clauses) may marginally not be associated with polarity features, being then unable to license S.

A. M. Martins

186

Portuguese and Spanish change in the same direction with respect to the fact that both interpolation and (unrestricted) IP-scrambling cease to be an option after the sixteenth century. As for clitic placement in Wnite main clauses, however, while in Modern Portuguese enclisis generalizes, in Modern Spanish proclisis becomes the only allowed pattern. Let us Wrst see how the OP and OS facts can be accounted for. We will then take into consideration the ensuing change. In aYrmative main clauses strong S attracts V. Proclisis (which is optional in this context) results from the formation of a complex head derived from movement of the clitic to S after AgrS, containing V, has moved to S: [S clitic[S [AgrS V, T,AgrS] S]]. Given the Edge Principle, the clitic must ‘meet AgrS’ in S. If the clitic would left adjoin to the AgrS head, AgrS would be unable to move into S carrying the clitic along without violating the Edge Principle. As for enclisis, it is derived by movement of the verb into S while the clitic targets (the most external) Spec of AgrS. In subordinate clauses (as well as in the main clauses that include C)11 proclisis with adjacency arises when the verb and the clitic are incorporated in AgrS—recall that strong S merges with V when neither C nor Neg are projected; whenever C is available, S is licensed by merging with C. Interpolation, on the other hand, involves movement of the clitic to (the most external) Spec,AgrSP. Interpolation emerges when other overtly Wlled Spec(s) of AgrS is/are projected and the verb does not move beyond AgrS. As for the kind of unrestricted IP-scrambling displayed by OP and OS, it directly correlates with interpolation and enclisis because it also depends on AgrS being associated with an Attract-all-F EPP feature, therefore allowing multiple SpeciWers. In both Modern Portuguese (MP) and Modern Spanish (MS), AgrS loses the ability to select multiple Specs (cf. Martins 2002). This change implies the concomitant losses of interpolation and (unrestricted) IP-scrambling. We will have to consider now how Portuguese and Spanish came to diverge from each other with respect to clitic placement in main clauses: the former maintains and generalizes enclisis; the latter loses it. Let us start with the easier case. In MS enclisis is never derived in Wnite clauses because in this language S lost the strong property which induced merger between S and V in aYrmative main clauses. In MS thus both the verb and the clitic left adjoin to AgrS, with the clitic standing in the edge of the AgrS head.12 11 12

See footnote 9 above. We are assuming with Kayne (1991, 1994) that adjunction is always left adjunction.

Clitic Placement, VP-Ellipsis, and Scrambling

187

As for the generalized enclitic pattern of MP, we could admit that it would reXect the emergence of excorporation as a grammatical option. In this scenario, enclisis would be derived by movement of the head [AgrS V, T,AgrS] into S, leaving the clitic incorporated in AgrS. Once excorporation came into play, the clitic would not be allowed to move into S—cf. Bosˇkovic´ (1997). However, I will take here Kayne’s (1991) view that excorporation is not a grammatical option. As an alternative approach, I make the hypothesis that in MP AgrS ceases to select an EPP feature; so AgrS does not project a Spec.13 This change makes possible that S merges with V postsyntactically, that is, in the Morphology component. Once this option becomes available, it appears to exclude the option for syntactic merger. Morphological merger takes place under adjacency (cf. Bobaljik 1995; Embick and Noyer 2001). Since at this stage there is no Spec intervening between S and ‘V-in-AgrS’, the two heads can in principle undergo morphological merger. If a clitic has left adjoined to AgrS, however, the clitic given its minimal/ maximal nature breaks the required adjacency between S and ‘V-in-AgrS’ inhibiting morphological merger. This is the reason why the clitic has to be ‘removed’ in order to permit merger of S with ‘V-in-AgrS’ under adjacency. I propose that enclisis is the outcome of ‘Local Dislocation merger with inversion’ (cf. Embick and Noyer 2001) between the clitic and AgrS. Then S undergoes Local Dislocation merger with ‘V-in-AgrS’.14 Local Dislocation merger is an operation that takes place after Linearization applies, converting the hierarchical structure received from Syntax into a linear structure. Therefore at this point (in the absence of syntactic hierarchical structure) the Edge Principle does not apply, being not violated by the inversion operation between clitic and Agr. 11.4.2 Catalan and Occitan

In Old Catalan (OC) and Old Occitan (OO), variation between enclisis and proclisis is found in aYrmative main clauses, as exempliWed by sentences (22) and (23) below. This feature is shared with OP and OS:

13

Or AgrS is simply not projected and T does not have an (obligatory) EPP feature. Cf. Costa (2003) and Costa and Martins (2003). 14 To be precise, the minimal/maximal nature of the clitic is relevant with respect to Lowering. Lowering is a pre-Linearization morphological operation that cannot take place across XPs. Local dislocation is a post-linearization morphological operation which requires strict adjacency—cf. Embick and Noyer (2001).

188

A. M. Martins

(22) e el clerge compta´ li e li dix tot lo feyt, axi and the clergyman told him and him said all the aVair so con yo vos he escrit. as I you-dat have written ‘And the clergyman told him everything about the aVair, so as I told you in my letter’ (Early fourteenth-century Catalan letter; Russell-Gebbett 1965: 143) (23)

a. Ilh recebia-os alegramentz she received-them happily (Occitan. Ste Douceline, fourteenth century; Vance 2001) b. Filha, ieu t’ aduc e t’apo`rti gazinh daughter I to-you-lead and to-you-bring merit (Occitan. Ste Douceline, fourteenth century; Vance 2001)

However OC and OO diVer from OP and OS in not allowing interpolation in subordinate clauses. In OC and OO there is always adjacency between clitic and verb. These facts can be accounted for if we admit that in OC and OO the EPP feature of AgrS has an Attract-1-F (not an Attract-all-F) nature. AgrS can thus project only one Spec position. This position (if not taken by the subject) may be Wlled with a clitic. In this case, movement of the verb to S (in aYrmative main clauses) will derive enclisis. As for interpolation, it cannot be derived because in interpolation structures both the clitic and an interpolated constituent target Spec positions selected by the AgrS head. Interpolation is only possible if AgrS can project multiple Specs. The unavailability of multiple Specs of AgrS also accounts for the fact that only one object per clause can undergo IP-scrambling in OC and OO, as illustrated by sentences (24) and (25) below.15 15 In Catalan scrambling may occur in negative clauses—see example (i) below—against what is predicted by Fischer’s analysis of clitic placement and ‘stylistic inversion’ (Fischer 2002). Note that I interpret as IP-scrambling what Fischer takes to be stylistic inversion. I believe that a comparative approach to word order in the Old Romance languages favours the IP-scrambling analysis. It may turn out however that IP-scrambling and stylistic inversion reduce to the same phenomenon, a matter which I will not go into here.

(i) que nos dar non.o deviem that we give-INF not-it should ‘that we should not give it’ (thirteenth century Catalan; Russell-Gebbett 1965: 108) In sentence (i) above S is licensed by C not by Neg. In Old Romance the negative marker may either directly merge with S or be adjoined to the higher inXectional head where the verb incorporates. It is the later option that is displayed by sentence (i) above (with V and Neg adjoined to AgrS). The order ‘negative marker—clitic’ (instead of ‘clitic—negative marker’, with the clitic at the edge of the relevant inXectional category) is presumably derived in the Morphology component. In example (ii) below the preverbal object of the subordinate negative clause appears to be focused, not scrambled.

Clitic Placement, VP-Ellipsis, and Scrambling (24)

189

yo cresegı´ lo feyt que veritat era I believed the aVair that truth was ‘I believed that the aVair (I was told about) was true’ (fourteenth-century Catalan; Russell-Gebbett 1965: 143–4)

(25) et Petrus de Lobeira dix que aquels C solidos and Petrus de Lobeira said that those hundred solidos feria pagar (he-)would-make pay ‘And Petrus de Lobeira said that he would make [A. de Prades] pay the hundred solidos (a monetary unit) that he owed [to Petrus Estrader].’ (twelfth or thirteenth century Catalan; Russell-Gebbett 1965: 91) The fact that besides the scrambled object an adverb—as in sentence (26) below—or the subject—as in sentence (27) below—can appear in preverbal position indicates that these constituents may occupy positions to the left of AgrSP (being adjunct or dislocated constituents). (26) de.so.qua axi gran honor los avia feta (167) because so great honour them he-had done ‘because in that way he had deeply honoured them’ (Catalan. Fourteenth century; Russell-Gebbett 1965: 167) (27)

Respos lo paire que el aquesta causa cresia fermament answered the father that he this thing believed Wrmly ‘His father answered that he Wrmly believed it’ (Occitan. Miracles; Vance 2001)

Modern Catalan (MC) and Modern Occitan (MO) display a pattern of clitic placement in Wnite clauses similar to the pattern of clitic placement found in Modern Spanish. The three Romance languages have in fact undergone the same type of change, that is, the weakening of the S functional head. Because the verb does not move to S, enclisis cannot be derived. As for the loss of IP-scrambling, I will leave undecided whether AgrS does not project a Spec in MC and MO or whether Spec,AgrSP can only be targeted by the subject. In any case there would have been a change in the featural make-up of AgrS with respect to its EPP property (cf. Martins 2002).

(ii) aqestes vergues no poden viure en nuyl loc si de l’ayga de la dita font non son reguades Those wickers not can live in any place if of the-water of the mentioned spring not are watered ‘Those wickers will die unless they are watered by the water of that particular spring’ (fourteenth-century Catalan; Russell-Gebbett 1965: 149)

190

A. M. Martins

11.4.3 French

Old French (OF) diVers from OP, OS, OC, and OO in not displaying variation between enclisis and proclisis in Wnite main clauses. The proclitic pattern is always found in OF in non-V1 clauses,16 as illustrated by sentences (28) and (29). (28) E cest a´faire li mustrad e´ sun conseil li dunad que . . . and this matter him showed and his advice him gave that . . . (Quatre Livre des Rois. Kok 1985: 75) (29) L’endemain se partirent de Wincester The-next day themselves left from Wincester (Mort Artu. Kok 1985: 75) As for interpolation, OF is similar to OC and OO (although a few cases of residual interpolation can be found in the earlier texts). OF also displays restricted scrambling, as shown by sentences (30) and (31) below. (30) devant que cil sera venuz qui ceste aventure doit eschoir before that he will-be come who this adventure must Wnish ‘before the coming of the one who must bring this adventure to a close’ (La Queste del Saint Graal; Vance 1997: 136) (31) que je devant mon frere viegne that I before my brother might-come ‘that I might come before my brother’ (La Queste del Saint Graal; Vance 1997: 136) In order to account for these facts, I propose that in OF AgrS is not associated with an EPP feature. Because AgrS does not project a Spec position neither enclisis (derived from movement of the clitic to Spec, AgrSP) nor interpolation is available. Although in OF, like in MP, AgrS does not project a Spec position, morphological merger between the verb and the clitic does not arise (cf. section 11.4.1. above); therefore enclisis is not derived in this way either. This is so because in OF (a V2 language) the verb targets a position higher than S, in the CP domain—cf. Poletto (2002), among others. Given the cyclic nature of head movement, the verb moves through S in its way to C. Thus strong S is always licensed in the syntax and there is no place for post syntactic licensing. 16

As for verb initial sentences (V1), enclisis is derived post-syntactically in order to avoid the phonological ban against Wrst position clitics observed by Tobler (1875, 1889) and MussaWa (1886). In this particular case enclisis is not an eVect of the need to license strong S.

Clitic Placement, VP-Ellipsis, and Scrambling

191

Restricted IP-scrambling (of one only constituent) is presumably derived in OF subordinate clauses with O in Spec,TP and V in T. Maybe OF does not have a split IP.17 So in OF the verb would target C in main clauses and T in subordinate clauses. The account given in the preceding sections with respect to clitic placement and lack of OV with full objects in Modern Spanish, Modern Catalan, and Modern Occitan naturally extends to Modern French.

11.5 Summary and Conclusion The main facts considered in this chapter are summarized in Table 11.1. Table 11.2 shows how such facts are explained under the views developed in the chapter. Note that although Old Italian is left out of the picture (Modern Standard Italian patterning with MS, MC, MO, and MF), it is likely that some varieties of ‘Old Italian’ (a cover term) will pattern with OC and OO, while other varieties of Old Italian will pattern with OF. Within the approach undertaken in the present investigation the availability of VP-ellipsis depends on the strong property of S. The availability of object IP-scrambling, on the other hand, depends on a certain type of EPP property of AgrS. As for clitic placement, it was shown that both the enclitic pattern and the proclitic pattern (in Wnite clauses) may be derived in two distinct ways. Enclisis is derived in the syntax or post-syntactically. In the former case enclisis is the outcome of verb movement to S plus clitic movement to Spec, AgrSP. This type of enclisis is unavailable in Modern Romance either because S is weak and does not induce V movement (MS, MC, MO, MF) or because there is no available Spec of AgrS (MP). Post-syntactic enclisis is a side eVect of morphological merger (i.e. local dislocation) between S and V. This morphological operation can only take place after clitic and verb change positions (through local dislocation merger with inversion), because the clitic breaks the required adjacency between S and V. Proclisis is obtained when the clitic and the verb left adjoin to the same inXectional head (C, S, AgrS, or T, depending on how high the verb moves and on whether there is a split or an unsplit IP). But proclisis may also be derived by clitic movement to Spec, AgrSP while the verb incorporates in AgrS. This option (which may feed interpolation) is restricted to grammars where the structure of the clause may include a Spec position available to 17 This means that in OF the Agr features would be introduced post-syntactically—i.e. in the Morphology component (cf. Embick and Halle 2003).

192

A. M. Martins

Table 11.1

Clitic placement in (non-V1) main clauses Clitic placement in subordinate clauses Object scrambling VP-ellipsis

Table 11.2

Comparative facts within Romance languages OP, OS

OC, OO

OF

MP

MS, MC, MO, MF

enclisis or proclisis

enclisis or proclisis

proclisis

enclisis

proclisis

adjacency

adjacency

adjacency

interpolation adjacency or adjacency

Unrestricted Restricted Restricted no no IP-scrambling IP-scrambling IP-scrambling IP-scrambling IP-scrambling allowed allowed allowed allowed not allowed

Explanation of these facts under the views developed in this chapter

S AgrS V movement in main clauses

OP, OS

OC, OO

OF

MP

MS, MC, MO, MF

strong multiple Specs to S

strong one Spec

strong no Spec

strong no Spec

weak not settled

to S

to a position above S

to a position below S

to a position below S

clitics. This does not appear to be the case with respect to any of the contemporary Romance languages. A last remark is due to conclude. This chapter is not intended directly to address the issue of linguistic change, concentrating instead on giving an integrated account of the set of grammars found in Romance with respect to clitic placement, VP-ellipsis, and scrambling. The loss of unrestricted IPscrambling in OP and OS is taken under scrutiny in Martins (2002). As for the weakening of S, the rationale of change is likely to emerge from a thorough diachronic inquiry on the patterns of answer to yes/no questions (an area where the collection of historical data is certainly a heavy, but maybe not hopeless, task).18

18

See on this matter Martins (1994b, 2002).

Clitic Placement, VP-Ellipsis, and Scrambling

193

Sources Blasco Ferrer, E. (1984). Grammatica Storica del Catalano e dei suoi dialetti con speciale risguardo all’Algherese. Tu¨bingen: Gunter Narr Verlag. Cejador y Frauca, J. (ed.) (1913). Fernando de Rojas, La Celestina. Madrid: Clasicos Castellanos—Ediciones de ‘La Lectura’. Costa, M. J. (2001). ‘Contributos para o estudo das perı´frases verbais com ‘‘aver’’ e ‘‘teer’’ no portugueˆs medieval’, in C. N. Correia, and A. Gonc¸alves (eds.), Actas do XVI Encontro Nacional da Associac¸a˜o Portuguesa de Linguı´stica. Lisboa: Associac¸a˜o Portuguesa de Linguı´stica, 179–86. Foulet, L. (1928, 1977, 1990). Petite Syntaxe de L’Ancien Franc¸ais. Paris: Librairie Honore´ Champion. Gifford, D. J., and Hodcroft, F. W. (1959). Textos Lingu¨´ısticos del Medioevo Espan˜ol. Oxford: Dolphin Book. Jensen, F. (1994). Syntaxe de l’Ancien Occitan. Tu¨bingen: Max Niemeyer Verlag. Keniston, H. (1937). The Syntax of Castilian Prose: The Sixteenth Century. Chicago: University of Chicago Press. Kok, A. de (1985). La Place du Pronom Personnel Re´gime Conjoint en Franc¸ais: Une E´tude Diachronique. Amsterdam: Rodopi. Martins, A. M. (2001). Documentos Portugueses do Noroeste e da Regia˜o de Lisboa: Da Produc¸a˜o Primitiva ao Se´culo XVI. Lisboa: Imprensa Nacional—Casa da Moeda. Mene´ ndez-Pidal, R. (ed.) (1978). Primera Cro´nica General de Espan˜a, editada por Ramo´n Mene´ndez Pidal con un estudio actualizador de Diego Catala´n. Vol. I. Madrid: Gredos. —— (1919). Documentos Lingu¨´ısticos de Espan˜a. I: Reino de Castilla. Madrid: Centro de Estudios Histo´ricos. Moignet, G. (1973). Grammaire de l’Ancien Franc¸ais: Morphologie – Syntax. Paris: Klincksieck. 1976 (2nd edn.). Pinkster, H. (1990). Latin Syntax and Semantics. London: Routledge. Russell-Gebbett, P. (1965). Mediaeval Catalan Linguistic Texts. Oxford: Dauphin Book. The Hamburg Corpus of Old French. Collaborative Center on Multilingualism, University of Hamburg. Vance, B. (1997). Syntactic Change in Medieval French Verb-Second and Null Subjects. Dordrecht: Kluwer. —— (2001). French and Occitan V2 systems in diachrony. Paper delivered at the Colloquium on Interaction of Language Systems. Hamburg, October.

12 Microvariation in French Negation Markers: A Historical Perspective FRANCE MARTINEAU A ND MARIE-THE´ R E` SE VINET

12.1 Introduction Studies on the evolution of markers of negation in the languages of the world have been very productive ever since Jespersen’s (1917) work on the negation cycle. The goal of the present study on the evolution of French negation markers will be restricted to the discussion of certain facts connected to negation markers and the left periphery which have been too often neglected and sometimes misinterpreted in the literature on historical change. First, we illustrate how the history of the absence of ne in certain clauses identiWed in the lower CP domain must be distinguished from the deletion of the discontinuous IP negation marker ne which developed in certain morphosyntactic surroundings in spoken French (Section 12.3). Second, we explore how semantic and syntactic factors may explain the absence of ne in interrogative clauses in certain grammars of French (Section 12.4). Most importantly, we want to emphasize the fact that the realization of French negation from a diachronic perspective has never been a homogeneous phenomenon. Just as contemporary grammars of French are numerous and varied, diachronic grammars of French were far from the homogeneous picture too often presented in many studies, even those reporting on the oral form of the language in the past. Such grammars, just like contemporary grammars of French, also displayed microvariation. We therefore demonstrate that French negation had diVerent realizations in diachronic grammars of the sixteenth, seventeenth, and eighteenth centuries. This situation is puzzling. Why, for instance, would the evolution of negation markers within interrogatives We follow alphabetical order for authors’ names. This research received a grant from SSHRC (Canada) (410-2001-0119). We would like to thank A. Lodge for allowing us to use his computerized corpus.

Microvariation in French Negation Markers

195

which display discursive eVects be distinct from the one observed in sentential negation? The diVerence here is clearly not linked to genre distinctions between formal and non-formal registers of speech. This general situation raises questions both for the cyclic development of French negation and for the motivation behind the evolution of negation.

12.2 Corpora of Familiar French from France and Quebec To get an overall view, we used, for Old, Middle, and sixteenth-century French, corpora from two databases: ARTFL (American Research on Tre´sor de la Langue Franc¸aise, University of Chicago) and TFA (Textes de Franc¸ais Ancien, University of Chicago). We also used corpora of familiar French from France and from Quebec. For European non-standard French, our study is based on Lodge’s seventeenth- and eighteenth-century corpora of plays, comedies, farces, or vaudevilles that included, among others, rural or urban characters from the lower strata of society, on Chavatte’s diary as it was edited by Ernst and Wolf (2002), and on letters written by low-educated people in the north-east of France from Martineau’s corpus. We also looked at He´roard’s transcript of the speech of Louis XIII as a young child (1605–11). For nineteenth-century French, we used comedies including not welleducated characters. For the beginning of the twentieth century, we looked at Barbusse, an author that includes informal speech in his novels (cf. ARTFL database). Concerning Quebec French, we used Martineau’s corpus of letters produced during the seventeenth, eighteenth, and nineteenth centuries (Martineau 2003); these family letters were produced by weakly literate people whose purpose was not to have their letters read by a large audience. We also examined plays and comedies written during the nineteenth century in Quebec in which low-stratum characters are represented, as well as Faribault and Portelance’s corpus of letters written to Father Bernier by lower-stratum people at the beginning of the twentieth century. It should be pointed out that the precious pieces of evidence gathered in the relatively few texts written by ‘weakly literate’ persons must be handled with care: they are not an exact picture of the oral form spoken at the time and they always remain a reconstitution of an image. However, as the oral form of ancient French is lost forever, it is clear that these handwritten texts are particularly important for reconstructing the evolution of spoken French (Ayres-Bennett 2002). In fact, we want to argue that microvariation may be hidden by lack of distinction in corpora registers, and that the use of this material allows us to have a better image of diachronic change.

196

F. Martineau and M.-T. Vinet

12.3 Absence of ne: Left Periphery and Sentential Domains 12.3.1 Jespersen’s cycle

The Jespersen’s cycle in (1) identiWed three diVerent strategies for the expression of sentential negation from a diachronic perspective. The basic claim was that the Xuctuation between diVerent types of negative markers is related to phonetic stress: (1) a. a weakly stressed negative marker becomes a clitic: non->ne; b. and is then strengthened by another element: ne . . . pas; c. which comes to be interpreted as the negative marker: pas. Research by Zanuttini (1997: 14) on negation and clausal structure has shown that these three diVerent strategies could also be found at the synchronic level, not only across contemporary Romance languages, but also in diVerent varieties of the same language, namely dialects spoken in Italy. She has indicated that central and northern varieties of Piedmontese use only a postverbal negative marker while other varieties of Piedmontese (towns on the border of Piedmont and Liguria) have a weakly stressed negative marker. The variety spoken in Cairo Montenotte has both a pre- and a postverbal negative marker. In the same manner, it is expected that these identiWed strategies were also observed at the diachronic level in French, where a variety of negative forms were used at diVerent stages depending often, in the case of French, on the interaction with other grammatical phenomena. 12.3.2 Pas: sentential, indeWnite, and presuppositional readings

Historical studies on ne deletion in non-standard spoken French are still quite divergent regarding the period at which pas comes to be interpreted as the only negative marker in the sentence, as in (2a), compared to (2b): (2)

a. Je veux pas y aller. ‘I don’t want to go.’ b. Je ne veux pas y aller.

Some have argued that ne deletion was already widespread in the seventeenth century (Valli 1983; Blanche-Benveniste and Jeanjean 1986; Hausmann 1992; Grieve 1984), while others have shown that its rise is a much more recent phenomenon (Ayres-Bennett 1994; Ashby 1981; Martineau and Mougeon 2003). However, studies on negation in diachrony have often disregarded the various interpretations connected to the use of ne or pas. In Old and Middle French, in addition to its negative reading, pas (and point, mie, neant) could have an indeWnite reading, in polarity contexts

Microvariation in French Negation Markers

197

(in interrogative contexts but also in hypothetical, comparative, etc.), as in (3): (3) a. Ave´s point d’amis? ‘Do you have a single friend?’ [Me´nard’s translation: ‘Auriez-vous un ami? / Avez-vous le moindre ami?’] (Romances et Pastourelles, 2. 10. 19; Me´nard 1976: 260) b. . . . devant ne ains, j Fors cest jor, ne l’avoit veu¨e, j Ne n’atocha a sa char nue j Dont ele fust pas empirie ‘Never before that day he had seen her or touched her naked skin and dishonoured her in any way.’ [Buridant’s translation: ‘Jamais avant ce jour il ne l’avait vue ni n’avait touche´ a` sa peau nue en la de´shonorant en quoi que ce soit.’] (Le Chevalier de la charrette, 4970–3; Buridant 2000: 721) In this context, pas behaves similarly to rien and aucun, which could also be used as polarity items in Old and Middle French, as in (4) and (5): (4)

a. Et dit qu’il me fera ennui j Si je de rien paroil a lui ‘And he says that he will make me regret that I speak to him about anything’; [Buridant’s translation: ‘Et il dit qu’il me fera ame`rement regretter que je lui parle de quoi que ce soit’] (Erec, 2965–6; Buridant 2000: 714) b. Le roi . . . doubtant aucun inconvenient, l’en Wst retraire ‘The king . . . afraid of any problem, made him go out.’ (Jehan de Saintre´, 88; Marchello-Nizia 1997: 186)

Whereas it has been shown that use of aucun and rien as polarity items tends to disappear during the eighteenth century and is nowadays restricted to some dialects, for instance Quebec French (QF), or to some syntactic contexts (De´prez and Martineau 2004; Martineau and De´prez forthcoming), no systematic research, to our knowledge, has been made on the use of pas as a polarity item. However, it seems clear that in the sixteenth century, its use is contextually restricted to the interrogative context, as in (5)—cf. Gougenheim (1984: 242). During the seventeenth century, following Muller (1991: 224), this indeWnite use of pas disappeared.1

1

Following Muller (1991), this use would have been maintained in QF in some contexts:

(i) C’ est le plus beau ble´ qu’il y a pas sur le marche´. dem is the most beautiful wheat that 3ps there has neg on the market ‘This is the most beautiful wheat that one can Wnd on the market.’

198 (5)

F. Martineau and M.-T. Vinet Et moy, suis-je dans un bain? Suis-je pas plus a` mon aise que toy? ‘And me, am I in a bath? Am I so less than anything more at ease than you are?’ [Muller’s translation: ‘Suis-je si peu que ce soit plus a` mon aise que toi?’] (Montaigne, Essais, iii. 6; Muller 1991)

The presuppositional use of pas in yes/no interrogatives contrasts with its use as an indeWnite. In Old and Middle French and in Classical French, pas (or point) is commonly used with a presuppositional reading without ne, in direct interrogative as in (6), and also, more rarely, in indirect interrogatives, as in (7)—cf. Haase (1969), Fournier (1998): (6) a. Middle French Que te semble de ma nouvelle Espousee? Est elle pas belle what you seems of my new wife is she neg nice Et honneste souYsanment? and honest enough ‘What do you think of my new wife? Isn’t she nice and honest enough?’ (Griseldis, v. 2430–2; Marchello-Nizia 1997: 306) b. Seventeenth-century French Piarot t’ es ban fou, sais tu pas qu’ nan ne poigera Piarot you is very fool know scl.2ps neg that they cl.neg pay poen de Taye . . . neg of taxes . . . ‘Piarot you are a fool, don’t you know that they won’t pay taxes . . .’ (Agre´ables Confe´rences, iv, l. 177; DeloVre 1999) c. Eighteenth-century French As-tu pas peur qu’ pendant s’ temps-la` on n’ are-scl.2ps neg scare that during this time 1pp neg mange ton manger que vla eat your food that here ‘Aren’t you afraid that during this time we won’t come and eat your food here’ (La Pipe casse´e, iv. 153–4; Wu¨est 1985: 252–3) (7)

Middle French . . . luy demanda s’il estoit point amoureux d’une telle damoiselle . . . ‘and (he) asked him if he wasn’t in love with a given young lady’ (Cent Nouvelles, 227. 21; Martin and Wilmet 1980: 33)

Microvariation in French Negation Markers

199

In these contexts, absence of ne seems to alternate with presence of ne, as in (8). As far as it can be seen, the choice is not triggered by formal or informal factors. For instance, we found absence of ne in these contexts in well-known seventeenth-century authors such as Se´vigne´, Vaugelas, Pascal, and Bossuet. (8) Medieval French N’es tu pas apreste´e? Qu’est ce?j Temps est que nous aillons a messe. ‘Are you ready to go? What is it? It’s time to go to church’ (Anonymous, Miracle de Saint Jehan Crisothomes, 271) In Old French, ne alone could also participate to the presuppositional reading of (9), and this last interpretation was sometimes reinforced by discursive elements such as donc or et. (9)

N’est la raı¨ne Ysolt ta amie?j -Oı¨l, par foi, je nel ‘Isn’t the queen Ysolt your friend? Yes, truly, she is and ni mie I don’t hide it’ (FolieTristOx, 386–7; Buridant 2000: 691)

Most studies on ne deletion (and statistical results also) have not pointed out the need to distinguish the structures in which pas contributes to sentential negation (cf. (2)) from those where pas does not trigger a negative force reading (cf. (3), (5)–(8)). In the next section, we examine statistical results with yes/no interrogatives. 12.3.3 Statistical results and yes/no interrogatives

Martineau and Mougeon (2003) have based their study of sentential negation on a corpus of non-standard French. Their results show clearly that deletion of ne, in sentential negation, is a relatively recent phenomenon; the rate of deletion is very low in the seventeenth and eighteenth centuries but increases markedly during the nineteenth century, as shown in Table 12.1, under the column Sentential negation. In Table 12.1, our own results show that the structures in which pas has a negative interpretation (as in imperatives or sentential negation IP structures) have distinct results from those where pas does not trigger a negative reading (as in yes/no interrogatives and expressive structures with discursive eVects). In the Agre´ables Confe´rences, a Wctitious text considered to be representative of the oral form of French in the seventeenth century, almost one yes/no interrogative out of two (45.4 per cent (10/22)) is found without ne, as in the question form in (10) below, with the presence of the positive reinforcer ban. Wu¨est (1985: 252–3) notes three examples from La Pipe casse´e

200

F. Martineau and M.-T. Vinet

Table 12.1 Rates of ne deletion according to the type of structure. Periods

Yes–no interrogatives

Negative imperatives

Sentential negation

Old and Middle French 16th century 17th century 18th century 19th century 20th century

15% (16/108) 26% (32/124) 26% (14/54) 25% (8/32) 40% (18/45) 27% (4/15)

0% (0/145) 0% (0/29) 0% (0/12) 0% (0/21) 19% (7/37) 58% (22/38)

0.1% (2/1691) 0.3% (4/1238) 0.4% (4/1171) 2% (18/848) 45% (326/723) 32% (559/1753)

Note : Rates under ‘Sentential negation’ for the 17th, 18th, and 19th centuries are extracted from Martineau and Mougeon (2003).

(eighteenth century) which appear without ne. All three are found exclusively in interrogative clauses; two of them present ban or ti (11): (10)

Te souvan tu pas ban de ce Carnaval . . . ? refl.2ps remember scl.2ps neg well of that Carnaval ‘Don’t you remember well that Carnaval . . .’ (Agre´ables Confe´rences, i, l. 40; DeloVre 1999) (11) a. he´ sc¸ay tu pas ban que c’est le Cardena? hey say scl.2ps neg well that it is the cardinal ‘Hey don’t you know that it is the cardinal?’ (Wu¨est 1985: 252–3) b. vous souvanti pas ban quan nout barbie´ mouti? 2pp remember.ti neg well when our barber died ‘Don’t you remember when our barber died?’ (Wu¨est 1985: 252–3) Our results in Table 12.1 show convincingly that deletion of ne was far more extended with these latter structures before the nineteenth century. This situation therefore raises questions: if absence or presence of ne in yes/no interrogatives is not constrained by phonetic weakening of ne as in sentential negation (triggered by aYxation of the subject clitic; cf. Harris 1978 and Ashby 1981, among others), what are the constraints on the choice between these two strategies?

Microvariation in French Negation Markers

201

12.4 Negation Connected to Tense, Mood, and Discursive EVects Linguistic research on negation has shown that negation often interacts with other grammatical phenomena in revealing ways. It can indeed be connected to tense, mood, and a variety of discursive or aVectedness eVects related to the left periphery of the clause. Morphemes which express negation, whether they display head or non-head behaviour, are always dependent upon surrounding semantic units for their meaning. As observed above, in yes/no interrogatives within previous stages of the French language, both ne alone or pas alone have been used without the ordinary interpretation of negation. As in Portner and Zanuttini (2000), we hold the view that the markers used to express the force of negation in yes/no interrogatives and in exclamative/expressive structures are always negative. They are real negatives, and their contribution to the meaning of the sentence is hidden by certain semantic features of the context. These forms need overt checking of additional projections in the left periphery. An empirical claim for the hypothesis that informative questions and presuppositional interrogatives must be checked in a special position in the structure can be found in contemporary Quebec French—see (13) and (14) below—illustrating how dialect variation can sometimes shed light on diachronic processes, as also claimed in Beninca` and Poletto (2005). Our working hypothesis is that similar strategies also existed within previous stages of the French language. Let us suppose for ease of exposition that the context in (6a) has a Question (Q) or wh-operator which must have scope from a lower left peripheral position, and the domain of this operator is always to its right. When the negation marker ne or pas alone scopes in the domain of this particular operator, it lacks Negative Force. Note that the expletive interpretation of the negation markers does not change the Xuctuation between the head or XP status of the negation markers. A theoretical representation of these structures is given through the general architecture of the left periphery (cf. Rizzi 1997), sketched in (12a). Studies on negation within the IP or CP domain give rise to a representation in terms of movement to diVerent layers of the clause, as initiated by Pollock (1989) and developed by Rizzi (1997), Poletto and Pollock (2004), and many others. It is assumed that the architecture of the left periphery has a Wxed component involving heads specifying Force which opens up the complementizer system and Finiteness which closes oV the system downward as in (12a). The operator Neg which appears in the scope of a Question operator identiWed from a left peripheral position through

202

F. Martineau and M.-T. Vinet

movement in CP of some lexical element(s) or a Remnant IP is then schematically represented as in (12b): (12) a. Force Top Foc wh Fin b. [CP . . . [wh/Q [IP . . . Neg . . . ]] The co-occurrence of two diVerent patterns or the concomitant variation in the use of either pas alone or ne . . . pas in yes/no interrogatives with discursive eVects is therefore attributed to the existence of microvariation systems within yes/no interrogatives. The source behind this syntactic microvariation in French grammars from these periods remains unclear but the regularity observed suggests that it is connected to the semantic features involved and to a type of clause bearing discursive eVects, such as presuppositional question forms and expressive structures which presuppose their content. Two constraints on these structures show the role of syntactic and semantic features, namely subject clitic inversion (SCL-I) and the presence of an explicit positive marker. As can be observed, in Old and Middle French and in the sixteenth, seventeenth, and eighteenth centuries, most yes–no interrogatives appear with SCL-I and they allow pas alone. In contemporary French, the modern equivalent of (6a) above (with SCL-I) is unacceptable with pas alone: (13) Contemporary French a. N’ est-elle pas belle? neg is-she neg lovely ‘Isn’t she lovely?’ b. *Est-elle pas belle? is-she neg lovely Whereas the same yes/no interrogative without clitic inversion is acceptable without ne: Elle est pas belle? or with ne: Elle n’est pas belle? Similar facts with SCL-I interrogative forms in contemporary Veneto dialects have also been found. Beninca` and Poletto (2005) have observed that in one dialect (Paduan), the preverbal negation marker is optional, whereas in the S. Anna dialect the presence of the preverbal marker is rather ruled out. Why has contemporary French developed this constraint on SCL-I and the presence of ne? We do not consider that the obligatory presence of ne is linked to the more formal style of the SCL-I in contemporary French. We rather assume that some parameter changes, connected to verb movement and rise of SV word order in interrogatives, must be implied. Similar facts have been observed in imperatives by Hirschbu¨hler and Labelle (2001)

Microvariation in French Negation Markers

203

who show that, with the pronouns en and y, ne is obligatory in negative imperatives if these pronouns are preverbal: N’en mange pas / *En mange pas. (‘Do not eat it’). On the contrary, some forms, which are never used in a more formal register, are impossible with the head ne, as in colloquial expressive structures with an emphatic aYrmative operator -tu (Vinet 2000, forthcoming) in contemporary expressive structures from QF: (14) Fak la`, il (* ne) part-tu pas a` crier suddenly he neg starts-tu pas yelling ‘Suddenly, he starts yelling’ Moreover, it can be observed in (15b) that pas is completely ruled out in informative questions with the enclitic interrogative marker -tu in contemporary QF (cf. Vinet 2000, 2001). The marker -tu, which can appear in various other discursive contexts in the grammar of QF, has been identiWed as a super positive or an aYrmative marker in Vinet (2000, forthcoming). The question form becomes acceptable only when an irrealis mood (like a conditional) turns it into a presuppositional interrogative form, as illustrated in (15c). The negation marker pas is obligatorily absent in informative questions—cf. (15b). This last sentence is interpreted as gibberish in QF: (15)

a. Ta me`re est-tu la`? your mother is-tu there ‘Is your mother in?’ b. *Ta me`re est-tu pas la`? your mother is-tu pas there ‘Is your mother in?’ c. Ta me`re serait-tu pas la` your mother should be-tu pas there (par hasard)? by chance ‘Isn’t your mother in by any chance?’

(QF: Inf. question)

(QF: Inf. question)

(QF: Presup. question)

In a certain way, the positive marker -tu in contemporary QF plays a role which is similar to the positive reinforcers ban (well), as exempliWed in (10, 11), or -ti in (16), in the seventeenth and eighteenth centuries: (16) a. Seventeenth-century French He la vela ti pas, la vela ti pas Ho 3pfs there ti neg 3pfs there ti neg

204

F. Martineau and M.-T. Vinet ‘Ho! There it is, there it is!’ (Journal de Jean He´roard: 1604; Foisil 1989) b. Eighteenth-century French C’est-il pas-la` comme le Pape ‘isn’t it somewhat like the Pope?’ (Les Sarcellades, from Lodge’s computerized corpus)

As is well known (cf. Pollock 1989; Laka 1990), aYrmative particles cannot cooccur with negative markers unless these markers lack a negative interpretation. This is exactly the situation observed in (10), (11), and (16) where pas lacks its negative value (or in QF: vla tu pas qu’elle se met a` pleurer). Just as for SCL-I and yes/no interrogatives, is there a free alternation between ne and ne . . . pas in these structures in older stages of French? Our data for the seventeenth and eighteenth centuries show prevalence of absence of ne, but presence of ne is not forbidden, as shown in (17): (17) Ne vlati pas la taye a cu ‘Isn’t there the . . .’ (Agre´ables Confe´rences, from Lodge’s computerized corpus) More research should be done on the exact role of ban and -ti in older stages of French as well as their relation to the negative marker ne.

12.5 Conclusion An understanding of the role played by phonological, syntactic, and semantic factors in the way French negative markers could combine can shed light on their evolution. Our Wndings suggest that diVerent realizations of negation within the IP/CP domain, as well as within various registers of speech, could have developed in partly independent ways in diachronic grammars. We have observed, for instance, that before the nineteenth century, many occurrences of preverbal negation were not present in yes/no interrogatives and this diVerence was not connected to a distinction between formal and non-formal registers. Recall that sequences as in (6a) are not possible in contemporary varieties of colloquial French. Remaining questions obviously concern (1) the exact motivation behind the concurrent use of both pas alone and ne . . . pas over a long period with these structures, and (2) the evolution of negation in the left periphery. To our knowledge, no attempt has been made to answer why these two diVerent strategies were used concomitantly.

Microvariation in French Negation Markers

205

Primary sources ARTFL Database, University of Chicago. Middle French and sixteenth-century French. Barbusse, H. Le Feu. (ARTFL Database, University of Chicago). Twentieth-century European French. Cent Nouvelles Nouvelles [1505] by Vigneulles (computerized version by P. Hirschbu¨hler). Middle French. Deloffre, J. (1999). Agre´ables Confe´rences de deux paysans de Saint-Ouen et de Montmorency sur les aVaires du temps (1649–1651). Slatkine reprints: Geneva. Seventeenth-century European French. Duval, E´. (1978). Anthologie the´matique du the´aˆtre que´be´cois au 19e sie`cle. Montreal: Lemeac. Nineteenth-century Quebec French. Ernst, G., and Wolf, B. (2002). Chavatte’s Diary. Niemeyer. Seventeenth-century European French. Foisil, M. (1989). Journal de Jean He´roard. Tome 1. Paris: Fayard. Seventeenth-century European French. Lodge, A., computerized corpus: Le Pe´dant joue´; La Nopce de Village; Agre´ables Confe´rences de deux paysans de Saint-Ouen et Montmorency sur les aVaires du temps; Les Mazarinades; Dom Juan ou le festin de Pierre; Lettres de Montmartre. Les Sarcellades. Seventeenth- and eighteenth-century European French. Martineau, F. Corpus de franc¸ais familier des XVIIe, XVIIIe et XIXe sie`cles. Seventeenth- and eighteenth-century European French. —— Corpus de franc¸ais familier des XVIIe, XVIIIe et XIXe sie`cle. www.uottawa. ca/academic/arts/lettres/nf. Seventeenth-, eighteenth-, and nineteenth-century Quebec French. Moreau, E´. (1912). Madame Sans-Geˆne. Paul OllendorV, Paris. Nineteenth- and twentienth-century European French. Te´miscouata corpus (F. Dupuis, M. Faribault and C. Portelance). Twentieth-century Quebec French Textes de Franc¸ais anciens (ARTFL): 79 texts. Old and Middle French.

13 Sı` and e as CP Expletives in Old Italian CECILIA POLET TO

13.1 Introduction In this chapter I examine the grammaticalization process of two CP elements in Old Italian: the particles sı` and e. Sı` will be claimed to have developed into an expletive occupying a Spec,Focus position (as originally suggested by Beninca` 1984, 1996) in the low part of the CP layer, while e will be shown to license a null category with the properties of a null constant (cf. Rizzi 1992) in its speciWer signalling the continuation of a Topic. More generally, this work contributes to shedding light on the complex phenomenon of V2 on the one hand and on the left periphery of the clause on the other, insofar as Old Italian represents the ideal case study to investigate how the V2 property interacts with a fully articulated left periphery. In order to shed light on the two particles taken into account here, I will Wrst address the question of the left periphery in some Old and Modern Romance languages in the light of recent developments of the CP structure that started with the seminal work of Rizzi (1997). I will Wrst examine the V2 and V3 sequences in another, less studied, V2 language, Modern Rhaetoromance, and explain the peculiar distribution of V3 in Old Italian adopting the split CP analysis proposed in Beninca` (2001, 2003), Poletto (2000, 2002), and further enriched in Beninca` and Poletto (2003), where the left periphery of the sentence is split into three ‘Welds’ each containing several projections, as illustrated in (1):

I am much indebted to Paola Beninca` and Nicola Munaro for fruitful discussion, and to Lorenzo Renzi and Giampaolo Salvi, who asked me to write a chapter on sentence structure in the Grammatica dell’Italiano antico, which gave me the original impulse to write this chapter. The database used here is the one kindly provided by the OVI (Opera del Vocabolario Italiano) CNR Institute and available on the web.

Sı` and e as CP Expletives (1)

207

[Hang. Topic [Scene Sett. [Left Disl. [List Interpr [[FOCUS.CP][wh-opCP]]] FRAME

THEME

FOCUS

In (1) the highest frame Weld contains a hanging topic position (HT) and a position for scene setting adverbials (Sc. Sett.). Left dislocated items (LD) are located higher than Focus, which is the lowest Weld of projections. Following Beninca`’s (2001) proposal for modern Italian, I will assume that no left dislocated element is located lower than Focus. More precisely, the following assumptions will play a central role in accounting for the peculiar distribution of multiple left dislocated constituents in Old Italian, and for the diVerence between Old Italian on the one side and Rhaetoromance on the other. Following Beninca` and Poletto (2005) I will assume that: (2) a. No LD element is located lower than Focus. b. In Old Romance Informational Focus is located in the CP. c. Hanging Topic and Scene Setting are located higher than (Declarative) Force,1 LD and Focus are lower. The chapter is organized as follows:2 in Section 13.2 I provide a description of the left periphery of Old Italian capitalizing on Beninca` (1995)’s work; in Section 13.3 I discuss a possible analysis of the diVerence between Old Italian and Rhaetoromance and I propose that in Old Italian V2 applies in the Focus projection, while in Rhaetoromance the ultimate target is Spec,Force, a much higher projection;3 in Section 13.4 the particle sı` is examined in detail and it will be shown that it is an expletive deprived of any lexical meaning which saturates the Spec,Focus position; Section 13.5 deals with the particle e, showing that it is used in non-coordinated clauses, and can be interpreted as a morphological marker of continuation of the Topic already set by the discourse; Wnally, Section 13.6 concludes the chapter.

1

What is meant here by Force is the position where the high declarative complementizer occurs in embedded inXected clauses, as originally proposed by Rizzi (1997). 2 In this work I will limit myself to the domain of Romance languages, although Germanic languages are clearly the Wrst comparison that comes to mind. The reason for this limitation is that it is not clear to me that German ‘Linksversetzung’ is the exact counterpart of Romance Left Dislocation, given that Romance languages have clitics, but German does not. 3 In this chapter I will not go into the problem of the trigger for V2. For the sake of the execution I will adopt Haegeman’s (1997a) and Robert’s (1999b) proposal that in V2 clauses the CP projection is endowed with EPP features. However, while these authors assume that the CP projection hosting the EPP feature is always the lowest one, I will try to capture the diVerences among languages by parametrizing the C8 head containing the EPP feature.

208

C. Poletto

13.2 The Left Periphery of Old Italian: V2 in a Complex CP As mentioned above I adopt here a layered structure of the CP component as the one originally proposed by Beninca` (2001) and further developed in Beninca` and Poletto (2001) already illustrated in (1). In principle, Old Italian should have the same left peripheral positions that Modern Italian displays, so the structure should be the same. However, Old Italian diVers from Modern Italian because it shows some of the phenomena typically connected to the V2 property: Old Italian displays subject inversion of the ‘Germanic’ type (i.e. the subject is found immediately after an auxiliary but in front of a past participle, of the direct object and more generally low adverbs in the sense of Cinque 1999): (3)

a. percio` che primieramente avea ella fatta a llui ingiuria because that for Wrst had she done to him oVence ‘because Wrstly she oVended him’ (Rettorica, p. 116, r. 15) b. quali denari avea Baldovino lasciati loro which money had Baldovino left them ‘how much money Baldovino left them’ (Doc. Fior., 1272–8, p. 437, r. 29)

(4) a. un giorno tolse questo re molto oro one day took-away this king a lot of gold ‘one day this king took a lot of gold’ (Nov., 142) b. cosı` ti manda lo nostro signore a dire so you sends the our lord to tell ‘our lord is sending you this message’ (Nov., 138) In (3) the subject is found between the auxiliary and the past participle, in (4) it is located before the object of IP complements, in a position that is ungrammatical in Modern Italian (which has in general very limited cases of VSO; see Belletti 1999). Moreover, Beninca` (1984) Wrst noted that the system of pro-drop licensing in Old Italian depends on V to C movement,4 in the sense that when no V to C movement applies (in general in embedded contexts) a subject pronoun is obligatory for both argumental and expletive subjects,5 as (5c) shows: 4

The same holds for Old French (see Adams 1987 and Roberts 1993a). Adams (1987) and Vance (1989) extend this analysis to Old French. Apparently Old Spanish does not share this property and pro-drop is licensed in the same way it is licensed in Modern Spanish. 5

Sı` and e as CP Expletives

209

(5) a. lo Wgliuolo lil domando` tanto ch’ elli l’ ebbe the son to.him.it asked so that he it got ‘the son asked him so long that he got it’ (Nov., 166) b. se lla natura domanda cio` ch’ ella ha perduto if the nature asks what that she has lost ‘if nature is asking what she lost’ (Nov., 135) c. ch’ elli era meglio that it was better ‘that it was better’ (FF, 8) Pro-drop licensed by V to C movement is possible both with expletives (6a) and with argumental subjects (6b,c). (6)

a. Gia` e` detto soYcientemente dell’ oYcio e della Wne already is said enough of.the work and of.the goal di rettorica of rhetorics ‘I already talked about the reason and goal of rhetorics.’ (Rettorica, p. 53, r. 6) b. Figliul mio, non posso star piu` teco son mine not canþ1sg stay anymore you.with ‘Oh my son, I cannot stay with you any longer.’ (Nov. 294) c. ma va’ sicuramente per via tra la gente. but goes safely in street among the people ‘but he goes safe in the street among people’ (Tesoretto, v. 1817–18)

However, V2 is possible in embedded contexts on a more general basis than in languages like German and it is preceded by a complementizer (7), as in Scandinavian languages, and also in some wh-contexts (8): (7) a. e credo che di tutte le cose t’ intendi and thinkþ1sg that of all the things yourself understand ‘and I believe that you know everything’ (Nov., 127)

210

C. Poletto b. fuli detto che in sua pregione avea lo sovrano maestro was.to.him told that in his prison had the king master ‘they told him that the king had a master in his prisons’ (Nov., 126) c. Dissemi che ad Alexandro andava perche´ li donasse told.me that to Alexandro went because to.him gaveþ3sg ‘He told me that he was going to see Alexandro to be given money.’ (Nov., 132–3). d. e istabilio` che un pane intero li fosse dato per giorno and established that a bread whole to.him was given each day ‘and he established that he was give one entire bread every day’ (FF, 127)

(8) a. Qual ragione ci mostri, che a colui che per sua which reason to.us showþ2sg that to the one that for his bonta` aveva guadagnato non desti e a colui che avea kindness had earned not giveþ2sg and to the one that had perduto per sua colpa e follia, tutto donasti? lost for his fault and folly everything gaveþ2sg ‘Which reason do you give us, that you did not give anything to the one that had earned something because of his goodness but you gave everything to the one that had lost everything because of his folly and fault?’ (Nov., 147) b. quando della pietra vi dissi . . . when about.the stone to.you toldþ1sg ‘when I told you about the stone . . .’ (Nov., 129) c. Perche´ sı` ti sconforte? why so yourself discourage? ‘Why are you so discouraged?’ (Vita nuova, 100) In general, embedded V2 is also found in relative or temporal clauses. The case reported in (8c) is much less frequent and it is not found in embedded domains (see Beninca` 2003). In addition, Old Italian displays other properties that are found in other V2 languages, like scrambling (see Poletto forthcoming): (9) a. i nimici avessero gia` il passo pigliato, the enemies had already the pace taken

Sı` and e as CP Expletives

211

‘the enemies had already walked away’ (Orosio, 88) b. ch’ egli avea il maleWcio commesso that he had the crime committed ‘that he had committed the crime’ (Fiore di rett., p. 31, r. 12–13) c. dice che poi a`e molto de ben fatto in guerra et in pace. says that then has a lot of good done in war and in peace ‘he said that he has done a lot of good during war and peace’ (Rettorica, p. 26, r. 22) d. Quelli rispuose ch’ avea tutto donato he answered that had everything given ‘He answered that he had given everything.’ (Nov., 167) e. il quale da che ebbe tutto Egitto vinto, . . . whom since hadþ3sg all Egypt won ‘whom, since he won the whole Egypt, . . .’ (Orosio, 83) f. Allora il cavalero, che ‘n sı` alto mestero avea la mente misa, . . . then the knight that in so high work had the mind set ‘Then the knight, who had set his mind so high, . . .’ (Tesoretto, v. 1975) An exceptional property displayed by Old Italian is that, as noted by Beninca` (1984), it does not respect the V2 linear restriction on ordering. This is documented in the very early texts and is shown by the fact that V3 and V* cases are very massively present in the corpus: (10) a. sao ko kelle terre per tali Wni trenta anni li knowþ1sg that those lands for those limits thirty years them possette Santi Patri Benedicti owned Santi Patri Benedicti ‘I know that those lands have been owned by the Santi Patri Benedicti, for thirty years.’ b. La reina cosı` fece tutto the queen so did everything ‘So the queen did everything.’ (Nov., 258) c. Quando lo Nostro Signore Gesu` Cristo parlava umanamente when the our lord Gesu` Cristo talked humanly

212

C. Poletto con noi, in fra l’ altre sue parole ne disse che . . . with us among the other his words of.them said that ‘When our Lord Jesus Christ spoke humanly to us, he said, among other things, that . . .’ (Nov., 177)

The situation seems at Wrst sight paradoxical, as Old Italian possesses only some but not all the properties that are generally associated with the V2 phenomenon. Old French is a much more ‘well-behaved’ language, as it shows inversion, pro-drop licensing only when I to C movement applies, and very rarely has V3 orderings. Roberts (1993a) notes that V3 is attested quite late and that it is generally found only in interrogatives: (11)

L’ aveirs Carlun est il appareillez?6 the treasure Carlun is it made ready ‘Is the treasure of Carlun ready?’ (Roland 1. 643; from Roberts 1993a: 10)

(12) Cest nostre rei por coi lessas cunfundre? this our king why letþ2sg overwhelm? ‘Why do you let our king be confused?’ (Roland 1. 2583; from Roberts 1993a: 10) In order better to understand which mysterious property makes Old Italian so peculiar among V2 languages, I will brieXy analyse a Modern Romance language that has maintained V2: the Rhaetoromance dialect spoken in the High Badia valley in northern Italy.7

13.3 Rhaetoromance V2 and V3 Orders The possibility of testing speakers’ intuitions will provide insights into the opposition between Old Italian and other V2 languages. In Rhaetoromance the XP in Wrst position in V2 clauses can only be an operator-moved argument or adverbial, while LD or HT cannot occur in a V2 construction:8 6 Note that this is apparently a V2 case, but on the standard assumption that yes/no questions have a null operator in the Spec,CP position, this case can be virtually assimilated to V3. 7 The data come from the dialect spoken in the village of S. Leonardo in the higher part of the valley. 8 This is unexpected given that even languages like German, which are rigidly V2, admit LD constructions of the type exempliWed in (i):

(i) Den Hans, den habe ich gesehen the Hans then have I seen See Poletto (2002) for an analysis of this contrast based on the diVerent X’ status of the resumptive pronoun in German and Rhaetoromance.

Sı` and e as CP Expletives (13)

213

a. L GIAT a-i odu. the cat have.I seen ‘I have seen THE CAT.’ b. Duman vagn-el. tomorrow comes.he ‘He is coming tomorrow.’

The following sentences, in which the object is resumed by a clitic pronoun, are totally excluded: (14) a. *L giat, l’ a-i odu. the cat it have.I seen ‘The cat, I saw it.’ b. *Giani, ti a-i bel baie`. Giani to.him have.I already spoken ‘I have already spoke to Giani.’ c. *De Giani, a-i bel baie´.9 of Giani have.I already spoken ‘I have already spoken about Giani.’ LD elements cannot be combined with V2 and are generally excluded also from V3 declarative contexts. As (15) shows, the left dislocated DP resumed by the clitic l ‘it’ cannot be found in declarative clauses in either Wrst or second position. (15)

a. *Da trai a Giani l’ ti a-i sometimes to Giani it to.him have.I ‘Sometimes I gave a book to Giani.’ b. *A Giani da trai l’ ti a-i to Giani sometimes it to.him have.I

de. given de. given

However, V3 is possible with the following restriction: the Wrst element is necessarily an HT or a scene setting adverb, and must be followed by V2: (16) ?Duman, giani vaighest? tomorrow Giani see.you ‘Will you see Giani tomorrow?’ (17)

L liber, a giani ti l’ a-i bel de´. the book to Giani to.him it have.I already given ‘I have already given it to Giani.’

9 This sentence is impossible as a left dislocation. If the PP is focalized, the construction is grammatical.

214 (18)

C. Poletto *De Giani cun piero a-i bel baie´. of Giani with Piero have.I already spoken ‘I have already talked with Piero about Giani.’

Example (16), in which V2 has been combined with a scene setting adverb, is felt to be strange but possible. Example (17), where the XP in Wrst position is in principle ambiguous between an HT and an LD constituent, is also possible. The ungrammaticality of (18) shows that LD is impossible and that, in consequence, (17) must be interpreted as a case of HT. This indicates that PPs are not grammatical even in V3 constructions. One of the major diVerences between HT and LD is precisely that LD reproduces the case or the preposition of the internal argument it is associated with, while HT does not. The ungrammaticality of (18) shows that the case copying procedure of LD is banned even from V3 structures. Hence, in V3 structures only hanging topics or scene setting adverbs are admitted. The situation radically changes when interrogative clauses are taken into account: focus and wh-movement cannot be combined, as is the general case in V2 languages:10 (19)

*L liber a che ti a de, Giani? the book to who to.him has given Giani ‘Whom did Giani give the book to?’

Less expectedly, a wh-element can be preceded not only by hanging topics or scene setting adverbs, but also by LD elements (or a combination of these). Cases of PPs in Wrst position followed by a wh-item11 are perfectly acceptable in contrast to (18). In all these cases the wh-item must be adjacent to the verb, as (20c), (21b), and (22b) show: (20)

a. De Giani, con che bai-la pa? of Giani with whom speak.she int.part. ‘With whom is she talking about Giani?’

10 Modern Italian, which is not a V2 language, also bans the co-occurrence between focus and whitems in main interrogatives. This has led several authors to the assumption that focus and wh-items occupy the same position in the CP layer. However, Cinque (p.c.) and Rizzi (2002) point out that in embedded interrogatives the combination is indeed possible and the ordering in focus wh-element, as shown in (i):

(i) Mi hanno chiesto a gianni chi fara` questo regalo. me have asked to Gianni who will.make this present ‘They asked me who will make this present to john.’ 11 In yes/no questions LD, HT, or scene setting is obviously in Wrst position. I assume here the standard view that yes/no questions have a null operator in the same position of overt wh-items.

Sı` and e as CP Expletives

215

b. Giani, ci o-l pa? Giani what wants.he int.part. ‘What does Giani want?’ c. *Ci Giani o-l pa? what Giani wants.he int.part. (21)

a. L liber chi l tol pa? the book who it takes int.part. ‘Who is going to take the book?’ b. *Chi l liber l tol pa? who the book it takes int.part.

(22)

a. Gonot ula va-al pa? often where goes.he int.part. ‘Where does he often go?’ b. *Ula gonot va-al pa? where often goes.he int.part.

When possible, LD displays exactly the same features that it displays in Old and Modern Italian: it can be embedded (23e) and there can be multiple LD elements which do not display any Wxed order, as (23a) and (23b) show: (23) a. Giani, inier, ci a-al pa fat? Giani yesterday what has.he int.part. done ‘What did Giani yesterday?’ b. Inier, Giani, ci a-al pa fat? yesterday Giani what has.he int.part. done ‘What did Giani do, yesterday?’ c. Giani, inier, l as-t ody? Giani yesterday him has.you seen ‘Did you see Giani yesterday?’ d. Inier, Giani, l as-t ody? yesterday Giani him has.you seen e. Al m a demanee Giani, can c al vagn a ciasa he me has asked Giani when that he comes at home ‘He asked me when Giani is coming home’ We can summarize the properties of Rhaetoromance in the following way: . LD and HT are ungrammatical in the V2 position. . V3 instances are possible in declaratives only if the Wrst constituent is an HT (or marginally a scene setting adverb) and neither the Wrst nor the second is an LD.

216

C. Poletto

. V3 in interrogative clauses is possible if the Wrst element is an HT, a scene setting adverb, or an LD item. . Interrogative clauses also admit V412 structures, given that it is possible to combine HT or scene setting with LD items and there are several LD positions. We can adapt the basic layering of CP projections given in (1) to the Rhaetoromance data in order to explain the unexpected distribution of LD constituents and the asymmetry between declarative and interrogative clauses. Let us adopt the standard view for Romance languages (see among others Rizzi 1997 for arguments in favour of this view) that the position of declarative force is located higher than LD and Focus but lower than HT and Scene Setting, while the position targeted by wh-items13 is located lower than LD. At this point we just need to ‘add’ the V2 feature to the picture and we obtain the Rhaetoromance pattern. Although the general claim in much recent work (see, among others, Haegeman 1997a and Roberts 1999b) is that V2 is the eVect of an EPP feature in the lowest CP projection, I will propose that the projection endowed with the EPP feature is not the same in all V2 languages.14 That V2 is not a unitary phenomenon is well known from the comparative work on Germanic languages (see Santorini 1989, Vikner 1995, Zwart 1997, among others). More speciWcally, Zwart proposes that even within the same language a subject in preverbal position is not located in the same projection as adverbials and objects. I would like to maintain Zwart’s basic intuition that V2 eVects can be achieved by targeting diVerent projections both cross-linguistically and language internally, although the technical execution proposed here is rather diVerent. Suppose that in Rhaetoromance the EPP feature that triggers V2 in declaratives is located in declarative Force,15 as illustrated in (24). (24) [HT [Scene setting [ Declarative Force V2 [LD . . . [LD [Focus [Wh-op WH]]]]]]]

12

What is meant here by V4 is not only V4, but given that there can be more than one LD element, V5, V6, and more are possible as well. 13 For the sake of clarity the position targeted by wh-items is termed here Wh-op. This does not mean that the structure of an interrogative CP must be squeezed inside a single projection (see Poletto and Pollock 2003, 2004, and Munaro et al. 2002, for a more detailed analysis of interrogative clauses on the basis of a layered low CP-area). 14 V2 could be also accounted for in terms of remnant movement and not as V to C8. Note, however, that the analysis presented here is tangential to this question, and goes through also within a theory that does not allow head movement at all. 15 However, focalized elements probably move to the Focus layer to check their Focus feature before reaching the speciWer of Declarative Force.

Sı` and e as CP Expletives

217

This means that DeclP has to contain an XP in its speciWer and the inXected verb in the head. Focalized XPs can move from Spec,Focus to Spec,DeclP,16 while scene setting adverbs can move through Spec,DeclP on their way to the Scene Setting position are always base generated inside the position where they occur, therefore they can never satisfy the V2 constraint in DeclP. If virtually all elements can move to DeclP and satisfy the EPP there, why are LD elements ungrammatical? As DeclP is higher than the LD position, the order ‘LD V2’ is always banned. However, the logical possibility that an LD element satisWes the V2 by moving from LD to Spec,DeclP remains. The reason why this is excluded has to do with the fact that, as originally proposed by Rizzi (1997), LD positions have some features in their head which ban the movement of the verb through the LD head. Therefore, if an LD element is present, verb movement to Decl8 is blocked (see Poletto 2002 for a detailed discussion on this subject). In interrogative clauses there is no declarative Force. Therefore, unless there is another trigger for V2, there should be no verb movement to the CP domain. However, it is also a well-known fact that residual V2 targets in general interrogative clauses: this is true both in Germanic, where English has maintained I to C movement in (main) interrogative clauses, and in Romance, where subject clitic inversion (usually analysed as I to C) is possible only in a subset of contexts that includes (main) interrogatives (this holds both in French and in northern Italian dialects). Given that V2 and residual interrogative V2 are ipso facto dissociated in Modern Romance and in English, it is natural to assume that this is the case also in Rhaetoromance, where the projection involved in interrogatives is the lower Wh-op position and not Declarative Force. This is enough to derive the Rhaetoromance pattern seen above: in declarative clauses, the only cases of V3 are those where the Wrst of the two elements occurring in front of the verb are an HT or a Scene Setting. There can be no LD in front of V2, because there is no LD position higher than Declarative Force. In interrogative clauses the wh-item and the verb target the lower Wh-op position, which is located lower than LD. Therefore, in front of the wh-item there can be LD elements, HT, scene setting adverbials, or a combination of these. Notice that the structure in (24) is not an ad hoc assumption made to explain the V2 pattern. It is just the structure proposed in Beninca` and Poletto (2005) as a modiWcation of Rizzi’s split-CP on the basis of Modern Italian.17 16

Note that the movement from the criterial position Spec,Focus to Spec,DeclP is not banned by the freezing eVect. As recently noted by Rizzi (2004), when the inXected verb moves creating a complex head (in this case FocusþDeclForce), the freezing eVect is circumvented. 17 Recent work on French shows that this is probably true of French as well, with some minor modiWcations (see Mathieu 2003 and Deutjes et al. 2003).

218

C. Poletto

The only additional assumption put forward here is one concerning the projection involved in V2 clauses in Rhaetoromance, an assumption that any theory on V2 has to make. Although it is tempting to extend this analysis of Rhaetoromance to Old French as well, I will not attempt to do it here, partly because Old French might have diVerent grammars according to the period chosen for the investigation, and rather concentrate on Old Italian. As originally noted by Beninca` (1984), Old Italian diVers from other Old Romance languages in admitting V3 and V4 in all clauses, although preserving the other typical features of Old Romance V2: ‘Germanic’ subject inversion and pro-drop licensing only when V2 has applied and no V2 in embedded interrogatives. I propose that Old Italian diVers from Rhaetoromance (and probably from Old French) in targeting a diVerent, much lower, projection for V2. In Old Italian the CP endowed with the EPP feature is not declarative Force but Focus: given that Focus is lower than LD, HT, and Scene Setting, it is always possible to violate the linear restriction and have V3 and V4 cases. Beninca` (1984) convincingly shows that when there are cases of V3 or V4, the Wrst XP located to the left of the inXected verb is always Focus, while the XPs located higher are always cases of HT, LD, and/or scene setting. The parameter distinguishing between Old Italian and Rhaetoromance (and probably Old French) thus concerns the projection endowed with the EPP feature, which obligatorily attracts an element in its speciWer and the inXected verb in its head.18 If this is correct, we expect to Wnd cases of Spec,Focus expletives: Beninca` (1984) brieXy mentions the possibility of having the adverbial form sı` ‘so’ functioning as a Spec,Focus expletive in Old Italian and works this hypothesis out in Beninca` (2003). Here the expletive usage of sı` is contrasted with the non-expletive one and with the adverb cosı`, clearly related to sı` in terms of meaning. This will conWrm our expectations and also provide a more detailed picture of how an adverb can become an expletive.

13.4 Sı` as a Spec,Focus Expletive In this section and in the following one I analyse two particles, sı` and e. I will show that their original meaning and their formal properties can favour the reanalysis into markers of Focus (in the case of sı`) and Topic (in the case of e).

18

Note that even inside a theory of V2 without head movement the basic intuition expressed here, namely that V2 in Old Italian is located lower than V2 in other Romance languages, remains valid, although the technical execution of the parameter would look diVerent.

Sı` and e as CP Expletives

219

In Old Italian, the adverb sı` has a number of usages and positions, some of which are the same as its homonym cosı` ‘so’.19 Before going into the expletive usage of sı`, it is convenient to describe brieXy the distribution of sı` and cosı`. In a number of cases the two elements seem interchangeable, but there are contexts in which only sı` is possible and others where cosı` is the only form found in the texts. Sı` can occur as an adjectival or adverbial modiWer: (25) a. fue sı` giusto e guardo` sı` le mani da . . . was so right and looked so the hands that ‘he was so right and he looked at his hands that . . .’ (FF, 110) b. comincio` a tremare sı` fortemente . . . began to tremble so strongly ‘he began to tremble so strongly . . .’ (Vita nuova, 6) The same is true of cosı`, as the following examples show: (26) a. quando io vi dissi del cavallo cosa cosı` maravigliosa, when I you told of.the horse thing so marvelous ‘when I told you such a wonderful thing about the horse’ (Nov., 129) b. onde picciolo guiderdone diedi a llui di cosı` ricco insegnamento so small reward gaveþ1sg to him of so rich teaching ‘I gave him such a small payback for what he had taught me’ (Nov., 145) c. Cavaliere, che ha’ tu misfatto a questi sergenti, che ti knight what have you badly-done to these people that you menano cosı` laidamente hitþ3sg so strongly? ‘Knight, what have you done to this people, that they are hitting you so much?’ This seems to be the basic context in which there is a free choice between the two particles. However, they are not completely interchangeable. First of all, cosı` appears to be an adverb anticipating the context of the following sentence, while sı` is not:

19

Although it would be tempting to connect the two homonymous forms to the same etymological origin and hypothesize that sı` is a short form of cosı`, Lorenzo Renzi (p.c.) pointed out to me that this is not the case, as sı` is most probably derived from Latin sic.

220

C. Poletto

(27) a. Allora il lapidaro si rallegro` e prese l’ una pietra then the stone-worker cheered up and took the one stone e miselasi in mano e disse cosı`: . . . and put.it in hand and said so ‘Then the stone-worker cheered up, took a stone, put it in his hand and said: . . .’ (Nov., 124) b. e disse cosı` a colui che ritenne i bisanti: . . . and said so to the one that kept the money ‘and said so to the one who kept the money: . . .’ (Nov., 151) Second, sı` is also found in contexts where cosı` is not: sı` is the only form used in the CP layer introducing a clause. Thus, it appears before come ‘as’ (28) and before che ‘that’ (29):20 (28) a. sı` come appare a chi lo intende so how appears to who it understands ‘so as it appears to those who understand it’ (Vita nuova, 23) b. . . . che sı` come tiranno distrugeva la terra that so like tyrant destroyed the earth ‘. . . who destroyed the earth like a tyrant’ (FF, 103) (29) a. sı` che quasi dal principio del suo anno non apparve so that almost from.the beginning of.the his year not appeared a me to me ‘so that it did not appear to me since the beginning of the year’ (Vita nuova, 6) b. sı` che li chiovi pareano 16 lettere so that the nails looked like 16 letters ‘so that the nails looked like 16 letters’ (Vita nuova, 7) Apparently, sı` can be used as a CP element combined with a complementizer che ‘that’ or come ‘as’, while cosı`, although it can be moved to Focus 20 Modern Italian has a single work in these cases, siccome, which is clearly a fusion of sı` and come. However, siccome corresponds to English ‘as’, not to ‘so as’, which is translated as cosı` come. It is clear that sı` has completely disappeared from the language and it is only found inside complex complementizers. I will not go into the question of the correspondence between Old and Modern Italian complementizers, which is tangential to the question examined here.

Sı` and e as CP Expletives

221

(see (31) below), like all XPs, cannot be base generated in CP. It is precisely this property that makes sı` ‘prone’ to a reanalysis as CP expletive. Moreover, when sı` is generated in the CP layer of an embedded clause, it has the interesting property of climbing to an operator position inside the main clause:21 (30)

a. e ho sı` saputo fare che li sudditi miei and haveþ1sg so been able doþinf that the subjects mine m’ hanno cacciato me have chased away ‘and I have been so good that my subjects chased me’ (Nov., 143) b. a chi mi sa sı` pregare che io lo diparta dagli to whom me knows so pray that I him split from.the altri others ‘to whom who can pray me so much that I can split him from the others’ (Nov., 167) c. Pero` quando mi tolle sı` il valore, che gli spiriti pare but when me took so the courage that the spirits seem fuggan via runþinf away ‘But when he took me my courage so that it seems that the soul is running away’ (Vita nuova, 121)

Notice that in all these cases, sı` is found within the IP area of the main clause, in a scrambling position lower than the inXected verb but higher than objects (cf. (30c)). As shown above, Old Italian displays scrambling positions. Such a property has gone unnoticed up to now, but it is quite massively present in the texts, where both cases of objects preceding the past participle and cases of scrambling of one internal argument over the other are found. I will not go into this here (see Poletto 2003), but I would like to notice only that sı` seems to have the property of climbing from within the embedded CP (probably from Spec,CP) to a position located in the low portion of the IP layer of the main clause.22 This is never the case for cosı`, which is also never found to the left of a complementizer. 21 As recently proposed by Belletti (2003) and Poletto (2003), the low phase of the clause (vP) is also associated to a left periphery including Topics and a Focus position, which is probably the target of the movement of sı` in these examples. 22 Cases like the following ones are also found in the texts:

222

C. Poletto

Cosı` can be found to the immediate left of the inXected verb, in a position that corresponds to Spec,Focus in the schema in (1) on a par with all adverbs: (31) E Guglielmo, vedendo che cosı` era sorpreso, parlo` e disse . . . and Guglielmo seeing that so was surprised spoke and said ‘And Guglielmo, seeing that he was so surprised, spoke and said . . .’ (Nov., 225) Notice that when Spec,Focus is occupied by sı`, the adverb becomes void of any semantic import, as the following cases show: (33)

a. onde lo ingannato amico di buona fede mi prese per la hence the betrayed friend of good faith me took for the mano, e traendomi fuori de la veduta di queste donne hand and taking.me out of the sight of these women sı` mi domando` che io avesse so me asked what I had ‘hence the sincere friend who had been betrayed took my hand and took me out from the sight of these women and asked me what was wrong with me’ (Vita nuova, 58) b. E parlandomi cosı`, sı` mi cesso` la forte fantasia and talking.to.me like that so me stopped the strong fantasy entro quello punto ch’ io volea dicere . . . within that point that I wanted sayþinf ‘and since he talked to me like this, my fantasy stopped just when I wanted to say’. . . (Vita nuova, 98)

(i) a. ‘Che e` cio`, messer Rinieri, che voi non siete partito di Sardigna?’ ‘Certo’ disse what is this Sir Rinieri that you not have left for Sardinia sure said messere Rinieri, ‘sı` sono; ma io sono tornato per li scappini delle sir Rinieri so am but I am come back because of.the laces of.the calze’ shoes ‘Why are you here, Sir Rinieri, why didn’t you leave for Sardinia? I did, said Sir Rinieri, but I came back because of the laces of my shoes.’ (Nov., 305) b. Or se’ tu ancor qui? non avestu la torta? Messer, sı` ebbi. now are you still here not hadþ2sg the pie Sir so had ‘Are you still here? Didn’t you have your piece of cake? Yes, Sir, I had it.’ (Nov., 309) I will not deal with these cases because they seem to be diVerent from the usage of sı` as an operator that we are examining. Probably this sı` is the homonymous aYrmative particle meaning ‘yes’, which in Modern Italian can be used as pro IP.

Sı` and e as CP Expletives

223

c. Poi che detta fue questa canzone, sı` venne a me uno, then that said was this song so came to me one lo quale . . . who ‘After this song, a man came to me, who . . .’ (Vita nuova, 133) d. E discacciato questo cotale malvagio desiderio, sı` si and chased away this really bad desire so rivolsero tutti li miei pensamenti a la loro themselves went all the my thoughts to the their gentilissima Beatrice very kind Beatrice ‘and once I had chased that horrible desire all my thoughts went to their kindest Beatrice’ (Vita nuova, 153) e. e contendendo col maestro, sı` fece aprire la bocca and discussing with.the master so made open the mouth dello š nfermo e, col dito estremo, li vi puose of.the sick and with.the Wnger to-him there puts l veleno the poison ‘and discussing with the master, he let the sick man open his mouth and put some poison with his Wnger . . .’ (Nov., 153) f. La volpe andando per un bosco sı` trovo` un mulo: e il the fox going in a wood so found a mule and the mulo sı` li mostro` il piede dritto, mule so to.her showed the foot right ‘While the fox was walking in the wood, it met a mule, and the mule showed it its right foot’ (Nov., 182) We can conclude that there are two main diVerences between sı` and cosı`: (i) when they appear in Spec,Focus cosı` is never void of meaning, while sı` is; and (ii) sı` can be generated in the CP layer (and also raise to the main clause) while cosı` cannot. (34)

E lo and the in quel in that

femine e mandolle re cosı` fece: tolse di belle king so did took of beautiful women and send.them modo nel campo way into.the camp

224

C. Poletto ‘And the king did so: he took a lot of beautiful women and sent them so into the camp’ (Nov., 212)

Note that sı`, due to its lack of semantic import, can be considered an expletive when it is located to the immediate left of the inXected verb, but not when it is found in the other contexts described, because it retains its original meaning as a manner adverb. This is perfectly in accordance with the behaviour of other expletive elements, such as German es, which is not only an expletive, but also the third person neuter pronoun and can be an argument,23 or English there, which is also a locative pronoun. On the other side, we have seen that the property of being able to be merged directly in the CP layer crucially distinguishes sı` from cosı`. If this is so, it is not only the semantic endowment of an element that counts for its reinterpretation as an expletive; its formal features also play a crucial role in triggering the reanalysis. The study of the distribution of expletive sı` reveals a number of interesting properties which strengthen the idea that sı` is an expletive located in the speciWer of the Focus projection, which in Old Italian, diVerently from other Romance V2 languages, is endowed with EPP features. We comment on these properties in the following paragraphs. A. When sı` is void of meaning, it always occurs to the immediate left of the inXected verb. The only elements that can separate sı` from the inXected verb are clitics and the preverbal negative marker: (35)

a. sı` s’ abacino` degli occhi so himself burnt the eyes ‘he burnt his eyes’ (FF, 105) b. e, parlando spezialmente alli spiriti del viso, sı` disse and speaking in particular to.the spirits of.the face so said queste parole these words ‘and talking in particular about the spirits of the face, he said these words’ (Vita nuova, 8)

23 I will not take a view concerning the position in which expletive sı` is merged, whether it is in IP, and then moved, or whether it is directly merged in the position where it surfaces.

Sı` and e as CP Expletives

225

c. e dette queste parole sı` disparve, e lo mio sonno and said these words so disappeared and the my sleep fue rotto was broken ‘and after he said these words, he disappeared so that my sleep was interrupted’ (Vita nuova, 24) B. There are a number of elements that can precede expletive sı`. They are instances of LD, HT, and/or Scene Setting adverbs. Sı` very often occurs after an embedded clause or after a temporal clause indicating anteriority or simultaneity, or after a Topic or the subject. Furthermore, it often co-occurs with the particle e at the beginning of the structure (see below for an analysis of the particle e): (36)

a. E, che avra` cuore nobile et intelligenzia sottile, sı` li and that will.have heart noble and intelligence subtle so to.him potra` simigliare per lo tempo che verra` will.be able lookþinf like for the time that will.come ‘And who will have high and subtle intelligence, will look like him for the future’ (Nov., 118) b. in questo Pittagora sı` comincio` in this P. so began ‘In this moment P. began . . .’ (FF, 104) c. Lo mperadore Federigo stando ad assedio a Melano, sı` li the emperor Federigo being in siege of Milan so to.him si fuggı` un suo astore dentro a Melano escaped a his hawk inside to Milan ‘While the emperor Federigo was besieging Milan, one of his hawks escaped into Milan’ (Nov., 177)

In (36a) there is an embedded clause preceding sı`, in (36b) there are two XPs, a Topic and the subject (which is a topic itself), and in (36c) the preceding constituent is a clause expressing a contemporary event. Both properties are captured by the analysis of V2 proposed above: the adjacency of sı` to the inXected verb is straightforwardly captured by the fact that sı` is in the speciWer of the Focus projection where the verb occurs. Given the structure in (1), LD, HT, Scene Setting, more generally the Topic and Frame Welds (where also embedded clauses are located; see Munaro 2002 for a

226

C. Poletto

typology of the embedded clauses occurring in the Topic Weld) are located higher than Focus; therefore, they can only precede expletive sı`, as (36) shows. C. Sı` can also occur in embedded contexts. In cases where it does, it maintains the properties seen above for main clauses: (37)

a. . . . che l ferro, se l’ aopere, sı` si logora, se no l’ aopere that the iron if it use so is battered if not it use la ruggine il consuma the rust it destroys ‘that if you use iron, it breaks, but if you do not use it, the rust destroys it’ (FF, 146) b. Leggesi del re Currado, del padre di Curradino, che, reads of.the king Currado of.the father of Curradino that quando era garzone, sı` avea in compagnia dodici garzoni di when was boy so had in company twelve boys of sua etade, che li faceano compagnia. his age that to.him kept company ‘It is said that when the king Currado, father of Curradino, was a boy he had twelve boys to keep him company’ (Nov, 232)

This is also expected since Old Italian shows V2 in embedded declaratives (see section 13.2). D. The last, and most revealing, feature is that sı` occurs only with proclisis: (38) e di cio` sı` ne fue contenzione and of this so of.it was matter for discussion ‘and this was a matter for discussion’ (FF, 106) (39) sı` si ne diede questa penitenza so to.himself of.it gave this punishment ‘he gave himself this punishment’ (FF, 108) This fact can be explained in terms of Beninca`’s (1995) analysis: she claims that proclisis is triggered when the Spec,Focus position is occupied by an XP, otherwise the inXected verb has to raise higher than Focus and enclisis is required. Therefore, when the verb is in initial position or when it is only preceded by topicalized elements, enclisis is required. If sı` is located in Spec,Focus, it is bound to occur with proclisis and never with enclisis. I will

Sı` and e as CP Expletives

227

not go into this topic any further here and refer the reader to Beninca` (1995, 2003) for a treatment of enclisis and proclisis.24 The hypothesis that sı` is a Spec,Focus expletive also derives why Modern Italian has completely lost this element, which is only found inside complex complementizers: given that Modern Italian is not a V2 language, it does not have a V2 expletive. Notice furthermore that the disappearance of expletive sı` coincides with the disappearance of the adverb sı`. Modern Italian only displays the adverb cosı`. The form sı` can be traced back only inside complex complementizers like (sicche´ ‘so’ or siccome ‘as’) but is no longer an independent work on its own. Probably two diVerent factors played a role in the disappearance of sı`: (i) the fact that it had a peculiar status as a ‘raising particle’ also in non-expletive contexts, and (ii) the fact that there was another adverb, cosı`, which was very close in terms of meaning to sı`. The reanalysis of an element as an expletive and its successive disappearance when the syntactic system changed are, thus, the result of diVerent factors, which do not only include their semantics. The formal properties of an element (in this case its capacity of being merged in CP and raised to the main clause) also play a crucial role in its evolution.

13.5 E as a Topic Marker In the preceding section I proposed an analysis of expletive sı` in Old Italian based on two assumptions, namely the structure of the split CP in (1), and the claim that Old Italian selects Focus as its V2 projection. We will see that, on a par with sı`, the formal properties of e are decisive to understand why it has been reanalysed as a Topic marker. Although e is the conjunction head, it occurs in a number of contexts where it is clearly not conjoining two phrases. (40) e quando avea forbiti i piedi ed elli tornava fuori and when had cleaned the feet and he came back outside e rinfangavalisi vie piu` e tornava a ricalpitare and got mudded more and more and came back to step on il letto. E partisi e disse a Platone: . . . the bed and went away and said to Plato ‘and when he had cleaned his feet went back outside, put mud on them, came back inside and went up onto the bed. He left and told Plato’: (FF, 124) 24

Old French also has an element that looks suspiciously like Old Italian sı`. Even if we consider that Old French selects a higher projection for V2, like Rhaetoromance, it still has an expletive which is located in Spec,Focus (like in Old Italian) and then moves to Declarative Force, while the inXected verb moves from the head of Focus to the head of Declarative Force. This would mean that in Old French two of the CP layer projections are endowed with an EPP feature: Focus and Declarative Force.

228

C. Poletto

Examples like the one in (40) show that e has a function which is not the typical one of conjoining two phrases. First, every clause in (40) is introduced by the particle e, even though generally the conjunction head is only realized before the last member of a coordination when more than two XPs are coordinated. Second, if e were the conjunction particle, Old Italian would have the peculiar property of conjoining embedded and main clauses,25 which is in general not an option in the most well-studied Romance and Germanic languages. Third, the particle e appears also in front of sentences which are separated by a full stop, which is not generally the case with conjuncts.26 This type of e is extremely frequent in the texts. (41)

25

a. Plauto fue uno grande savio, cortese in parlare. E scrisse Plauto was a great wise kind in speech and wrote queste sentenze these sayings ‘Plauto was a great wise man, who spoke very kindly. He wrote these sayings.’ (FF, 104) b. Scipio Africano fue consolo di Roma e fue tagliato di Scipio Africano was consul of Rome and was cut of corpo a la madre e per cio` fue chiamato Cesare. body from the mother and therefore was called Caesar E dice uno Wlosafo che . . . and says a philosopher that (FF, 140) ‘Scipio Africano was console in Rome, he was born with a Caesarean cut and for this reason he was named Caesar. A philosopher says that . . .’ c. Stando lo ’mperadore Federigo- e facea dare l’ acqua being the emperor Federigo and let giveþinf the water alle mani e, innebriato il pane dell’ odore che n’ to.the hands and putting the bread in.the smell that of.it uscia, del mangiare, e quelli lo mordea, e cosı` il came out of.the food and he it bite and so it consumo` di mangiare, ricevendo il fumo e mordendolo. Wnished to eat getting the smoke and biting.it.

The phenomenon is known as paraipotassi in philological studies of Old Italian. This is not very strong evidence, because in several cases the punctuation is inserted by the editor of the manuscript. 26

Sı` and e as CP Expletives

229

‘While Emperor Federigo was standing there he commanded to bring water for the people to clean their hands, and he was putting the bread close to the smell which came out of the meal and ate it up’ (Nov., 177) d. e poi, quando tutto ebbe dato, et elli si fece and then when all hadþ3sg given and he himself let vendere, sellþinf ‘and then when he had given everything he let himself be sold’ (Nov., 162) e. quando entro` nella chiesa, et uno parlo` e disse: . . . when got into.the church and one spoke and said ‘when he entered the church one spoke and said . . .’ (Nov., 189) Some cases are particularly revealing: in (41c), e is found in front of both an embedded and a main clause, and the embedded clause is an inWnitival one. Furthermore, there are cases like (41d) where two instances of e are realized and, at Wrst sight, it appears that the conjunction head is doubled. However, if we assume that e can also have another function, these cases are not special doubling cases at all: the Wrst e is the real conjunction particle, but the second one is clearly not. A further test for this is the translation into Modern Italian. The Wrst e in (41d) is perfectly grammatical, while the second is completely impossible: (42)

(e) poi quando ebbe dato tutto, (* e) egli si fece and then when hadþ3sg given all and he himself let vendere (Modern Italian) sellþinf ‘and then when he had given everything he let himself be sold’

The same is true with respect to (41e), where the particle e would be completely excluded in Modern Italian. Although sequences like (42) are completely ungrammatical in Modern Italian, there are still residual cases of e which look similar to the Topic marker we are dealing with here for Old Italian. While sı` has completely disappeared from the language, Modern Italian has retained the Topic marker e when the CP layer is activated, as in interrogative clauses: (43)

a. E io? ‘What about me?’

230

C. Poletto b. E adesso? ‘Now what?’

(44)

a. E cosa potrebbe fare in un frangente simile? RQ interpretation ‘And what could (he) do in such a case?’ b. E viene quando, allora? non-echo wh- in situ ‘And come when, then?’

Moreover, the presence of e indicating the existence of a Topic favours the interpretation of the sentence as a rhetorical question, as in (44a). E can also trigger non-echo wh- in situ in Italian (cf. (44b)), which is probably only possible when a Topic is present. Hence, in non-echo wh- in situ clauses in Italian, e is necessary to license the null Topic. Gemma Rigau (p.c.) pointed out to me that the same is true in Spanish (for an analysis of these cases in Modern Italian, see Poletto forthcoming). Apparently, the Topic marker e in Modern Italian can only be parasitic on an already existing CP layer. It would be tempting to assume that e can only represent the continuation of a Topic if there is a Focus layer active, which is always the case in Old Italian, being an ‘I to Focus’ language, while in Modern Italian this is true only of some structures. If this is so, we predict that e is possible also in any other case in which a wh-item is present, as in exclamative clauses:27 (45) E che vestito che ti sei comprato! and what dress that yourself are bought ‘What a dress you bought!’ Another construction involving focalization in the CP layer is the so-called ‘anaphoric anteposition’ (see Beninca` 1988), where the Topic marker e seems to be obligatory. Consider the following example: (46)

Speaker A: Gianni voleva comprarsi un castello. ‘Gianni wanted to buy a castle.’ Speaker B: E un castello si e` comprato. ‘And a castle he bought.’

Here the particle e cannot be a conjunction. It is pronounced by the second speaker and signals the continuation of the same Topic plus the focalization of the DP un castello ‘a castle’.

27 In some northern Italian dialects (i.e. Taglio di Po), the particle e has become a sentence type marker for exclamative clauses.

Sı` and e as CP Expletives

231

However, the Topic marker is excluded in contrastive focalizations, which are also analysed as involving a Focus projection in the CP layer: (47)

??E IL VESTITO ha comprato, non il cappello. and the dress has bought not the hat

Cases like (47) are very strange and diYcult to interpret. They constitute a problem for the hypothesis that the Topic marker e can only be present when Focus has been activated. However, the contrastive Focus constructions might be excluded on independent grounds. For instance, they might already have a null Topic of a diVerent kind. This could also be the case for Old Italian, given the lack of detailed studies on contrastive focus in this language (probably due to the diYculty of circumscribing the relevant contexts). Therefore, I propose that the particle e occurring in these contexts is not the usual conjunction head, but a diVerent element, which, as we will see, displays diVerent properties. If it is not a conjunction, what is it? As anticipated above, the pragmatic context in which the particle occurs suggests that e is a form of Topic marker; more precisely it marks the continuation of the same Topic present in the preceding discourse. The Wrst argument in favour of this interpretation is that this type of e never appears after a Topic element, but always at the very beginning of the clause. The only elements that can precede the particle are temporal clauses. Moreover, if e is immediately followed by the verb, it obligatorily triggers enclisis. In the section about sı`, Beninca`’s analysis of enclisis and proclisis has been brieXy illustrated: if Spec,Focus is occupied, there is proclisis; if Spec,Focus is empty and the only elements preceding the inXected verb are Topics or the verb is in Wrst position, then enclisis is triggered. We saw that sı` being an expletive in Spec,Focus, it can only occur with proclisis. If e is indeed a Topic marker, and as such occurs in a Topic position, we expect to Wnd enclisis when it is immediately followed by the verb (i.e. when Spec,Focus is empty). This prediction is borne out, as the particle e only triggers enclisis when it immediately precedes the inXected verb: (48) a. e and (FF, b. e and (FF,

tenerlo keep.it 135) bevenne drank.of.it 134)

The properties of the particle e examined so far are thus the following:

232

C. Poletto

. The Topic particle e only occurs at the beginning of a clause. . The Topic particle e can occur in front of both main and embedded clauses (giving the impression of a strange coordination structure with a main and an embedded inWnitival or gerundive clause). . The Topic particle e can co-occur with the conjunction head e (giving the impression of ‘conjunction doubling’). . The Topic particle e only triggers enclisis when it is immediately followed by the inXected verb, like all Topics. At this point one might ask why just the item e is used to mark the continuation of a Topic. Is there any link between this usage and the coordination usage? In other words, are there two distinct items with diVerent properties listed in the lexicon, or is there just one item and the diVerent properties depend on other factors? In order to be more precise, I will adopt the analysis of conjunction structures originally proposed by Kayne (1994), namely that conjunction also conforms to X’ theory in taking a conjunct as its complement and another conjunct as its speciWer, with the restriction that the speciWer and the complement must be the same type of category: (49) [CoP XP [Co8 e [XP]]] The particle e is the head of the conjunction structure, taking the same type of XP in its complement and in its speciWer position. Suppose further that when it is a Topic marker, e also takes the same structure: (50) [TopicP (Null)Top [Topic8 e [TopicP [CP]]]] The only diVerence between (49) and (50) is that the label XP has been substituted by TopicP. Thus, e maintains the property of taking a speciWer and a complement which are of the same type. The complement TopicP is the highest projection of the whole clause occurring after the particle, labelled CP in (50). In the speciWer position there is a null TopicP licensed by the particle e. This null Topic must be interpreted and recovers its features from the preceding discourse: the sentence is understood as having the same topic as the preceding one. Therefore, the fact that e can function as a Topic marker depends on its original formal property of taking a speciWer and a complement which must be of the same type, not only on its semantics or on some other independent mechanism which ‘creates’ new particles. In other words, the fact that e can be used as a Topic marker depends, on the one hand, on the formal properties of e, and, on the other, on general features of the language. In the case under

Sı` and e as CP Expletives

233

discussion, the relevant point is the fact that Old Italian was a V2 language in which the CP layer was always active. An independent argument in favour of the idea that null Topics exist in Old Italian is provided by cases like the following ones, whose interpretation is diVerent from the cases licensed by e: (51)

Uno cavaliere pregava un giorno una donna d’ amore e A knight was praying a day a woman of love and diceale intra l’ altre parole com’ elli era gentile e ricco e told her among other things how he was kind and rich and bello a dismisura, ‘e ’l vostro marito e` cosı` laido come voi very handsome, ‘and your husband is so ugly as you sapete’; e quel cotal marito era dopo la parete della know’; and precisely that husband was behind the wall of the camera. room Ø Parlo` e disse: ‘Eh, messer, per cortesia: He spoke and said: ‘Eh, sir, please, acconciate li fatti vostri e non isconciate li altrui’. mind your own business and do not spoil the others’ (Nov., 231)

(52) ‘Iscrivi’ disse quel re cortese ‘ch’ io obligo l’ anima write said that king kind that I oblige the soul mia a perpetua pregione inWno a tanto che voi pagati siate’. mine to eternal prison until that you paid are Ø Morı`o. Questi, dopo la morte, andaro al padre died they after the death went to.the father suo e domandaro la moneta his and asked for the money (Nov., 171) The examples in (51) and (52) have a null category (represented with Ø) at the very beginning of the sentence, immediately followed by the inXected verb. Cases like these are completely ungrammatical in Modern Italian, where either the subject has to be repeated, or the subject can be null but there must be another element (preferably a temporal adverb) before the verb. Even more, to a native speaker of Modern Italian these cases seem rather opaque. (51) for instance is hard to understand if the whole context is not attentively evaluated. These kinds of cases strongly recall the phenomenon of Topic drop

234

C. Poletto

in Germanic languages (like German), or at least one type of Topic drop typically involving the subject as discussed by Cardinaletti (1996). As for the status of this entity, in the literature on null elements there is a category that has precisely the properties of the null Topic licensed by e: the null constant proposed by Rizzi (1992) for the pro-drop phenomenon of child language. This category can only occur in Wrst position, its content is retrieved from the context, and it is never c-commanded by anything. The same is true for cases like (50): the null Topic occurs at the very beginning of a sentence, its content is retrieved from the context, and, although it occurs in embedded domains, the clause containing it is not a complement clause. All the examples examined correspond to temporal clauses or to gerundive or inWnitival clauses that express contemporary events. If the temporal clause is placed before the main clause, in an anti-symmetric approach like the one adopted here, the null constant is clearly not c-commanded by anything. This would apply also to constructions where the non-Wnite verbal form follows the main clause. In these cases the inWnitival or gerundive form is probably not c-commanded by elements internal to the main clause. At this point a cautionary note is in order: before we can decide whether the null Topic in structures like (50) is really Rizzi’s null constant, we should know more about the way these sequences of clauses, which are not connected through selection but simply express contemporary events, are layered, a problem that cannot be solved here. However, for now the null constant is the best candidate to account for the distribution of null Topics in Old Italian.

13.6 Conclusion In this chapter I have compared the left periphery of Old Italian with the left periphery of Modern Rhaetoromance and proposed that V2 is not a unitary phenomenon in the sense that the projection targeted by V2 can change. More precisely, V2 can target Focus or Force: in the Wrst case declarative clauses admit V3 and V4 with left dislocated constituents, in the latter case only interrogatives can have V3 or V4. On the basis of Beninca`’s idea that Old Italian V2 is quite low in the structure of the CP area, I have shown that the adverb sı`, contrary to cosı`, is an expletive of the Focus projection because it can be directly merged inside the CP layer. Such hypothesis derives the distribution of sı` in Old Italian on the basis of its semantic and formal properties. Another particle merged in the left periphery of the clause is e, which marks the continuation of a Topic by means of a null Topic in its speciWer and the whole clause (headed by a TopicP) in its complement position. This Topic

Sı` and e as CP Expletives

235

marker has been lost in Modern Italian declaratives, although some restricted usages have been maintained when the CP layer is already independently active (i.e. interrogatives). Rizzi’s null constant can account for the properties of the null Topic licensed by e, provided the sequences of clauses illustrated in the examples are not cases of complementation. While the particle sı` has been totally lost (and this loss probably also caused the disappearance of the corresponding adverbial), e has not only retained its meaning as a conjunction, but also its usage as a Topic marker when the appropriate structural conditions are met (namely the presence in the left periphery of a Focus layer containing some lexical material). This leads to the general conclusion that once the formal properties of an element meet the appropriate structural conditions (in our case V2), the reanalysis of that element as a Focus or Topic marker is straightforward. Primary Sources Alighieri, D. Vita Nuova, ed. M. Barbi (1932). Florence: Bemporad. [Abbr. Vita nuova] Anon. Fiori e vita di WlosaW e d’altri savi e d’imperatori, ed. A. D’Agostino (1979). Florence: La Nuova Italia. [Abbr. FF] —— Novellino, ed. G. Favati (1970). Genova: Bozzi. [Abbr. Nov] —— Testi Fiorentini del Dugento e dei primi del Trecento, ed. A. SchiaYni (1926). Florence: Sansoni. [Abbr. Doc. Fior] Giamboni, B. Delle Storie contra I Pagani di Paolo Orosio. Libri VII, ed. F. Fontani (8th edn.) (1885). Florence: Marenigh. [Abbr. Orosio] —— Fiore di rettorica, ed. G. B. Speroni (1994). Pavia: Dipartimento di Scienza della Letteratura e dell’Arte medioevale e moderna. [Abbr. Fiore di rett] Latini, B. La Rettorica, ed. F. Maggini (1968). Florence: Le Monnier. [Abbr. Rettorica] —— Tesoretto. In G. Contini (1960), Poeti del Duecento. Milano–Napoli: Ricciardi, 175–277. [Abbr. Tesoretto]

14 Split wh-Constructions in Classical and Modern Greek: A Diachronic Perspective ERIC MATHIEU AND IOANNA SITARIDOU

14.1 Introduction The present chapter is about split wh-constructions and provides an answer to the following two questions: (i) why split wh-constructions were widespread in Classical Greek, but are now impossible in Modern Greek; (ii) why splitting of tinos and pianu, both meaning ‘whose’, are exceptions to the rule against wh-splitting in Modern Greek. This study is part of a larger investigation on split constructions (Mathieu and Sitaridou 2002, 2003; Butler and Mathieu 2004). It can be seen as the companion paper of Mathieu and Sitaridou (2004), where we deal with adjectival split constructions and give ample details about the synchronic issues related to DP splitting. In this contribution, we concentrate on wh-split and on the diachronic issue. Section 14.2 introduces some basic facts about hyperbaton in Classical and Modern Greek and outlines the diachronic puzzle. Section 14.3 presents our account of the licensing mechanism behind DP splitting. Section 14.4 provides a diachronic account of the loss of split wh-constructions in Modern Greek. Section 14.5 discusses the question as to why splitting of tinos and pianu is possible. We conclude in Section 14.6.

We would like to thank Artemis Alexiadou, Joa˜o Costa, Þo´rhallur Eyþo´rsson, Giuliana Giusti, Anthony Kroch, Ian Roberts, Hanneke Van Hoof, and Roberto Zamparelli for their questions and remarks. Special thanks to Anna Roussou for comments on an earlier version. We also wish to thank Me´lanie Jouitteau, whose help with the Wnal stage of the chapter has been very precious. Thanks also for their help with the Greek data to: Christos Nifadopoulos, Phoevos Panagiotidis, Lukas Pietsch, Athina Sioupi, Tasos Tsangalidis, Stavroula Tsiplakou, and Stela Vergi. Usual disclaimers apply.

Split wh-Constructions in Greek

237

14.2 Hyperbaton in Classical and Modern Greek: The Diachronic Puzzle Classical Greek consistently allowed a process by which continuous strings became discontinuous. Traditionally labelled ‘hyperbaton’, this operation typically splits XPs otherwise treated as a unit with respect to Case theory and u-theory. The Wrst observation is that in Classical Greek, a wh-element need not raise together with its associated nominal. Whereas in (1a) both tina (the accusative form of tis ‘who/what’) and dynamin ‘power’ raise to the sentence initial position; in the hyperbaton case (1b), only tina raises, stranding the nominal:1 (1) a. Tina dynamin echei? what.acc.fem.sg power.acc.fem.sg have.3sg ‘What power does it have?’ (Plato, Laws, 643a) b. Tinai echei ti dynamin? (Plato, Republic, 358b; in Devine and Stephens 2000: 5)

(CG)

In Classical Greek, wh-words and indeWnites had the same morphological make-up, i.e. tis. In interrogatives, the indeWnite was stressed and obligatorily raised to a clause initial position. Otherwise they were interpreted as simple indeWnites with the interpretation some-x or any-x (Roussou 1998; Roberts and Roussou 2003). On the other hand, Modern Greek has a distinct class of wh-words, polarity items, and existential quantiWers, and uses the pios/pia/pio elements for whstructures.2 However, the two Classical Greek forms ti ‘which/what’ and tinos ‘whose’ have survived. Note that in Modern Greek, tinos is the only remaining Case-marked form within the paradigm ti, which otherwise is no longer declinable. 1 The original transliteration of Classical Greek examples as quoted by the authors has not always been preserved. This is for reasons which have to do with the harmonization of the transliteration systems across diVerent authors. 2 Roberts and Roussou (1999) argue that the change from the general use of indeWnites to a distinct class of wh-phrases was triggered by a number of factors. First, on the existential/polarity interpretation, the Classical Greek indeWnites started being replaced productively by the numerical heis, mian, hen ‘one’. Second, the negative quantiWer oudhen ‘nothing’, which was part of a distinct paradigm of negative quantiWers, started replacing the sentential negator ou, contributing to the loss of no-words and the consequent development of any-words (e.g. polarity item tipota ‘anything’). Roberts and Roussou’s (1999) proposal is that when the indeWnite is systematically found in a wh-dependency, tis is reanalysed by the language learner as associated with the head of the dependency, i.e. as a designated wh-word. This follows from the Lexical Subset Principle: Interpret lexical items as being susceptible of occurrence in the smallest set of contexts consistent with the input.

E. Mathieu and I. Sitaridou

238 Table 14.1

wh-elements in Classical Greek Singular

Plural

Case

Masc./Fem.

Neuter

Case

Masc./Fem.

Neuter

Nom. Gen. Dat. Acc.

tiB (tis) tinoB (tinos) tini (tini) tina (tina)

ti (ti) tinoB (tinos) tini (tini) ti (ti)

Nom. Gen. Dat. Acc.

tineB (tines) tinvn (tinon) tisi(n) (tisi) tinaB (tinas)

tina (tina) tinvn (tinon) tisi(n) (tisi) tina (tina)

Table 14.2

wh-elements in Modern Greek Singular

Case

Masculine

Feminine

Neuter

All genders

Nom. Gen. Acc.

pios piu/pianu pion

pia pias/pianis pia(n)

pio piu/pianu pio

ti tinos —

pia pion/pianon pia

— (tinon) —

Plural Nom. Gen. Acc.

pii pion/pianon pious/pianus

pies pion/pianon pies/pianes

The second observation is that, although ti could be separated from its associated nominal in Classical Greek, at present it is no longer an alternative, as shown by (2b). (2)

a. Ti dinami exi? what power.acc.fem.sg have.3sg ‘What power does he/she/it have?’ b. *Tii exi ti dinami?

(MG)

The third observation is that the genitive Case-marked form tinos ‘whose’ continues to allow splitting in Modern Greek (a fact originally discussed by Horrocks and Stavrou 1987). It must be noted, however, that sentences such as (3b) are not accepted by all speakers. In Section 14.5, an attempt will be made to account for the variation in those dialects/registers. (3)

a. Tinos to vivlio eferes? (MG) whose.gen.sg the.acc.neut.sg book.acc.neut.sg bring.past.2sg ‘Whose book did you bring?’ b. Tinosi eferes ti to vivlio? (Horrocks and Stavrou 1987: 89)

Split wh-Constructions in Greek

239

The fourth fact with which the present contribution is concerned is that, somehow surprisingly (compare (4b) with (5b)), Modern Greek pianu ‘whose’ does behave like tinos in that it permits splitting (again, in some but not all dialects/registers). Obviously, what tinos and pianu have in common is the fact that they are both genitive. Thus, the possibility of splitting from their associated nominals must stem from that fact (see Section 14.5 for details). (4)

a. Pianu to vivlio eferes? (MG) whose.gen.masc.sg the.acc.neut.sg book.acc.neut.sg bring.past. 2sg ‘Whose book did you bring?’ b. Pianui eferes ti to vivlio?

On the other hand, the behaviour of the rest of the pios ‘who/which’ paradigm is consistent with the observation that split wh-constructions are no longer available in Modern Greek (cf. (2b)). This is exempliWed in (5b). (5) a. Pia dinami exi? which.acc.fem.sg power.acc.fem.sg have.3sg ‘Which power does it have?’ b. *Piai exi ti dinami?

(MG)

While extraction of wh-elements other than genitive interrogatives is not permitted in Modern Greek, a whole series of adjectives, quantiWers, and demonstratives can be extracted without the need to pied-pipe the nominal. In each case, the dislocated element bears stress and is contrastively focused. Beginning with adjectives, it is clear from the context in which they appear that such discontinuous structures in Classical Greek involved a set of alternates, i.e. involved strong focus. This is well documented in Devine and Stephens (2000). In (6b) the interpretation is: ‘private, not public’ (stress/ emphasis is indicated by capital letters). (6)

a. Kateskeuakasin oikias tas (CG) build.pluperfect.3pl house.acc.fem.pl the.acc.fem.pl idias private.acc.fem.pl ‘They have built their private homes.’ b. tas idiasi kateskeuakasin ti oikias ‘They have built their private homes (i.e. not their public ones).’ (Demosthenes 23.208; in Devine and Stephens 2000: 236)

Turning now to Modern Greek, the same pattern is found. Whereas in (7a) the adjective kokino ‘red’ is adjacent to the noun it modiWes, i.e. forema ‘dress’, in (7b) the adjective has been extracted on its own. If the condition that the

240

E. Mathieu and I. Sitaridou

extracted element must receive stress and is interpreted contrastively is not met, the sentence is ungrammatical, as illustrated by (7c): (7)

a. Agorase to kokino forema (MG) buy.past.3sg the.acc.neut.sg red.acc.neut.sg dress.acc.neut.sg ‘She bought the red dress.’ b. to kokino agorase forema ‘She bought the RED dress (i.e. not the blue one).’ c. *To kokino agorase forema

The type of example shown in (7b) was Wrst introduced by Androutsopoulou (1997: 2). A variant of (7b), originally discussed by Horrocks and Stavrou (1987: 91), appears in (8). The sentence in (8) is identical to (7b) apart from the additional determiner present to the left of the stranded nominal; a process dubbed Determiner Spreading (the choice to have an additional determiner appears to be dialectal/a matter of register): (8) To kokino agorase to (MG) the.acc.neut.sg red.acc.neut.sg buy.past.3sg the.acc.neut.sg forema dress.acc.neut.sg ‘She bought the red dress (i.e. not the blue one).’ In Classical Greek, the situation was more straightforward, since in the canonical word order (adjective-noun, as in the modern form of the language), a doubled determiner never appeared in this kind of construction (cf. Devine and Stephens 2000 and Manolessou 2000).3 Next, (9b) and (10b) demonstrate extraction of a quantiWer in Classical and Modern Greek respectively, whereas (11b) and (12b) illustrate extraction of a demonstrative. (9)

a. Eiche elpidas pollas have.past.3sg hope.acc.fem.pl many.acc.fem.pl ‘He had many hopes.’ b. pollasi eiche ti elpidas ‘He had many hopes (i.e. not a few).’ (Herodotus 5.36; in Devine and Stephens 2000: 33)

(10)

a. Ixe poles elpides have.past.3sg many.acc.fem.pl hope.acc.fem.pl ‘He had many hopes.’

(CG)

(MG)

3 In Classical Greek, it appears that Determiner Spreading was only possible when an adjective followed a noun, but the process was optional, and not obligatory like the case of Modern Greek.

Split wh-Constructions in Greek

241

b. polesi ixe ti elpides ‘He had many hopes (i.e. not a few).’ (11) a. Tous chroˆmenous teˆ tauteˆ (CG) the.acc.masc.pl use.part.acc.masc.pl the.dat.fem.sg this. dat.fem.sg ergasia profession.dat.fem.sg ‘Those that exercise this profession.’ b. Tous taut^e i chroˆmenous teˆ ti ergasia ‘Those that exercise this profession (i.e. not that one).’ (Aeschines 1.119; in Devine and Stephens 2000: 33) (12)

a. Ida afto to forema (MG) see.past.1sg this.acc.neut.sg the.acc.neut.sg dress.acc.neut.sg ‘I saw this dress.’ b. aftoi ida ti to forema ‘I saw this dress (i.e. not that one).’

Finally, in (13b) and (14b) extraction of a negative quantiWer is exhibited: (13)

a. Hymas pothein akousai oudemian prophasin (CG) you.acc.pl desire.inf hear.inf.past no.acc.fem.sg excuse. acc.fem.sg ‘That you desire to hear no excuse.’ b. oudemian i hymas pothein akousai ti prophasin ‘That you desire to hear no excuse.’ (Lysias 14.1; in Devine and Stephens 2000: 33)

(14) a. Den thelo na akuso kamia dikeologia (MG) not want.1sg prt.subj listen.1sg no.acc.fem.sg excuse.acc.fem.sg ‘I don’t want to hear any excuse.’ b. kamiai den thelo na akuso ti dikeologia ‘I want to hear no excuse.’ In short, Modern Greek is not wholly consistent with respect to whether it allows split-DPs. Some DPs can split (tinos, pianu, adjectives, indeWnites, and negative elements), while others cannot (ti, pios). There thus cannot be such thing as a split-DP versus non split-DP language. Cross-linguistically, diVerent combinations arise. For example, bare combien extraction is possible in French, as shown by (15b) (Obenauer 1976, 1983, 1994), whereas the equivalent construction is not possible in Modern Greek (16b). On the other hand,

E. Mathieu and I. Sitaridou

242

whereas bare demonstrative extraction is available in Modern Greek (12b), it is impossible in French (17b). (15) a. Combien de livres as-tu lus? how many of books have.you read.masc.pl ‘How many books have you read?’ b. Combien as-tu lus de livres?

(French)

(16)

a. Posa vivlia diavases? (MG) how many.acc.neut.pl book.acc.neut.pl read.past.2sg ‘How many books did you read?’ b. *Posa diavases vivlia? (17) a. J’ai achete´ cette robe (French) I.have bought this.fem.sg dress.fem.sg ‘I have bought this dress.’ b. *cette j’ai achete´ de robe 4 this.fem.sg I.have bought of dress.fem.sg ‘I have bought this dress.’ The complexity of the data and the intra-language idiosyncracies in extraction patterns might therefore be seen as a challenge for parametric accounts of split-DPs: a single parameter (e.g. + extraction) cannot capture the existing variability. However, although it is true that diYculties may arise under a Government and Binding view of parametrization, the Minimalist Programme (cf. Chomsky 1995 and subsequent papers) provides us with a theory of parametrization which can account for the apparent wild variation. According to minimalism, variation between languages is restricted exclusively to diVerences in the lexicon. Languages may thus not be wholly consistent with respect to a particular parametric choice: one lexical item may choose one value; a second lexical item may choose a diVerent value. This means that a binary account is not completely out of reach after all (for similar arguments, see Gavruseva 2000, as well as Grosu 1974).

14.3 The Mechanics of Split-Constructions Before we provide a diachronic account for the loss of split wh-constructions in Modern Greek, we Wrst summarize our synchronic approach to the phenomenon. In doing so, we follow essentially Mathieu and Sitaridou (2004). For DP-splitting to be possible the following condition is essential: rich agreement (where agreement is part of a collection of f-features that includes 4

The presence of de ‘of’ is usually required in split constructions in French.

Split wh-Constructions in Greek

243

Case) on the extracted element. From this, other conditions follow, for example the availability of adjectives accompanied by null heads, and the possibility of so-called ‘Determiner Spreading’. In all the examples of the previous section, the extracted element bears rich agreement. Note, however, that in (17b) the extracted element has no overt case marking, which may explain why it cannot split (but see Section 14.4). Following and adapting a series of unrelated proposals (Fanselow 1988; Androutsopoulou 1997; van Geenhoven 1998; den Dikken 1998; Devine and Stephens 2000), we put forward the hypothesis that in hyperbaton what raises is not a head or a left branch, but a constituent; namely, a D’ (thus the Left Branch Condition is not violated). We adopt Androutsopoulou’s (1997) basic insight that DP splitting is possible in Modern Greek because null head adjectives/DP fragments are available independently (in Mathieu and Sitaridou 2004, we add to this generalization the connected fact that Determiner Spreading is also possible because of the availability of noun ellipsis; see below). In English very few null head modiWers/DP fragments are possible, which in our opinion explains why DP splitting is virtually impossible. Whereas (18a) is possible in Modern Greek, (18b) is not possible in English (one support is needed). There are apparent exceptions, but see Mathieu and Sitaridou (2004) for details. (18)

Speaker A Which dress did you buy? Speaker B a. to kokino the.acc.neut.sg red.acc.neut.sg b. *the red

No agreement marking is present on red in English, thus noun ellipsis is not possible (the empty nominal cannot be licensed). Consequently, split constructions are not possible in that language, since a prerequisite for their availability is an XP with an empty noun.5 This is equivalent to Devine and Stephens’s (2000) notion of ‘null head modiWer’, i.e. a modiWer that can stand by itself in place of a noun phrase without the support of a noun or an overt pronoun. The idea that it is a constituent that moves rather than a part of a constituent is most obvious in the case of bare adjectival extraction, since not only the adjective raises, but so does the determiner (cf. (7b) and (8)). We follow den Dikken’s (1998) Small Clause (SC) analysis of subjectpredicate DP constructions. The subject of the SC is a bare noun. Technically it is not in fact an argument, but starts out as a predicate (for full details about 5 On noun ellipsis in Romance and other languages, see Lobeck (1995), Kester (1996), and Sleeman (1996).

244

E. Mathieu and I. Sitaridou

the predicative nature of the stranded nominal, see Mathieu 2002, 2004; Mathieu and Sitaridou 2004). The predicate of the SC is an adjective. The predicate undergoes predicate inversion. Motivation for such a structure comes from the fact that splitting is possible only with predicative adjectives. Non-predicative adjectives cannot be split, as shown by (19):6 (19)

*ton ipotithemenoi ida ti dolofono the.acc.masc.sg alleged.acc.masc.sg see.past.1sg murderer. acc.masc.sg ‘I saw the alleged murderer (i.e. not the real one).’

(MG)

Besides, only predicative adjectives can be used as fragments, as (20a) shows: (20) Speaker A Which criminal did you arrest? Speaker B a. *ton ipotithemeno the.acc.masc.sg alleged.acc.neut.sg b. *the alleged. The technical implementation of our proposal is as follows. In (21a) the subject of the SC is forema ‘dress’, while kokino ‘red’ is the predicate. In those dialects/registers where to does not surface, we assume that the head of the SC is simply not spelled out. Across dialects/registers, the nominal copula is not spelled out either in Greek. On the other hand, in languages like French the nominal copula surfaces as de (den Dikken 1998). We also give the structure for the Classical Greek example (1b) in (21b). In this case, tina is the predicate while dynamin is the subject of the SC. AgrP is the equivalent of den Dikken’s (1998) FP; it is an agreement projection.7

6 Androutsopoulou (1997) introduces a similar example (i). Interestingly, other languages (e.g. Polish) have the same constraint on splitting: non-predicative adjectives cannot split (Nowak 2000; Partee 2002).

(i) a. Miso tin kathari adhikia. injustice hate-1sg the clear ‘I hate pure justice.’ b. *Tin kathari miso adhikia. ‘I hate PURE justice.’ 7

In so-called ‘N of N’ constructions, the nominal copula de surfaces in French. However, the head of the Small Clause is not spelled out (i). In English, both are available (translation of (i) and (ii)). On the other hand, as shown by (ii) in Modern Greek, although the head of the SC is spelled out (as o, i.e. before jatros), the nominal copula is not morphologically marked (there is no element between vlakas and the second occurrence of o). According to den Dikken (1998) and Moro (1997), a copula is obligatory when predicate inversion arises. However, this does not appear to be the case in Greek. We discuss facts about the nominal and the clausal copula and its non-obligatoriness in Greek in Mathieu and Sitaridou (2004).

Split wh-Constructions in Greek (21)

245

a. [DP D to[SC Spec-SC [NP forema] SC8 to [AP kokino]]] b. [DP D Ø[SC Spec-SC [NP dynamin] SC8 Ø [AP tina]]]

The predicate undergoes inversion as shown in (22a) and (22b). Adjectival agreement in Modern Greek is rich and therefore the movement is triggered by an EPP feature on Agr. (22)

a. [DP to [AgrP Spec, Agr [AP kokino]i [Agr ∅ + SC toj [SC Spec, SC [NP forema] [ tj ti]]]]] b. [DP ∅ [AgrP Spec, Agr [AP tina]i [Agr ∅ + SC ∅j [SC Spec, SC [NP dynamin] [ tj ti]]]]] c. [Spec,Top [Agr
[Agr ∅ + SC toj [SC [NP forema] [tj ti]]]]]k [D
D to [AgrP [AP kokino]i tk]] d. [Spec,Top [Agr
[Agr ∅ + SCj [SC [NP dynamin] [tj ti]]]]]k [D
D ∅ [AgrP [AP tina]i tk]] e. [Spec,Top [Agr
[Agr ∅ + SC toj [SC [NP forema] [tj ti]]]]]k [Spec,Foc [D
D to [AgrP [AP kokino]i tk]]]m tm f. [Spec,Top [Agr
[Agr ∅ + SCj [SC [NP dynamin] [tj ti]]]]]k [Spec,Foc [D
D ∅ [AgrP [AP tina]i tk]]]m tm

Next, Agr’ and the material contained in it raises to the speciWer of a DP internal topic position (we thus propose to split the DP domain on a par with the split of the CP domain; cf. Rizzi 1997). Thus, like many others we take this speciWer to be an A’-position in Modern Greek (for DP-internal focus and topics in Greek, see Horrocks and Stavrou 1987; and for other languages, Cinque 1980; Szabolcsi 1983; Stowell 1991; and Giorgi and Longobardi 1991). The conWguration obtained is shown in (22c) and (22d). Then, the D’ complex raises to the speciWer of a DP internal focus position (that occurs lower than TopP, but higher than D). The conWguration is shown in (22e) and (22f). Without any further movement (22e) is basically the conWguration for (i) Cet idiot de docteur! that idiot of doctor (ii) Aftos o vlakas o jatros! that a idiot a doctor ‘That idiot of a doctor!’ French is consistent in that respect in that the nominal copula also surfaces in bare adjectival constructions. (iii) Le grand, je veux de cafe´ latte! the tall I want of cafe´ latte ‘I want the large cafe´ latte (i.e. not the small one).’

246

E. Mathieu and I. Sitaridou

so-called ‘Determiner Spreading’ structures. Movement of the Agr’ complex is forced by an EPP/strong D feature present on D. This proposal is a variant of what Alexiadou and Wilder (1998) have proposed for predicative adjectives. The diVerence between our account and theirs is that the Wrst to is the spellout of the head of an SC; only the to associated with the adjective is a determiner. Determiner Spreading is thus a misnomer and does not exist (at least not in Greek). It must be noted that after movement of the Agr’ complex to Spec,TopP, a trace is left behind. We take this trace to be the equivalent of Devine and Stephens’s null head and Androutsopoulou’s empty nominal category. After movement of the nominal to Spec,TopP, the D’ complex that has itself raised higher in the DP to the speciWer of a Foc projection can raise higher up in the clause. Movement is thus movement of a determiner together with an adjective as well as an empty category standing for a nominal. (23)

a. [CP [D
to KOKINO tk]i b. [CP [D
∅

agorase [spec,TopP to foremak [Spec,FocP ti ]]]

tina tk]i echei [spec,TopP ∅ dynamink [Spec,FocP ti ]]]

To recap, the steps of a split-DP derivation are the following: . The extracted element must involve an adjective, possibly accompanied by a determiner, but the extracted element cannot be a sole determiner. . Overt marking of f-features on the adjective is compulsory so that the empty noun can be licensed. . The adjective undergoes inversion and raises to Spec,AgrP because of an EPP feature on Agr8. . The nominal raises to Spec,TopP and the predicate raises to Spec,FocP. . The predicate constituent now containing the trace of forema, which contains a trace of the predicate, further raises to the C domain.

14.4 The Loss of wh-Split in Modern Greek In this section, we show how wh-splitting ceased to be operative in Modern Greek. It is argued that this is because rich agreement has been lost. For example, ti in Modern Greek has one form for all genders whereas in Classical Greek it had diVerent forms for diVerent genders, and it carried Case. However, it must be noted that rich agreement is not a suYcient condition. The

Split wh-Constructions in Greek

247

relevance of the pios series should be obvious here (full declination is available in this case). We thus propose that the category of the element on its own crucially depends on the categorial status of that element. The idea is that, as a result of the reanalysis of tis and the subsequent creation of distinct classes, ‘relabelling’ (in the sense of Whitman 2000) from a modiWer to a determiner of the wh-element took place. Determiners may carry overt agreement features (the case of French ce/cet/cette, and the case of pios), but because they are determiners, they cannot split. A crucial test for the determiner or adjective status of a given element is that of combinatorial possibilities with another determiner. For example, in French the string *ce le ‘this the’ is not possible, suggesting that ce is a determiner. Similarly, possessives like mon/ma/mes ‘my’, although rich in agreement, are determiners: unlike Italian, for example, they cannot appear together with a determiner (French *le mon livre ‘my book’ versus Italian il mio libro ‘my book’). Those possessives that can appear with a determiner are adjectives: le tien/le mien/le sien, etc. ‘yours/mine/his, etc.’ In English, we know that wh-elements like what are determiners because they cannot appear together with words like the: *what the. To resume, Modern Greek wh-elements are no longer adjectives, but determiners. Our proposal is consistent with Roberts and Roussou’s (1999) analysis of tis in Classical Greek. As pointed out in Section 14.2, there is a lot of evidence that tis was a simple indeWnite with no quantiWcational properties of its own in Classical Greek. We simply follow traditional assumptions according to which tis, although an indeWnite, was also an adjective in the string tina dynamin (cf. (1)), and (even more clearly) in the string tina teˆn dynamin (cf. (24)). Evidence for the hypothesis that relabelling took place comes from the grammaticality of Classical Greek (24), where a determiner accompanies the wh-element, and the ungrammaticality of Modern Greek (25), where the juxtaposition of a wh-element with a determiner is not possible: (24)

teˆn dynamin eichon Tina what.acc.fem.sg the.acc.fem.sg power.acc.fem.sg have.past.3pl ouk egnoˆn (CG) not know.past.3pl ‘They didn’t know what power they had.’

(25) a. *Pia tin dinami exi? (MG) what.acc.fem.sg the.acc.fem.sg power.acc.fem.sg have.3sg ‘What power does it have?’ b. *Ti to gliko thelis? what the.acc.neut.sg sweet.acc.neut.sg want.2sg ‘What kind of sweet do you want?’

248

E. Mathieu and I. Sitaridou

Further evidence that ti is a determiner in Modern Greek comes from the fact that it can appear together with an extracted bare adjective (very much as in the case of (7b) and (8)). The complex ti þ other thus forms an ellipsed structure. (26) is a genuine example spontaneously produced by an informant. (26)

thelis ja ta xristujena Ti alai what other.acc.neut.pl want.2sg for the Christmas glika? sweet.acc.neut.pl ‘What other sweets do you want for Christmas?’

ti (MG)

Finally, quantiWers and demonstratives are often considered modiWers in Classical and Modern Greek, so they are on a par with adjectives. To conclude this section, there is independent evidence that determiners originate from adjectives (Greenberg 1978; Haspelmath 1995). Crosslinguistically it is common to Wnd indeWnite articles originating from numeral adjectives (French: un ‘one’ > un ‘a’). More generally, it is traditionally assumed that lexical items grammaticalize and become functional via a process of reanalysis. For example, serial verbs change into prepositional phrases (Whitman 2000) and nominals into negative markers (Roberts and Roussou 1999).

14.5 The Case of wh-Possessor Raising In this section, it is argued that the mechanism behind possessor extraction is diVerent from the one involved in other (non-possessor) split wh-constructions. A language may allow genitive wh-phrases to split across the board without necessarily allowing any other types of wh-elements to undergo bare movement without pied-piping of the nominal. Modern Greek tinos (like Modern Greek ti) and pianu are determiners, and not adjectives. In addition, whereas splitting of a non-genitive wh-phrase very much depends on the categorial status of the extracted element (i.e. whether it is an adjective or a determiner), splitting of a genitive wh-phrase does not depend on this condition. It is a single process that a language may or may not have at its disposal. Moreover, no predicate raising is necessary in that case. To account for wh-possessor extraction, we essentially follow Horrocks and Stavrou’s (1987) analysis (see also Mouma 1993 and Gavruseva 2000), according to which genitive wh-phrases can extract because N assigns Case to its complement. The wh-possessor thus does not need to raise to an agreement position (as in the case of non-wh-possessor split elements), since it is already

Split wh-Constructions in Greek

249

Case-marked. No predicate raising occurs, and no DP-internal movement to Spec,TopP and Spec,FocP is necessary. Spec,DP is available as an escape hatch as the default case, D8, is associated with an EPP feature. The possessor can thus move higher in the clause.8 However, in some registers it is impossible to split tinos and pianu. In order to account for this, we argue that in this case there is no EPP feature associated with D. We would like to suggest that the trigger for the loss of the feature is the potential ambiguity between a genitive and a dative reading of tinos and pianu in Modern Greek. (3b) and (4b) can also mean: ‘To whom did you bring the books?’ When a verb like ferno ‘to bring’ or other ditransitive verbs are used, the direct object is Case-marked with accusative and the indirect object is Case-marked with genitive or prepositional accusative (with diVerent word order restrictions in each case). In interrogative contexts, the indirect object is replaced by tinos, since the latter—albeit in genitive—is the Case form that replaces the indirect object.9 In addition, for another group of speakers tinos (and by analogy pianu) has an archaic Xavour and marks ‘high’ registers on the split interpretation; and its use as a split element is disappearing completely. For others it is used as a learned construction (in the sense of Pountain 1998; see also Sitaridou 2002), and it coexists with its non-split counterpart. The case of tinos could be captured in terms of ‘competing grammars’, in the sense of Kroch (1989), namely that the individual is dealing with two grammars. However, it is not clear at all whether learned constructions highlight this mental state or rather the one whereby there is one grammar with diVerent options activated upon diVerent conditions on use (for more details see Sitaridou 2005).

14.6 Conclusion The idea that we put forward in this chapter is that wh-splitting ceased to be operational in Modern Greek because agreement on wh-elements is not

8

As pointed out by Horrocks and Stavrou (1987), in a rhetorical in situ interrogative tinos can appear post-nominally. Presumably, this is possible because tinos gets case from the noun. (i) Eferes to vivlio tinos? bring.past.2sg the.acc.neut.sg book.acc.neut.sg whose.gen.sg ‘You have brought whose book?’ 9

This happens with all clitics that replace indirect objects in Modern Greek. Although they replace dative arguments their morphological form is the genitive. This is due to the loss of dative case, an innovation of Modern Greek discussed extensively in the literature.

250

E. Mathieu and I. Sitaridou

uniformally rich; consequently ti is no longer an adjective in Modern Greek, but a determiner. The fact that splitting is possible with tinos and pianu for some speakers is due to the fact that it involves a diVerent process, namely one that does not involve DP-internal predicate inversion. Finally, it was argued that for those speakers who do not accept splitting with tinos and pianu, no EPP feature is associated with D, and thus no escape hatch (i.e. Spec,DP) is available. This is because tinos and pianu are being reinterpreted solely as dative arguments in constructions involving dislocation of such lexical items.

15 The Syntax of Objects in Old English SUSAN PINTZUK

15.1 Introduction In this chapter, I present evidence against three commonly held views of Old English syntax and verb-object order: (1) the increase in verb-object order during the Old English period can be attributed to a gradual increase in the frequency of extraposition, an increase which is not necessarily tied to internal grammatical factors; (2) all object types (negative, quantiWed, and nonnegative non-quantiWed) have the same syntax in Old English; (3) Old English can be analysed as a head-initial language, with head-Wnal order derived by the leftward movement of various constituents. The arguments against these views all rely on quantitative evidence. In the discussion that follows, I will make a more general methodological point: in diachronic syntax studies, much of the historical data are structurally ambiguous and open to a number of diVerent interpretations. It is therefore necessary to analyse in detail the patterns of variation in order to determine the structural changes that underlie them. Old English, for example, exhibits variation in the linear order of verbs and their objects and in the order of Wnite and non-Wnite verbs in both main and subordinate clauses, as illustrated in (1) through (4), where Aux is the Wnite auxiliary, V is the non-Wnite main verb, and O is the nominal object: Some of the material in this chapter has been presented before audiences at the 7th Diachronic Generative Syntax Conference (University of Girona, 2002), the 2nd York-Holland Symposium on the History of English Syntax (University of Leiden, 2003), and NWAVE 32 (University of Pennsylvania, 2003). I thank these audiences for helpful comments and discussion. All errors and omissions remain my own responsibility. This work has been supported by AHRB Research Leave Grant 4727 APN 13199 and a matching grant from the Department of Language and Linguistic Science, University of York; this support is gratefully acknowledged. Finally, thanks are due to Beth Randall for CorpusSearch, the software that makes it possible to search large annotated corpora.

252

S. Pintzuk

(1) Aux O V þæt hi mihton heora fynd oferwinnan so-that they could their foes overcome ‘. . . so that they could overcome their foes.’ (cobede,Bede-1: 9.44.11.372) (2)

Aux V O þæt he mot ehtan godra manna that he might persecute good men ‘. . . that he might persecute good men . . .’ (cowulf,WHom-4: 37.122)

(3)

O V Aux hu he his agene unðeawas ongietan wille how he his own faults perceive will ‘. . . how he will perceive his own faults . . .’ (cocura,CP:0.23.21.78)

(4) V Aux O þæt ænig mon atellan mæge ealne þone demm that any man relate can all the misery ‘. . . that any man can relate all the misery . . .’ (coorosiu,Or-2:8.52.6.998) There are at least four analyses available in principle to account for these data: head-Wnal structure with rightward movement, head-initial structure with leftward movement, free word order, and grammatical competition between head-initial and head-Wnal structure. It will be demonstrated, however, that only the last analysis, grammatical competition, accounts for all of the patterns of variation that exist in the data. I make the following two assumptions about Old English syntax. First, verb seconding in Old English is movement of the Wnite verb to a position lower than C, call it I. Verb movement to I occurs in both main and subordinate clauses; verb movement to C occurs only in some exceptional main clause types. This analysis of verb seconding in Old English is by now uncontroversial (van Kemenade 1997; Kroch and Taylor 1997; Pintzuk 1999). Second, projections in Old English, in particular IP and VP, can vary in headedness. Since verbal inXection is strong in Old English, movement of the Wnite verb to I is obligatory, and the headedness of IP is indicated by the position of the Wnite auxiliary verb. For example, (1) is an I-initial clause with the object before the non-Wnite main verb, (4) is an I-Wnal clause with the object after the non-Wnite main verb. Since both IP and VP can vary in headedness, four

The Syntax of Objects in Old English

253

possible structures can be generated, as shown in (5). To simplify the structures and the exposition, it will be assumed here that both I and V take VP complements, with no functional projections between IP and VP. (5) a. head-initial IP, head-Wnal VP1 and VP2, deriving (1) Aux O V [IP AuxiþI [VP1 [VP2 OV] ti ]] b. head-initial IP, head-initial VP1 and VP2, deriving (2) Aux V O [IP AuxiþI[VP1ti [VP2 VO]]] c. head-Wnal IP, head-Wnal VP1 and VP2, deriving (3) O V Aux [IP [VP1 [VP2 OV]ti ] AuxiþI] d. head-Wnal IP, head-initial VP1 and VP2, deriving *V O Aux *[IP [VP1 ti [VP2 VO]] Auxi þI] Notice that although there are four possible structures, only three occur with any frequency in Old English.1 I assume that (5d) is in fact ungrammatical, and the fact that it does not occur in Old English must simply be stipulated.2 To this variation in headedness of structure, we must add optional rightward movement rules to derive clauses like (4), with constituents after the V-Aux cluster. The derivation is sketched in (6), with postposition as rightadjunction to IP. (6)

head-Wnal IP, head-Wnal VP1 and VP2, with postposition, deriving (4) V Aux O [IP [IP [VP1 [VP2 tj V] ti ] AuxiþI] Oj ]

For ease of exposition, I assume variation in structure in Sections 15.2 and 15.3, and then present evidence supporting it in Section 15.4. In the analyses and discussion that follows, three types of non-pronominal objects are distinguished: negative, quantiWed, and non-negative non-quantiWed (henceforth ‘positive’). The database used for this study is the York-Toronto-Helsinki Parsed Corpus of Old English Prose, henceforth the YCOE (Taylor et al. 2003). All texts in the YCOE were used except for duplicate manuscripts, almost 1.5 million words from eighty-three Old English texts.

1 Instances of the order in (5d) are quite rare: I have found only three clauses in Old English prose texts in which a constituent appears between the non-Wnite verb and the Wnite auxiliary that is unambiguously in clause-Wnal position. In all three clauses the constituent is an adjunct adverbial or prepositional phrase. 2 Fuss and Trips (2002) and Kiparsky (1996) explain the non-existence of structures like (5d) by proposing that in clauses with head-Wnal VPs, there is no rightward movement of the Wnite verb. Space constraints prevent the discussion of their proposals here.

254

S. Pintzuk

15.2 View 1: The Increase in Verb-Object Order During the Old English Period Can be Attributed to a Gradual Increase in the Frequency of Extraposition An increase in the frequency of extraposition was the traditional explanation of the increase in verb-object order during the Old English period; see the discussions in Lightfoot (1979: 392–5) and Fischer et al. (2000: 173–4). This view can be interpreted in diVerent ways, depending upon the assumptions made about the syntax of Old English. For example, if Old English is uniformly head Wnal (van Kemenade 1987), then all postverbal objects are derived by rightward movement, and an increase in their frequency must be due to an increase in frequency of rightward movement. If Old English is uniformly head initial (Nunes 2002; Roberts 1997; van der WurV 1999), then postverbal objects are derived by the non-application of some type of leftward movement, i.e. postverbal objects are those that remain in their merged position. An increase in their frequency must in this case be due to a decrease in the frequency of leftward movement. In this section I ignore the diVerences between these two analyses, and simply use the term ‘extraposition’ to refer to the syntactic movement or lack thereof that derives postverbal objects. If the hypothesis of an increase in extraposition is correct, then we should see the same increase in VO surface order in all clause types, in particular I-initial and I-Wnal clauses. In addition, we should see the same factors inXuencing the position of objects in I-initial and I-Wnal clauses. However, this prediction is not borne out: the frequency of postverbal objects shows a signiWcant increase in I-initial clauses during the Old English period, but not in I-Wnal clauses. In order to demonstrate this, I brieXy present the results of a multivariate analysis of unambiguously I-initial and I-Wnal clauses with Wnite auxiliaries, non-Wnite main verbs, and positive non-pronominal objects; in other words, clauses like (1) to (4) above.3 The Wve independent factors investigated in the analysis are listed below. Space constraints prevent a full discussion of these factors and their signiWcance and eVect; see Pintzuk and Taylor (2003) for more details. (i) Length of the object, measured in words. The eVect of length, which is signiWcant for all object types and all clause types, may reXect a processing constraint against the centre-embedding of long and complex material. 3 Clauses with Wnite main verbs were not included, since movement of the Wnite main verb to I obscures the underlying order of verbs and their objects. Pronominal objects were not included, since pronouns in Old English exhibit clitic-like behaviour. I-initial clauses with the object before the Wnite auxiliary were not included, because it is not clear that objects in pre-auxiliary position are derived in the same way as objects between the auxiliary and the main verb.

The Syntax of Objects in Old English

255

(ii) Case of the object. DPs in the YCOE are marked accusative, dative, or genitive if case marking is unambiguous on any element of the DP; otherwise, objects are not marked for case. For object types and clause types where this factor is signiWcant, accusative objects favour preverbal position, while dative and genitive objects and objects not marked for case favour postverbal position. It is not clear how to interpret this eVect, but it may be linked to a diVerence in thematic role, which reXects a diVerence in structure in most generative frameworks: accusative objects are usually themes, while dative and genitive objects are usually goals and sources.4 (iii) Clause type. The clauses were divided into four categories: main, conjoined main, subordinate, and conjoined subordinate. Conjoined clauses were considered separately from non-conjoined clauses, since it has frequently been suggested that conjoined main clauses in Old English behave syntactically more like subordinate clauses, while conjoined subordinate clauses without overt complementizers may have the structure of main clauses. For object types and clause types where this factor is signiWcant, main clauses show the highest frequency of postverbal objects, subordinate clauses the lowest frequency, and conjoined clauses fall between these two types. (iv) Date of composition. The texts were divided into two groups: early (before 950) and late (after 950). It would have been desirable to date each of the texts precisely so that date of composition could be used as a continuous rather than a discrete variable. However, the assignment of an exact date to each of the texts was not always possible: estimates of the date of composition for many Old English texts span 20 to 100 years or more, and the midpoint of the interval would not have been a meaningful estimate. The decision was therefore made to simply group the texts into two periods and examine the diVerence between early and late texts. (v) InXuence of Latin original. The inXuence of a Latin source on its Old English translation is relevant to a study of object position, because the unmarked word order in many of the Latin originals is verb Wnal, with objects necessarily preverbal. If the translator followed his source closely, the text might exhibit a higher frequency of OV order than Old English texts not translated from the Latin. The complete results of the multivariate analysis for positive objects are summarized in Table 15.1.

4 It is interesting and clearly signiWcant that the same eVect of thematic role is seen in Middle English (Pintzuk and Taylor 2003), after overt case marking has been lost.

256

S. Pintzuk

Table 15.1 Factors having a signiWcant eVect on the position of positive nominal objects in Old English clauses with auxiliary verbs I-Wnal clauses p p

Length of object Case of object Clause type Date of composition InXuence of Latin

p

I-initial clauses p p p p

Two important conclusions can be drawn from the table. First, only two of the factors inXuencing the position of objects in the two clause types are the same; the other three are diVerent. For both I-Wnal and I-initial clauses, length of object and case of object have a statistically signiWcant eVect, but these are the only factors that the two clause types have in common. I-Wnal clauses show a signiWcantly higher frequency of preverbal objects in texts translated from Latin than in texts not translated from Latin, while I-initial clauses show no such eVect. And the position of objects in I-initial clauses is signiWcantly inXuenced by clause type and date of composition, while these factors are not signiWcant for I-Wnal clauses. These results indicate that postverbal objects are derived diVerently in the two clause types. Second, the inXuence of date of composition is particularly important for the view of Old English that I am arguing against in this section. Date has a statistically signiWcant eVect on the position of the object in I-initial clauses but not in I-Wnal clauses; in other words, there is a signiWcant increase in surface VO in I-initial clauses but no such increase in I-Wnal clauses. If the increase in VO order over the Old English period were due to an increase in extraposition (however extraposition is interpreted), we would expect the eVect of date to be signiWcant for both I-initial and I-Wnal clauses. We can interpret the eVect of date of composition in terms of variation in headedness of structure in the following way. In I-initial clauses, VPs may be either head initial or head Wnal. Therefore postverbal objects in these clauses have two possible derivations, underlying structure and rightward movement, as shown in (7a) and (7b).5 In contrast, postverbal objects in I-Wnal clauses have only one derivation, rightward movement, as shown in (7c). Since the frequency of rightward movement remains constant over the Old English period, the frequency of postverbal objects derived in this way will not increase in either I-initial or I-Wnal clauses; in other words, the movement 5

Traces of the raised Wnite verb are not shown.

The Syntax of Objects in Old English

257

shown in (7b) and (7c) will occur at about the same frequency in all texts during the Old English period. However, I-initial clauses have an additional source of VO surface order: head-initial VP, as in (7a). If the frequency of head-initial VP structure is increasing during the Old English period, then the frequency of postverbal objects will increase in I-initial clauses. The data exhibit exactly these patterns. (7) a. head-initial IP and VP godra manna]] þæt [IP he [I mot ] [VP ehtan that he might persecute good men ‘. . . that he might persecute good men . . .’ (cowulf,WHom-4:37.122) b. head-initial IP, head-Wnal VP þæt [IP he [I mot] [VP ti ehtan ]] [DP godra manna ]i c. head-Wnal IP and VP þæt [IP ænig mon [VP ti atellan ] [I mæge]] [DP ealne þone demm ]i that any man relate can all the misery ‘. . . that any man can relate all the misery . . .’ (coorosiu,Or-2:8.52.6.998) Since the factors signiWcantly inXuencing the position of objects are diVerent in I-initial and I-Wnal clauses, I conclude that diVerent syntactic processes are at work in the two clause types. In addition, since the frequency of postverbal objects does not signiWcantly increase in I-Wnal clauses, I conclude that the increase in VO surface word order during the Old English period cannot be attributed to an increase in the frequency of extraposition, but is due instead to an increase in the frequency of head-initial structure. It should be emphasized that only a detailed statistical analysis of the patterns of variation in the data permit these conclusions to be drawn.

15.3 View 2: Negative, QuantiWed, and Positive Objects All Have the Same Syntax in Old English In a series of articles, Foster and van der WurV (1995, 1997) and van der WurV (1997, 1999) demonstrate that OV order continues at a low frequency throughout the Middle English period, and that preverbal negative and quantiWed objects can still be found in Early Modern English at a time when positive objects are categorically postverbal. Van der WurV (1997, 1999) assumes a uniform head-initial grammar with verb movement to AgrO, and analyses OV order in Old and Middle English as the leftward movement of objects of all types to Spec, AgrOP, as shown in (8):

258

S. Pintzuk

(8) Ac he sceal [AgrOP [þa sacfullan]i gesibbianj þAgrO[VP tj ti ]] But he must the quarrelsome reconcile ‘But he must reconcile the quarrelsome . . .’ (colwstan 1,þALet-2-[Wulfstan-1]: 188.256) Van der WurV (1999) suggests that at the end of the fourteenth century, children reinterpret OV orders as derived by two new rules: raising to Spec,NegP and overt quantiWer raising to the left periphery of VP. Since these rules aVect only negative and quantiWed objects, no positive objects can appear in preverbal position. However, an alternative analysis of these data is possible: not all preverbal objects in Old and Middle English were derived by movement to Spec,AgrOP; instead, preverbal negative, quantiWed, and positive objects had diVerent derivations. If this alternative analysis is correct, there was no reinterpretation at the end of the fourteenth century, but simply the loss of whatever mechanism was responsible for preverbal positive objects. Evidence in favour of this alternative analysis would be a demonstration that negative, quantiWed, and positive objects behaved diVerently from each other in Old English. In fact, the distribution of the data shows that there are clear diVerences in the behaviour of the three types, as shown in Table 15.2. Notice Wrst that the frequency of postverbal negative objects is much lower than the frequency for positive and quantiWed objects in both I-Wnal and I-initial clauses. This, however, does not necessarily reXect a diVerence in derivation, since diVerent syntactic contexts often show diVerences in frequencies (Kroch 1989). More importantly than the diVerence in relative frequencies, there are no postverbal negative objects at all in I-Wnal clauses. This categorical behaviour contrasts with that of positive and quantiWed objects, which show a healthy frequency of postverbal position in I-Wnal clauses, about 15 per cent. The absence of postverbal negative objects in

Table 15.2 The position of positive, quantiWed, and negative nominal objects in Old English clauses with auxiliary verbs Type of clause I-initial

I-Wnal

Type of object

Postverbal

N

% Postverbal

positive quantiWed negative positive quantiWed negative

1,815 150 25 154 25 0

3,893 376 143 1,070 167 81

46.6 39.9 17.5 14.4 15.0 0.0

The Syntax of Objects in Old English

259

I-Wnal clauses cannot be attributed to a length eVect (recall that heaviness is one factor that favours postverbal position for objects in Old English): 10 of the 81 objects are four words or longer. Based on the frequency of postverbal position for positive and quantiWed objects of diVerent lengths, we would expect 11 of the 81 negative objects to appear postverbally. Notice that within a framework where the headedness of projections can vary, the lack of postverbal negative objects in I-Wnal clauses receives a natural explanation: Old English is a negative concord language. Postverbal objects in I-Wnal clauses can only be derived by rightward movement from head-Wnal structure (see (7c)), and rightward movement creates an island that blocks negative concord. Thus positive and quantiWed objects can move rightward from preverbal position over the non-Wnite verb, but negative objects cannot. Now consider the case of clauses with Wnite auxiliaries, non-Wnite main verbs, and postverbal pronominal objects, particles, and stranded prepositions. Regardless of the assumptions made about underlying structure, these clauses are unambiguously head initial: light elements like pronouns, particles, and stranded prepositions cannot postpose from head-Wnal structure, probably because of a heaviness constraint on rightward movement (see Pintzuk 2002 for discussion). I call these elements ‘postverbal diagnostic elements’, since they are diagnostic of head-initial structure. The distribution of nominal objects in clauses with postverbal diagnostic elements is shown in Table 15.3; because of the small numbers, negative and quantiWed objects are grouped together. Three out of four quantiWed objects appear preverbally in these clauses, while positive objects rarely do (only one out of 40). Compare these numbers to those for the position of objects in I-initial clauses without postverbal diagnostic elements, also shown in Table 15.3: the number of preverbal positive objects is quite high, more than 50 per cent.

Table 15.3 The distribution of nominal objects in clauses with auxiliary verbs in Old English: the eVect of postverbal diagnostic elements Preverbal Clauses with postverbal diagnostic elements Non-quantiWed objects QuantiWed objects Clauses without postverbal diagnostic elements Non-quantiWed objects QuantiWed objects

Postverbal

Total

% Preverbal

1 3

39 1

40 4

2.5 75.0

2,077 341

1,789 175

3,866 516

53.7 66.1

260

S. Pintzuk

Table 15.4 The distribution of objects in clauses with auxiliary verbs in Early Middle English: the eVect of postverbal pronouns Preverbal Clauses with postverbal pronouns Non-quantiWed objects QuantiWed objects Clauses without postverbal pronouns Non-quantiWed objects QuantiWed objects

Postverbal

Total

% Preverbal

1 4

19 9

20 13

5.0 30.8

244 60

565 85

809 145

30.2 41.4

Source : Data from Kroch and Taylor 2000: tables 6.9, 6.10, and 6.11.

Although the amount of data in Table 15.3 is small, Early Middle English shows the same pattern, as can be seen in Table 15.4. This supports the conclusion that negative and quantiWed objects behave diVerently from positive ones in the early stages of English. To summarize, positive, quantiWed, and negative objects behave in diVerent ways during the Old English period: under the assumption of variation in the headedness of the VP, with objects of all types generated in either preverbal or postverbal position, positive and quantiWed objects can postpose from preverbal position over the non-Wnite verb, while negative objects cannot; quantiWed and negative objects can prepose from postverbal position over the nonWnite verb, while positive objects cannot. Thus the surface position of Old English objects is the result of a complex interaction of variation in underlying structure and the application of diVerent types of optional movement rules. If this analysis is correct, we would expect to see diVerences among the object types when we consider the change in their position over time: since the preverbal vs. postverbal position of the three types is derived in diVerent ways, the rate of the loss of OV order should be diVerent for each type. These diVerences can be seen in part by simply comparing the frequency of preverbal position of the three object types in the two Old English periods, as shown in Table 15.5. Notice Wrst that the main diVerence between positive and quantiWed objects is the higher frequency of quantiWed objects in preverbal position, which can be attributed to the fact that quantiWed objects can prepose from postverbal position while positive objects cannot. The change over time, i.e. the loss of preverbal position between the early and the late periods, is about the same for the two object types, contrary to expectations. However, the lack of statistical diVerence is probably due to a combination of three factors:

The Syntax of Objects in Old English

261

Table 15.5 The position of positive, quantiWed, and negative nominal objects in Old English I-initial clauses with auxiliary verbs, by period Type of object Positive QuantiWed Negative

Period

Preverbal

N

% Preverbal

Early Late Early Late Early Late

803 1,165 113 101 45 65

1,416 2,310 178 179 49 83

56.7 50.4 63.5 56.4 91.8 78.3

(1) the frequency of underlying head-initial VP structure is less than 50 per cent in Old English (33.3 per cent in the early period and 40.7 per cent in the late period, according to Pintzuk and Taylor 2003)–it is only the head-initial clauses which can show a diVerence between positive and quantiWed objects, since it is only these clauses that can be aVected by preposing; (2) the change in the frequency of underlying head-initial structure between the early and the late periods is not very large; (3) the rate of preposing of quantiWed objects is only 20.4 per cent in the early period and 14.7 per cent in the later period (see Table 15.6). In contrast, it is clear that rate of decrease of negative objects in preverbal position is much higher than for positive and quantiWed objects. The diVerent behaviour of negative objects is not simply due to the smaller amount of data. Pintzuk and Taylor (2003) use the frequency of OV and VO surface word order and the frequency of postposition to estimate the frequency of headinitial and head-Wnal structure and the rate of preposing for negative objects and quantiWed objects during Old and Middle English. As shown in Table 15.6, the rate of preposing for the two types of objects is diVerent, and the rate of preposing of negative objects is decreasing faster than the rate of preposing of quantiWed objects. In summary, negative, quantiWed, and positive objects are behaving diVerently during the Old English period; this conclusion is based on distributional evidence in clauses with unambiguous head-initial or head-Wnal structure, and supported by diVerential rates of loss of the preposing rules that aVect the surface position of negative and quantiWed objects. Contra van der WurV (1999), it is clear that the change in late Middle English is not a reinterpretation of the structure of preverbal objects, but rather the loss of preverbal positive objects. Under an analysis incorporating variation in the headedness of the VP, this loss can be interpreted as the loss of head-Wnal VP structure.

S. Pintzuk

262 Table 15.6

Rate of preposing for quantiWed and negative objects in Old English (%)

Period

QuantiWed

Negative

Early Late

20.4 14.7

75.4 46.7

15.4 View 3: Old English Can be Analysed as a Head-Initial Language, with Head-Final Order Derived by the Leftward Movement of Various Constituents In this section I brieXy present a head-initial analysis of Old English based on Nunes (2002) and Roberts (1997), and discuss the data that present problems for such an account. Recall that the absence of V O Aux surface orders required a stipulation under the variation in headedness analysis. I will demonstrate that uniform head-initial accounts require the same stipulation, and moreover that the distribution of objects and light elements cannot be derived except by additional stipulations. For the sake of concreteness, I use the terminology of Nunes, but the syntactic movements are the same for both Nunes and Roberts except where noted. All projections are head initial, and auxiliary verbs take complements that contain at least two functional projections, as shown in (9a). If there is no movement within the lowest VP, (9a) derives (2). Movement of the object to the outer Spec,vP for case-checking purposes, as shown in (9b), derives (1). Remnant vP movement to Spec,XP, the projection associated with Aux, derives clauses like (4) with uniform head-Wnal surface order, as shown in (9c). (9) a. underlying structure, deriving (2) Aux V O [XP Aux [YP [vP [VP V O ]]]] b. case checking, deriving (1) Aux O V [XP Aux [YP [vP Oi [VP V ti ]]]] c. remnant vP movement, deriving (3) O V Aux [XP [vP Oi [VP V ti ]]k Aux [YP tk ]] To derive (4), with objects after the V-Aux verb cluster, objects optionally scramble out of the vP before remnant vP movement occurs, as shown in (10): (10)

a. underlying structure [XP Aux [YP [vP [VP V O ]]]] b. case checking [XP Aux [YP [vP Oi [VP V ti ]]]]

The Syntax of Objects in Old English

263

c. scrambling [XP Aux [YP Oi [vP ti [VP V ti ]]]] d. remnant vP movement, deriving (4) V Aux O [XP [vP ti [VP V ti ]]k Aux [YP Oi tk ]] Let us consider more closely the possibility that objects may remain within the VP. Roberts (1997: 412) suggests that focused objects may escape case checking and remain in postverbal position within the VP. We must then prevent remnant vP movement with focused objects, so that V-O-Aux order is not derived from (11a) as in (11b). (11) a. underlying structure [XP Aux [YP [vP [VP V O ]]]] b. remnant vP movement, deriving * V-O-Aux  [XP [vP [VP V O ]]k Aux [YP tk ]] Roberts (1997: 418) proposes that focused elements must raise at LF, but that elements cannot raise out of a constituent on a left branch; thus the interpretation of (11b) is blocked when the object is focused. However, this proposal is not viable. Postverbal position in Old English cannot be directly linked to focus, because the frequency of postverbal objects increases during the Old English period, as was shown in Section 15.2; and it is of course unlikely that this increase is due to speakers using focused objects more frequently as time goes by. This means that movement for case-checking purposes is simply optional or perhaps driven by feature strength. We are left with the same problem: if remnant vP movement applies when the object has not moved, the unattested V-O-Aux order is derived as in (11b). In contrast to Roberts, Nunes (2002: 304 n. 6) suggests that object movement to the outer Spec,vP is obligatory rather than optional: under the copy theory of movement, the object in (9b) can be pronounced either in preverbal position, to derive (1), or in postverbal position, to derive (2); focal stress forces the pronunciation of the lower copy. But if movement for case-checking purposes is obligatory, and if either copy of the object can be pronounced in (9b), then surely either copy of the object can be pronounced in (9c). Again, in the case of focused objects, the lower copy is pronounced, deriving the unattested V-O-Aux order. The head-initial analysis with remnant vP fronting as presented thus far derives all of the same constituent orders that the variation-in-headedness analysis does, with both analyses needing a stipulation to block the derivation of V-O-Aux surface order. However, there are additional constraints on constituent order which cannot be derived under the uniform head-initial

264

S. Pintzuk

account. As we will see below, the basic problem is that the position of complements is not independent of the positions of the Wnite auxiliary and non-Wnite main verb. Certain processes and constituent orders are permitted in head-initial structures which are not permitted in head-Wnal structures, and vice versa. The Wrst type of problematic data consists of positive objects in clauses with postverbal diagnostic elements, discussed in Section 15.3: positive objects do not generally appear preverbally in clauses with head-initial IPs and VPs. Nunes (2002: 305) suggests that the long-distance scrambling (i.e. movement to preverbal position) of non-quantiWed objects and remnant vPs can be blocked by Relativized Minimality, assuming that pronominal objects move to a high position above vP. But the movement of pronominal objects is not relevant here: in the data under consideration, the diagnostic elements (including but not limited to pronominal objects) remain in postverbal position and therefore cannot block the scrambling of non-quantiWed objects in the way that Nunes suggests. I have seen no uniform head-initial analysis that can account for this constraint on scrambling in clauses with postverbal diagnostic elements. Under the variation in headedness account, this constraint may fall under Holmberg’s Generalization; it is comparable in its eVect to the constraint that blocks the scrambling of non-quantiWed objects to preverbal position in Modern Icelandic. The second problem for a head-initial analysis is the fact that certain types of constituents cannot appear in postverbal position in V þ Aux structures. We have seen in Section 15.3 that negative objects cannot appear in this conWguration. Pronominal objects, particles, and stranded prepositions behave in exactly the same way: they can appear preverbally or postverbally in I-initial clauses, but only preverbally in I-Wnal clauses; see Pintzuk (2002) for discussion and examples. If V þ Aux clauses are underlyingly head Wnal, there is an obvious explanation for the distribution: these constituents are base generated preverbally and cannot move rightward to postverbal position. The reasons for the constraint on movement are of course diVerent for the diVerent constituents: negative objects cannot move rightward because they cannot enter into negative concord from the island created by postposition; pronouns, particles, and stranded prepositions cannot move rightward because postposition applies only to heavy and/or stressed constituents. In order to demonstrate the problem that this constraint poses for a uniform head-initial analysis, consider the constituent orders in (12), with the structures that would be derived from a head-initial analysis. (12a-c) are possible for all types of XP. (12d) is possible for positive and quantiWed objects but not for negative objects or prosodically light elements.

The Syntax of Objects in Old English (12)

a. b. c. d.

265

Aux V XP [XP Aux [YP [vP [VP V XP ]]]] Aux XP V [XP Aux [YP [vP XPi [VP V ti ]]]] XP V Aux [XP [vP XPi [VP V ti ]]k Aux [YP tk ]] V Aux XP [XP [vP ti [VP V ti ]]k Aux [YP XPi tk ]] (* for XP¼ negative object, pronominal object, particle, stranded preposition)

Let us adapt in the following way Nunes’s suggestion that pronoun scrambling blocks long-distance remnant vP movement. Assume that the movement of complements of all types to Spec,vP is optional; thus we correctly derive both (12a) and (12b). Once complements are in Spec,vP, they can remain there and be moved to pre-Aux position by remnant vP movement, as in (12c). However, if they move out of Spec,vP before remnant vP movement, then they must raise to a position high enough (i.e. higher than positive objects) to block remnant vP movement by Relativized Minimality. This is sketched in (13), following Nunes (2002: 305). (13)

a. underlying structure [XP Aux [WP [YP [vP [VP V O-pro]]]]] b. case checking (movement of O-pro to Spec,vP) [XP Aux [WP [YP [vP O-proi [VP V ti ]]]]] c. scrambling (movement of O-pro to Spec,YP) [XP Aux [WP [YP O-proi [vP ti [VP V ti ]]]]] d. pronoun scrambling (movement of O-pro to Spec,WP) [XP Aux [WP O-proi [YP ti [vP ti [VP V ti ]]]]] e. remnant vP movement (movement of vP to Spec,XP) is blocked *[XP [vP ti [VP V ti ]]k Aux [WP O-proi [YP ti tk ]]]

However, this account cannot be maintained, for both empirical and formal reasons. Space constraints prevent a full discussion of the data here, but three problems are worth mentioning. First, it is only particles that exhibit the categorical behaviour required by this account; particles in preverbal position are always adjacent to the verb. In contrast, pronominal objects can raise further, but they do not necessarily raise higher than nominal objects, as shown in the example in (14): (14) þæt hi woldon þære byrgene hine eft befæstan that they would the grave him again entrust ‘. . . that they would again entrust him to the grave.’ (cogregdC,GD-2-[C]:24.154.21.1843)

266

S. Pintzuk

Second, it is not at all clear that pronoun scrambling and remnant vP movement are triggered by the same type of features, and therefore that the Wrst can block the second. Biberauer and Roberts (2003), for example, propose that vP moves to satisfy the EPP feature of TP in Old English. If this is true, then remnant vP movement is A-movement rather than A’-movement and therefore, like subject movement to Spec,TP, cannot be blocked by scrambled pronouns under Relativized Minimality. And third, movement rules must be ordered so that pronoun scrambling in (13d) occurs before remnant vP movement in (13e); it is diYcult to see how rule ordering of this type can be enforced. To summarize, head-initial accounts do not succeed in deriving the word order patterns of Old English, because head-initial accounts cannot use the distinction between Aux- . . . -V clauses and V-Aux clauses to block some types of movement and permit others.

15.5 Conclusions In this chapter I have presented evidence against three commonly held views of Old English syntax, evidence which is primarily quantitative in nature. I have shown that the patterns of variation in the Old English data are best explained by an analysis of Old English as a language exhibiting competition between head-initial and head-Wnal structure, both in the IP and in the VP. The change from OV to VO in the history of English can then be seen as the loss of head-Wnal structure, combined with the loss of preposing of negative and quantiWed objects.

16 Clause Type Asymmetries in Old English and the Syntax of Verb Movement E R I C H A E BE R L I

16.1 Introduction Old English exhibits main clause/subordinate clause word order asymmetries comparable to those found in modern West Germanic languages such as Dutch or German. Finite verbs generally occur near the beginning of a main clause whereas they tend to occur towards the end of subordinate clauses. Furthermore (in contrast to modern West Germanic), it has been observed that conjoined main clauses often have subordinate clause word order in Old English. This chapter addresses the question of how these clause type asymmetries can be accounted for. Section 16.2 presents the main word order properties of Old English and some quantitative data concerning the distribution of Wnite verbs in diVerent types of clauses. Recent proposals for the structural analysis of the Old English word order patterns are presented in Section 16.3, and some problems for the treatment of subordinate clauses are discussed. Section 16.4 presents a solution to these problems in terms of Bobaljik and Thra´insson’s (1998) analysis of verb movement in modern Germanic. The remainder of the chapter then explores some consequences of this approach with respect to the analysis of conjoined clauses and the development of verb movement in the history of English. This chapter owes its existence to several people whose response to presentations of some of my earlier work was: ‘What about subordinate clauses?’ In particular, I would like to thank Tho´rhallur Eytho´rsson, Richard Ingham, Susan Pintzuk, and Anthony Warner for comments which made me pursue this topic and later on for discussions on the proposals made here. Thanks also go to the audiences at the York-Holland Symposium on the History of English Syntax (University of York, April 2002), the 7th Diachronic Generative Syntax Conference (University of Girona, June 2002), and the 17th Comparative Germanic Syntax Workshop (University of Iceland, August 2002) where earlier versions of this chapter were presented.

268

E. Haeberli

16.2 The Distribution of Finite Verbs in Old English As often observed in the literature, one of the characteristic properties of Old English (OE) word order is the variation in the distribution of Wnite verbs in diVerent types of clauses (cf. e.g. Bacquet 1962; Mitchell 1985; Traugott 1992; van Kemenade 1987). In main clauses, the verb tends to occur in a position near the beginning of the clause. This leads to frequent Verb Second (V2) orders as shown in (1).1 (1) a. [He] com on Breotone mid fyrde [SV. . . ] he came to Britain with army (Bede 1: 5.32.16.257) b. [Ðas gifu] sealde seo ceasterwaru on Tharsum Apollonio this gift gave the citizens in Tharsus Apollonius þam tiriscan [OV. . . ] the Tyrian ‘The citizens of Tharsus gave this gift to Apollonius the Tyrian.’ (ApT: 10.16.181) c. And [egeslice] spæc Gregorius be ðam . . . [AdvV. . . ] and sternly spoke Gregorius about that ‘And Gregorius spoke sternly about that . . .’ (WHom 10c: 48.865) In subordinate clauses, however, there is a tendency for the verb to be placed towards the end of the clause. This is illustrated by the following clauses (from van Kemenade 1987: 16/19). (2)

a. þæt ic þas boc of Ledenum gereorde to Engliscre spræce awende that I this book from Latin language to English tongue translate ‘. . . that I translate this book from Latin to English.’ (ÆCHom I [Pref]: 174.48.5) b. þæt Darius hie mid gefeohte secan wolde that Darius them for battle visit wanted ‘. . . that Darius wanted to seek them out for a battle.’ (Or 2: 5.45.30.867)

Finally, with respect to conjoined clauses, and in particular second conjuncts introduced by conjunctions, it has been observed that they ‘tend to be verb-Wnal, like subordinate clauses’ (Traugott 1992: 272). Illustrations for this 1

The Old English data are taken from The York-Toronto-Helsinki Parsed Corpus of Old English Prose (Taylor et al. 2003) and follow the referencing conventions of that corpus.

Clause Type Asymmetries in Old English

269

observation are given in (3) where the Wnite verbs in the Wrst conjunct occur in second position and in the second conjunct in Wnal position. (3) a. [þa tungel witgan of eastdæle] cuomon to þon þæt hie Crist the star prophets from East came to that that they Christ weorþedon & [þa cild on Bethlem] [ofslægene] wærun . . . honour and the children in Bethlehem killed were ‘The wise men from the East came to honour Christ and the children in Bethlehem were killed . . .’ (ChronA: 2.1.59/60) b. . . . [ða] aras he hal & gesund. Ond [he] [for his then arose he uninjured and healthy. And he for his hælo] [eft] [Dryhtne] [þonc] [secgende] wæs . . . health again God thank saying was ‘Then he got up uninjured. And he repeatedly thanked God for his health.’ (Bede 4: 32.380.16.3796/7) c. [Sume hy] forleton þæt unalyfede þing. And [mid clænnysse] some they gave-up that unlawful thing. And with purity [Criste] þenedon . . . Christ served ‘Some gave that unlawful thing up and served Christ with purity.’ (Ælet 2 [Wulfstan 1]: 79.124/5) As is well known, the contrasts shown in (1) to (3) are tendencies rather than systematic diVerences among clause types. Thus, main clauses can occasionally have verb-Wnal word order (cf. e.g. Koopman 1995), or in subordinate clauses various constituents can follow the Wnite verb (cf. e.g. van Kemenade 1987; Pintzuk 1999). Let us therefore brieXy examine how strong the word order tendencies in (1) to (3) are in quantitative terms. As a test case, I will consider clauses like (2b), (3a), and (3b), which contain a Wnite verb and a non-Wnite verb. In terms of the observations made above, we would expect Wnite verbs generally to precede non-Wnite verbs in main clauses, whereas we should Wnd frequent occurrences of the inverted order (non-Wnite–Wnite) in subordinate and conjoined main clauses. Table 16.1 shows the relevant quantitative data for a sample of texts from The York-Toronto-Helsinki Parsed Corpus of Old English Prose (Taylor et al. 2003) (Aux ¼ Wnite verb; V ¼ non-Wnite verb; MC ¼ non-conjoined declarative main clause;

270

E. Haeberli

þMC ¼ declarative main clause introduced by a conjunction; SC ¼ subordinate clause).2 An important observation based on Table 16.1 is that conjoined main clauses and subordinate clauses do not have the same status with respect to the distribution of Wnite verbs. While it is correct that in general the frequency of the order ‘non-Wnite–Wnite’ is higher in conjoined main clauses than in non-conjoined main clauses (four times higher in the total percentage), there is also a substantial diVerence between conjoined main clauses and subordinate clauses in that the latter have a much higher frequency of ‘non-Wnite– Wnite’ orders (six times higher). Thus, we obtain a scale concerning the occurrence of Wnite verbs in Wnal or near-Wnal position in OE clauses with subordinate clauses having such orders more frequently than conjoined main clauses and conjoined main clauses having such orders more frequently than non-conjoined main clauses. This scale is conWrmed if we look at other word

Table 16.1 The order of Wnite auxiliaries and non-Wnite main verbs in diVerent texts and clause types in Old English MC AuxV

MC VAux

þMC AuxV

þMC VAux

SC AuxV

Bede

409

140

367

Cura Pastoralis

463

76 35.2% 2 1.2% 1 0.5% 6 8.5% 2 1.4% 10 2.9% 2 8.3% 4 2.6% 103 7.9%

217

Boethius

38 8.5% 6 1.6% 2 0.4% 2 2.9% 0 0.0% 2 0.3% 0 0.0% 1 1.0% 51 2.1%

Chronicle A

67

Ælfric’s Letters

218

Ælfric’s Lives of Saints

692

Apollonius of Tyre

35

Wulfstan’s Homilies

102

total

2

2,353

159 188 65 137 338 22 150 1,199

267 545 23 109 404 22 105 1,692

SC VAux 449 67.4% 248 48.2% 313 36.5% 41 64.1% 65 37.4% 216 34.8% 19 46.3% 87 45.3% 1,438 45.9%

In Table 16.1, only clauses with an overt subject are counted in which the non-Wnite verb does not precede the subject. Similar percentages are obtained if clauses without an overt subject (e.g. conjoined clauses with an omitted subject in the second clause) are included.

Clause Type Asymmetries in Old English

271

order patterns. For example, in clauses containing an overt (pronominal or non-pronominal) subject which precedes both a Wnite main verb and a full DP object, the orders SOV and SVO are distributed as follows in the texts examined in Table 16.1: (a) subordinate clauses: 57.1 per cent SOV—42.9 per cent SVO; (b) conjoined main clauses: 36.1 per cent SOV—63.9 per cent SVO; (c) non-conjoined main clauses: 15.0 per cent SOV—85.0 per cent SVO. With respect to these Wgures, conjoined main clauses fall exactly between the other clause types (21 per cent diVerence to both of them). In summary, we can observe a three-way distinction among clause types with respect to the distribution of Wnite verbs in OE: (i) subordinate clauses where the Wnite verb frequently occurs in Wnal or near-Wnal position; (ii) conjoined main clauses where such word orders occur regularly but less frequently; (iii) main clauses where such word orders are even less frequent. The question that arises then is how these diVerences between the diVerent clause types can be accounted for.

16.3 The Structural Analysis of Old English The main/subordinate asymmetry shown in (1) and (2) is reminiscent of the one found in the modern West Germanic SOV/Verb Second (V2) languages like Dutch or German. A Wrst possibility would therefore be to analyse OE along the lines of what has been proposed for Dutch or German (cf. van Kemenade 1987). This would mean that OE has systematic verb movement to C and XP movement to Spec,CP in main clauses, and that it has head-Wnal projections below C (I and V) which give rise to verb-Wnal word orders with Wnite verbs in subordinate clauses because the presence of a complementizer blocks verb movement to C. However, it has been shown that OE has certain syntactic properties which are problematic for a parallel treatment of Dutch/German and OE. First, the V2 syntax in main clauses is less consistent in OE than in Dutch/German. For example, in contexts where a non-subject constituent is fronted, subject pronouns systematically occur between the fronted constituent and the Wnite verb, thereby giving rise to Verb Third (V3) (cf. e.g. Fourquet 1938; van Kemenade 1987; Pintzuk 1999). In the same context, we can also frequently Wnd V3 with full DP subjects, although V2 is the majority pattern here (cf. Koopman 1998; Haeberli 2002a). Systematic subject-verb inversion (i.e. rigid V2) with pronoun and full DP subjects occurs only in contexts that have been referred to as operator fronting contexts (i.e. questions or negation, but also with less operator-like fronted elements such as þa or þonne ‘then’). Thus, the V2 syntax of OE is substantially diVerent from the one found in Dutch or

272

E. Haeberli

German, and the standard V-to-C analysis of main clauses proposed for these languages cannot satisfactorily account for the situation in OE. Subordinate clauses raise additional problems for the extension of the analysis of modern West Germanic to OE. As shown by Pintzuk (1993, 1999) and Haeberli and Haegeman (1995), aspects of the syntax of particles, pronouns, and negative constituents suggest that the structure below C cannot be analysed uniformly in terms of head Wnal projections, contrary to what has traditionally been proposed for modern West Germanic. Instead, the projection hosting the Wnite verb in subordinate clauses seems to be at least optionally head-initial. Finally, another problem that a parallel treatment of OE and Dutch/German raises is the behaviour of conjoined main clauses in OE. Given that conjoined main clauses do not exhibit verb-Wnal word order in modern West Germanic, it remains unclear how the occurrence of such word orders could be explained in OE. Given the problems discussed in the previous paragraph, alternative structural analyses of OE have been proposed in much recent work. With respect to the analysis of main clauses, there has been a certain consensus in the literature. Based on the distributional properties of nominal and pronominal subjects in contexts where another constituent is fronted (V2/V3), two main assumptions are generally made (cf. e.g. Cardinaletti and Roberts 1991, Fischer et al. 2000; Haeberli 2000, 2002b; Hulk and van Kemenade 1997; Kroch and Taylor 1997; Pintzuk 1993, 1999): (i) Finite verbs move to two potential landing sites, namely to C when an operator occurs in clause-initial position and to the head of a headinitial inXectional projection below C when a non-operator is in initial position. As for the nature of the lower target of V-movement, various proposals have been made in the literature. Here, I will adopt the proposal in Haeberli (2000) where this head is identiWed as Agr on the basis of some observations related to the Middle English dialect variation discussed by Kroch and Taylor (1997). (ii) DiVerent types of subjects occur in diVerent structural positions. Pronouns, being clitics or weak pronouns, have to occur in a high position (Spec,AgrP). Full DP subjects can (at least optionally) remain in a lower position (Spec,TP). Finally, several of the authors cited above make a third assumption: (iii) Non-operators (like operators) in clause-initial position occur in Spec,CP.

Clause Type Asymmetries in Old English

273

In terms of these assumptions, we obtain the following general picture for the OE main clause syntax (targets of head movement in italics): (4)

[CP XPC [AgrP SU1 (þ/–pronominal) Agr [TP SU2(–pronominal) T. . . ]]]

(4) accounts for the fact that pronominal subjects precede the Wnite verb when a non-operator XP is fronted (V in Agr) but follow the Wnite verb when an operator is in initial position (V in C). Non-pronominal subjects, however, generally follow the Wnite verb in both contexts but can optionally precede the verb in non-operator fronting contexts. Although the structure in (4) provides an explanation for the main word order phenomena found in OE main clauses, it raises two important problems for the analysis of subordinate clauses and an additional problem for the analysis of conjoined main clauses. First, according to (4), we would expect the following scenario to be possible for subordinate clauses: (a) C is Wlled by a complementizer; (b) the Wnite verb moves to Agr; (c) a non-pronominal subject can remain in SU2. The consequence of (a) to (c) would be a subordinate clause with the word order ‘complementizer-Wnite verb-full DP subject’. This word order would even be expected to occur very frequently in OE subordinate clauses, because in main clauses with V in Agr full DP subjects occur more often in SU2 than in SU1 (i.e. more V2 than V3). Yet, this expectation is not borne out. The order ‘complementizer-Wnite verb-full DP subject’ is not a productive word order pattern in OE.3 The second problem for (4) is raised by the main/subordinate asymmetry. Let us assume, as in (b) above, that the Wnite verb moves to Agr in subordinate clauses. While verb-Wnal orders suggest that Agr is head Wnal, certain other constructions (involving particles, pronouns, or negation) suggest that it is sometimes head initial. Following Pintzuk (1993, 1999), we can analyse this ambiguous status of AgrP in terms of the double base hypothesis, i.e. the hypothesis that AgrP can be either head Wnal or head initial (grammar competition). The diVerent word order patterns discussed in Section 16.2 then suggest that head-Wnal AgrP is frequent in subordinate clauses but very rare in main clauses. In the data Pintzuk (1999: 223) discusses, head-Wnal AgrP (IP in her analysis) occurs in 16 per cent of the main clauses and in 53 per cent of the subordinate clauses. But as Pintzuk (1999: 223) admits, this considerable 3 This word order is not entirely absent from the Old English corpus, but it is very rare. Among all the subordinate clauses (excluding relative clauses) in The York-Toronto-Helsinki Parsed Corpus of Old English Prose, I have found only 120 examples with the Wnite verb in initial position after the element(s) in CP (0.2 per cent of the total). Presumably, this word order is best analysed as a case of main clause syntax in an embedded context (i.e. V1 due to CP-recursion).

274

E. Haeberli

contrast remains unexplained in terms of the double base hypothesis. Thus, the main/subordinate asymmetry is problematic from the point of view of the structure in (4). A similar problem arises for conjoined main clauses. The higher frequency of verb-Wnal word order in conjoined main clauses as compared to nonconjoined main clauses would suggest that the frequency of head-Wnal AgrP is higher in conjoined main clauses. But it again remains unexplained why there should be such a diVerence in frequencies. In the next two sections, I will show how the problems raised by subordinate and conjoined clauses can be dealt with in terms of the structural analysis of OE summarized in (4).4

16.4. The Main Clause/Subordinate Clause Asymmetry As observed earlier, asymmetries between main and subordinate clauses in the Modern Germanic languages have been argued to be the result of the (un)availability of certain projections as the landing sites for verb movement in the diVerent clause types. In this section, I will propose an analysis of OE that diVers from the standard analysis of modern Germanic but is based on a very similar idea. More precisely, pursuing proposals made by Bobaljik and Thra´insson (1998), I will argue that in OE main clauses the Wnite verb moves higher than in subordinate clauses. Bobaljik and Thra´insson’s analysis of verb movement in Modern Germanic is based on the following main assumptions made within the Minimalist framework (Chomsky 1995): (i) Elements move for the purposes of feature checking. InXectional heads and V have features which require checking against one another. 4 One potential explanation for the absence of ‘complementizer-Wnite verb-full DP subject’ is discussed by Kroch and Taylor (1997: 306 V.). Their proposal is based on the assumption that there is a V2 constraint holding at the level of AgrP in OE which, in main clauses, is satisWed by the topic XP on its way to Spec,CP and, in subordinate clauses, generally by the subject due to discourse factors which disfavour fronting of non-subjects. However, this approach is rather idiosyncratic as there does not seem to be any cross-linguistic evidence for a V2 constraint at a structural level (AgrP) below the target of topics (CP). Furthermore, the clause type asymmetries shown in Table 16.1 would remain unexplained. An alternative explanation for the rarity of ‘complementizer-Wnite verb-full DP subject’ is explored in an earlier version of this chapter (cf. Haeberli 2001: 209–15). There I considered the possibility of dealing with this issue in terms of main clause structure in subordinate contexts (i.e. in terms of a Split CP or CP-recursion). Although such an analysis may indeed hold for some subordinate clauses (possibly of a certain type like complements of bridge verbs), it would again raise the problem that it cannot account for the clause type asymmetries. The goal of the analysis outlined below is to address all the problematic issues raised in this section, and not just the absence of ‘complementizer-Wnite verb-full DP subject’.

Clause Type Asymmetries in Old English

275

(ii) Features are checked in any type of local conWguration (head speciWer, head-adjoined head, and, contra Chomsky 1995, head complement). (iii) The splitting of INFL is parametrized (Split IP Parameter (SIP)). Some languages have the clause structure AgrSP-TP-AgrOP-VP whereas others have the structure IP-VP. With respect to the distribution of verbs, these assumptions have the following consequences. According to Bobaljik and Thra´insson, English is a language with a simple IP-VP structure. Even though INFL may have a feature which has to be checked against V, the verb does not have to move to INFL in English. The reason for this is that under the standard Minimalist assumption that the features of a projection (e.g. VP) are those of its head (V), a checking relation between V and InX can be established in the head-complement conWguration I-VP according to hypothesis (ii) above. In other languages, however, verbs do move and, within Bobaljik and Thra´insson’s framework, movement is related to a split IP structure (postulated for e.g. Icelandic). Given a split IP of the form AgrSP-TP-AgrOP-VP, the occurrence of verb movement out of the VP can be accounted for by assuming that T and V have features which require checking against one another. Given that AgrOP intervenes between T and VP in a split IP structure, in situ checking is not possible here. The only option for establishing a checking relation between T and V is therefore V movement out of the VP so that V can enter a local relation with T. For our purposes, an additional point made by Bobaljik and Thra´insson with respect to V movement in Icelandic will be crucial. Bobaljik and Thra´insson argue that, in cases where the verb does not move to C in Icelandic, V movement does not target the highest inXectional head (AgrS) but only the head below AgrS, i.e. T. Evidence for this claim comes from two diVerent domains of the grammar of Icelandic. First, in subordinate clauses which generally do not license embedded topicalization (i.e. without V-to-C), Bobaljik and Thra´insson (1998: 63) identify two subject positions above the Wnite verb, and they suggest that these subject positions are Spec,AgrSP and Spec,TP. In terms of this analysis of subjects, the Wnite verb thus has to occur in T when it does not move to C. And secondly, again in embedded clauses which do not license V-to-C, certain adverbs can either follow the Wnite verb or immediately precede it and thus intervene between the subject and the verb. This phenomenon is illustrated in (5) (examples from Bobaljik and Thra´insson 1998: 64).

276 (5)

E. Haeberli a. Marı´a las kvæðið þegar hu´n (loksins) keypti (loksins) Mary read poem-the when she Wnally bought Wnally bo´kina (Icelandic) book-the ‘Mary read the poem when she Wnally bought the book.’ b. það er nu´ það sem e´g (ekki) veit (ekki) That is now it that I not know not ‘That’s exactly what I don’t know.’

If we assume that Wnite verbs only move to T, the variation in (5) can be accounted for by proposing that the subject is in Spec,AgrSP and the verb in T, and that there is variation with respect to the placement of the adverb (TP adjoined or VP adjoined). I will adopt Bobaljik and Thra´insson’s basic proposals here for the analysis of OE word order. But before doing so, an additional point should be discussed brieXy. Bobaljik and Thra´insson assume that V movement is triggered by formal features on an inXectional head and V which require checking against one another. Although Bobaljik and Thra´insson do not discuss this issue in much detail, the claim that Icelandic verbs only move to T and not to AgrS may suggest that only T has a feature which requires checking against V, whereas AgrS does not have such a feature. However, there would be an alternative option, namely that all inXectional heads have a feature which requires checking against V. For languages like English with a simple IP-VP structure, the consequences of this assumption are straightforward, as discussed above. But what about Icelandic, which has a rich IP structure (AgrSPTP-AgrOP-VP) and V movement to T? I propose that the idea behind the analysis of English can be extended to this case as well. If the features of a projection are those of its head, we can assume that this also holds for complex heads derived through head movement. Thus, a T-head containing V after V movement has features of both T and V and the maximal projection could then be argued to contain features of both of these heads. As a consequence, AgrS can enter a checking relation with verbal features through the head-complement conWguration with TP, and V movement to AgrS is therefore not necessary. In other words, V movement to T is suYcient in Icelandic not because AgrS lacks features which require checking with V, but because, by moving to T, the verb is suYciently close to enter a checking relation with AgrS due to feature percolation within TP. Given these proposals let us now return to Old English. As discussed in Section 16.3, the assumption made in most recent work on OE is that in main clauses, V movement can target two possible positions, a head position in the CP domain and an inXectional head below C. The two head positions are given in italics in (4) above, repeated here as (6).

Clause Type Asymmetries in Old English (6)

277

[CP XPC[AgrP SU1 (þ/pronominal) Agr [TP SU2(-pronominal) T . . . ]]]

A Wrst question that arises in terms of (6) is why verbs that do not move to C move to the inXectional head below C in OE. In terms of the proposals made above, movement of the verb to Agr cannot be triggered by features of Agr (since these features could be checked by V-to-T movement), but it has to be triggered by a head in the CP domain which requires checking by V. If we assume a split CP structure of the type ForceP-TopP/FocP-FinP above AgrP (cf. Rizzi 1997), the crucial head is the Wniteness head Fin. We conclude then that Fin has a feature which has to be checked by V and that the verb therefore moves to Agr. The features of V percolate up to AgrP and Fin can enter a checking relation with the verb in the head-complement conWguration FinAgrP. The second question raised by (6) is how we derive V-to-C movement, which, as observed in Section 16.3, generally occurs with operator fronting to CP in OE. There are two main options here. First, extending the feature checking approach outlined above, we could propose that a head above FinP (i.e. the head of the projection hosting the operator) contains a feature which attracts V and thus gives rise to verb movement into the CP domain. A potential alternative to this analysis would be to derive V-to-C in terms of a structural well-formedness criterion for operators along the lines of Rizzi’s (1996) wh-criterion. Having considered main clauses, let us now turn to subordinate clauses in OE. As pointed out in Section 16.3, the main/subordinate asymmetry with respect to the distribution of Wnite verbs would be surprising if the verb occupied the same structural position in both types of clauses. The theoretical approach discussed in this section now provides the basis for a solution to this problem. Along the lines of Bobaljik and Thra´insson ’s analysis of (5), we may assume that in OE subordinate clauses the verb generally moves only to T rather than to the main clause targets Agr or C.5 Furthermore, in order to 5 That V does not move as high in subordinate clauses as in main clauses has been independently proposed by Fuss and Trips (2002). They use the occurrence of the word order ‘C-SU(pronoun)-AdvV(þWn) . . . V(-Wn) . . .’ in embedded clauses and the rarity of such orders in main clauses as evidence for distinct positions for the Wnite verb (2002: 190 V.). However, this evidence is not conclusive, as the contrast could be accounted for in terms of a unique landing site by assuming that the projection whose head the verb occupies is frequently head Wnal in subordinate clauses (Pintzuk 1999) and that the material to the right of the Wnite verb has undergone Verb Projection Raising. It therefore seems that the main arguments for an asymmetric V-movement analysis in Old English come from the problems raised in Section 16.3. Fuss and Trips’s analysis diVers from the one proposed here in that it is based on an Agr-less structure where the verb moves to T in main clauses and remains within the verbal domain in subordinate clauses. However, as shown in Haeberli and Ingham (forthcoming), there is evidence

278

E. Haeberli

account for the ambiguous status of the projection hosting the Wnite verb in subordinate clauses (head initial/head Wnal; cf. Section 16.3), we can adopt Pintzuk’s (1999) double base hypothesis and propose that it is TP which can be head Wnal or head initial (rather than AgrP/IP as in Pintzuk’s analysis). The frequent verb-Wnal word orders in subordinate clauses are thus the result of the frequent use of head-Wnal TP. Given that in main clauses the verb generally moves on at least to Agr, a head-initial projection, the presence of a head-Wnal TP does not have any consequences for main clause word order. Hence, the main/subordinate asymmetry with respect to the placement of Wnite verbs can be accounted for.6 A V-to-T analysis of OE subordinate clauses also immediately solves the Wrst problem raised in Section 16.3. The occurrence of the Wnite verb in T in (6) means that a subject generally precedes the verb in subordinate clauses regardless of whether it is pronominal or non-pronominal. This explains the fact that ‘complementizer-Wnite verb-full DP subject’ is not a productive word order option in OE. One additional issue remains to be addressed at this point. Why is it that the verb only moves to T in subordinate clauses, whereas it has to move at least to Agr in main clauses? As proposed above, verb movement to Agr in main clauses can be analysed as the result of the feature checking requirements of the Fin-head. As a consequence, Fin must have a diVerent status in subordinate clauses. I propose that the reason for this is that Fin is the head from Early Middle English suggesting that the Wnite verb in subordinate clauses moves not as high as in main clauses but nevertheless out of the VP (to T) because it can occur to the left of the negator not, which is generally assumed to mark the left edge of the VP. The minimal assumption would then be that V-to-T in Early Middle English subordinate clauses is simply a continuation of Old English V-toT. Furthermore, the analysis proposed here has the advantage of allowing us to integrate Early English into the larger picture of the syntax of verb movement in Germanic outlined by Bobaljik and Thra´insson. 6 Alternatively, within a framework which bans head-Wnal projections (cf. Kayne 1994), we could assume that, as sometimes proposed in recent work, verb-Wnal word order is derived through remnant leftward movement of a large constituent and, more speciWcally, that this remnant movement targets the speciWer position of a functional projection immediately below AgrP. Grammar competition then would not concern the directionality of inXectional projections but variation with respect to whether this leftward movement occurs or not. As pointed out by Susan Pintzuk (p.c.), some statistical considerations may raise a potential problem for the proposal that the verb generally does not occupy the same position in main and subordinate clauses. Pintzuk (1999: 223–4) shows that, in terms of her structural interpretation of the corpus she examined, the frequencies of head-Wnal structures in main and subordinate clauses change in parallel in the Old English period. If distinct landing sites were involved in the two types of clause, such a constant rate eVect would be surprising as it would mean that the two inXectional projections are somehow interdependent and it would not be clear why this should be the case. For the time being, I will have to leave this issue open. To address it fully, a detailed statistical evaluation of the analyses proposed here and in Pintzuk’s work would be necessary, a task which we are planning to carry out in future joint work.

Clause Type Asymmetries in Old English

279

where the complementizer is inserted in OE (cf. also Rizzi 1997: 288 on the option of inserting a complementizer in Fin, and Haeberli 2002c: 234 for this proposal for the modern Germanic languages). The presence of a complementizer in Fin can then be argued to have one of the following two possible consequences. First, we could assume that the complementizer actually satisWes Fin’s feature checking requirements itself. V therefore does not have to move into a local conWguration with Fin, but only with Agr, and it therefore moves to T. Alternatively, we could argue that the Fin-head which allows insertion of a complementizer has syntactic properties that diVer from those of the empty Fin-head. It is only the latter which bears a feature requiring a checking relation with V, whereas the former does not bear such a feature. Again the result would be that the Wnite V remains in T. I will have to leave it open for the moment whether there are any arguments in favour of one approach or the other. What is common to both of them is the idea that the insertion of a complementizer interferes with the movement properties of Wnite verbs. In this respect, they are similar to the traditional analyses of main/subordinate asymmetries as found, for example, in Dutch or German.7

16.5 Conjoined Main Clauses In this section, I propose that the approach outlined above can also shed light on the observation that OE conjoined main clauses exhibit subordinate clause word order more frequently than other main clauses. In order to obtain this result, one main assumption is necessary, namely the assumption that the input categories for coordination are not always full CP structures but that main clauses in OE can be conjoined either at the CP level or at the AgrP-level

7 At Wrst sight, there seems to be a substantial diVerence, however. In the traditional analyses of Dutch or German, the complementizer and the verb compete for the same head position, which is not the case for the proposal in the text. Yet, if we assume that V-to-C is triggered by a feature on a C-head, it is not clear whether the traditional intuition is suYcient to explain the situation in Dutch/German. A complementizer in subordinate clauses does not necessarily block V movement. Instead, the verb could adjoin to C, thereby creating a complex Comp-Vor V-Comp head. Given that such heads do not occur in Dutch/German, we have to conclude that there is an independent reason why verbs do not move to C in subordinate clauses. Adopting the proposals made in the text, we could say either that a complementizer can check the relevant feature in the CP domain or that the presence of a complementizer alters the feature speciWcation of the CP domain. Given these observations, the situation in Old English may indeed be directly comparable to the situation in languages such as Dutch/German. As observed in notes 3 and 4, it is conceivable that main clause structure can occur in (some types of) subordinate clauses. For the proposals made in the text, this would mean that certain complementizers can be optionally inserted in a higher C-head (Force in Rizzi’s Split-CP). The empty Fin-head then attracts the verb to Agr as in main clauses. In what contexts this option is available and how frequently it occurs is an issue for future research.

280

E. Haeberli

(cf. also Kiparsky 1995: 149 for a similar proposal). It is the second option which provides the basis for the analysis of verb-Wnal conjoined main clauses. Let us assume (following e.g. Johannessen 1998) that conjunction involves a conjunction phrase (CoP) with the two conjuncts as the speciWer and the complement respectively. If two AgrPs are conjoined, CoP is the complement of Fin, and the Wrst AgrP occupies Spec,CoP while the second AgrP occupies the complement position of Co. Within the AgrPs, the Wnite verbs undergo movement to T in order to enter a checking relation with Agr (cf. Section 16.4). But the Fin-head taking CoP as a complement also has a feature that requires checking by a verb. To establish this checking relation, Fin selects the Wnite verb in the Wrst conjunct, which can be argued to be the closer one of the two Wnite verbs in the conjuncts.8 The Wnite verb therefore moves to Agr in the Wrst conjunct. As proposed in Section 16.4, the verbal feature then percolates to AgrP. Assuming furthermore (following Johannessen 1998: 110) that Co inherits the syntactic features from its speciWer conjunct by speciWerhead agreement, CoP satisWes the checking requirements of Fin. Thus, we obtain an asymmetry between the Wrst conjunct and the second conjunct when two AgrPs are conjoined. While the Wnite verb has to move to Agr in the Wrst conjunct, it remains in T in the second conjunct. Given the proposal made in Section 16.4 that TP is frequently head Wnal, we can now account for the increased frequency of verb-Wnal word order in conjoined main clauses as compared to other main clauses. As for the fact that verb-Wnal in conjoined main clauses is not as frequent as in subordinate clauses, it can be explained in terms of variation with respect to the categorial status of the conjoined constituents. As proposed above, conjoined clauses can also be CPs and these clauses do not give rise to verb-Wnal word order because each of them contains a Fin-head that requires checking by V. The AgrP/CP variation with conjoined main clauses is independent of subordinate clause syntax and the distinct properties of the two types of clause are therefore expected. The syntax of coordination raises various additional issues, but reasons of space do not allow me to consider these and the extensive literature dealing with them in any detail here. In general, the approach outlined by Johannessen (1998) seems to give the desired results for our purposes, the main exception being that while Johannessen (1998: 204 V.) argues that the input 8

Note that this result cannot be obtained by deWning closeness in terms of c-command (as proposed by Chomsky 1995: 358) because the verbal features are on TP in the two conjuncts and neither of the two TPs c-commands the other. The verbal feature in the Wrst conjunct can nevertheless be argued to be closer to Fin if, instead, we simply deWne closeness in terms of the number of the intervening structural nodes (cf. Haeberli 2002c: 45 n. 23), Co’ being the additional intervening node between Fin and the Wnite verb in the second conjunct.

Clause Type Asymmetries in Old English

281

categories for coordination are always full CP structures, the OE word order asymmetries suggest that they can at least sometimes be AgrPs. Apart from this, we can adopt Johannessen’s proposals. For example, we can account for reduced second conjuncts in terms of her deletion approach. Furthermore, her claim (Johannessen 1998: ch. 6) that extraction out of a conjunct is possible allows us to deal with a remaining open issue concerning AgrP coordination. If, as proposed above, CoP is the complement of Fin in this case, we would expect the entire CP-domain above FinP to be activated as well. But, as observed in Haeberli (2001: 210–11), it is plausible to assume that some constituent always has to be fronted to the CP domain in OE main clauses (i.e. to the XP position in (6)), and the question therefore arises as to how this requirement can be satisWed with a CP above a CoP. Assuming that extraction out of a conjunct is possible, the answer to this question is simple. The requirements of the CP can be met through movement of the initial constituent in the Wrst AgrP conjunct to CP.9

16.6 The Loss of Verb Movement in the History of English Having discussed the analysis of clause type asymmetries in Old English, let us conclude by brieXy considering a possible diachronic consequence of our proposals. As often discussed, verb movement was lost in the history of English (cf. e.g. Roberts 1985; Kroch 1989) and this loss was the source of the rise of do-support. The standard analyses of the loss of V-movement in English assume that there was a unique underlying change in the grammar of English, namely loss of V-movement to an inXectional head. However, Han (2000) and Han and Kroch (2000) provide evidence suggesting that the loss of V-movement is actually (at least) a two-step process. First, at the beginning of the Wfteenth century, movement from one inXectional head to a higher inXectional head starts being lost. And second, at the end of the sixteenth century, movement from V to the inXectional domain starts being lost. Although this sequential loss scenario seems attractive given the evidence Han and Kroch provide, they leave this development unexplained. In particular, it remains unclear why the loss of V-movement between inXectional heads should have begun at the beginning of the Wfteenth century. 9 Among the basic distributional options for Wnite verbs discussed in Section 16.2, there is one that has not been discussed yet and that is the (rare) verb-Wnal order in non-conjoined main clauses. For completeness’s sake let us brieXy consider two options to account for this pattern:

(i) AgrP (like TP) has variable directionality and verb-Wnal non-conjoined main clauses have a head-Wnal AgrP. AgrP diVers from TP in that it is head Wnal very rarely. (ii) Main clauses optionally (but rarely) lack a CP layer. The verb then only moves to T (no feature checking with Fin) and can occur clause Wnally if TP is head Wnal.

282

E. Haeberli

The analysis discussed in this chapter may shed some light on this issue. The basic hypothesis to be made is that the presence of a feature on Fin that requires checking by the verb (as proposed for OE) is subject to parametric variation. Learners of OE received clear-cut evidence for such a feature in the form of the pronoun/full DP contrast in contexts of non-operator fronting discussed in Section 16.3. This contrast identiWes a V-position (i.e. Agr in (6)) between the subject position for pronouns and a lower subject position which can only be occupied by full DP subjects. This type of evidence for V-to-Agr and hence for the presence of a feature on Fin that attracts V thus depends on the ability of full DP subjects to remain in a lower position than subject pronouns. In the period following OE, i.e. in Early Middle English (EME), full DP subjects maintain this property and the proposals made in this chapter can be directly extended to EME (cf. Haeberli and Ingham forthcoming). However, this type of evidence is disappearing by around 1400. As argued in Haeberli (2002a, 2002b), full DP subjects cannot remain in Spec,TP any more in later Middle English because the Wllers of Spec,AgrP in OE/EME, i.e. empty expletives, are being lost in this period. Thus, full DP subjects have to move to AgrP as well. Subject-verb inversion in non-operator fronting contexts is therefore lost and subjects always precede the Wnite verb regardless of whether the subject is pronominal or non-pronominal. In other words, the crucial evidence for the feature on Fin which requires checking by V and, hence, the crucial evidence for V movement to Agr is lost. I propose that, as a consequence, the landing site of V movement in main clauses can be reanalysed as a lower inXectional head and arguably even must be reanalysed as a lower head for reasons of economy (assuming that less feature checking and, hence, less movement is more economical). Hence, the parameter value for Fin with respect to attraction of V is changed. This development has the eVect of destabilizing the V-movement system outlined before, and it can be argued to have led to the Wrst step in the loss of V movement around the beginning of the Wfteenth century, as suggested by Han and Kroch.

16.7 Conclusion In this chapter, I proposed an analysis of the clause type asymmetries found with respect to the distribution of Wnite verbs in Old English. Following a proposal made by Bobaljik and Thra´insson (1998) on verb movement in Icelandic subordinate clauses and developing their theoretical approach to the syntax of verb movement, I argued that Wnite verbs in Old English generally move only to T in subordinate clauses rather than to Agr/C as in main clauses. If we assume that TP has variable directionality in Old English

Clause Type Asymmetries in Old English

283

whereas AgrP/CP are head initial, the contrast in the landing site of verb movement can account for the higher frequency of verb-Wnal order in subordinate clauses as compared to main clauses. The frequency diVerences between conjoined and non-conjoined main clauses can also be accounted for in terms of this approach because second conjuncts trigger verb movement to T rather than to Agr when coordination involves two AgrPs. Finally, I sketched a potential consequence of the analysis developed here for the diachronic development of verb movement in the history of English.

References Abney, S. (1987). ‘The English NP in its Sentential Aspect’. Ph.D. dissertation, MIT. Ackerman, F. (1992). ‘Complex Predicates and Morphological Relatedness: Locative Alternation in Hungarian’, in I. Sag and A. Szabolcsi (eds.), Lexical Matters. Stanford, Calif.: CSLI Publications (CSLI Lecture Notes 24), 55–83. Adams, M. (1987). ‘From Old French to the Theory of Prodrop’, Natural Language and Linguistic Theory, 5: 1–32. Alexiadou, A., and Wilder, C. (1998). ‘Adjectival ModiWcation and Multiple Determiners’, in A. Alexiadou and C. Wilder (eds.), Possessors, Predicates and Movement in the DP. Amsterdam: Benjamins, 303–32. Andersen, H. (1973). ‘Abductive and Deductive Change’, Language, 49.4: 765–93. Anderson, S. R. (1996). ‘How to Put your Clitics in their Place, or Why the Best Account of Second-Position May be Something Like the Optimal One’, Linguistic Review, 13: 165–91. Andrade, J. C. (2002). ‘As construc¸o˜es causativas do portugueˆs do Brasil na perspectiva gerativa’. MA dissertation, Universidade de Brası´lia. Androutsopoulou, A. (1997). ‘Split DPs, Focus, and Scrambling in Modern Greek’, West Coast Conference on Formal Linguistics, 16, 1–16. Ashby, W. (1981). ‘The Loss of the Negative Particle ne in French: A Syntactic Change in Progress’, Language, 57: 674–87. Ayres-Bennet, W. (1994). ‘Negative Evidence: Or Another Look at the Non-Use of ne in 17th Century French’, French Studies, 48: 63–85. —— (2000). ‘Voices from the Past’, Romanische Forschungen: 323–48. —— (2002). ‘Socio-historical Linguistics and the History of French’, Journal of French Language Studies, 11.2: 159–78. Bacquet, P. (1962). La Structure de la phrase verbale a` l’e´poque alfre´dienne. Paris: Les Belles Lettres. Barbosa, P. (1993). ‘Clitic Placement in Old Romance and European Portuguese’, Papers from the Twenty-Ninth Regional Meeting of the Chicago Linguistic Society. Chicago: Chicago University Press. —— (1996). ‘Clitic Placement in European Portuguese and the Position of Subjects’, in A. L. Halpern and A. M. Zwicky (eds.), Second Position Clitics and Related Phenomena. Stanford, Calif.: CSLI Publications, 1–40. —— (2000). ‘Clitics: A Window into the Null Subject Property’, in Costa 2000: 31–93. Bartra, A., Brucart, J. M., and Rigau, G. (2002). ‘ ‘‘Per que` lo prec, si.l plats, que.m perdo´ de mos pecats’’: vige`ncia de l’alternanc¸a acusativa en catala`’, Estudi general, 22: 213–32. Also available at http://seneca.uab.es/ggt/reports.htm.

References

285

Batllori, M., and Roca, F. (2000). ‘The Value of Def Determiners from Old Spanish to Modern Spanish’, in S. Pintzuk, E. Tsoulas, and A. Warner 2000: 241–54. Belletti, A. (1999). ‘ ‘‘Inversion’’ as Focalization and Related Questions’. MS, University of Siena. —— (2004). ‘Aspects of the Low IP Area’ in L. Rizzi (ed.), The Structure of CP and IP, New York: Oxford University Press, 16–51. Bender, E. (2000). ‘The Syntax of Mandarin ba: Reconsidering the Verbal Analysis’, Journal of East Asian Linguistics, 9.2: 105–45. Beninc, P. (1984). ‘Un’ipotesi sulla sintassi delle lingue romanze medievali’, Quaderni Patavini di linguistica, 4: 3–19. —— (1988). ‘L’ordine degli elementi della frase e le costruzioni marcate’, in L. Renzi (ed.), Grande grammatica italiana di consultazione, Bologna: II Mulino, 129–94. —— (1995). ‘Complement Clitics in Medieval Romance: The Tobler–MussaWa Law’, in A. Battye and I. Roberts (eds.), Clause Structure and Language Change. New York: Oxford University Press, 325–44. —— (1996). ‘La struttura della frase esclamativa alla luce del dialetto padovano’, in P. Beninca`, G. Cinque, T. De Mauro, and N. Vincent (eds.), Italiano e dialetti nel tempo: Saggi di grammatica per Giulio Lepschy. Rome: Bulzoni, 23–43. —— (2001). ‘The Position of Topic and Focus in the Left Periphery’, in G. Cinque and G. Salvi (eds.), Current Studies in Italian Syntax: Essays OVered to Lorenzo Renzi. Amsterdam: Elsevier (North-Holland Linguistic Series), 39–64. —— (2003). ‘On the Functional Structure of the Left Periphery: Evidence from Medieval Romance’. MS, University of Padua. —— (2005). ‘On Some Descriptive Generalizations in Romance’, in G. Cinque and R. Kayne (eds.), The Oxford Handbook of Comparative Syntax. Oxford: Oxford University Press, ch. 6 —— and Poletto, C. (2004). ‘Topic, Focus and V2: DeWning the CP Sublayers’, in L. Rizzi (ed.), The Structure of CP and IP, New York: Oxford University Press, 52–75. Bennett, P. A. (1981). ‘The Evolution of Passive and Disposal Sentences’, Journal of Chinese Linguistics, 9.1: 61–89. Berlinck, R. (2000). ‘Complementos preposicionados: variac¸a˜o e mudanc¸a no portugueˆs brasileiro’, Letras, 56, 159–75. Biberauer, T., and Roberts, I. (2003). ‘The EPP and Word Order Change in Middle English’. Talk given at the 18th Comparative Germanic Syntax Workshop, University of Durham. Bispo, K. I. (forthcoming). ‘On the Syntax of Indirect Objects in Brazilian Portuguese’. Universidade de Brası´lia. Blanche-Benveniste, C., and Jeanjean, C. (1986). Le Franc¸ais parle´: transcription et e´dition. Paris: Didier. Blasco Ferrer, E. (1984). Grammatica storica del catalano e dei suoi dialetti con speciale risguardo all’algherese. Tu¨bingen: Gunter Narr Verlag. Bobaljik, J. D. (1995). ‘Morphosyntax: On the Syntax of Verbal InXection’. Ph.D. dissertation, MIT.

286

References

Bobaljik, J. D., and ThrÆinsson, H. (1998). ‘Two Heads Aren’t Always Better than One’, Syntax, 1: 37–71. Borer, H. (1984). Parametric Syntax. Dordrecht: Foris. Boƒkovic´ , Zˇ. (1996). ‘Selection and the Categorial Status of InWnitival Complements’, Natural Language and Linguistic Theory, 14: 269–304. —— (1997). The Syntax of NonWnite Complementation: An Economy Approach. Cambridge, Mass.: MIT Press. —— (1999). ‘On Multiple Feature Checking: Multiple wh-Fronting and Multiple Head Movement’, in S. D. Epstein and N. Hornstein (eds.), Working Minimalism. Cambridge, Mass.: MIT Press, 159–87. —— (2001). On the Nature of the Syntax–Phonology Interface: Cliticization and Related Phenomena. Amsterdam: North-Holland. Botelho-Pereira, M. Aˆ. (1974). ‘Aspectos da oposic¸a˜o modal indicativo-subjuntivo no portugueˆs contemporaˆneo’. MA dissertation, Universidade Federal do Rio de Janeiro. —— and Roncaratti, C. N. (1993). ‘O caso do sujeito em orac¸o˜es inWnitivas introduzidas por para no portugueˆs popular do Brasil’, DELTA, 9.1: 15–30. Boucher, P. (2003). ‘Determiner Phrases in Old and Modern French’, in M. Coene and Y. D’hulst (eds.), From NP to DP, i: The Syntax and Semantics of Noun Phrases. Amsterdam: John Benjamins, 47–70. Brunot, F., and Bruneau, C. (1949). Pre´cis de grammaire historique de la langue franc¸aise. Paris: Masson & Cie. Bruyn, A., Muysken, P., and Verrips, M. (1999). ‘Double-Object Constructions in the Creole Languages: Development and Acquisition’, in DeGraV (ed.): 329–73. Buridant, C. (2000). Grammaire nouvelle de l’ancien franc¸ais. Paris: E´ditions Sedes. Burzio, L. (1986). Italian Syntax: A Government-Binding Approach. Dordrecht: Kluwer. Butler, A., and Mathieu, E. (2004). The Syntax and Semantics of Split Constructions: A Comparative Study. Basingstoke: Palgrave-MacMillan. Cardinaletti, A. (1996). ‘Subjects and Clause Structure’, University of Venice Working Papers in Linguistics, 6.1: 55–95. —— and Roberts, I. (1991). ‘Clause Structure and X-Second’. MS, Universita` di Venezia and Universite´ de Gene`ve. Carlier, A., and Goyens, M. (1998). ‘De l’ancien franc¸ais au franc¸ais moderne: regression du degre´ ze´ro de la de´termination et restructuration du syste`me des articles’, CILL, 24.3–4: 77–112. Carstens, V. (2002). ‘Antisymmetry and Word Order in Serial Constructions’, Language, 78.1: 3–50. Cavalli Sforza, L. L. (1996). Geni, popoli e lingue. Milan: Adelphi. —— Menozzi, P., and Piazza, A. (1994). History and Geography of Human Genes. Princeton: Princeton University Press. Chao, Yuen Ren (1968). A Grammar of Spoken Chinese. Berkeley and Los Angeles: University of California Press.

References

287

Chen Xilong (1993). ‘On the Syntax of Serial Verb Constructions in Chinese’. Ph.D. dissertation, University of Georgia. Chomsky, N. (1964). ‘Current Issues in Linguistic Theory’, in J. Fodor and J. Katz (eds.), The Structure of Language. Englewood CliVs, NJ: Prentice Hall, 50–118. —— (1981). Lectures on Government and Binding. Dordrecht: Foris. —— (1986). Knowledge of Language: Its Nature, Origin and Use. New York: Praeger. —— (1994). ‘Bare Phrase Structure’, MIT Occasional Papers in Linguistics. Repr. in G. Webelhuth (ed.) (1995), Government and Binding Theory and the Minimalist Program. Oxford: Blackwell, 383–439. [Included in Chomsky 1995.] —— (1995). The Minimalist Program. Cambridge, Mass.: MIT Press. —— (2000). ‘Minimalist Inquiries: The Framework’, in R. Martin, D. Michaels, and J. Uriagereka (eds.), Step by Step: Essays on Minimalist Syntax in Honor of Howard Lasnik. Cambridge, Mass.: MIT Press, 89–155. —— (2001). ‘Derivation by Phase’, in M. Kenstowicz (ed.), Ken Hale: A Life for Language. Cambridge, Mass.: MIT Press, 1–52. Cinque, G. (1980). ‘On Extraction from NP in Italian’, Journal of Italian Linguistics, 5: 47–99. —— (1999). Adverbs and Functional Heads: A Cross-Linguistic Perspective. Oxford: Oxford University Press. Clark, R., and Roberts, I. (1993). ‘A Computational Model of Language Learnability and Language Change’, Linguistic Inquiry, 24.2: 299–345. Collins, C. (1993). ‘Topics in Ewe Syntax’. Ph.D. dissertation, MIT. —— (1997). ‘Argument Sharing in Serial Verb Constructions’, Linguistic Inquiry, 28.3: 461–97. Colussi, G. (1987). Glossario degli antichi volgari italiani (vol. iii/3). Helsinki: Helsinki University Press. Coromines, J. (1971). Lleures i converses d’un Wlo`leg. Barcelona: Club Editor. Costa, J. (1998). Word Order Variation: A Constraint-Based Approach. The Hague: Holland Academic Graphics. —— (ed.) (2000). Portuguese Syntax: New Comparative Studies. Oxford: Oxford University Press. —— (2003). ‘Null vs. Overt Spec,TP in European Portuguese’, in J. Quer, J. Schroten, M. Scorretti, P. Sleeman, and E. Verheugd (eds.), Romance Language and Linguistic Theory 2001: Selected Papers from ‘Going Romance’, Amsterdam, 6–8 December 2001. Amsterdam: John Benjamins, 31–47. —— and Martins, A. M. (2003). ‘Clitic Placement across Grammar Components’. Paper delivered at Going Romance 2003, Nijmegen, 20–2 November. Costa, M. J. (2001). ‘Contributos para o estudo das perı´frases verbais com ‘‘aver’’ e ‘‘teer’’ no portugueˆs medieval’, in C. N. Correia and A. Gonc¸alves (eds.), Actas do XVI Encontro Nacional da Associac¸a˜o Portuguesa de Linguı´stica Coimbra 2000. Lisbon: Associac¸a˜o Portuguesa de Linguı´stica, 179–86. Crisma, P. (2000). ‘Sintassi medievale e tipologia: l’articolo in inglese antico’, Archivio glottologico italiano, 85, Fasc. I: 38–84.

288

References

Crisma, P. (forthcoming). ‘Genitive Constructions in the History of English’, in G. Banti, P. Di Giovine, and P. Ramat (eds.), Typological Change in the Morphosyntax of the Indo-European Languages. Munich: Lincom Europa. Cui, Gui-bo (1984). ‘Zhuzi yulei de suo biaoxian de ji-ge baihua yufa xianxiang’ [Some Grammatical Phenomena of the Vernacular Language in the Zhuzi yulei]. MA thesis, National Taiwan University. Cummings, S., and Roberge, Y. (2003). ‘Null Objects in French and English’. Paper presented at LSRL, Indiana University. Cyrino, S. M. L. (1993). ‘Observac¸o˜es sobre a mudanc¸a diacroˆnica no portugueˆs do Brasil: objeto nulo e clı´ticos’, in Roberts and Kato 1993. —— (1997). O objeto nulo no portugueˆs do Brasil: um estudo sinta´tico-diacroˆnico. Londrina: Editora UEL. Demonte, V. (1991). Detra´s de la palabra. Madrid: Alianza Editorial. DeGraff, M. (ed.) (1999). Language Creation and Language Change: Creolization, Diachrony and Development. Cambridge, Mass.: MIT Press. Delfitto, D., and Schroten, J. (1991). ‘Bare Plurals and the Number AYx in DP’, Probus, 3.2: 155–85. de Murc¸ in (1815). Serments preˆte´s a` Strasbourg en 842 par Charles-le-Chauve, Louis-leGermanique et leurs arme´es respectives. Paris: Didot l’Aıˆne´. den Dikken, M. (1998). ‘Predicate Inversion in DP’, in A. Alexiadou and C. Wilder (eds.), Possessives, Predicates and Movement in the DP. Amsterdam: John Benjamins, 177–214. DØprez, V., and Martineau, F. (2004). ‘Microparametric Variation and Negative Concord’, in J. Auger, C. Clements, and B. Vance (eds.), Contemporary Approaches to Romance Linguistics. Amsterdam: John Benjamins, 139–58. Deutjes, J., et al. (2003). ‘Left Dislocation and Topicalization in French’. Talk delivered at the Berlin Workshop on the Left Periphery, Berlin, 29–30 November. Devine, A. M., and Stephens, L. D. (2000). Discontinuous Syntax: Hyperbaton in Greek. Oxford: Oxford University Press. Diesing, M. (1992). IndeWnites. Cambridge, Mass.: MIT Press. Di Sciullo, A.-M. (1996). ConWgurations: Essays on Structure and Interpretation. Somerville: Cascadilla Press. Dixon, R. M. W. (1971). ‘A Method of Semantic Description’, in D. D. Steinberg and L. A. Jakobovitz (eds.), Semantics. London: Cambridge University Press, 436–70. Djamouri, R. (1988). ‘E´tude des formes syntaxiques dans les e´crits oraculaires grave´s sur os et carapaces de tortue’. Ph.D. dissertation, EHESS, Paris. Dobrovie-Sorin, C. (2001). ‘Ge´nitifs et de´terminants’, in G. Kleiber, B. Laca, and L. Tasmowski (eds.), Typologie des groupes nominaux. Rennes: Presses universitaires de Rennes 2, 205–34. Dresher, B. E. (1999). ‘Charting the Learning Path: Cues to Parameter Setting’, Linguistic Inquiry, 30: 27–67. Duarte, E. L. (1993). ‘Do pronome nulo ao pronome pleno: a trajeto´ria do sujeito no portugueˆs do Brasil’, in Roberts and Kato 1993.

References

289

—— (2000). ‘The Loss of the Avoid Pronoun Principle in Brazilian Portuguese’, in M. Kato and E. Negra˜o (eds.), Brazilian Portuguese and the Null Subject Parameter. Frankfurt: Vervuert-IberoAmericana, 17–36. Duarte, M. E. (1992). ‘A perda da ordem Verbo Sujeito em interrogativas-qu no portugueˆs do Brasil’, DELTA 8, special issue: 37–52. Dufresne, M., Dupuis, F., and Longtin, C. M. (2001). ‘Un changement dans la diachronie du franc¸ais: la perte de la pre´Wxation aspectuelle en a-’, Revue que´be´coise de linguistique, 29.2: 33–54. —— —— and Tremblay, M. (2003). ‘Preverbs and Particles in Old French’, in G. Booij (ed.), Yearbook of Morphology. Dordrecht: Kluwer Academic Publishers, 30–60. Dupuis, F. (1989). ‘L’Expression du sujet dans les subordonne´es en ancien franc¸ais’. Ph.D. dissertation, Universite´ de Montre´al. Embick, D., and Halle, M. (2003). ‘Latin InXections’. Paper delivered at Going Romance 2003, Nijmegen, 20–2 November. —— and Noyer, R. (2001). ‘Movement Operations after Syntax’, Linguistic Inquiry, 32: 555–95. Emonds, J. (1976). A Transformational Approach to English Syntax. New York: Academic Press. —— (1993), ‘Projecting Indirect Objects’, Linguistic Review, 10.3: 211–63. Epstein, R. (1995). ‘L’Article de´Wni en ancien franc¸ais: l’expression de la subjectivite´’, Langue franc¸aise, 107: 58–71. Fanselow, G. (1988). ‘Aufspaltung von NPn und das Problem der ‘‘freien Wortstellung’’ ’, Linguistische Berichte, 115: 91–113. Ferguson, K. S. (1996). ‘Shortest Move and Object Case Checking’, in W. Abraham et al. (eds.), Minimal Ideas. Amsterdam: Benjamins, 97–111. Fischer, O. (1988). ‘The Rise of ‘‘for NP to V’’ Constructions: An Explanation’, in G. Nixon and J. Honey (eds.), An Historical Tongue: Studies on English Linguistics in Memory of Barbara Strang. London: Routledge, 67–88. —— van Kemenade, A., Koopman, W., and van der Wurff, W. (2000). The Syntax of Early English. Cambridge: Cambridge University Press. Fischer, S. (2002). The Catalan Clitic System: A Diachronic Perspective on its Syntax and Phonology. Berlin: Mouton de Gruyter. —— (forthcoming). ‘Stylistic Fronting: A Contribution to Information Structure’, in G. Fanselow and R. Vogel (eds.), MLC-EVects in Syntax. Berlin: Mouton de Gruyter. Folgar, C. (1993). Diacronı´a de los objetos directo e indirecto (del latı´n al castellano medieval): verba, anuario galego de Wloloxı´a, Anexo 37. Santiago de Compostela: Universidade de Santiago de Compostela. Fontana, J. M. (1993). ‘Phrase Structure and the Syntax of Clitics in the History of Spanish’. Ph.D. dissertation, University of Pennsylvania. —— (1996). ‘Phonology and Syntax in the Interpretation of the Tobler–MussaWa Law’, in A. Halpern and A. Zwicky (eds.), Approaching Second: Second Position Clitics and Related Phenomena. Stanford, Calif.: CSLI Publications, 41–83.

290

References

Fontana, J. M. (1997). ‘On the Integration of Second Position Phenomena’, in A. van Kemenade and N. Vincent (eds.), Parameters of Morphosyntactic Change. Cambridge: Cambridge University Press, 207–49. Foster, T., and van der Wurff, W. (1995). ‘From Syntax to Discourse: The Function of the Object-Verb Order in Later Middle English’, in J. Fisiak (ed.), Studies in Middle English Linguistics. Berlin: Mouton de Gruyter, 135–56. Foulet, L. (1928, 1974, 1977, 1990). Petite Syntaxe de l’ancien franc¸ais. Paris: Librairie Honore´ Champion. Fournier, N. (1998). Grammaire du franc¸ais classique. Paris: Belin. Fourquet, J. (1938). L’Ordre des ´ele´ments de la phrase en germanique ancien: e´tudes de syntaxe de position. Paris: Les Belles Lettres. Franks, S. (1999). ‘Clitics at the Interface: An Introduction to Clitic Phenomena in European Languages’, in F. Beukema and M. den Dikken (eds.), Clitic Phenomena in European Language. Amsterdam: John Benjamins, 1–46. Fuss, E., and Trips, C. (2002). ‘Variation and Change in Old and Middle English: On the Validity of the Double Base Hypothesis’, Journal of Comparative Germanic Linguistics, 4: 171–224. Galves, C. (2000). ‘Agreement, Predication and Pronouns in the History of Portuguese’, in Costa 2000: 143, 168. —— (2001). Ensaios sobre as grama´ticas do portugueˆs. Campinas: Editora da Unicamp. —— (2003). ‘InWnitivo Xexionado e legitimac¸a˜o de agr’, Ensaios sobre as grama´ticas do portugue´s. Campinas: Editora da UNICAMP, 181–94. Gavruseva, E. (2000). ‘On the Syntax of Possessor Extraction’, Lingua, 110: 743–72. Geenhoven, V. van (1998). Semantic Incorporation and IndeWnite Descriptions: Semantic and Syntactic Aspects of Noun Incorporation in West Greenlandic. Stanford, Calif.: CSLI. Giacalone Ramat, A. (1992). ‘The Pairing of Structure and Function in Syntactic Development’, in M. Gerritsen and D. Stein (eds.), Internal and External Factors in Syntactic Change. Berlin: Mouton de Gruyter, 317–39. Gianollo, C. (forthcoming). ‘Struttura e storia del sintagma nominale latino’. Ph.D. dissertation, Universita` di Pisa. Giorgi, A., and Longobardi, G. (1991). The Syntax of Noun Phrases: ConWguration, Parameters and Empty Categories. Cambridge: Cambridge University Press. Giusti, G. (1993). La sintassi dei determinanti. Padua: Unipress. Goldberg, A. E. (1995). Constructions: A Construction Grammar Approach to Argument Structure. Chicago: Chicago University Press. Gomes, C. A. (2003). ‘Variac¸a˜o e mudanc¸a na expressa˜o do dativo no portugue´s brasileiro’, in M. C. Paiva and M. E. L. Duarte (eds.), Mudanc¸a lingu¨´ıstica em tempo real. Rio de Janeiro: FAPERJ/ContraCapa, 81–96. Goria, C. (2002). ‘The Complexity of the Left Periphery’, Syntax, 5: 89–115. Gougenheim, G. (1984). Grammaire de la langue franc¸aise du seizie`me sie`cle. Paris: A. and J. Picard.

References

291

Goyens, M. (1994). E´mergence et e´volution du syntagme nominal en franc¸ais. Bern: Peter Lang Greenberg, J. (1978). ‘How Does a Language Acquire Gender Markers’, in J. Greenberg (ed.), Universals of Human Language, vol. iii. Stanford, Calif.: Stanford University Press, 47–82. —— (1987). Language in the Americas. Stanford, Calif.: Stanford University Press. —— (2000). Indoeuropean and its Closest Relatives: The Eurasiatic Language Family. Stanford, Calif.: Stanford University Press. Grewendorf, G., and Sabel, J. (1999). ‘Scrambling in German and Japanese: Adjunction versus Multiple SpeciWers’, Natural Language and Linguistic Theory, 17: 1–65. Grieve, A. (1984). ‘Remarques sur l’histoire du franc¸ais parle´’, CILL, 10: 1–3, 65–76. Grosu, A. (1974). ‘On the Nature of the Left Branch Condition’, Linguistic Inquiry, 5: 308–19. Guardiano, C. (2003). ‘Struttura e storia del sintagma nominale nel Greco Antico: ipotesi parametriche’. Ph.D. dissertation, Universita` di Pisa. —— Gianollo, C., and Longobardi, G. (forthcoming). ‘A Guide to Nominal Parameters’. —— and Longobardi, G. (forthcoming). ‘Parametric Syntax as a Source of Historical-Comparative Generalisations’, in G. Banti, P. Di Giovine, and P. Ramat (eds.), Typological Change in the Morphosyntax of the Indo-European Languages. Munich: Lincom Europa. Haase, A. (1969). Syntaxe franc¸aise du XVIIe sie`cle. Paris: Delagrave. Haeberli, E. (2000). ‘Adjuncts and the Syntax of Subjects in Old and Middle English’, in Pintzuk, Tsoulas, and Warner 2000: 109–31. —— (2001). ‘Speculations on the Syntax of Subordinate Clauses in Old English’, Reading Working Papers in Linguistics, 5: 201–29 (www.rdg.ac.uk/AcaDepts/cl/ slals/workingpapers/haeberli.pdf). —— (2002a). ‘Observations on the Loss of Verb Second in the History of English’, in J.-W. Zwart and W. Abraham (eds.), Studies in Comparative Germanic Syntax. Amsterdam: John Benjamins, 245–72. —— (2002b). ‘InXectional Morphology and the Loss of Verb Second in English’, in Lightfoot 2002c : 88–106. —— (2002c). Features, Categories and the Syntax of A-Positions: Cross-Linguistic Variation in the Germanic Languages. Dordrecht: Kluwer. —— and Haegeman, L. (1995). ‘Clause Structure in Old English: Evidence from Negative Concord’, Journal of Linguistics, 31: 81–108. —— and Ingham, R. (forthcoming). ‘The Position of Negation and Adverbs in Early Middle English’, Lingua. Haegeman, L. (1997a). ‘Negative Inversion and the Structure of CP’, paper presented at the Linguistic Colloquium, University of Wuppertal. —— (ed.) (1997b). Elements of Grammar: Handbook in Generative Syntax. Dordrecht: Kluwer.

292

References

Haegeman, L. (2000). ‘Negative Preposing, Negative Inversion, and the Split CP’, in L. R. Horn and Y. Kato (eds.), Negation and Polarity. Oxford: Oxford University Press, 21–61. Hale, K., and Keyser, S. J. (1993). ‘On Argument Structure and the Lexical Expression of Syntactic Relations’, in K. Hale and S. J. Keyser (eds.), The View from Building 20: A Festschrift for Sylvain Bromberger. Cambridge, Mass.: MIT Press. —— —— (1997). ‘On the Complex Nature of Simple Predicators’, in A. Alsina, J. Bresnan, and P. Sells (eds.), Complex Predicates. Stanford, Calif.: CSLI Publications (CSLI Lecture Notes, 64), 29–95. —— —— (1998). ‘The Basic Elements of Argument Structure’, MITWPL 32 (Papers from the UPenn/MIT Roundtable on Argument Structure and Aspect): 73–118. Hale, M. (1998). ‘Diachronic Syntax’, Syntax 1: 1–18. Halle, M., and Marantz, A. (1993). ‘Distributed Morphology and the Pieces of InXexion’, in K. Hale and S. J. Keyser (eds.), The View from Building 20: Essays in Linguistics in Honor of Sylvain Bromberger. Cambridge, Mass.: MIT Press, 111–76. Halpern, A. (1995). On the Placement and Morphology of Clitics. Stanford, Calif.: CSLI Publications. —— and Fontana, J. M. (1994). ‘X0 Clitics and Xmax Clitics’, Proceedings of the West Coast Conference on Formal Linguistics WCCFL XII. Stanford, Calif.: CSLI Publications. Hamacher, E. (2002). ‘La Pre´Wxation verbale en ancien franc¸ais: le recul d’un proce´de´ morphologique’. MA thesis, Queen’s University. Han, C. (2000). ‘The Evolution of do-Support in English Imperatives’, in Pintzuk, Tsoulas, and Warner 2000: 275–95. —— and Kroch, A. (2000). ‘The Rise of do-Support in English: Implications for Clause Structure’, paper presented at NELS 30. Harley, H. (2002). ‘ACD, WCO and QR of DPs’, Linguistic Inquiry, 33.4: 659–64. —— and Noyer, R. (1999). ‘Distributed Morphology’, Glot International, 4.4: 3–9. Harris, M. (1978). The Evolution of French Syntax: A Comparative Approach. London: Longman. Hashimoto, A. (1971). ‘Mandarin Syntactic Structures’, Unicorn, 8: 1–149. Hashimoto, M. J. (1988). ‘The Structure and Typology of the Chinese Passive Construction’, in M. Shibatani (ed.), Passive and Voice. Amsterdam: Benjamins, 329–54. Haspelmath, M. (1995). ‘Diachronic Sources of ‘‘All’’ and ‘‘Every’’ ’, in E. Bach, E. Jelinek, A. Kratzer, and B. Partee (eds.), QuantiWcation in Natural Languages. Dordrecht: Kluwer, 363–82. Hausmann, F.-J. (1992). ‘L’Aˆge du franc¸ais parle´ actuel: bilan d’une controverse allemande. Grammaire des fautes et franc¸ais non conventionnels’, Actes du 4e colloque international organise´ par le Groupe d’E´tude en Histoire de la Langue Franc¸aise. Paris: Presses de l’E´cole Normale Supe´rieure, 355–62. Higginbotham, J. (1986). ‘IndeWniteness and Predication’, in E. Reuland and A. Ter Meulen (eds.), The Representation of IndeWniteness. Cambridge, Mass.: MIT Press, 43–70.

References

293

Hirschbhler, P., and Junker, M.-O. (1988). ‘Remarques sur les sujets nuls en subordonne´es en ancien et en moyen franc¸ais’, Revue que´be´coise de linguistique the´orique et applique´e, 7.3: 63–84. —— and Labelle, M. (2000). ‘Evolving TM EVects in the Placement of French Clitics’, in S. Dworkin and D. Wanner (eds.), New Approaches to Old Problems: Issues in Romance Historical Linguistics. Amsterdam: John Benjamins, 165–82. —— —— (2001). ‘La Position des clitiques par rapport au verbe a` l’impe´ratif dans l’e´volution du franc¸ais’, RLV 30. E´tudes diachroniques: 13–37. Holmberg, A. (1986). ‘Word Order and Syntactic Features in the Scandinavian Languages and English’. Ph.D. dissertation, University of Stockholm. Hopper, P. J., and Traugott, E. C. (1993/2003). Grammaticalization. Cambridge: Cambridge University Press. Horrocks, G., and Stavrou, M. (1987). ‘Bounding Theory and Greek Syntax: Evidence for wh-Movement in NP’, Journal of Linguistics, 23: 79–108. Huang, C.-T. James (1982). ‘Logical Relations in Chinese and the Theory of Grammar’. Ph.D. dissertation, MIT. —— (1999). ‘Chinese Passives in Comparative Perspective’, Tsing Hua Journal of Chinese Studies (New Series), 29.4: 423–509. Hulk, A., and van Kemenade, A. (1997). ‘Negation as a ReXex of Clause Structure’, in D. Forget, P. Hirschbu¨hler, F. Martineau, and M.-L. Rivero (eds.), Negation and Polarity: Syntax and Semantics. Amsterdam: John Benjamins, 183–207. Jackendoff, R. (1990). Semantic Structures. Cambridge, Mass.: MIT Press. Jaeggli, O., and Safir, K. (eds.) (1989). The Null Subject Parameter. Dordrecht: Kluwer. Jarad, N. (1997). ‘The Origin and Development of for-InWnitives’. Ph.D. dissertation, University of Wales. Jensen, F. (1990). Old French and Comparative Gallo-Romance Syntax. Tu¨bingen: Niemeyer. —— (1994). Syntaxe de l’Ancien Occitan. Tu¨bingen: Max Niemeyer Verlag. Jespersen, O. (1917). Negation in English and Other Languages. Copenhagen: A. F. Host. Johannessen, J. B. (1998). Coordination. Oxford: Oxford University Press. Joseph, B., and Salmons, J. C. (eds.) (1998). Nostratic, Sifting the Evidence. Amsterdam: Benjamins. Kayne, R. (1989). ‘Null Subjects and Clitic Climbing’, in Jaeggli and SaWr 1989: 239–61. —— (1991). ‘Romance Clitics, Verb Movement and PRO’, Linguistic Inquiry, 22: 647–86. [Repr. in Kayne 2000a.] —— (1994). The Antisymmetry of Syntax. Cambridge, Mass.: MIT Press. —— (2000a). Parameters and Universals. Oxford: Oxford University Press. —— (2000b). ‘Prepositional Complementizers as Attractors’, in Kayne 2000a: 282–313. —— (2001). ‘Prepositions as Probes’. MS, New York University.

294

References

Kayne, R. (2002). ‘On Some Prepositions that Look DP-internal: English of and French de’, Catalan Journal of Linguistics, 1: 70–115. Keenan, E. (1994). ‘Creating Anaphors. An Historical Study of the English ReXexive Pronouns’. MS, UCLA. van Kemenade, A. (1987). Syntactic Case and Morphological Case in the History of English. Dordrecht: Foris. —— (1997). ‘V2 and Embedded Topicalization in Old and Middle English’, in A. van Kemenade and N. Vincent 1997: 326–52. —— and Vincent, N. (eds.) (1997). Parameters of Morphosyntactic Change. Cambridge: Cambridge University Press. Kempchinsky, P.-M. (1986). ‘Romance Subjunctive Clauses and Logical Form’. Ph.D. dissertation, University of California. Keniston, H. (1937). The Syntax of Castilian Prose: The Sixteenth Century. Chicago: University of Chicago Press. Kenstowicz, M. (ed.) (2001). Ken Hale: A Life in Language. Cambridge, Mass.: MIT Press. Kerstens, J. (1993). The Syntax of Number, Person and Gender: A Theory of PhiFeatures. Berlin: Mouton de Gruyter. Kester, E.-P. (1996). The Nature of Adjectival InXection. Utrecht: OTS. Kiparsky, P. (1995). ‘Indo-European Origins of Germanic Syntax’, in A. Battye and I. Roberts (eds.), Clause Structure and Language Change. Oxford: Oxford University Press, 140–69. —— (1996). ‘The Shift to Head-Initial VP in Germanic’, in H. Thra´insson, S. D. Epstein, and S. Peter (eds.), Studies in Comparative Germanic Syntax, vol. ii (Studies in Natural Language and Linguistic Theory, Volume 38). Dordrecht: Kluwer, 140–79. Klipple, E. (1997). ‘Prepositions and Variation’, in A.-M. Di Sciullo (ed.), Projections and Interface Conditions: Essays on Modularity. New York: Oxford University Press, 74–108. Knittel, M.-L. (2002). ‘Existe-t-il un DP en turc?’, Recherches linguistiques de Vincennes, 31: 79–96. Kok, A. de (1985). La Place du pronom personnel re´gime conjoint en franc¸ais: une e´tude diachronique. Amsterdam: Rodopi. Koopman, W. (1995). ‘Verb-Final Main Clauses in Old English Prose’, Studia Neophilologica, 67: 129–44. —— (1998). ‘Inversion after Single and Multiple Topics in Old English’, in J. Fisiak and M. Krygier (eds.), Advances in English Historical Linguistics. Berlin: Mouton de Gruyter, 135–49. Kroch, A. (1989). ‘ReXexes of Grammar in Patterns of Language Change’, Journal of Language Variation and Change, 1: 199–244. —— (1994). ‘Morphological Variation’, in K. Beals et al. (eds.), Papers from the 30th Regional Meeting of the Chicago Linguistic Society: Parasession on Variation and Linguistic Theory, 180–201.

References

295

Kroch, A. (2000). ‘Syntactic Change’, in. M. Baltin and C. Collins (eds.), The Handbook of Contemporary Syntactic Theory. Oxford: Blackwell. —— and Taylor, A. (1997). ‘Verb Movement in Old and Middle English: Dialect Variation and Language Contact’, in A. van Kemenade and N. Vincent 1997: 297–325. —— —— (2000). ‘Verb-Complement Order in Middle English’, in Pintzuk, Tsoulas, and Warner 2000: 132–63. Khner, R., and Stegman, C. (1914). Ausfu¨hrliche Grammatik der lateinischen Sprache. Hanover: Verlag Hahnsche Buchhandlung. [ModiWed versions from 1955, 1976; repr. 1992.] Laka, I. (1990). ‘Negation in Syntax: On the Nature of Functional Categories and Projections’. Ph.D. dissertation, MIT. Larjavaara, M. (2000). Pre´sence ou absence de l’objet: limites du possible en franc¸ais contemporain. Helsinki: Academia Scientiarum Fennica. Larson, R. (1991). ‘Some Issues in Verb Serialization’, in C. Lefebvre (ed.), Serial Verbs: Grammatical, Comparative and Cognitive Approaches. Amsterdam: Benjamins, 185–210. Larson, R. K. (1988). ‘On the Double Object Construction’, Linguistic Inquiry, 19.3: 335–91. Lasnik, H. (1999). Syntactic Structures Revisited: Contemporary Lectures on Classic Transformational Grammar. Cambridge, Mass.: MIT Press. Legendre, G. (1996). ‘Clitics, Verb (Non-movement, and Optimality in Bulgarian)’, Technical Report JHU CogSci-96–5. Baltimore: Johns Hopkins University. —— (1998). ‘Second Position Clitics in a Verb-Second Language: ConXict Resolution in Macedonian’, in J. Austin and A. Lawson (eds.), ESCOL ’97. Proceedings of the Eastern States Conference on Linguistics. Ithaca, NY: Cornell University, CLC Publications, 139–49. Lehmann, C. (1995). Thoughts on Grammaticalization: Revised and Expanded Version. Munich: Lincom Europa. —— (2002). ‘New ReXections on Grammaticalization and Lexicalization’, in I. Wischer and G. Diewald (eds.), New ReXections on Grammaticalization. Amsterdam: John Benjamins, 1–18. Lemle, M., and Naro, A. J. (1977). Competeˆncias ba´sicas do portugueˆs. Relato´rio Wnal de pesquisa apresentado ao MOBRAL e Fundac¸a˜o FORD. Rio de Janeiro: MOBRAL (CETEP/SEPES). Li, C., and Thompson, S. (1973). ‘Serial Verb Constructions in Mandarin Chinese: Subordination or Coordination?’, in C. Corum, T. C. Smith-Stark, and A. Weiser (eds.), You Take the High Node and I’ll Take the Low Node: Papers from the Comparative Syntax Festival. Chicago: Chicago Linguistic Society, 96–103. —— —— (1974). ‘An Explanation of Word Order Change SVO > SOV’, Foundations of Language, 12: 201–14. Li Jinxi and Shiru Liu (1955). Zhongguo yufa jiaocai [Teaching materials on Chinese grammar]. Beijing: Wushinian chubanshe. Li, Y.-H. Audrey (1985). ‘Abstract Case in Chinese’. Ph.D. dissertation, USC.

296

References

Li, Y.-H. Audrey (1990). Order and Constituency in Mandarin Chinese. Dordrecht: Kluwer. —— (2001). ‘The ba Construction’. MS, available at www.usc.edu/dept/LAS/ealc/ chinling. Lightfoot, D. (1979). Principles of Diachronic Syntax. Cambridge: Cambridge University Press. —— (1991). How to Set Parameters: Arguments from Language Change. Cambridge, Mass.: MIT Press. —— (1999). The Development of Language: Acquisition, Change and Evoution. Cambridge, Mass.: Blackwell. —— (2002a). ‘Myths and the Prehistory of Grammars’, Journal of Linguistics, 38: 113–36. —— (2002b). ‘Introduction’, in Lightfoot 2002c: 1–19. —— (ed.) (2002c). Syntactic EVects of Morphological Change. Oxford: Oxford University Press. —— and Rodrigues, C. (2002). ‘Subject Inversion and Clitic Climbing in Romance’. Paper delivered at the 7th Diachronic Generative Syntax Conference, Girona, 27–9 June. Lobato, L. (1988). ‘Sobre a regra de anteposic¸a˜o do verbo no portugueˆs do Brasil’, DELTA, 4.1: 121–47. Lobeck, A. (1995). Ellipsis: Functional Heads, Licensing and IdentiWcation. Oxford: Oxford University Press. —— (1999). ‘VP Ellipsis and the Minimalist Program: Some Speculation and Proposals’, in S. Lappin and E. Benmamoum (eds.), Fragments: Studies in Ellipsis and Gapping. New York: Oxford University Press, 98–123. Longa, V. M. (1994). ‘The Galician InXected InWnitive and the Theory of UG’, Catalan Working Papers in Linguistics, 4.1: 23–44. Longobardi, G. (1994). ‘Reference and Proper Names: A Theory of N-Movement in Syntax and Logical Form’, Linguistic Inquiry, 25: 609–65. —— (2000). ‘The Structure of DPs: Principles, Parameters and Problems’, in M. Baltin and C. Collins (eds.), Handbook of Syntactic Theory. Cambridge, Mass.: Blackwell, 562–603. —— (2001a). ‘On Parameters and Parameter Theory’. MS, Universita` di Trieste. To appear in U. Wandruszka and E. Stark (eds.), Akten des Romanistentages Mu¨nchen 2001. —— (2001b). ‘Formal Syntax, Diachronic Minimalism and Etymology: The History of French chez’, Linguistic Inquiry, 32.2: 275–302. —— (2003a). ‘Determinerless Nouns: A Parametric Mapping Theory’, in M. Coene and Y. D’hulst (eds.), From NP to DP, i: The Syntax and Semantics of Noun Phrases. Amsterdam: John Benjamins, 239–54. —— (2003b). ‘Methods in Parametric Linguistics and Cognitive History’. MS, Universita` di Trieste.

References

297

Los, B. (1999). InWnitival Complementation in Old and Middle English. The Hague: Thesus. L Shuxiang et al. (eds.) (1980/1995). Xiandai hanyu babaici [Eight Hundred Words of Modern Chinese]. Beijing: Shangwu chubanshe. Lyons, C. (1999). DeWniteness. Cambridge: Cambridge University Press. Ma Beijia (2003). ‘Zai hanyu li shi fenxi zhong ruhe qufen dongci he jieci [Methods to Distinguish between Verbs and Prepositions in the Historical Analysis of Chinese]’, Zhongguo yuwen, 2003.1: 59–65. McCawley, J. (1992). ‘Justifying Part of Speech Assignments in Mandarin Chinese’, Journal of Chinese Linguistics, 20.2: 211–45. McMahon, A., and McMahon, R. (2003). ‘Finding Families: Quantitative Methods in Language ClassiWcation’, Transactions of the Philological Society, 101.1: 7–55. Madeira, A. M. (1995). ‘Topics in Portuguese Syntax: The Licensing of T and D’. Ph.D. dissertation, University College London. Manolessou, I. (2000). ‘Greek Noun Phrase Structure: A Study in Syntactic Evolution’. Ph.D. dissertation, University of Cambridge. Marchello-Nizia, C. (1995). L’E´volution du franc¸ais: ordre des mots, de´monstratifs, accent tonique. Paris: Colin. —— (1997). La Langue franc¸aise aux XIVe et XVe sie`cles. Paris: Nathan Universite´. Martin, R. (1971). Temps et aspect: essai d’interpre´tation des temps narratifs en franc¸ais. Paris: Klincksieck. —— and Wilmet, M. (1980). Manuel du franc¸ais du moyen aˆge. ii: Syntaxe du moyen franc¸ais. Bordeaux: Sobodi. Martineau, F. (2003). ‘Un corpus de textes franc¸ais pour l’analyse de la variation diachronique et dialectale’, Lexicometrica, www.cavi.univ-paris3.fr/lexicometrica/ thema/thema7/Texte-Martineau.pdf. —— and Deprez, V. (forthcoming). ‘Pas aucun/pas rien en franc¸ais classique: variation dialectale et historique’. Langue franc¸aise. —— and Mougeon, R. (2003). ‘Sociolinguistic Research on the Origins of ne Deletion in European and Quebec French’, Language, 79.1: 118–52. Martins, A. M. (1994a). ‘Clı´ticos na histo´ria do portugueˆs’. Ph.D. dissertation, University of Lisbon. —— (1994b). ‘Enclisis, VP-Deletion and the Nature of Sigma’, Probus, 6: 173–205. —— (2000a). ‘A Minimalist Approach to Clitic Climbing’, in Costa 2000: 169–90. [Repr., with minor revisions, from A. M. Martins (1995). ‘A Minimalist Approach to Clitic Climbing’, Papers from the 31st Regional Meeting of the Chicago Linguistic Society, 215–39.] —— (2000b). ‘Polarity Items in Romance: UnderspeciWcation and Lexical Change’, in Pintzuk, Tsoulas, and Warner 2000: 191–219. —— (2001). Documentos portugueses do noroeste e da regia˜o de Lisboa: da produc¸a˜o primitiva ao se´culo XVI. Lisbon: Imprensa Nacional—Casa da Moeda. —— (2002). ‘The Loss of IP-Scrambling in Portuguese: Clause Structure, Word Order Variation and Change’, in Lightfoot 2002c : 232–48.

298

References

Martins, A. M. (2003a). ‘DeWcient Pronouns and Linguistic Change in Portuguese and Spanish’, in J. Quer, J. Schroten, M. Scorretti, P. Sleeman, and E. Verheugd (eds.), Romance Language and Linguistic Theory 2001: Selected Papers from ‘Going Romance’, Amsterdam, 6–8 December 2001. Amsterdam: John Benjamins, 213–30. —— (2003b). ‘From Unity to Diversity in Romance Syntax: Portuguese and Spanish’, in K. Braunmu¨ller and G. Ferraresi (ed.) Multilingualism in European Language History. Amsterdam: John Benjamins, 201–33. Mateu, J. (2001a). ‘Preverbs in Complex Denominal Verbs: Lexical Adjuncts or Core Predicates?’, Catalan Working Papers in Linguistics, 9: 37–51. —— (2001b). ‘Small Clause Results Revisited’, in N. Zhang (ed.), ZAS Papers in Linguistics, 26: Proceedings of the Workshop on the Syntax of Predication. —— (2002). ‘Argument Structure: Relational Construal at the Syntax-Semantics Interface’. Ph.D. dissertation, Universitat Auto`noma de Barcelona. —— and Rigau G. (2002), ‘A Minimalist Account of ConXation Processes. Parametric Variation at the Lexicon–Syntax Interface’, in A. Alexiadou (ed.), Theoretical Approaches to Universals. Amsterdam: John Benjamins, 211–36. Mateus, M. H. M., et al. (2003). Grama´tica da lı´ngua portuguesa. Lisbon: Caminho. Mathieu, E. (2002). ‘The Syntax of Non-Canonical QuantiWcation: A Comparative Study’. Ph.D. dissertation, University College London. —— (2003). ‘French Topics’. MS, University of Cyprus. —— (2004). ‘The Mapping of Form and Interpretation: the Case of Optional whMovement in French’, Lingua, 114.9–10: 1090–132. —— and Sitaridou, I. (2002). ‘Split DP Syntax and Defocalised Nominals: A View on the Syntax–Discourse Interface’. Talk presented at GLOW 2002, Workshop on ‘The Syntax-Discourse Interface: Linguistic and Psycholinguistic Issues’, 5–7 April,. Utrecht. —— —— (2003). ‘Split wh Constructions in Classical and Modern Greek’, in A. Alexiadou, S. Fischer, and M. Stavrou (eds.), Proceedings of the Workshop on Language Change and Generative Grammar (Aristotle University of Thessaloniki). Linguistics in Potsdam, 19: 143–82. —— —— (2004). ‘Noun Ellipsis, Determiner Spreading and Split-DPs’. Submitted. Matisoff, J. (1990). ‘On Megalocomparison’, Language, 66.1: 106–20. Meillet, A. (1912). ‘L’E´volution des formes grammaticales’, Scientia (Rivista di scienza), 12. 26.6. Reprinted in Antoine Meillet, Linguistique historique et linguistique ge´ne´rale. Paris, Champion, 1958, 130–48. —— (1958). Linguistique historique et linguistique ge´ne´rale. Paris: Champion. MØnard, P. (1976). Manuel du franc¸ais du moyen aˆge: syntaxe de l’ancien franc¸ais. Bordeaux: Sobodi. Miller, P. (1991). ‘Clitics and Constituents in Phrase Structure Grammar’. Ph.D. dissertation, Rijksuniversiteit te Utrecht. Mitchell, B. (1985). Old English Syntax. Oxford: Clarendon Press. Moignet, G. (1973). Grammaire de l’ancien franc¸ais: morphologie—syntax. Paris: Klincksieck.

References

299

Morais, M. A. T. (1993). ‘Aspectos diacroˆnicos do movimento do verbo, estrutura da frase e caso nominativo no portugueˆs do Brasil’, in Roberts and Kato 1993, 263–306. Moreno Cabrera, J. C. (1998). ‘On the Relationships between Grammaticalization and Lexicalization’, in A. G. Ramat and P. J. Hopper (eds.), The Limits of Grammaticalization. Amsterdam: John Benjamins, 211–27. Moro, A. (1997). The Raising of Predicates: Predicative Noun Phrases and the Theory of Clause Structure. Cambridge: Cambridge University Press. Mouma, E. (1993). ‘On Some Properties of DPs in Modern Greek’, UCL Working Papers in Linguistics, 5: 75–101. Muller, C. (1991). La Ne´gation en franc¸ais. Geneva: Droz. Munaro, N. (1997). ‘Proprieta` strutturali e distribuzionali dei sintagmi interrogativi in alcuni dialetti italiani settentrionali’. Ph.D. dissertation, University of Padua and University of Venice. —— (1998). ‘L’evoluzione diacronica del sintagma interrogativo che cosa nei dialetti veneti settentrionali’, in P. Ramat and E. Roma (eds.), Sintassi storica. Rome: Bulzoni, 267–92. —— (2001). ‘Grammaticalization as Indirect Trigger for the Loss of Movement: A Case Study in wh-Syntax’, GLOW Newsletter, 46: 77–8. —— (forthcoming). ‘La frase interrogativa’, in L. Renzi and G. Salvi (eds.), ItalAnt– Per una grammatica dell’italiano antico. Bologna: Il Mulino. —— and Obenauer, H.-G. (2002). ‘On the Semantic Widening of UnderspeciWed wh-Elements’, in M. Leonetti, O. Ferna´ndez Soriano, and V. Escandell Vidal (eds.), Current Issues in Generative Grammar. Universidad de Alcala´—Universidad Nacional de Educacio´n a Distancia—Universidad Auto´noma de Madrid, 165–94. —— Poletto, C., and Pollock, J. Y. (2001). ‘Eppur si muove! On Comparing French and Bellunese wh-Movement’, in P. Pica and J. Rooryck (eds.), Linguistic Variation Yearbook 1. Amsterdam: John Benjamins, 147–80. —— —— —— (2000). ‘Eppur si muove: On Comparing French, Portuguese and Bellunese wh-Movement’, in J. Roorick and P. Pica (eds.), Linguistic Variation Yearbook. Amsterdam: John Benjamins, 147–80. —— and Pollock, J.-Y. (2005). ‘Qu’est-ce que qu’ est-ce que? A Case Study in Comparative Romance Interrogative Syntax’, in G. Cinque and R. Kayne (eds.), Handbook of Comparative Syntax. NewYork: Oxford University Press, ch. 13. Mussafia, A. (1886). ‘Una particolarita` sinttatica della lingua italiana dei primi secoli’, in G. I. Ascoli et al. (eds.), Miscellanea di Wlologia e linguistica in memoria di Napoleone Caix e Ugo Angelo Canello. Florence: LeMonnier, 255–61. Newmeyer, F. (1998). Language Form and Language Function. Cambridge, Mass.: MIT Press. —— (1998). Language Form and Language Function. Cambridge, Mass.: MIT Press. Nowak, A. (2000). ‘On Split PPs in Polish’. MS, University of Massachusetts, Amherst. Nunes, J. (1993). ‘Direc¸a˜o de cliticizac¸a˜o, objeto nulo e pronome toˆnico na posic¸a˜o de objeto em portugueˆs brasileiro’, in Roberts and Kato 1993, 207–22.

300

References

Nunes, J. (2002). ‘VO or OV? That’s the Underlying Question: Commentary on Pintzuk’, in Lightfoot 2002c: 300–6. Obenauer, H.-G. (1976). E´tudes de syntaxe interrogative du franc¸ais. Tu¨bingen: Niemeyer. —— (1983). ‘Une quantiWcation non-canonique: la quantiWcation a` distance’, Langue franc¸aise, 58: 66–88. —— (1994). ‘Aspects de la syntaxe A-barre’. The`se de doctorat d’e´tat, Universite´ de Paris VIII. Ohta, Tatsuo (1958). Chuugokugo rekishi bunpoo [History of Chinese Grammar]. Tokyo: Hooyu shoten. [Repr. 1981.] Pagotto, E. G. (1993). ‘Clı´ticos, mudanc¸a e selec¸a˜o natural’, in Roberts and Kato 1993, 185–206. Pancheva, R. (forthcoming). ‘The Rise and Fall of Second-Position Clitics’, Natural Language and Linguistic Theory. Par, A. (1923). Sintaxi catalana segons los escrits en prosa de Bernat Metge. Halle (Saale): Verlag von Max Niemeyer. Partee, B. (2002). ‘Privative Adjectives: Subsective plus Coercion’. MS, University of Massachusetts, Amherst. Paul, W. (2002). ‘Proxy Categories in Phrase Structure Theory and the Chinese VP’, Cahiers de linguistique—Asie orientale, 31.2: 137–74. Paykin, K., and van Peteghem, M. (2002). ‘DeWniteness in a Language without Articles: A Case-Study of Russian’, Recherches linguistiques de Vincennes, 31: 97–112. PØrez Saldanya, M. (1996). ‘Aproximacio´ cognitiva al complement indirecte en catala`’, in A. Bover I Font, J. Martı´-Olivella, and M. A. Newmann (eds.), Actes del Sete` Colloqui d’Estudis Catalans a Nord-Ame`rica. Berkeley, 1993. Barcelona: Publicacions de l’Abadia de Montserrat, 191–209. Perlmutter, D. (1971). Deep and Surface Structure Constraints in Syntax. New York: Holt, Rinehart & Winston. Perrot, J. (1966). ‘Le Fonctionnement du syste`me des cas en latin’, Revue de philologie, 40: 217–27. [Repr. in K. Strunk (1973), Probleme der Lateinischen Grammatik. Darmstadt: Wissenchaftliche Buchgesellschaft, 302–17]. Peyraube, A. (1985). ‘Les Structures en ba en chinois me´die´val et moderne’, Cahiers de linguistique: Asie orientale 14.2: 193–213. —— (1996). ‘Recent Issues in Chinese Historical Syntax’, in C.-T. J. Huang and Y.-H. Li (eds.), New Horizons in Chinese linguistics. Dordrecht: Kluwer, 161–213. Philippi, J. (1997). ‘The Rise of the Article in the Germanic Languages’, in A. van Kemenade and N. Vincent 1997: 62–93. Pinkster, H. (1990). Latin Syntax and Semantics. London: Routledge. Pintzuk, S. (1993). ‘Verb Seconding in Old English: Verb Movement to InX’, Linguistic Review, 10: 5–35. —— (1999). Phrase Structures in Competition: Variation and Change in Old English Word Order. New York: Garland.

References

301

—— (2002). ‘Verb–Object Order in Old English: Variation as Grammatical Competition’, in Lightfoot 2002c: 276–99. —— and Taylor, A. (2003). ‘The Loss of OV Order in the History of English’, in A. van Kemenade and B. Los (eds.), Blackwell Handbook of the History of English. Oxford: Blackwell Publishing. —— Tsoulas, G., and Warner, A. (eds.) (2000). Diachronic Syntax: Models and Mechanisms. Oxford: Oxford University Press. Pires, A. (2001). ‘The Syntax of Gerunds and InWnitives: Subjects, Case and Control’. Ph.D. dissertation, University of Maryland. —— (2002). ‘Cue-Based Change: InXection and Subjects in the History of Portuguese InWnitives’, in Lightfoot 2002c: 143–59. Poletto, C. (2000). The Higher Functional Field in the Northern Italian Dialects. Oxford: Oxford University Press. —— (2002). ‘Sı` and e as CP Expletives in Old Italian’. Paper delivered at the 7th Diachronic Generative Syntax Conference, Girona, 27–9 June. [This volume] —— (forthcoming). ‘Internal and External Topics: on Scrambling in Old Italian’. —— and Pollock, J.-Y. (2003). On the Left Periphery of Some Romance wh-questions. Oxford: Oxford University Press. —— —— (2004). ‘On wh-Clitics, wh-Doubling and Apparent wh in situ in French and Some North Eastern Italian Dialects’ MS, University of Amiens, ISTC-CNR Padua. Pollock, J.-Y. (1989). ‘Verb Movement, UG and the Structure of IP’, Linguistic Inquiry, 20: 365–425. —— (1997). ‘Notes in Clause Structure’, in Haegeman 1997: 237–79. Polo, C. (2003). ‘Word Order in Latin, Italian and Slovene: Between Morphology and Syntax’. Ph.D. dissertation, Universita` di Padova. Portner, P., and Zanuttini, R. (2000). ‘The Force of Negation in wh-Exclamatives and Interrogatives’, in L. Horn and Y. Kato (eds.), Negation and Polarity: Syntactic and Semantic Perspectives. New York: Oxford University Press, 191–231. Pountain, C. J. (1998). ‘Learned Syntax in the Romance Languages: The ‘‘Accusative and InWnitive’’ Construction with Declarative Verbs in Castilian’, Transactions of the Philological Society, 96.2: 159–201. Pyles, T., and Algeo, J. (1993). The Origins and Development of the English Language. New York: Harcourt Brace Jovanovich (4th edn.). Radford, A. (1988). Transformational Grammar. Cambridge: Cambridge University Press. Ramos, J. (2002). ‘Dativos de posse: o percurso diacroˆnico da perda de [a NP] pleno’. Talk presented at the Annual Meeting of CELSUL, Curitiba. —— (forthcoming, a). ‘El re`gim verbal: ana`lisi contrastive catala`—castella`’, Quaderns de Wlologia: estudis lingu¨´ıstics, 9. Vale`ncia. —— (forthcoming, b). ‘El complement indirecte: L’alternanc¸a datiu/acusatiu’, Estudis roma`nics, 26. Raposo, E. (1987). ‘Case Theory and INFL-to-Comp: The InXected InWnitive in European Portuguese’, Linguistic Inquiry, 18: 85–109.

302

References

Raposo, E. (2000). ‘Clitic Positions and Verb Movement’, in Costa 2000: 266–97. Rappaport Hovav, M., and Levin, B. (1998). ‘Building Verb Meanings’, in M. Butt and W. Beuder (eds.), The Projection of Arguments: Lexical and Compositional Factors. Stanford, Calif. CSLI, 97–134. Renzi, L. (1985). Nuova introduzione alla Wlologia romanza [in collaboration with G. Salvi]. Bologna: Il Mulino. Riiho, T. (1988). La redundancia pronominal en el iberorromance medieval. Tu¨bingen: Max Niemeyer Verlag. Rivero, M. L. (1986). ‘La tipologı´a de los pronombres a´tonos en el espan˜ol medieval y en el espan˜ol actual’, Anuario de lingu¨´ıstica hispa´nica, 2: 197–220. —— (1991). ‘Clitic and NP Climbing in Old Spanish.’, in H. Campos and F. Martı´nezGil (eds.), Current Studies in Spanish Linguistics. Washington, DC: Georgetown University Press, 241–82. —— (1997). ‘On Two Locations for Complement Clitic Pronouns: Serbo-Croatian, Bulgarian and Old Spanish’, in A. van Kemenade and N. Vincent (eds.), Parameters of Morphosyntactic Change. Cambridge: Cambridge University Press, 170–206. —— and Terzi, A. (1995). ‘Imperatives, V-movement and Logical Mood’, Journal of Linguistics, 31: 301–32. Rizzi, L. (1982). Issues in Italian Syntax. Dordrecht: Foris. —— (1986). ‘Null Objects in Italian and the Theory of pro’, Linguistic Inquiry, 17: 501–57. —— (1992). ‘Early Null Subjects and Root Null Subjects’, in T. Hoekstra and B. Schwartz (eds.), Language Acquisition Studies in Generative Grammar. Amsterdam: John Benjamins, 151–76. —— (1996). ‘Residual Verb Second and the wh-Criterion’, in A. Belletti and L. Rizzi (eds.), Parameters and Functional Heads. Oxford: Oxford University Press, 63–90. —— (1997). ‘The Fine Structure of the Left Periphery’, in Haegeman 1997: 281–337. —— (1999). ‘Broadening the Empirical Basis of Universal Grammar Models: A Commentary’, in DeGraV 1999: 453–72. —— (2001). ‘On the Position Int(errogative) in the Left Periphery of the Clause’, in G. Cinque and G. Salvi (eds.), Current Studies in Italian Syntax OVered to Lorenzo Renzi. Amsterdam: Elsevier (North-Holland Linguistic Series), 287–96. —— (2002). ‘Locality and Left Periphery’, in A. Belletti (ed.), Structures and Beyond: The Cartography of Syntactic Structures, vol. iii. Oxford: Oxford University Press. —— (2004), ‘On the Form of Chains: Criterial Positions and ECP EVects’, MS, University of Siena. Roberge, Y. (1998). ‘Les Pre´positions orphelines dans diverses varie´te´s de franc¸ais d’Ame´rique du Nord’, in P. Brasseur (ed.), Franc¸ais d’Ame´rique: variation, cre´olisation, normalisation. Avignon: Universite´ d’Avignon, Centre d’E´tudes Canadiennes, 49–60. Roberts, I. (1985). ‘Agreement Parameters and the Development of English Auxiliaries’, Natural Language and Linguistic Theory, 3: 21–58. —— (1993a). Verbs and Diachronic Syntax: A Comparative History of English and French. Dordrecht: Kluwer.

References

303

—— (1993b). ‘Posfa´cio’, in Roberts and Kato 1993, 409–25. —— (1993c). ‘A Formal Account of Grammaticalization in the History of Romance Futures’, Folia linguı´stica historica, 13.1–2: 219–58. —— (1997). ‘Directionality and Word Order Change in the History of English’, in A. van Kemenade and N. Vincent 1997: 397–426. —— (1999a). ‘Verb Movement and Markedness’, in DeGraV 1999: 287–327. —— (1999b). ‘The C-System in Brythonic Celtic Languages, V2 and the EPP’. MS, University of Cambridge. —— and Kato, M. (eds.) (1993). Portugueˆs brasileiro: uma viagem diacroˆnica. Homenagem a Fernando Tarallo. Campinas: Editora da Unicamp. —— and Roussou, A. (1999). ‘A Formal Approach to Grammaticalization’, Linguistics, 37.6: 1011–41. —— —— (2002). ‘Where Do Complementizers Come from?’ Paper presented at the 7th Diachronic Generative Syntax Conference, Girona, 27–9 June. —— —— (2003). Syntactic Change: A Minimalist Approach to Grammaticalization. Cambridge: Cambridge University Press. Rocha, R. C. F. (1997). ‘A alternaˆncia indicativo/subjuntivo nas orac¸o˜es subordinadas substantivas em portugueˆs’. MA dissertation, Universidade de Brası´lia. Rohlfs, G. (1969). Grammatica storica della lingua italiana e dei suoi dialetti: sintassi e formazione delle parole. Milan: Einaudi. Ross, C. (1991). ‘Coverbs and Category Distinctions in Mandarin Chinese’, Journal of Chinese Linguistics, 19.1: 79–115. Ross, J. R. (1967). ‘Constraints on Variables in Syntax’. Ph.D. dissertation, MIT. Rossi, M. A. L. (1993). ‘Estudo diacroˆnico sobre as interrogativas do portugueˆs do Brasil’, in Roberts and Kato 1993, 307–42. Rothstein, S. (1983). ‘The Syntactic Form of Predication’. Ph.D. dissertation, MIT. Roussou, A. (1998). ‘Wh-Interrogatives from Classical Greek to Modern Greek’, in S. Lambropoulou (ed.), Proceedings of the 12th International Symposium of Theoretical and Applied Linguistics, Vol. 1 Thessaloniki: University of Thessaloniki, 109–26. —— (2000). ‘On the Left Periphery: Modal Particles and Complementizers’, Journal of Greek Linguistics, 1: 65–94. Ruhlen, M. (1994). On the Origin of Languages. Stanford, Calif.: Stanford University Press. Said Ali, M. (1921/2001). Grama´tica histo´rica da lı´ngua portuguesa. Brası´lia: Editora da Universidade de Brasilia/Sa˜o Paulo: Melhoramentos. Salles, H. M. L. (1997). ‘Prepositions and the Syntax of Complementation’. Ph.D. dissertation, University of Wales. —— (1999). ‘Orac¸o˜es inWnitivas no portugueˆs do Brasil’, Revista do Grupo de Estudos Lingu¨´ısticos do Nordeste (GELNE), 1.2: 71–5. —— (2003). ‘InWnitive Clauses as Substitutes for Subjunctive Clauses’, in M. T. Pe´rezLe´roux and Y. Roberge (eds.), Romance Linguistics—Theory and Acquisition: Selected Papers from the 32nd LSRL. Amsterdam: J. Benjamins, 197–208.

304

References

Salles, H. M. L., and Scherre, M. M. P. (2003). ‘Indirect Objects in Brazilian Portuguese and in English’, in R. Nu´nez-Cedeno and L. L. Cameron (eds.), A Romance Perspective on Language Knowledge and Use: Selected Papers from the 31st LSRL. Amsterdam: J. Benjamins, 151–65. Salvi, G. (1990). ‘La sopravvivenza della legge di Wackernagel nei dialetti occidentali della Peninsola Iberica’, Medioevo romanzo, 15: 177–210. —— (1991). ‘Difesa e illustrazione della legge di Wackernagel applicata alle lingue romanze antiche: la pozicione delle forme pronominali clitiche’, in Per Giovan Battista Pellegrini: scritti degli allievi padovani. Padua: Unipress, 439–62. —— (1993). ‘Ordine delle parole e struttura della frase nelle lingue romanze antiche’, Alfa, 37: 187–203. —— (1997). ‘From Latin Weak Pronouns to Romance Clitics’, in R. V. Mattos e Silva (ed.), Estudos lingu¨´ısticos e litera´rios 19. Salvador: Universidade Federal da Bahia, 85–104. Santorini, B. (1989). ‘The Generalization of the Verb-Second Constraint in the History of Yiddish’. Ph.D. dissertation, University of Pennsylvania. Schroten, J. (1991). ‘Interaction between Morphological Structure and Logical Form Interpretation: Bare Plural Nouns in Spanish and Dutch’, Recherches de Linguistique franc¸aise et romaine de l’Universite´ d’Utrecht, 10.30: 29–39. —— (2001). ‘L’Absence de de´terminant en espagnol’, in G. Kleiber et al. (eds.), Typologie des groupes nominaux. Rennes: Presses Universitaires de Rennes 2, 189–204. Shen, Pei (1992). Jiagu buci yuxu yanjiu [Research on the Word Order in Oracle Bone Inscriptions]. Taipei: Wenjin. Silva, M. C. F. (1996). A posic¸a˜o do sujeito no portugueˆs brasileiro: frases Wnitas e inWnitivas. Campinas: Editora da UNICAMP. Sitaridou, I. (2002). ‘The Synchrony and Diachrony of Romance InWnitives with Nominative Subjects’. Ph.D. dissertation, University of Manchester. —— (2005). ‘Language Contact-Induced Change or Internal Evolution? The Loss of Null Subjects in the History of French’. Paper submitted. Skrup, P. (1975). Les Premie`res Zones de la proposition en ancien franc¸ais. Copenhagen: Akademisk Forlag. Sleeman, P. (1996). ‘Licensing Empty Nouns in French’. Ph.D. dissertation, University of Amsterdam. Stowell, T. (1991). ‘Determiners in NP and DP’, in K. LeVel and D. Bouchard (eds.), Views on Phrase Structure. Dordrecht: Kluwer Academic, 37–56. SuÆrez Martnez, P. M. (1999). ‘El valor de los casos en los sintagmas preposicionales latinos’, Faventia, 21/2: 65–74. Sybesma, R. (1992). ‘Causatives and Accomplishments: The Case of Chinese ba’. Ph.D. dissertation, Leiden University. —— (1999). The Mandarin VP. Dordrecht: Kluwer. Szabolcsi, A. (1983). ‘The Possessor that Ran away from Home’, Linguistic Review, 1: 89–102.

References

305

Talmy, L. (1985). ‘Lexicalization Patterns: Semantic Structures in Lexical Forms’, in T. Shpen (ed.), Language Typology and Syntactic Description III: Grammatical Categories and the Lexicon. Cambridge: Cambridge University Press, 57–149. Tang, C.-C. Jane (1990). ‘Chinese Phrase Structure and the Extended X-Bar Theory’. Ph.D. dissertation, Cornell University. Taraldsen, K. T. (1978). ‘On the NIC, Vacuous Application and the ‘‘That-Trace Filter’’ ’. Indiana: Indiana University Linguistics Club. Taylor, A., Warner, A., Pintzuk, S., and Beths, F. (2003). The York–Toronto– Helsinki Parsed Corpus of Old English Prose. York: Department of Language and Linguistic Science, University of York (www-users.york.ac.uk/lang22/YcoeHome1.htm). Tenny, C. (1987). ‘Grammaticalizing Aspect and AVectedness’. Ph.D. dissertation, MIT. —— (1989). The Aspectual Interface Hypothesis. Cambridge, Mass.: Center for Cognitive Science. —— (1992). ‘The Aspectual Interface Hypothesis’, in I. Sag and A. Szabolcsi (eds.), Lexical Matters. Stanford, Calif.: CSLI Publications (CSLI Lecture Notes, 24), 1–27. Teyssier, P. (1976). Manuel de langue portugaise: Portugal-Bre´sil. Paris: Klincksieck. ThrÆinsson, H. (1996). ‘On the Non-Universality of Functional Categories’, in W. Abrahams, S. D. Epstein, H. Thra´insson, and C. J.-W. Zwart (eds.), Minimal Ideas: Syntactic Studies in the Minimalist Framework. Amsterdam: John Benjamins, 253–81. Tobler, A. (1875). Review of ‘J. Le Coultre, De l’ordre des mots dans Chre´tien de Troyes’, Go¨ttingische gelehrte Anzeigen, 34: 1057–82. [Repr. in Vermischte Beitra¨ge zur franzo¨sischen Grammatik v (1902 and 1971). Amsterdam: Rodopi]. —— (1889). ‘Pronominales Objekt zu InWnitiv oder Partizipium’, Zeitschrift fu¨r romanische Philologie, 13: 186–91. [Repr. in Vermischte Beitra¨ge zur franzo¨sischen Grammatik ii. (1902 and 1971). Amsterdam: Rodopi]. —— and Lommatzsch, E. (1966). Altfranzo¨sisches Wo¨rterbuch. Wiesbaden: Franz Steiner Verlag. Traugott, E. C. (1992). ‘Syntax’, in R. Hogg (ed.), The Cambridge Encyclopedia of the English Language, i: The Beginning to 1066. Cambridge: Cambridge University Press, 168–289. Tremblay, M., Dupuis, F., and Dufresne, M. (2002). The ‘Reanalysis of the French Prepositional System: A Case of Grammaticalization in Competing Grammars’. Paper presented at DIGS VII, Girona. [This volume]. —— —— —— (2003). ‘Les pre´positions dans l’histoire du franc¸ais: transitivite´, grammaticalisation et lexicalisation’, Verbum, 25.4: 549–62. Uriagereka, J. (1995a). ‘Aspects of the Syntax of Clitic Placement in Western Romance’, Linguistic Inquiry, 26.1: 79–123. —— (1995b). ‘An F Position in Western Romance’, in K. Kiss (ed.), Discourse ConWgurational Languages. New York: Oxford University Press, 153–75.

306

References

Valent Fiol, E. (1960). Sintaxis latina. Barcelona: Bosch. [Catalan version by J. Medina, Sintaxi llatina, reelaborada per Pere-Lluı´s Cano. Barcelona: Curial, 1979.] Valli, A. (1983). ‘Un exemple d’approche du proble`me des variantes syntaxiques en linguistique diachronique’, Recherche sur le franc¸ais parle´, 5: 125–46. Vance, B. (1989). ‘Null Subjects and Syntactic Change in Medieval French’. Ph.D. dissertation, Cornell University. —— (1997). Syntactic Change in Medieval French Verb-Second and Null Subjects. Dordrecht: Kluwer. —— (2001). ‘French and Occitan V2 Systems in Diachrony’. Paper delivered at the Colloquium on Interaction of Language Systems, Hamburg, October. van der Wurff, W. (1997). ‘Deriving Object-Verb Order in Late Middle English’, Journal of Linguistics, 33: 485–509. —— (1999). ‘Objects and Verbs in Modern Icelandic and Fifteenth-Century English: A Word Order Parallel and its Causes’, Lingua, 109: 237–65. Vikner, S. (1995). Verb Movement and Expletive Subjects in the Germanic Languages. Oxford: Oxford University Press. Vincent, N. (1997). ‘The Emergence of the D-System in Romance’, in A. van Kemenade, and N. Vincent 1997: 149–69. —— (1999). ‘The Evolution of C-structure: Prepositions and PPs from IndoEuropean to Romance’, Linguistics, 37: 1111–53. Vinet, M.-T. (2000). ‘Feature Representation and -tu pas in Quebec French’, Studia Linguistica, 54.3: 381–411. —— (2001). D’un franc¸ais a` l’autre: la syntaxe de la microvariation. Montreal: Fides. —— (forthcoming). ‘-Tu pas as a Superpositive Marker in Quebec French’, in B. Mann-Kampe and R. Bok (eds.), Going Romance 2002. Amsterdam: Benjamins. Wang Li. (1958/1988). Hanyu shi gao (zhong) [Outline of the History of Chinese, vol. ii]. Beijing: Kexue chubanshe. [Reprinted in: Wang Li (1988), Wang Li wenji [Collected Works of Wang Li], vol. ix] Wanner, D. (1987). The Development of Romance Clitic Pronouns. Berlin: Mouton de Gruyter. Whitman, J. (2000). ‘Relabelling’, in Pintzuk, Tsoulas, and Warner 2000: 220–38. Williams, A. (2000). ‘Null Subjects in Middle English Existentials’, in Pintzuk, Tsoulas, and Warner 2000: 164–87. Woolford, E. (1993). ‘Symmetric and Asymmetric Passives’, Natural Language and Linguistic Theory, 11: 679–728. —— (1997). ‘Four-way Case Systems: Ergative, Nominative, Objective and Accusative’, Natural Language and Linguistic Theory, 15: 181–227. West, J. (1985). ‘Le Patois de Paris et l’histoire du franc¸ais’, Vox Romanica, 44: 234–58. Wu, Meng (1982). ‘Ba zi yongfa er li’ [Two Illustrations for the Usage of ba], Zhongguo yuwen, 1982.6: 434. Yvon, H. (1946–7). ‘L’Assiette du nom des origines au XIV sie`cle’, Romania, 49: 289–316.

References

307

Zanuttini, R. (1994). ‘Re-examining Negative Clauses’, in G. Cinque, J. Koster, J.-Y. Pollock, L. Rizzi, and R. Zanuttini (eds.), Paths towards Universal Grammar: Studies in Honor of Richard S. Kayne. Washington, DC: Georgetown University Press, 427–51. —— (1997). Negation and Clausal Structure: A Comparative Study of Romance Languages. New York: Oxford University Press. —— and Portner, P. (2002). ‘Exclamative Clauses: At the Syntax–Semantics Interface’, Language, 79.1: 39–81. Zhu Minche (1957). ‘Lun chuqi chuzhi shi [On the Early Disposal Form]’, Yuyanxue luncong, 1: 17–33. Zou, Ke (1993). ‘The Syntax of the Chinese ba Construction’, Linguistics, 31.4: 715–36. Zribi-Hertz, A. (1983). ‘Pre´positions orphelines et pronoms nuls’, Recherches linguistiques, 12: 46–91. Zwart, J.-W. (1997). Morphosyntax and Verb Movement. Dordrecht: Kluwer.

This page intentionally left blank

Index The letter n indicates a textual note and t, a table. absolute constructions: in Romance languages 138 accusatives see double accusative: in Latin 132–6 accusative/dative alternations: in Romance languages 124–45 adverbial clauses: and postverbal clitics 65–7 agreement 242–9; Agr in C 76; AgrO 257–8; AgrP 246, 273–4; AgrS 275, 276 Anti-Babelic principle: language change 161–2 Arabic 12, 13n, 14; chronological separation in 164t, 165t, 167, 168, 169, 170, 171, 172 aspect/directionality: in Old French 113–14 aux-subject inversion: loss of 80 ba in Chinese 83–9 bare noun phrases (BNPs) 95, 97–9, 100, 101 Basque: predicates in 4 bei (passive) in Chinese 92–4 biological options 150, 152–4 Brazilian Portuguese: complementation system in 74–7; exceptional case marking in 72–81; Wnite (subjunctive) clauses in 72–4; (inXected) inWnitives in 76–7; interrogatives in 52; verb movement and clitics in 49–50, 51–9 see also European Portuguese Bulgarian: chronological separation in 164t, 165t, 168, 171, 172 Canadian French: complex prepositions in 122; negation markers in 197 n, 203 cases 106–8; in Latin 140, 141; and number 97–9; in Old English 255 see also accusative/dative alternations: in Romance languages; case changes: in Romance languages 124–145 case marking see exceptional case marking: in Brazilian Portuguese

Catalan: accusative alternations in 130, 138–9; case changes in 124–7; determiner phrases in 8; predicates in 4; proclisis in 187–9; sigma in 180–81 chance 151 Chinese: syntactic changes in 82–94 chronological separations 167–71 Classical Greek: chronological separations in 163t, 164t, 165t, 168, 169, 170, 171, 172, 173; interrogative phrases in 33n; split interrogative constructions in 236–39, 240–41, 246; word order in 237–42 clause types 15n; in Old English 255–7, 267–83 clitic placements: in Brazilian Portuguese 54–5; in complex verb constructions 56–8; in European Portuguese 49, 50, 51; in Galician 49, 51; in Latin 50; in Modern Greek 249n; in Old French 61–71; in Old Portuguese 182–3; in Romance languages 49–50, 175–6n, 183–4 see also enclisis; object clitics: in Old French; proclisis coeYcients 160–61 comparative method: historicalcomparative method 150–51 complementation system: in Portuguese 74–80 complementizer phrase expletives: in Old Italian 206–11, 218–35 complementizer phrases (CPs): structural layers 9, 11–15, n, 279n complementizers see CP-layer complex prepositions: in Old French 114–16, 121 complex verb constructions: clitic placement in 56–8 computer corpora 251, 253 cong (preposition) ‘from’ in Chinese 90, 91

310

Index

conjoined clauses: in Old English 268–9, 270, 273, 274, 279–81 conservancy of structure 82 control 74, 80 coordination: and object clitics 67–9 CP-layer 38; attitude phrase 43–4; complementizer level 9, 11; declarative force 207, 216–17; DiscourseP 11n; F position 50, 51, 53–4, 55, 57–8, 206, 207; FinitenessP 11, 61, 64, 68, 70; focus 207, 216–17, 218; focus phrase (FP) 13, 50–51, 53, 57–8, 60, 69; force phrase 11; frame; hanging topic 11, 207; frame: scene setting 207, 216, 217, 218; ground phrase 38; illocutionary force 11; informational focus 207; interrogative phrase 12n; interrogative force phrase 38–43; new information phrase (NIP) 38; theme: left dislocation 207; theme: list interpretation 207; topic markers 227–34; topic phrase 11–12, 13, 38, 61, 64, 70, 71; wh-op CP 207 cross-linguistic variation 17 dative case see accusative/dative alternations: in Romance languages deWniteness 96–108 determiner phrases (DPs) 8, 9–10 determiners 106–7, 247–8 diachronic development: in Veneto dialects 29–39 dialects see Galician; Veneto dialects Diesing’s Mapping Hypothesis 99, 100 directionality see aspect/directionality: in Old French directive reading 72, 73, 80 discourse feature 60, 61 double accusatives: in Latin 132–6 double-base hypothesis 2, 227 DP-splitting 242–6 Dutch: complementizers in 279n; determiners in 106–7; compared with Old English 271 enclisis: in Romance languages 176–8, 184, 187–92 English: chronological separations in 163t, 164t, 165t, 168, 169, 171, 172; complementizers in 11; determiners in 107, 108; predicates in 4; verb

movements in 281–2 see also Middle English; Old English epistemic modality 72–3, 76, 80 European Portuguese: chronological separations in 163t, 164t, 168, 171; clitic placements in 49, 50, 51; complementation systems in 74–80; enclisis in 176–7; Wnite (subjunctive) clauses in 73, 74; inXected inWnitives in 74–6; verb-subject order in 52 see also Brazilian Portuguese; Old Portuguese exceptional case marking: in Brazilian Portuguese 72–81 expletives see complementizer phrase expletives: in Old Italian; there expletives Extended Projection Principle (EPP) 7–8 external language 2, n, 3 extraposition: in Old English 254 features 177, 183, 184, 186–7, 188–9 see also sigma; properties of 7–9 see also functional categories Wnite (subjunctive) clauses: in Portuguese 72–4 Wnite verbs: in Old English 268–71 focus: in Brazilian Portuguese 53–4 French: bare noun phrases in 100–101; cases in 107–8; chronological separations in 163t, 164t, 165t, 168, 171, 172; complementizer phrases in 12; deWnite articles in 95–101; determiners in 247; negation markers in 194–204; interrogatives in 199–204; nominal copulas in 244–5n; null objects in 110; null-subject parameters in 4; object drops and prepositions in 120–22; parametric settings in 49–50, 51; prepositions in 109–12, 115–16, 122; sentential negation in 199–200 see also Canadian French; Middle French; Old French; Renaissance French functional categories 9–15 Galician: clitic placements in 49, 51; verb-subject order in 52 Gallo-Romance: noun phrases in 102

Index genetics see population genetics: and language taxonomy German: complementizers in 279n; as a verb-second language 212n, 216 Germanic languages: bare noun phrases in 97, 98; chronological separations in 163t, 164t, 165t, 169, 171, 172, 173; determiners in 106; compared with Old English 271 see also Dutch; German; Icelandic; Modern Germanic grammars: acquisition of 2–3, 16, 37n, 38 (Brazilian Portuguese 49–50, 51–9; French 195, 202; Modern Greek 249) grammatical competition 248–9, 251–3 Greek: interrogatives in 33n, 236–50 see also Classical Greek; Modern Greek head-initial languages: Old English 262–6 hyperbaton: in Greek 237–42, 243 Icelandic: verb movements in 275–6 illocution: in Old French 64, 67, 68 illocutionary features 60, 64, 67, 68–9 indeWnites 194–204, 241, 247 inertial theory 156, 186, 188, 190 inXected inWnitives: in Portuguese 74–7 inXection phrases (IPs): structural layers 9–11 interlocutory force 15n internal language 2, n, 3 interpolation 185, 192n interrogatives: in Brazilian Portuguese 52; in French 199–204; in Greek 33n, 236–50; in Italian 214n; in Rhaetoromance 217; in Veneto dialects 29–46 irrealis modality 73, 77n, 78 Italian: bare noun phrases in 98; chronological separations in 163t, 164t, 168, 171; complementizer phrases in 13–14, 15n; determiners in 247; interrogatives in 214n; negation markers in 196, 202; Topic and Focus Welds in 12, 13n, 14; Topic markers in 229–30, 233, 234–5 see also Old Italian; Veneto dialects Jespersen’s cycle: in French 196 jiang in Chinese 83n, 84, 85

311

language change see Anti-Babelic principle: language change language taxonomy 150–53, 154–74 Latin: cases in 140, 141; chronological separations in 163t, 164t, 165t, 167, 169–70, 172, 173, 174; clitic placements in 50; double accusatives in 132–6; inXuence on Old English translations 255, 256 left branch condition: violations of 243 left periphery 11, 32–3, 38–9, 206, 208–12, 234 lexical and syntactic comparisons 154–5 lexical argument structures 140–42 lexical doublets: in Renaissance French 116–18 lexical item: loss of 92n linguistics see parametric linguistics logical forms (LF) 17 Mandarin see Chinese merge 6, 17, 78 Middle English: syntax in 257, 258, 261 Middle French: negation markers in 196–7, 198–9; prepositions in 115 minimalist programme: and syntactic variations 82; Principles and Parameters theory 5–7, 242; and syntactic changes 15–18 Modern Germanic: clauses in 274–5 Modern Greek: split interrogative constructions in 236–50; word order in 237–42 moods: and negation: in French 201–4 morphosyntactic information see functional categories movement: loss of 17–18 negation markers: in French 194–204 nominal copulas: in French 244–5n noun phrases see bare noun phrases (BNPs) null objects: in French 110 number and case 97–9 object clitics: in Old French 69–70 object drops and prepositions: in French 120–22 object markers: in Chinese 83–9 object-verb order: loss of 260

312

Index

objects: in Old English 251–66 Occitan: proclisis in 187–9; sigma in 179–80, 182 Old Catalan see Catalan Old English: clause types in 255–7, 267–83; conjoined clauses in 268–9, 270, 273, 274, 279–81; Wnite verbs in 268–71; compared with Germanic 271; inXuence of Latin on 255, 256; null-subjects in 5; syntax in 251–66; texts, dates of 256, 256 Old French: accusative alternations in 129n; aspect/directionality in 113–14; bare noun phrases in 98, 101–5; cases in 107; clitic placements in 56–8, 69–70; deWniteness in 96–108; negation markers in 196–7, 199; null-subjects in 5; particles in 113, 114, 115; prepositions in 110–15, 116–17, 121; proclisis in 190; pronouns in 105–6; sigma in 179–80, 181; as a Topic/Focus language 61; as a verb-second language 61–71, 226–7n; verb-third structures in 212 Old Italian: complementizer phrase expletives in 206–211, 218–35; 218–225; particles in 218–225; pro-drop theory in 208–12; Topic Markers in 227–34; verb-second structures in 212–18 Old Occitan see Occitan Old Portuguese: clitic placements in 182–3; syntactic variations in 184–7 word order in 182–3 Old Romance: accusative alternations in 124–45 Old Spanish: sigma in 177, 179; syntactic variation in 185–7 Old Testament Greek see Greek parameters see Principles and Parameters theory parametric change 17–18 parametric comparison 155–162 parametric linguistics 149–50, 153–4 parametric settings: in Romance languages 49–51 particles: in Old French 113, 114, 115; in Old Italian 218–225 passive markers: in Chinese 92–4 passive sentences: in Catalan 138–9

phonetic forms (PF) 17 phylogenetic grouping: language taxonomy 162–71 population genetics: and language taxonomy 152–3 Portuguese see Brazilian Portuguese; European Portuguese possessors: in Modern Greek 248–9 predicates: in accusative structures 134; in English 4; in Modern Greek 244–6; lexical entries 3–4; in Old French 111 preWxed verbs: in Latin 135–6 prepositions: in French 109–12, 115–16, 122 Principles and Parameters theory 3–9, 15 pro-drop theory: in Old Italian 208–12; Principles and Parameters theory 4, 5 proclisis 56; in Romance languages 186–92 pronominalization: in Romance languages 139 pronouns: as determiners 10; in Old French 105–6 pruning: and structure 92 quantiWcational element 31, 97 quantiWers 237, 239 reanalysis 2, 18, 29, 33, 46, 54, 73, 74 relabelling 82, 89, 247 Renaissance French: lexical doublets in 116–18 Restricted QuantiWcation Constraint (RQC) 97, 98, 99–100 Rhaetoromance: interrogatives in 217; verb-second and verb-third orders in 212–18 Romance languages: accusative alternations in 124–45; bare noun phrases in 98; cases in 107, 124–145; chronological separations in 163t, 164t, 165t, 168, 169, 172, 173–4; clitic placement in 49–50, 175–6n, 183–4, 183–4; enclisis in 176–8, 184, 185n, 187, 192; proclisis in 56, 186–92; pronouns in 10; sigma in 178–82; verb phrase ellipsis in 176–82 see also under individual languages Russian: deWniteness in 97 semantic information see functional categories sentential negation: in French 199–200

Index

313

Serbo-Croatian: chronological separations in 163t, 164t, 167, 168, 171, 172; Tobler-MussaWa eVects in 62, 63 Shang bone inscriptions 83 sigma: in Romance languages 176–82 Slavic languages: language taxonomy 168, 170 see also Bulgarian; Russian; Serbo-Croatian Spanish: bare noun phrases in 98; chronological separations in 163t, 164t, 165t, 168; complementizer phrases in 12; dative alternations in 131n; enclisis in 177; Extended Projection Principle feature in 7–8; null-subject parameters in 4; that-trace Wlters in 4n see also Catalan; Galician split interrogative constructions: in Greek 236–50 strong F feature 176 structural change see syntactic change structural layers 9–15 subjunctive clauses see Wnite (subjunctive) clauses: in Portuguese subordinate clauses: in Old English 268, 273, 277–9 syntactic change: in Chinese 82–94; and minimalist programme 15–18 syntactic comparisons see lexical and syntactic comparisons syntactic variations: in Old Spanish 185–7 syntax: in Old English 251–66 see also lexical and syntactic comparison

Veneto dialects: interrogative items in 29–46 verb-Wnal 268–71 verb-Wrst clauses: and Tobler-MussaWa eVects 61–5, 176n verb-fourth structures: in Rhaetoromance 216 verb movements: and clitics 48–59; in English 281–2; in Icelandic 275–6; loss of: loss of V-to-C 49, 54; loss of V-to-F 48–59, 281–2 verb-object order: in Old English 251–66 verb phrase ellipsis: in Romance languages 176–82 verb phrases (VPs): structural layers 9–10 verb-second structures: in German 212n, 216; in Old French 61–71, 226–7n; in Old Italian 212–18; in Rhaetoromance 212–18 verb-subject order: in European Portuguese 52; loss of 49–51 verb-third structures: in Old French 212; in Rhaetoromance 212–18 verbs: case change of in Romance languages 124–145; Wnite: in Old English 268–71 see also complex verb constructions: clitic placement in

taxonomy see language taxonomy tenses: and negation: in French 201–4 texts, Old English: dates of composition 255, 256 that-trace Wlters: in Spanish 4, n there expletives 5–6 Tobler-MussaWa eVects: and verb-Wrst clauses 61–5, 176n Topic and Focus Welds: complementizer phrases 11–12, 14 Topic/Focus languages: Old French 61; Old Italian 227–34 Turkish: deWniteness in 96–7, 98

Wanner, D. 50 word order: in Chinese 82; in Classical Greek 237–42; in Old Portuguse 182–3 see also clause types: in Old English; verb-Wrst clauses: and Tobler-MussaWa eVects; verb-fourth structures: in Rhaetoromance; verb movements: and clitics; verb-object order: in Old English; verb-second structures; verb-subject order: in European Portuguese; verb-third structures

Uniformity of Theta Assignment Hypothesis (Baker 1988) 133 Universal Grammar 3, 16 Uriagereka, J. 49, 50, 51, 54n