The Bantu-Romance Connection: A comparative investigation of verbal agreement, DPs, and information structure (Linguistik Aktuell Linguistics Today)

The Bantu–Romance Connection Linguistik Aktuell/Linguistics Today (LA) Linguistik Aktuell/Linguistics Today (LA) prov...

Author: Cecile De Cat | Katherine Demuth

96 downloads 895 Views 4MB Size Report

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

Report copyright / DMCA form

DOWNLOAD PDF

The Bantu–Romance Connection

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

General Editors Werner Abraham

University of Vienna / Rijksuniversiteit Groningen

Elly van Gelderen

Arizona State University

Advisory Editorial Board Cedric Boeckx

Christer Platzack

Guglielmo Cinque

Ian Roberts

Günther Grewendorf

Lisa deMena Travis

Liliane Haegeman

Sten Vikner

Hubert Haider

C. Jan-Wouter Zwart

Harvard University University of Venice

J.W. Goethe-University, Frankfurt University of Lille, France University of Salzburg

University of Lund

Cambridge University McGill University

University of Aarhus University of Groningen

Volume 131 The Bantu–Romance Connection. A comparative investigation of verbal agreement, DPs, and information structure Edited by Cécile De Cat and Katherine Demuth

The Bantu–Romance Connection A comparative investigation of verbal agreement, DPs, and information structure

Edited by

Cécile De Cat University of Leeds

Katherine Demuth Brown University

John Benjamins Publishing Company Amsterdam / Philadelphia

8

TM

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

Library of Congress Cataloging-in-Publication Data The Bantu-Romance connection : A comparative investigation of verbal agreement, DPs, and information structure / edited by Cécile De Cat and Katherine Demuth. p. cm. (Linguistik Aktuell/Linguistics Today, issn 0166-0829 ; v. 131) Includes bibliographical references and index. 1. Bantu languages--Grammar, Comparative--Romance. 2. Romance language-Grammar, Comparative--Bantu. I. De Cat, Cécile. II. Demuth, Katherine. PL8025.1.B36 2008 496'.39045--dc22 isbn 978 90 272 5514 3 (Hb; alk. paper)

2008023397

© 2008 – John Benjamins B.V. No part of this book may be reproduced in any form, by print, photoprint, microfilm, or any other means, without written permission from the publisher. John Benjamins Publishing Co. · P.O. Box 36224 · 1020 me Amsterdam · The Netherlands John Benjamins North America · P.O. Box 27519 · Philadelphia pa 19118-0519 · usa

Table of contents

Acknowledgements

vii

List of contributors

ix

Introduction

xi

Part 1. Clitic and agreement Concepts of structural underspecification in Bantu and Romance Lutz Marten, Ruth Kempson & Miriam Bouzouita

3

On different types of clitic clusters Anna Cardinaletti

41

Pronominal object markers in Bantu and Romance Marie Labelle

83

The Bantu-Romance connection in verb movement and verbal inflectional morphology Carolyn Harford

111

Part 2. The structure of DPs DP in Bantu and Romance Vicki Carstens

131

On the interpretability of φ-features Roberto Zamparelli

167

Agreement and concord in nominal expressions Giuliana Giusti

201

A unified syntactic analysis of Italian and Luganda nouns Franca Ferrari-Bridgers

239

 The Bantu-Romance connection: A comparative investigaton

Part 3. Information structure The fine structure of the topic field Mara Frascarelli

261

Focus at the interface: Evidence from Romance and Bantu João Costa & Nancy C. Kula

293

Agreement in thetic VS sentences in Bantu and Romance Jenneke van der Wal

323

Index of languages

351

General index

353

Acknowledgments We wish to acknowledge the financial support of the European Science Foundation (grant #EW05-242), the British Academy (grant #BCG-41902) and the Linguistic Association of Great Britain which helped bring to life the Bantu-Romance Connection ESF Exploratory Workshop (Leeds, May 2006). Every stage of the workshop and this volume was extensively reviewed. We would especially like to express our gratitude to the team of Bantu and Romance scholars who provided extensive comments on the chapters of the book. Each chapter was reviewed by both a Bantuist and a Romanist. Thanks also to the contributors and participants for having embarked with us on this project, and for having made it such a stimulating experience.

List of contributors Miriam Bouzouita King’s College London Philosophy Department The Strand London, WC2R 2LS United Kingdom [email protected] Anna Cardinaletti Ca’ Foscari University of Venice Dipartimento di Scienze del Linguaggio Ca’ Bembo – Dorsoduro 1075 30123 Venezia Italy [email protected] Vicki Carstens University of Missouri-Columbia Department of English 107 Tate Hall Columbia, MO 65211 USA [email protected] João Costa Departamento de Linguistica Faculdade de Ciências Sociais e Humanas Av. de Berna, 26 C 1069-061 Lisbon Portugal [email protected] Cécile De Cat Department of Linguistics & Phonetics School of Modern Languages and Cultures University of Leeds Leeds LS2 9JT UK [email protected]

Katherine Demuth Department of Cognitive & Linguistic Sciences Brown University Box 1978 RI 02912 Providence USA [email protected] Franca Ferrari-Bridgers Department of Speech and Communication Studies IONA College 18 President Street, New Rochelle, NY [email protected] Mara Frascarelli Università degli Studi Roma Tre Dipartimento di Linguistica Via Ostiense, 236-00146 Roma Italy [email protected] Giuliana Giusti Università Ca’ Foscari Venezia Dipartimento di Scienze del Linguaggio Ca’ Bembo – Dorsoduro 1075 30123 Venezia Italy [email protected] Carolyn Harford Midlands State University P Bag 9055 Gweru Zimbabwe [email protected] Ruth Kempson King’s College London



List of contributors Philosophy Department The Strand London, WC2R 2LS United Kingdom [email protected] Nancy Kula Department of Language & Linguistics University of Essex Wivenhoe Park Colchester CO4 3SQ UK [email protected] Marie Labelle Université du Québec à Montréal Département de linguistique et de didactique des langues Université du Québec à Montréal C.P. 8888, succ. Centre-Ville Montréal, QC H3C 3P8 Canada [email protected]

Lutz Marten School of Oriental and African Studies Africa Department Thornhaugh Street Russell Square London, WC1H 0XG United Kingdom [email protected] Jenneke van der Wal Leiden University Centre for Linguistics van Wijkplaats 4 2311 BX Leiden The Netherlands [email protected] Roberto Zamparelli Dipartimento di Scienze della Cognizione e della Formazione Università degli Studi di Trento Via Matteo del Ben, 5 38068 Rovereto Italy [email protected]

Introduction The study of Romance linguistic structures has a long and fruitful tradition, with issues in comparative Romance being hotly debated today. However, the study of Bantu languages, especially from a more theoretical perspective, has received much less attention. There are approximately 500 Bantu languages in Africa (Nurse & Philippson 2003). These form a minor sub-branch of Niger-Khordofanian languages, though geographically they cover much of central, eastern and southern Africa. These languages are morphologically and syntactically similar in many respects (much as Romance or Germanic languages are). However, they also exhibit significant linguistic diversity, providing an especially rich (but as yet largely untapped) resource for understanding linguistic structure and language change. Recent funding by the British Academy for the project Bantu Grammar: Theory and Description, has brought together Bantu researchers from the School for Oriental and African Studies (SOAS) in London, the University of Leiden, and the Zentrum für Allgemeine Sprachwissenschaft (ZAS) in Berlin, further stimulating research and discussion regarding Bantu linguistic structure. The study of Bantu linguistic structures, and the import this has for the field of linguistics more generally, is therefore gradually becoming more widely appreciated. Of particular note is that several aspects of Bantu morphological and syntactic structure appear to be similar to structures in many Romance languages. Like most Romance languages (except French), Bantu languages permit null subjects, with subject agreement being marked on the verb. Likewise, when the object is pronominalized, both groups of languages show object clitics on the verb. As a result of these rich “agreement” systems, lexical subjects and objects can appear in various surface word orders, or not at all. Compare, for example, the Sesotho and Spanish sentences in (1) and (2) below. (1) (Thabo) o-rata basadi. (Thabo) agr-like women ‘Thabo likes the women.’

Oa-ba-rata, Thabo. agr-obj-like, Thabo ‘He likes them, Thabo.’

(2) (Juán) conoce a las mujeres. (Juan) know/agr prt det women ‘Juan knows the women.’

Las conoce, Juán. obj-knows/agr Juan ‘He knows them, Juan.’

(Sesotho)

(Spanish)

Recent research on the syntax of issues of subject and object clitics and variability in word order has concentrated primarily on comparative Bantu or comparative Romance, often in separate African or Romance conferences (e.g., WOCAL – World

 The Bantu-Romance connection: A comparative investigation

Conference on African Languages, Going Romance). There has been little discussion of these issues across these research communities. At more general conferences there are typically only a few papers presented on these topics, with no critical mass present to enhance in-depth discussion or comparative study. The only project to date that has explored similarities between Bantu and Romance languages (led by Bantuists at SOAS and funded by the AHRB) focused exclusively on morphological issues. The time is right to take Bantu research to the broader linguistic community and to extend the field of comparative investigation to syntax and information structure, providing a breeding ground for new collaborative research. The purpose of The Bantu-Romance Connection Exploratory Workshop (funded by the European Science Foundation), in which the contributions to this volume are rooted, was therefore to foster further communication between Bantu and Romance research communities in an effort to better understand the nature of linguistic structure, and its diversity and constraints. In particular, participants were invited to address several questions regarding the surface similarities found in these two language groups, and to identify the linguistic tests needed to determine if the syntactic structures underlying these surface similarities are actually the same or not. Presenters were invited to discuss the following (partially overlapping) topics: 1. Clitics, agreement, object drop • What is the decisive evidence for the agreement/pronominal status of subject and object markers/clitics? • What is the grammatical status of overt expletives, and how are these realized? • What are the discourse/syntactic restrictions on null objects? • What is the connection between object agreement and specificity? 2. DP structure • What is the structure of the DP? • At what level of structure are determiners and/or noun class (pre)prefixes specified? • How does this interact with definiteness and/or specificity? • What is the structure of concord? • How does this explain, within language groups and across groups, similarities and differences in surface morpho-syntactic structure? 3. Focus, Topic and Information Structure • What evidence bears on the structural analysis of lexical subjects? Are these actually “topics” rather than “subjects”? • What is the structure for topicalization, clefting, right and left dislocation? What are the syntactic/discourse restrictions on such structures?

Introduction 

It was hoped that the outcome of the workshop would help establish a framework for fostering future communication between scholars of Bantu and Romance languages. One of the ways it could have a lasting effect on the field was by encouraging presenters to enter into Bantu/Romance collaborative teams. We hoped that the resulting volume would provide a framework for future workshops and publications of this type, and would be of interest as a teaching text for higher-level linguistics seminars. The intriguing similarities in between these Romance and Bantu languages, in both their agreement systems and various aspects of their syntactic word-order possibilities and resulting information structures, provide a new, exciting field of research that should lay the groundwork for forging a better understanding of the structure of language. In particular it will provide a deeper understanding of interactions at the morphology-syntax interface. This will in turn contribute to our understanding of language change, and ultimately the evolution and acquisition of language itself. In the following section we briefly discuss the theoretical highlights of each of the chapters.

Part I: Clitics and agreement Marten, Kempson & Bouzouita explore word order and subject/object clitic similarities between Bantu (Otjiherero) and Romance (Latin, Spanish). This is couched within the framework of Dynamic Syntax (DS) (Cann, Kempson, Marten 2005). First, the authors provide a detailed analysis of the different word order possibilities in Bantu and Romance, showing how DS provides a uniform framework for analysis for word order variation across the two language groups. They then argue that Bantu subject and object clitics can be analyzed as being similar to Romance (object) clitics, where unfixed nodes have to be construed within a tightly locally restricted domain. Support for their analysis comes from restrictions on the presence of object markers in passive and locative inversion constructions in Otjiherero, which follow from independent constraints on the availability of unfixed nodes within a given domain. Thus, despite differences in surface morphology between the two language groups, Bantu and Romance exhibit striking parallels with respect to the way lexical information and general structure-building principles of interact within a theory of DS. Cardinaletti brings to light a restriction on clitic clusters hitherto ignored in the literature, determining which clusters can appear as proclisis and which as enclisis. The latter is shown to be available to fewer types of clitic clusters, while the former is available to all. Another distinguishing property is vowel change (in Italian): clitic clusters that exhibit vowel change can occur in both positions, but only those with no vowel change can appear as enclisis. These restrictions are

 The Bantu-Romance connection: A comparative investigation

argued to derive from partially independent properties of clitic clusters involving their internal structure and whether they are inserted as lexical units or independent words. Cardinaletti argues that the two types of clusters (i.e., those that can appear anywhere vs. those restricted to the proclitic position) result from different types of adjunction: adjunction of one clitic to another, or adjunction to different functional heads. Only the former structure can give rise to enclitic clusters. Two clitic positions are distinguished: one in the IP layer and one in the VP layer. Different features are checked in each position, and only clitic clusters resulting from adjunction on the same head are possible in the VP-internal position. Case is shown to only play a limited role in explaining the cluster-internal ordering of clitics. The data are mainly from Italian, but comparison with Bantu leads to a refinement of the proposal with respect to the (possibly universal) merging order of direct object and indirect object clitics. Differences between the two language groups are attributed to independent differences in clause structure. Labelle investigates the status of object markers in Romance and Bantu. She contends that the traditional split in the literature between analyses in terms of affix vs. clitic pronoun needs to be overcome if a unified analysis of object markers is ever to be attained. Based on evidence from French (including substantial diachronic evidence) and Chicheŵa, she shows that many aspects of the distribution of object markers cannot be captured by a strictly morphological analysis and that these elements must be visible to the syntactic component. She goes on to propose a tentative unified analysis. Critically, she proposes that there is a continuum from independent word to affix, with object clitics situated more on the clitic side in Romance than in Bantu. By recognizing that clitics have a mixed status, Labelle shows that one can account for their morphological properties without having to overlook crucial syntactic properties. Harford examines a number of morphological and syntactic properties of WH-extraction constructions and verbal inflectional morphology in two Bantu languages (Isizulu and Chishona), and two Romance languages (standard French and Italian). She analyses these properties in terms of a common right-branching Split-INFL structure, in which the lexical projection of V is dominated by functional projections of inflectional morphemes. She suggests that the relatively higher degree of fusion in verbal inflectional morphology in the Romance languages can be understood as a constraint that minimizes the number of suffixal inflectional morphemes whose postverbal directionality clashes with the basic order of functional and lexical morphemes. This reinforces the idea that Head-to-Head movement may be used to account for the directionality and relative fusion of verbal inflectional morphology. It also suggests that favoring or disfavoring movement, as expressed in relative rankings for the Optimality Theoretic STAY, cut across syntactic and morphological constructions. Thus, despite their surface morphological

Introduction 

differences, Harford suggests that Romance and Bantu languages are typologically very similar.

Part II: The structure of DPs Carstens compares several aspects of Bantu and Romance DP structure, drawing especially on data from Swahili and Spanish. First, she argues that the Bantu Noun Class is a gender system very much like that of Romance. Drawing on data from Swahili diminutive/augmentative processes, she argues that apparent derivational properties of Noun Classes and their prefixes are in fact due to derivational zero morphemes in Bantu with their own gender specifications, much like the phraselevel null element pro found in both Bantu and Romance, with concomitant recoverability/identification requirements. She suggests that Bantu has a greater number and diversity of null elements than Romance because the larger number of genders enables them to be unambiguously identified, facilitating recoverability. Carstens then proposes that Romance and Bantu DPs share a common architecture, and that both groups of languages exhibit noun raising (though with different landing sites). She then argues against Cinque’s (2005) anti-symmetric approach to postnominal modifier order, showing that a symmetrical base-generation approach to modifiers is preferable. Finally, she shows the Agree relation (Chomsky 2000, 2001) in both Bantu and Romance can easily account for the agreement within the DP (concord). Carstens concludes that all aspects of DPs in the two languages are therefore highly similar, in keeping with Universal Grammar. This paper therefore provides the basis for a much larger comparative analysis of these issues, both within and between each language group. Zamparelli’s paper explores the relation between the ± interpretable and ± valued status of features, within the Chomskian Minimalist program of generative grammar. He argues that uninterpretable features are not automatically unvalued, and shows that at least some semantic features must be visible to Agree (an operation traditionally considered to handle syntactic information exclusively). The paper opens with a discussion of the notions of default and interpretability, and the handling of lexical “exceptions”. It then explores the status of the person feature (pers) in light of what has been said about defaults, and proposes a distinction between a genuine 3rd person value for this feature and the assignment of pers = 3rd to categories without person specification. Turning to gender (gend), Zamparelli contends this feature should not be part of “core” grammar as it is not autonomously interpretable in spite of being valued. As for number agreement (num), this is shown to involve semantic as well as syntactic features. The facts are argued to require a modification of the Agree system. Zamparelli draws extensively

 The Bantu-Romance connection: A comparative investigation

on Romance evidence (especially from Italian) and identifies how Bantu data might bear on the issues discussed and advance this research program. Giusti investigates feature sharing in Nominal Expressions, drawing on data from Romance (mainly Romanian, Italian and Spanish) and Bantu (mainly Swahili and Xhosa). Concord on possessors and modifiers in Romance is considered to be parallel to prefixes in Bantu, and the determiner system in Romance as parallel to pre-prefix spreading and multiple demonstratives in Bantu. Giusti contends that feature sharing involves two distinct structural relations: Agreement and Concord. Concord involves modifiers, is triggered by Number, Gender or Class features, and never results in movement. Agreement involves subjects, is triggered by Person features, and is argued to consist of the operations probe and move. However in both Romance and Bantu, Agreement in Nominal Expressions is shown to seldom result in movement: this only happens with pronominal and adjectival possessives, which are argued to have pronominal reference. In this case the goal is exclusively formed by the probed features and pied-piping to the specifier of the probe takes place. Additional support for the distinction between Agreement and Concord is found in the co-occurrence of these two relations on the same element (the genitival article in Romanian and connectors in Bantu). Ferrari-Bridgers argues in favor of a unified syntactic analysis of Luganda and Italian simple nouns. Specifically, she shows that, in both languages, all nouns are formed via the merger of the nominalizer head [n] with a nominal stem [LP], yielding the nominal structure [nP [n [LP]]]. She further claims that syntactic movement is necessary in the formation process of Italian nouns to derive the correct morpheme order. In order to demonstrate that the structure [nP [n [LP]]] is representative for both languages, she shows that the nominalizer head [n] corresponds to both the Italian gender feature and the Luganda class feature and that, therefore, gender and class are the same feature with identical inflectional and derivational functions: At the inflectional level, gender and class trigger VP and DP agreement, whereas at the derivational level, gender and class function as n-marked heads whose merger with an XP yields a noun. She further argues that, unlike Luganda nouns in general and Italian masculine nouns, Italian feminine nouns are derived by movement. Drawing on data from diminutive formation, she also argues that XP movement is the preferred type of movement in derived nominal formation. Thus, by reinterpreting the gender/class as a nominalizer, it is possible to provide a uniform treatment for the derivation of both Bantu and Romance nouns.

Part III: Information structure Frascarelli argues that a one-to-one mapping between syntax, phonology and discourse underlies the encoding of topics and subjects. Three types of topics are

Introduction 

distinguished (aboutness-shift, familiar, and contrastive) and shown to correspond to dedicated functional projections within the C-domain, each of which is associated with distinct intonational patterns. Syntax plays a central role in this approach: it determines the interpretation (scope effects, contrast, discourse functions), the word order and the prosody. PF operations are ruled out: prosodic patterns are determined by the phonological output of syntactic structures. On this basis, apparently right-peripheral elements are argued to occupy the position of familiar topics in the left-periphery, the rest of the clause being moved to GroundP via IP-inversion. The strict one-to-one mapping of syntax onto prosody and discourse is shown to also account for subjects in Bantu and Romance. Referential/ specific preverbal subjects are argued to be topics (hosted in the C-domain), while indefinite/non-specific preverbal subjects occupy the canonical A-position. Each is associated with particular intonation and interpretive effects. Postverbal subjects are also argued to be of two types, one occupying a low A-position and resumed by an expletive subject, the other occupying the Familiar Topic position and resumed by a thematic pro. In both cases the word order is derived by remnant movement, and distinct prosodic patterns obtain. Evidence is mainly drawn from a corpus study of spoken Italian and a review of the Bantu literature (especially on Kinande). Costa and Kula explore the syntactic and prosodic realization of new information focus in Bantu and Romance. They argue that syntax alone is insufficient to encode information structure and determine its prosodic manifestations. Rather, they propose that the syntax generates all possible structures, and that these are then filtered out at the interface with the phonological component of the grammar. Contrary to Frascarelli, they contend that the prosodic effects arising from information structure should not be tied to dedicated syntactic projections. According to Costa and Kula, the locus of cross-linguistic variation lies in the way prosodic prominence is assigned. In stress-based languages (Romance), a single position of prominence may be identified syntactically; in tone-based languages (Bantu), phonological phrasing determines where focus is assigned. The first section of the paper presents arguments for a prosodic account of focus movement in Romance (drawing essentially on evidence from Portuguese) and points out shortcomings of syntactic account of these facts. The second section discusses evidence from Bantu indicating that the position of focus is predictable from prosody, most notably at the right edge of the phonological phrase. Van der Wal investigates the status of thetic, or non-topical, presentational constructions with post-verbal subjects in Bantu and Romance. She shows that type 1 languages (Sesotho and French) exhibit default (expletive) “agreement” in thetic constructions, whereas type 2 languages (Makhuwa and Italian) exhibit verbal agreement with the postverbal subject. Previous treatment of these facts attributed these differences to different parameter settings of the Agree system (e.g., Collins 2004; Carstens 2005; Baker 2008), but van der Wal shows that these



The Bantu-Romance connection: A comparative investigation

proposals cannot account for the Makhuwa facts. Rather, she shows how issues involving Case, Binding Theory and Information Structure demonstrate that these language differences are due to the status of subject-verb agreement: In the type 1 languages she suggests that subject-verb “agreement” is actually pronominal, as evidenced by the use of the “expletive” class 17 marker in Sesotho and the obligatory subject clitic in French. In contrast, subject agreement in the type 2 languages shares agreement features with the postverbal subject. The paper presents a nice treatment of comparative Bantu prosodic structure to demonstrate the different position of the logical subject with respect to the VP, laying the groundwork for further comparative research in this area. In sum, this volume has wide empirical coverage across Bantu and Romance languages, providing some support for universalist claims regarding: the status of subject and object clitics (Marten, Kempson & Bouzouita), the order of insertion for DO and IO clitics (Cardinaletti), the syntactic restrictions on the nature of Inflection (Harford), and the existence of a continuum of graded rather than discrete notions of clitic vs. affix (Labelle). Several of the papers also use data from both Bantu and Romance languages to explore current theoretical issues, especially relating to the understanding of the Agree relation. Some suggest that Agree is not sufficient to account for feature sharing in Nominal Expressions, and should be complemented with Concord (Giusti). Others suggest that Agree should be rethought in light of the lack of automatic correlation between unvalued and uninterpretable features (Zamparelli). Still others propose that its current formulation nicely accounts for Bantu/Romance similarities, though the details of the analysis vary (Carstens, Ferrari-Bridgers). Finally, several of the papers contribute to current theoretical debates in the field, such as the role played by syntax in the overall architecture of grammar. Some suggest that it is the syntax that determines discourse structure and its prosodic reflexes (Frascarelli), whereas others argue that interfaces between syntax and discourse and/or syntax and phonology (Costa & Kula) as well as the pronominal/agreement status of subject clitics (van der Wal) play a more prominent role. It is hoped that these contributions, with their new data and comparative Bantu/Romance analysis, will stimulate further debate about the nature of grammar. Cécile De Cat Leeds, UK April 2008

Katherine Demuth Providence, RI

Introduction 

References Baker, M. 2008. The Syntax of Agreement and Concord. Cambridge: CUP. Cann, R., Kempson, R. & Marten, L. 2005. The Dynamics of Language. Oxford: Elsevier. Carstens, V. 2005. Agree and EPP in Bantu. Natural Language & Linguistic Theory 23: 219–279. Chomsky, N. 2000. Minimalist inquiries: The framework. In Step by Step: Essays on Minimalist Syntax in Honor of Howard Lasnik, H. Lasnik, R. Martin, D. Michaels & J. Uriagereka (Eds), 89–156. Cambridge MA: The MIT Press. Chomsky, N. 2001. Derivation by phase. In Ken Hale: A Life in Language, M. Kenstowicz (Ed.), 1–52. Cambridge MA: The MIT Press. Cinque, G. 2005. On deriving Greenberg’s universal 20. Linguistic Inquiry 36(3): 315–332. Collins, C. 2004. The agreement parameter. In Triggers, A. Breitbarth & H. van Riemsdijk (Eds), 115–136. Berlin: Mouton de Gruyter. Nurse, D. & Philippson, G. 2003. The Bantu Languages. London: Routledge.

part 1

Clitics and agreement

Concepts of structural underspeciﬁcation in Bantu and Romance Lutz Marten

Department of Africa, School of Oriental and African Studies, University of London

Ruth Kempson & Miriam Bouzouita

Philosophy Department, King’s College, London

The paper explores parallelisms between Bantu (speciﬁcally Otjiherero) and Romance (through Latin and Spanish) with respect to left and right peripheries, and subject and object clitics. The analysis is formulated in Dynamic Syntax (DS, Cann et al. 2005) and centrally involves notions of structural underspeciﬁcation. Through providing detailed analyses of diﬀerent word order possibilities in the Bantu and Romance languages discussed, we show how DS concepts of structural growth over initially underspeciﬁed tree relations, such as the building of linked structures and unﬁxed nodes, provide a uniform basis for analyses of word order variation across the two language groups. We then extend our analysis to include Bantu subject/object markers, which we analyze by employing the same formal tools as used in the analysis of Romance (object) clitics, namely unﬁxed nodes which have to be construed within a tightly locally restricted domain. Empirical support for our analysis comes from restrictions on the presence of object markers in passive and locative inversion constructions in Otjiherero, which we show to follow from independent constraints of the availability of unﬁxed nodes within a given domain. The analyses of Bantu and Romance presented show that despite diﬀerences in surface morphology between the two language groups, both exhibit a striking parallel with respect to the way lexical information and general structure building principles of DS interact. The diﬀerence between Romance clitic systems and the agglutinative morphology of Bantu subject and object makers is thus seen to be comparatively superﬁcial, while the DS analysis brings out the strong structural parallelism between the two language groups.1

. We are grateful to Jekura Kavari, Nancy Kula, Clara Simango, Nhlanhla Thwala and audiences in Groningen, Leiden, Oxford and Leeds for helpful comments on parts of this paper, as well as to Helsinki University Library for granting access to their Swahili corpus. Parts of



Lutz Marten, Ruth Kempson & Miriam Bouzouita

1. Introduction In this chapter, we present an explanation of structural parallels between Bantu and Romance from the perspective of Dynamic Syntax, a grammar formalism which argues that natural language syntax can be explained as humans’ ability to build structured representations of meaning from words in context, on a left-toright basis. The model thus purports to reﬂect both natural language structures in relation to the linear order of words, and the way information progressively emerges during the incremental parsing of those words. Given this perspective, left and right periphery constructions are of speciﬁc interest, and we begin by illustrating similarities of periphery constructions in Romance and Bantu. In Section 2, we introduce the DS analysis of the left and right periphery in more detail, with reference to Spanish and Latin, based on previous DS analyses. Section 3 shows how the DS analysis of Romance can be extended to left and right periphery constructions in Bantu, which we illustrate from Otjiherero, Nsenga, siSwati, Swahili and Tumbuka, and demonstrates how the DS model provides a straightforward analysis of these constructions, without stipulation of construction-speciﬁc principles, which extends equally to Romance and Bantu. In the following Section 4, we probe further the parallelism between Romance and Bantu, and the formal space in which this is expressed in the DS model. We develop an analysis of Bantu subject agreement markers along the lines of Romance (object) clitics, which in our analysis reﬂect the more liberal word-order of Latin, and in particular local scrambling. Employing formal DS concepts, we construct an analysis of Bantu subject (and object) markers, which allows them freedom of construal within a very tightly locally restricted domain – what we call local underspeciﬁcation. This analysis contradicts the view of Bantu verbal structure as morphologically ﬁxed, but has the advantage of oﬀering a principled analysis of subject and object-marking restrictions in passive and locative inversion constructions, which we illustrate from Otjiherero (Section 5). Our argument thus starts from the more familiar left and right periphery analyses of Romance, and proceeds to show how they are matched in Bantu. In a second step, we focus on a less obvious parallel, namely the type of process induced by Bantu (subject) agreement markers and Romance clitics (those that are syncretic in form), and show that this parallel can be brought out through the DS formalization of local underspeciﬁcation, with the shared restrictions of locative inversion and passive as evidence.

the research reported in this paper have received ﬁnancial support from the AHRC (B/RG/ AN8675/APN16312) which is hereby gratefully acknowledged.

Concepts of structural underspeciﬁcation in Bantu and Romance

1.1 Romance-Bantu similarities on the left and right periphery It has often been proposed that Bantu and Romance share many structural characteristics despite diﬀerences in morphology. For example, objects can be fronted, with agreement and clitic doubling:2 (1) Alenje njˆuchi zi-ná-wá-lum-a. 2.hunters 10.bees sm10-past-om2-bite-fv ‘The hunters, bees bit them.’ (2) El coche, María lo compró. the car Maria cl bought.3sg ‘The car, Maria bought it.’

[Chichewa]

[Spanish]

Subjects can be postposed, with either a focus or backgrounding eﬀect, or when postposed to the right of the full verb phrase, characteristically associated with contrast: (3) Zi-ná-wá-lum-a alenje njûchi. sm10-past-om2-bite-fv 2.hunters 10.bees ‘They bit the hunters, the bees.’ (4) Canta muy bien María. sing.3sg very well, Maria ‘She sings very well, Maria.’

[Chichewa]

[Spanish]

There are also object-subject inversion eﬀects in both language families, associated with contrastive interpretation: (5) Alenje zi-ná-wá-lum-a njûchi. 2.hunters sm10-past-om2-bite-fv 10.bees ‘The hunters, they bit them, the bees.’ (6) Un coche compró María. a car bought.3sg Maria ‘A car, Maria bought.’

[Chichewa]

[Spanish]

. The term clitic doubling is here used non-technically to mean use of an NP with co-construed clitic for both right-and left-peripheral NPs. We use the following abbreviations in the glosses: a = “personal a”; acc = accusative; appl = applicative; cl = clitic; fut = future; fv = ﬁnal vowel; hab = habitual; nom = nominative; om = object marker; pass = passive; pl = plural; sg = singular; sm = subject marker; 1, 2, 3, = noun-class number. We are grateful to Jekura Kavari, Clara Simango and Nhlanhla Thwala for providing the Otjiherero, Nsenga and siSwati examples. Chichewa examples are from Bresnan and Mchombo (1987).





Lutz Marten, Ruth Kempson & Miriam Bouzouita

In all these cases there is variation. For example, fronted objects require doubling in most Bantu languages, while in Spanish, clitic doubling is obligatory only if the subject intervenes between the object and the verb. However, despite variation both within and across the language families, the overall parallel between Bantu and Romance is clear. Object agreement markers on the verb, whether an identiﬁably separate clitic or an aﬃx, result in increased word-order freedom; subjects and objects broadly diﬀer in the referentiality implications associated with co-occurrence of the full NP with the agreement marker; and the subject, if rightperipherally placed, may be placed either externally to the VP or, in some languages, immediately following the verb: (7) Zi-ná-wá-lum-a njûchi alenje. sm10-past-om2-bite-fv 10.bees 2.hunters ‘They bit them, the bees, the hunters.’ (8) Compró María un coche. bought.3sg Maria a car ‘She bought, Maria, a car.’

[Chichewa]

[Spanish]

In both language families there is variation as to the extent to which “agreement” for non-subject expressions patterns alongside subject marking or more like a regular anaphoric process, hence giving rise to referentiality eﬀects. In Spanish, for example, the dative construction behaves more like an agreement phenomenon than an anaphoric linkage between clitic and full NP that it doubles; and in Bantu, languages vary as to whether or not object marking, unlike subject marking, is subject to a referentiality restriction, or is an invariant agreement-like device. Such left and right periphery eﬀects in Romance have been the subject of considerable research, both in orthodox frameworks where the work is well known (Rizzi 1997 and others following; Monachesi 2005 in HPSG) and in the emergent Dynamic Syntax framework (DS; Cann et al. 2005; Kempson et al. 2006, 2007); and some work has been done extending these claims to the Bantu case (see Cocchi 2001 for a minimalist analysis; Cann et al. 2005; Marten 2007; Marten & Kempson 2002 for DS analyses). Certainly, the parallelism between the two language families even down to the level of individual-language variation suggests the phenomena should emerge as the consequence of interaction of general principles. The relevance of the DS perspective in this connection is the claim that syntax is no more than the progressive construction of semantic representation. In particular, the concept of building “unﬁxed” nodes as part of the ongoing construction process which is the heart of the DS claim, deﬁnes a family of relatively weak relations; and these provide the basis for an integrated characterization of left and right periphery eﬀects on a broad cross-linguistic basis (Kempson et al. 2007). There is in particular: (i) the construction of a highly restricted, structurally underspeciﬁed relation

Concepts of structural underspeciﬁcation in Bantu and Romance

where a node is introduced as unﬁxed but necessitates update within an individual predicate-argument array, the basis for local scrambling; (ii) the construction of an unﬁxed node requiring update within an individual tree, the basis for long-distance dependency and long-distance scrambling; and (iii) the building of paired (“linked”) trees whose relation to each other is not that of mother/daughter at all but merely a relation of anaphoric connectedness. These three forms of structural underspeciﬁcation are put together with the fact that pronouns themselves only provide a partial speciﬁcation of node decorations, and so can be used as a basis for unifying any weak structural relation to provide the necessary update that will ﬁx any such initially “unﬁxed” node. The result is an account of a range of intermediate eﬀects such as clitic and pronoun doubling. With just the additional and independently motivated assumption that pronoun decorations may vary as to whether they allow further structural development of the node they decorate (the DS account of expletives: Cann et al. 2005), a further range of intermediate eﬀects is obtained, in particular associated with expletives, and with doubling as displayed by Spanish datives. The overall result is that interactions between general processes of anaphora construal and the construction and update of structural relations provide a principled explanation of what are otherwise taken to be topic and focus eﬀects, while nevertheless providing an analysis which is suﬃciently ﬁne-grained to provide a basis for the full range of cross-linguistic eﬀects. Indeed Kempson et al. (2006, 2007) argue that the various topic and focus eﬀects should be seen as grounded in such interactions rather than requiring these notions as syntactic primitives. The account thus makes a bid to be explanatory in a way that an account in terms of stipulating as many discrete formal structures (such as features, categories, or functional projections) as there are distinct patternings, as in many current alternative analyses, fails to match (e.g., Rizzi 1997; Cardinaletti this volume; Rivero forthcoming). However, there are idiosyncracies associated with the interaction of object marking and local clause-internal variation in Bantu not displayed in Romance, which threaten this claim, suggesting that the commitment to a single set of universally available principles for inducing the relevant data may not be sustainable. First of all, in the majority of Bantu languages, there can only be one object marker, whether direct or indirect object marking: (9) Ngi-m-nik-e kudla. sm1sg-om1-give-past 15.food ‘I gave him/her food.’ (10) Ngi-ku-nik-e Jabulani. sm1sg-om15-give-past 1.Jabulani ‘I gave it to Jabulani.’

[siSwati]





Lutz Marten, Ruth Kempson & Miriam Bouzouita

(11) *Ngi-ku-m-nik-e. (12) *Ngi-m-ku-nik-e.

Secondly, in the majority of Bantu languages where only one object clitic is allowed, passives preclude the presence of object markers, even when the verb is di-transitive, an equally puzzling restriction (although see Woolford 1995 for exceptions): (13) òmbápírà y-á-tjàng-èr-w-á òvá-nátjè (í Kàténáà). 9.letter sm9-past-write-appl-pass-fv 2-children (by Katenaa) ‘A letter was written to the children (by Katenaa).’ (14) *òmbápírà y-é-và-tjàng-èr-w-á. 9.letter sm9-past-om2-write-appl-pass-fv Intended: ‘The letter was written to them.’

Thirdly, as is known from Bresnan and Kanerva (1989) and others, locatives undergo inversion, and in so doing, induce “subject” agreement on the verb, suggesting that the concept of subject in Bantu is exceptional in that canonical subject and the locative share a potential to act as “subject”: (15) m-Òn-djúwó mw-á-hìtí é-rúngà. 18–9-house sm18-past-enter 5-thief ‘Into the house entered the thief.’ (or, ‘There was a thief entering the house.’)

Finally there is the puzzle that unlike the canonical subject and like a passive subject, such pre-posed locatives cannot co-occur with object markers: (16) *m-on-djuwo mw-a-ri-hiti. 18–9-house sm18-past-om5-enter Intended: ‘He/she entered the house’

This suggests that there may be cross-linguistically distinct bases for relations we informally understand as subject and object, so that the concepts of subject and object themselves continue to elude us even though they are so very familiar. However, we shall argue to the contrary in this paper that the family of concepts of structural underspeciﬁcation which are deﬁned in DS as replacing concepts of movement equally apply to explain these apparent Bantu idiosyncracies, while retaining a universalist methodology on the syntactic processes themselves. In so doing, we shall provide an integrated analysis of Bantu passive and locative inversion that is not available in other frameworks. We will argue that the Bantu subject and object markers should be seen as associated with underspeciﬁed treerelations which are restricted to requiring resolution within a local domain, an underspeciﬁcation which, until resolved, debars any other such underspeciﬁed

Concepts of structural underspeciﬁcation in Bantu and Romance

tree relation. Moreover, as we shall show, this turns out to constitute a further parallelism with the Romance data. Kempson and Cann (2007) have argued that the family of patterns displayed by Romance clitics involves the construction of locally underspeciﬁed tree relations as a reﬂex of earlier Latin scrambling, an underspeciﬁcation which is most clearly displayed by syncretic forms (e.g., Spanish me, te). The non-co-occurrence of such syncretic forms within an individual cluster is explained by the tree-logic restriction that only one token of a structurally underspeciﬁed relation is possible at any one time.3 What we shall argue is that exactly the same explanation carries over to the non-occurrence of more than one Bantu object clitic, and indeed to the restriction on both passive and locative inversion that object clitics are generally precluded. The consequence is that these arguments will provide strong and novel endorsement of the parallelism between Romance and Bantu, while at the same time providing an integrated account of what have been taken to be heterogeneous and puzzling Bantu data. Given that the concept of syntactic subject deﬁned as a locally underspeciﬁed tree-relation will be new both within and beyond the DS framework (though see Wu 2005 for a related conclusion), we close by reﬂecting on what this tells us about the concept of subject in general. However, before setting out our analysis, we provide a short introduction to DS in the following section, using a combination of Latin and Contemporary Castilian Spanish as the languages of illustration for the various concepts of paired “linked” trees, unﬁxed tree relations and consequent characterizations of such structures as Hanging Topic Left Dislocation, long-distance scrambling, pronoun doubling, short-distance scrambling. We will also look at the account provided within DS of the morphological template behavior of clitic clusters.4

2. The dynamics of language processing The methodology implicit in the Dynamic Syntax formalism is to take the constructs used in semantics and deﬁne them as tree-structure representations with . Potential complications for this account arise with se, which can co-occur with both me and te and may give rise to clitic clusters (Cuervo 2002; Heap 2005). But there is reason to think that in all such cases se is an ethical dative use, for which there is considerable evidence of its requiring an independent adjunct analysis (Cuervo 2002). . It is not possible in the space provided to give a full introduction to DS, nor to provide full justiﬁcation for the Romance analyses we discuss. See Cann et al. 2005; Kempson et al. 2006; Kempson & Cann 2007; Bouzouita 2007; Bouzouita in preparation; Bouzouita & Kempson 2006; which present more detail.



 Lutz Marten, Ruth Kempson & Miriam Bouzouita

a concomitant tree-growth process of establishing some interpretation in context. The central claim is that syntactic properties of natural language reside exclusively in the progressive growth of such tree-structure representations strictly following the dynamics of left-right processing: no additional level of syntax is needed. The type of tree-growth process we assume as the process of building up an interpretation for a string such as (17) involves a growth from just a single-node tree with a requirement to build-up a propositional structure, as in the left below in (18) through a sequence of transitions to yield a ﬁnal tree, being the tree on the right in (18):5 (17) María compró un coche. Maria bought.3sg a car ‘Maria bought a car.’

[Spanish]

(18) Parsing María compró un coche in (17)

Ty(t), Comprʹ(Cocheʹ)(Maríaʹ),

?Ty(t),

→

Maríaʹ, Ty(e)

Comprʹ(Cocheʹ), Ty (e → t) Cocheʹ, Comprʹ, Ty(e) Ty(e → (e → t))

In example (18), diagrammatically displaying this process, a lot of implicit content is packed into , which symbolizes the concept of tree growth. Central to the concept of tree-growth is the concept of requirement: ?X for any decoration X. Decorations on nodes such as ?Ty(t), ?Ty(e), ?Ty(e → t) etc. express requirements to construct formulae of the appropriate type on the nodes so decorated (propositions, terms and predicates respectively), and these drive the subsequent treeconstruction process.6 The general dynamic is ﬁrst to unfold a tree structure imposing such requirements following a mixture of top-down general tree-growth strategies and bottom-

. By convention, nodes decorated by functor types are on the right, nodes decorated by argument types are on the left. In all tree displays, we give only such tree decorations as are needed to demonstrate the point in question. . In any partial tree, there is one node indicated by a pointer, ◊, as the node under development. In this framework all noun phrase construals are taken to be of type e, matching arbitrary names manipulated in natural deduction proofs. Accordingly, the terms onto which words map are lambda terms within the epsilon calculus (the epsilon calculus provides the formal study of arbitrary names, see Meyer-Viol 1995). In general, we ignore the internal structure to be assigned to such type e names.

Concepts of structural underspeciﬁcation in Bantu and Romance

up lexical actions which contribute concepts and other aspects of structure, and then compositionally to combine those concepts in a strictly bottom-up fashion to yield the overall interpretation, in which all requirements must have been satisﬁed. For any language, the process of growth of structure is strictly monotonic, and for any one interpretation, hence wellformedness, there must be at least one sequence of progressively enriched partial trees between input tree and resulting logical form in which all requirements are met. The formal system underpinning the partial trees which are constructed is a logic of ﬁnite trees (LOFT: Blackburn and Meyer-Viol 1994). There are two basic modalities, 〈↓〉 and 〈↑〉, such that 〈↓〉α holds at a node if α holds at its daughter; and the inverse, 〈↑〉α, holds at a node if α holds at its mother. Functor and argument relations are distinguished by deﬁning two types of daughter relation: 〈↓0〉 for argument daughters, 〈↓1〉 for functor daughters (with their inverses 〈↑0〉, 〈↑1〉). There is also an additional link operator, 〈L〉, which relates paired trees, with a link relation from a node in one tree to the top node of another. This tree language plays a critical role in deﬁning the individual steps of tree growth; and procedures are deﬁned for step-wise building up of such structures either by computational actions or by lexical or even pragmatic actions. All are deﬁned in the same vocabulary, a set of context-relative actions for updating representations of interpretation. Such formal tree languages by deﬁnition provide characterizations of such structural relations as dominate; and in LOFT, the concept of dominate is deﬁned in the following terms: a node can be described as dominated by a node Tn(a) when 〈↑∗〉Tn(a) holds at that node, that is when the node identiﬁed as Tn(a) is along some sequence of mother relations from the present node. Such structural relations will play an important part in what follows, but we start with the characterization of how predicates are built from lexical speciﬁcations. 2.1 Lexical information provided by verbs As in other frameworks, verbs are the major projector of structure, for which actions are deﬁned that induce some or even all of the propositional template they express. (19) Result of running lexical actions of compró

?Ty(t)Tns() U,Ty(e), ? x.Fo(x),

?Ty(e → t) ?Ty(e)

Ty(e → (e → t)), Comprʹ,[↓]⊥



 Lutz Marten, Ruth Kempson & Miriam Bouzouita

In Spanish, a subject pro-drop language, verbs project full propositional structure, in which: (i) a subject argument node is decorated with a placeholder i.e., a metavariable of the form U,V, ... of type e, which stands for some value to be assigned either from context or from the construction process; and (ii) an object node which requires subsequent development, as expressed by the requirement ?Ty(e) in (19). Taking ﬁrst the subject node decoration, these meta-variables all have an associated requirement, ?∃x.Fo(x) which guarantees they must be updated. The pointer is positioned at this node, since this allows the subject to be identiﬁed contextually as the very next step. Strings made up of either a SVO sequence or a VSO sequence are thus taken to be unmarked (we shall see in the next section how left-peripheral expressions may be treated as decorating distinct nodes, external to the propositional structure itself and identiﬁed solely through replacement of the meta-variable).7 The eﬀect is that verbs induce a sequence of actions which might equivalently be expressed by a pairing of verb plus subject pronoun. The decoration of argument nodes with a meta-variable, for example, is the intrinsic property of pronouns, underspeciﬁcation with respect to content being their hallmark. Whether provided by parsing a pronoun or a pro-drop verb, all such place-holding devices must be supplied with an assigned value (notice the requirement for a full formula ?∃x.Fo(x)); and diﬀerent types of anaphoric expression can be deﬁned according to the diﬀerent constraints on that process which they impose. Values for reﬂexive pronouns have to be updated within a given single predicate-argument structure, values for other pronouns outside such locally deﬁned structure. Meta-variables projected as part of the intrinsic speciﬁcation of the verb, on the other hand, lack any such restriction, and can have their value identiﬁed either locally or nonlocally, indexically, or from some term provided later on in the construction process, where the choice of value involves general cognitive constraints such as relevance (Sperber & Wilson 1995). The object node, on the other hand (the other node induced by the verb’s actions in (19)), has only a requirement that the node be developed so further linguistic input is essential (with no provided meta-variable, there is no license to use context-provided values). By virtue of the grounding in LOFT, trees are not however taken as syntactic primitives as in other frameworks, but are built up by explicit tree-growth procedures, deﬁned as actions which induce the structure in (19), and then feed into

. As Zagona (2002: 27) points out, the acceptability judgments for the VSO order in ﬁnite declaratives vary from speaker to speaker, some reporting it as archaic.

Concepts of structural underspeciﬁcation in Bantu and Romance 

whatever structural or pragmatic operations are suitably triggered. The so-called “bottom-node restriction”, [↓]⊥, which is part of the decoration of the predicate node in (19), is deﬁnitive of regular lexical speciﬁcations that the node to which the formula decoration is provided constitute a terminal node in the resulting tree, closing oﬀ development of the node so decorated, so that no further expansion is possible. As we shall see, this restriction may get lost in some words. 2.2 Context-dependence and lexical speciﬁcations for pronouns The process of providing such a meta-variable with a value is a substitution process, its replacement having to be selected from the set of terms made available during the construction process, characteristically from the context. For example, in (20), lo is naturally understood as picking out the same individual as picked out by the antecedent use of Pablo: (20) ¿Quién ama a Pablo? María lo ama. who love.3sg A Pablo Maria cl love.3sg ‘Who loves Pablo? Maria loves him.’

The concept of the context-dependence of anaphoric expressions in language is familiar enough. What is less orthodox is the assumption that it is to be deﬁned as a tree-update process; and with contexts also represented as (partial) trees, anaphora resolution can apply equally to the update of a pronoun from antecedent terms within the structure under construction: (21) Pablo cree que María lo ama. Pablo think.3sg that Maria cl love.3sg ‘Pablo thinks that Maria loves him.’

For the identiﬁcation of lo in (21), the context relative to which that interpretation process takes place includes the partial structure containing the subject node with its decorations. So there is no distinction between grammar-internal and discourse uses of pronouns: both are analyzed alike as a tree update process in which the meta-variable gets to be provided with a term as value by substitution. In canonical uses of pronouns, like all other content words, the pronoun has ‘the bottom restriction’ that whatever value it is assigned must be taken as decorating a terminal node in the tree. However, by adopting a representationalist stance, with pronoun construal seen as a substitution process, we can integrate anaphoric and apparently nonanaphoric uses of pronouns into a single form of explanation. In Spanish, as already brieﬂy discussed above, there is widespread clitic doubling of dative expressions which occur either pre- or post-verbally (e.g., the pre-verbal dative NP

 Lutz Marten, Ruth Kempson & Miriam Bouzouita

is doubled by les in (22)), sometimes referred to as an agreement phenomenon (Franco 2000 inter alia): (22) A familias de pocos medios les ofrecieron queso. to families of small means cl oﬀered.3pl cheese ‘To low-income families, they oﬀered cheese.’

From a DS perspective, such clitic uses can be seen as anaphoric devices of just the same sort as more regular pronouns, with just a minor loosening of the deﬁning properties of what it is to correspond to a regular lexical item of the language: they have lost this bottom restriction, with the result that the meta-variable they provide, in (23) represented as U, can be updated by structure as part of the construction process.8 As a result, these dative clitics have broader distribution, allowing identiﬁcation either from context or from the construction process, retaining their anaphoric properties but in a modiﬁed form. (23) Parsing A familias de pocos medios les ofrecieron queso in (22)

?Ty(t) ?Ty(e → t)

Funcionariosʹ Quesoʹ

?Ty(e → (e → t)), FPMʹ, U

Ofrecʹ

The eﬀect, however, is that doubling in such cases will not impose a referentiality constraint, as the meta-variable provided by the clitic can be given a value by any term whatever of appropriate type, even allowing indubitably quantiﬁed NPs to be doubled: (24) A nadie le devolvió María su manuscrito to nobody cl returned.3sg Maria his/her manuscript ‘Maria didn’t return anyone their manuscript.’

This renders them equivalent to the decorations provided by the verb for its subject in subject pro-drop languages, hence their supposed agreement-displaying properties.

. Details of individual terms are omitted, including the internal structure of the composite term projected from familias de pocos medios, diagrammatically represented as projecting the predicate FPM'. We assume that the subject position is annotated with Funcionarios', supplied from the context.

Concepts of structural underspeciﬁcation in Bantu and Romance 

2.3 The dynamics of long-distance dependency The dynamics of initially constructing some incomplete speciﬁcation and subsequently providing its update apply also to the projection of structure. The core claim of DS is that all syntactic mechanisms can be seen in terms of tree growth and update, and in particular, discontinuity phenomena are modelled by deﬁning structural concepts of underspeciﬁcation and update as a direct analogue of the formula underspeciﬁcation and update, taken to be the underpinning of anaphora resolution. In (26), for example the expression un coche is construed as providing a term for the resulting logical form, but the node which it decorates does not yet have its relation to the root, Tn(0), ﬁxed within the overall structure (indicated by the dashed line):9 (25) un coche compró a car bought.3sg ‘A car, he/she bought.’ (26) Parsing un coche in (25)

Tn(0), ?Ty(t), Cocheʹ, ·≠*ÒTn(0), ? x.Tn(x)

Once this unﬁxed node is decorated, the actions of the verb can be used to project a full template of propositional structure; and, with the subject having been contextually identiﬁed, say as someone called Maria, the pointer returns to the object node for further development. This move provides the necessary input for ﬁxing the unﬁxed node, with this delayed update also solving the subcategorization requirement of the two-place predicate projected by the verb: (27) Parsing un coche compró in (25) ?Ty(t),Tn(0)

Cocheʹ , ·≠*ÒTn(0), ? x.Tn(x)

?Ty(e Æ t)

Marίaʹ, Ty(e) ?Ty(e ),

Comprʹ

. Formally, this is deﬁned using the Kleene star operation deﬁned on the daughter-mother relation: 〈↑∗〉Tn(a) is a node dominated by a node Tn(a), where Tn(a) is along an arbitrary sequence of mother relations from the current node to Tn(a). Adding the requirement ?∃x.Tn(x)

 Lutz Marten, Ruth Kempson & Miriam Bouzouita

In addition, given that the only restriction on this uniﬁcation process is that the result be an update, we expect that it can also apply in the presence of a pronoun decorating the object node, as long as that pronoun has no bottom restriction. So we predict the availability of long-distance dependency eﬀects in Spanish with a preposed dative expression and dative doubling, as in (22): the so-called Clitic Left Dislocation (CLLD) phenomenon. Like the process of anaphora resolution, the process of constructing an unﬁxed node is in principle available at any point, allowing the construction of an unﬁxed node at some relatively late point in the parse. There is an asymmetry between early and late application of such construction processes, however, by virtue of the tree either being radically underspeciﬁed (early in a parse) or all but complete (late on in the parse process). In particular, DS assumptions lead us to expect a more restrictive result once an emergent tree is complete apart from completing the formula decorations on its nodes. General restrictions on tree development dictate that once the pointer is at a node with some imposed requirement(s), it cannot move up from that node except as particular licence is deﬁned to enable it do so. One such case is in the parsing of pronouns. In parsing pronouns, although the formula value may remain to be provided at some late stage, a type value is established, and so the construction process can proceed. However, the delay in completing the decorations on the node which this gives rise to has to be essentially local, as a strict bottomup compositionality requirement on the containing predicate-argument structure must nevertheless be satisﬁed. In consequence, though the pointer can move away from a type-decorated node as long as it has some interim place-holder decoration, it will have to return to the partially decorated node so that a ﬁxed value for that node is provided when the semantic information on the tree is compiled. This late construction of a subtree can be achieved by a process of building an unﬁxed node of a type that matches the node from which it is built, decorating it suitably, and then unifying the two (so-called Late*Adjunction). This sequence of steps will, for example, apply in deriving a construal for (28): (28) Les ofrecieron queso a familias de pocos medios. cl oﬀered.3pl cheese to families of small means ‘They oﬀered cheese to low-income families.’

With the clitic pronoun decorating a third argument node of ofrecieron early on in the parse, Late*Adjunction can induce a node of type e (Ty(e)) for the parse

as an additional decoration on that node imposes the requirement that in all successful completions of the tree, this underspeciﬁed characterization is replaced by a ﬁxed tree relation. Cf. Kaplan & Zaenen (1989) for use of the Kleene star in deﬁning the related concept of “functional uncertainty” in LFG.

Concepts of structural underspeciﬁcation in Bantu and Romance 

of a familias de pocos medios which, once established, uniﬁes with that node. The result of this introduction and decoration of an unﬁxed node at this point in the derivation yields as a consequence the very strict locality restriction at the right periphery (sometimes called “The Right Roof Constraint”), which, although well known, has never been satisfactorily explained. As a bonus, this account provides a basis for analyzing the inversion phenomenon characteristic of the Romance languages: (29) un coche, compró

María.

a car bought.3sg Maria ‘A car, Maria bought.’

Early on in the derivation, an unﬁxed node is introduced, decorated by the parse of un coche. This weak tree relation is not updated immediately but only after the verb has been processed, hence with a delay in its construal. The projection of content for the subject node is also delayed, this node only being introduced as part of the information provided by the verb, with its content provided once a predicate has been compiled from verb and object in combination. Once the pointer has returned to the subject node, Late*Adjunction can then be used to provide an intervening node for parsing the end-placed subject expression and so establishing the construal of that subject. (30) Parsing Un coche compró María in (29)

?Ty(t) U,Ty(e), ? x.Fo(x), ʹ ʹ Maria

Comprʹ(Cocheʹ), Ty(e Æ t) Cocheʹ

Comprʹ

This ﬁnally leads to a tree no diﬀerent from that which could have been derived by parsing (17) (see the the right-hand side tree in (18)). As mentioned above, wellformedness in DS is not deﬁned by the ﬁnal tree of a derivation, but rather by the monotonic transition from an initial to a ﬁnal tree. There is, however, an important constraint, which is an immediate consequence of the concept of partial trees. Just like any individual ﬁxed node, any unﬁxed node is nevertheless identiﬁed by its tree-node value – in these very weak cases solely by the dominate relation that deﬁnes it. But this means that there can eﬀectively never be more than one unﬁxed node at a time, because any two such identiﬁed nodes cannot be distinguished, and so would lead to incompatible tree decorations. The derivation of (29) notably meets this constraint: at no point in the sequence of trees over which its interpretation is built up is there ever more than one unﬁxed node.

 Lutz Marten, Ruth Kempson & Miriam Bouzouita

2.4 Constructing trees in tandem The one missing part of the DS jigsaw of interacting mechanisms is the details of how paired so-called ‘linked’ trees are constructed. These are licensed by a mechanism which induces paired trees on the assumption that they must share a term which occurs in both. The two trees are, in eﬀect, introduced in tandem, using LOFT modal requirements to ensure an anaphoric substitution process across the pair of trees. Relative clause construal provides a core example (see Kempson et al. 2001; Cann et al. 2005). But this very same device applies equally to provide a basis for Hanging Topic Left Dislocation (HTLD) structures (Anagnostopoulou et al. 1997). The left-peripheral expression is analyzed as providing the trigger for introducing a tree linked to the rootnode to be of type e, which that expression duly decorates. There is an additional restriction that the root node with requirement ?Ty(t) must now be constrained to contain a copy of the term projected from the left-peripheral noun phrase so as to satisfy the sharing-of-terms requirement dictated by the LINK relation. This modal form of requirement determines the presence of a suitably construed pronoun in the twinned structure, which the facts of HTLD corroborate: (31) Este coche, Pablo lo compró this car Pablo cl bought.3sg ‘This car, Pablo bought it.’ (32) *Este coche, Pablo compró this car Pablo bought.3sg

The only exception to this required presence of a lexical pronoun is the case of subject pro-drop structures: i.e., just those cases where the verb projects its argument nodes decorated with only a meta-variable as formula value, exactly as though a lexical form of pronoun had been present, since these can satisfy that same requirement by identifying the meta-variable appropriately. The result is that a range of strategies is available in the opening stages of a parsing process, any one of which can apply. This gives us a basis for explaining the blurring eﬀect associated with the subject position in pro-drop languages: that it is able to be construed either as a backgrounding device, or as a focussed term, or more neutrally, with concomitant diﬃculty in some cases of diﬀerentiating what have been distinguished as HTLD and CLLD structures (De Cat 2007). In particular we expect both the availability of building an unﬁxed node, decorated, and presumed to be incorporated into the single emergent structure as it unfolds, and the building of a pair of linked structures, with the second structure suitably construed as having an interpretation dictated by the decoration on the ﬁrst, as shown below. It is notable that, if these are the only two options available for parsing preverbal subject expressions in Spanish, we derive the result observed in Zubizarreta (2001) that Spanish subjects are invariably in some sense external to the clause:

Concepts of structural underspeciﬁcation in Bantu and Romance 

(33) Martín lo escribió. Martin cl wrote.3sg ‘Martin wrote it.’ (34) Parsing Martín (a) as decorating an unﬁxed node or (b) as a linked structure (a)

Tn(0), ?Ty(t),

Martίnʹ, ·≠*ÒTn(0)

(b) Martίnʹ, ·LÒTn(0)

·L-1ÒTn(n), ?Ty(t) ? ·Ø*ÒMartίnʹ,

The two strategies of building linked structures or an unﬁxed node within an individual structure are in principle also available at the right periphery, though to rather diﬀerent eﬀect. If an independent linked structure is constructed, it will be the value assigned to the meta-variable of the pronoun that is identiﬁed from context: the construal of the end-placed NP will then have to be ﬁxed to be coextensive with value of the meta-variable in order to match the linked-structure requirement of shared term in the paired structures. Hence the reported backgrounding eﬀect that can be conveyed by so-called Pronoun Doubling. Alternatively, an unﬁxed node can be introduced by Late*Adjunction whose decoration provides a value for the meta-variable. This typically gives rise to contrastive or new information eﬀects by virtue of that late construction (see Kempson & Cann 2007). This predicted ﬂexibility at both left and right peripheries is a bonus of the parsing perspective since, unlike in more conventional grammar formalisms, there is no commitment to a single assignment of structure for an unambiguous string. On the contrary, this framework provides a range of tree-growth strategies, which may feed each other, giving rise to a mixed array of eﬀects. Taking a step back from the details, what is striking in the set of explanations which this account makes possible is how few stipulations there are: no single identiﬁable structure is deﬁned by a mechanism individual to that structure. Instead we have general principles of tree development interacting with general principles for anaphora construal: it is these together that determine the range of eﬀects associated with HTLD, CLLD, Pronoun Doubling, expletives and so on. 2.5 Scrambling and locality constraints on structural under-speciﬁcation Locality constraints on actions for tree-growth are not all merely an epiphenomenon of the stage in the construction process at which update must have taken place. In free word-order languages, in particular, there is evidence of local processes of tree construction in anticipation of the verb. To express this in a principled way, we push the parallelism with anaphora resolution yet further, and extend the articulation of diﬀerent locality restrictions to structural processes of tree growth, articulating analogous restrictions on update of tree growth, deﬁning a sub-type of structural

 Lutz Marten, Ruth Kempson & Miriam Bouzouita

underspeciﬁcation which requires update within a single propositional domain (socalled Local*Adjunction). Local*Adjunction applies to a type-t-requiring node. It involves the construction of an unﬁxed node which, in eﬀect, has to be updated to a ﬁxed relation within a given local scope domain. What the rule induces is one ﬁxed argument daughter node immediately dominated by a node whose relation to the node of introduction is an underspeciﬁed relation across functor relations ∗ (〈↑1〉 picks out the functor spine along which argument nodes can be constructed – the unﬁxed tree-relation is diagrammatically indicated by a dashed line): (35) The eﬀect of Local*Adjunction

Tn(a), ... ?Ty(t)

·≠1*ÒTn(a)

·≠0Ò·≠1*ÒTn(a), ?Ty(e), ? x.Tn(x),

Proposition node

Unfixed functor node

Argument node

The node introduced by these actions has a requirement for an argument term (of type e), a description of its tree relation to the point of departure, and a requirement for a ﬁxed value. As we shall see, this rule is used to induce structure for local scrambling eﬀects. To see the general application of this, we need to revert to Latin, the source language from which Romance languages developed, as this displays the relatively free word order variation symptomatic of free use of Local*Adjunction, applying in conjunction with constructive use of case. What the case speciﬁcation ensures is immediate update of any such unﬁxed node, ﬁxing the structural relation of the node decorated by the expression well before the occurrence of the verb, thereby allowing the rule to apply again to introduce a further unﬁxed node. So as in (36), we can license the building of structure from ﬁrst servum and then Xerxes: (36) Serv-um Xerxes cecidit slave-acc Xerxes.nom killed.3sg ‘Xerxes killed the slave.’

This sequence of words can be in any order. This is where the restriction imposed by the system that there be only one unﬁxed structural relation of a type at a time has a role to play, since any duplication of the process without any such update would lead to immediate collapse of the two unﬁxed tree relations, yielding just one argument node for a predicate simultaneously characterized as a subject and an object argument. The eﬀect of case as a constructive mechanism for ﬁxing any such underspeciﬁed tree relation is therefore essential

Concepts of structural underspeciﬁcation in Bantu and Romance 

to the eﬀective re-use of Local*Adjunction as a strategy.10 Of course, once any one relation is ﬁxed, (as in (37b) below), another unﬁxed node can be introduced, following through on the same sequence of actions; and the interim result (37d) – is a partial tree with a set of argument nodes but as yet lacking any predicate node with which to combine. Notice in particular the pair of argument nodes (without having parsed a verb) in (37d), a pattern reﬂected by the later clitic clusters that emerged in the Romance languages. The verb then follows, ﬁlling out the remainder of the propositional structure to yield the appropriate output tree with Xerxes' as subject argument Serv' as object argument to the predicate Caed'. Unlike two case-distinguished unﬁxed nodes, either subject or object nodes induced by actions of the verb harmlessly collapse with those introduced as unﬁxed and updated through constructive use of case (Nordlinger 1998), because annotations provided by the verb are compatible with those provided by computational actions used in parsing the NPs: the formula decorations provided on the verb-induced argument nodes are meta-variables, compatible with all formula updates. So the tree projected by the lexical actions of cecidit can be constructed by applying those actions to the tree (37d). This allows “free” word order eﬀects without any necessary interpretational diﬀerence, with pragmatic constraints free to determine preferential orderings. (37) (a)

Locally unfixed node

(b)

Tn(0), ... ?Ty(t)

(c)

Parsing Serv-um Tn(0), ... ?Ty(t),

·≠1*ÒTn(0)

·≠1ÒTn(0) ?Ty(e Æ t),

·≠0Ò·≠1*ÒTn(0), ?Ty(e),

·≠0Ò·≠1ÒTn(0), Ty(e), Servʹ

Locally unfixed node Tn(0), ?Ty(t) ·≠1*ÒTn(0)

·≠0Ò·≠1*ÒTn(0), ?Ty(e),

·≠1ÒTn(0), ?Ty(e Æ t)

·≠0Ò·≠1ÒTn(0), Ty(e), Servʹ

(d)

Parsing Xerxes Tn(0), ?Ty(t),

·≠0ÒTn(0), Ty(e), Xerxesʹ

·≠1ÒTn(0), ?Ty(e Æ t)

·≠0Ò·≠1ÒTn(0), Ty(e), Servʹ

. Instead of, or in addition to, case speciﬁcations there may be pragmatic and prosodic clues, which are, however, likely to be less secure.

 Lutz Marten, Ruth Kempson & Miriam Bouzouita

In the modern Romance languages, with case no longer being expressible in the morphology of the full NPs, the eﬀects of the actions of Local*Adjunction are now restricted to its use with clitic pronouns.11 Firstly, in ﬁnite clauses, they invariably occur in an early position, prior to the verb, exactly as the eﬀects of multiple applications of Local*Adjunction plus case update (see Bouzouita 2007, in preparation; Bouzouita & Kempson 2006; and Kempson & Cann 2007 for a diachronic account of clitic placement). Furthermore, being stored as a lexical device, each is associated with a distinct scrambling mechanism. For example, the third person accusative clitics ﬁx their structural relation to the dominating type-t-requiring node immediately. However, many of the clitics display syncretic properties, e.g., French me, te and Castilian Spanish le. These fail to identify whether the argument introduced is a direct or an indirect object, and are assigned lexical actions which induce the construction of a locally unﬁxed node without immediate update. The syncretic form thus matches the weakeness of the update they provide. It is these actions of which the agreement forms of Bantu are redolent; so we will return to this type of update action in due course. The characteristic rigid pre-verbal positioning, allowing only other clitics to intervene between the clitic and the verb, echoes the multiple applications of Local*Adjunction plus update displayed in scrambling, a grouping which over time became calciﬁed in the lexicon as a single look-up, hence eventually a single lexical entry. In some languages, e.g., Italian, these are written as one word, glielo; in other languages, e.g., Spanish, they may become associated with idiosyncratic and noncompositional forms of interpretation.12 (This eﬀect is missing from the Bantu languages where the almost invariant case syncretism has not led to any such multiple clustering.) The composite eﬀect achieved by the Dynamic Syntax analysis is a characterization of clitic template phenomena as a lexicalization of the earlier free word order system. The update actions which had licensed ﬂexibility of NP ordering in the earlier Latin system, where they were freely available, are now lexically associated with individual clitics or clitic clusters. In this view, it is the retention of case speciﬁcations only in the lexicon with the clitic pronouns that ensured that the eﬀects of Local*Adjunction are reﬂected in clitic placement. Furthermore, because these are now all individual lexical stipulations, each can only reﬂect one of the possible mechanisms which give rise to scrambling eﬀects.

. We leave on one side whether enclitic forms are directly lexically speciﬁed or induced via Local*Adjunction. . “I gave it to them.” or “I gave them to him/her.” or “I gave them to them.”

Concepts of structural underspeciﬁcation in Bantu and Romance 

3. Left and right periphery eﬀects in Bantu The structural possibilities at the right and the left periphery available in Romance are mirrored by similar structures in Bantu, as already shown in Section 1. We can thus use the Romance analyses of the previous sections for developing analyses of similar structures in Bantu, and so in this section, we give a cross-Bantu survey to show how the range of variation encountered is similar to variation across Romance. For instance, the use of *Adjunction and LINK is exploited in Bantu as well, given the optional doubling phenomenon at least with object markers; and, as in Romance, these general mechanisms interact with the tree-update actions which the subject and object markers provide (see Marten 2007). Bantu subject and object markers are sometimes referred to as agreement markers, but are better analyzed from a semantic perspective as quasi-independent pronominal elements (e.g., Bresnan & Mchombo 1987; Marten & Kempson 2002). They behave like pronouns in terms of their referential properties much as Romance clitics, and although they appear to be more morphologized in Bantu than in Romance in terms of positional restrictions, we will argue in the following section that, like the Romance clitics, they should be analyzed by employing the concept of locally unﬁxed node, but with greater systematicity, hence demanding a more general form of explanation. At the left periphery, in some languages (such as Tumbuka (38)), object-argument nodes can be introduced through *Adjunction without co-referring object marker; although, in many languages, the more common, or sometimes only, strategy is to have an object marker (39): (38) Ngóoma ti-zamu-limilír-a namchéero. 9.maize sm1pl-fut-weed-fv tomorrow ‘Maize we will weed tomorrow.’ (Downing 2006: 62) (39) Q: Ba-ntfwana, ba-ba-nik-e-ni? 2-children sm2-om2-give-past-what ‘What did they give to the children?’

[Tumbuka]

[siSwati]

A: Tin-cwadzi, ba-ti-nik-e ba-ntfwana. 10-books sm2-om10-give-past 2-children ‘Books, they gave (them) to the children.’

In the siSwati example, tincwadzi ‘books’ is projected onto a linked structure, hence the object marking. In addition, as we will see shortly, siSwati object markers have not lost their bottom restriction, indicative of the node having to be a terminal node in the resulting tree. At the right periphery, subject and object expressions are found, and the phenomenon of clitics losing their bottom restriction may seem to have extended further than in Romance since, in some languages, the

 Lutz Marten, Ruth Kempson & Miriam Bouzouita

doubling phenomenon is generalised to all object clitics, and even in some cases internally to the structure under construction (rather than between pairs of trees). First, there is the distribution which is widespread in Romance: in siSwati, object expressions tend to occur with object clitics only when they are right dislocated, as the adverb placement in (40) indicates. But in Swahili, object clitics can occur with co-referential objects internally to the verb-phrase sequence, indicating that in Swahili object clitics have lost the bottom restriction: (40) Ng-a-yi-bon-a kahle inja. sm1sg-past-om10-see-fv well 10.dog ‘I saw the dog well.’ (41) *Ng-a-yi-bon-a (42)

inja

[siSwati]

kahle.

Gidyoni a-li-kuwa h-a-ja-mw-on-a huyo ki-jana Gidyoni sm1-past-be neg-sm1-perf-om1-see-fv dem 17-youth vizuri. well ‘Gidyoni had not seen the youth well.’ (Mvungi n.d.: 126)

[Swahili]

However, in apparent marked contrast to the Romance pattern, many Bantu languages have a restriction on the number of object markers available in the verbal template. Thus languages like siSwati, Swahili and Otjiherero allow only one object marker per verb, although multiple object markers are found, for example, in Tswana, Rundi and Kinyarwanda. While the restriction on the number of object markers is often taken as morphological, we will argue below that it is in fact syntactic, in that object markers project locally unﬁxed nodes, from which the restriction to only one object marker at a time follows immediately.13 The analysis involving locally unﬁxed nodes predicts furthermore that there should be no signiﬁcant restrictions on the case or thematic role of the object clitics, in contrast to Romance, where clitics are a reﬂection of an older case system, and this prediction is borne out: (43) ú-térék-èr-à òvá-éndà ònyàmà p-òngàndà. sm1-cook-appl-fv 2-guests 9.meat 16-9.house ‘S/he cooks meat for the guests at home.’

[Otjiherero]

(44) ú-vé-térék-èr-à ònyámà p-òngàndà. sm1-om2-cook-appl-fv 9.meat 16-9.house ‘S/he cooks them meat at home.’

. Multiple object markers, under this view, are only possible if they are taken to induce a complex of nodes built from a single intermediate propositional node, itself unﬁxed (a characteristic of scrambling languages: Kiaer 2007), or if the node each one induces is ﬁxed immediately, involving a pragmatic notion of constructive case (McCormack 2008). However, we will

Concepts of structural underspeciﬁcation in Bantu and Romance 

(45) ú-í-térék-èr-à òvá-éndà p-òngàndà. sm1-om9-cook-appl-fv 2-guests 16-9.house ‘S/he cooks it for the guests at the house.’ (46) ú-pé-térék-èr-à òvá-éndà ònyàmà. sm1-om16-cook-appl-fv 2-guests 9.meat ‘S/he cooks meat for the guests there.’

Although Otjiherero allows only one object marker per verb, object markers can mark the dative (44), direct (45) or locative (46) complement of the verb. Object markers are formally distinguished for diﬀerent noun classes, but not for case. This is directly explicable on the assumption that the object markers themselves induce the unﬁxed node and provide it with a meta-variable decoration, without any structural update of that node relation taking place. In eﬀect, the only structural information they provide is that they are an argument of the predicate corresponding to the verb. Subject expressions, too, as in the Romance languages, are found in post-verbal, inverted position and often carry presentational focus by virtue of this late placement, or alternatively, may be associated with an afterthought interpretation (48): (47) à-léndò à-fwík-à. 2-visitors sm2.past-arrive-fv ‘(The) guests have arrived.’

[Nsenga]

(48) à-fwík-à à-léndó. sm2.past-arrive-fv 2-visitors ‘Guests have arrived.’/‘They have arrived, the guests.’

Nsenga subject markers have lost their bottom restriction, and so update both through *Adjunction and by constructing a linked node is possible, accounting for the two diﬀerent readings. Again, we ﬁnd variation across Bantu, as, for example, in Otjiherero, subject markers cannot be associated with update by Late*Adjunction (49), and post-verbal subjects with agreeing subject marker can only be introduced through a LINK structure yielding (50), with its associated co-referring afterthought interpretation. However, since the diﬀerent update possibilities of Otjiherero clitics are speciﬁed lexically, we would expect variation within the language as well as cross-linguistic variation, and indeed, update within the propositional structure is possible with (grammaticalized) locative subject markers (51) indicating that they have lost their bottom restriction: (49) *v-á-hìtí òvá-ndú. sm2-past-enter 2-people

[Otjiherero]

focus on one object marker languages, more speciﬁcally on Otjiherero in the following sections, and will leave multiple object markers to one side for the present.

 Lutz Marten, Ruth Kempson & Miriam Bouzouita

(50) v-á-hìtí, òvà-ndú. sm2-past-enter 2-people ‘They entered, the people.’ (51) p-á-hìtí òvá-ndú. sm16-past-enter 2-people ‘There entered people.’

Bantu languages thus make use of the same strategies as the Romance languages, with minor variations. Overt NPs, both subjects and objects, can be introduced into the parse sequence early or late, with the nodes that they serve to decorate either being introduced within a single tree by *Adjunction or being taken to be a linked structure. As in Romance, these diﬀerent modes of introduction interact with the lexical speciﬁcations of the corresponding subject and object markers, introducing locally unﬁxed nodes so that these can be uniﬁed with an already constructed unﬁxed node, or, by virtue of some bottom restriction, allowing no such update and forcing any co-referring NP to be processed as decorating an independent linked structure. As expected, these diﬀerent strategies can be used together. As a result, an object expression can be processed by early application of *Adjunction to create an unﬁxed node, which can then come to be ﬁxed as the object node once that is introduced by actions of the verb. The subject marker, on the other hand, which provides a type value but only a meta-variable as formula value, will need update from application of Late*Adjunction after the object node is completed; and the decorations of that subject node will then be provided by using a post-verbal subject-marked expression to decorate the introduced unﬁxed node. An illustration of this is (52), with class 10 subject marking and post-posed agreeing subject, under the assumption that subject markers in Tumbuka have lost their bottom restriction and that the locative term is construed as argument of the predicate: (52) pa-mu-páanda zi-ka-dúk-a mbúuzi. [Tumbuka] 16-3-wall sm10-past-jump-fv 10.goats ‘Over the wall jumped goats.’ (‘The goats jumped over the wall.’) (Downing 2006: 62)

The data presented in this section have served to show the parallelism of Romance and Bantu in terms of word-order freedom (through the application of *Adjunction and the building of linked structures), with restrictions on this freedom imposed by the lexical constraints encoded by subject and object markers. However, we have not yet addressed the question of the representation of these subject and object markers, having concentrated merely on their interaction with early and late placed NPs. In the following section, we address this question and argue, taking the conceptual underspeciﬁcation of Bantu clitics as a starting point, that they are also structurally underspeciﬁed, inducing a locally unﬁxed

Concepts of structural underspeciﬁcation in Bantu and Romance 

node which they decorate with a place-holder, with both structural relation and formula value needing to be updated. 4. Inducing locally unﬁxed nodes: Otjiherero subject markers We argued above that Romance clitics may decorate locally unﬁxed nodes, the clitic system overall being a reﬂex of the historic case system of Latin and its constructive use in scrambling. We will employ the same notion of locally unﬁxed nodes for an analysis of subject and object markers in Otjiherero, arguing synchronically with evidence from passive and locative inversion constructions that the Bantu subjectmarking system parallels the scrambling-induced actions of Romance.14 From this it also follows that in contrast to subject pro-drop in Spanish, where we have analyzed the verbal actions as providing both a subject node and a meta-variable as decoration, we analyze subject “pro-drop” in Bantu as resulting from the lexical actions of the subject markers, similar to the actions provided by pronominal object clitics in Romance. Since subject markers are obligatory cross-Bantu, this means that overt NP subjects are always taken either to decorate a linked structure, with a copy of the term they provide having to be constructed in the primary structure, or an unﬁxed node, in which case that node will have to unify with the node decorated by the subject marker. The analysis is motivated initially through the parallelism with Romance, and as formal reﬂex of the observation that Bantu clitics are pronoun-like. However, crucial to the analysis to be given is that the subject marker induces the building of a locally unﬁxed node which it decorates with a meta-variable as formula value without updating that structural relation. The following steps of the derivation of (53) illustrate the analysis. (53) Kàtènáà w-á-kòtòk-á. 1a.Katenaa sm1-past-return-fv ‘Katenaa returned.’

[Otjiherero]

(54) Parsing Kàtènáà w- in (53)

Tn(0), ?Ty(t), Kàtènáàʹ , Ty(e), ·≠*ÒTn(0)

Uw,Ty(e), ·≠0Ò·≠1*ÒTn(0)

. The notion of locally unﬁxed nodes is independent of the notion of case, and so we are not proposing that Bantu languages have grammatical case, since it is widely known that they do not. On the other hand, there are in fact isolated instances of morphological diﬀerences between subject and object markers, e.g., Proto-Bantu class 1 (3rd sing) subject *´a vs. object *m`u, 1st pl. subject *t`u vs. object *t´u (Meeussen 1967), so the point might be worth revisiting.

 Lutz Marten, Ruth Kempson & Miriam Bouzouita

We assume that the subject NP decorates an unﬁxed node introduced by *Adjunction. We could equally have chosen to introduce an independent linked structure into the emergent overall structure, since both strategies are available. However, as mentioned above, under the locally unﬁxed node analysis for subject markers, full NP subjects will never be taken to provide the decorations for a ﬁxed node because no ﬁxed node is available if the subject precedes the verb. Following the construction of the unﬁxed node decorated by the preposed expression, we assume, as with the Romance object clitics, that it is the subject marker that induces the building of a locally unﬁxed node, to which it adds a meta-variable decoration. We assume, furthermore, that the subject marker encodes as formula value a pronominal meta-variable restricted by the associated class information (in this case expressed as a subscripted W), and we leave to one side the problem of how this information is spelled out (see Cann et al. 2005: chapter 7). Note that the two unﬁxed nodes can unify at this stage; and indeed in this derivation they do so, with the actions induced by the subject marker having initiated the local domain within which the left-peripheral expression will be interpreted. The next step is to scan the lexical information from the tense morpheme. We assume that the tense marker not only provides semantic information about the time of the event (an annotation on the root node), but also provides a skeletal predicate frame with subject- and predicate-requiring nodes, reﬂecting the probable historical origin of many Bantu tense markers as verbs: (55) Parsing Kàtènáà w-á in (53)

Tn(0), ?Ty(t),Tns()

Kàtènáàʹ , Ty(e), ·≠0Ò ·≠*1ÒTn(0)

·≠0ÒTn(0) ?Ty(e)

·≠1ÒTn(0), ?Ty(e Æ t),

With the new presence of a ﬁxed subject node, the unﬁxed node-relation could now be ﬁxed by unifying the two nodes, and in this simple case they do so (with the passive and locative cases that we shall see, this option is not taken up).15 Hence the decoration of the subject is completed at this stage, and the information from the verb then annotates the predicate node as the next step. The ﬁnal

. By deﬁnition of the Kleene star operator, the set of relations denoted can be empty, in this case, the possible sequence of functor relations being null, allowing enrichment as the subject relation.

Concepts of structural underspeciﬁcation in Bantu and Romance 

step in the derivation is the parsing of the ﬁnal vowel, which induces the eventual construction of the predicate, and then all semantic information in the tree is duly compiled up progressively so that a ﬁnal propositional formula of type t is derived as decoration to the top node, satisfying the overall requirement. While the eventual tree for (53) looks like an ordinary subject-predicate structure, it is important to keep in mind that its derivation involved construction steps at which ﬁrst one unﬁxed node and then one locally unﬁxed node were part of the tree. One of the reasons for analyzing subject markers in Otjiherero as triggering the building of unﬁxed nodes was to extend a parallelism between Romance and Bantu clitics so that it applies also to the Bantu subject clitic (which not all Romance languages display).16 We have so far discussed similarities between the two language groups with respect to the diﬀerent word-order possibilities of full NPs, which we have analyzed as resulting from the interplay of linked structures, *Adjunction and diﬀerent lexical speciﬁcations of co-referential pronominal elements. Now we propose that in terms of structural and referential properties of their pronominal elements also, the two language groups are similar in that, in both groups, pronominal elements are typically positioned close to the verb stem because they trigger the building of locally unﬁxed nodes. However, whereas in Romance, the projection of locally unﬁxed nodes for a clitic coincides with remnants of a case system, so that locally unﬁxed nodes can be ﬁxed if enough case information is available, thereby licensing the occurrence of multiple clitics, in Otjiherero, locally unﬁxed nodes are only ﬁxed if a ﬁxed tree node address is provided independently, either by lexical information from the tense marking, or from the verb. This distinction provides the basis for the constraint in Otjiherero, as in many Bantu languages, that only one object marker can be present in the inﬂected verb form. Moreover, it explains why, in many noun-classes in the Bantu languages, there is no diﬀerence in morphological marking between subject and object marker: all the marker does is to initiate and decorate a locally unﬁxed node, with some other expressions providing its decoration, and construal as subject or not is only ensured either by the following tense markers (in the case of the subject), or by the following verb in the case of the object marker. Conﬁrmation of this analysis now comes from passive and locative inversion in Otjiherero, which pose the additional puzzle of precluding object preﬁx-marking, which this analysis now promises to solve.

. In French, there is clear evidence that the pronoun does not decorate an unﬁxed node, as the referentiality restriction on subject clitic doubling remains completely undisturbed. In the Northern Italian dialects however, an account of subject clitic doubling might arguably follow lines similar to those developed here (cf. Poletto 2000).

 Lutz Marten, Ruth Kempson & Miriam Bouzouita

5. Passive and locative inversion The passive construction was one of the constructions ﬁrst argued in generative grammar to involve movement, based on the older descriptive observation that the logical object of the verb becomes the grammatical subject in the corresponding passive. Accordingly, most analyses of the passive involve the matching of the function of the relevant NP at two diﬀerent levels: the logical or semantic level, and the grammatical level. The challenge in Dynamic Syntax for an analysis of the passive is that no level of grammatical function is deﬁned: the trees built in Dynamic Syntax are logical or semantic trees, and thus the eventual tree structure associated with both the active and the passive will be a transitive structure, although in the passive the formula value for the logical subject might be an existentially quantiﬁed term (“someone”). However the semantic representation of the logical object will be the same for both active and passive, as it will be associated with an argument node below the predicate node. The question then is how to derive such a transitive structure from the passive form of the verb. In the light of the preceding discussion, a natural hypothesis to explore is associating the passive suﬃx with a delay in unifying the unﬁxed node provided by the subject marker with the logical-subject node provided by the tense marker.17 Delay in unifying the unﬁxed node is unproblematic: all rules are constraints so not taking up the option of uniﬁcation is always a possibility. So the unﬁxed node, now identiﬁed as locally unﬁxed, can remain unﬁxed until a point at some later stage in the parse when another putative uniﬁcation site arises, such as, for example, the logical object node supplied by the lexical information from the verb. Thus, we assume that in a parse of (56) the structural option to unify the locally unﬁxed node after the introduction of the tense marker which supplied the ﬁxed subject node, is NOT taken up in passives, but rather the pointer is moved directly to the predicate node as in (57):

(56) òmbàpírà y-á-tjàng-w-á (í òvá-nátjè). 9.letter sm9-past-write-pass-fv by 2-children ‘The letter was written (by the children).’ (57) Parsing òmbàpírà y-á in (56)

Tn(0), ?Ty(t),Tns()

Òmbàpίràʹ,Ty(e), ·≠0Ò ·≠*1ÒTn(0)

·≠0ÒTn(0) ?Ty(e)

·≠1ÒTn(0), ?Ty(e Æ t),

. An alternative analysis in the system would be to invoke steps of inference between diﬀerent, albeit related, concepts and their associated propositional structures. For further discussion of analyses of passives in DS, see Cann & Wu (2006).

Concepts of structural underspeciﬁcation in Bantu and Romance 

The parsing of the transitive verb stem results in the building of a node for the twoplace predicate, and also for the object argument, this node still requiring a Ty(e) expression, all exactly as in the regular transitive cases – indeed at this point in the parse process there is still no indication of any passive form of construal: (58) Parsing òmbàpírà y-á-tjàng in (56)

Tn(0), ?Ty(t),Tns()

Òmbàpίràʹ,Ty(e), ·≠0Ò ·≠*1ÒTn(0)

?Ty(e), ·≠0ÒTn(0)

?Ty(e Æ t), ·≠1ÒTn(0) ?Ty(e ),

Ty(e Æ (e Æ t)), Tjàngʹ

However, since the object node has a requirement for an expression of type e, and the locally unﬁxed node is of Ty(e), the locally unﬁxed node can unify with the object node, and in so doing ﬁx the construal of that front-placed expression as the logical object. As a result, the logical object decorated by òmbàpírà ends up in the correct tree position, all without having to invoke lexical inference or grammatical function changing rules as proposed in LFG (e.g., Mchombo 2005). The interpretation of the logical subject is still outstanding however, and we propose that it is provided as part of the lexical tree-update actions deﬁning the passive marker. That is, the passive suﬃx is deﬁned as providing an annotation for the subject node, and this forces the compilation of semantic information associated with the predicate in order that the pointer can indeed return to the subject node to provide it with the decorations encoded by the passive suﬃx. Finally, there is the ﬁnal vowel -a whose update actions are the indication that the derivation is completed. It should be noted that in this analysis Bantu “subject markers” are in fact not logical-subject markers at all, as they can be used to encode logical objects, as for example in the passive. They are more correctly thought of as mere markers that introduce some local domain whose infra-structure will be built up in what follows. Since the lexically supplied formula speciﬁcation of the subject is quite weak, it can be updated with information from the context, or by using the agent expression to induce a linked structure (upon a construal of this as background information). Note that this analysis involves the assumption that Bantu so-called subject markers are in fact clitics with referential and positional freedom comparable to Romance clitics; and, like them, they decorate locally unﬁxed nodes, with one additional lexical stipulation that the passive morpheme provides an annotation for the logical subject node as a relatively late step in the derivation. In support of this, it is noteworthy that across almost all Bantu languages, the

 Lutz Marten, Ruth Kempson & Miriam Bouzouita

passive morpheme comes last in the series of derivational suﬃxes, a fact which has been much commented on, but which has not been satisfactorily explained (e.g., Hyman 2003 proposed a ‘morphological template’ to account for the suﬃx order). Under this analysis, the ﬁnal position of the passive reﬂects its particular function: it does not induce a local action modifying the semantics of the verb, but provides annotations for the logical subject node which the pointer can return to only by compiling up the predicate node, hence providing a last set of steps before the tree as a whole can be completed. In eﬀect, the passive marker thus acts as a closure of the predicate term. Further support for this analysis comes from the unavailability of object marking in passives in many Bantu languages (Woolford 1995), including Otjiherero. While in passives of transitives, no requisite object exists, and so the absence of object marking is not surprising, the absence of object marking in ditransitives is more puzzling: (59) òvà-nátjè v-á-tjàng-ér-w-á òmbápírà (í Kàténáà). 2-children sm2-past-write-appl-pass-fv 9.letter (by Katenaa) ‘The children were written a letter (by Katenaa).’ (60) òmbàpírà y-á-tjàng-ér-w-á óvá-nátjè (í Kàténáà). 9.letter sm9-past-write-appl-pass-fv 2-children (by Katenaa) ‘A letter was written to the children (by Katenaa).’ (61) *òvá-nátjè v-é-ì-tjàng-ér-w-á. 2-children sm2-past-om9-write-appl-pass-fv (62) *òmbàpírà 9.letter

y-é-và-tjàng-ér-w-á. sm9-past-om2-write-appl-pass-fv

As these data show, Otjiherero allows both the benefactive and the theme object to become subjects and corresponding passives, with the remaining object following the verb. However, neither of the remaining objects may be expressed by an object marker. This fact has sometimes been associated with properties of the relevant objects or the verb, e.g., the unavailability of case marking for object markers in passives. However, as DS does not have a case theory as argument licensing, such an analysis is not readily available in the framework. Instead, the ungrammaticality of object markers in passive structures follows directly from the analysis in terms of building a locally unﬁxed node to be decorated by what is morphologically the subject. The object marker in active structures is licensed because the locally unﬁxed node introduced at the outset of the parse is ﬁxed once the tense marker provides a ﬁxed subject node, and so a new locally unﬁxed node can be built. However, in passives, the initial locally unﬁxed node is not ﬁxed until the verb has been parsed, as it is the actions of the verb which provide the ﬁxed

Concepts of structural underspeciﬁcation in Bantu and Romance 

object node. But since only one locally unﬁxed node at a time is normally possible, it follows that no further locally unﬁxed node, and hence no object marker, can be built before the verb is parsed. Once the locally unﬁxed node has been ﬁxed, it is of course possible to introduce the second objects of the ditransitives after the verb, as the examples above show. This analysis of passive constructions in Otjiherero thus conﬁrms our more general enterprise of uncovering the parallelism between Romance and Bantu, and speciﬁcally to describe the variation encountered within the same structural architecture. Furthermore, our analysis brings out parallels between passive and locative inversion, which are often taken as two distinct problems: passive is seen as a valency changing operation, with distinct morphology and involving some interaction between grammatical function and thematic roles, while locative inversion, on the other hand, has no associated morphology and does not change the valency of the verb. Yet, what is similar about the two constructions is that in both cases, “subject” agreement is with an NP which is not the logical subject of the verb. Incidentally, both constructions also share de-topicalization of the logical subject, but we will leave this aside here (see Marten 2006 for more discussion of Otjiherero and Bantu locative inversion): (63) è-rúngá r-á-hìtí m-ón-djúwó. 5-thief sm5-past-enter 18–9-house ‘The thief entered (into) the house.’

[Otjiherero]

(64) m-òn-djúwó mw-á-hìtí é-rúngà. 18-9-house sm18-past-enter 5-thief ‘Into the house entered the thief.’ (or, ‘There was a thief entering the house.’)

In the locative inversion construction in (64), the locative phrase mòndjúwó precedes the verb and the clitic agrees with it. The logical subject follows the verb, and there is no special morphology associated with the verb. Our analysis of the Otjiherero passive can easily be extended to locative inversion if we assume that the locative is projected as an argument term internal to the predicate. As we have seen above, there is some evidence for this view from object marking, where locative expressions behave like other complements. Similarly, locatives in interaction with applicatives show complement-like behaviour. If we take this as evidence of the argument-hood of the locative term to be constructed, then we can analyze locative inversion exactly like passives with but one minor diﬀerence. This is that the interpretation of the logical subject is not provided by any morphological marker, there being no analogue to a stem-ﬁnal passive marker in locatives: the logical subject thus has to be introduced by parsing of lexical input, which cannot be omitted. This follows because the tense marking, though it provides a subjectpredicate skeleton, does not provide any type or formula speciﬁcation to satisfy

 Lutz Marten, Ruth Kempson & Miriam Bouzouita

the type requirement on the introduced subject node. In the general case this is provided by the subject-marker but in these cases this is uniﬁed with the node decorated by the locative, so there is nothing as yet satisfying the requirement on the subject node. Accordingly, the post-posed subject expression in such constructions is obligatory: (65) *m-òn-dju´wó mw-a´-hìtí. 18-9-house sm18-past-enter

Otherwise, the analysis is exactly as in the passive. A parse of the initial locative expression is taken to decorate an unﬁxed node, and the subject marker provides a locally unﬁxed node, which duly uniﬁes with the initially constructed unﬁxed node. The tense speciﬁcation then provides a skeletal subject predicate structure, but without any decoration of the subject node it introduces other than assigning it a requirement ?Ty(e). By placing the pointer at the predicate node, however, it is the predicate that is then constructed, as in the passive by the parsing of the verb stem. There is no object marker, as we would now expect, since there remains an outstanding locally unﬁxed node in the construction waiting to be resolved. This can, however, unify with the locative, on the assumption that these can be taken as arguments of the predicate. This then completes that predicate, almost leading to a complete propositional formula. However, the subject node still has its type requirement outstanding as the tense marker does not project any pronominal-like meta-variable. Hence, having compiled up a predicate interpretation, the pointer has to return to the subject node, at which stage the subsequent ‘postposed’ subject expression can duly decorate this node as one of the very last steps in the construction process. The main diﬀerence between our analysis of the passive and locative inversion is that the passive marker provides the annotation for the subject, while in locative inversion constructions, the subject decoration is provided only as a very late step in the construction process. As is often noted, this structure is characteristically associated with a focus construal of the subject, and we would take this to be a consequence of its highly marked late placement, indicating that everything is in some sense part of the context which it updates. Structurally however, both constructions follow a similar strategy. Both involve the construction of a locally unﬁxed node at the outset of the parse, which is only uniﬁed at the object node once this has been introduced by the verb. Thus, as in the passive, no object markers are possible in locative inversion as long as the verb has not been parsed: (66) *m-on-djuwo mw-a-ri-hiti. 18-9-house sm18-past-om5-enter

The reason for this is the same as for the passive, namely that no locally unﬁxed node can be built as long there is another locally unﬁxed node present. Since the

Concepts of structural underspeciﬁcation in Bantu and Romance 

locally unﬁxed node with the locative clitic is uniﬁed in object position, the postverbal subject can be expressed by a clitic: (67) m-òn-djúwó mw-á-hìtí-rò. 18-9-house sm18-past-enter-om5 ‘He/she entered the house.’

It is sometimes claimed that a crucial characteristic of locative inversion is that it is restricted to a certain set of predicates, characterized by their thematic roles, for example unaccusatives in Chichewa (Bresnan & Kanerva 1989). Since Dynamic Syntax does not have any notion of “thematic role”, this appears to be problematic for our analysis. However, whatever the situation in Chichewa may be, as it turns out, in Otjiherero no such restriction exists (cf. Marten 2006). In (68), the transitive predicate tj`anga “write” is used in a locative inversion construction. It is marked by an applicative which in this case introduces the locative argument, further buttressing our analysis of construing the locative within the predicate. (68) p-òn-djúwó p´e-tjéng-èr-à òvá-nátjè òmbàpírà. 16-9-house sm16.hab-write-appl-fv 2-children 9.letter ‘At the house write children a letter.’

As in our analysis of the Otjiherero passive, we have shown that locative inversion can be analysed by assuming that initial clitics in Otjiherero are taken to induce a locally unﬁxed node which can be uniﬁed either with the subject node or, as in passives and locative inversion, with the object node. The analysis brings out both the similarities of the constructions, in particular the fact that the verb-initial clitic agrees with an object NP, and the diﬀerences, in particular that in the passive the interpretation of the subject results from lexical information from the passive marker, while in locative inversion, the logical subject is found in immediate postverbal position.

6. Reﬂections and directions for the future In this paper, we have taken relatively familiar data from Romance and summarized the basis for a typology of left and right periphery eﬀects, using the Dynamic Syntax notions of LINK and *Adjunction, in conjunction with restrictions encoded in the lexical speciﬁcations of diﬀerent clitics, to derive diﬀerent wordorder eﬀects. After illustrating the analysis with respect to Spanish, we extended it to Bantu, in particular Otjiherero and showed how word-order variation in Bantu can be captured with the very same strategies introduced for Romance, deriving

 Lutz Marten, Ruth Kempson & Miriam Bouzouita

similar eﬀects on the right and left periphery, through similar interaction between lexically encoded restrictions on the interpretation of pronominal elements and the processes of update involved in the construction of unﬁxed nodes and linked structures. While this gave an account of word-order variation found in the two language families, for the analysis of clitics (in Romance) and agreement markers (in Bantu), we introduced in addition the notion of local underspeciﬁcation, a mechanism deﬁned to reﬂect incremental processing in very free constituent order systems. While in Romance we have analysed the construction of locally unﬁxed nodes for clitics as remnant reﬂexes of the case system of ProtoRomance/Latin as it applied in licensing scrambling eﬀects, in Bantu we have simply assumed that the projection of locally unﬁxed nodes is part of being a Bantu clitic. By applying this analysis to active, passive and locative inversion constructions in Otjiherero, we have shown how a number of structural properties of Otjiherero are predicted by the analysis: the availability of only one object marker per verb, unrestricted as to case or thematic role; the absence of pre-verbal object markers in passive and locative inversion; the ﬁnal position of the passive morpheme in the string of derivational suﬃxes; the obligatory post-verbal position of the logical subject in locative inversion, and the association with presentational focus of locative inversion constructions. After having shown the empirical advantages of our analysis, we turn in this ﬁnal section brieﬂy to some conceptual implications. In Bantu linguistics, it is usually assumed that subject and object markers are part of the verbal morphology. Our analysis departs from this assumption, by proposing an analysis which essentially accords these markers much more structural freedom. We hope with this to have opened a more general discussion about agreement in Bantu, which might shed light not only on passives and locative inversion, but also, for example, on subject-object reversal constructions (to which the present analysis could in principle be extended), topic agreement, and default expletive agreement markers found in many Bantu languages. Based on the analysis proposed here, the question of multiple object markers can be more precisely addressed, where it is worth exploring whether multiple object markers may be analyzed as complex locally unﬁxed nodes, or as a mixture of ﬁxed and unﬁxed nodes. In this connection, there is also the possibility that these reﬂect some earlier scrambling stage more directly, and involve the construction of complexes of argument nodes, parallelling the Romance clitics and their mirroring of earlier Latin scrambling. By providing a new way of thinking about Bantu subject and object agreement marker, we hope to have opened avenues for thinking about a number of issues in Bantu grammar from a diﬀerent perspective. With regard to the DS approach, the analyses presented here have addressed two main issues. Firstly, a key element in our analysis is the use of underspeciﬁcation of diﬀerent kinds. Underspeciﬁcation of content is relevant for the analysis of

Concepts of structural underspeciﬁcation in Bantu and Romance 

pronominal elements which encode an underspeciﬁed formula value in need of contextual enrichment. Structural underspeciﬁcation is at the core of our use of *Adjunction which allows the introduction of some information at a node whose position in the tree is not fully speciﬁed at the time it is introduced, and thus needs to be ﬁxed at a later stage. These two diﬀerent forms of underspeciﬁcation are combined in the analysis of Romance clitics and Bantu agreement markers as encoding both underspeciﬁed pronominal content and the construction of locally unﬁxed nodes. By comparing Spanish as illustrative of Romance and Herero as illustrative of Bantu, we have shown how these diﬀerent forms of underspeciﬁcation play a role in both language families, while being encoded and restricted in diﬀerent idiosyncratic ways to yield a range of variation within an overall principled analysis. Secondly, the discussion presented addresses the problem of the representation of thematic roles and grammatical functions within the theory. Without making use of thematic roles or grammatical functions as primitives of the theory, both passive and locative inversion structures seem to pose a problem for Dynamic Syntax, since these notions are often thought to be essential for a successful analysis. However, we have shown that by employing locally unﬁxed nodes, we derive the appropriate semantic representations in a left-to-right fashion, without invocation of these concepts. It remains to be seen whether this line of thinking can be extended to the analysis of other constructions often associated with thematic roles and grammatical functions in other languages. As a ﬁnal remark, we address the question of the more general signiﬁcance to be drawn from our proposed account of the “surface subject” in Bantu. The concept of locally unﬁxed node, central to the Bantu analysis, was originally developed for scrambling eﬀects (Cann et al. 2005), and applied to model behavior of case-marked Romance clitics (Kempson & Cann 2007). However, as we have shown, the mechanisms needed for such an account indubitably apply to the case of subject, as we have shown, enabling one to provide a new basis for grappling with the otherwise puzzling concept of syntactic subject. Though the behavior of locatives in Otjiherero is idiosyncratic to the Bantu languages, what it suggests more generally is that the concept of syntactic subject may be heterogeneous, a genuine epiphenomenon. As both our Bantu and Romance analyses show, subjects can be introduced into the parse through a variety of means, as decorating ﬁxed, unﬁxed or linked nodes, and as being coreferential with pronouns, clitics or agreement markers. From a parsing perspective, with alternative strategies available at any point, and with diﬀerent calciﬁcations over time consolidating to give rise to subtly diﬀerent patterns, what is common to all such subject marking is a parsing identiﬁcation of a distinct emergent local predicate-argument structure. This new light on the reducibility of the concept of syntactic subject to a number of diﬀerent forms of tree growth update, integral as it is to a procedural

 Lutz Marten, Ruth Kempson & Miriam Bouzouita

perspective, opens up new ways of looking at a well-known but mysterious phenomenon. Indeed, the general dynamics of the parsing perspective promises to provide new solutions to the family of puzzles associated with the interaction of structural processes in natural language and the anaphoric devices such as clitics and agreement markers which languages make available.

References Anagnostopoulou, E., van Riemsdijk, H. & Zwarts, F. (Eds). 1997. Materials on Left Dislocation. Amsterdam: John Benjamins. Blackburn, P. & Meyer-Viol, W. 1994. Linguistics, logic and finite trees. Bulletin of the Interest Group for Pure and Applied Logic 2: 3–29. Bouzouita, M. in preparation. The Diachronic Development of Spanish Clitic Placement. Ph.D. Dissertation, King’s College London. Bouzouita, M. 2007. Processing factors in syntactic variation and change: Clitics in Medieval and Renaissance Spanish. In Historical Linguistics 2005, J. Salmons & S. Dubenion-Smith (Eds), 51–71. Amsterdam: John Benjamins. Bouzouita, M. & Kempson, R. 2006. Clitic placement in Old and Modern Spanish: A Dynamic Account. In Competing Models of Linguistic Change, O. Nedergaard Thomsen (Ed.), 253–268. Amsterdam: John Benjamins. Bresnan, J. & Kanerva, J. 1989. Locative inversion in Chichewa. Linguistic Inquiry 20: 1–50. Bresnan, J. & Mchombo, S.A. 1987. Topic, pronoun, and agreement in Chichewa. Language 63: 741–782. Cann, R. & Wu, Y. 2006. The dynamic syntax of Chinese passive constructions. MS., University of Edinburgh and University of Hong Kong. Cann, R., Kempson, R. & Marten L. 2005. The Dynamics of Language. Oxford: Elsevier. Cardinaletti, A. this volume. On clitic clusters. In The Bantu-Romance connection. A comparative investigation of verbal agreement, DPs and information structure, C. De Cat & K. Demuth (Eds). Amsterdam: John Benjamins. Cocchi, G. 2001. Free clitics and bound affixes: Towards a unitary analysis. In Clitics in Phonology, Morphology and Syntax, B. Gerlach & J. Grijzenhout (Eds), 85–119. Amsterdam: John Benjamins. Cuervo, C. 2002. Spanish clitics: Three of a perfect pair. MS., MIT. De Cat, C. 2007. Dislocation without movement. Natural Language and Linguistic Theory 25: 485–534. Downing, L. 2006. The prosody and syntax of focus in Chitumbuka. In Papers in Bantu Grammar and Description [ZAS Papers in Linguistics 43], L. Downing, L. Marten & S. Zerbian (Eds), 55–79. Berlin: ZAS. Franco, J. 2000. Agreement as a continuum. In Clitic Phenomena in European Languages, F. Beukema & M. den Dikken (Eds), 147–188. Amsterdam: John Benjamins. Heap, D. 2005. Constraining Optimality: Clitic sequences and Feature Geometry. In Clitic and Affix Combinations, L. Heggie & F. Ord´on ˜ ez, 81–102. Amsterdam: John Benjamins. Hyman, L.M. 2003. Suffix ordering in Bantu: A morphocentric approach. In Yearbook of Morphology 2002, G. Booij & J. van Marle (Eds), 245–281. Berlin: Mouton. Kaplan, R. & Zaenen, A. 1989. Long-distance dependencies, constituent structure, and functional uncertainty. In Alternative Conceptions of Phrase Structure, M. Baltin & A. Kroch (Eds), 17–42. Chicago: Chicago University Press.

Concepts of structural underspeciﬁcation in Bantu and Romance 

Kempson, R. & Cann, R. 2007. Dynamic Syntax and dialogue modelling: preliminaries for a dialogue-driven account of syntactic change. In Historical Linguistics 2005, J. Salmons & S. Dubenion-Smith (Eds), 73–101. Amsterdam: John Benjamins. Kempson, R., Cann, R. & Kiaer, J. 2006. Topic, focus and the structural dynamics of language. In The Architecture of Focus, S. Winkler & V. Molnár (Eds), 59–82. Berlin, New York: Mouton de Gruyter. Kempson, R., Kiaer, J. & Cann R. 2007. Periphery effects and the dynamics of tree growth. In Dislocated Elements in Discourse, B. Shaer, W. Frey & C. Maienborn (Eds), 141–170. London: Routledge. Kempson, R., Meyer-Viol M. & Gabbay D. 2001. Dynamic Syntax. Oxford: Blackwell. Kiaer, J. 2007. Processing and Interfaces in Syntactic Theory: the case of Korean. Ph.D. dissertation, King’s College London. Marten, L. 2006. Locative inversion in Otjiherero: More on morpho-syntactic variation in Bantu. In Papers in Bantu Grammar and Description [ZAS Papers in Linguistics 43], L. Downing, L. Marten & S. Zerbian (Eds), 97–122. Berlin: ZAS. Marten, L. 2007. Focus strategies and the incremental development of semantic representations: evidence from Bantu. In Focus Strategies in African Languages, E.O. Aboh, K. Hartmann & M. Zimmermann (Eds), 113–135. Berlin/New York: Mouton de Gruyter. Marten, L. & Kempson, R. 2002. Pronouns, agreement, and the dynamic construction of verb phrase interpretation: A Dynamic Syntax approach to Bantu clause structure. Linguistic Analysis 32: 471–504. Mchombo, S. 2005. The Syntax of Chichewa. Cambridge: Cambridge University Press. McCormack, A. 2008. Object marking in Tswana: A Dynamic Syntax analysis. Ph.D. dissertation, School of Oriental and African Studies. Meeussen, A.E. 1967. Bantu grammatical reconstruction. Africana Linguistica 3: 80–122. Meyer-Viol, W. 1995. Instantial Logic: An investigation into reasoning with instances. Ph.D. dissertation, University of Utrecht. Monachesi, P. 2005. The Verbal Complex in Romance, Oxford: Oxford University Press. Mvungi, M. n.d. Lwidiko. Dar es Salaam: Tanzania Publishing House. Nordlinger, R. 1998. Constructive Case: Evidence from Australian languages. Stanford: CSLI. Poletto, C. 2000. The Higher Functional Field: Evidence from northern Italian dialects. Oxford: Oxford University Press. Rivero, M-L. 2008. Oblique subjects and person restrictions in Spanish: a morphological approach. In Agreement Restrictions, R. D’Alessandro, S. Fischer & G.H. Hrafnbjargarson (Eds). Berlin: de Gruyter. Rizzi, L. 1997, The fine structure of the left periphery. In Elements of Grammar, L. Haegeman (Ed.), 289–330. Dordrecht: Kluwer. Sperber, D. & Wilson, D. 1995. Relevance: Communication and Cognition. Oxford: Blackwell. Woolford, E. 1995. Why passive can block object marking. In Theoretical Approaches to African Languages, A. Akinlabi (Ed.), 199–215. Trenton, NJ: Africa World Press. Wu, Y., 2005. The Dynamics of Left and Right Dislocation on Chinese. Ph.D. dissertation, University of Edinburgh. Zagona, K. 2002. The Syntax of Spanish. Cambridge: Cambridge University Press. Zubizarreta, M.L. 2001. The constraint on preverbal subjects in Romance interrogatives: a Minimality effect. In On Romance Inversion, A. Hulk & J-Y Pollock (Eds), 183–204. Oxford: Oxford University Press.

On different types of clitic clusters* Anna Cardinaletti

Ca’ Foscari University of Venice In this paper, Italian and Bantu clitic clusters are analyzed and compared. I claim that in both language families, pronouns check case in a low clitic position, and in so doing they reverse the order of arguments. In Italian, clitics move to a high clitic position where they check person and number features. Different types of Italian clusters are individuated: they differ syntactically, phonologically, and morphologically. As in Kayne’s (1994) proposal, clusters can be formed either on one single functional head or on adjacent heads; the former type can appear both in enclisis and proclisis, the latter only in proclisis. Clitic pronouns can end with an epenthetic vowel or a class marker/ inflectional morpheme; the former (consonantal clitics) can appear in clitic clusters in any position, the latter (morphologically complex clitics) cannot be the first element in clitic clusters dominated by one single head. Finally, some clusters are inserted as lexical units, others are two independent words: only the former display the linking vowel [e]. The intricate interplay of these (partially) independent properties explains a number of restrictions on clusters found in Italian (and other Romance languages). The hierarchy of person and number features in the high clitic position is also discussed, which explains other restrictions on clusters.

1. Introduction The purpose of this paper is twofold. On the one hand, it shows that in spite of the many previous discussions of Italian and Romance clitic clusters, some

*The original nucleus of this analysis was presented at the LSA Summer Institute workshop on Perspectives on Clitic and Agreement Affix Combinations held at the University of Illinois, Urbana, on July 28 1999, at the University of Thessaloniki in September 1999, at the University of Geneva in February 2000, at the workshop on Antisymmetry, held in Cortona on May 15–17 2000, and at the workshop on The Bantu-Romance Connection, held in Leeds on May 25–27 2006. The audiences are kindly thanked for questions and comments. Many thanks also go to Valentina Bianchi, Franca Ferrari, Richie Kayne, Andrew Nevins, and two anonymous reviewers for very stimulating comments and criticism.

 Anna Cardinaletti

roperties have gone unnoticed in the literature. Mainly using Italian data, I discuss p the properties listed below: i. not all clitic clusters occurring in proclisis are also possible in enclisis. In Italian, proclisis seems to be more liberal than enclisis.1 Following Kayne (1994), I make the hypothesis that (Italian) clusters can be of two main types, in accordance with the LCA: in one type, the pronouns in the cluster occur on one and the same functional head; in the other, the clitics occur on different (adjacent) heads. Due to verb movement, enclisis is possible only with clusters which form a single constituent; ii. assuming that there are two clitic positions in the clause, one very high in the IP layer (INFL, the “clausal” clitic position) and one very low in the VP layer (the “lexical” clitic position) (Cardinaletti & Shlonsky 2004), only clitic clusters occurring on one and the same head are possible in the low clitic position, which only allows enclisis. The clausal clitic position must instead be thought of as a series of adjacent heads; iii. in the spirit of Bianchi (2006a), among others, clitic pronouns check person and number features in the clausal clitic position. Person and number features are not checked in the low clitic position, where case is checked; iv. clusters containing 3rd person indirect object (IO) clitics, such as Italian cluster glielo “to-him it”, do not behave differently from clusters containing 1st and 2nd person IO clitics such as me lo “to-me it” and te lo “to-you it”. It is thus not necessary to assume a special syntax for glielo, pace Laenzlinger (1993), Heggie & Ordóñez (2005: 26), Bianchi (2006a: 2038), among others. I suggest that the similarity stems from the fact that Italian 3sg IO gli, not being marked for number features, is a person clitic like mi, ti, si, etc. Since person and number seem to be mutually exclusive features, French lui and Spanish le “to him”, which are marked for number, are not person clitics. On the other hand, I show that in extending the comparison to a different language family such as Bantu, some of the conclusions drawn from the analysis of Romance clusters should be revised. In particular: i. contrary to apparently abundant evidence from Romance, IO – DO is presumably not the order in which clitics are merged, but DO – IO is, as is clearly shown by Bantu languages, where the argument order is reversed by clitic pronouns;2 . Italian differs from Greek, where enclisis is more liberal than proclisis (Terzi 1999). . Following Dryer (1983), Bresnan and Moshi (1990), Krifka (1995), and Cocchi (2000a: 88ff), Bantu so-called object markers are taken here to be clitic pronouns like Italian/Romance ones. Differently from Hyman & Duranti (1982), I do not distinguish between clitic object markers (e.g., those occurring with no overt object and in Left dislocation) and “true” object agreement (i.e., those co-occurring with an in situ object). This case can be taken to be an instance of clitic

On different types of clitic clusters 

ii. Italian clusters like me lo and glielo are not telling on the internal order of clitic clusters since there is some evidence that they are merged as lexical units. Their internal order rather complies with the person hierarchy in the high clitic position; iii. the universal status of the Person Case Constraint (Bonet 1991) should be reconsidered. The order of the clitics inside the cluster seems to be important and should be taken into consideration to understand this constraint. Although I will not give a full account of the differences between Romance and Bantu, I will try to relate the contrast between Italian and Bantu to an independent difference between the two language families: in Italian, clitics appear before the finite verb, in Bantu they occur between the tense marker and the lexical verb. It is uncontroversial that Italian proclitic pronouns occur in the high clitic position. Following Barrett-Keach (1986) and Krifka (1995: 1412–4), I assume that in Bantu, the Tense affix is in INFL. Object clitics must thus utilize what I have called the low clitic position. In Italian, person restrictions seem to be operative also in the low clitic position. This can be accounted for by establishing a relation between the two clitic positions, which is visible in the case of clitic climbing. Since in Bantu, clitics do not raise to the high clitic position, no relation between the low and the high clitic position is established.

2. Cluster internal restrictions Not all languages that have clitic pronouns display clitic clusters. In those which do, like Italian and some of the Bantu languages (Bresnan & Moshi 1990), clitic clusters display order restrictions not displayed by full arguments. These restrictions may be of a different sort. We start from one type of these restrictions.3

doubling similar to what is found e.g., in Spanish (Jaeggli 1982; Suñer 1988; Uriagereka 1995): it displays similar properties, e.g., sensitivity to grammatical function, definiteness, referentiality, etc. (Krifka 1995: 78f; Cocchi 2000a: 89f). As for subjects, Bantu displays systematic subject clitic doubling: subject clitics are obligatory, and double a dp subject when present (similarly to what happens in Northern Italian dialects, Cocchi 2000a: 100); see also Kinyalolo (2003: 346). . When writing about clitic clusters, it must be mentioned that there is much poorly understood Italian-internal variation. Some readers will not agree with some judgments reported in this paper. Nor do I agree with some data reported in the literature. Just to give two examples, I find Gianni ce ne parlerà “Gianni there on-it will-talk” (Laenzlinger 1993: 265) ungrammatical, and I am among the speakers who do not accept Ti/Vi ci affideranno “[they] you to-us willentrust” marked as % in Bianchi (2006a: 2039, n. 34).

 Anna Cardinaletti

The order of two clitic pronouns may vary depending on the clitic pronouns found in the cluster. For instance, in Italian, the unmarked order of full complements is “direct object (DO) – locative (LOC)”, (1a), but in clitic clusters, the locative may follow or precede the other clitic depending on its feature specification: cf. mi ci (1b) vs. ci si/ce lo (1c).4 A similar restriction is found in French. While 1st and 2nd person IO clitics precede DO clitics (2a), the order is reversed when the IO is 3rd person, (2b): (1) a.

Metterà me/sé/lui in quel posto. [he] will-put me/himself/him in that place

b. Mi ci metterà. [he] me there will-put

DO – loc

c.

Ci si/Ce lo metterà. [he] there himself/it will-put

loc – DO

(2) a.

Jean me/te le donne. Jean to-me/to-you it gives

IO – DO

b. Jean le lui/leur donne. Jean it to-him/to-them gives

DO – IO

These differences are also relevant in a comparative perspective. Languages may differ with respect to the order displayed by clitics inside the cluster. Italian and French differ with respect to the order of DO and IO clitics (e.g., the order in (2b) is opposite to the one of the Italian counterpart glielo) and of accusative and locative clitics, (3a) vs. (3b): (3) a.

Ce lo metterà. [he] there it will-put

b. Jean les y a rencontrés. Jean them there has met

loc – DO DO – loc

As shown in (4) and (5), Italian and Bantu differ with respect to the order of IO and DO clitics (in (5) and throughout, the glosses of the Bantu examples are those found in the quoted works): (4) a.

Me lo/Glielo darà. [he] to-me it/to-him it will-give

IO – DO

b. *Lo gli/mi darà.

. In example (1c) and throughout, the 3sg.masc DO lo stands for the whole DO clitic series, containing the four possible combinations of number and gender features: lo (3sg.masc), la (3sg.fem), li (3pl.masc), le (3pl.fem).

(5) a.

On different types of clitic clusters 

A-ka-bi-mú-h-a. he-P3-them-him-give ‘He gave them to him.’

DO – IO (Haya, Hyman & Duranti 1982: 221)

b. Umugabo y-a-ki-ba-haa-ye. DO – IO (Kinyarwanda, Dryer 1983: 132) man he-past-it-them-give-asp ‘The man gave it to them.’ c.

Mukaji u-tshi-mu-p-a. woman 1.su-7.do-1.IO-give-I ‘The woman gives it to him.’

DO – IO (Tshiluba, Cocchi 2000b: 50)

Bantu languages also differ from Romance languages in that differences like those seen in (1) and (2) are not found. The DO – IO order is also found with 1st and 2nd person clitics: (6) a.

A-ka-mu-ku-léét-el-a. he-P3-him-you-bring-app ‘He brought him to you”.

DO – IO (Haya, Hyman & Duranti 1982: 231)

b. Nu-mu-m-pe. you:pl-1.DO-1sg.IO-give ‘Give her to me!’

DO – IO (Tshiluba, Willems 1949)

I start by presenting a proposal concerning the data in (1)–(3) based on the case properties of clitic pronouns. I will conclude that case is not the (only) property responsible for the word order found inside clitic clusters. We will then move on to the main empirical result of this paper: namely a restriction never noticed before as to which clusters can appear in proclitic and enclitic position in Italian, and to the discussion of this restriction. This will allow us to address the different clitic order in Italian and Bantu pointed out above.

3. On a case-approach to clitic clusters In Romance languages, 1st and 2nd person clitics are not marked for case, being possible both as DO and IO, while 3rd person clitics are marked for case. Two different forms for dative and accusative exist: cf. Italian gli (to-him) vs. lo (him), le (to-her) vs. la (her). A classification of French and Italian clitic pronouns is reported in (7) and (8), respectively:5

. Differently from French leur, which is a clitic pronoun, Italian 3pl IO loro is a postverbal weak pronoun (Cardinaletti 1991) and does not enter the present discussion.

 Anna Cardinaletti

(7) a. clitics+case: b. clitics–case:

le, la, les, lui, leur, y, en me, te, nous, vous, se

(8) a. clitics+case: b. clitics–case:

lo, la, li, leacc, gli, ledat, ne mi, ti, ci, vi, si

Assuming that the licensing of case-marked clitics (selected incorporation) is different from that of non-case-marked clitics (free incorporation), Laenzlinger (1993) suggests that: (i) two clitics marked for case cannot combine on the same node; and (ii) case-marked clitics must be closer to the host than non-case-marked clitics. This proposal explains the possibility of clusters like French me le in (2a) and their order: me le vs. *le me. The clusters le lui/le leur in (2b) circumvent the constraint in (i) in that lui/leur are taken to be adjoined to a head lower than the one le adjoins to. Locative clitic y and partitive clitic en behave like lui and leur: they give rise to the clusters in (9) where they follow other clitics. Y and en are also taken to occur in a lower head (Laenzlinger 1993: 261f) (notice that lui and y cannot co-occur):6 . An alternative way of phrasing Laenzlinger’s proposal is to say that lui, leur, y and en are a different type of pronoun, not clitic but weak in the sense of Cardinaletti and Starke (1999). In French, these pronouns would not occupy the clitic position discussed so far, but a lower position. There is some evidence that these French pronouns are weak rather than clitic. Consider first the fact that lui in (2b) patterns like postverbal moi in (ib) in that it follows the DO clitic: (i)

a. b.

Il me le donne. Donne-le-moi!

clitic + clitic clitic + weak

Following previous work (Cardinaletti & Starke 1999: 221, n. 32), I consider IO moi as a weak pronoun, which is morphologically different from both clitic me and strong à moi. The same analysis can be suggested for IO lui, which is morphologically different from strong à lui in the same way. If this is correct, the series lui – à lui would not have a clitic counterpart, differently from the tripartite series me – moi – à moi:

(ii) a. b.

clitic weak strong me moi à moi – lui à lui

Evidence for the weak status of y comes from the fact that in some fixed expressions, as in (iiia), it can occur in the position preceding a past participle, a context which is not available to clitic pronouns, as shown by Italian (iiib): (iii) a. y compris/inclus … there included b. *vi compreso … c. ivi compreso …

weak pronoun clitic pronoun weak pronoun

Interestingly, (formal) Italian allows (iiic), where the locative pronoun has the longer form ivi.

On different types of clitic clusters 

(9) a.

Jean les y a rencontrés. Jean them there has met

b. Jean l’en informera. Jean him about-it will-inform c.

Marie lui en parle. Marie to-him about-it speaks

d. Jean y en parlera. Jean there about-it will-speak

Laenzlinger’s proposal allows us to understand why in Italian, as in French, sequences like me lo, te lo, ce lo, ve lo, se lo (to-me/to-you:sg/to-us/to-you:pl/ to-refl it) are possible and occur in that order. It can also explain the different order between French les y and Italian ce lo (see (3) above): since Italian locative ci is the same lexical item as 1pl ci (see fn. 9), it differs from French y in that it is not marked for locative case. It can adjoin to an accusative clitic and must precede it. This proposal seems to find support in Bantu languages, where clitics are not marked for case and restrictions like those seen in Romance are not found. Two 3rd person clitics can combine, as shown in (5). That the two clitics are not specified for case is confirmed by the fact that in some Bantu languages, sentences like (10) are ambiguous: (10)

A-ka-bi-ba-léét-el-a. he-P3-them-them-bring-APPL ‘He brought them (bi) to them (ba).’ ‘He brought them (ba) to them (bi).’

(Haya, Hyman & Duranti 1982: 232)

Laenzlinger uses his analysis also to rule out the ungrammatical sequence in (11). In Spanish, two 3rd person clitics, one accusative and one dative, are banned. In the grammatical cluster, dative le is replaced by non-case-marked se (spurious se): (11) *le(s) lo

→

se lo

‘to-him/them it’

Ivi can be taken to be the weak counterpart of the locative clitic pronoun vi (see note 10), analyzed as i + vi = support morpheme + clitic pronoun (see Cardinaletti 1994 and Cardinaletti & Starke 1999: 193f for support morphemes). Due to the choice principle of Cardinaletti and Starke (1999), vi is always chosen over ivi because it has a smaller structure. In (iiic), the weak pronoun ivi is ruled in because with past participles, clitic vi is independently ruled out, (iiib). The phenomenon is not productive, however. For instance, the weak form ici (built on locative clitic ci) does not exist. For some differences between Italian and French as regards coordination of clitic pronouns, which might support the above analysis, see Benincà and Cinque (1993: 2323).

 Anna Cardinaletti

A problematic case for a case-based approach is, however, provided by the Leismo dialects of Spanish, in which le is not case-marked (being used for both accusative and dative 3rd person animates); yet it is replaced by se in cases like (11) (see Nevins 2007 for relevant discussion). Italian provides other problematic cases for a case-based approach to clitic clusters: (i) the clusters gli/le si in (12), where case-marked IO clitics gli and le combine with impersonal and reflexive si, respectively; and (ii) the cluster lo si in (13), where case-marked DO clitic lo combines with impersonal si: (12) a.

Gli/Le si è parlato. to-him/to-her imp has spoken

b. Gli/Le si è presentato. [he] to-him/to-her refl has introduced (13) Lo si è visto. him imp has seen ‘One has seen him.’

In (12) and (13), the case-marked clitics gli, le, and lo occur further from the host than the non-case-marked clitic si, in contrast with Laenzlinger’s condition (ii). The question thus arises as to how gli, le, and lo are licensed in (12) and (13) (also see gli ci clusters). Furthermore, it is clear that what matters in the different word orders found with locative ci (see (1) above) is not case but other properties of the pronouns in the cluster. In conclusion, there are some problematic clusters for a case-based approach to clitic clusters. In the following sections, we turn to two other problematic cases. 3.1 On the apparent special status of Italian glielo Laenzlinger (1993: 254) suggests that the ungrammaticality of the Italian clusters *le lo and *le ne is due to the same constraint that rules out (11) in Spanish: two case-marked clitics cannot combine on one and the same node. To get a grammatical sequence, feminine dative le is replaced by (masculine) gli: (14) a. *ledat lo b. *ledat ne

→ →

glielo gliene

‘to-her it’ ‘to-her of-it’

Since gli is also a case-marked pronoun (see (8a)), this change is, however, unexpected. To explain it, Laenzlinger (1993: 253) suggests that glielo forms “a unique clitic constituent at the structural level, as opposed to me lo which are two independent clitic constituents”. He provides the impossibility of (15) as an argument for the clitic compound analysis:

On different types of clitic clusters 

(15) *Glielo o la presenterà? [he] to-him him or her will-introduce?

(Laenzlinger 1993: 254, fn.13)

But this test does not distinguish glielo from other “IO – DO” clusters, which are also impossible in coordination, (16). Something different should be said about Italian glielo: (16) *Me lo o la presenterà? [he] to-me him or her will-introduce?

Laenzlinger’s claim that glielo has a special status among Italian clitic clusters is shared by other researchers. For instance, Heggie & Ordóñez (2005: 26) have recently proposed that glielo is an “amalgamation”, to be best analyzed in the morphology. Bianchi (2006a: 2038) assumes that in glielo, the two clitics adjoin to one another and move as a unit, while in e.g., me lo, the two clitic pronouns move separately (see section 5.1 below for the analysis of different cluster types in terms of Kayne’s 1994 LCA). The common property of these accounts is that clusters containing 3rd person IOs are treated differently from those containing 1st and 2nd person IOs. But the two sequences me lo and glielo must be treated alike. As I show below, they share all syntactic and phonological properties: they build the same type of cluster, which is possible in both proclisis and enclisis, and in both cases the final vowel of the first clitic is [e] instead of [i]. There is no reason to treat glielo differently from e.g., me lo. In conclusion, Italian glielo seems to be another problematic case for a case-based approach to clusters in that it unexpectedly behaves like e.g., me lo. 3.2 Mi ti combinations A further complication is provided by mi ti combinations, which are possible for some speakers of Italian (Seuren 1976: 32; Evans et al. 1978: 160; Monachesi 1995: 42; Bianchi 2006a: 2040) and display a difference between proclisis and enclisis not encountered so far. Out of the six combinations of grammatical functions found in proclisis, only three can appear in enclisis, namely DO – IO sequences:7 (17) a.

Lui mi ti presentò/affidò. he me to-you introduced/entrusted

miDO tiIO

. Among 1st and 2nd person clitics, mi ti is the only possible combination. The other ones are ungrammatical, in either order and with either interpretation: *mi vi/*vi mi “me you:pl”, *ti ci/*ci ti “you:sg us”, *vi ci/*ci vi “you:pl us”. Two of these combinations are exemplified by *Ti/*Vi ci affideranno in note 3. This contrast suggests, as in Kayne (2000), that mi and ti form a natural class, while ci and vi do not belong to the very same class. *Ti vi/*Vi ti and *mi ci/*ci mi are cases of overlapping reference.

 Anna Cardinaletti

b. Lui pensa di presentarmiti/affidarmiti. he thinks to introduce/entrust me to-you

miDO tiIO

(18) a. ?Lui mi ti presentò/affidò quando eri piccolo. he to-me you introduced/entrusted when [you] were a child

miIO tiDO

b. *Lui pensa di presentarmiti/affidarmiti. he thinks to introduce/entrust to-me you

*miIO tiDO

(19) a.

Io mi ti presenterò/affiderò. I myself to-you will-introduce/will-entrust

mirefl tiIO

b. Io penso di presentarmiti/affidarmiti. I think to introduce/entrust myself to-you

mirefl tiIO

(20) a.

Tu mi ti presenti/affidi così? you to-me yourself introduce/entrust so

miIO tirefl

b. *Tu pensi di presentarmiti/affidarmiti così? you think to introduce/entrust to-me yourself this-way

*miIO tirefl

(21) a.

miDO tirefl

Tu mi ti prendesti come segretaria, non come baby-sitter. you me to-yourself took as secretary, not as baby-sitter

b. Tu pensavi di prendermiti come segretaria. you thought to take me to-yourself as secretary

miDO tirefl

(22) a. ?Io mi ti prendo come segretaria. I to-myself you take as secretary

mirefl tiDO

b. *Io penso di prendermiti come segretaria. I think to take to-myself you as secretary

*mirefl tiDO

This is another problematic case for a purely case-based approach to clitic clusters. Before suggesting a different approach, based on both case and person/number properties of clitic pronouns, I turn to the discussion of the different types of clitic clusters found in Italian, some of which display an asymmetry between proclisis and enclisis similar to what we have just seen with mi ti sequences.

4. Many different types of clitic clusters in Italian In Italian, many different types of clitic clusters exist. They display both syntactic and morphophonological differences. Firstly, while all types of clitic clusters can occur in proclitic position, only some of them are possible in enclitic position. Secondly, only some clitic clusters display morphological changes on the pronouns entering the cluster.

On different types of clitic clusters

4.1 Type 1: Unrestricted clusters with vowel change In Type 1 clusters, the first clitic pronoun changes its form when it appears in the cluster. The final vowel is not [i], but [e] (23b,b') (see section 5.3 for discussion). This type of cluster is possible in both proclitic and enclitic position. Proclisis is found with finite verbs, (23a,a',b,b'); enclitic clusters occur with infinitival and imperative verbs, as shown in (23c,c') and (23d,d'), respectively:8 (23) a.

Mi ha dato un libro. a'. Mi ha dato tre libri. [he] to-me has given a book [he] to-me has given three books

b. Me lo ha dato. b'. Me ne ha dati tre. [he] to-me it has given [he] to-me of-them has given three c. Pensa di darmelo. c'. Pensa di darmene tre. [he] thinks to give to-me it [he] thinks to give to-me of-them three d. Dammelo! d'. Dammene tre! give to-me it give to-me of-them three

Combinations of this type contain 1st and 2nd clitics in their personal and reflexive usages: e.g., me lo in (23) and me ne in (23) and (24), 3rd person dative gli (glielo), reflexive si as in (25), locative ci as in (26), and impersonal si as in (27) and (28): (24) a.

Mi libero/Ti liberi/Si libera/Ci liberiamo/Vi liberate/Si liberano di questo. [I]/[you:sg]/[he]/[we]/[you:pl]/[they] refl get-rid of this

b. Me ne libero/Te ne liberi/etc. c.

Penso di liberarmene/etc. [I] think to get-rid refl of-it

(25) a.

Si è preso un libro. [he] to-refl has taken a book

a'. Si è preso tre libri. [he] to-refl has taken three books

b. Se lo è preso. b'. Se ne è presi tre. [he] to-refl it has taken [he] to-refl of-them has taken three c. Pensa di prenderselo. c'. Pensa di prendersene tre. [he] thinks to take for-refl it [he] thinks to take for-refl of-them three (26) a.

Ci ho messo un libro. [I] there have put a book

a'. Ci ho messo tre libri. [I] there have put three books

. From here on, only examples with infinitives will be provided. Enclisis is also possible with gerunds and absolute past participles. I will not discuss these cases here.



 Anna Cardinaletti

b. Ce l’ho messo. [I] there it have put a book

b'. Ce ne ho messi tre. [I] there of-them have put three

c.

Penso di mettercelo. c'. Penso di mettercene tre. [I] think to put there it [I] think to put there of-them 3

(27) a.

Si è già parlato di questo. imp has already spoken about this

b. Se ne è già parlato. imp of-it has already spoken (=One has already spoken about it) c.

Potrebbe/Ritengo essersene già parlato. [it] could/[I] believe [to] have imp about-it already spoken too much

(28) a.

Si sono comprati troppi mobili, quest’anno. imp have bought too-many furniture, this year

b. Se ne sono comprati troppi. imp of-them have bought too-many c.

Potrebbe/Ritengo essersene comprati già troppi. could/[I] believe have imp of-them bought already too.many

Type 1 combinations are the following: (29)

Type 1: okproclisis, okenclisis, okvowel change a. mi/ti/ci/viIO loDO/negen/part b. mi/ti/ci/virefl.IO loDO/negen/part c. gliIO loDO/negen/part d. sirefl.IO loDO/negen/part e. ciloc loDO/negen/part f. mi/ti/ci/virefl.DO negen g. sirefl.DO negen h. siimp negen/part

4.2 Type 2: Unrestricted clusters with no vowel change Type 2 clusters are possible in both proclitic and enclitic position, but they do not display any vowel change on the linearly first clitic. Some examples are provided below:9 . In (30)–(32), the following combinations with locative ci are missing because they are independently ungrammatical: (i) a. *Ci ci mette. [he] us there puts

(cf. Ci mette lì/(Lì) ci mette noi) [he] us puts there/(there) [he] there puts us

b. *Ci ci metteremo. (cf. Ci metteremo lì.) [we] ourselves there will-put [we] ourselves will-put there

(30) a.

On different types of clitic clusters 

Mi/Ti/Vi ci metterà. [he] me/you:sg/you:pl there will-put

b. Pensa di mettermici/mettertici/mettervici. [he] thinks to put me/you:sg/you:pl there (31) a. b.

Mi ci metterò/Ti ci metterai/Vi ci metterete. [I] myself/[you:sg] yourself/[you:pl] yourselves there will-put Penso di mettermici/Pensi di mettertici/Pensate di mettervici. [I] think to put myself there/[you:sg] think to put yourself there/ [you:pl] think to put yourselves there

(32) a.

Mi/Ti/Vi/Gli ci metterà un po’ di latte. [he] to-me/you:sg/you:pl/him there will-put some milk

b. Pensa di mettermici/mettertici/mettervici/metterglici un po’ di latte. [he] thinks to put to-me/you:sg/you:pl/him there some milk (33) a.

Gianni ci si metterà. Gianni there himself will-put

b. Gianni pensa di mettercisi. Gianni thinks to put there himself

Type 2 combinations are the following. They all contain locative ci:10 (34) Type 2: okproclisis, okenclisis, *vowel change a. mi / ti / viDO ciloc b. mi / ti / virefl.DO ciloc

c. *Ci ci metterà un po’ di latte. [he] to-us there will-put some milk I take the two cis to be one and the same lexical item which can perform several functions: pronominal and reflexive 1pl, locative, comitative, instrumental. Since they are one and the same lexical item, two cis cannot enter one and the same numeration, which explains the ungrammaticality of (i) (see Grimshaw’s 1997 analysis in terms of the Obligatory Contour Principle). A similar analysis holds for other clitics which can perform more than one function, such as vi (pronominal and reflexive 2pl, and locative, see note 10), si (3rd person reflexive, impersonal, passive), and ne (partitive, genitive, source). Sequences of two vis, two sis and two nes are also ungrammatical (see Cinque 1995: 195f. For *si si see note 13). Ne is the only clitic which allows haplology (Cardinaletti & Giusti 2006: 83): e.g., Se ne sono andati via molti “refl ne are gone away many”. . Formal/Literary registers of Italian also possess locative clitic vi, as in Gianni vi si oppose “Gianni there refl opposed”, Nulla e nessuno sembra avere la forza di opporvisi “nothing and nobody seems [to] have the strength to oppose there refl” (il Manifesto, 11.7.2000, p. 13), (see Evans et al. 1978: 157, n.6). Locative vi is more constrained than locative ci: e.g., it cannot combine with other clitics: compare *Mi/*Ti vi metterà with (30). This pronoun is nowadays used very rarely and is not discussed in this paper.

 Anna Cardinaletti

c. d. e. f.

mi / ti / viIO mi / ti / virefl.IO gli ciloc

ciloc ciloc ciloc sirefl.DO

4.3 Type 3: Combinations which are only possible in proclitic position Type 3 combinations, which contain reflexive si and what Cinque (1988) calls [+argumental] impersonal si, display no vowel change and cannot occur in enclitic position (b. sentences and (39d)). The sentences are possible if the object is realized with a strong pronoun, as in (35c)–(39c), or a full XP, (39e), (40c). These data show that si can occur in enclitic position by itself, but not inside a cluster:11 (35) a.

Gianni mi/ti/ci/vi/gli/le si è rivolto in inglese. Gianni to-me/you:sg/us/you:pl/him/her refl has addressed in English

b. *Gianni pensa di rivolgermisi/rivolgertisi/rivolgercisi/rivolgervisi/ rivolgerglisi/rivolgerlesi in inglese. Gianni thinks to address to-me/you:sg/us/you:pl/him/her refl in English c.

Gianni pensa di rivolgersi a me/a te/a noi/a voi/a lui/a lei in inglese. Gianni thinks to address refl to me/you:sg/us/you:pl/him/ her in English

(36) a.

Gianni mi/ti/ci/vi si è preso a benvolere. Gianni me/you:sg/us/you:pl refl has taken a liking

b.

*Gianni vorrebbe prendermisi/prendertisi/prendercisi/prendervisi a benvolere, ma non ci riesce. Gianni would-like [to] take a liking on me/you:sg/us/you:pl refl, but [he] cannot

c.

Gianni vorrebbe prendersi me/te/noi/voi a benvolere, ma non ci riesce. Gianni would-like [to] take a liking refl [on] me/you:sg/us/ you:pl, but [he] cannot

(37) a.

Gianni ci si metterà un po’ di latte. Gianni there to-himself will-put some milk

. In (36), past participle agreement with 1pl ci and 2pl vi is optional (Ci/Vi si è presi a benvolere), as is always the case with these forms: Gianni ci / vi ha visto / visti “Gianni has seen us/ you:pl”(see note 12 and 29). As for the combinations in (39), note the ungrammaticality of *Non ci si parlò con la dovuta attenzione “not to-us imp spoke with the due attention” and the grammatical counterpart with a strong IO pronoun: Non si parlò a noi con la dovuta attenzione “not imp spoke to us with the due attention”. Impersonal si gets a 1pl interpretation (Cinque 1988) and is thus incompatible with a 1pl clitic pronoun. For cirefl siimp see note 13.

On different types of clitic clusters 

b. *Gianni pensa di mettercisi un po’ di latte. Gianni thinks to put there to-himself some milk c.

Gianni pensa di metterci un po’ di latte per sé. Gianni thinks to put there some milk for himself

(38) a.

Le ci metterà un po’ di latte. [he] to-her there will-put some milk

b. *Pensa di metterleci un po’ di latte. [he] thinks to put to-her there some milk c.

Pensa di metterci un po’ di latte per lei. [he] thinks to put there some milk for her

(39) a.

Non mi/ti/vi/gli/le si parlò con la dovuta attenzione. not to-me/to-you:sg/to-you:pl/to-him/to-her imp spoke with the due attention

b. *Non sembra essermisi/essertisi/esservisi/esserglisi/esserlesi parlato con la dovuta attenzione. [it] not seems [to] have to-me/to-you:sg/to-you:pl/to-him/to-her imp spoken with the due attention c. ?Non sembra essersi parlato a me/a te/a voi/a lui/a lei con la dovuta attenzione. [it] not seems [to] have imp spoken to me/to you:sg/to you:pl/to him/to her with the due attention d. *Ritengo esserglisi parlato di questo. [I] believe [to] have to-him imp spoken e.

Ritengo essersi parlato di questo a Gianni. [I] believe [to] have imp spoken about this to Gianni

(40) a.

(Su quel progetto) non ci si è lavorato abbastanza. (on that project) not there imp has worked enough ‘One hasn’t worked enough on that project.’

b. *(Su quel progetto) sembra/ritengo non essercisi lavorato abbastanza. (on that project) [it] seems/[I] believe not [to] have there imp worked enough c.

Ritengo/Sembra non essersi lavorato abbastanza su quel progetto. [I] believe/[it] seems not [to] have imp worked enough on that project

Type 3 combinations are the following: (41)

Type 3: okproclisis, *enclisis, *vowel change a. mi/ti/ci/viIO sirefl.DO b. gli/leIO sirefl.DO c. mi/ti/ci/viDO sirefl.IO d. ciloc sirefl.IO

 Anna Cardinaletti

e. f. g. h.

leIO mi/ti/ci/viIO gli/leIO ciloc

ciloc siimp siimp siimp

4.4 T ype 4 and 5: Combinations which are independently impossible in enclitic position Some combinations are banned from the enclitic position for independent reasons. They all contain what Cinque (1988) calls [–argumental] impersonal si, which absorbs Nominative case and is independently ruled out in infinitivals. The first case is provided by the cluster “impersonal si - source ne” which in proclisis looks like Type 1 clusters in that it undergoes vowel change, (42a). As shown by (42c) with a full PP, [–argumental] si cannot occur in infinitivals (unaccusative verbs such as uscire are only compatible with this type of si). The properties of this combination are summarised in (43): (42) a.

Da quella situazione, se ne uscirà presto. from that situation imp from-there will-go-out soon

b. *Sembra/Ritengo essersene già usciti. [it] seems/[I] believe [to] have imp from-there already gone-out c. *Sembra/Ritengo essersi già usciti da quella situazione. [I] believe [to] have imp already gone-out from that situation (43) Type 4: okproclisis, *enclisis, okvowel change siimp nesource

The same restriction operates on clusters which look like Type 2, as in (13) and (44a)–(46a). These clusters are banned from the enclitic position, b. sentences. The independent restriction against [–argumental] impersonal si with infinitives is illustrated by the ungrammaticality of (44c) and (46c) with a strong pronoun and (45c) with a full DP:12 (44) a.

Mi/Ti/Vi si è invitato spesso, ultimamente. me/you:sg/you:pl imp has invited often, lately

b. *Ritengo essermisi/essertisi/esservisi invitato spesso, ultimamente. [I] believe [to] have me/you:sg/you:pl imp invited often, lately

. Past participle agreement with 2pl vi is optional: Vi si è invitati spesso, ultimamente is possibile alongside (44a) (see note 29). As discussed in note 11, 1pl ci cannot combine with impersonal si: *Ci si è invitato/invitati spesso, ultimamente. Differently from (45c), the following sentence is possible with past participle agreement: Ritengo essersi vendute poche automobili. In this case, si is [+argumental] and can occur in infinitival clauses, as in (39c,e) and (40c) (see Cinque 1988).

On different types of clitic clusters 

c. *Ritengo essersi invitato solo me/te/voi, ultimamente. [I] believe [to] have imp invited only me/you:sg/you:pl, lately (45) a.

Le si è vendute bene. them imp have sold well

b. *Ritengo esserlesi vendute ad un prezzo eccessivo. [I] believe [to] have them imp sold well

(Cinque 1988: 557)

c. *Ritengo essersi venduto poche automobili. [I] believe [to] have imp sold few cars

(Cinque 1988: 556)

(46) a.

Ci si va spesso, ultimamente. there imp goes often, lately

b. *Ritengo andarcisi spesso, ultimamente. [I] believe [to] go there imp often, lately c. *Ritengo andarsi spesso lì, ultimamente. [I] believe [to] go imp often there, lately

The same restriction applies to the sequence cirefl siimp, which necessarily contains [–argumental] si (see Cinque 1995: 197f for discussion and the examples in (47)):13 (47) a.

A Beirut ci si è sparati addosso tutta la notte. in Beirut refl imp has fired all night long

b. *A Beirut sembra essercisi sparati addosso tutta la notte. in Beirut [it] seems imp to have fired at each other all night long c. *Ritengo essercisi aiutati inutilmente. [I] believe imp to have helped each other in vain

Type 5 combinations, which contain [–argumental] impersonal si, are the following: (48)

Type 5: okproclisis, *enclisis, *vowel change a. mi/ti/viDO siimp b. loDO siimp c. ciloc siimp d. cirefl siimp

4.5 Summary The following table summarizes the different types of clusters which are found in Italian.14 While all of them occur in proclisis, some are banned from the enclitic

. Ci si is suppletive for *si si, as in Ci si è arrabbiati “refl imp has got-angry” (= One has got angry), and Ci si è aiutati “refl imp has helped” (= One has helped each other). . See Monachesi (1995) for other clusters which enter the types discussed here.

 Anna Cardinaletti

position. Another property which differentiates Italian clusters is vowel change on the first clitic: (49)

+ enclisis

– enclisis

+ vowel change – vowel change

Type 1 (e.g., me lo) Type 2 (e.g., mi ci)

Type 4 (e.g., se ne) Type 3 (e.g., mi si) Type 5 (e.g., lo si)

If types 4 and 5 are not taken into consideration (since they cannot occur in enclisis for independent reasons), a correlation arises between phonological and syntactic properties: (50)

+ enclisis

– enclisis

+ vowel change – vowel change

Type 1 (e.g., me lo) Type 2 (e.g., mi ci)

– Type 3 (e.g., mi si)

Clusters that exhibit vowel change are possible in both proclitic and enclitic position. The reverse is not true, however: some clusters with no vowel change can appear in enclisis, while others cannot.

5. Ingredients for the analysis 5.1 The representation of clitic clusters in antisymmetry In order to differentiate the clitic sequences which appear in an unconstrained manner from those which can only appear in the proclitic position, I propose that the two types instantiate different ways of adjunction to their host, namely the two possibilities made available by LCA (Kayne 1994). The first type of clitic clusters (Type 1, 2 and 4), instantiate the structure in (51), where the two clitic pronouns are adjoined one to the other and thus occur inside one and the same functional head (Kayne 1994: 20). The other type of clitic sequence (Type 3 and presumably Type 5), is represented as in (52a), where the two clitic pronouns are adjoined to two distinct (adjacent) functional heads (Kayne 1994: 21). As shown in (52b), there can also be sequences of three adjacent heads, as in Le ci si può mettere un po’ di zucchero “to-her there imp can put some sugar”:15

. Our conclusions differ from Terzi’s (1999), who denies the existence in Romance languages of the structure in (52) (see, however, Ordóñez 2002 for Spanish). As for Type 5 clusters, the comparison with Type 3 suggests that they also utilize structure (52), although the enclitic test is unavailable for independent reasons (section 4.4).

On different types of clitic clusters 

(51) adjunction of one clitic to the other XP X cl2

cl1 me/glie/se/ce mi ci

(52) a.

X cl2 lo/ne ci si

adjunction to distinct functional heads XP X

YP

cl1 X mi/ti/ci/vi/gli/le

Y

cl2 si

Y

b. adjunction to three distinct functional heads XP X cl1 le

YP X Y cl2 ci

ZP Y cl3 si

Z Z

The combination of the two possibilities in (51) and (52), hypothesized but not made explicit by Kayne (1994: 21), is found in other sequences containing three clitics (or more). In (52), the material adjoined to the higher functional head or the material adjoined to the lower functional head can be itself a clitic cluster. In Me lo/Glielo si dice spesso “to-me it/to-him it imp says often”, the cluster me lo/glielo is adjoined to the higher functional head as in (51), while impersonal si is adjoined

 Anna Cardinaletti

to the lower one (the cluster in Quando ce le si è comprate, … “when to-refl them imp has bought”, from Cinque 1995: 195, see note 13, is of the same type). In Mi/Gli ce ne vorranno due “to-me/him there of-them will-want two” (= I/He will need two of them), it is the reverse: mi/gli is adjoined to the higher head, while the cluster ce ne is adjoined to the lower functional head as in (51). An anonymous reviewer suggests the following example of a sequence with more than three clitics: Mi ci se ne mette molti “to-me there imp of-them puts many”. Here, mi ci forms a unit that adjoins to the higher head, while se ne adjoins to the lower head. In the next section, we discuss why only structure (51) gives rise to enclitic clusters. 5.2 Proclisis vs. enclisis Following many previous proposals (Kayne 1991, 1994; Belletti 1993, 1999; Rizzi 2000), I assume that enclisis arises via verb movement across the clitic (cluster), specifically via adjunction of the verb to the clitic (cluster) itself. In structure (51), the verb adjoins to the cl1 inside the cluster, and enclisis is produced, as in (53): (53)

XP X cl2 cl1 V metter

X

… VP metter

cl2 ci cl1 mi

In structure (52) there is no way for the verb to attach to the clitic pronouns. If it adjoins to the clitic in the higher head, it skips the lower clitic. If it adjoins to the lower clitic, it should move further pied-piping the clitic, but this is only possible in languages where the higher pronoun is not clitic but weak (for Greek and some Spanish varieties, see Ordóñez 2002: 216–223).16 The structure in (53) allows us to understand Benincà and Cinque’s (1993: 2325) claim that Italian has “one word” encliticization. Among other properties, their proposal relies on the observation that the verb and the enclitic(s) are written as

. Differently from Greek and Spanish varieties, Italian does not display different ordering possibilities for one and the same combination of clitics.

On different types of clitic clusters 

one single word (for orthographic conventions, see also section 5.5). This claim is to be interpreted in structural terms: in order for enclisis to be possible, the verb must adjoin to a cluster that is formed via structure (51). The verb and the clitic pronoun(s) end up being dominated by one and the same head, as shown in (53). Benincà & Cinque’s (1993) claim cannot be interpreted in phonological terms, however. The sequence “verb + enclitic pronoun(s)” is not a single phonological word. Phonological processes which apply word-internally do not apply in enclitic structures. The process of s-sonorization, for instance, which is found word-internally in intervocalic contexts (as in ca[z]a “house”), does not take place between the verb and an enclitic pronoun: *mettendo[z]i “put.ger refl”. The verb and the enclitic(s) are different phonological words. In Selkirk’s (1995) terminology, (en)clitic pronouns are neither “internal clitics” nor “affixal clitics”, but “free clitics”, which are adjoined to the same Phonological Phrase as the verb (see Cardinaletti & Repetti 2008 for subject enclitics).17 5.3 Vowel change Vowel change can be captured via the lowering rule in (54) (Cinque 1995: 194):18 (54) [a …i] → [a …e]/____ [nasal or liquid], where A is a clitic.

Since it does not apply whenever the phonological context is found, vowel change is not a purely phonological rule. For instance, it does not apply between a clitic

. S-sonorization does not apply in proclisis either: Lo [s]o/*[z]o “[I] it know”. . Vowel change is not some sort of vowel harmony. It applies in front of all 3rd person clitics, which display low (me la), middle (me lo, me le) and high vowels (me li). Desouvrey (2005: 63) differs from the standard analysis in considering me in e.g., (23b) to be the same as the strong pronoun me. A number of considerations speak against this analysis. Firstly, a strong DO pronoun never occurs between the subject and the verb, (ia) and (iia). Secondly, vowel change is also possible with 1pl ci and 2pl vi, whose strong counterparts are noi and voi, not ce and ve, (iib) vs. (iic): (i) a. *Gianni me vede. b. Gianni me sees (ii) a. *Gianni noi vede. b. Gianni us sees

Gianni vede me. Gianni sees me Gianni vede noi. c. *Gianni vede ce. Gianni sees us Gianni sees us

Desouvrey (2005: 78, n.30) notes the potential problem raised by glielo, where [e] is found which is not present in the corresponding strong form lui, but suggests that this “is due to the phonological harmonization of the third person cluster with clusters including first and second person pronouns”. The more traditional analysis in terms of /i/ to [e] change does not need to make any of these controversial assumptions.

 Anna Cardinaletti

pronoun and the verb, as in (55), or between a clitic determiner and the noun, as in (56) (Kaisse 1985):19 (55) Mi/*Me lava/nuoce. [he] me washes/harms

(56) i/*e laghi/nidi the lakes/nests

Vowel change is a morphophonological process which only applies cluster internally. As we have seen, it applies inside a particular type of clitic sequences: namely (some of) those corresponding to structure (51), where the two clitics are dominated by one and the same head. The rule in (54) can be made more precise as follows: (57) [a …i] → [a …e]/____ [coronal sonorant], where A is a clitic pronoun in structure (51).

Things are, however, more complex in clusters composed of three clitics. As noted by Kayne (2000: 154, n.10), some speakers also allow vowel change on the clitic which is not immediately followed by a clitic beginning with a coronal sonorant. See his example Me ce ne vorranno due “to-me there of-them will-want two” (= I will need two of them) and (58). For some speakers, the same holds for enclitics, as shown in (59):20 (58) a.

Me se ne sono presentate due. to-me refl of-them have introduced two ‘Two of them introduced themselves to me.’

b. A Mario, lo zucchero, nel caffè, non glie ce l’ho messo. to Mario, the sugar, in-the coffee, [I] not to-him there it have put (59) a.

Hanno deciso di presentarmesene due. have decided to introduce to-me refl of-them two

b. Ho deciso di non mettergliecelo. [I] have decided to not put to-him there it

The rule in (54) cannot, however, be simplified as in (60) because the same speakers do not accept vowel change in Type 2 clusters, e.g.,*Me ci metterà “[he] me there will-put”, *Ha deciso di mettermeci “[he] has decided to put me there”:21

. In many Italian dialects, both in the North and in the Center/South, clitic pronouns display final –e (Rohlfs 1968: 151ff), and the starred form in (55) is possible. In these dialects, [e] is the epenthetic vowel. For epenthetic [i] in Italian, see below in the text. . The enclitic clusters in (59) must have a representation similar to (51), in which the three clitics are dominated by one and the same head. . In the varieties mentioned in note 19, such as the Central variety I speak, both clitics would end in –e: Me ce metterà, Ha deciso de mette(r)mece.

On different types of clitic clusters 

(60) [a …i] → [a …e] where A is a clitic pronoun in structure (51).

Thus, it seems necessary to have a different analysis, which does not make reference to the phonological environment, but still captures the correlation observed in (50) above. Notice that Type 1 clusters display the same vowel which is found in the combinations of preposition and determiner such as in + il > nel “in the” or di + il > del “of the”, where the linking vowel is [e] and not the epenthetic vowel [i] which appears in clitic pronouns (mi, ti, etc.), prepositions (in, di), and determiners (il) when they occur in isolation (for epenthesis, see Vanelli 1992; Repetti 2003; Cardinaletti & Repetti 2008, and the references cited there). As we will see below, the syntactic properties of this type of cluster seem to suggest that they are merged as lexical units. These units consist of a consonantal clitic (m-, t-, s-, c-, v-, gl-, see Kayne 2000: 135; Cardinaletti & Shlonsky 2004: 534), the linking vowel [e], and the accusative (lo) or partitive/genitive (ne) clitic. We conclude that some clusters are lexical units, and that [e] appears instead of [i] only in this type of clusters. The absence of s-sonorization in cluster ci si, both in proclisis (Ci [s]i/*[z]i va “there imp goes”) and enclisis (metterci[s]i/*metterci[z]i “[to] put there refl”) confirms that although Type 2 clusters are a constituent in structure (51), they are not lexical units, hence they do not display the linking vowel [e].22 5.4 On the replacement of le by gli The above proposals also provide a way of understanding why le is replaced by gli when it combines with accusative and partitive/genitive clitics, as in (14) above. This replacement does not occur when le combines with other clitics, such as locative ci, reflexive si, and impersonal si, as shown in (61): (61) a.

Le ci vuole un’ora per prepararsi. to-her loc takes an hour to prepare-refl

b. Gianni le si è rivolto in inglese. [he] to-her refl has addressed in English c.

(see (38a)) (= (35a))

Le si affidò completamente. [he] to-her refl entrusted completely

. Rohlfs (1968: 168, n.1) makes the interesting observation that vowel change only obtains with clitics which derive from vowel-initial pronouns (ILLU > lo, INDE > ne) and had to have an initial –e in the earliest stages of Italian. In Old Italian, the form with [e] was not obligatory, but it is by far the most frequent form found in enclisis (Cardinaletti 2004b), where the two clitics and the verb form a constituent as in Modern Italian.

 Anna Cardinaletti

d. Non le si parlò con la dovuta attenzione. to-her imp spoke with the due attention

(= (39a))

e.

(= (45a))

Le si è vendute bene. them imp has sold well

The comparison between (14) and (61) suggests that in proclisis, the replacement of le by gli only occurs in clusters which are inserted as lexical units. In (61), le and ci and le and si instead occur as separate constituents, as shown by the fact that the clusters are impossible in enclisis (see (41b,e,g) and (48b)); as expected, s-sonorization does not apply: le *[z]i (see sections 5.2–3). Suppose that like 3rd person accusative clitics, le is bi-morphemic (l+e). Differently from word-final -i in gli and other clitics, -e in le cannot be deleted in front of vowel-initial verbs and auxiliaries: cf. Le/*L’apro la porta “[I] to-her open the door”, Le/*L’ha aperto la porta “[he] to-her has opened the door” vs. Gli/Gl’apro la porta, Gli/Gl’ha aperto la porta (an observation due to Franca Ferrari, p.c.; also see Garrapa 2008). Thus, while -i is an epenthetic vowel, -e is not; I suggest that it is a class marker.23 If this is correct, le is morphologically too complex to be the first element in single word clusters, which contain mono-morphemic consonantal clitics (see section 5.3). In (14), le is replaced by the other (less marked) dative clitic pronoun, namely gli. In enclisis, le is replaced by gli in one more case, namely when the combination le ci is ungrammatical (38b): A Maria vorrei metterglici un po’ di latte (see (32b)). The constraint on le is thus more general. Morphologically complex clitics cannot be the first constituent in clusters which are dominated by one and the same head (whether they are lexical units – *le lo → glielo, or not, *le ci → gli ci). These must contain mono-morphemic consonantal clitics.24 Interestingly, the replacement of le by gli does not occur when le is not adjacent to the clitic with which it usually forms a unit, as in A Maria, di zucchero, nel caffè, le ce ne metto/le se ne mette sempre troppo “to Mary of sugar in-the coffee, [I] to-her there of-it put/to-her imp of-it puts too much”, and A Maria, di questo, le se ne è parlato spesso “to Mary of this to-her imp of-it has spoken often”. In these

. In Italian, -e appears in the nominal declension which does not display gender distinctions (e.g., felice “happy.sg” vs. felici “happy.pl”) (see Harris 1995 for class marker -e in Spanish pronouns). This morpheme is probably also found on strong pronouns me and te (Kayne 2000: 131, 145). . Spanish 3rd person IO clitics le/les “to-him/her/them” are also morphologically complex. The ungrammaticality of *le(s) lo in (11) can be attributed to the same constraint that prevents Italian *le lo, *le ne in (14) and enclitic le ci in (38b). Spurious se is not complex: final -e is presumably an epenthetic vowel like Italian -i, and does not combine with plural -s (*ses lo, Harris & Halle 2005: 204).

On different types of clitic clusters 

cases, le and ce ne/se ne occur in independent functional heads in structure (52), and no morphological restriction applies. Notice that final -e in ne has properties similar to -e in dative le: Ne/*N’apro molti “[I] of-them open many”. In both ungrammatical clusters *lo ne/*ne lo “it of-it/them”, the linearly first clitic is morphologically too complex to enter the cluster.25 5.5 An aside on orthographic conventions Since the analyses which differentiate me lo from glielo (see section 3 above) utilize the observation that glielo is written as one word, while all other combinations of clitics are not (me lo, te lo, etc.), it is worth considering a few observations about orthography. I take the implications concerning orthography to go one way only. If two elements are written as one word, they are a single constituent. If two elements are not written together, they may be one constituent or different constituents. For sequences of Italian clitic pronouns, this means that enclitic clusters, which are written as one word with the verb, necessarily arise by structure (51), as shown in (53), while proclitic sequences may be one constituent in structure (51) (e.g., me lo, mi ci) or distinct constituents in structure (52) (e.g., mi si). The different orthography of (proclitic) me lo and glielo (two orthographic words vs. one) does not point to a different syntactic and morphophonological behaviour. As I have shown in section 4.1, the two sequences behave in the same way as far as syntax and morphophonology are concerned. It is also interesting to note that Italo Calvino used to write glielo as two orthographic words: Il fucile glie l’avevo procurato io “the gun to-him it had got I” (Il barone rampante, Mondadori 1993). The question is not why glielo is written as one word, but rather why the other sequences which behave like glielo are not written in the same way: i.e., melo, telo, etc. Interestingly, this is a typical mistake of young children when they start writing.26 Finally, a note on the spelling of glielo is in order. Differently from e.g., the mi vs. me lo case, it seems that in glielo, final [i] is not replaced by [e], but [e] seems to be added to [i]. This is illusory, however. Vowel change occurs as in the other

. A phonological clue allows Italian speakers to individuate consonantal clitics and morphologically complex clitics: the former consist of sounds different from the coronal sonorant sounds [l] and [n] that characterize the latter (Cardinaletti 2004a). See also 5.6. . Variable spelling is also found with compounds which can be written as single words (cassapanca), hyphenated words (partito-azienda) or two independent words (nave traghetto), although they are very similar or identical on the morphological level.

 Anna Cardinaletti

Type 1 clusters. The 〈i〉 is needed to graphically represent the palatal sound [ ]: [ e:lo]. If 〈i〉 were not present, the sequence would represent the consonantal cluster [gl] as in [gle:ba] “glebe”.27 5.6 On person and number feature checking In section 3, we have seen that a proposal based on the case properties of clitics, which seems adequate for French, runs into problems in a comparative perspective. If case were the only property which regulates the order inside clitic clusters, some Italian clusters should be impossible, contrary to fact. The discussion of the different types of clusters confirms this conclusion. In Type 1 and Type 2 clusters, either a DO or an IO person clitic can appear first when the linearly second clitic in the cluster is a genitive or locative pronoun, respectively, (29a–c,f) and (34a-e). In Type 3 clusters, either a DO or an IO clitic can appear first when the linearly second clitic in the cluster is a reflexive pronoun, (41a,c). It thus seems that other properties matter for the construction of clusters. Kayne (2000) suggests that clitic pronouns can be distinguished into the two classes of person and non-person clitics: the former contains 1st and 2nd person clitics and si, while the latter contains 3rd person clitics. There is a long debate as to whether 3rd person pronouns are also marked for person or not. For instance, Bianchi (2006a) and Nevins (2007) have recently argued that they do. Adger and Harbour (2007: 20f, 25) and Anagnostopoulou (2005: 211) instead take 3rd person IO clitics to be marked with a person feature, albeit negative [–person], while 3rd person DOs lack a person feature altogether. I share this view and suggest the following classification for Italian: (62) a. Person clitics: b. Non-person/Number clitics:

mi, ti, si, ci, vi, gli, ledat lo, la, li, leacc, ne

I hypothesize that person and number are mutually exclusive features and person and number clitics belong to two different paradigms. Apart from dative le, person clitics are mono-morphemic consonantal clitics, while number clitics are morphologically complex (see sections 5.3–4 and note 25). 3rd person accusative clitics and partitive/genitive ne are not taken to contain any person feature (see also Harley & Ritter 2002). They are marked for the φ-features number and gender only, and share the property of triggering past participle agreement obligatorily (see note 29 and Belletti 2006; Cardinaletti & Giusti 2006: 68, 77). In partitive ne, these features are left unspecified (see note 38). . It is thus not true that “ie is spelled as e”, as Desouvrey (2005: 78, n. 30) claims.

On different types of clitic clusters 

As pointed out by Kayne (2000: 140), lack of number (and gender) distinctions is typical of person clitics. Since Benveniste (1966), it is common to observe that 1pl is not semantically the plural of 1sg, nor is 2pl the plural of 2sg. Reflexive si never displays number distinctions (Kayne 2000: 145f), and the same is true of impersonal si and locative ci. Since Italian gli and le do not have a plural clitic counterpart (3pl dative loro is not a clitic pronoun, note 5), they are not marked [singular] and can be analyzed as person clitics. Furthermore, in (colloquial) Italian and many Italian dialects gli does not express any number (or gender) distinctions: a sentence like Gianni gli dà un libro can mean “Gianni to-him/her/them gives a book”. The same is true of le in other varieties (e.g., in some Veneto varieties).28 The above classification holds for both the pronominal and reflexive usage of 1st and 2nd person clitics, which always behave in the same way and must be considered as one and the same lexical item (see note 9). Since, according to the classification in (8a), 3rd person IO clitis (e.g., Italian gli and le) are case-marked clitics (dative), person and case features do not exclude each other. Neither do number and case in non-person clitics. As suggested by Bianchi (2006a: 2049), person features are encoded in clausal head positions, and clitic pronouns move to check their person features against these heads. She formulates a Person licensing requirement: “A personal argument of the verb must license its person feature in the functional structure of the clause” (see also Rezac 2005). In Bianchi’s (2006a: 2059) analysis, the objects licensing area is located between VP and TP (while the subjects licensing area is between FINP and TP). I suggest instead that both areas result from splitting the INFL projection, which is traditionally taken to host clitic pronouns. As shown by Northern Italian dialects which have both subject and object clitics, subject clitic heads are higher than object clitic heads (and are often separated by the clitic negation). Since non-person clitics do not contain person features, but still appear in the high clitic field together with person clitics, they must adjoin to head position(s) different from the person heads. Since non-person clitics express number distinctions, I take the head to which they cliticize to be a Number head. See Shlonsky (1989), Taraldsen (1995), Sigurðsson (2004), Bianchi (2006b) among others for the proposal that person and number features are encoded in separate functional heads.

. By contrasting with 3pl clitic leur, French lui is marked [singular] and cannot be a person clitic. The same can be said of Spanish singular IO le vs. plural IO les. As expected, feature specification of single lexical items can be different across languages. This is also true of French locative y: differently from Italian ci, it is not a person clitic.

 Anna Cardinaletti

A reasonable hypothesis is that the heads against which clitic pronouns check person and number features are criterial in the sense of Rizzi (2006). If criterial heads have freezing effects (ibid.), we also understand why clitic movement is not long distance. In section 6 we will see how person and number heads are hierarchically organised; this hierarchy accounts for some aspects of the word order found inside clusters.29 5.7 Clitic climbing and two clitic positions inside the clause In restructuring contexts, the high clitic position discussed so far is targeted in clitic climbing. Following Cinque (2004), restructuring implies a monoclausal structure where the higher verb (a modal, aspectual, causative, perception, or motion verb) is merged in a functional head associated with the infinitival lexical verb. Clitic climbing is thus only apparently long distance: it takes place inside one and the same clause. The question naturally arises as to where clitics occur when clitic climbing does not apply. Sentences like (63a) show that it is necessary to posit a low clitic position, different from the high clitic position usually identified with INFL and discussed in the previous section (Cardinaletti & Shlonsky 2004). Auxiliary switch, a hallmark of restructuring, applies not only when clitic climbing has applied, as in (63b), but also when the clitic appears on the infinitive, (63a) (Rizzi 1976: 48, n. 18; Cinque 2004):30

. Non-person/Number clitics also express gender distinctions. Number features are encoded differently from gender features: while the former are part of the syntactic (functional) structure associated with the (lexical) noun, the latter are part of the lexical entry of the noun (Ritter 1991, 1993, 1995; De Vincenzi & Di Domenico 1999). This proposal has consequences for other categories: only number features are part of the syntactic (functional) structure associated with past participles and pronouns, while gender is parasitic on number (Di Domenico 1997: 131f). Cardinaletti & Chinellato (2005) suggest the following restrictive theory of number, with the consequence that past participle agreement is number agreement only: masculine singular: no number features (default -o as morphological epenthesis – Ferrari 2005; Cardinaletti and Repetti 2008; or as a word marker – Kayne 2000: 140); feminine singular: [–number]/parasitic gender features (-a) (Chinellato 2004); masculine plural: [+number] (-i); feminine plural: [+number]/ parasitic gender features (-e). Since person clitics are not specified for number, past participle agreement must be so-called semantic agreement. It differs from number agreement in that it is optional (see notes 11 and 12); notice that lack of agreement is the preferred option for many speakers. . In non-restructuring cases like Avrei voluto [esserci già andato] “[I] would-have liked bethere already gone”, ci occurs in the high clitic position of the embedded clause.

(63) a.

On different types of clitic clusters 

Sarei voluto andarci. [I] would-be wanted [to] go-there

b. Ci sarei voluto andare.

The low clitic position is located in the lexical domain (see also Manzini & Savoia 2005). Since person and number heads, like the other criterial heads discussed by Rizzi (2006), only occur in the high portion of the clause, clitics must move to the low clitic position to check other features. Following many previous works (Laenzlinger 1993; Belletti 1993, 1999; Cardinaletti & Starke 1999, among others), I suggest that the first step of clitic movement is motivated by case checking. This proposal finds some support in data from Bantu languages. Differently from Romance proclitics, which appear higher than tense-inflected verbs, Bantu object clitics follow tense markers. As in Romance, Bantu tense markers can be taken to occur in INFL (Barrett-Keach 1986; Krifka 1995: 1412–4).31 It is tempting to hypothesize that Bantu object clitics occur structurally lower than their Italian proclitic counterparts. Whether or not their position is the same as the Italian low clitic position needs further research which I am not able to address here. As in Italian, however, this position must be located somewhere above the VP, above some of the aspectual heads associated with the lexical verb (Cinque 2004: 178, n.47) and iterated for some of these heads (quasi-functional restructuring verbs, Cardinaletti & Shlonsky 2004). As shown by the following Kinyarwanda example from Dryer (1983: 130), the verb which follows the two object clitics is inflected for aspect and has presumably raised to some functional head above VP:32 (64) Yohaâni y-a-yi-mw-oher-er-eje. John he-past-it-her-send-ben-asp ‘John sent it to her.’

. See Cocchi (2000a: 101) and (2003: 11) for the different view that in Bantu, tense and aspect markers lexicalize C. . An anonymous reviewer suggests that two clitic positions should be hypothesized for Bantu as well. Object clitics can also be postverbal, as shown by the following Tshiluba example for the DO clitic tshio: n-aka-mu-pa-tshio “1.su-T-1.IO-give-13.DO” (= I have given it to him) (from Willems 1949, pointed out by the reviewer). An account of Bantu in terms of more than one clitic field (in the framework of Manzini & Savoia 2001, 2005) is indeed proposed by Cocchi (2000a,b), (2003). Notice, however, that the postverbal DO clitic tshio differs from the preverbal one: tshi in (5c). This difference could be used to argue that the pronoun has a different status in the two positions: preverbal tshi is a true clitic, while postverbal tshio is a weak or strong pronoun (for weak pronouns in Swahili and the contrast between clitic wa, weak o and tonic wao “they’’, see Cocchi 2003: 3–4). If this is correct, the analysis proposed in the text can be kept as it is.

 Anna Cardinaletti

Differently from Italian, where clitics can appear both in the high and the low clitic position, Bantu clitic pronouns occupy the low clitic position where person features are not checked and clusters generally appear not to be sensitive to person features (see (5) and (6) above).33 In Italian, the low clitic position only hosts enclitics. The clusters which can appear in the low clitic position are the same as those which appear in enclisis in the high clitic position (those which I have called Type 1 and Type 2 in section 4). 5.8 On the clitic (cluster) derivation Restructuring contexts make visible a relationship between the high and the low clitic positions. In other words, clitics undergo a two step derivation: they move from their argument positions to the low clitic position in order to check case and, in doing so, reverse the order of the full arguments they pronominalize (Laenzlinger 1993: 264f). As we will see, this effect is very clear in Bantu and masked in Italian Type 1 clusters. From the low position, pronouns move to the clausal clitic position to check person or number features. In Italian, the low clitic position only hosts enclitics. In order to get enclisis, clitic clusters must form a constituent so that the infinitive verb can adjoin to it, as in structure (53). As we have seen for the high clitic position, there are two types of clusters which do so: those with vowel change (Type 1), which are inserted as lexical units (section 5.3), and those without vowel change (Type 2), which arise by adjunction of one clitic to the other inside one and the same head. If clitics have formed a constituent in the low clitic position, they move as a unit to the clausal clitic position, where they can appear both in proclisis and in enclisis (depending on whether the verb is finite, imperative, or infinitival, see (23) and note 8). In Type 3 clusters, the clitics move independently to the high clitic position and end up in distinct (person and number) heads in configuration (52). In this case, only proclisis is possible, and enclisis is disallowed. Notice that if in restructuring contexts, no overt climbing takes place and clitic pronouns are spelled out in the low clitic position, they can be taken to covertly raise to check their person and number features in the high person field (or to agree with these heads long distance). In what follows we see how the proposed system derives the different types of clitic clusters individuated above.

. Some cases where they are are discussed in Cocchi (2000a: 114, n. 29) and Hyman & Duranti (1982: 231ff).

On different types of clitic clusters 

6. On the derivation of the different types of clusters 6.1 Type 1 clusters with IO – DO clitics As we have proposed in section 5.3, Type 1 clusters in (29a–c), which display the IO – DO order, are merged as lexical units. I suggest that this order is not due to the order of merging, but is lexically determined (for the ban against morphologically complex clitics in first position, see section 5.4). Evidence from French (2b) and Bantu (5), where no such lexical units as in Italian exist, suggests that DO > IO is the merging order of object clitics. The French cluster le lui, in which neither pronoun has a person feature (see note 28), reflects the merging order in a pure way.34 The order inside the clitic cluster is the opposite of the argument order, which is IO > DO (see Anagnostopoulou 2005: 211; Bianchi 2006a: 2037, among many others). The argument closer to the functional head (IO) is attracted first, while the DO is moved second and adjoined to IO in structure (51). This is also very clear in Bantu, where clitic pronouns occupy the low clitic position where person features are not checked and clusters are generally not sensitive to person features (see (5) and (6) above). In Bantu, the order of arguments is IO > DO, as in (65) (Xhosa, Bearth 2003: 127), and is reversed in clusters, which display DO > IO (Dryer 1983: 132; Bresnan & Moshi 1990; Krifka 1995: 1407; Bearth 2003: 126f, among others):35 (65) Ndì-n íkà úmfâzì úmntwáná. I-am-giving woman child ‘I am giving a child to the woman.’

IO – DO

The difference between Italian and Haya in (4) and (5a), repeated here as (66), is particularly telling. Since the two languages manifest both word orders in the case of full XPs, as shown in (67), the difference in (66) is surprising: (66) a. b. c. (67) a.

Gliele darà. *Le gli darà. A-ka-bi-mú-h-a. he-P3-them-him-give

IO – DO DO – IO

Darà banane al bambino. DO – IO he.will.give bananas to-the child

. For a similar proposal framed in his theory of silent clitics, see Kayne (2006: 7). . If it turns out that the merging order of arguments is the opposite (DO > IO), the generalization will be that clusters maintain this order (as in Bianchi’s 2006a analysis, but with different assumptions); the same observation holds for the respective order of DO and loc, discussed in section 6.2.

 Anna Cardinaletti

b. Darà al bambino banane. c.

IO – DO

A-ka-h’ èbitook’ ómwáana. he-P3-give bananas child ‘He gave bananas to the child.’

d. A-ka-h’ ómwáán’ èbitooke. ‘He gave the child bananas.’

DO – IO (Hyman & Duranti 1982: 220)

IO – DO (Hyman & Duranti 1982: 218)

Notice that the order manifested by Bantu clitic pronouns (DO – IO) is more frequent in OV languages (like Latin or German), but Bantu languages are VO (Bearth 2003; Kinyalolo 2003: 345, and the references quoted there). I conclude that, as clearly shown in Bantu, DO > IO corresponds to the merging order of clitic pronouns. No further requirement is operative in Bantu since object clitics do not reach the high clitic position.36 The particular order seen in Italian Type 1 clusters (IO – DO) complies with the person field in the high portion of the clause, where the cluster checks person and number features in the order in (68). A clitic cluster like me lo first checks the number feature of lo (the higher label in structure (51)) against the Number head, and then further moves to check the person feature of the adjoined 1st person clitic against the 1st person head; a similar derivation holds for clusters containing 2nd (te lo) and 3rd person clitics (glielo) against the relevant person heads, (68a). By +Person in (68a), I mean person heads marked for both person features, which, following Nevins (2007) among others, I take to be [±Participant] and [±Author]. 1pl ci, which is also a locative clitic (see n. 9), and 2pl vi are probably not specified for one of the person features: [+Participant, 0Author] and [0Participant, –Author], respectively (see Nevins 2007 for the possibility that pronouns have no specification for the person features, where the 0 value is compatible with both + and – values). I express this by saying that ci and vi target a –Person head, as in (68b): (68) a.

+Person 1st > 2nd > 3rd me lo/te lo/ glielo

b. –Person ve lo/ce lo

+Number me lo / te lo / glielo +Number ve lo / ce lo

. In Bantu, subject clitics precede the tense marker, (5), and must occupy the high clitic position(s). Subject clitics are not found in the low clitic position (Cardinaletti and Shlonsky 2004). For verb movement across a subject clitic, as in ba-ntu b-aka-tuma-ye “cl.2-men 2.relT/A-send-1.su”(= The men that he has sent), see Cocchi (2003: 10).

On different types of clitic clusters 

In the high clitic position, these clusters are also possible in enclisis. Since they involve one single functional head, the verb can adjoin to them providing enclisis, as in (53). The data concerning mi ti combinations seen in section 3.2 confirm this approach to the order inside clitic clusters. Mi ti sequences are possible in enclisis when they display the order DO – IO, i.e., the order of adjunction of the two clitics to the low clitic position. The two clitics occur in one and the same head, and enclisis is possible. The order inside the constituent complies with the person hierarchy in the high clitic position where the 1st person head is higher than the 2nd person head (see (68a)). The IO – DO mi – ti sequences only found in proclisis instead require a change in the order of clitics which is not motivated in the low clitic position, but can arise in the high clitic position via the need of person feature checking: the IO mi ends up before the DO ti because in Italian, the 1st person head is higher than the 2nd person head (see (68a)). The two clitics occur in different, adjacent heads, and are not possible in enclisis in the high clitic position either. 6.2 Type 1 and Type 2 clusters with locative ci Type 1 and Type 2 clusters containing locative ci are both possible in the low enclitic position, but the order is different: LOC > DO in ce lo, DO > LOC in mi/ti/ vi ci (see (1), (29e), (34a,b)). I take the two clusters to be minimally different: one must be the order in which the two clitics are merged to check their case features; the other must be derived in some minimally different way. Evidence from French les y in (9a) indicates that the order of adjunction is DO > LOC, which reverses the argument order LOC > DO. I take the cluster mi/ti/ vi ci to be the one formed in the syntax: the argument that is closer to the functional head (LOC) is attracted first, while the DO is moved second and adjoined to LOC. The infinitival verb adjoins to the cluster in configuration (53), and an enclitic cluster is obtained. A similar analysis holds for Type 2 clusters mi/ti/vi/gli ci in (34c-e), where the order of adjunction is IO > LOC, which reverses the argument order LOC > IO. Since they target one single head, these clusters are possible in enclisis, (30b)–(32b). In the high clitic field, these clusters first target the –Person head to check the feature of ci (see (68b)), and then move to the +Person head to check the person feature of mi, etc. (for the observation that 1pl and locative ci target the same position, see also Bianchi 2006a: 2039, n.34). The cluster vi ci shows that –Person heads are also ordered, (69b): (69) a.

+Person 1st > 2nd > 3rd mi ci/ti ci/gli ci

–Person mi / ti / gli ci

 Anna Cardinaletti

b. –Person vi ci

–Person vi ci

As we have seen for the clusters in (29a–c), Type 1 cluster ce lo is merged as a lexical unit. If this analysis is correct, the word order inside the cluster is not telling on the order of merging of the two clitic pronouns. The cluster-internal order reflects the order of person and number heads in the high portion of the clause. The cluster first checks the number feature of lo and then targets the –Person head to check the feature of ci, as in (70) (see (68b)):37 (70) –Person ce lo

+Number ce lo

Being merged in one single head, the cluster is possible in enclisis. The fact that in Thsiluba, a locative clitic precedes a DO clitic (muana u-mutshi-di-a “boy 1.su-18.loc-7.do-eat-I” = The boy eats it there), apparently reversing the argument order DO > LOC (muana u-di-a tshimuma mu nzubu “boy 1.sueat-I fruit in house” = The boy eats fruit at home) (Cocchi 2000b: 50), seems to be problematic for my proposal. Bantu locatives, however, behave similarly to objects with respect to many phenomena (Bresnan & Moshi 1990; Cocchi 2000b: 44, and references cited there), and in some Bantu languages, locative arguments come in the order loc > (IO >) DO (Umwáana y-a-taa-yé-mo amáazi igitabo “child he-past-throw-asp-in water book” = The child has thrown the book into the water, Kinyarwanda, Kimenyi 1976, quoted in Dryer 1983: 134). This could be an intermediate step in Tshiluba cliticization, and the order displayed by Tshiluba clitic clusters is indeed compatible with my proposal above. 6.3 Another Type 2 cluster with locative ci As for the Type 2 cluster ciloc sirefl/DO in (34f), the two clitics adjoin to the low clitic position starting from the following order of arguments: reflexive si – locative argument ci. That reflexive si is higher than the locative argument is compatible with Manzini & Savoia’s (2001: 237) proposal according to which reflexive si is linked to the external argument position (and makes the clause become unaccusative, Kayne 2000: 144). The cluster moves to the clausal position to check the features of reflexive si against an unspecified person head (Grimshaw 1997) or 0-person head (Kayne 1993: 16), (whose person features are both unspecified, . Like lui, French y is not a person clitic (see note 28) and need not occur as high in the clitic cluster as its Italian counterpart (see (9a) above).

On different types of clitic clusters 

[0Participant, 0Author], like impersonal si in Nevins 2007), and then targets the –Person head seen in (68)–(70) above: (71) –Person ciloc sirefl

0Person ciloc sirefl

6.4 Clusters with impersonal si Consider now those clusters that cannot occur in enclitic position. One type contains [–argumental] impersonal si, which gives rise to what I have called Type 4 and Type 5 clitic clusters. In Cinque’s (1988) proposal, [–argumental] si is merged in INFL and provides INFL with the relevant features to get an impersonal reading. Translating this into our proposal, [–argumental] si is merged in one of the positions of the high clitic field. If this is true, it follows that [–argumental] si never gives rise to enclitic clusters in the low clitic position (see also Cinque 2004: 176, n. 42). This is the simplest case to explain away. Impersonal si gives rise to proclitic clusters like lo si, discussed in (13) and (45) above (see (48b)). It must target a position lower than the Number projection which hosts 3rd person accusative clitics, presumably the same head as reflexive si in (71): (72) +Number 0Person lo siimp

If impersonal si targets such a low position in the string of clitic heads, it is expected that it follows all person clitics (both +Person and –Person): see mi/ti/vi siimp in (48a) and cirefl siimp in (48c,d). I suggest that the same position is targeted by [+argumental] impersonal si. We expect that it also follows person clitics: see mi/ti/ci / vi / gli / leIO siimp in (41f,g) and ciloc siimp in (41h): (73) +Person –Person +Number 0Person 1st > 2nd > 3rd mi ti gli/le vi > ci lo siimp

Differently from [–argumental] si, [+argumental] impersonal si can occur in the low clitic position by itself, but in this position it does not give rise to clusters like the proclitic ones in (41f–h). It can be suggested that si is merged in a position higher than the low clitic position, which, following Cinque (2004: 178, n.47), can be identified with an impersonal Voice higher than personal Voice. It follows that si cannot give rise to any enclitic clusters, unless they are merged as lexical units, such as se ne in (29h). What is the location of partitive/genitive/source ne, which follows all proclitics (for reflexive and impersonal si see (29g,h) and (43))? French (9) provides

 Anna Cardinaletti

evidence that en occupies the lowest position in the clitic string. The same can be said of Italian ne. Since ne is not specified for number and gender [0Number, 0Gender], we might suppose that it targets a 0Number head, which must be taken to be the lowest in the string of clitic heads.38 Cluster se ne in (29h) moves to the high clitic position to check the number feature of ne and the person feature of si, in this order (a similar derivation holds for the other clusters containing ne in (29)): (74) 0Person se ne 0Number se ne

6.5 Type 3 clusters with reflexive si Another set of clusters which cannot occur in enclitic position are Type 3 clusters containing reflexive si. What prevents the clusters in (41a–d) from occurring in the low clitic position? What is wrong in the clusters (41a, b, d) is the fact that an io and a loc precede a DO (reflexive si being also linked to the internal argument position, Manzini & Savoia 2001: 238). These orders cannot be obtained in the low clitic position. I must admit that (41c) is unexpectedly wrong. The observation is, however, of some relevance, in that these clusters are rare and for many speakers they are also marginal in proclisis, as pointed out by the two reviewers. An independent restriction must account for their properties. According to our proposal in 5.2 above, clitics which are not possible in enclitic clusters occur in proclisis in distinct functional heads in configuration (52). Type 3 clusters involve combinations of the person heads seen above, in the order in (75):39 (75) +Person 1st > 2nd > 3rd mi ti gli/le

–Person

0Person

vi > ci

si

The orders in (41a,c) are obtained if the heads containing 1st and 2nd person pronouns are higher than the one containing reflexive si, which targets the same position as impersonal si. 3rd person io clitics are also higher than reflexive si (see (41b)). The cases in (41e) and (41d) confirm that 3rd person io clitics are higher than locative ci (see (69)) and locative ci is higher than reflexive si (see (71)). . Partitive ne inherits number (and gender) features from Q and triggers past participle agreement (Cardinaletti & Giusti 2006: 68, and note 29). For some speakers, agreement with partitive ne is however impossible. . Also in Spanish varieties, clusters me se enter structure (52) (Ordóñez 2002: 215).

On different types of clitic clusters 

A final remark concerns Type 1 cluster se lo in (29d), where IO reflexive si must target a position higher than the Number head where accusative clitics check their number feature. I take it to move to the –Person head which hosts 1pl and locative ci, (68b) and (69)–(71). IO si is similar to IO gli in that it has a 3rd Person feature [–Participant] (see section 5.6), but it is unspecified for the other feature [0Author]. This might also explain why ci is the form that replaces reflexive si when it combines with impersonal si (see note 13). The cluster in (41d) shows that ci precedes si. For Italian, we end up with the following field of hierarchically ordered functional heads in the high portion of the IP layer, located between the subject position and the position reached by the inflected verb: (76) +Person –Person +Number 1st > 2nd > 3rd mi ti gli/le vi > ci > siIO lo

0Person

0Number

si

ne

In conclusion, the way in which clitic pronouns adjoin to the functional heads in the high portion of the clause depends on their feature specification, namely person and number features. Whether they end up as a constituent on a single head, as in (51), or adjoin to adjacent heads (two or more, see (52)) depends on their derivational history: the former case arises if the clitics cluster in the low clitic position; the latter when this does not happen, and the clitics move independently to the clausal clitic field. The following feature specification of Italian person clitics is tentatively proposed: i. +Person (both features specified): 1sg mi [+Participant, +Author], 2sg ti [+Participant, –Author], 3rd person gli, le [–Participant, –Author, (gender)]; ii. –Person (one feature specified): 2pl vi [0Participant, –Author], 1pl ci [+Participant, 0Author], IO reflexive si [–Participant, 0Author]; iii. 0Person (no feature specified): impersonal and DO reflexive si [0Participant, 0Author]. (The combinations [–Participant, +Author], [0Participant, +Author] are excluded.) 7. On the mi gli constraint In Italian and other Romance languages, a 1st or 2nd person do clitic cannot co-occur with a 3rd person io clitic (the mi gli constraint or Person Case Constraint, PCC, Bonet 1991), as shown in (77a) for Italian. Data are however different

 Anna Cardinaletti

in other languages, such as Old Italian and Bantu, as shown in (77b) and (78), respectively: (77) a. b.

*Ti gli/*Gli ti sei data. e dì come gli ti se’ tutta data … and say how [you] to-him yourself are all given ‘And say how you gave yourself all to him.’

(78) A-ka-mu-ku-léét-el-a. he-P3-him-you-bring-app ‘He brought you to him.’

(Dante, Fiore; 173,2)

(Haya; Hyman & Duranti 1982: 231)

Several syntactic analyses have recently been given for the ungrammaticality of (77a) in languages like Italian (Anagnostopoulou 2005; Rezac 2005; Bianchi 2006a; Nevins 2007). On the basis of the assumption that the verb itself contains the φ-features person and number to check and IOs are always specified for person, Anagnostopoulou (2005) suggests that the person feature on the verbal head is checked by the IO; the DO clitic could only check the number feature on the verbal head, but it does not match it since it contains a person feature; hence the ungrammaticality of the sequence. We cannot adopt this analysis. Surprisingly, no violation arises in Italian with reflexive si, a pronoun specified for [+person] under Anagnostopoulou’s (2005: 211) assumptions: see Mi/Gli si è rivolto in inglese in (35a) above. These sequences should be ungrammatical, as they are in French (*Elle se lui est donnée entièrement “she refl to-him is given entirely”, quoted in Anagnostopoulou 2005: 204), but they are not. A different approach should therefore be assumed which takes into account the order of DO and IO inside the cluster. Notice that in Old Italian and Bantu, the 3rd person IO clitic precedes the 1st or 2nd person do clitic. Additionally, in the possible Italian cluster gli si, the 3rd person io clitic comes first (differently from French).40 Neither can the data be understood by adopting Bianchi’s (2006a) account of the mi gli constraint. Under the hypothesis that IOs are merged higher than DOs,

. As for (78), remember that Haya allows IO > DO alongside DO > IO (see (5a), (6a) and (10)). A reviewer finds a contrast between *ti gli and ??gli ti, which recalls the order in (77b). That the clitic order matters for the mi gli constraint is confirmed by the milder effect found in Czech, according to Rezac’s (2005: 125) judgment: Ukážu mu ho/??tĕ zítra “[I] show to-him him/you tomorrow”. In Czech, the two clitics occur in the order seen in (77b) and (78). See, however, Spanish Ella se le entregó en cuerpo y alma “she herself to-him gave in body and soul” (Rivero 2004: 498) and Haya *A-ka-ku-mu-léét-el-a “he P3-you-him-bring-app” (Hyman & Duranti 1982: 231).

On different types of clitic clusters 

Bianchi takes a 1st or 2nd person DO clitic moved higher than a 3rd person IO one, as in mi gli, to give rise to a nested paths configuration. According to her assumptions, nested paths are banned. If, however, the order of merging is DO – IO, as I have proposed in section 6.1, this analysis cannot be adopted either. This issue will not be pursued further here. Here it suffices to conclude that the picture is complicated by those languages (e.g., Old Italian and Bantu), in which 1st and 2nd person clitics do not need to appear first. 8. Conclusion We conclude with a new view of the serialization of clitic pronouns in clitic clusters. Differently from descriptive grammars and templatic approaches (e.g., Calabrese 2001; Perlmutter 1971), which usually report the surface serialization of clitic pronouns, I have shown that the superficial orders of clitic pronouns can be partially different from the order of merging (which is LOC > IO > DO in Italian/Romance and Bantu), as is shown by some of the clusters possible in the low clitic position. Other requirements sensitive to the feature make-up of clitic pronouns may vary this order since in Romance, clitics must move to the high clitic position(s) to check person and number features. In the high portion of the clause, there can be many (adjacent) heads to host clitic pronouns. The high clitic position must be thought of as an articulated clitic field. In proclitic clusters, clitics can appear in these adjacent heads or inside one and the same head. In enclitic clusters, clitics can only be dominated by one and the same head. This is also true of the low clitic position, where only enclisis is found in Italian. More cluster possibilities are thus possible in proclisis than enclisis in this language. In conclusion, the properties of clitic clusters can illuminate both the order in which arguments are merged in the lower portion of the clause and the order of functional projections in the higher portion of clause structure.

References Adger, D. & Harbour, D. 2007. The syntax and syncretisms of the person case constraint. Syntax 10(1): 2–37. Anagnostopoulou, E. 2005. Strong and weak person restrictions. In Heggie & Ordóñez (Eds), 199–235. Bearth, T. 2003. Syntax. In The Bantu Languages, D. Nurse & G. Philippson (Eds), 121–142. London: Routledge. Barrett-Keach, C.N. 1986. Word-internal evidence from Swahili for Aux/Infl. Linguistic Inquiry 17(3): 559–564.

 Anna Cardinaletti Belletti, A. 1993. Case checking and clitic placement. Three issues on Italian/Romance clitics. In Clitics: Their Origin, Status ands Position, H. van Riemsdijk & L. Hellan (Eds), EUROTYP Working Papers 6: 1–18. Belletti, A. 1999. Italian/Romance clitics: Structure and derivation. In van Riemsdijk (Ed.), 543–579. Belletti, A. 2006. (Past) participle agreement. In Everaert & van Riemsdijk (Eds), Vol. III, 491–521. Benincà, P. & Cinque, G. 1993. Su alcune differenze tra enclisi e proclisi. In Omaggio a Gianfranco Folena, 2313–2326. Padova: Editoriale Programma. Benveniste, E. 1966. Problèmes de linguistique générale. Paris: Gallimard. Bianchi, V. 2006a. On the syntax of personal arguments. Lingua 116: 2023–2067. Bianchi, V. 2006b. Number agreement and event pluralization: A case study. In Phases of Interpretation, M. Frascarelli (Ed.), 213–235. Berlin: Mouton de Gruyter. Bonet, E. 1991. Morphology after Syntax. Pronominal clitics in Romance. PhD dissertation, MIT. Bresnan, J. & Moshi, L. 1990. Objects asymmetries in comparative Bantu syntax. Linguistic Inquiry 21(2): 147–185. Calabrese, A. 2001. I pronomi personali. In Grande grammatica italiana di consultazione, L. Renzi, G. Salvi & A. Cardinaletti (Eds), Vol. 1, 563–606. Bologna: il Mulino. Cardinaletti, A. 1991. On pronoun movement: The Italian dative loro. Probus 3: 127–153. Cardinaletti, A. 1994. On the internal structure of pronominal DPs. The Linguistic Review 11: 195–219. Cardinaletti, A. 2004a. L’italiano contemporaneo: Cambiamento in atto e competenza dei parlanti. In Intorno all’italiano contemporaneo, A. Cardinaletti & F. Frasnedi (Eds), 49–75. Milano: FrancoAngeli. Cardinaletti, A. 2004b. Il pronome personale obliquo. To appear in Grammatica dell’italiano antico, L. Renzi & G. Salvi (Eds). Bologna: il Mulino. Cardinaletti, A. & Chinellato, P. 2005. Clitic pronouns and verbal agreement in two deaf adults, Poster presented at GALA 2005, Siena. Cardinaletti, A. & Giusti, G. 2006. The syntax of quantified phrases and quantitative clitics. In Everaert & van Riemsdijk (Eds), Vol. V, 23–93. Cardinaletti, A. & Repetti, L. 2008. The phonology and syntax of preverbal and postverbal subject clitics in Northern Italian dialects. To appear in Linguistic Inquiry 39(4). Cardinaletti, A. & Shlonsky, U. 2004. Clitic positions and restructuring in Italian. Linguistic Inquiry 35(4): 519–557. Cardinaletti, A. & Starke, M. 1999. The typology of structural deficiency: A case study of the three classes of pronouns. In van Riemsdijk (Ed.), 145–233. Chinellato, P. 2004. Disturbi di sintassi nell’afasia non fluente: Un’analisi dell’agrammatismo italiano e dialettale. Ph.D. dissertation, University of Padua. Cinque, G. 1988. On si constructions and the theory of pro. Linguistic Inquiry 19(4): 521–581. Cinque, G. 1995. Italian Syntax and Universal Grammar. Cambridge: CUP. Cinque, G. 2004. “Restructuring” and functional structure. In Structures and Beyond. The Cartography of Syntactic Structures, Vol. 3, A. Belletti (Ed.), 132–191. Oxford: OUP. Cocchi, G. 2000a. Free clitics and bound affixes: Towards a unitary analysis. In Clitics in Phonology, Morphology and Syntax, B. Gerlach & J. Grijzenhout (Eds), 85–119. Amsterdam: John Benjamins. Cocchi, G. 2000b. Locative constructions in Bantu. Quaderni del Dipartimento di Linguistica – Università di Firenze 10: 43–54.

On different types of clitic clusters 

Cocchi, G. 2003. Copulas in Bantu: Hints for verb movement. Quaderni del Dipartimento di Linguistica – Università di Firenze 13: 1–14. Desouvrey, L.-H. 2005. Romance clitic clusters. The case connection. In Heggie & Ordóñez (Eds), 33–79. De Vincenzi, M. & Di Domenico, E. 1999. A distinction among φ-features: The role of gender and number in the retrieval of pronoun antecedents. Rivista di Linguistica 11(1): 41–74. Di Domenico, E. 1997. Per una teoria del genere grammaticale. Padova: Unipress. Dryer, M.S. 1983. Indirect objects in Kinyarwanda revisited. In Studies in Relational Grammar 1, D.M. Perlmutter (Ed.), 129–140. Chicago IL: The University of Chicago Press. Everaert, M. & van Riemsdijk, H. (Eds) 2006. The Blackwell Companion to Syntax. Oxford: Blackwell. Evans, K.J., Lepschy, G.C., Morris, S.C., Newman, J. & Watson, D. 1978. Italian clitic clusters. Studi italiani di linguistica teorica ed applicata VII(1–2): 153–168. Ferrari, F. 2005. A Syntactic Analysis of the Nominal Systems of Italian and Luganda: How Nouns Can Be Formed in the Syntax. Ph.D. dissertation, New York University. Garrapa, L. 2008. L’applicazione dell’elisione vocalica ai pronomi nell’italiano parlato e l’interazione fra morfologia, fonologia e frequenza, Abstract, IGG, February 2008, Padua Grimshaw, J. 1997. The best clitic: Constraint conflict in morphosyntax. In Elements of Grammar, L. Haegeman (Ed.), 169–196. Dordrecht: Kluwer. Harley, H. & Ritter, E. 2002. Person and number in pronouns: A feature-geometric analysis. Language 78(3): 482–526. Harris, J. 1995. The morphology of Spanish clitics. In Evolution and Revolution in Linguistic Theory, H. Campos & P. Kempchinsky (Eds), 168–197. Washington DC: Georgetown University Press. Harris, J. & Halle, M. 2005. Unexpected Plural Inflections in Spanish: Reduplication and Metathesis. Linguistic Inquiry 36(2): 195–222. Heggie, L. & Ordóñez, F. (Eds). 2005. Clitic and Affix Combinations. Amsterdam: John Benjamins. Hyman, L.H. & Duranti, A. 1982. On the object relation in Bantu. In Studies in Transitivity [Syntax and Semantics 15], P.J. Hopper & S.A. Thompson (Eds), 217–239. San Diego CA: Academic Press. Jaeggli, O. 1982. Topics in Romance Syntax. Dordrecht: Foris. Kaisse, E. 1985. Connected Speech. The Interaction of Syntax and Phonology. San Diego CA: Academic Press. Kayne, R.S. 1991. Romance clitics, verb movement, and PRO. Linguistic Inquiry 22: 647–686. Kayne, R.S. 1993. Toward a modular theory of auxiliary selection. Studia Linguistica 37: 3–31. Kayne, R.S. 1994. The Antisymmetry of Syntax. Cambridge MA: The MIT Press. Kayne, R.S. 2000. Person morphemes and reflexives in Italian, French, and related languages. In Parameters and Universals, 131–162. Oxford: OUP. Kayne, R.S. 2006. Silent clitics. Ms, New York University. Kinyalolo, K.K.W. 2003. The case for null subject-verb agreement morphology in Bantu. In Proceedings of North East Linguistic Society 34, Stony Brook University, Vol. 2, K. Moulton & M. Wolf (Eds), 345–356. Amherst MA: GLSA. Krifka, M. 1995. Swahili. In Syntax. Ein internationales Handbuch zeitgenössischer Forschung, J. Jacobs, A. von Stechow, W. Sternefeld & T. Venneman (Eds), 2. Halbband, 1397–1418. Berlin: Walter de Gruyter. Laenzlinger, Ch. 1993. A syntactic view of Romance pronominal sequences. Probus 5 (3): 241–270. Manzini, M.R. & Savoia, L.M. 2001. The syntax of object clitics: Si in Italian dialects. In Current Studies in Italian Syntax. Essays offered to Lorenzo Renzi, G. Cinque & G. Salvi (Eds), 233–264. Amsterdam: Elsevier.

 Anna Cardinaletti Manzini, M.R. & Savoia, L.M. 2005. I dialetti italiani e romanci. Alessandria: Edizioni Dell’Orso. Monachesi, P. 1995. A Grammar of Italian Clitics [TILDIL Dissertation series]. Tilburg: Tilburg University. Nevins, A. 2007. The representation of third person and its consequences for person-case effects. Natural Language and Linguistic Theory 25: 273–313. Ordóñez, F. 2002. Some Clitic Combinations in the Syntax of Romance. Catalan Journal of Linguistics 1: 201–224. Perlmutter, D. 1971. Deep and Surface Constraints in Syntax. New York NY: Holt, Reinhart and Winston. Repetti, L. 2003. The masculine singular definite article in Italian and Italian dialects. Ms., Stony Brook University. Rezac, M. 2005. The syntax of clitic climbing in Czech. In Heggie & Ordóñez (Eds), 103–140. van Riemsdijk, H. (Ed) 1999. Clitics in the Languages of Europe [EALT/EUROTYP 20–5]. Berlin: Mouton de Gruyter. Ritter, E. 1991. Two functional categories in noun phrases: Evidence from modern Hebrew. In Perspectives on Phrase Structure [Syntax and Semantics 25], S. Rothstein (Ed.), 37–62. San Diego CA: Academic Press. Ritter, E. 1993. Where’s gender? Linguistic Inquiry 24: 795–803. Ritter, E. 1995. On the syntactic category of pronouns and agreement. Natural Language and Linguistic Theory 13(3): 405–443. Rivero, M.L. 2004. Spanish quirky subjects, person restrictions, and the person-case constraint. Linguistic Inquiry 35(3): 494–502. Rizzi, L. 1976. Ristrutturazione. Rivista di Grammatica Generativa 1: 1–54. Rizzi, L. 2000. Some notes on Romance cliticization. In Comparative Syntax and Language acquisition, 96–121. London. Routledge. Rizzi, L. 2006. On the form of chains: Criterial positions and ECP effects. In Wh-Movement. Moving On, L. Cheng & N. Corver (Eds), 97–133. Cambridge MA: The MIT Press. Rohlfs, G. 1968. Grammatica storica della lingua italiana e dei suoi dialetti. Vol. II: Morfologia. Torino: Einaudi. Selkirk, E. 1995. The prosodic structure of function words. In Papers in Optimality Theory UMOP 18, 439–469. Amherst MA: GLSA. Seuren, P. 1976. Clitic pronoun clusters. Italian Linguistics 2: 17–36. Shlonsky, U. 1989. The hierarchical representation of subject verb agreement. Ms., University of Haifa. Sigurðsson, H. 2004. Agree in syntax, agreement in signs. Ms, University of Lund. Suñer, M. 1988. The role of agreement in clitic doubled constructions. Natural Language and Linguistic Theory 6: 391–434. Taraldsen, T. 1995. On Agreement and nominative objects in Icelandic. In Studies in Comparative Germanic Syntax, H. Haider, S. Olsen & S. Vikner (Eds), 307–327. Dordrecht: Kluwer. Terzi, A. 1999. Clitic combinations, their hosts and their ordering. Natural Language and Linguistic Theory 17: 85–121. Uriagereka, J. 1995. Aspects of the syntax of clitic placement in Western Romance. Linguistic Inquiry 26(1): 79–123. Vanelli, L. 1992. Da lo a il: Storia dell’articolo definito maschile singolare in italiano e nei dialetti italiani. Rivista italiana di dialettologia 16: 29–66. Willems, E. 1949. Le Tshiluba du Kasayi pour débutants, 2nd edn. Hemptinne St. Benoit (Congo).

Pronominal object markers in Bantu and Romance* Marie Labelle

Université du Québec à Montréal

Romance pronominal clitics and Bantu object markers vary in gender and number, replace arguments, and surface to the left of the verbal root in declarative clauses. Both types of morphemes are regularly analyzed as affixes on the verb. It is argued that both have syntactic properties that justify treating them differently from lexical affixes. The argument is first made for French unstressed pronominal objects like me, le, en, drawing from their distribution in contemporary French as well as from their historical behavior. It is shown that they are syntactic objects, and that their morphophonological affix-like properties should be treated independently of their syntactic behavior. Then evidence is presented pointing to the fact that Bantu object markers are also syntactic objects of some sort. The last section asks whether one can arrive at a unified treatment of both elements. The conclusion is that by focusing on the morphophonological properties of these pronominal elements, one runs the risk of overlooking their syntactic properties.

1. The Bantu-Romance connexion Romance and Bantu languages mark objects on the verb, and in both language families the object markers share remarkable properties: they replace arguments; they vary in gender and number; and they surface, in declarative clauses, on the left of the verb while the normal object position is to the right. In the present paper we will use the term “object marker” as a theoretically neutral term for these elements. In Romance in general, and in French in particular, object markers are traditionally treated as clitic pronouns (Grevisse 1993; Kayne 1975; Zwicky 1977), but the affixal approach has deep roots in the literature (Tesnière 1959/1969: 85; BlancheBenvéniste 1975: 41; Blanche-Benvéniste et al. 1984: 70; Auger 1995; Borer 1986;

*Many thanks to Lucie Kearns and anonymous reviewers for their comments on a previous version of this work.

 Marie Labelle

Roberge & Vinet 1989; Roberge 1990; Miller 1992; Miller & Sag 1997; for a recent review, see Miller & Monachesi 2003). In Bantu languages, object markers are traditionally considered affixes (Johnston 1919/1922; Baker 1988), but some authors treat them as clitics (Duranti 1979; Hyman & Duranti 1982; Mchombo 2001, 2002) or speak of cliticization (Kimenyi 1980: 179) often on the basis of their apparent similarity with Romance object markers: they are pronominal in nature (Bresnan & Mchombo 1987) and they surface to the left of the verb root in a rigid order. On what criteria can we base ourselves to decide between the two types of analyses, and is a common analysis of Bantu and Romance object markers possible? In order to discuss the difference between clitics and affixes, we need to have working definitions of these concepts. The traditional definitions are that affixes are pieces of morphology serving as building blocks for word construction in the lexicon while clitics are syntactic objects with postlexical morphophonological attachment to a host. However, nowadays, the theoretical distinction between clitics and affixes is extremely difficult to draw. Not only is it possible in some models to generate bound flexional morphology under syntactic heads distinct from their base (e.g., Chomsky 1981; Pollock 1989), but also, some authors propose that clitics are phrasal affixes (Anderson 2005; Miller 1992). The purpose of this paper is not to argue for or against any of these models but to show that both Romance and Bantu object markers have properties that place them within the syntactic domain in some respects. Hence, if we consider that there is a continuum from independent word to clitic to affix, in both language families object markers are somewhat towards the clitic side, but perhaps more so for Romance than for Bantu. The distinction between clitic and affix, if we use the traditional definitions, is relatively clear in principle; but when it comes to object markers which attach to a verb, and which do not occupy the usual position of free pronouns, this distinction is difficult to apply without considering the facts in more detail. It is not a priori obvious that object markers are independent syntactically and semantically from the verbs on which they appear, because they often spell out the arguments of these verbs. Bantu object markers have affixal properties: they intervene between tense and modality prefixes and the verbal root, and in languages which allow more than one object marker, these appear in a fixed order (but see section 4). Similarly, Romance object markers have morphophonological properties that are affix-like. In contemporary French for example, their host is always a verb; they appear in a fixed order; there are constraints on cooccurrence, for example the sequence *me lui is excluded; it is impossible to find sequences of two identical object markers having a different function, as illustrated in (1b) with en; they are subject to arbitrary gaps, for example the sequence le lui is usually replaced in colloquial French by lui; they are subject to the morphophonological rule of elision in front of a vowel (Miller 1992: 173 ss.).

(1) a.

Pronominal object markers in Bantu and Romance 

Je crois l’auteur de ce livre1 capable de ce méfait2. I believe the author of this book capable of that misdeed

b. *J’en1 en2 crois l’auteur capable.

Given that an affixal approach seems to account for both language families, what is the motivation for a clitic approach to Romance and Bantu object markers? The answer I want to propose is that to treat these elements as clitics is to claim that they have properties that place them within the set of (morpho)syntactic elements. In what follows, I will concentrate on French, the language I know best, and on Chicheŵa, as representing their respective language families, while recognizing that, given the variety of phenomena within both language families, what is said of one of these languages does not necessarily hold for all of them. In the case of French, I argue that the syntactic computational system is involved in either the positioning of object markers or in constraining the relation between them and their thematic position. This to me justifies treating them as clitics, that is, as syntactic elements, even though these elements are phonologically dependent. In the case of Bantu object markers, there is, to my knowledge, no evidence of the same type as in that offered for French, but Mchombo (1993, 2002) offers interesting evidence that they are syntactic elements. I review the arguments in section 3. Accepting this evidence, I ask in section 4 whether a unified account is possible for both languages.

2. Object markers in French Given the ambiguous nature of French object markers, theories differ in the way they are treated. The Chomskyan framework has tended to place more emphasis on their morphosyntactic properties, and to favor the clitic analysis (in particular Sportiche 1995, 1998). Work in HPSG (starting with Miller 1992) gives more weight to the morphophonological properties of these elements and tends to favor the affixal approach. I argue here that, whatever the precise account one may want to adopt, French object markers are visible to the syntactic component. I agree with Gerlach (2002) that clitics are neither “words” nor “affixes”, but a special category which requires a special treatment. My personal position is that it is productive to approach them with a combination of Sportiche’s (1995, 1998) general type of approach and Distributed Morphology (Bonet 1991; Halle & Marantz 1993, 1994): they are generated as independent bundles of features within the functional layer of the clause, and they, or an element in their specifier, may bind a thematic role within the clause; the bundles of features are then spelled-out in a postsyntactic component. Be that as it may, my aim here is not to argue in favor of a precise analysis, but to emphasise the syntactic properties of those elements that make them special.

 Marie Labelle

A simple analysis of French object markers is to treat them as affixes reducing the valency of the verb on which they appear (Borer 1986; Suñer 1988; Roberge & Vinet 1989; Roberge 1990). For example, from the trivalent verb dire “to say” selecting an Agent, a Theme, and a Goal, one may derive in the lexicon the bivalent verb le-dire “it-say” selecting only an Agent and a Goal, and with an object marker binding the Theme argument of the root. If object markers are operators on the argument structure of the verb to which they attach, they can be considered derivational affixes of a sort. An apparent problem with this approach is that the position of Romance object markers is not always to the left of the verb. Miller & Sag (1997) propose to encode the information in the morphology, a move that is feasible for contemporary French, because object markers are preverbal everywhere (il le-dit “he says it”) except in positive imperatives (dis-le “say it”). Thus, Miller & Sag’s analysis stipulates that the lexicon generates the object marker as a suffix with verbs having positive imperative morphology, and as a prefix with all other verb forms. Well known facts go against such a simple approach to French object markers. Firstly, unselected complements may surface as object markers on the verb, as shown in (2). Because adjuncts are not lexically selected by a predicate, the fact that they may be realized as object markers on the verb is not accounted for, without additional assumptions, by the idea that object markers absorb the verb’s arguments in the lexicon. (2) a.

Il y mange souvent. he loc eats often1 He eats there often.

b. Jean leur construit une maison. Jean 3dat.pl. builds a house He is building a house for them

Secondly, and more importantly, French object markers have the displacement property: they often surface in a position distant from the one where they are interpreted. (Displacement should be interpreted here as implying the necessity of establishing a syntactic link between the position of the object marker and either

. Abbreviations used in this paper. For the Romance examples: number = person of the object marker; acc = accusative; dat = dative; sg = singular; pl = plural; loc = locative; gen = genitive; neg = negative particle; refl = reflexive; for the Bantu examples, we have retained the glosses of the originals, but harmonized the presentation of class information: number = word class; sm = subject marker; om = object marker; refl = reflexive; t/a = tense/aspect; tns = tense; pres = present; pst = past; hab = habitual; dir m = directional modality; caus = causative; appl = applicative; recip = reciprocal; fv = final vowel; indic or i = indicative; DO = direct object; IO = indirect object. Other abbreviations are spelled out in the text.

Pronominal object markers in Bantu and Romance 

a thematic position or the head selecting the object in question.) From a parsing point of view, (3) requires the parser to store the marker lui in memory and wait until it reaches the verb parlé to be able to interpret it. (3) Jean lui a souvent longuement parlé __. Jean 3dat.sg has often long talked Jean often talked to him for a long time

Displacement is particularly clear when the object marker is not an argument of the lexical verb, but of a complement (of a complement) of that head. Consider the following examples, where the verbs are underlined and the thematic position of the object is between angled brackets. In example (4), the object marker is thematically the complement of a verb embedded under a causative, but it appears on the auxiliary of the causative, three verbs higher in the clause; in (5), the object marker is the argument of the adjective complement of the verb; and in (6), it is a noun complement, the noun itself being the argument of the complement of a causative verb. (4) Les citoyens les ont tous très souvent laissé bêtement démolir 〈les〉. The citizens 3acc.pl have all very often let stupidly demolish The citizens have all very often let them stupidly demolish. (5) Luc le considère [〈le〉 Luc 3acc.sg considers Luc considers him intelligent.

intelligent.] intelligent

(6) Lui qui aimait tellement cet air, il en aura au moins entendu jouer he who liked so-much this air, he 3gen will-have at least heard play [les premières notes 〈en〉] the first notes

avant de s’éteindre. before of refl-dying

He, who liked this air so much, he will have at least heard (someone) play the first notes of it before dying.

On the basis of such facts, Labelle (1985) argues that object markers do not lexically, or pre-syntactically, operate on the verb on which they appear, and that therefore they must be syntactic elements. However, Miller (Miller 1992; Miller & Sag 1997; Miller & Monachesi 2003; also Monachesi 1999, 2005 for other languages), working within the HPSG framework, shows that an affixal approach can account for the facts if the verb affixed in the lexicon with an object marker adopts the argument structure of all the predicates that compose with it, up to and including the one selecting the argument spelled-out by the object marker. Hence, the pronominal affix affects the valency of the verb on which it appears, because this verb’s argument structure is extended to include the argument corresponding to the object. This approach accounts for the cliticization of adjuncts by assuming

 Marie Labelle

that they are additions to the argument structure of a predicate. Notice that, while treating French object markers as lexical affixes, this account requires the construction of a sequence of dependency relations between the host verb and the predicate selecting the object, a process not purely lexicon internal, but dependent on syntactic combinatory operations. Hence the account has a syntactic component to it; it does not treat object markers as syntactic objects, but it builds a complex syntactic object out of the various predicates involved. A third argument for treating French pronominal object markers as syntactic objects is that they are subject to constraints on syntactic A-bar movement (Godard 1992; Haverkort 1994; Sportiche 1995, 1998; Tellier 1991: 90). For example, an object marker cannot be extracted from a tensed clause (7), and it cannot be extracted if the embedded infinitive has a specified subject (8). (7) *Jean les veut [C que je voie 〈les〉]. Jean 3acc.sg wants that I see 〈3acc.sg〉 Jean wants me to see them. (8) a. *Jean lui a laissé [Pierre parler 〈lui〉]. Jean 3dat.sg has let [Pierre speak 〈3dat.sg〉] Jean let Pierre speak to him.(Sportiche 1998: 254, ex. 14c also 1995, ex. 13c) b. *Cela leur fera [téléphoner Jean 〈leur〉]. This 3dat.pl will-make [telephone Jean 〈3dat.pl〉] This will make Jean telephone them. (Kayne 1975: 231)

Moreover, extraction of the genitive marker en, replacing complements in de X (“of X” or “from X”), occurs in the same contexts as extraction of the WH de quoi (“of what”). While extraction from a definite or indefinite DP is possible (examples 6, and 9), it is excluded if the DP is introduced by a possessive determiner (a “specified subject”) (10); similarly it is impossible to extract the noun complement if it is locative rather than genitive (11) (Sportiche 1995).

(9) Extraction from a DP possible:

a.

Paul en a vu [une photo 〈en〉]. (= de la gare) Paul of-it has seen a picture 〈of-it〉 (= of the station) Paul has seen a picture of it.

b. De quoi Paul a-t-il vu [une photo 〈de quoi〉]? of what Paul has he seen a picture 〈of what〉? Of what did Paul see a picture? (10) …except when the DP has a possessive determiner (subject): a. *Paul en a vu [ma photo 〈en〉]. Paul of-it has seen my picture 〈of-it〉 Paul has seen my picture of it.

Pronominal object markers in Bantu and Romance 

b. *De quoi Paul a-t-il vu [ma photo 〈de quoi〉]? of what Paul has he seen my picture 〈of what〉 Paul saw my picture of what? (11) …and when the complement has a locative value: a. *Jean en a déchiré [une dépêche 〈en〉]. (= de Paris) Jean from-there has torn a dispatch 〈from-there〉 (= from Paris) Jean tore a dispatch from there. b. *D’où Jean a-t-il déchiré [une dépêche 〈d’où〉]? from where Jean has he torn a dispatch 〈from where〉 Jean tore a dispatch from where? compare with: c.

Jean en a déchiré [la première page 〈en〉] (= du livre) Jean of-it has torn the first page 〈of-it〉 (= of the book) Jean tore the first page of it.

d. De quoi Jean a-t-il déchiré [la première page 〈de quoi〉]? of what Jean has he torn the first page 〈of what〉 Jean tore the first page of what?

To my mind, examples (10) and (11) are crucial, as they demonstrate that the relation between the object marker and its thematic position is of the same type as that governing syntactic movement. It is not clear how the affixal approach briefly outlined above can account in a principled way for the parallel between object markers and WH constituents: why can a verb absorb the argument structure of its noun complement only if this noun has a genitive, rather than a locative, complement, and only if the noun is not introduced by a possessive determiner? Sportiche (1998) and Valois (1991) account for the pattern illustrated in (9)–(11) by appealing to the notion of antecedent government: the thematic position associated with what is for them a clitic must be antecedent governed, i.e., properly bound. In Sportiche (1995, 1998), the binder is not the clitic itself, but an empty pronominal in the specifier of a Clitic Phrase. In either case, the crucial point is that there is a syntactic relation between the clitic or its specifier and the corresponding thematic position.2 I conclude that French object markers have syntactic properties that must be accounted for, and that cannot be ignored by proponents of an affixal approach. I now turn to evidence from the history of French object markers which shows quite clearly that the morphophonological properties of these elements should not be taken as an indication of their status over other considerations relating to the syntactic .

This, of course, when the clitic is thematic.

 Marie Labelle

behavior of these elements. The data to be presented here was published in joint work with Paul Hirschbühler over some years (Hirschbühler & Labelle 1994, 2000, 2003, 2004; Labelle & Hirschbühler 2001a,b, 2005; Labelle 2007). Before starting, it is important to establish that in Old French, i.e., from the 9th century to the 13th century, unstressed object pronouns (the ancestors of contemporary object markers, or pronominal clitics) passed nearly all the tests used to argue for the affixal character of pronominal object markers in contemporary French. These properties of Old French weak object pronouns, discussed in de Kok (1985) and Foulet (1968 [1928]) are summarized here; they clearly show that Old French weak object pronouns shared the main properties of contemporary French object markers on the verb. – Selection of the host: in the Old French texts, weak pronouns always appeared next to the verb whatever the position of the verb in the clause; – Constraints on cooccurrence: just as in contemporary French, the sequence me lui (or rather *lui me, since the normal order in Old French was le me) was excluded; – Rigid order within a clitic group (even though the ordering changed during Middle French – between the 13th and the 16th centuries – Galambos 1985); – Haplology: one did not find sequences of two identical pronouns having a different function; – Arbitrary gaps: the sequence le li (nowadays le lui, i.e., 3acc-3dat) was replaced by li (lui), a gap which survives in many dialects, but not in standard written French; – Weak object pronouns could not be coordinated; – Weak object pronouns were subject to morphophonological rules, such as elision in front of a vowel: je l’aime ‘I him love’; in the oldest texts, they also elided in post-verbal position in front of a vowel initial DP (Donne m’un feu ‘give me a fief ’), which suggests that this rule was initially linked to lack of stress. If we limit ourselves to these morphophonological properties (with the exception perhaps of the last one), we would apparently have to conclude that Old French weak object pronouns were not clitic elements but affixal object markers on the verb. However, a number of empirical facts regarding the grammar and history of Old and Middle French weak object pronouns lead us to reject this position. Firstly, for Old French, the position of weak object pronouns with respect to the verb depended on the syntactic context, and the changes in their position with respect to the verb are best explained by changes in clausal syntax. Consider the placement of weak pronouns in tensed main clauses. Old French was a verb second or V2 language in main clauses, a fact that may be accounted for by the presence of a tense feature [+T] on the Fin node (the first node of the CP layer), triggering movement of the verb to this node, and by the necessity to fill a discourse-related position (Topic or

Pronominal object markers in Bantu and Romance 

Focus) to the left of the verb. In V2 clauses, pronominal objects always occurred to the left of the finite verb. This word order follows if the pronoun had a [+T] feature which adjoined it to the element bearing this feature, i.e., Fin in main clauses and the inflectional head T elsewhere. This analysis is schematized in (12). (12)

V2 : [TopP XP [FinP le fait [ … 〈fait〉… [vP… 〈fait〉…]]]] +Top +T CP layer

TP

(13) Toutes ces choses te presta Nostre Sires. all these things 2dat.sg lent Our Lord Our Lord lent you all those things. (Queste: 68, 28 in de Kok: 74)

Apart from V2 main clauses, Old French had a number of verb initial, or V1 main clauses. In the oldest texts, weak object pronouns were postverbal in V1 main clauses. This alternation between preverbal vs postverbal position was clearly dependent on verb position, i.e., on the syntax of the clause. The postverbal position follows if in V1 main clauses the verb raised over the pronoun to the first position of the CP layer, where it satisfied the language’s requirement for a discourse-related element. This is schematized in (14). (14) Stage 1: w eak object pronoun post-verbal in CP initial position; preverbal elsewhere: V→Top (et) [TopP Fait [FinP le 〈fait〉 [TP … ]]]] CP layer

TP

(15) Fuit li li sans et la colour. Leave 3dat.sg the blood and the color She loses her blood and color. (Piramus & Tisbé. ed. C. de Boer. Paris: Champion 1921: 664–6)

Around 1170, weak object pronouns started to appear preverbally in coordinated V1 main clauses, while remaining postverbal in non coordinated ones, as schematized in (16a–b). We assume that independent main clauses had the structure in (12) and (16a), while main clauses introduced by a coordinator were now generated without a discourse-related higher head where the verb could raise, as illustrated in (16b), a precursor perhaps of the breakdown of the V2 grammar. (16) Stage 2 (c. 1170): weak object pronoun preverbal after coordinator: a.

[TopP fait [FinP le … [...]

b.

[coord et

[FinP le fait [TP … ]]]]

but CP

CP

TP

 Marie Labelle

(17) Tele vie mena com ot fait Acháb et li suen [et se forfist such life led as have done Achab and the his and reflcommit-a-mis deed vers notre Seignor]. against our Lord He led a life like that led by Achab and his people, and sinned against our Lord. (QLR: 190, 27, in de Kok: 92)

Throughout that period, the preverbal vs postverbal alternation in weak object pronoun position held for all illocutionary types of clauses: declaratives, imperatives, questions. It is clear that the alternation was dependent on the syntactic context, and not on verb morphology. For independent reasons, weak object pronouns became preverbal in absolute CP initial position in declaratives and questions during the 13th century; after that, the possibility for an alternation remained only in imperatives. In imperatives, however, the alternation between preverbal and postverbal position according to the syntax of the clause survived until the 18th century: Dis-le (“say it”), but … et le dis (“and say it”). If the structures outlined above are correct, the change in the position of the weak object pronoun with respect to the verb results from a change in the position of the verb with respect to that occupied by weak pronouns in main clauses, i.e., whether the verb raised above the pronoun to a discourse-related head or not. The syntactic position of weak pronouns was not affected by the change. This analysis entails that weak object pronouns were syntactically independent from the verb, as they remained in place when the verb moved. Because they were also phonologically dependent elements, we are led to conclude that Old French weak object pronouns were already clitic elements. Secondly, until the 17th century, weak object pronouns could have scope over coordinated VPs, a fact indicating that they were not lexical affixes. (18) Et Gauchier Le Fevre de Orgelet le and Gauchier Le Fevre de Orgelet him

[reçut 〈le〉 et norri 〈le〉 et [received 〈him〉 and fed 〈him〉 and

aleva 〈le〉 en sa meson par douze ans]. raised 〈him〉 in his house for twelve years] (Mir. Louis 50: 8, in de Kok: 2) (1283)

Thirdly, until the end of Middle French (16th century), one finds clitic climbing out of infinitives. Crucially, weak object pronouns could even climb out of infinitives introduced by a preposition (19a) or by a WH element (19b) (Martineau, 1990, 1991).3 . Similar examples are found in Old French texts; the following one is from Aucassin & Nicolette (end 12th c.): (1) se je le seusc u trover 〈le〉, /Je ne l’ eusce ore mie a querre 〈l’〉. (XL) if I 3s knew where find /I neg 3s had hence not to look-for If I knew where to find her, I would not have to look for her any more.

(19) a.

Pronominal object markers in Bantu and Romance 

il la commence [CP a sonner 〈la〉]. he it starts to ring 〈it〉 He starts to ring it. (CNNA [1468] : 452/103; in Martineau 1990)

b. On ne le sçavoit [CP à qui demander 〈le〉]. One neg it knew to whom (to) ask 〈it〉 One didn’t know to whom to ask for it. (CNNV [1505–15]: 84/27; in Martineau 1990)

In these examples, the pronoun raises out of a CP, something not expected if it is a lexical affix on the verb at that time. In contemporary French, clitic climbing survives only with causative and perception verbs. Consider now the phenomenon of interpolation: that is, the possibility of a clitic being separated from the infinitival verb. In the oldest stage of French, weak object pronouns associated with an infinitival verb usually climbed out of the infinitive to appear on the main verb, as illustrated above; in the rare cases where pronouns appeared in the infinitival clause, they were full pronouns or they occurred in immediate post-verbal position (de Kok 1985: 243 ss.). It is only during the 14th century that one starts finding preverbal weak object pronouns on the infinitival verb (de Kok 1985: 326 ss.). The first governed negated infinitive with ne and a pas/point-type support negative to be reported in the literature dates from 1372 (de Kok 1985: 335), and it can be seen that both the pronoun and the verb precede the negative VP adverbial point. (20) Pour ce, (…) est il bon de ne se haster point. For that, (…) is it good of neg refl hasten not For that, (…) is it good not to hasten. (La Tour Landry [1372]: 102; in de Kok 1985: 335)

However, in the same text and in texts of the same period, one sees elements intervening between the pronoun and the verb in non-negative clauses. (21) l’on ne se puet trop humilier envers Dieu, ne le trop one neg refl can too-much humiliate towards God, nor 3acc.sg too-much servir. serve One cannot humiliate oneself too much towards God, nor serve him too much. (La Tour Landry [1372]: 153, in de Kok 1985: 339)

The intervening element could be an XP or a strong adverb: (22) a. et son père, pour la de son duel gecter, ly parla. and his father, for 3acc.sg of her grief throw, 3dat.sg spoke and his father spoke to her to distract her from her grief (Saintré [1385–1460]: 5,1; in Pearce 1990: 223)

 Marie Labelle

b. (…) se prindrent toutes a le tresgrandement loer. (…) refl took all to 3acc.sg very-much praise (they) all started to praise him exceedingly (Saintré [1385–1460]: 129,11, ibidem) c.

et marchoit vers sa femme pour la tresbien frotter. and walked towards his wife fo 3acc.sg very-well rub and (he) went to his wife to rub her well (CNNA [1468]: 241; in de Kok: 339)

From the examples given in de Kok, interpolation by a non-negative adverb seemed to have been possible without interruption from the end of the 14th century. For negative adverbs, de Kok (1985: 335) gives the following evolution: the word order in (20) was the norm until the beginning of the 17th century, when it was replaced by the order pronoun-negative adverb-verb. (23) pour ne se point souiller ny l’esprit ny le corps. for neg refl not taint neither the spirit nor the body In order not to taint one’s spirit or body. (Colletet: 109, 47, in de Kok p. 336)

The position of weak object pronouns throughout this period appears to be the T node of the infinitival, to the left of adverbs and next to the negative marker ne in negative clauses. Here again, the change may be explained by a change in verbal behavior: the infinitival verb, but not the pronouns, stopped raising to T, with a time-lag between positive and negative clauses perhaps because the negative marker ne attracted the verb (on the evolution of verb position in negative clauses, see Hirschbühler & Labelle 1994). (24) [TP cl-V [(XP)[VP …]]] → [TP cl [(XP/adv)[VP V …]]]

Of course, even though the pronouns were separated from the verb, the other properties mentioned above held (rigid order, etc.); this tells us that the clitic group forms a morphological unit, whether cliticized on the verb or not. Interpolation with strong XPs and heavy adverbs became rare during the 15th century, perhaps because a scrambling position intervening between VP and T disappeared at that time, but interpolation with a small set of VP adverbs was the norm until the 18th century: pas “not”, point “not”, plus “no-more”, jamais “never”, rien “nothing”, bien “well”, mieux “better”, trop “too-much” (Galet, 1971), as exemplified in (25). (25) a.

Je vous ordonne de ne le point voir au parloir. I 2dat.pl order of neg 3acc.sg not see at the parlor I order you to not see him at the parlor. (Diderot [1784] La religieuse, 10/18, p. 228.)

b. Et ai-je lieu de n’ en pas espérer un succès aussi favorable? and have-I reason of neg gen not hope a success as favorable And should I not hope for as favorable a success from it ? (Cyrano [1655])

Pronominal object markers in Bantu and Romance 

c.

…pour compenser sa mollesse et me mieux faire avaler …to compensate his weakness and 1dat.sg better make swallow

la pilule. the pill

…to compensate his weakness and make me better swallow the pill (Lesage A.-R. [1732], Histoire de Gil Blas de Sant: 745.) (In Romanciers du 18E S., Ed. Etiemble, T.1. Paris, Gallimard, 1960.)

It is only during the 18th century that weak object pronouns stopped raising systematically to T in infinitivals and remained with the verb within the VP, the grammar of Contemporary French. Whatever the exact account of these facts, the interpolation of an element between the pronoun and the verb indicates that, even though Old French weak object pronouns had morphophonological properties which were similar to those used to argue for an affixal approach to contemporary French object markers, and even though they were always adjacent to the verb during the oldest period of French, they were not affixes on the verb. Their placement under T was independent of that of V, so that when the infinitival verb stopped raising to T, pronouns continued to do so for centuries. Hence, affix-like morphophonological properties should not be taken as necessarily reflecting lexical affixation, as the elements involved may have autonomous syntactic behavior. The evidence discussed leads us to conclude that at least until the 18th century, French weak object pronouns, or object markers, were manipulated by the computational system independently of the verb and were clitic elements rather than affixes. This conclusion is strengthened by the observation that en and y did not always undergo the changes at the same time as the other object markers, suggesting that object markers are positioned according to their own individual specifications, something which, again, is difficult to reconcile with an affixal approach. For example, while we said that weak object pronouns remained preverbal in coordinated main clauses until well into the 18th century, as can be seen in (26), en and y anticipated the postverbal position in this context in imperatives, as evidenced by the comment of the grammarian Restaud, cited by de Kok (1985: 315). Restaud says that the position for weak object pronouns in imperatives is normally preverbal, but he adds the commentary in (27) stating that en and y are “more normal” in postverbal position after a coordinator (for more examples and a discussion, see Hirschbühler & Labelle 2003): (26) Sois homme et me fais voir le fils d’ Agamemnon. be man and 1dat.sg make see the son of Agamemnon Be a man and let me see the son of Agamemnon (Guimond de la Touche [1758], Iphigenie En Tauride. (p. 240); in Repertoire Gen. du Theatre Fr., T.26. Paris, Menard & Raymond 1813.)

 Marie Labelle

(27) Restaut [1766]: 241; in de Kok, 1985: 315): ‘il est plus ordinaire de dire, écoutez ma proposition, et réfléchissez-y; recevez ma remontrance, & profitez-en; que y réfléchissez, en profitez.’ ‘it is more normal to say écoutez ma proposition, et réfléchissez-y; recevez ma remontrance, & profitez-en than to say y réfléchissez, en profitez.’

On other occasions, en and y displayed a behavior more conservative than the other object markers. For example, they climbed out of infinitives introduced by modals for a longer time than the other clitics. Such examples are found until the 19th century. (28) Si l’ on n’ y doit pas céder 〈y〉 if one neg dat must not yield If one must not yield to-it always

toujours. always

(Sainte-Beuve Ch [1840], Port-Royal T. 1

(Paris, Hachette 1860.))

(29) Oh! Ce livre, le monde entier l’ attend! Oh! this book, the whole world is waiting for it! Et il n’ en veut rien dire 〈en〉, pas même and he neg gen wants nothing say, not even

le titre! the title

And he doesn’t want to say anything about it, not even the title! (Pailleron E. [1869], Le Monde où l’on s’ennuie. In Pieces de Theatre, T.3. Paris, Calmann-Levy 1895.)

Interpolation of an adverb between the object marker and the infinitival verb is occasionally found until well into the 20th century with en and y, but not with the other object markers. A particularly revealing example is that in (31), dating from 1966, where en precedes the negative adverb, but se follows it. (30) a.

Je préfèrerais n’ en rien dire 〈en〉. I would-prefer neg gen nothing say I would prefer not say anything about it.

b. J’ai regretté de n’y pas être allé 〈y〉. I have regretted of neg loc not be gone I regretted not to have gone there. (31) …de mourir et de pleurer à n’en plus finir à n’ en plus se …to die and to cry to neg gen more finish, to neg gen more refl réveiller jamais. awaken ever …to die and to cry endlessly, to the point of never waking up ever (René Fallet [1966]. Banlieue Sud-Est. Denoël 221)

These peculiar properties of en and y tell us that French object markers are not treated as a homogeneous set of affixes, but as a heterogeneous set of pronominal

Pronominal object markers in Bantu and Romance 

elements; when a syntactic change affects some of them, not all of them follow suit immediately, even though they tend to eventually align their behavior. To summarize, we have surveyed a number of phenomena which are best accounted for if French object markers have always been manipulated by the syntactic computational system, even though their morphophonological properties were and are affix-like in many respects. In consequence, they are truly clitic elements. This is not denying the fact that the changes undergone by French pronominal object clitics have been towards less and less autonomy with respect to the verb. In tensed clauses, their preverbal vs postverbal position is now limited to positive imperative morphology, and in infinitival clauses, they tend to remain on the infinitival verb instead of climbing to the tensed verb or to the T node of the infinitive. But the facts reviewed at the beginning of this section show that they must be visible to the syntax, as they are subject to constraints governing syntactic elements. If we consider that there is a continuum between clitic and affix, French object markers are still somewhere towards the clitic side. 3. Object markers in Bantu To my knowledge, the type of phenomena reviewed for French indicating that object markers must be visible to the syntactic component or that they are manipulated independently of the verb by the computational system has been very little discussed for Bantu languages. However, Bresnan and Mchombo (1987) treat Bantu object markers as incorporated pronouns, while Mchombo (1993, 2001, 2002) goes further in claiming that they are syntactic objects, and more specifically “clitics”. Bresnan and Mchombo (1987) and Bresnan (1997, 2001) show that Chicheŵa object markers are incorporated object pronouns, or bound pronominals. This is because an object NP can appear in the clause only if there is no object marker on the verb. Conversely, when there is no object marker, a full NP is required, and it must appear within the VP to the immediate right of the verb. (32) Njûchi zi-ná-lúm-a alenje. 10.bees 10.sm-past-bite-indic 2.hunters The bees bit the hunters. (Bresnan & Mchombo 1987, ex.1) (33) Njûchi zi-na-wa-lum-a. 10.bees 10.sm-past-2.om-bite-indic. The bees bit them.

Thus, there is complementary distribution between object markers and object NPs. When there seems to be cooccurrence of an object marker with an NP, as in (34), the authors show that this NP is a VP external topic.

 Marie Labelle

(34) Njûchi zi-na-wa-lum-a 10.bees 10.sm-past-2.om-bit-indic The bees bit them, the hunters.

alenje. 2.hunters (Bresnan & Mchombo 1987, ex. 2)

Hence Chicheŵa object markers are pronominal in nature, and not agreement markers.4 Bresnan (1997, 2001) further makes a distinction between full pronouns and reduced pronouns, the latter including bound pronouns and clitics. She argues that reduced pronouns are universally specialized for topic anaphoricity; they are always anaphoric to a preceding discourse referent, whereas free pronouns are used by default to introduce new topics or for contrast, and they may refer anaphorically only where no bound pronoun is available. In this typology, Chicheŵa object markers are bound pronouns, appearing as affixes on the verb. As a result, (35) is ungrammatical, because the lion is given information, hence discourse anaphoric, and consequently a bound pronoun/object marker would be required instead of a full pronoun: (35) *Fîsi a-na-dy-á hyena sm-rec.past-eat-indic The hyena ate it, this lion.

í’wo mkángó uwu. 3.it 3.lion this (Bresnan & Mchombo 1987, ex. 16b)

Now, having the semantics of a pronominal and being in complementary distribution with a full NP is not evidence of syntactic autonomy. Bresnan (1997) considers Chicheŵa object markers as pronominal affixes, and not as clitics. From what we have said until now, nothing prevents us from assuming that the Chicheŵa pronominal object marker is a type of derivational morpheme absorbing one argument on the argument-structure of the verb in the manner described in section 2. Such a valency-absorbing approach would account for the complementary distribution between the object marker and the NP, and it may also account for (35), given a proper theory of the referential properties of internally bound arguments. Mchombo (1993, 2001, 2002), however, provides evidence from Chicheŵa suggesting that object markers are syntactic objects; in his 2001 and 2002 papers, he further claims that they are clitics. In Chicheŵa and closely related Bantu languages, argument structure changing affixes (passive, causative, applicative, etc.) are extensions appearing as suffixes after the verbal root. To the left of the root are found tense and modality prefixes, the subject marker, and object markers. (36) [Subject Marker – Tense – Mood – Object Marker [[VB [Verbal Root] Extensions] fv]]

. The authors show that some Bantu languages are now undergoing grammaticization of object markers into agreement markers (Bresnan & Mchombo 1987: 777).

Pronominal object markers in Bantu and Romance 

The suffixal extensions attach to the verb root first, forming the verbal base (VB). To the base is affixed a final vowel (fv) sometimes glossed as indicative/subjunctive mood; the base with the final vowel forms the verb stem. The verb stem is the locus of various processes like vowel harmony, tone inheritance, and nominalization, which suggests that it is a lexical domain. The prefixes do not participate in these processes, a fact taken by Mchombo (2002) to argue that they are post-lexical clitics. Crucially, object markers pattern like the other prefixes in this respect. Notice that this already hints at the fact that the language does not treat them as argumentstructure changing elements, contrary to what we might expect if we believe them to be valency absorbers. Consider now the reflexive and reciprocal morphemes. The two morphemes appear in different positions of this schema. The reflexive morpheme dzi is an object marker. It immediately precedes the verb stem (37), and in those Bantu languages where only one object marker is possible, such as Chicheŵa, it cannot co-occur with another object marker (38). (37) Mkângo u-na-dzí-súpul-a. 3.lion 3.sm-pst-refl-bruise-fv The lion bruised itself.

(Mchombo 1993, ex. 4a)

(38) *Mkângo u-na-dzí-wa-pats-a alenje. 3.lion 3.sm-pst-refl-2.om-give-fv 2-hunters The lion gave itself (them) the hunters.

(Mchombo 1993, ex. 5a)

In contrast, the reciprocal marker an follows the verb like the argument structure changing extensions (39). Mchombo (1993) says that the reciprocal interacts with the other extensions; for example, (40) is interpreted as [cause [hit-recip]]: the reciprocal affects the subject of the root verb, and not that of the derived causative verb. (39) Mbidzi nd’ nkhandwe zi-ku-mény-an-a. 10.zebras and 10.foxes 10.sm-pres-hit-recip-fv The zebras and the foxes are hitting each other.

(Mchombo 1993, 15b)

(40) Mbidzi zi-ku-mény-an-its-a nkhandwe. 10.zebras 10.sm-pres-hit-recip-caus-fv 10.foxes The zebras are making the foxes fight one another.

(Mchombo 1993, 16)

Mchombo (1993) shows that the reflexive and reciprocal morphemes have properties corresponding to their respective positions. The reciprocal is involved in the phonological processes affecting the Verb Stem like vowel harmony. It also participates in nominalization, a process whereby a verb stem is turned into a noun by replacing the final vowel a with either i or o, then prefixing the appropriate class-marker.

 Marie Labelle

(41) teng-an-a ‘take each other’ mténg-an-o ‘death pact’ (42) gwiriz-an-a ‘agree’

chigwiriz-an-o ‘cooperation’

(Mchombo 1993, 11)

The reciprocal is reduplicated with the verbal stem; crucially neither the object markers nor the reflexive reduplicate with the stem (Mchombo gives no translation for (44)–(45)): (43) Alenje a-na-[tém-ér-ân-a][tém-ér-ân-a] nzimbe. 2.hunters 2.sm-pst-cut-appl-recip 9.sugar-cane The hunters cut for each other totally the sugar-cane. (Mchombo 1993, 13b) (44) *Mbidzi 10.zebras

zi-ku-[i-gw-ets-a][ i-gw-ets-a] mitengo mnkhalango. 10.sm-pres-om-fall-cause 4.trees 18.9.forest (Mchombo 1993, 14a)

(45) *Mbidzi zi-ku-[dzi-thamangitsa][dzi-thamangitsa] mnkhalango. 10.zebras 10.sm-pres-refl-chase 18.9.forest (Mchombo 1993, 14c)

The reciprocal appears in bare imperatives, a construction formed with a bare verb stem. (46) Mang-ir-an-a-ni nyumbá zólîmba. build-appl-recip-fv-pl 10.houses 10.sm-strong Build one another firm houses.

(Mchombo 1993, 19a)

The reflexive morpheme, like other object markers, is excluded from the same construction, and, according to Mchombo, it can only form an imperative with a subjunctive form of the verb, where the final vowel is e. (Here, a reviewer points out that other non reflexive object markers are possible in the imperative, even without the subjunctive, suggesting that the reflexive has perhaps a different status, and undermining Mchombo’s argument.) (47) Dzi-máng-ir-é nyumbá zólîmba. refl-build-appl-subjunctive 10.houses 10.sm-strong Build yourself firm houses. (Mchombo 1993, 19b)

The above morphophonological arguments clearly suggest that the reflexive, and by extension the other object markers, have a status distinct from that of the reciprocal morpheme. Mchombo (1993) further shows that they contribute differently to the interpretation, over and above the difference between reflexive and reciprocal meaning. The reflexive sentence in (48) is ambiguous between strict and sloppy identity: it can mean that the hunters despise themselves more than the fishermen despise them, that they despise themselves more than they despise

Pronominal object markers in Bantu and Romance 

the fishermen, or that they despise themselves more than the fishermen despise themselves. The reciprocal in (49) is unambiguous, allowing only the sloppy identity reading: the hunters despise each other more than the fishermen despise each other. (48) Alenje á-ma-dzi-nyóz-á kupósá asodzi. 2.hunters 2.sm-hab-refl-despise-fv exceeding 2.fishermen The hunters despise themselves more than the fishermen. (Mchombo 1993, 20a) (49) Alenje á-ma-nyóz-an-á kupósá asodzi. 2.hunters 2.sm-hab-despise-recip-fv exceeding 2.fishermen The hunters despise each other more than the fishermen. (Mchombo 1993, 20b)

Mchombo argues that the difference is explained if the reciprocal is a derivational morpheme, deriving intransitive reciprocal verbs. The only argument of the reciprocal is its subject. This accounts for the unambiguity of (49): in the second part of the comparative, asodzi “fishermen” can only be the subject of the verb. The ambiguity of the reflexive, on the other hand, follows if the verb is transitive. The reflexive behaves like an argument, specifically an anaphor that needs to be bound to the subject. Contrary to the reciprocal, it does not derive an intransitive verb. In consequence, in the second part of the comparative, the argument can be either the subject or the object of the transitive verb. The conclusion reached by Mchombo is that the reflexive morpheme is a syntactic argument, a pronominal object of the verb, and the same holds for object markers. This is why he treats these elements as clitics. The reciprocal morpheme, on the other hand, is an argument-changing lexical operator. In Chicheŵa, the verbal base, including the reciprocal morpheme, is a lexical element; the object markers, however, have syntactic properties. To this evidence, we may add a few other points. Firstly, Kimenyi (1980: 180) points out that in Kinyarwanda the locative may be “incorporated”, “even without being promoted to direct object status”, meaning, if I understand correctly, that there is no applicative morpheme justifying its presence. If this is correct, it means that at least some unselected adjuncts may appear as object markers on the Bantu verb in some languages. This would indicate that object markers are not simply thematic argument binders on the verb, an observation that parallels the one we made for French in section 2. Also, in Kinyarwanda, a Bantu language accepting more than one object marker on the verb, there is a rigid order of object markers according to a thematic hierarchy: patient/theme > goal > beneficiary, with exceptions, e.g., the first person appears next to the verb root (but in front of the reflexive) whatever its thematic role (Duranti 1979; Kimenyi 1980). However, the position of the

 Marie Labelle

locative object marker is mobile within the sequence of object markers (Kimenyi 1980: 183): (50) a.

Ba- ra- ha- zi- tu- gu- he- er- a. 2.sm-t/a- 16.om-10.om-1pl.om-2sg.mo-give-appl-vf

b. Ba- ra- zi- ha- tu- gu- he- er- a. 2.sm -t/a-10.om -16.om -1pl.om -2sg.mo -give-appl-vf c.

Ba- ra- zi- tu- ha- gu- he- er- a. 2.sm -t/a-10.om - 1pl.om -16.om -2sg.mo -give-appl-vf

d.

Ba- ra- zi- tu- gu- ha- he- er- a. 2.sm -t/a-10.om- 1pl.om - 2sg.mo - 16.om -give-appl-vf They are giving them to us for you. (Kimenyi 1980: 183, ex. 18)

In Kihaya the instrumental also seems to have some mobility (Duranti 1979: 41). This mobility of some object markers suggests that the sequence of object markers is not lexically specified, but constructed in accordance with constraints that may be thematic, related to the person of the object marker or others. If an object marker does not interact with the constraints, its position is free. While this mobility can be derived in the lexicon, it is more typical of syntactic computation. (See Bresnan & Mchombo 1995 for the syntactic nature of locative class markers in the nominal domain in Chichewa.) 4. A re Romance clitics and Bantu object markers amenable to a unified analysis? I have shown that French clitics and Chicheŵa object markers are syntactic elements. That does not mean that they are the same type of element. In contrast to Romance clitics, Bantu object markers appear to always attach to a verb base which includes the head selecting the corresponding argument (except perhaps for some locatives). Bantu object markers do not have the displacement property of Romance clitics. Their position with respect to the verb stem is also always to the left, while that of Romance clitics alternates between left and right. Despite these distinct properties, I want to show in this section that, instead of focusing on the differences, some analyses bring out the similarities between these arguments, and by doing so offer insights into the crosslinguistic similarities between reduced pronominals. Consider the following tree, apparently proposed for Bantu by Gadelii (2000) – cited in Petzell (2004) – on the basis of Mchombo (1997: 191).

Pronominal object markers in Bantu and Romance  NEGP NEG

AGRSP AGRS

TP MOODP

T MOOD

AGROP AGRO

VP V +EXT

si á ku dzá ngo u phîk-its-ir- á      3. cook--- He is not merely coming to get pumpkins cooked for the lion.

XP

XP

ma-úngu m-kángo 6.pumpkin 3.lion

This structure is similar in many respects to the one proposed in various papers for Romance. To take one example, consider the derivation proposed for Spanish by Franco (2000: 181), schematized in (52), simplifying from Franco, who constructs inflectional morphology in syntax by having the verb raise from its VP internal position all the way up to AgrS. (52)

[AGRSP [TENSEP [AGRIOP me [AGRDOP la me it will-send (He) will send it to me

[VP enviará pro pro]]]]]

In this structure, clitics are generated under agreement heads between Tense and the VP. The tensed verb, passing through these heads, picks up the clitics on its way to the tense and subject agreement nodes. It should be clear that the position of object clitics in (52) is comparable to that of object markers in (51). This suggests the interesting possibility that Romance and Bantu object markers are generated in the same structural position. The difference between Romance and Bantu would then follow from the fact that in Romance, but not in Bantu, object markers raise to a higher head, either independently, as was certainly the case from Old French until at least the 18th century, or with the verb, as in (52). A second difference is that, in Romance, the various projections are syntactically and morphologically autonomous; adverbs

 Marie Labelle

may intervene between them. Mchombo (2002: 193) presents for Bantu a tree very similar to the one above, but he claims that the elements to the left of the verbal stem are clitic heads, not dominated by XP, but forming a complex verb with the stem. In this perspective, the various heads are specified as being proclitics on the verbal stem, preventing any item from intervening. As for the position of negation, Zanuttini (1997) discusses two types of negative markers in Romance, pre-Infl and postInfl. Pre-infl negative markers, found in Spanish and Italian (53a), occupy a position parallel to that in (51) (to the left of the weak agreement subject clitics found in some Romance varieties). A post-Infl negative marker is best observed in Quebec French (and other colloquial varieties), where pas surfaces after the tensed auxiliary (53b). For standard French, Pollock (1989) assumes that the negative marker ne is basegenerated next to pas, and that it cliticizes to the left of the clitic group. (53) a.

Maria non glielo ha dato. Maria neg 3.dat.3.acc aux given

(Italian, Zanuttini 1997: 18, ex. 3a)

b. Marie lui a pas donné. (colloquial French; le+lui reduced to lui) Marie 3.dat aux neg given Marie didn’t give it to him.

Hence, a structure like (51) may account for the position of Bantu and Romance clitic groups. If we consider now the order of clitics within the clitic group, we may turn to Cocchi (2000), who views Bantu object markers as representing “the morphologically bound counterpart of the free Romance clitics, which have a very similar syntactic behaviour” (p. 90) and who develops a unitary analysis of clitic groups within the framework of Manzini and Savoia (1998, 2002). In this framework, clitics are viewed as bundles of features including person, number, definiteness, categorical, and aspectual features. In a nutshell, Manzini and Savoia postulate (54) as the rigidly ordered universal sequence of clitic shells occurring between the complementizer C0 and the inflectional head I0. (54) C0 d0 num0 n0 p0 or0 loc0 meas0 i0 (where d = definiteness; num = number; n = noun; p = person; or = originator; loc = locative; meas = measure)

The rightmost clitic heads, or(iginator), loc(ative) and meas(ure) are aspectual in the sense of Borer (1994) and Tenny (1994), and represent respectively the (proto-) Agent of the event, the delimitation of the event in the locative sense, and the incremental Theme. It is also assumed that the head meas0 combines two different aspectual features: Extension (path) and Terminus (goal, receiver, final destination), making it compatible with Themes and Goals. The four clitic heads above these are inflectional, consisting of Phifeatures – person (p) and number (num) – , definiteness (d) and categorial information (N). This clitic sequence may be reduplicated within the complementizer system.

Pronominal object markers in Bantu and Romance 

On the basis of their feature content, each clitic head is compatible with certain clitics: meas0 will be lexicalized by 3rd person Patient/Theme or Goal arguments, or by partitives; p0 by 1st and 2nd person clitics; n0 and num0 by 3rd persons (in line with the idea that 3rd persons are default or non-persons, hence incompatible with p0), loc0 by locatives, and d0 by subject expletives and locatives which do not exhibit phifeatures. The bundle of features corresponding to a clitic being compatible with various heads, it is assumed that the specific heads which a clitic may lexicalize is determined parametrically by each language. A clitic will lexicalize the most prominent among its features, modulo the micro-parameters fixed in the grammar of the language. This flexibility allows the authors to account for the limited crosslinguistic variation among clitic groups. The inflectional features are considered semantically uninterpretable, contrary to the aspectual features; consequently, if a clitic occurs under an inflectional head, it needs to attract an aspectual feature to be semantically identified. Take for example, the difference in clitic order between French and Italian illustrated in (55)–(56) (Cocchi 2000: 96, ex. 13–14). In French and in Italian, the 3rd person dative clitic lui/glie is assumed to lexicalize N0 (and attract the aspectual feature +Term(inus) of Meas0), while the 3rd person accusative clitic le/ lo would lexicalize the +Ext(ension) feature of Meas0 in Italian, but it would be parametrically specified as lexicalizing Num0 in French (a head compatible with 3rd person clitics), from where it would attract +Ext. This would account for the difference in clitic sequence between the two languages. (55) c0 d0 num0 n0 p0 or0 loc0 meas0 I0 glie lo dà (Italian) (+Term) (+Ext) (56) c0 d0 num0 n0 p0 or0 loc0 meas0 I0 le lui donne (French) (+Ext) (+Term) gives it to him

Cocchi proposes that Bantu languages where only one object marker is possible are parameterized as allowing only one head to be lexicalized; it is assumed that this head is Meas0, and that Themes and Goals (referred to by Cocchi as Direct Objects – DO – and Indirect Objects – IO) compete to lexicalize this position (compatible with both thematic roles). (57) a.

Mtoto a-na-m-nunu-li-a mwanamke matunda. boy 1.sm-tns-1.IO-buy-appl-I woman fruit The boy buys the woman fruit. (Cocchi 2000: 87, ex. 4a)

b. Mtoto a-na-ki-nunu-a (kitabu). boy 1.sm-tns-1.DO-buy-I (book) The boy buys it (the book). (Cocchi, 2000: 97, ex. 15)

(Swahili)

(Swahili)

 Marie Labelle

(58) c0 d0 num0 n0 p0 or0 loc0 meas0 I0 m ki

nunulia (cf. 56a) nunua (cf. 56b)

Bantu languages allowing more than one object marker make the whole set of clitic heads parametrically available for lexicalization. To account for the order of object markers in the Tshiluba examples in (59)–(60), the direct object (Patient/Theme) tshi is said to lexicalize N0 (and attract +Ext), the 3rd person indirect object (Goal) mu to lexicalize Meas0 (its +Term feature), and the 1st–2nd persons ku to lexicalize P0 (and attract +Term). (59) c0 d0 num0 n0 p0 or0 loc0 meas0 I0 tshi mu (+Ext) (+Term) buys it for him (60) c0 d0 num0 n0 p0 or0 loc0 meas0 I0 tshi ku pa (+Ext)(+Term) give it to you

sumbila

(Tshiluba)

(Tshiluba)

Whether this approach is viable or not, it has the merit of spelling out explicitly the differences and similarities between the clitic groups in the various languages, and of trying to account for the rigid order within clitic sequences. 5. Conclusion In this paper I have shown that neither in French nor in Chicheŵa, a simple account of object markers as lexical valency-absorbing derivational affixes is sufficient to account for the properties and distribution of these elements. Whatever the precise analysis one might want to propose for Bantu and Romance object markers, it is important to realise that their affix-like morphophonological properties should not deter us from studying their syntactic properties. Clitics do have a mixed status. As shown in section 2, elements that have affix-like morphophonological properties may have autonomous syntactic behavior. By focusing the debate on whether clitics or object markers are affixed in the lexicon or not, one runs the risk of losing sight of the characteristics of their distribution. This is why I tend to favor an approach in which clitic features are manipulated by the syntactic component and then spelled out in a morphological component. This approach allows one to focus on their syntactic distribution, without overlooking their morphological characteristics. I do not know if it is possible to develop a unified analysis for object markers in the two language families, but trying to do so may lead to interesting insights into their properties, and eventually to a universal characterization of reduced pronominals. From this perspective, it is worth pursuing.

Pronominal object markers in Bantu and Romance 

References Anderson, S.R. 2005. Aspects of the Theory of Clitics. Oxford: OUP. Auger, J. 1995. Les clitiques pronominaux en français parlé informel: Une approche morphologique. Revue québécoise de linguistique 24(1): 21–60. Baker, M. 1988. Incorporation. A Theory of Grammatical Function Changing. Chicago IL: University of Chicago Press. Blanche-Benvéniste, C. 1975. Recherches en vue d’une théorie de la grammaire française: Essai d’application à la syntaxe des pronoms. Paris: Honoré Champion. Blanche-Benvéniste, C., Deulofeu, J., Stefanini, J. & Van den Eynde, K. 1984. Pronoms et syntaxe: l’Approche pronominale et son application au français. Paris: SELAF (Société d’études linguistiques et anthropologiques de France). Bonet, E. 1991. Morphology after Syntax: Pronominal Clitics in Romance. Ph.D. dissertation, MIT. Borer, H. 1986. The syntax of pronominal clitics. An introduction. In The Syntax of Pronominal Clitics [Syntax and Semantics 19], H. Borer (Ed.), 1–11. Orlando FL: Academic Press. Borer, H. 1994. The projection of arguments. In Functional Projections [University of Massachusetts Occasional Papers in Linguistics 17], E. Benedicto & J. Runner (Eds), 19–47. Amherst MA: GSLA. Bresnan, J. 1997. The emergence of the unmarked pronoun: Chicheŵa pronominals in optimality theory. Berkeley Linguistic Society 23. (Accessible at the LFG archive: http://www-lfg. stanford.edu/archive/archive.html) Bresnan, J. 2001. The emergence of the unmarked pronoun. In Optimality-Theoretic Syntax, G. Legendre, J. Grimshaw & S. Vikner (Eds), 113–142. Cambridge MA: The MIT Press. Bresnan, J. & Mchombo, S. 1987. Topic, pronoun and agreement in Chicheŵa. Language 63: 741–782. Bresnan, J. & Mchombo, S.A. 1995. The lexical integrity principle: Evidence from Bantu. Natural Languages and Linguistic Theory 13(2): 181–252. Chomsky, N. 1981. Lectures on Government and Binding. Dordrecht: Foris. Cocchi, G. 2000. Free clitics and bound affixes: Towards a unitary analysis. In Clitics in Phonology, Morphology and Syntax, B. Gerlach & J. Grijzenhout (Eds), 85–110. Amsterdam : John Benjamins. Duranti, A. 1979. Object clitic pronouns in Bantu and the Topicality Hierarchy. Studies in African Linguistics X: 31–45. Foulet, L. 1928. Petite syntaxe de l’ancien français, édition 1968. Paris: Honoré Champion. Franco, J. 2000. Agreement as a continuum: The case of Spanish pronominal clitics. In Clitic Phenomena in European Languages, F. Beukema & M. den Dikken (Eds), 147–189. Amsterdam: John Benjamins. Gadelii, K.E. 2000. Why are Chomskyan linguists not interested in Bantu languages? Paper presented at the 3rd World Congress of African Linguistics, August 2000, Université du Bénin-Lomé, Togo. Galambos, S.J. 1985. Mechanisms of change in the position of object pronouns: From classical Latin to modern French. In Selected papers from the XIIIth Linguistic Symposium on Romance Languages, L.D. King & C.A. Maley (Eds), 99–116. Amsterdam: John Benjamins. Galet, Y. 1971. L’évolution de l’ordre des mots dans la phrase française de 1600 à 1700. Paris: Presses universitaires de France. Gerlach, B. 2002. Clitics between Syntax and the Lexicon. Amsterdam: John Benjamins. Godard, D. 1992. Extraction out of NP in French. Natural Language and Linguistic Theory 10: 233–277.

 Marie Labelle Grevisse, M. 1993. Le Bon Usage – Grammaire française. Édition refondue par André Goosse, 13th Edn, 1993–1997. Louvain-la-Neuve : DeBoeck-Duculot. Halle, M. & Marantz, A. 1993. Distributed morphology and the pieces of inflection. In The View from Building 20, K. Hale & S.J. Keyser (Eds), 111–176. Cambridge MA: The MIT Press. Halle, M. & Marantz, A. 1994. Some key features of distributed morphology. In MIT Working Papers in Linguistics 21, A. Carnie & H. Harley (Eds), 275–288. Cambridge MA: MIT Department of linguistics and philosophy. Haverkort, M. 1994. Parameters in the syntax of clitics. Proceedings of Berkeley Linguistic Society 20: 234–245. Hirschbühler, P. & Labelle, M. 1994. Changes in verb position in French negative infinitival clauses. Language, Variation and Change 6: 149–178. Hirschbühler, P. & Labelle, M. 2000. Evolving Tobler-Mussafia effects in the placement of French clitics. In New Approaches to Old Problems: Issues in Romance Historical Linguistics. Selected Papers from the LSRL XXIX Parasession [Current issues in linguistic theory 210], S.N. Dworkin & D. Wanner (Eds), 165–182. Amsterdam : John Benjamins. Hirschbühler, P. & Labelle, M. 2003. Residual Tobler-Mussafia effects in French dialects. In Romance Linguistics: Theory and Acquisition, A.M. Pérez-Leroux & Y. Roberge (Eds), 149–164. Amsterdam: John Benjamins. Hirschbühler, P. & Labelle, M. 2004. Enclisis and proclisis of object pronouns at the turn of the 17th century: The speech of the future Louis XIIIth. In Historical Romance Linguistics: Retrospective and Perspectives, D. Arteaga & R. Gess (Eds), 187–208. Amsterdam: John Benjamins. Hyman, L.M. & Duranti, A. 1982. On the object relation in Bantu. In Studies in Transitivity [Syntax and Semantics 15], P. Hopper & S. Thompson (Eds), 217–239. San Diego SA: Academic Press. Johnston, [Sir] H.H. 1919/1922. A Comparative Study of Bantu and Semi-Bantu Languages. 2 vols. Oxford: Clarendon Press. Kayne, R. 1975. French Syntax: The Transformational Cycle. Cambridge MA: The MIT Press. Kimenyi, A. 1980. A Relational Grammar of Kinyarwanda. Berkeley CA: University of California Press. Kok, A. de. 1985. La place du pronom personnel régime conjoint en français. Amsterdam: Rodopi. Labelle, M. 1985. Caractère post-lexical de la cliticisation française. Lingvisticae Investigationes XI(1): 83–96. Labelle, M. 2007. Clausal architecture in Old French. Lingua 117(1): 289–316. Labelle, M. & Hirschbühler, P. 2001a. La position des clitiques par rapport au verbe à l’impératif dans l’évolution du français. Recherches linguistiques de Vincennes 30: 13–38. Labelle, M. & Hirschbühler, P. 2001b. Les clitiques arguments en serbo-croate et dans l’histoire du français. In Clitiques et cliticisation, actes du colloque de Bordeaux, C. Muller, P. de Carvalho, L. Labrune, F. Lambert & K. Ploog (Eds), 109–132. Paris: Honoré Champion. Labelle, M. & Hirschbühler, P. 2005. Changes in clausal organization and the position of clitics in Old French. In Grammaticalization and Parametric Variation, M. Batllori, M.-L. Hernanz, C. Picallo & F. Roca (Eds), 60–71. Oxford: OUP. Manzini, M.R. & Savoia, L.M. 1998. Clitics and auxiliary choice in Italian dialects: Their relevance for the person ergativity split. Recherches linguistiques de Vincennes 27: 115–138. Manzini, M.R. & Savoia, L.M. 2002. Clitics: Lexicalization patterns of the so-called 3rd person dative. Catalan Journal of Linguistics 2002(1): 117–155. Martineau, F. 1990. La montée des clitiques en moyen français: une étude de la syntaxe des constructions infinitives. Ph.D. dissertation, University of Ottawa.

Pronominal object markers in Bantu and Romance 

Martineau, F. 1991. Clitic climbing in infinitival constructions in Middle French. In New Analyses in Romance Linguistics, D. Wanner & D. Kibbee (Eds), 235–252. Amsterdam: John Benjamins. Miller, P. H. 1992. Clitics and Constituents in Phrase Structure Grammar. New York NY: Garland. Miller, P.H. & Sag, I.A. 1997. French clitic movement without clitics or movement. Natural Language and Linguistic Theory 15(3): 575–639. Miller, P.H. & Monachesi, P. 2003. Les pronoms clitiques dans les langues romanes. In Langues romanes, problèmes de la phrase simple, D. Godard (Ed.), 67–123. Paris: Editions du CNRS. Mchombo, S.A. 1993. On the binding of the reflexive and the reciprocal in Chicheŵa. In Theoretical Aspects of Bantu Grammar, S.A. Mchombo (Ed.), 181–207. Stanford CA: CSLI. Mchombo, S. 1997. Contributions of African languages to generative grammar. In African Linguistics at the Crossroads: Papers from Kwaluseni, R.K. Herbert (Ed.), 179–206. Köln: Rüdiger Köppe. Mchombo, S. 2001. Effects of head-marking on constituent order in Chichewa. In Proceedings of the LFG 2001 Conference, M. Butt & T.H. King (eds), CSLI Online Publications, http://cslipublications.stanford.edu/LFG/ Mchombo, S. 2002. Affixes, clitics and Bantu morphosyntax. In Language Universals and Variation, M. Amberber & P. Collins (Eds), 185–210. Westport CT: Prager. Monachesi, P. 1999. A Lexical Approach to Italian Cliticization. Stanford CA: CSLI. Monachesi, P. 2005. The Verbal Complex in Romance. Oxford: OUP. Pearce, E. 1990. Parameters in Old French Syntax: Infinitival Complements. Dordrecht: Kluwer. Petzell, M. 2004. LFG vs transformational theories: A comparison of certain phenomena in certain Bantu languages. Africa & Asia 4: 151–162. Göteborg University: Dept. of Oriental and African languages. Pollock, J.-Y. 1989. Verb movement, Universal Grammar and the structure of IP. Linguistic Inquiry 20(3): 365–424. Roberge, Y. 1990. The Syntactic Recoverability of Null Arguments. Kingston-Montreal: McGillQueen’s University Press. Roberge, Y. & Vinet, M.-T. 1989. La variation dialectale en grammaire universelle. Montréal: Presses de l’Université de Montréal. Sportiche, D. 1995. Clitic constructions. In Phrase Structure and the Lexicon, L. Zaring & J. Rooryck (Eds), 213–276. Dordrecht: Kluwer. Sportiche, D. 1998. Partitions and Atoms of Clause Structure. London: Routledge. Tellier, C. 1991. Licensing Theory and French Parasitic Gaps. Dordrecht: Kluwer. Tenny, C. 1994. Aspectual Roles and the Syntax-Semantics Interface. Dordrecht: Kluwer. Suñer, M. 1988. The role of agreement in clitic doubled constructions. Natural Language and Linguistic Theory 6: 391–434. Tesnière, L. 1959. Éléments de syntaxe structurale. 2nd Edn 1969. Paris: Klincksieck. Valois, D. 1991. The internal syntax of DP. Ph.D. dissertation, UCLA. Zanuttini, R. 1997. Negation and Clausal Structure: A Comparative Study of Romance Languages. Oxford: OUP. Zwicky, A.M. 1977. On Clitics. Bloomington IN: Indiana University Linguistics Club.

The Bantu-Romance connection in verb movement and verbal inflectional morphology Carolyn Harford

Department of English and Communication, Midlands State University

Romance and Bantu languages show a range of contrasting morphological and syntactic properties, ranging from WH extraction strategies and the existence of V2 to directionality of affixation and degree of fusion and agglutination in verbal inflectional morphology. Using Isizulu and Chishona (Bantu) and standard French and Italian (Romance) as examples, this paper correlates these contrasts in terms of a common right-branching Split-INFL structure, and contrasting preferences for movement (STAY, in Optimality Theoretic terms). It is proposed that the same preference which motivates WH extraction strategies in the four languages also motivates the contrast between the relatively prefixal and agglutinative verbal inflectional morphology of Isizulu and Chishona and the suffixal and more fusional verbal inflectional morphology of French and Italian. The conclusions reinforce intuitions that languages belonging to the Romance and Bantu language families are typologically similar.

1. Introduction Recent theorizing concerning syntactic movement has suggested that it is subject to competing Optimality Theoretic constraints.1 One example comes from WHextraction: a constraint Op-Spec (syntactic operators must be in specifier position; Grimshaw 1997; Baković 1998) favors moving operators such as WH interrogatives to specifier position, whereas a second constraint STAY (trace is not allowed; ibid.) favors leaving such operators in situ.2 The V→I (→C) movement associated

. I would like to express my gratitude to the following: Cécile De Cat, Katherine Demuth (as always), João Costa, Advice Viriri, for assistance with the Chishona data, and two anonymous reviewers. . An anonymous reviewer points out that these OT constraints reformulate Huang’s (1982a, 1982b) concepts of movement at S-structure and Logical Form.

 Carolyn Harford

with subject inversion in WH constructions is also subject to these two constraints. Languages vary in terms of ranking of these constraints such that, inter alia, ex situ or in situ strategies for WH constructions, with or without subject inversion, are preferred or dispreferred. Another instance of verb movement is the V→I movement in which verbs incorporate their verbal inflectional morphemes (Baker 1988; Baker et al. 1989). This type of verb movement has generally been conceived of as subject to different motivations: whereas movement of WH interrogatives to SPEC and V→I (→C) movement are motivated to enable interrogatives to c-command their domains, and to fill the head positions of filled SPECs (attributed first to den Besten 1983, see Roberts 2001 for overview), V→I movement has been seen as necessary to ensure that verbs are united with obligatory bound morphemes (Chomsky 1986, 1993; Pollock 1989). This paper explores the possibility of correlating these two instances of movement, comparing data from Bantu and Romance languages. Given the wide range of internal typological variation within these two language families, the range of comparison will be restricted to two from each: Isizulu and Chishona (Bantu) and standard French and Italian (Romance). In spite of this restriction, one of the goals of this paper is to provide some validation for the intuition that there are fundamental similarities in the syntactic and morphological systems of Romance and Bantu languages. The paper compares a variety of contrasting morphological and syntactic properties, ranging from WH extraction strategies and the existence of V2 to directionality of affixation and degree of fusion and agglutination in verbal inflectional morphology. These contrasts are summarized in Table 1. Table 1. Differences between Isizulu/Chishona and French/Italian in the Syntax and Morphology of Verbs Isizulu, Chishona

French, Italian

(1) WH-extraction by clefting

(1) WH-extraction by movement to SPEC

(2) no V2

(2) V2

(3) verbal inflectional morphology is predominantly prefixal

(3) verbal inflectional morphology is suffixal

(4) agrS precedes T (where morphemes (4) T precedes agrS (where morphemes are not fused) not fused) (5) verbal inflectional morphology is less (5) verbal inflectional morphology is more fusional fusional

The paper correlates these contrasts in terms of a common right-branching Split-INFL structure and contrasting preferences for movement; i.e., contrasting

The Bantu-Romance connection in verb movement and verbal inflectional morphology 

rankings of STAY, which, it is suggested, is more highly valued in the two Bantu languages than in the two Romance languages. The overall proposal is that the same ranking which motivates certain WH extraction strategies in the four languages also motivates the contrast between the relatively prefixal and agglutinative verbal inflectional morphology of Isizulu and Chishona and the suffixal and more fusional verbal inflectional morphology of French and Italian. The paper is organized as follows. Sections 1 and 2 compare the formation of WH interrogatives and certain properties of verbal inflectional morphology, respectively, in the four languages. Section 3 proposes a structural analysis of verbal inflectional morphology in the four languages, based on a common rightbranching Split-INFL structure and contrasting preferences for V-movement (STAY). Section 4 correlates these preferences in the four languages with their respective strategies for WH-extraction as described in Section 1. Section 5 suggests an explanation for the relatively greater degree of fusion in verbal inflectional morphology in the two Romance languages. Section 6 concludes and considers theoretical implications.

2. WH interrogatives in Bantu and Romance Ex situ WH questions in French and Italian have been analyzed as involving movement of the interrogative phrase to SPEC/C′ at the left periphery of the clause (Rizzi 1996; Rizzi & Roberts 1996), as illustrated in the following examples from the two languages. (1) Che cosa ha detto il direttore? what thing has said the director ‘What has the director said?’ (2) Quel livre Jean a-t-il lu? which book J. has-he read ‘Which book John has he read?’

(Rizzi 1996: 79; glosses added)

(Rizzi & Roberts 1996: 91; glosses added)

In example (1), the interrogative phrase che cosa “what thing?”, which is thematically the object of the verb detto “said” (in past participle form), has been moved to Spec/C”. In conjunction with the fronted interrogative, the auxiliary verb ha “has” appears in second position, showing the verb second (V2) effect widely attested in Germanic and Romance. The placement of the verb in second position requires inversion of the subject il direttore “the director” to a postverbal position (following the past participle), whereas, in a non-inversion context, it could precede the verb. In the French complex inversion construction in example (2), as analyzed in Rizzi & Roberts (1996), the lexical subject Jean occupies preverbal position,

 Carolyn Harford

whereas it is its corresponding subject pronoun which is inverted and cliticized to the auxiliary verb. Rizzi & Roberts (1996: 102) propose that Jean is left-adjoined to C’, and that a-t-il occupies C0, the V2 position. In contrast to French and Italian, the pattern in Isizulu and Chishona is that ex situ WH questions are formed by fronting the interrogative phrase as the focus of a cleft construction, as illustrated in examples (3) and (4) from Isizulu and Chishona.3 Examples (5) and (6), from Chishona, indicate that fronting of the WH interrogative without clefting is not possible, regardless of the order of subject and verb in the following clause.4 (3) Yi- li- phi i:- hashi e- ngi- zo- ku- li- gibel-a? cop- pro5- which N5- horse rel5- agrs1s- fut- inf- agro5-ride-fv ‘Which is the horse that I shall ride?’ (Doke 1992: 346; glosses added) (Isizulu) (4) Ndi- aní wa- v- a- ón- a va- kadzi? cop-who rel1- agrs 2- pst- see- fv n2-women ‘Who did the women see?’ (lit. ‘It is who whom the women saw?’)

(Chishona)

(5) *Ani va- kadzi v- a- on- a? who n2-women agrs2- pst- see- fv (6) *Ani v- who agrs2-

a- on- a va- kadzi? pst- see- fv N2-women

In examples (3) and (4), the interrogatives liphi i:hashi “this horse” and ani “who” appear as foci of cleft phrases headed by the copulative particles yi and ndi, respectively. These phrases are followed by relative clauses whose heads are the interrogatives. In the Chishona example (4), there is inversion of the subject within

. Extraction of an interrogative without clefting is occasionally found in other Bantu languages, such as Kiswahili and Chicheŵa (Bresnan & Mchombo 1987). At the same time, clefting is possible, and sometimes preferred, in Romance, as exemplified by French que est-ce que tu fais? “What is it that you do?/What are you doing?”. I am indebted to the two anonymous reviewers for drawing my attention to these facts. . Abbreviations in the glosses are as follows: agro: object morpheme, agrs: subject morpheme, ap: associative particle, cop: copulative particle, excl: exclusive, fut: future, fv: final vowel, hab: habitual, imperf: imperfective, inf: infinitive, n: noun prefix, neg: negative, perf: perfective, pro: pronoun, pst: past, rel: relative. Numbers in the glosses refer to noun class numbers according to the Bleek-Meinhof numbering system standardly used for Bantu noun classes, except for 1S, 2S, 1P, 2P, which refer to 1st person singular, 2nd person singular, 1st person plural, 2nd person plural, in both the Bantu and Romance glosses. Acute accent in the Chishona examples indicates high tone; no accent indicates low tone. Isizulu examples are taken from Doke (1992) and follow his practice of not marking tone.

The Bantu-Romance connection in verb movement and verbal inflectional morphology 

the relative clause, which follows the verb which it normally precedes in non-focus contexts. Can this be V2? There are two reasons for thinking that it is not. Firstly, apart from the relative clause illustrated in example (4), subject inversion does not occur in any other context in which V2 would be expected. For example, the appearance of a Topic phrase at the left periphery does not trigger subject inversion, as seen in example (7). (7) Marí y- achó Tawanda ha- á- na- y- ó. 9.money pro9- this T. neg- agrS1- have- pro9- pro ‘This money, Tawanda doesn’t have it.’ (Chishona)

In example (7), the Topic phrase marí yachó “this money” appears at the left periphery of the sentence, followed immediately by the subject Tawanda, followed in turn by the verb haánayó “(s/he) doesn’t have it”, without subject inversion. Secondly, subject inversion in Bantu languages is correlated with monosyllabic relative morphemes, such as the morpheme a which is used to mark object relative clauses in Isizulu and Chishona5 (Harford & Demuth 1999). The monosyllabic relative morpheme cliticizes to the verb to avoid violating a constraint disfavoring monosyllables as independent words (Harford 1999). Subject inversion, as in example (4) above, is a strategy to permit the relative morpheme to appear immediately adjacent to the verb to which it cliticizes. It is unnecessary to assume that it is a consequence of V2 (Harford & Demuth 1999). To summarize at this point, WH-movement to SPEC with concomitant V2 is possible in French and Italian, whereas movement without clefting is impossible in Isizulu and Chishona. The next section describes verbal inflectional morphology and subject pronouns in these languages.

3. Verbal inflectional morphology in Bantu and Romance The four languages under discussion show patterns of subject exponence which are widely attested in their respective language families. For example, verbal inflectional morphemes indicate the person, number, and (in Bantu) the noun class of the subject of the clause. Examples are given in (8a)–(11a).

. The relative morpheme a, which has a cognate in Isizulu, is referred to as possessive a in Fortune 1985, p. 15. See Visser (2007) for description of the same morpheme in Isixhosa, which is closely related to Isizulu.

 Carolyn Harford

(8) a.   ngi- thand- ile6 (Isizulu)   agr1s- love- pst   ‘I loved’ (Doke 1992: 169; glosses added)

agrS - T- V

b. *thand-ile (9) a.   a- no- d- a7 (Chishona)   agr1- hab- love- fv   ‘S/he loves/wants’

agrS - T- V

b. *no-d-a (10) a.   romp- er- ai (French) V - T - agrS   break-fut- agr1s   ‘I will break’ (Alexiadou & Fanselow n.d.: 9; translation added) b. *romp-er8 (11) a.   am- av- o (Italian)   love-pst-agr1s   ‘I loved’ (Giorgi & Pianesi 1997: 69)

V - T - agrS

These subject inflectional morphemes are obligatory in the verb forms in which they occur,9 as indicated in examples (8b)–(11b), which are ungrammatical because the morphemes have been omitted. Also, in these four languages, T morphemes,10 whether prefixal or suffixal, appear immediately adjacent to the verb stem. In Isizulu and Chishona, the subject morpheme appears to the left of the verb stem and any T morpheme; in French and Italian, the subject morpheme appears to the right of V-T. (In certain forms, to be discussed below, T and subject morphemes are fused.) This order is invariant in all four languages. These observations are also illustrated in examples (8a)–(11a).

. Unlike Clements’ (1984) analysis of the morpheme cognate to ile in Gikuyu, I assume that the final e of ile is not a separate morpheme. It does not alternate with any other final vowel and it is standardly reconstructed in Proto-Bantu as part of a single morpheme *ide (Bastin 1983, p. 9; Hyman 1998). See footnote 7. . The final vowel a of a-no-d-a is a morpheme separate from the verb stem, as indicated by the observation that it alternates with the final vowels i and e in negative and subjunctive forms, respectively: ha-a-d-i “s/he doesn’t love/want”, a-d-e “(that) s/he love/want”. It thus contrasts with the final vowel of the past morpheme ide in Isizulu (see footnote 6). . The phonological string romper-(orthographically rompre) is grammatical as an infinitive. . In Bantu languages there are limited exceptions to this generalization in habitual and consecutive forms. . T morphemes may also encode aspect and modality.

The Bantu-Romance connection in verb movement and verbal inflectional morphology 

In Isizulu (8a) and Chishona (9a), the order of morphemes is agrS-T-V, whereas in French (10a) and Italian (11a), the order is the mirror image V-T-agrS. Thirdly, verbal inflectional morphology in French and Italian tends to be more fusional than in Isizulu and Chishona. For example, in the following imperfect indicative paradigm from French, the verb has single inflectional suffixes encoding aspect, person and number. (12) a.

je march-ais I walk- imperf.agr1s ‘I was walking’

b. tu march-ais you (sg.) walk- imperf.agr2s ‘You (sg.) were walking’ c.

il march-ais he walk- imperf.agr3s ‘He was walking’

d. nous march-ons we walk- imperf.agr1s ‘We were walking’ e.

vous march-iez you (pl.) walk- imperf.agr2p ‘You (pl.) were walking’

f.

ils march-aient they walk- imperf.agr3p ‘They were walking’ (Alexiadou & Fanselow 2002: 220 glosses and translations added)

(13) a.

nd- a- tor- a agrS1s- pst- take- fv ‘I took’

In contrast, in a comparable recent past paradigm from Chishona, person/number and tense are encoded in separate morphemes, illustrated as follows.

b. w- a- tor- a agrS2s- pst- take- fv ‘You (sg.) took’ c.

a- ø- tor- a11 agrS3s- pst- take- fv ‘S/he took’

. The tense morpheme a is realized as null following the identical vowel of the subject morpheme a.

 Carolyn Harford

d. t- a- tor- a agrs1p- pst- take- fv ‘We took’ e.

m- a- tor- a agrs2p- pst- take- fv ‘You (pl.) took’

f.

v- a- tor- a agrs3p- pst- take- fv ‘They took’

A restricted instance of fusion in the two Bantu languages may be may be seen in perfective forms in Isizulu, which are examples of the widespread Bantu process of imbrication (Bastin 1983; Hyman 1998, for Isizulu, see Doke 1992 and Botne & Kershner 2000; for Chishona, see Fortune 1949 and Harford 2001). (14) ngi- sa- lele agrS1s- excl- rest.perf ‘I am still resting’ (Doke, Malcolm, Sikakana & Vilikazi 1990: 445)

In example (14), the perfective morpheme (which is identical to the past tense morpheme in example (8a) above) has fused with the verb stem lal “rest”. The properties of verbal inflectional morphology in the four languages as described thus far in this section may be summarized as follows: (15) a. Obligatory verbal inflectional morphology encodes person, number, and noun class (in the Bantu languages). b. T0 is immediately adjacent to the verb stem, suffixal in the Romance languages, prefixal in the Bantu languages. c. agrs appears to the left of T0 in the Bantu languages, to the right of T0 in the Romance languages (where morphemes are not fused). d. Fusion in verbal inflectional morphology in the four languages occurs in suffixes but not prefixes.12

This section concludes the description of the contrasts in WH-extraction strategies (Section 1) and verbal inflectional morphology between Isizulu/Chishona and French /Italian set out in Table 1 above. The next section provides a general account

. An anonymous reviewer suggests that fusional object clitics in Romance languages (Bonet 1995, and references cited there) constitute a counterexample to the generalization stated in the text, that suffixal morphemes in the relevant languages are more prone to fusion than prefixal ones. In the analysis developed in Section 3 below, object clitics would not form part of the functional projection of V″ to which this generalization is intended to apply.

The Bantu-Romance connection in verb movement and verbal inflectional morphology 

of these contrasts in terms of an analysis whose basic components are a common basic Split-INFL structure and contrasting preferences for V-movement. 4. F ormal structure and V-movement in verbal inflectional morphology in Bantu and Romance I propose to account for these contrasting patterns in the four Bantu and Romance languages under discussion in terms of a common basic structure and differences with respect to the possibility of V-movement. The common basic structure is a right-branching Split-INFL structure (Chomsky 1986; Pollock 1989), with separate projections for all, and only, separate morphemes, following the assumptions in Bobaljik & Thráinsson (1998). The order of functional projections preceding V″ is agrs″- T″ (Demuth & Gruber 1995). In terms of this structure, the morphemes of the French and Italian verb forms given in (10a) and (11a) above are structured as in the tree shown in example (16) (The upper morpheme under each node is French, the lower Italian). (16)

s′′ s′

SPEC s0 ai o

T′′ T′

SPEC T0 r av

V′′ SPEC

V′ V0 romp am

In this tree, V-movement takes place according to standard assumptions, under which  V0 moves through each successively higher node and incorporates its morpheme. According to Chomsky (1993), this movement takes place to enable featurechecking between the verb and its inflections to take place. The effect of incorporation is suffixation. As a result, taking Italian as an example, V0 am moves to T0 av, incorporating and suffixing it to produce am-av. In turn, this combination moves to o, which is accordingly incorporated and suffixed to produce am-av-o. The same process produces French romp-r-ai. The resulting structures are given in example (17).

 Carolyn Harford

(17)

s′′ s′

SPEC s0

T′′

rompi-rj-ai

T′

SPEC T0

ami-avj-o

V′′

tj

V′

SPEC

V0

tj

ti ti

Turning to the Bantu languages, the morphemes of the Isizulu example given in (8a) may be arrayed in the same basic structure as shown in the tree in example (18).13 (18)

s′′ s′

SPEC s0 ngi

T′′ T′

SPEC T0 ile

V′′ SPEC

V′ V0 thand

In this tree, the morpheme ile is dominated by T0. V-movement takes place in the same way as for French and Italian, with the difference that it proceeds no farther than T0. Thus, the T0 morpheme ile is incorporated and suffixed to the verb stem thand, as shown in the tree in example (19). . The four languages also have subject pronouns, which encode the same person, number, and gender/noun class features as verbal inflectional morphology, and it might be considered more profitable to compare the Bantu subject morphemes to Romance subject pronouns rather

The Bantu-Romance connection in verb movement and verbal inflectional morphology 

(19)

s′′ s′

SPEC s0 ngi

T′′ T′

SPEC T0

thandi-ile SPEC

V′′ V′ V0 ti

The subject morpheme ngi is left unincorporated and is realized prefixed to the VT combination thand-ile.14 How is the subject morpheme prefixed to V-T without incorporation? I suggest that it is associated with the verb prosodically, not syntactically, following proposals in Myers (1995), in which the Prosodic Phonology of Selkirk (1986) is extended to the verbal inflectional morphology of Chishona (without the Split-INFL tree structure assumed here). In particular, Myers, utilizing Selkirk’s distinction between functional and full (essentially, lexical) categories, proposes that a phonological word bracket is inserted at the right edge of the verb stem, a full category, combining it into a phonological word with the verbal inflections on its left. If this proposal is applied to the Isizulu example as represented in the tree in (19), the result is the labeled bracketing shown in example (20). (Phonological word boundaries are in bold). (20) ]PW ngi]agrS″thand-ile]V″]PW

In this structure, a phonological word is formed from the Isizulu verb stem, with the incorporated T0 morpheme, plus its preceding inflections. The leftmost phonological word bracket is the right bracket of the phonological word preceding the inflected verb, indicating that the verbal inflections following it belong to the same than to verbal inflectional morphology. On the other hand, Romance subject pronouns differ in being generally optional and lacking the same invariant prefixal or suffixal relationship to V, properties which they share with D″. (The French subject pronoun may invert, unlike its Italian equivalent, a contrast which Rizzi & Roberts (1996) attribute to its being a clitic, unlike in Italian.) For these reasons, I prefer to analyze subject morphemes in Isizulu and Chishona in the same theoretical terms as verbal inflectional morphemes in French and Italian; i.e., as heads of functional projections rather than as maximal projections. . I am indebted to an anonymous reviewer for suggesting this line of analysis.

 Carolyn Harford

phonological word as the verb stem. In this way, the observation that the subject morpheme forms part of the same phonological word as the V-T combination may be accounted for without V-movement to AGRs0 and incorporation. For the Chishona example in (9a), I assume an additional projection for the final vowel, labeled Mood″ (nothing hinges on this particular labeling) illustrated as follows. (21)

s′′ s′

SPEC

s0

T′′ T′

a SPEC T0

Mood′′

no SPEC

Mood′

Mood0

V′′ V′

a SPEC

V0 d

In the tree in (21), V-movement proceeds no farther than Mood0, as illustrated in the tree in (22). (22)

s′′ s′

SPEC s0

T′′ T′

a SPEC T0

Mood′′

no SPEC

Mood′

Mood0 di-a SPEC

V′′ V′ V0 ti

The Bantu-Romance connection in verb movement and verbal inflectional morphology 

In this tree, the FV a is suffixed to the verb stem, and the T0 morpheme no and AGRS0 morpheme a are prefixed prosodically, in the order in which they occur as terminal nodes, in accordance with Kayne’s (1995) Linear Correspondence Axiom. This example demonstrates how the mirror image AGRS0-T0-V0 prefixal order in the Bantu languages is produced by prosodic association, as opposed to the V0-T0AGRS0 suffixal order via incorporation in the Romance languages. These analyses of the Isizulu and Chishona examples follow recent research (Julien 2003; Buell 2005) in assuming that prefixation of morphemes in Bantu languages indicates absence of V-movement. To summarize at this point, on the assumption that suffixation of morphemes indicates incorporation whereas prefixation indicates prosodic association, the overall pattern that emerges is that, assuming a common structure for inflected verbs in the four languages, V-movement is preferred in this structure in the two Romance languages, whereas it is dispreferred in the two Bantu languages. An interesting point which also emerges is that, in spite of the disfavoring of V-movement in the Bantu languages, the verb does move leftward by exactly one step, to the next higher head position, and no farther. This pattern suggests that it is desirable for the verb to move to its functional projection, even if it is only to the minimal extent possible to achieve this. One may speculate that this minimal movement is motivated by a need to prevent the verb from being separated from its functional projection by a D″ in the intervening SPEC/V″. This section has proposed an account of the order and directionality of verbal inflectional morphemes in the four languages in terms of a single basic Split-INFL structure and contrasting preferences for V-movement. The next section extends these correlations to the WH-extraction strategies described in Section 1. 5. F ormal structure and V-movement in WH-extraction strategies in Bantu and Romance In the previous section, it was suggested that V in the two Romance languages moves to all head positions in the functional projections of V″, whereas, in the two Bantu languages, V moves only to the head position of the functional projection immediately dominating it. This section suggests that a wider correlation may be established between the favoring, or non-favoring, of movement in the inflected verb and the possibility of WH-extraction strategies based on movement in the four languages. As described in Section 1 above, one possible extraction strategy in the two Romance languages under discussion is movement to SPEC and concomitant V movement to a head, whereas in the two Bantu languages, this type of WH-construction is not possible: an interrogative at the left periphery occupies

 Carolyn Harford

the focus position of a cleft construction, followed by a relative clause. The appearance of a non-clefted interrogative in a left peripheral position has been standardly assumed to involve movement to a specifier position; this assumption is expressed in the Optimality Theoretic constraint Op-Spec (operators appear in specifier position (Grimshaw 1997; Baković 1998). At the same time, concomitant V-movement has also been assumed to be to the head of the maximal projection of this specifier (den Besten 1983; Roberts 1993; Vikner 1995, and references cited there). Whichever maximal projection is involved (C″ or one of the articulated projections of I), there are two components of this extraction strategy which involve movement: the operator to SPEC, and V to the head position. On the other hand, the clefting strategy offers fewer instances of movement. In particular, the structure of the cleft may be assumed to be as in example (23), in which the copulative morpheme in the two Bantu languages is structured as the head of a maximal projection, with the interrogative as its complement. (The upper morpheme under each node is Isizulu, the lower Chishona). (23)

COP′′ SPEC

COP′ COP0 Yi Ndi

N′′ SPEC

N′ N0 liphi ani

According to usual assumptions, A′-movement would not take place to the complement position of a head, such as the N0 position in the tree in (23). If this is the case, then clefting contrasts with movement to SPEC in not involving movement at all. The requirement for clefting in the Bantu languages may then be seen as a strategy for minimizing movement. It thus dovetails with the pattern shown in the inflected verb, which also minimizes instances of movement.15 In contrast, the

. This is not the only explanation for the existence of the clefting strategy in Bantu and need not be seen as ruling out any others. Another possibility is the functionally based explanation offered in Bresnan & Mchombo (1987) and Demuth & Johnson (1989). There may also be a need to distance the interrogative from the pronominal subject morpheme to prevent binding problems. All of these may be seen as exerting pressure in favor of one type of strategy in preference to another.

The Bantu-Romance connection in verb movement and verbal inflectional morphology 

two Romance languages reveal a pattern of tolerating movement, or favoring it, in WH-extraction and the inflected verb. This section has proposed that the conclusion reached in Section 3 (that the inflected verb in the two Romance languages shows movement to a greater extent than its equivalent in the two Bantu languages) may be extended to include patterns of movement, or the lack of it, in WH-extraction strategies in the same languages. The next section suggests an account of fusion in verbal inflectional morphology in the four languages.

6. Fusion in verbal inflectional morphology in Bantu and Romance In Section 2 above, it was proposed that suffixes in verbal inflectional morphology in the four languages may be fusional, but prefixes may not be. This section proposes an account of this difference. According to the Split-INFL Hypothesis, and also Grimshaw’s (1991) account of functional and lexical projections, functional projections embed lexical projections in a right-branching structure, and thus material in functional projections precedes material in lexical projections in the structural order of terminal nodes. Prefixal verbal inflectional morphology in Bantu languages reflects this structure in terms of linear order more transparently than in Romance languages, which are, to the best of my knowledge, strictly suffixal. In Romance, then, to a greater degree than in Bantu, the order and positioning of morphemes is inverse to the basic phrase structure. I propose that the greater degree of fusion in Romance verbal inflections is a response to this mismatch; an attempt to cover it up by reducing the number of separate morphemes whose linear order is the reverse of the basic structural order. This idea suggests that fusion may be an effect of Kayne’s (1995) Linear Correspondence Axiom. This idea is also reflected in the relatively rare instances of suffixal verbal inflectional morphology in Bantu, where there is a greater degree of fusion, such as in the process of imbrication (Bastin 1983; Hyman 1998; Botne & Kershner 2000) illustrated in example (14) above from Isizulu, repeated here as example (24). (24) ngi- sa- lele agr1s- excl- rest.perf ‘I am still resting’ (Doke, Malcolm, Sikakana & Vilakazi 1990: 445)

(Isizulu)

In example (24), the perfective morpheme (which is identical to the past tense morpheme in example (8a) above) has fused with the verb -lal- “rest”. Imbrication in suffixal verbal inflectional morphology thus contrasts with prefixal agglutinative morphology in Bantu. I suggest that this contrast arises from the same source

 Carolyn Harford

as fusion in Romance verbal suffixes: namely, the preference for minimizing the number of separate inflectional suffixes, as opposed to prefixes. This section has proposed that there is a correlation between suffixal and relatively fusional morphology and prefixal and relatively agglutinative morphology, with Romance as an example of the former and Bantu as an example of the latter. It has been argued that, since suffixal verbal inflectional morphology reverses the basic structural order of syntactic projections, in which functional projections precede lexical ones, fusion is favored as a strategy to reduce the number of separate verbal inflectional morphemes following the verb stem. The next section summarizes and considers theoretical implications.

7. Summary and theoretical implications In summary, this paper has examined a number of morphological and syntactic properties of WH-extraction constructions and verbal inflectional morphology in two Bantu languages, Isizulu and Chishona, and two Romance languages, standard French and Italian. These properties have been analyzed in terms of a common right-branching Split-INFL structure, in which the lexical projection of V is dominated by functional projections of inflectional morphemes. Within this structure, the languages representing these two language families show differing preferences or tolerances for movement. It has also been suggested that the favoring or disfavoring of movement covers the degree of movement in both WH-extraction constructions and the inflected verb. Furthermore, the relatively higher degree of fusion in verbal inflectional morphology in the Romance languages has been attributed to a preference for minimizing the number of suffixal inflectional morphemes, whose postverbal directionality clashes with the basic order of functional and lexical morphemes indicated by the basic structure. There are also a number of theoretical implications. Firstly, these conclusions reinforce the idea that the theory of Head-to-Head movement (Travis 1984; Baker 1988) may be used to account for the directionality and relative fusion of verbal inflectional morphology, in addition to the data it was originally devised to account for. Secondly, they suggest that the favoring or disfavoring of movement, as expressed in relative rankings for the Optimality Theoretic STAY, cut across syntactic and morphological constructions. Finally, these conclusions about four Bantu and Romance languages suggest that intuitions about typological similarities between languages belonging to these two families are not misplaced.

The Bantu-Romance connection in verb movement and verbal inflectional morphology 

References Alexiadou, A. & Fanselow, G. 2002. On the correlation between morphology and syntax: The case of V-to-I. In Studies in Comparative Germanic Syntax, J.-W. Swart & W. Abraham (Eds.), Amsterdam: John Benjamins. Baker, M. 1988. Incorporation: A Theory of Grammatical Function Changing. Chicago IL: The University of Chicago Press. Baker, M.C., Johnson, K. & Roberts, I. 1989. Passive arguments raised. Linguistic Inquiry 20: 219–252. Baković, E. 1998. Optimality and inversion in Spanish. In Is the Best Good Enough? Optimality and Competition in Syntax, P. Barbosa, D. Fox, P. Hagstrom, M. McGinnis & D. Pesetsky (Eds), 35–58, Cambridge MA: The MIT Press. Bastin, Y. 1983. La finale–IDE et l’imbrication en bantou [Annales, Série IN-8, Sciences Humaines, No. 114]. Tervuren: Musée Royal de l’Afrique Centrale. Besten, H. den. 1983. On the interaction of root transformations and lexical deletive rules. In On the formal syntax of the Westgermania, W. Abraham (Ed.), 47–131. Amsterdam: John Benjamins. Bobaljik, J.D. & Thráinsson, H. 1998. Two heads aren’t always better than one. Syntax 1(1): 37–71. Bonet, E. 1995. Feature structure of Romance clitics. Natural Language and Linguistic Theory 13(4): 607–47. Botne, R. & Kershner, T.L. 2000. Time, tense and the perfect in Zulu. Afrika und Übersee 83(2): 161–180. Bresnan, J. & Mchombo, S. 1987. Topic, pronoun and agreement in Chichewa. Language 63(4): 741–782. Buell, L. 2005. Issues in Zulu Verbal Morphosyntax. Ph.D. dissertation, University of CaliforniaLos Angeles. Chomsky, N. 1993. A minimalist program for linguistic theory. In The View from Building 20, Essays in Linguistics in Honor of Sylvain Bromberger, K. Hale & S.J. Keyser (Eds), 1–52. Cambridge MA: The MIT Press. Chomsky, N. 1986. Barriers. Cambridge MA: The MIT Press. Clements, G.N. 1984. Principles of tone assignment in Kikuyu. In Autosegmental Studies in Bantu Tone, G.N. Clements & J. Goldsmith (Eds), 281–347. Dordrecht: Foris. Demuth, K. & Gruber, J. 1995. Constraining XP-Sequences. In Niger-Congo Syntax and Semantics No. 6, V. Manfredi & K. Reynolds (Eds), 3–30. Boston MA: Boston University African Studies Center. Demuth, K. & Johnson, M. 1989. Interaction between discourse functions and agreement in Setawana. Journal of African Languages and Linguistics 11: 21–35. Doke, C.M. 1992. Textbook of Zulu Grammar, 6th Edn. Cape Town: Maskew Miller Longman. Doke, C.M., Malcolm, D.M., Sikakana, J.M.A. & Vilakazi, B.W. 1990. English-Zulu, Zulu-English Dictionary. Johannesburg: Witwatersrand University Press. Fortune, G. 1985. Shona Grammatical Constructions, Vol. 1, 3rd Edn. Harare: Mercury Press. Fortune, G. 1949. The conjugation of inchoative verbs in Shona. African Studies: 132–140. Giorgi, A. & Fabio, P. 1997. Tense and Aspect: From Semantics to Morphosyntax. Oxford: OUP. Grimshaw, J. 1991. Extended Projection. Ms. Brandeis University. Grimshaw, J. 1997. Projection, heads, and optimality. Linguistic Inquiry 28(3): 373–422.

 Carolyn Harford Harford, C. 2001. Perfective stems in Shona. Zambezia 30(ii): 190–203. Harford, C. 1999. Disyllabification in Shona and the prosody-syntax interface in optimality theory. Malilime: Journal of Malawian Linguistics 1: 22–40. Harford, C. & Demuth, K. 1999. Prosody outranks syntax: An optimality approach to subject inversion in Bantu relatives. Linguistic Analysis 29(1–2), S. Mchombo (Ed.), 47–68. Huang, C.-T. J. 1982a. Move wh in a language without wh-movement. The Linguistic Review 1: 369-416. Huang, C.-T. J. 1982b. Logical Relations in Chinese and the Theory of Grammar. Ph.D. Dissertation, Massachusetts Institute of Technology. Hyman, L. 1998. Positional prominence and the ‘prosodic trough’ in Yaka. Phonology 15(1): 41–75. Julien, M. 2003. Syntactic Heads and Word Formation. Oxford: OUP. Kayne, R. 1995. The Antisymmetry of Syntax. Cambridge MA: The MIT Press. Myers, S. 1995. The phonological word in Shona. In Bantu Phonology and Morphology, F. Katamba (Ed.), 69–92. München: Lincom. Pollock, J.Y. 1989. Verb movement, Universal Grammar and the structure of IP. Linguistic Inquiry 20(3): 365–424. Rizzi, L. 1996. Residual verb second and the wh-criterion. In Parameters and Functional Heads: Essays in Comparative Syntax, A. Belletti & L. Rizzi (Eds), 63–90, Oxford: OUP. Rizzi, L. & Roberts, I. 1996. Complex inversion in French. In Parameters and Functional Heads: Essays in Comparative Syntax, A. Belletti & L. Rizzi (Eds), 91–116. Oxford: OUP. Roberts, I. 2001. Verb movement and markedness. In Language Creation and Language Change, Creolization, Diachrony and Development, M. DeGraff (Ed.), 287–327. Cambridge MA: The MIT Press. Roberts, I. 1993. Verbs and Diachronic Syntax: A Comparative History of French and English. Dordrecht: Kluwer. Selkirk, E. 1986. On derived domains in sentence phonology. Phonology Yearbook 3: 371–405. Travis, L. 1984. Parameters and Effects of Word Order Variation. Ph.D. dissertation, MIT. Vikner, S. 1995. Verb Movement and Expletive Subjects in Germanic Languages. Oxford: OUP. Visser, M. 2007. DP structure in Isixhosa: Definiteness properties and the occurrence of the noun class preprefix. Paper presented at The Bantu-Romance Connection Workshop, An Exploratory Workshop sponsored by the European Science Foundation, University of Leeds, UK.

part 2

The structure of DPs

DP in Bantu and Romance* Vicki Carstens

University of Missouri-Columbia

The paper considers several aspects of Bantu and Romance DPs, including: (i) Noun Class and grammatical gender; (ii) apparently derivational properties of the two; (iii) ordering among nouns and their modifiers; and (iv) concord in DPs. Several conclusions are argued for. Firstly, Bantu Class is a gender system like that of Romance, with gender-specific Spell-Out of number features. Secondly, despite some surface evidence to the contrary, gender/Class is an uninterpretable feature, without derivational functions. Thirdly, DPs of the two languages share a common architecture; and fourth, in both families concord is the result of the Agree relation. Thus many properties of nouns and DPs are common to both language groups, as the hypothesis of UG leads us to expect.

1. Introduction Languages of the Bantu and Romance families exhibit striking similarities, including several points of correspondence within the DP: members of both groups partition nouns into genders or classes; categories within DP agree with the head noun in this feature and in number; and movement places both Bantu and Romance nouns to the left of their base positions. This paper compares and contrasts these aspects of DP morpho-syntax in the two families, delineating both underlying unities and points of true divergence. I begin in 2.1 with an analysis of Bantu Noun Class as a gender system very much like that of Romance. The remainder of section 2 explores some apparently derivational functions of Noun Class that lack precise parallels in Romance. I show that, contrary to appearances, neither Noun Classes nor their prefixes are derivational; rather there are derivational zero morphemes in Bantu with their own gender specifications. These are kin to the larger, phrase-level null element pro found in both Bantu and Romance, and like pro they are subject to a recoverability or identification requirement. Bantu has a greater number and diversity of * My thanks to Josephat Rugemalira for helpful discussion of the Swahili data in section 3 of this paper.

 Vicki Carstens

null elements than Romance does because its larger number of genders enables them to be identified unambiguously and hence to be recoverable. This is a modest difference, consistent with the assumption that the underlying grammatical principles at work are very much the same. Section 3 compares the syntax of Romance and Bantu DPs. I argue that the architecture is shared, and that both groups of languages exhibit noun raising, though they differ in the landing sites of nouns. I consider and reject the antisymmetric approach to post-nominal modifier order presented in Cinque (2005) in favor of symmetric base-generation of modifiers. Section 4 shows that in both groups of languages the Agree relation (cf. Chomsky 2000, 2001) can readily account for the agreement of categories within the DP with the head noun (often labeled concord). Section 5 concludes that all aspects of DPs in the two languages are indeed highly similar: a welcome finding, supportive of the theory of Universal Grammar. For the sake of brevity I rely as much as possible on representative examples from just two languages: Swahili for illustrations of Bantu phenomena, and Spanish for illustrations of Romance phenomena. Because the paper deals with broad questions this is generally sufficient. 2. Words and morphemes 2.1 Noun Class as grammatical gender As is well-known, Romance languages generally partition nouns into two grammatical genders, masculine and feminine. Most nouns bear a suffixal wordmarker whose shape correlates fairly consistently with the gender of the noun (see 1).1 In Spanish, a plural suffix may follow it, as in (2). Nouns denoting humans are distributed among the two genders on the basis of the sex of their referents (see 3): (1) a.

cas-a house-F ‘house’

[Spanish]

. Harris (1991) argues convincingly that Spanish word markers should not be viewed as gender morphology; rather, their distribution is determined by redundancy rules in which gender plays a limited role. I abstract away from this and gloss the markers -a and -o as F(eminine) and M(asculine) respectively, as a convenience, while assuming that gender is a lexical property of nouns. Also, I will not discuss the presence of neuter gender in some Romance languages.

DP in Bantu and Romance 

b. libr-o book-M ‘book’ (2) a.

cas-a-s house-F-PL ‘houses’

b. libr-o-s book-M-PL ‘books’ (3) a.

chic-a child-F ‘girl’

b. hij-o offspring-M ‘son’

Bantu nouns are similarly partitioned, though there are several points of contrast. Firstly, natural or sex gender is not a factor in determining the Classes of nouns; rather, nouns denoting humans or, in many Bantu languages, animates, constitute a single Class. Secondly, Bantu Noun Classes are indicated by prefixes; and thirdly, they greatly outnumber the Romance genders, especially as the Classes are conceptualized in the Bantuist linguistic tradition. I provide representative examples from Swahili in (4). The numbers of the Noun Classes are traditional ones, based on Meinhoff ’s reconstructions for proto-Bantu. Classes 12 and 13 of Proto-Bantu are absent in Swahili;2 on locative Classes 16–18, see section 2.4.

(4) Noun Classes: Swahili

Class

Example

Gloss

1 2 3 4 5 6 7

m-tu wa-tu m-ti mi-ti gari ma-gari ki-atu

person people tree trees car cars shoe

. Bantu languages differ as to which and how many of the proto-Bantu classes are preserved. There are also language-particular deviations from the pairings 1/2, 3/4, etc.: Swahili nouns in Class 11 take their plurals in Class 10, proto-Bantu Class 12 being absent; a reviewer points out that in Sesotho some Class 1 stems take plurals in Class 6; some Class 5 ones take plurals in both Class 6 and 10; and so forth.

 Vicki Carstens

8 9 10 11 14 15 16 17 18

vi-atu n-yumba n-yumba u-bao u-kweli ku-soma N+loc. suffix N+loc. suffix N+loc. suffix

shoes house houses board truth to read; reading specific place general place inside place

It is clear that pairs of prefixes for many Classes function as singular/plural morphology for particular groups of stems. For example every noun stem in Swahili which takes the Class 1 singular prefix also takes the Class 2 prefix in the plural (see 5); every noun which takes the Class 7 singular prefix takes the Class 8 prefix in the plural (see 6); and so on, through Class 10. Nouns in Class 11 are exceptional to a very modest degree, in that their plurals are in Class 10, like the plurals of Class 9 nouns. (5) a.

mtoto/watoto 1child/2child ‘child/ren’

[Swahili]

b. mjinga/wajinga 1fool/2fool ‘fool/s’ c.

msichana/wasichana 1girl/2girl ‘girl/s’

(6) a.

kiatu/viatu 7shoe/8shoe ‘shoe/s’

b. kikombe/vikombe 7cup/8cup ‘cup/s’ c.

kikapu/vikapu 7basket/8basket ‘basket/s’

It is a weakness of the traditional system depicted in (4) that the regular relationships among pairs of Classes are not formally reflected; though it has the virtue of familiarity and conserves some space in glosses, the system is theoretically uninsightful. This is demonstrated in (7) through its hypothetical extension to a

DP in Bantu and Romance 

Romance language; I replace the gender and number designations of the Spanish nouns from (1) and (2) with Bantu-style Class numbers. A feminine noun like casa “house” or chica “girl” is analyzed as belonging to two different classes, 1 and 2, depending on whether it is singular or plural; similarly for libro “book” or hijo “son”, whose singulars are Class 3 and plurals Class 4. We lose the insight that there are two groups of stems, not four: (7) a.

casa b. house1 ‘house’

chica girl1 ‘girl’

c.

casas d. house2 ‘houses’

chicas girl2 ‘girls’

(8) a.

libro b. book3 ‘book’

hijo offspring3 ‘son’

c.

libros d. book4 ‘books’

hijos offspring4 ‘sons’

[Spanish]

Because plurality has a consistent realization as the suffix -s in the Spanish data, the approach is transparently misguided. In Bantu, on the other hand, the realization of plural (and singular) varies by gender. But this is a difference in surface detail, not in underlying categories or concepts.3 I accordingly follow Carstens (1991) in analyzing Noun Class as a gender system, along the lines of Romance (see also Corbett 1991). This proposal reduces Noun Classes 1–11 to six genders, as shown in (9).4,5 Prefixes are gender-specific number morphology, which I assume to be added to nouns by late Spell-Out rules . Italian plurals are rather like Bantu Class-marked nouns in that the morphology of an Italian plural noun correlates fairly predictably with the noun’s grammatical gender, and there is no separate word marker for gender on plurals: casa/case house/s(f), pizza/pizze pizza/s(f); versus ragazzo/ragazzi boy/s(m), tempo/tempi time/s(m). It is nonetheless standard to analyze interpretable number as a feature distinct from gender, the latter a lexical property of Italian nouns. My analysis extends this reasoning to Bantu. . Though I adopt this analysis, my glosses will continue to follow the traditional numbering system for consistency with Bantuist practice. . Though I use the term stem to designate the form which Noun Class prefixes attach to, I assume that gender is a lexical property of noun roots. This follows from the fact that the gender is largely arbitrary; it must be part of the lexical listing, since it cannot be predicted (see discussion of 18).

 Vicki Carstens

(cf. Halle & Marantz 1993).6 The approach is neutral on the question of phrasal projections for inflectional features; I will argue in 2.2.1 that no such projection is motivated for gender; in contrast, interpretable number features may head functional projections (see section 3).

(9) Bantu Genders (Carstens 1991)

Gender A: stems of Classes 1/2 Gender B: stems of Classes 3/4 Gender C: stems of Classes 5/6 Gender D: stems of Classes 7/8 Gender E: stems of Classes 9/10 Gender F: stems of Classes 11/10

(10) Sample Spell-Out rules yielding Swahili Noun Class Prefixes7 [Singular] 〈–〉 /m- / /__N Gender A [Singular] 〈–〉 /ki- / /__N Gender D [Plural] 〈–〉 /vi- / /__N Gender D [Plural] 〈–〉 /wa-/ /__N Gender A

In contrast, a Spell-Out rule for singular number in Spanish is superfluous, since it would consistently add Ø; and since the realization of plural is not gender-specific in Spanish, a single Spell-Out rule suffices for all regular plurals: (11) [Plural] 〈–〉 /-s /

/N__

Spell-Out for the Spanish plural suffix

Summing up, I have sketched out a treatment of Noun Class in line with the following general conclusions: (12) a. Bantu and Romance both have grammatical gender. b. Bantu has a greater number of genders than Romance. c. Bantu expresses number in gender-particular prefixes, while Romance concatenates markers of gender and number as suffixes.

. An alternative is to suppose that gender and number are two distinct morphemes underlyingly; but in Bantu, gender and number undergo the operation fusion (Halle & Marantz 1993). Since I am assuming gender is strictly a lexical property of nouns I leave this possibility aside. . In Gender D, all forms of agreement are homophonous with the prefixes on nouns (save for some phonologically-conditioned allomorphy). This is not the case in all genders; for example, in Gender A singular, subject agreement is a- rather than m, and the Class morphology for demonstratives is a third form, yu-. Category-specific Spell-Out rules like these are therefore needed.

DP in Bantu and Romance 

d. Animacy or humanness has a gender correlation in Bantu languages; biological sexes have such correlates in the genders of Romance.

The differences between Romance gender and Bantu Noun Class are fairly trivial under this approach, and this is a desirable conclusion, consistent with the view that deep unities underlie surface variation among languages. But for the analysis of Noun Class as gender to succeed, Classes 14–18 must be given a satisfactory account, as must a set of derivational phenomena associated with certain Classes. I turn to these matters next. 2.2 Noun Class and derivation 2.2.1 Representing Class Obscuring the simplicity of the Bantu system and the unity of Bantu and Romance are some much-noted alternations in which Noun Class prefixes seem to play a derivational role. Observe that replacing a noun’s usual Class prefixes with those of Classes 7 and 8 yields diminutive meanings in Swahili; switching to those of Classes 5 and 6 produces augmentative meanings (augmentatives are also formed in Classes 3/4, not exemplified here): (13) a.

msumari/misumari b. 3nail/4nail ‘nail/s’

c.

sumari/masumari 5nail/6nail ‘big (ugly/nasty) nail/s’

kisumari/visumari 7nail/8nail ‘little nail/s’

[Swahili]

When the prefixes are replaced in this way, agreement with the noun also shifts, reflecting the Class of the replacement prefix.8 (14) a.

Msumari mmoja u-me-anguka. 3nail one 3agr-perf-fall ‘One nail has fallen’.

[Swahili]

b. Kisumari kimoja ki-me-anguka. 7nail 7one 7agr-perf-fall ‘One little nail has fallen’. c.

Masumari mawili ya-me-anguka. 6nail 6big 6agr-perf-fall ‘Two big nails have fallen’.

. In Class 3 (Gender B Singular), though the Noun Class prefix is m-, the subject agreement prefix is u-. See note 7.

 Vicki Carstens

Two conclusions have been drawn from this state of affairs. Firstly, many researchers have argued from such facts as (13) that Noun Class is entirely a property of the Noun Class prefixes, which are heads, selecting N(P)s (cf. Sproat 1985; Myers 1987; Bresnan & Mchombo 1995). Thus the representation of a noun is as shown in (15), in these works: β (Class 3)

(15) mClass 3

(β and α = N or NP)

α sumari nail

Secondly, alternations like (14) are taken as evidence that Noun Class is at least partially derivational in nature (Mufwene 1980; Myers 1987; and Ferrari 2005). In what follows I will argue that there are empirical reasons to reject both of these views. Consider first the status of stems with respect to Class information. Although they are treated as unmarked for Class in Sproat (1985), Myers (1987), Bresnan & Mchombo (1995), this cannot be correct. We must assume that stems are specified for Class (in the framework adopted here, for gender) to account for combinatorial restrictions like (16) and for the default pattern of singular and plural formation for each noun. Thus something like (9) seems to be unavoidable. (16) a. *n-tu C9-person ‘person’ b. *mi-atu C3-shoe ‘shoe’

One might solve this problem by specifying the stems for gender while still maintaining the analysis of Noun Class prefixes as selecting heads. But this creates an undesirably redundant system, since each member of a singular/plural prefix pair must be specified for identical selectional features (see 17): m- wa-

[_+αGender A] [_+αGender A]

denotes singular denotes plural

b. m- mi-

[_+αGender B] [_+αGender B]

denotes singular denotes plural

c.

[_+αGender C] [_+αGender C]

denotes singular denotes plural

(17) a.

Ø ma-

DP in Bantu and Romance 

d. ki- vi- etc.

[_+αGender D] [_+αGender D]

denotes singular denotes singular

An alternative to considering the specific prefixes to be selecting heads might be to suppose that gender as a general category is a head selecting nominal complements, or perhaps that each of the Bantu genders proposed in (9) is a head selecting stems with matching gender specifications (see 18). The prefixes themselves would spell out singular or plural on a gender-specific basis, at the point where number features are added, as shown in (10). (18)

GenP Gen° α gender

N or NP N α gender

This proposal too has serious drawbacks. Chomsky (1995) proposes that uninterpretable features must be deleted from the representations sent to the semantic interface, and that it is therefore unlikely that they head syntactic categories, introducing structure which must subsequently be eliminated. Though the specific concern of his proposal is agreement features, the line of reasoning extends naturally to gender, arguing against a syntactic representation like (18), if gender is an uninterpretable φ-feature as agreement features are. There are good reasons to think that gender is, in fact, uninterpretable. Despite some pockets of semantic unity in the groups of nouns assigned to one gender or another, membership in a grammatical gender is recognized to be, on balance, arbitrary. There is no semantic reason why, for example, the Swahili noun kikapu “basket” should belong to Gender D, while the noun ndoo “bucket” is Gender E; or why the Italian patata “potato” should be feminine while pomodoro “tomato” is masculine.9 Even where there is some semantic consistency to gender, as in the assignment of animates to Swahili Gender A or females to Romance feminine gender, this does not entail that the gender itself has any semantic content; rather, some semantic properties of the nouns seem to serve as sorting criteria, similar to what is involved in a household decision to keep plates on one shelf and glasses on another. No one considers that essential properties of plates and glasses are intrinsic to the shelves on which we choose to store them; similarly, threads of semantic unity in grammatical genders do not indicate that the genders themselves have semantic content.

. My thanks to an anonymous reviewer for pointing these examples out.

 Vicki Carstens

These considerations argue in favor of viewing gender as an uninterpretable feature of nouns, and hence against adopting (18). It has often been claimed that consistent meanings like diminutive and augmentative are properties of specific Bantu Noun Classes and/or their prefixes. If this should turn out to be true, (18) might yet be a possibility. I will argue in 2.2.2 that these meanings are instead properties of zero-morphemes of particular genders. I conclude that gender is not a selecting head, and should not have a projection of its own. 2.2.2 The zero-affixation analysis In this section I will provide two arguments that diminutives and augmentatives are formed by a zero-affixation process distinct from the attachment of Noun Class prefixes, contra Sproat (1985); Myers (1987); Bresnan & Mchombo (1995); Mufwene (1980); and Ferrari (2005). The first argument is based on a minimum word size requirement and an epenthetic noun prefix ji-. Lexical categories in Swahili must consist of at least two moras: (19) Swahili Two Mora Constraint: *[N µn], n < 2 (cf. Carstens 1991; Park 1997; Njogu 1994)

Noun class prefixes enable many monomoraic stems to meet this requirement: (20) a.

mtu/watu b. 1person/2person ‘person/people’

c.

mti/miti 3tree/4tree ‘tree/s’

kiti/viti 7chair/8chair ‘chair/s’

[Swahili]

Class 5 (= the singular of Gender C in 9) generally lacks any prefix (see 21). Monomoraic stems in this Class surface bearing the prefix ji-(see 22), absent on polymoraic Class 5 nouns. I accordingly analyze ji- as an epenthetic element with the function of “rescuing” stems which would otherwise violate the Two Mora Constraint: (21) a.

bega/mabega 5shoulder/6shoulder ‘shoulder/s’

b. tawi/matawi 5branch/6branch ‘branch/es’

[Swahili]

(22) a.

DP in Bantu and Romance 

jicho/macho 5eye/6eye ‘eye/s’

b. jiwe/mawe 5stone/6stone ‘stone/s’ (23) ji-insertion: Prefix ji- to a noun which fails to satisfy the Two Mora Constraint.10

Ji-epenthesis interacts with derivation in revealing ways. Recall that diminutive nouns belong to Gender D, and thus bear ki/vi prefixes; augmentatives belong either to Gender C (Classes 5 and 6) or Gender B (Classes 3 and 4) of which all but the singular Class 5 have overt Noun Class Prefixes. Yet any monomoraic stem made diminutive or augmentative must bear ji- as a kind of “inner” prefix, despite the presence of the Noun Class prefixes of its derived Class: (24) a.

kijicho/vijicho b. *kicho/vicho 7eye/8eye 7eye/8eye ‘envy; jealous glance/s’ ‘envy; jealous glance/s’ (lit: little eye/s)

[Swahili]

b. mjicho/mijicho d.  *mcho/micho 3/eye/4eye 3eye/4eye ‘big eye/s’ ‘big eye/s’

As Carstens (1991) points out, two assumptions allow us to make sense of this distribution of ji-: (a) the Two Mora Constraint and ji-epenthesis apply in derived environments;11 and (b) the derivational processes forming diminutives and augmentatives are zero-affixation processes distinct from the attachment of Noun Class prefixes. I assume that they involve silent morphemes with their own semantics and gender specifications, as shown in (25); since overt derivational affixes are

. Ji-epenthesis is also employed in two other contexts. The first is in the formation of superdiminutives and super-augmentatives, a fact that Carstens (1991) relates to the Zero-Morpheme licensing requirement (see 28), since a doubly zero-affixed root needs a means to identify the outer zero morpheme and distinguish it from the inner one. Secondly, there is a constraint prohibiting vowel-initial nouns and adjectives that can be satisfied by ji-epenthesis. See Carstens (1991: 219–238) for details. . In the Optimality framework there are no derivations; hence “derived environment” effects are explained without reference to derivations. See Lubowicz (2002); McCarthy (2003) and references cited there. I use the traditional term as an expository convenience.

 Vicki Carstens

suffixes, in Swahili, I treat the zero-affixes as suffixes also.12,13 The diminutive affix selects the noun stem and is therefore head of the derived word, determining its gender specification: N  D

(25)

N

Af

sumari nail  B

Ø DIMIN  D

Under these assumptions, a noun stem made diminutive or augmentative is subject to the Two Mora Constraint because zero-affixation creates a derived environment. On the other hand, the monomoraic roots in (21), which bear no derivational affixes, are assessed for compliance with the Two Mora Constraint only when the addition of singular or plural Noun Class prefixes creates a derived environment. The Two Mora Constraint therefore applies only to the prefixed form in such cases, and is satisfied without ji-insertion. An analysis equating diminutive and augmentative formation with the prefixes of the derived forms would be hard-pressed to account for these facts, since the prefixes ki-, vi-, m- and mi- and ma- do not otherwise require that the stems they attach to meet the Two Mora Constraint (see 20b,c).14 One could only maintain

. An anonymous reviewer points out that diminutives are formed with overt suffixes in some Bantu languages. . It has been argued that words have underlying (Spec)-head-complement structure and order (see Kayne 1994; Roeper 1999; Di Sciullo 2005 for various implementations of this idea). Applied to the case under consideration, this would mean that the diminutive affix is Merged to the left of the stem, and the stem reaches its surface position by leftwards movement. This interesting possibility has no obvious bearing on the issues I address here (the nature of Class, and its non-involvement in derivation) so I will leave it aside. . In the reconstructed proto-Bantu, diminutives belonged to special Classes, usually 12/13, and this system persists in many of the modern languages including Kikuyu. Note that Kikuyu diminutive formation is also subject to the 2 Mora minimum word requirement; a monomoraic stem retains the prefix of its lexical class as an inner prefix, while most polymoraic stems do not. Thus the zero-morpheme analysis is motivated for a language with Class 12 diminutives: (i) a.

ka-iriitu < 12-girl ‘little girl’

mu-iriitu 1-girl ‘girl’

b.

ka-mu-ti < 12–3-tree ‘stick’

mu-ti 3-tree ‘tree’

DP in Bantu and Romance 

the approach by supposing that diminutives and augmentatives are formed, rather redundantly, by means of pairs of prefixes which happen to be identical and homophonous to existing singular and plural prefixes, but which differ from them in having the semantics [diminutive singular] and [diminutive plural], [augmentative singular] and [augmentative plural]; these prefixes also happen, exceptionally, to require that the stems they attach to meet the constraint. This is neither economical nor advantageous in any other way that I can think of, so I reject it in favor of the simpler and more elegant zero-affixation account. I turn to the second argument that diminutive and augmentative formation processes are distinct from their associated genders and prefixes. Though ordinary nouns of Classes 7/8 bear ki-/vi- prefixes like diminutives nouns, they do not necessarily denote small things (see 26 a,b), nor must those of Classes 5/6 denote large things (see 26c,d). It follows that the diminutive and augmentative semantics do not reside in the relevant genders or their associated prefixes, contra Mufwene (1980) and Ferrari (2005). The analysis in (25) is consistent with this state of affairs. (26) a. Kiingereza b. 7English ‘English language’

kiapo 7oath ‘oath’ (deverbal)

c. maneno d. 6word ‘words’

matawi 6branch ‘branches’

[Swahili]

In fact, ordinary nouns of Classes 7/8 are not even ambiguous between neutral and diminutive readings, nor are those of Classes 5/6 ambiguous between neutral and augmentative readings (see 27):15 (27) a. kitabu/vitabu b. gari/magari 7book/8book 5car/6car ‘book/s’ ‘car/s’    *‘little book/s’    *‘big car/s’

[Swahili]

There is evidence that minimum word requirements do not apply cyclically in all languages (see discussion of Navajo in Carstens 1991: 239–244). It would therefore not be surprising to find Bantu languages in which zero-affixation of the diminutive morpheme did not lead to epenthesis in the stem; this would not indicate that the zero-affixation approach is incorrect for such languages. . There are ways of getting around this problem. Many Class 5/6 nouns form augmentatives in Classes 3 and 4; the epenthetic “inner prefix” morpheme ji- salvages otherwise impossible diminutives and some of the blocked augmentatives. Thus while kitabu ‘book’ has no diminutive reading, kijitabu “little book” is fine. Carstens (1991: 223–238) broadens the ji-epenthesis rule so it is triggered by violations of nominal well-formedness conditions generally, including the zero-morpheme licensing principle proposed below.

 Vicki Carstens

I follow Carstens (1991) in relating these facts to independently motivated licensing requirements on zero-morphemes (cf. Pesetsky 1990), which can in turn be related to general recoverability requirements. The augmentative and diminutive morphemes are subject to (28): 16 (28) Zero-morpheme licensing principle: Zero-morphemes must be identified.

Pesetsky (1990) argues that inflection and certain derivational morphemes can identify zero-morphemes in English; Carstens (1991) proposes that the requirement is satisfied for Bantu zero-morphemes by their introduction of a gender specification that contrasts with that of the root, as in (29a):17 (29) a. ok [[sumari Gender B] Ø Gender D] – ‘little nail’ b. * [[tabu Gender D] Ø Gender D] – *‘little book’ c. Zero-morpheme identification condition: a zero-morpheme is identified if its features are unambiguously represented on the derived word.

I will propose in section 2.5 that this requirement is kin to the well-known licensing requirement on pro. In contrast, it is not at all clear how the facts in (26) and (27) can be accounted for under the assumption that Noun Classes or their prefixes have a derivational function.

. One might alternatively conceive of the derivation of Swahili diminutives and augmentatives as a rule or process adding the relevant gender feature, rather than as a morpheme. But since, as noted in the text, Genders C and D do not have diminutive and augmentative meanings, their addition to nouns cannot suffice to indicate the derived noun’s altered semantics; there must additionally be a specification as diminutive or augmentative introduced: (ii)

a. b.

Diminutive formation: Nα gender → N-dimin Gender D Augmentative formation: Nα gender → N-augmentGender C

I take these specifications to be indistinguishable from zero-morphemes; thus the analyses are notational variants. . An appealing alternative idea is that the Noun Class prefixes themselves identify the zeromorphemes. But this approach falls short of explaining the facts in (iii); only a gender feature distinct from that of the stem suffices: (iii) a. ok

ki-[ tabu Gender D] 7- book ‘book’

b.

ki-[[tabu Gender D] 7- book ‘little book’

*

Ø Gender D] dimin

DP in Bantu and Romance 

2.3 Discussion I have argued that Bantu Noun Class prefixes are gender-specific number morphology, and that derivational properties of the Classes and their prefixes are apparent rather than real. Diminutives and augmentatives in Bantu are zero-affixation processes, where the zero-affixes have gender specifications. The Swahili diminutive morpheme differs from those found in Romance languages in that it is silent, and in that Romance diminutive morphemes are unspecified for gender. Under the morphological percolation conventions of Selkirk (1982), a derived word inherits the features of its head and any non-clashing features of its non-head. Thus, absent a competing gender feature in the affix, a Spanish noun made diminutive inherits the gender of its non-head member, the stem (see 30):18,19 (30) a.

el carr-o b. the car (masc) ‘the car’

c.

la caj-a d. la caj-it-a the box (fem) the box-dimin (fem) ‘the box’ ‘the little box’

e.

el carr-it-o the car-dimin (masc) ‘the little car’

[Spanish]

N [α gender] Ν [α gender]



Bantu diminutives are more like those of German, which are always neuter (see 31), or Faroese, where diminutives are uniformly masculine (32). We can assume that the German diminutive morpheme has a neuter specification which is inherited . The proposed percolation from non-heads of complex words has no correlate in Minimalist syntax, as an anonymous reviewer points out. Since the final vowel of the noun in Romance correlates fairly regularly with gender, an alternative account comes to mind, namely that the gender feature of Romance diminutives is added in the outermost layer, in the suffix, and this is why the diminutive affix itself has no impact on the gender of the derived form. But the suffix and the nominal root must match for gender specification, a fact that leads to the same conclusion I have argued for: that when the diminutive is affixed to a nominal root, the gender of the latter determines the gender of the derived form. Hence the diminutive affix itself must lack a gender feature, and the stem’s gender percolates. . An anonymous reviewer points out that gender variation with a single root in some Romance languages can be related to size dimension (i.e., Spanish barca-barco ‘small boat-boat’). Yet, both forms can also take diminutive-augmentative suffixes (barquita/barcaza; barquito/barcote). The suffixes are represented as in (30e). An account of connections between gender and semantics is sketched out in (38).

 Vicki Carstens

by the derived noun, while that of Faroese has a feminine feature. (32) illustrates schematically diminutives of Faroese, German, and Bantu: (31) a.

die maus b. the mouse (fem) ‘the mouse’

das maus- chen [German] the mouse- dimin (neuter) ‘the little mouse’

c.

der hase d. the hare (masc) ‘the hare’

das has- chen the hare- dimin (neuter) ‘the little hare’

(32) a.

ketta b. cat (fem) ‘cat’

kett-lingur cat- dimin (masc) ‘kitten’

(33)

N [α gender]

[Faroese]

Bantu, German, and Faroese diminutives

Ν  [β gender] [α gender]

The Bantu pattern of diminutive formation thus differs from that of Romance in several ways. But under this line of analysis for derivational phenomena associated with Bantu Noun Classes, some otherwise exotic-seeming functions of Class dissolve, leaving us with an understanding of Class as a simple gender system which differs from that of Romance only in minor details.20 My approach to Bantu diminutives and augmentatives can be extended to abstract noun formation (34) and gerund formation (35), though these processes change categories; we need only assume that the zero morpheme head of the derived noun contributes a nominal categorial feature, as many derivational morphemes do (cf. destroy/destruction; happy/happiness, etc.). Abstract nouns and gerunds have nominal morphology and agreement paradigms distinct from other Classes, so I posit two additional genders for them, Genders G and H: (34) a. c.

kweli b. u- kweli true 14- true ‘true’  ‘truth’

[Swahili]

N  G Adj

AfN

kweli true

Ø  G

. Ferrari (2005) argues that Romance has derivational uses of gender like those she posits for Bantu. I consider these briefly below.

DP in Bantu and Romance 

(35) a. c.

soma b. read ‘readV’

ku-soma 15-read ‘readingN’

N  H V

AfN

soma read

Ø  H

There are, in Romance languages, some correlations of semantics and gender which Ferrari (2005) cites in evidence of her claim that gender is semantic and derivational. As previously noted, biological gender of the referent correlates with grammatical gender in Romance nouns referring to humans: (36) a.

la niña b. the.fem child.fem ‘the girl’

el niño the.masc boy.masc ‘the boy’

c.

la señora d. el señor the.fem woman.fem the.masc man.masc ‘the woman’ ‘the man’

[Spanish]

Ferrari (2005) describes a number of additional correlations of grammatical gender and semantic feature: there is a count vs. mass association for masculine vs. neuter gender in Ripano; and several patterns of semantic correlation to gender choice in Cantabrian (Ferrari 2005: 39–44), among them male/female, dark/light, coarse/smooth, vertical/horizontal, narrow/wide. Ferrari also provides examples of several systematic gender-meaning correspondences in Italian, which I reproduce below:21 (37)

a. ferramento ‘iron tool’ ~ ferramenta ‘hardware’ [count/mass] b. granolo ‘grain’ ~ granola ‘the sifted whole wheat grains’ [count/mass] c. raccolto ‘harvest/crop’ ~ raccolta ‘harvesting, collection’ [± collective] d. scarico ‘unloading of a weapon’ ~ scarica ‘a volley of rifle-fire’

e.

[±collective]

taglio ‘cut’ ~ taglia ‘ransom, tally, body size’ [concrete/abstract]

. Ferrari follows Contini-Morava (2000) in using the term “animate” to designate “entities with vitality”; that is, having to do with life or its sustenance; hence granito “stone” is inanimate but the grated-ice drink granita is animate because it is a food and food is life-sustaining. I find the over-arching semantic classifications unconvincing and it is not clear to me, if these components of nominal semantics reside in the genders, how it happens that a food item like pomodoro “tomato” is masculine.

 Vicki Carstens

f. g. h. i. j. k. l. m. n. o. p.

rancio ‘ration’~ rancia ‘the distribution of the ‘ration’ [concrete/abstract] mestolo ‘small ladle’ ~ mestola ‘big ladle’ [small/big] buco ‘small hole’ ~ buca ‘pit or big hole’ [small/big] pozzo ‘well’ ~ pozza ‘puddle’ [small/big] granito ‘granite’~granita ‘grated-ice drink’ [inanimate- animate (food)] ciuccio ‘pacifier’~ciuccia ‘breast’ [inanimate –animate (body part)] melo ‘apple tree’ ~ mela ‘apple’ [inanimate – animate (food)] lucciolo ‘sequin’ ~ lucciola ‘fire-fly’ [inanimate-animate (insect)] girello ‘small disk’ ~ girella ‘spinning wheel’ [small/big] ninfeo ‘temple of pagan gods’ ~ ninfea ‘a kind of plant’ [+/-animate] terrazzo ‘balcony’ ~ terrazza ‘terrace’ [small/big]

The co-existence of different strands of meaning within the same gender (for masculine: count or male or small or concrete or inanimate) presents a challenge to analyzing genders as contributing any meaning. Ferrari addresses this by proposing that semantic properties of the gender interact with those of the root to yield the correct meaning. I believe there is a simple and very plausible interpretation of how root and gender interact that is consistent with the status of gender as an uninterpretable inflectional feature. An epicene noun (borrowing Ferrari’s term) can be viewed as one whose lexical meaning contains an open variable (underspecification, represented as Ø in (38) to avoid confusion with uninterpretability, often represented as u). Each such variable ranges over a pair of oppositions out of a fixed and finite set: male/female; mass/count; small/large; concrete/abstract; and so forth on a language-particular basis. The variability of such nouns is grammatically encoded in underspecification for gender (also represented as Ø). A set of redundancy rules maps the nouns to genders, according to the value given to the semantic variable upon lexical selection: (38) Redundancy rules assigning gender to epicene nouns Root

Gender mapping rule

NØsex Øgender

→ →

NØcount/mass → Øgender

feminine if female22 masculine if male masculine if count feminine if mass

. This rule is much like Harris’s (1991) Human Cloning Rule, which operates on any noun with the semantic specification “human” and no specification for sex. Human Cloning turns such nouns into pairs of nouns specified ‘male’ and ‘female’. Then a redundancy rule ‘female’ → f/[__human] maps female sex to feminine grammatical gender.

DP in Bantu and Romance 

NØsize → Øgender

masculine if small feminine if large

etc.

Under this approach, the genders themselves contribute no meaning. It is therefore quite natural that the collection of semantic features in (38) which map to a given gender is an arbitrary set, just as the assignment of ordinary nouns to genders is a semantically arbitrary assignment. Before leaving this topic, let me summarize the significant contrasts between Romance epicene nouns on the one hand and Bantu formation of diminutives, augmentatives, abstract nouns and gerunds on the other. Romance epicene nouns are a specific, finite set of nouns. While common themes are observable, the semantics involved in gender choice are determined by the roots themselves. In contrast, the Bantu processes are regular, productive derivational processes in which specific, predictable meanings can be linked to specific affixes. The analyses I have provided reflect these differences. 2.4 Locatives A final aspect of Bantu Noun Class which distinguishes it from Romance gender is the phenomenon of Locative Noun Classes. Many Bantu languages have three locative Classes, with characteristic agreement and Noun Class prefixes. I illustrate with Chichewa: (39) a.

Mu-nyumba mu-ku-nunkha. 18–9house 18agr-asp-stink-fv ‘Inside the house stinks’

[Chichewa]

b. Ku-nyumba ‘ku ndi ku-tali 17–9-lake 17dem cop 17agr-far ‘That house and its environs are far away’ c.

Pa-nyumba pa-ku-on-ek-a ngati pa-ku-psy-a 16–9house 16 agr-asp-see-stat-fv like 16agr-asp-burn- fv ‘The house and surrounding yard look like they’re burning’

I present here a sketch of the properties of the locative classes, and show that while they have no counterpart in the system of Romance genders, their syntax crucially involves sub-phrasal gaps which occur in Romance as well. In both language families the gaps can be licensed by “rich” agreement, indicating that the grammatical principles underlying the two systems are truly the same. The structure of locative phrases in Bantu is revealed by the agreement and interpretation of modifiers within them. In many languages, including Chichewa

 Vicki Carstens

and Shona, modifiers of locativized nouns can agree in either the locative Class or the Class of the noun stem. Though Myers (1987) asserts that there are no semantic correlates to this choice, Carstens (1997) shows otherwise. Items bearing locative agreement are semantically linked to the locative content of the phrase, unlike those bearing non-locative agreement. Thus in (40a), where the quantificational adjective -ri -onse “every” agrees with chiseko “door”, the meaning is “every door”; in (40b), where -ri -onse bears locative agreement, it is interpreted as quantifying over places on the door, so the meaning is “every part of the door, or “all over the door”: (40) a.

pa-li nchenche pa-chiseko chirichonse. 16-be 10fly 16–7door 7every ‘there are flies on every door’

[Chichewa]

b. pa-li nchenche pa-chiseko pariponse. 16-be 10fly 16–7door 16every ‘there are flies all over the door’

These facts argue that in addition to the overt noun stem within locative phrases there is also a sort of “place” noun present, which takes modifiers of its own, and controls agreement on them. Bresnan and Mchombo (1995) analyze locative prefixes themselves as nouns, given that they seem to control agreement as nouns do, and, based on evidence like conjoinability, that they are independent words rather than true prefixes (see 41): (41) Ndi-na-jámbúl-a zithunzi mu ndí pa madengu. 1sg-sb-rec.pst-draw-ind 8picture 18 and 16 6basket ‘I drew pictures in and on baskets.’

[Chichewa]

Carstens (1997) argues against the analysis of locative prefixes as nouns, since the prefixes violate the Two Mora Constraint to which nouns are subject, and fail to trigger “of ” insertion before their nominal complements (otherwise obligatory for complements of nouns). Functional categories are exempt from the Two Mora Constraint, so locative prefixes pattern as functional elements in this regard; “of ” insertion does not apply between a Case-“assigning” category and its complement; thus the locative prefixes behave like Case-“assigners”. Carstens concludes that locative prefixes are gender-specific Case-markers, each selected by one of three silent “place” nouns of distinct genders.23 The silent locative noun adjoins to D as . An anonymous reviewer points out that Case-markers might reasonably be viewed as having only uninterpretable content; thus by the logic of section 2.1, KPs could be argued to lack syntactic legitimacy, like agrPs. I accordingly assume that they have some prepositional semantic content (like ‘in’ or ‘at’) compatible with the meanings of the null nouns they combine with.

DP in Bantu and Romance 

is standard in Bantu (see section 3). Word order indicates that the Case-marker and its complement (KP in 42) raise to the left edge of the locative DP, thus to its Spec. In this position the richly agreeing preposition licenses the empty noun by making its content recoverable (see 43): (42)

DP

Spec

D'

KP K

D DP

N

pa chiseko [16 place] 16 7door ‘all over the door’

NP D

AP

NP

pariponse 16all

tN tKP

(43) An empty locative nominal is licensed by “rich” agreement

(Carstens 1997)

2.5 Discussion While Romance languages do not have locative genders, they do have welldocumented sub-phrasal gaps which are disallowed in English, and which therefore seem to be connected with the availability of “rich” agreement on nearby categories: (44) Me gusta el /la [e] de Juan 1sg.dat please the.masc /the.fem of *’I like the of Juan’

[Spanish]

(45) Q: Que libro quieres? Which book want.2S ‘Which book do you want?’

[Spanish]

A: El grande [e]. the.masc big *‘The big’.

It seems likely that these gaps, like the null Bantu “place” nouns, are available at least in part because agreement makes their content recoverable.24 I follow Carstens . An anonymous reviewer points out that agreement does not suffice in every case, since *Me gusta el ‘I like the’ is ill-formed. Contreras (1989) attributes this to a clitic property of the determiner.

 Vicki Carstens

(1997) in concluding that Bantu place nouns and these Romance gaps are subphrasal counterparts to pro; assuming that zero-morphemes are sub-word-level pro, the whole set of phenomena described in the section can receive a unified treatment; see the generalized licensing requirement in (46). (46) The φ-features of a silent nominal element must be identified.

Rich agreement (or, in the case of zero morphemes, an identifying feature of a “rich” gender system) licenses silent elements at all levels of structure. Bantu has more gender distinctions than Romance, and hence can identify unambiguously a greater number of null elements. 3. Word order and phrase structure in DP In this section I turn to DP-internal syntax, showing that Bantu and Romance DPs have a shared architecture and many common properties. Nouns move higher in Bantu than in Romance, however. In both languages, modifiers can be adjoined to the left or to the right of N’s base position, yielding variability in post-nominal modifier order. 3.1 The noun-raising approach Bantu languages generally lack overt articles. In a DP containing a lexical possessor, agent, or theme argument in an “of ” phrase, standard word order is N-AP-ofP:25 In contrast to lexical arguments, genitive pronouns are positioned between the noun and any APs: (47) a.

picha nzuri ya Halima 9picture 9nice 9of ‘Halima’s nice picture’

b. picha yake 9picture 9his/her ‘her nice picture’

[Swahili]

nzuri 9nice

Carstens (1991, 1993, 1997) follows Ritter’s (1988, 1992) analysis of Hebrew in deriving Bantu N-initial order via N-raising to D. Genitive pronouns occupy Spec of a functional category in the DPs middle field, perhaps NumberP. Updating the . Judgments vary on DPs containing two lexical arguments in ‘of ’Ps. Though several speakers I have consulted accept them, others find them degraded (cf. Rugemalira p.c.). Pursuit of this topic lies outside this paper’s scope.

DP in Bantu and Romance 

analysis a bit, I assume that agents originate in Spec of “little” n, the counterpart to v* of transitive clauses. Possessors are arguments of a different variety of n; themes are complements to N, and adjective phrases adjoin to nP. (48) [DP picha+D [NumP yake t Num [nP nzuri tposs [NP tN]]] picture her nice

In contrast, Romance languages have overt articles. Both articles and genitive pronouns precede the noun, though adjectives generally follow (see 49a,b). Researchers of the early 1990s argued that Romance nouns raise as far as a functional category between D and N, perhaps NumP (see Cinque 1995; Valois 1991). Updating their proposals slightly we have (49b):26 (49) a.

la casa bonita the.fem house(f) pretty.fem ‘the pretty house’

[Spanish]

b. mi casa my.fem car(f) ‘my house’ c. [DP (la)D [NumP mi casa [nP bonita tposs [NP tN]]]] the my house pretty

These analyses of Bantu and Romance yield a simple and elegant account of their differing word orders, paralleling the classical analysis of sentence-level word order differences in terms of the presence or absence of V-movement to I and beyond. They are based on the assumption that elements are Merged in the same hierarchical positions universally; word order variation results from simple and constrained movements. 3.2 Modifier order in Cinque (2005) Recent developments have led to a more complex picture of DP-syntax. It is wellknown that languages in which modifiers precede the noun exhibit consistent ordering among the modifiers, while those with nouns left-peripheral in DP often have either the same order or its mirror image. These generalizations are expressed in Greenberg’s Universal 20, reproduced below (cf. Greenberg 1963).

. The facts are more complex than this brief sketch suggests. In French and Spanish, articles and genitive pronouns cannot co-occur. In Italian, many adjectives are pre-nominal. While these points of variation are significant, space considerations preclude exploration here. See Cinque (1995), Valois (1991) for in-depth discussion of the Italian and French facts respectively.

 Vicki Carstens

(50) Greenberg’s Universal 20: When any or all of the items (demonstrative, numeral, and descriptive adjective) precede the noun, they are always found in that order. If they follow the noun the order is either the same or its exact opposite.

Universal 20 has proved to be too strong in that other post-nominal orders exist (see Cinque 2005 for extensive discussion). But the claimed inflexibility of pre-nominal modifiers has proved accurate. The contrast between pre- and post-nominal modifiers in this regard, and the various post-nominal orders, present a challenge addressed in much current research on noun phrase syntax. Cinque (2000, 2005) argues that to obtain all and only the licit orders, it is necessary to assume that the pre-nominal order is universally base-generated: (51) Universal Merge Order: Dem > Numeral > Adjective > N

Cinque observes that if noun-raising accounted for all variation in word order, and if head-positions intervened between modifiers, we would expect only the following orders: (52)

a.

Dem Numeral Adj N

b.

N Dem Numeral Adj t (complete N-raising)

c.

Dem N Numeral Adj t

d.

Dem Numeral N Adj t (N raises 1 level)

(no movement)

(N raises 2 levels)

This leaves unaccounted for the common mirror-image order, and 9 additional possibilities which are attested (cf. Cinque 2005). Cinque accordingly adopts a different approach, arguing that there is no head-movement at all in grammar. A complex array of leftwards NP-movements and roll-up movements produces the attested orders among nouns and their modifiers (Cinque 2005: 321–324). With this much established I turn to modifier order within Bantu DPs. Rugemalira (2006) reports that Swahili allows the orders Noun-Demonstrative and Demonstrative-Noun, with free word order among the modifiers to the right of these items (53a–d adapted from Rugemalira 2006: 12–13). The order NumeralAdjective in (53a) and (53c) is consistent with (52), but the order Adjective-Numeral

DP in Bantu and Romance 

in (53b,d) is a mirror-image order (though a partial one, since the demonstrative is not final): (53) a.

wale 2those

watu wawili wazuri 2person 2two 2good

[Dem-N-Num-Adj]

b. wale 2those

watu wazuri wawili 2person 2good 2two

[Dem-N-Adj-Num]

c.

watu wale 2person 2those

wawili wazuri 2two 2good

[N-Dem-Num-Adj]

d. watu wale wazuri wawili 2person 2those 2good 2two ‘those two good people’

[N-Dem-Adj-Num]

If we apply the derivations in Cinque (2005) to the Swahili orders in (53), here is how they work. The Merge order is “those two good people”. Two surface orders are produced by NP moving all the way up (see 54), or partially, ending just to the left of the numeral (55) (I bracket only moved constituents): (54) [NP people] those two good tNP

(= 53c; complete NP-movement)

(55) those [NP people] two good tNP

(= 53a; partial NP-movement)

The orders in (53b) and (53d) are derived by a sequence of movements. First, the NP raises to the left of the adjective. It does not surface in this position, however; the projection containing NP, adjective and the NP’s trace (= XP in (56)) subsequently moves leftwards across the numeral: (56) a. those two [NP people] good tNP b. those [XP people good tNP] two tXP

→ (= 53b)

(53d) follows up on this derivation with one final step, moving the NP alone again, this time across the demonstrative: (57) [NP people] those [XP tNP good tNP] two tXP

(= 53d)

Cinque acknowledges that motivation for the movements he posits remains to be found. 3.3 An alternative In what follows I will sketch out an alternative approach to these phenomena that is more in keeping with surface appearances. I adopt the hypotheses in (58): (58) a. Spec, head, complement are Merged, always in the same hierarchical and left-to-right arrangement (cf. Kayne 1994).

 Vicki Carstens

b. Modifiers are Adjoined, always in the same hierarchical arrangement whether on the left or right edge of the constituent they modify.27

Universal Hierarchy (adapted from 51): Dem > Numeral > Adjective > N

c.

Movement is always to the left.

(58) maintains a traditional distinction between the outputs of the Merge operation on the one hand, and an Adjunction operation adding modifiers on the other; in the terminology of Chomsky (2000), these are the Set-Merge and PairMerge processes respectively. Under (58), syntax is weakly antisymmetric, in the sense of Takano (2003): Spec-head-complement material conforms to some version of the Linear Correspondence Axiom of Kayne (1994), mapping hierarchical relations into left-to-right order;28 but adjuncts are distinct from specifiers, contra Cinque (1999, 2005), and can appear to the left or to the right edge of the constituent they modify. Hypothesis (58b) thus reflects the variability of modifier order in DPs, while positing hierarchical rigidity among modifiers. As for (58c), we may attribute it to movement’s status as a variety of Set-Merge (cf. Takano op cit). (59) sketches out representations for two of the acceptable Swahili word orders (53c,d). APs are adjuncts to nP, the nominal counterpart to vP. I leave open the identity of XP, the projection modified by demonstratives; for arguments that demonstratives are adjuncts below DP in Swahili, see Carstens (1991). I take numerals to be universally adjuncts to Num(ber)P; in Swahili, they have the flexibility to be adjoined either to the right or to the left (59a versus 59b).29 Head-movement applies cyclically to move the noun to the left edge of DP:30

. This would mean that the LCA of Kayne (1994) does not apply to adjuncts. I comment below. . There have been several alternatives to the original LCA proposed in the syntactic literature; I will not explore these here. Representations which follow are not intended to conform to the LCA in all technical details, but assume that Spec-head-complement order of Merged material is universal. . In their concordial prefixes, numerals pattern with adjectives. I leave their category an open question. . See Matushansky (2006) for an analysis of head-movement as raising to a Spec of the root, followed by a morphological merger process (m-merger).

DP in Bantu and Romance 

(59) a.

DP

= (53c), N-Dem-Numeral-Adj

D X Num n

XP D DemP

X

wale those

XP 〈X〉

NumP

Num

YP

wawili 〈Num〉 two

n

N

NumP

watu people

nP nP

AP wazuri 〈n〉 good

NP 〈watu〉

b.

DP D X Num n

N watu people

n

XP XP

D DemP X

Num

= (53d), N-Dem-Adj-Numeral

wale those

〈X〉

NumP YP

NumP 〈Num〉

nP

AP wazuri 〈n〉 good

wawili nP NP 〈watu〉

If a language permits demonstratives to adjoin on the right edge of XP, this will add two additional orders to the four described. Rugemalira (2006: 14) describes these orders as alternatives available in closely-related Safwa.

 Vicki Carstens

(53a,b) differ from (59a,b) respectively only in the appearance of the demonstrative in initial position. I follow Carstens (1991) in analyzing Bantu initial demonstratives as appearing in Spec, DP. Ashton (1944) states that Swahili prenominal demonstratives are semantically different from their post-nominal counterparts, close in interpretation to definite determiners, picking out something aforementioned (Ashton 1944: 59 & 181–183). While it is in principle possible that they are slightly different lexical items with different Merge positions, there is a more intriguing interpretation that does not rely on a proliferation of homophonous lexical items. Just as definite DPs tend to move to high positions (cf. Diesing 1992), raising a definite demonstrative to Spec, DP plausibly correlates with a definite interpretation for it. (60)

DemP

D' D

wale those X Num n N

= (53a), Dem-N-Numeral-Adj

DP

n

watu people

D DemP X

Num

XP XP

〈wale〉 〈X〉 those

NumP YP

NumP

wawili 〈Num〉 two

nP

AP wazuri 〈n〉 good

nP NP 〈watu〉

3.4 Discussion Symmetric base-generation of modifiers and some basic leftward movements account simply for the ordering possibilities among nouns and their modifiers in Swahili. These proposals fit into a constrained model of grammar: the ban on rightwards movement and the universal hierarchy governing adjunction proposed in (58b) restrict the possibilities to only those which are found in the languages of the world (see Cinque 2005: 319–20). Assuming also that movement respects

DP in Bantu and Romance 

Relativized Minimality (Rizzi 1990) or its minimalist successor Attract Closest, the universal hierarchy is consistently maintained. Cinque describes fourteen attested orderings of the elements Noun, Demonstrative, Adjective, and Numeral. Assuming that head positions intervene between each modifier, where nouns might surface in different languages, twelve word orders are generated by the combination of symmetric modifier adjunction possibilities and noun-movement. Only two orderings in the languages of the world have no obvious account in terms of variable adjunction direction and head-movement: [Adj-N-Dem-Numeral] and [N-Adj-Dem-Numeral]. These would seem best accounted for by raising the NP containing Adjective and Noun as Cinque proposes, in a kind of predicate-fronting comparable to the clauselevel operation of VP-raising. 3.5 Romance DPs Adjectives in Romance are often found post-nominally, and the same issues arise in connection with their relative ordering. Laenzlinger (2005) reports that postnominal APs in French exhibit either the sequence in (61a) or mirror-image order (62a). Laenzlinger, following Cinque’s approach, concludes that French sometimes moves the smallest NP ([NP chaise] in 60b), but can also do roll-up movement as shown in (62) (moved constituents bracketed): (61) a. b.

[French]

une ____ ronde rouge [chaise]

(62) a.

une voiture rouge2 splendide1 vs. a beautiful1 red2 car une splendide ______ rouge [voiture]

b.

[French]

c.

une chaise ronde1 rouge2 a chair round red

une ______ splendide [voiture rouge t]

As in the case of Bantu, an alternative analysis exists; it seems in principle possible that splendide, though hierarchically superior to rouge, can be adjoined to the right. Future research will no doubt uncover compelling reasons to reject one of these approaches to word order variation in grammar. This review of the theoretical issues and the Bantu and Romance facts illustrates how closely parallel

 Vicki Carstens

are the structures and processes of these genetically unrelated groups of languages. We can be confident, I think, that a single theoretical model will account for them all.

4. Agreement in DP The final topic of this comparative paper is the phenomenon of agreement or concord among the elements within DP. In both Bantu and Romance languages, modifiers and arguments in DP inflect for the gender and number features of the head noun: (63) a.

kikombe changu cheupe 7cup 7my 7white ‘my white cup’

b. vikombe vyangu 8cup 8my ‘my white cups’ (64) a.

[Swahili]

vyeupe 8white

mtoto huyu mdogo 1child 1this 1small ‘this small child’

b. watoto hawa wadogo 2child 2this 2small ‘these small children’ (65) a.

el niño pequeño the.masc child.masc small.masc ‘the small boy’

b. la niña the.fem child.fem ‘the small girl’

[Spanish]

pequeña small.fem

In the framework of Chomsky (1986) and (1992), agreement was analyzed as a Spec-head relation (see also Koopman 1992). Thus subject agreement occurs because the features of a DP in Spec, TP are shared with the head T(ense). (66) a. [[Spec TP a man] T+be in the room] b. [[Spec TP a man] T+be+3Sagr in the room] c. A man is in the room

This approach did not readily extend to concord in DP. Based on the Bantu pattern of concord Carstens (1991) accordingly argued that agreement could be licensed either

DP in Bantu and Romance 

through Spec-head feature-sharing or through feature-sharing in the government relation, the latter yielding concord. Such a disjunctive approach is theoretically undesirable, but seemed unavoidable. But in Minimalist theory, agreement is licensed in the Agree relation (Chomsky 2000, 2001) which is based on closest c-command. The agreement features (the probe) must c-command the category they agree with (the goal): (67) a. Agree (α,β) iff α c-commands β; α,β have matching features; there is no γ with matching features which is closer to α than β. b. α probe uF

... β ... goal F

The canonical case again is subject agreement (SA). T(ense) enters the syntax bearing the unvalued, uninterpretable φ–features that underlie SA. The closest potential goal for these is the subject in its base position, Spec, vP. The subject therefore provides T’s φ–features with values for person, number, and, in Bantu, gender. T'

(68)

vP

Tuφ probe DPφ goal

...

Agree

English expletive constructions illustrate that closest c-command underlies agreement, since subject agreement can reflect the features of a DP that never raises to Spec, TP (see 69). Alternatively, the agreed-with category can move to Spec, TP (see 70) to satisfy the requirement that TP have something in that position (= the EPP). But agreement is available whether subject raising applies or not: There T uφ be [two men] in the room

(69) a.

Agree

b. There T3PL+be [two men] in the room c.

There are two men in the room

 Vicki Carstens

(70) a. Two men are in the room. b. [Two men] T3PL+be ttwo men in the room

Returning to the topic of concord, Carstens (2001) shows that Agree accounts for it straightforwardly. The agreeing categories in DP always c-command the head noun in its Merge position, prior to noun-raising; hence Agree predicts that their features can licitly be valued: (71) a.

la casa bonita The(f) house(f) pretty(f) ‘the pretty house’

[Spanish]

DP

b. D PROBE #3 >

la

NumP nP

Num

nP

AP PROBE #2 >

bonita

n

NP N

GOAL >

casa

Just one theoretical modification is needed for Agree to account for concord in Universal Grammar. Chomsky (2000, 2001) proposes that a goal must have an undeleted Case feature, and that every Agree relation which values φ-features also deletes Case. If this were true, multiple agreement with a single category would be impossible. I follow Carstens (2001, 2003, 2005) in concluding that only a head lexically specified with the Case-“assigning” property deletes Case in the Agree relation. Under this view, concord is an entirely unexceptional instance of agreement in grammar. 5. Conclusion Resemblances between Bantu and Romance languages have always been apparent in their gender systems and extensive agreement phenomena. In this paper I have looked closely at several aspects of DP syntax in languages from the two groups, including: their nominal gender systems; word order and phrase structure in

DP in Bantu and Romance 

DP; DP-internal concord; and the ability of such concord to license sub-phrasal nominal gaps. The comparison and the results are summarized in (72): (72) Phenomena Genders: Nominal Morphology: DP structure: N-raising: Concord: Empty nouns:

Romance Bantu 2–3 Numerous (in Swahili, 6 plus locative, gerundive, etc.) Suffixal, often Prefixal and fusional concatenating gender and number markers Identical for the two families To DP’s middle field To D For both families, in gender & number, via the Agree relation Licensed by concord in both families

Thus, close investigation confirms that Bantu and Romance families are indeed highly similar in many respects. This in turn confirms the validity and utility of Universal Grammar as a hypothesis and a set of analytical tools.

References Ashton, E.O. 1944. Swahili Grammar Including Intonation. London: Longman. Bresnan, J. & Mchombo, S. 1995. The lexical integrity hypothesis: Evidence from Bantu. Natural Language and Linguistic Theory 13: 181–254. Carstens, V. 1991. The Morphology and Syntax of Determiner Phrases in Kiswahili. Ph.D. dissertation, UCLA. Carstens, V. 1993. On nominal morphology and DP structure. In Theoretical Aspects of Bantu Grammar, S. Mchombo (Ed.). Stanford CA: Stanford University Press. Carstens, V. 1997. Empty nouns in Bantu locatives. The Linguistic Review 14: 361–410. Carstens, V. 2001. Multiple agreement and case deletion: Against φ-(in)completeness. Syntax 3: 147–163. Carstens, V. 2003. Rethinking complementizer agreement: Agree with a case-checked goal. Linguistic Inquiry 34(3): 393–412. Carstens, V. 2005. Agree and EPP in Bantu. Natural Language and Linguistic Theory 23: 219–279. Chomsky, N. 1986. Barriers. Cambridge MA: The MIT Press. Chomsky, N. 1992. Some notes on economy of derivation and representation. In Principles and Parameters in Comparative Grammar, R. Friedin (Ed.), 417–454. Cambridge MA: The MIT Press. Chomsky, N. 1995. The Minimalist Program. Cambridge MA: The MIT Press. Chomsky, N. 2000. Minimalist inquiries: The framework. In Step by Step: Essays on Minimalist Syntax in Honor of Howard Lasnik, H. Lasnik, R. Martin, D. Michaels & J. Uriagereka (Eds), 89–156. Cambridge MA: The MIT Press.

 Vicki Carstens Chomsky, N. 2001. Derivation by phase. In Ken Hale: A Life in Language, M. Kenstowicz (Ed.), 1–52. Cambridge MA: The MIT Press. Cinque, G. 1995. Italian Syntax and Universal Grammar. Cambridge: CUP. Cinque, G. 1999. Adverbs and Functional Heads. Oxford: OUP. Cinque, G. 2000. On Greenberg’s universal 20 and the semitic DP. In University of Venice Working Papers in Linguistics, L. Bruge (Ed.), 10.2: 45–61. Venice: University of Venice. Cinque, G. 2005. On deriving Greenberg’s universal 20. LI 36(3): 315–332. Contini-Morava, E. 2000. Noun class and number in Swahili. In Between Grammar and Lexicon, E. Contini-Morava & Y. Tobi, (Eds), 3–30. Amsterdam: John Benjamins. Contreras, H. 1989. On Spanish empty N’ and N. In Studies in Romance Linguistics, C. Kirschner & J. DeCesaris (Eds), 83–95. Amsterdam: John Benjamins. Corbett, G. 1991. Gender. Cambridge: CUP. Diesing, M. 1992. Indefinites. Cambridge MA: The MIT Press. Di Sciullo, A.M. 2005. Asymmetry in Morphology. Cambridge MA: The MIT Press. Ferrari, F. 2005. A Syntactic Analysis of the Nominal System of Italian and Luganda – how nouns can be formed in the syntax. Ph.D. dissertation, New York University. Greenberg, J. 1963. Some universals of grammar with particular reference to the order of meaningful elements. In Universals of Language, J. Greenberg (Ed), 73–113. Cambridge MA: The MIT Press. Halle, M. & Marantz, A. 1993. Distributed morphology and the pieces of inflection. In The View from Building 20, K. Hale & S.J. Keyser (Eds), 111–176. Cambridge MA: The MIT Press. Harris, J. 1991. The exponence of gender in Spanish. Linguistic Inquiry 22(1): 27–62. Kayne, R. 1994. The Antisymmetry of Syntax. Cambridge MA: The MIT Press. Koopman, H. 1992. On the absence of case-chains in Bambara. Natural Language and Linguistic Theory 10: 555–594. Laenzlinger, 2005. French adjective ordering: Perspectives on DP-internal movement types. Lingua 115: 645–689. Lubowicz, A. 2002. Derived environment effects in optimality theory. Lingua 112: 243–280. McCarthy, J. 2003. Comparative markedness. Theoretical Linguistics 29: 1–51. Meinhoff, C. 1899. Grundriss einer Lautlehre der Bantusprachen. Berlin: Deutsche Morgenländische Gesellschaft, Abhandlungen für die Kunde des Morgenlandes. Mufwene, S. 1980. Bantu Class prefixes: Inflectional or derivational? Papers from the 16th Regional Meeting. Chicago IL: CLS. Mutashansky, O. 2006. Heads-movement in linguistic theory. Linguistic Inquiry 37(1): 69–109. Myers, S. 1987. Tone and the Structure of Words in Shona. Ph.D. dissertation, University of Massachusetts at Amherst. Njogu, K. 1994. The minimal size condition in KiSwahili. Paper presented at the 25th Annual Conference on African Linguistics, Rutgers University, New Brunswick. Park, Jae-Ik. 1997. Disyllabic requirement in Swahili morphology. U. Penn Working Papers in Linguistics 4(2): 245–259. Pesetsky, David. 1990. Experiencer predicates and universal alignment principles. Ms, MIT. Ritter, E. 1988. A head-movement approach to construct state noun phrases. Linguistics 26: 909–929. Ritter, E. 1992. Cross-linguistic evidence for number phrase. The Canadian Journal of Linguistics 37(2): 197–218. Rizzi, L. 1990. Relativized Minimality. Cambridge MA: The MIT Press.

DP in Bantu and Romance 

Roeper, T. 1999. Leftward movement in morphology. In MIT Working Papers in Linguistics 34: 35–66. Rugemalira, J. 2006. The structure of the Bantu noun phrase. Ms, Dar es Salaam. Selkirk, E. 1982. The Syntax of Words. Cambridge MA: The MIT Press. Sproat, R. 1985. On deriving the lexicon. Ph.D. dissertation, MIT. Takano, Y. 2003. How antisymmetric is syntax? Linguistic Inquiry 34(3): 516–526. Valois, D. 1991. The Internal Syntax of DP. Ph.D. dissertation, UCLA.

On the interpretability of φ-features* Roberto Zamparelli1 The paper reconsiders the evidence for the interpretable status of φ-features within DP in Romance languages, focusing on Italian. The conclusions are that, given a reasonably local interpretation of what counts as interpretable, gender is not an interpretable feature, and number (or plur, in the sense of Heycock & Zamparelli 2005) is interpretable, but is not sufficient to drive the process of verbal and predicative agreement. The notion of “default”, particularly with respect to the Person feature, is discussed and various alternatives are explored.

1. Introduction The way grammatical features like gender or number are selected and shared in linguistic structures has always played a prominent role in lexicalist theories of grammar such as Lexical Functional Grammar (LFG), Generalized Phrase Structure Grammar (GPSG) or Head-driven Phrase Structure Grammar (HPSG). In the last ten years, features have started to play an increasingly central role also in the Chomskyan framework, in particular in the Minimalist Program (Chomsky 1995, 2000, 2001, 2004), where they regulate agreement phenomena and drive overt and covert movement (the latter reduced to feature attraction without remerge of phonetically overt material). The technical apparatus of the Minimalist Program introduces an important distinction between interpretable features, i.e., those which contribute semantic information at the Conceptual-Intensional interface (C-I), and non-interpretable ones (written UF, e.g., UT), which play a role in the morphosyntactic derivation but have no semantics and must be deleted at

* This paper owes much to the audiences of the Bantu-Romance Connection E.U.-Funded

Exploratory Workshop in Leeds and of the Stuttgart Symposium on the syntax and semantics of grammatical features, and to two anonymous referees. Special thanks to Denis Delfitto, Valentina Bianchi, Paolo Acquaviva, Roberta D’Alessandro and Viery Samek-Lodovici for helpful hints, discussions and judgments. All errors are of course my own. 1. Università di Trento, DISCOF/CIMeC, Via Matteo del Ben 5b, 38060 Rovereto (TN), Italy. Email: [email protected]

 Roberto Zamparelli

the end of certain cyclic domains, and anyway before the C-I interface. The (non-) interpretable status of features is relativized at projections: for instance, T(ense) features are interpretable at Tns, but non-interpretable at V. Another major ingredient of the minimalist feature system is a mechanism to allow the percolation of feature values throughout appropriate local domains. This system, called Agree (Chomsky 2001), starts from the assumption that uninterpretable features enter the derivation without a value, while interpretable ones are always inserted with a value. Unvalued features are “active”: they function as “probes” seeking a “goal”, i.e., a compatible feature with a value specified. When a goal is found within a certain local domain, the value of the goal is transmitted to the probe which becomes “valued”. Features which remain unvalued at the end of the derivation lead to grammatical failure. As pointed out in Adger (2003); Asudeh & Toivonen (2006) and others, this mechanism of feature value transmission closely resembles the mechanism of Feature Unification adopted in GPSG or HPSG (Sells 1985; Pollard & Sag 1994). An unvalued feature is analogous to a feature whose value is underspecified, where “underspecification” can be expressed as the disjunctive set of all the values the feature can take.2 In general, Agree and unification between an unvalued/underspecified feature and a valued/specified one will yield the same result (e.g., (1)), and fail in the same cases (e.g., (2)), but for different reasons: in (2), the unification mechanism does not find a non-empty intersection, while the Agree operation never takes place with respect to the feature num, since agreement is always triggered by unvalued features and here none is present.3 (1) a. Agree(X[num = plur],Y[num = øJ])⇒X[unum = plur],Y[num = plur] b. Unification(X[num = plur], Y[num = plur,sing])⇒X[num = plur], Y[num = plur]

(2) Agree/Unification(X[num = plur], Y[num = sing]) ⇒ failure!

2. One difference is that expressing underspecification as disjunction allows us to specify exactly which values will count as possible specifications for the feature. Thus gend = masc/fem/ neuter cannot be specified by num = plur since the intersection of their values is empty, while ugend = ø could in principle be specified by num = plur unless this possibility is independently ruled out. A second difference is that unification allows partial specification, as in the unification of num = Sing/Plur/Dual and num = Plur/Dual/Paucal, which gives num = Plur/Dual. This point is, of course, irrelevant if all features are taken to be privative or binary, but can become important if feature structures are used. 3. I will write a/b/…/z for {a, b, …, z}, glossing over the distinction between individual values and sets thereof. I will likewise use Num as a generic number feature, setting aside for the time being the distinction between semantic and syntactic number discussed in Section 5, and write sing/plur/ masc/fem as shorthands for num = plur, etc. When not specified, examples are from Italian.

On the interpretability of φ-features 

Beyond these technical similarities, the minimalist approach in Chomsky (2000), (2001) makes a stronger claim than a simple unification system, since it enforces a relation between having an underspecified value and lacking a semantic interpretation. (3) Valuation/Interpretability Biconditional (Chomsky 2001, 5) A feature F is uninterpretable iff F is unvalued.

The goal of this paper is to take issue with this position, and argue, on the basis of Italian and English data, that unvalued and uninterpretable features are not coextensional. Thus, this paper brings support to alternative conceptions on the relation between valuation and interpretation, such as the one put forth in Pesetsky and Torrego (2007), where ±valued and ±interpretable are distinct (meta)features and all four possible value combinations are claimed to occur. However, the semantic agreement facts discussed in Section 5.3 remain a problem even for this proposal. In Chomskyan minimalism, no feature which receives a value from another feature can give a semantic contribution at C-I or can be preserved past the end of the phase. Vice versa, an interpretable feature must be fully specified in value from the moment it is inserted with a lexical item. This predicts that the following cases should not arise: (4) a. An uninterpretable feature takes a value from another uninterpretable one. b. An unvalued feature takes its value from semantic, rather than syntactic specifications. I will argue that both of these cases take place within the familiar domain of φ-features.4

1.1 Preliminary issues Any analysis of features and feature semantics requires as a methodological preliminary a brief discussion of three issues. Firstly, what do we mean by interpretable? Secondly, how do we intend to represent the notion of default value and its semantic mapping? Thirdly, how do we cast the influence of lexically idiosyncratic information on the interpretation of features? The notion of “interpretable (at a projection F)” is of course at the heart of the matter, and regretfully, it has not received the attention it deserves within the linguistic community at work on the Minimalist Program (though see at least Brody

4. Objections to (3) have already been raised in Watanabe (2000), on the basis of switch reference and complementizer agreement data.

 Roberto Zamparelli

1997; Borer 2004; Fox 1999). A minimal requirement is that the interpretation be local. In other terms, the meaning conveyed by F to a head H should not make reference to the relation between H and other projections. Suppose, for instance, that we suspect that the number agreement on past participles is uninterpretable – a mere reflex of a syntactic relation. If num = plur at the participle had the nonlocal interpretation “this head is in an agreement relation with a DP whose head denotes a plurality”, this feature would seem to be interpretable.5 But by similar means, all features at all projections could be made interpretable, emptying the set of non-interpretable features. The concrete position I will adopt here is in (5):

(5) A feature F is interpretable at a head H of category K iff

a. The value of F at H contributes to determine the denotation of H (in the case of a functional head, the type of function the head denotes) b. The semantic effect of F is not completely dependent on the lexical properties of the various members of K.

Basically, a feature should have a systematic effect on the way words interact with the semantic model, and it should bring information which does not change depending on the idiosyncratic lexical properties of the words within a category. This second point excludes the case of a hypothetical feature with meaning, say, “definite” in association with the noun dog but “indefinite” in association with table, but leaves open the possibility that features might give different semantic contributions across categories, or on specific functional items; for instance, it is plausible and indeed unsurprising that the feature for number should have different effects on words as different as the, some and dog (though not on dog and rabbit). This formulation does not exclude the possibility that the semantic effect of a feature might be partially dependent on the value of other features, though we would expect at least a family resemblance in the output. For instance, Heycock and Zamparelli (2005) define a feature for semantic plurality whose exact effect depends on whether the noun is morphologically plural (as with count nouns) or singular (as in general with mass nouns like water); however, in both cases its application leads to a lattice-structured denotation (see Sec. 5 for further details). A feature of this sort should count as interpretable according to (5). The idea of default or unmarked values is pervasive in linguistics since the seminal work of Greenberg (1966). In the perspective of the C-I interface and the Agree system, the main question is how a feature system can express the idea that “3rd person”, “singular” and “masculine” are default values for pers(on), num(ber) and gend(er), as often proposed for Romance. 5. After all, such information might not be useless, from a parsing perspective.

On the interpretability of φ-features 

At least two notions of “default” must be distinguished here. First, there is a notion of default as the most normal choice among a set of values: “default gender” might thus be the one that most nouns in a language have, or that attracts the largest number of borrowings. Alternatively, a complexity-based criterion could be used (the default form would be the one with the simplest morphology, the simplest semantics according to some metrics), or an inclusiveness-based metrics, (“default form” is the one whose meaning is a superset of the meanings of all the marked forms). These criteria (which, however, do not always agree) point to the fact that one among multiple values is “less marked” or “more basic” than the others. Let us call this, with Corbett (2006), “normal case default” or simply “normal default”. The second notion of default value (called here “exceptional default”, see Corbett 2006, 5.2) embodies the idea of last resort: when a category which does not normally use (and thus, presumably, have) a certain feature appears in a position where that feature is required, a “default value” appears. A good example is the φ-feature values assigned to CPs, VPs, citations or other non-DP elements in subject position. In Italian and many other languages these forms trigger agreement for “3rd person”, “singular” and “masculine”. Typically, the values that emerge from these two notions of default coincide, but this is not always the case: in Russian, most nouns and most nominal borrowings are masculine, but clausal and other non-DP subjects are marked neuter (Corbett & Fraser 1993). The question is now how and to what extent a formal features system should represent these two notions. One concrete possibility is to collapse them together using a binary system of privative features: the absence of a feature is understood as the unmarked value; its presence as the marked one. Alternatively, when the feature is present a value can be added. This way, multiple marked values can be specified. In a system of this sort both the N dog and the CP subject that he left would have no feature for number, and in both cases this absence would be read as “singular”. This situation can be complemented by a default at the C-I interface: when the interpretation system receives no information about a certain feature it carries out some default computation (often the identity function: simply pass the meaning built so far on for further processing). This simple solution, represented in (6a), is limited in that it does not allow for two different default values to be specified, as in the case of Russian.6 Any feature system that wants to preserve the distinction must somehow make dog and a subject CP come out differently. For instance, singular nominals could have

6. More exactly, it should be possible to state a conditional rule such as: “If the root is nominal, the lack of gender feature means masculine’, otherwise, it means neuter”. This seems to me to run against the spirit of defaults as unconditioned, bottom-line, when-all-else-fails devices.

 Roberto Zamparelli

a num feature but no value for it, while clausal subjects would really have no num. This is illustrated in (6b). (6)

a. {[A yellow dog]/[that he left]} isunum = ø not so strange b. {[A yellow dognum]/[that he left]} isunum = ø not so strange c. {[A yellow dognum = sing]/[that he left]} isunum = ø not so strange d. {[A yellow dognum = sing]/[that he left]num = sing (by except. default)} is unum = ø not so strange

A more radical possibility is to say that the notion of normal default is an epiphenomenon that emerges from statistical properties of language, which should simply have no place in its grammatical description. In this view, the fact that singular is more basic or common should have no reflex in its featural representation, while the φ-feature-defective status of non-DP subject does have a place in grammar. This would lead to the representation in (6c). I find this solution plausible, but there are still two drawbacks, one technical, the other more substantial, which apply to all of these solutions. Firstly, according to the logic of Agree, most cases in (6) should crash because the unvalued, uninterpretable feature of the verb (marked unum = ø) is not valued by the CP or the DP. In (6b) Num with no value must somehow be kept distinct from the unvalued feature unum = ø on V. One solution might be to fill out all the missing features and feature values with defaults before a word is merged, or at least before Agree. This solution (in (6d)) saves Agree, but obliterates the fact that some values are default. The substantive problem for the use of privative features is the possibility of disjunctive specifications (“feature F has value A or B”) when one of the alternatives is the unmarked choice. In Haitian Creole (HC), for instance, a bare noun is ambiguous between singular or plural. Only the context, or world-knowledge, disambiguates (Deprez 2005). (7) Jan achte liv pou Pòl. John bought book for Paul ‘Jan bought a book/books for Paul.’

This fact can be given the following interpretations: (a) liv has no num (and the absence of a feature does not indicate a default, but the union of all its possible values). This solution predicts, incorrectly, that CPs and similar non-DP categories, which have no φ-features for completely different reasons, might similarly be plural or singular; (b) liv has num = {plur, sing}: an explicit disjunctive system, incompatible with the idea that a default is an absence of features or values; (c) liv has num = plur, but plural includes singular (cf. footnote 10). The difference between HC and e.g., English is that HC has no overt marking for singularity, so the singular meaning of plurals is not blocked by the existence of a singular form.

On the interpretability of φ-features 

However, HC does have a way to specify “a book” (yon liv), so a scalar account might not be so directly implemented. This brief discussion shows that it is not easy to design a feature system that can simultaneously represent movement (via Agree), the different notions of default and the possibility of disjunctive feature values. For the time being I will assume that default values (and default features) are filled in before a word is merged. This makes all feature values available for syntactic processes, but hides the difference between values which come from the lexicon and values which are filled in by default. I will return to a problematic consequence of this choice in Section 3 and 5.2. A final issue is how to handle the effect of ‘exceptional’ lexical items on feature interpretation. Consider (8)–(10). (8) Pisae debellatae sunt. Pisaplur defeatedpiur areplus ‘Pisa was defeated’. (9) I viveri scarseggano. the food_suppliesplur are_scarceplur ‘Food supplies are scarce’. (10) Vous2pl êtes2pl loyalsing . you aux loyal ‘You(= one, formal address) have been loyal’.

In the Latin (8) a plural name triggers agreement on the aux and participle, yet the denotation is a single city: the plural feature value must not be interpreted or we would get the wrong semantics.7 In the Italian (9), viveri is a plural mass noun (there is no singular *vivero and it is impossible to use a cardinal number with the plural form: *tre viveri “three food supplies”). In (10), a case of formal address in French, the plural feature on vous codes the “wrong” semantics: not number (the addressee can be singular), but formality. The picture that emerges is the familiar hierarchy where more specific information blocks less specific information (an application of Kiparsky’s 1982 Elsewhere Condition). (11) Root semantics > Marked feature values > Default feature values.

In what follows I will assume that something like (11) is in place, without trying to formalize how it must be implemented.

7. Other examples are Athenae, Syracusae, Thebae, nuptiae “wedding”, aedes “house”, etc. Many have been borrowed from Greek, where they were already plural. In some cases the syntax might historically be derived from a real plurality, but it is unlikely that the plural meaning survived the borrowing.

 Roberto Zamparelli

The rest of this paper is organized as follows. After an introduction to the distinction between internal and external agreement, I will briefly examine the status of the pers feature (Section 3), not from the standpoint of interpretability, which is not at issue, but with respect to the putative absence of 3rd as a real value. In Section 4 I will argue that, given a properly restricted notion of interpretability, gend is not independently interpretable, since the mapping between the grammatical feature and its semantics (say, biological gender) is too heavily dependent on the meaning of the root gender associates with. This seems to be true even in null-nominal or pronominal constructions (Section 4.2), where gender is shown to depend on the semantics of the missing N or of the antecedent/binder. Section 5 deals with the status of num, and in particular, the number agreement that appears on verbs and primary predicates. The conclusions will be that external number agreement must be influenced both by syntactic and semantic factors. Section 6 will draw some final conclusions and point to directions for further work. 2. φ-features A traditional idea in the literature on agreement is that agreement is a directional process: agreement is not simply a systematic covariance in grammatical feature values among different linguistic elements, but a process where a feature value originates at a source element (the controller) and is assigned or received by one or more other target elements (Lapointe 1988; Corbett 2000, 2006). We can thus say that “X agrees with Y in F”, X the controller, Y the target. In the minimalist system the target is in some sense “unsaturated”, and thus “active” in probing for a value, while the controlled is the passive element. The literature on agreement usually distinguishes between “internal” and “external” agreement (see e.g., Lehmann 1988). The former is the agreement in number, gender, case, but never person, which obtains within the nominal phrase, specifically between N, its preposed or postposed adjectival and numeral modifiers, the quantifiers and determiners, plus in some languages the possessor (with the possessed, as e.g., in Chamorro), the preposition that introduces the noun’s complement and the relative pronoun (e.g., Arabic). “External” agreement is the correspondence in number, gender and person, but never Case, which obtains between a verb (auxiliary and/or main) and its arguments (subject, object, indirect object). (See Lapointe 1988 for a more detailed classification). For external agreement, it is normally assumed that the controller is the nominal, and the verb or the T/aux node is the target. Indeed, φ-features on DPs are usually counted among interpretable features (see e.g., Hornstein, Nunes &

On the interpretability of φ-features 

Grohmann 2005), just like tense features on T. The inventory of uninterpretable features, on the other hand, includes φ-features at T, small v and Adj, and T features on DPs. For internal agreement the situation is more complex and, in pre-DP terms, it was sometimes difficult to identify target and controller for each feature. In an extended DP structure containing at least distinct heads for determiners/quantifiers (D), cardinal numerals (num) and the noun (n), it becomes possible to try to identify the controller with a specific head or projection, and to pinpoint its semantic contribution.8 3. Person pers is primarily a property of (overt and covert) pronouns, and it is widely acknowledged to be related (at least) to conversational roles. A common position is that 1st person codes the Speaker, 2nd person the Hearer/Addressee, and the 3rd, any other entity salient in the discourse (see Bianchi 2003; Schlenker 2003 for various positions and a review of the literature, and Delfitto in press; Longobardi 2006 for alternative views). This “everything else” character of 3rd person has prompted the idea (originally advanced in Benveniste 1966), that 3rd should be the default value for pers: a nominal is “3rd” when it is not otherwise marked for person. This view has been frequently picked up in the literature (it is used, for instance, to explain the Person-Case constrain in Anagnostopoulou 2003 and Adger & Harbour 2003) and sometimes implemented with the proposal that 3rd-person-marked elements lack the pers feature altogether. Harley & Ritter (2002), for instance, start from the feature geometry in (12) and propose that 3rd person corresponds to the absence of 8. In Chomsky (2001) Case is treated differently from other features, since its assignment is linked to the notion of φ-completeness. In general, Case is not taken to be interpretable on DPs, so I will not analyze it in the rest of this paper. Descriptively speaking, it suffices to say that structural Case is assigned to a high DP position (possibly the DP layer itself, identified with a “Kase Phrase” in Giusti 1992, 1999). Evidence comes from the fact that in various languages (e.g., German, Russian) the Case assigned to the whole DP is only visible on determiners and/or pronouns: the noun either bears no Case (German for nominative/accusative) or bears a Case marking dependent on the quantifier, like the genitive required by numerals in Russian (i) (Lehmann 1988). (i)

tri three-nom-plur

svietlye light-nom-plur

comnaty room-gen-sing

This non-homogeneous agreement does not obtain with oblique Case like instrumental (see Babby 1988 for discussion).

 Roberto Zamparelli

the whole participant node (if the node is present but with no further specifications, it defaults to 1st person). (12)

Referring Expression PARTICIPANT Speaker

INDIVIDUATION

Addressee Minimal Group

Class

Augmented Animate Inanimate/Neuter Masc.

Fem.

Moving from an implementation of this sort, Matushansky (2006) concludes that the feature Pers cannot be a trigger for the attraction of the DP, and that it is impossible to find a φ-feature common to all kinds of DPs (for instance, in a Chinese 3rd person DP with no number marking). The solution Matushansky subscribes to is that A-movement is not driven by φ-features, but by categorial N-features.9 This radical approach to person introduces, however, an asymmetry which seems to go beyond the fact that person is minimally a 3-way distinction (not counting inclusive/exclusive), while number is often a two-way choice. In the Harley and Ritter’s feature geometry, a 3rd person singular element has no node for participant, but it does have a node for individuation, which defaults to singular (13). (13) Referring Expression individuation

Harley and Ritter’s typological classification does include languages with no individuation node (Berik, Maxakali, Kwakiutl): these are languages that have no number marking on (most) pronouns, full nouns or verbs. Given this meaning for “no individuation node”, one would naturally expect that the lack of the participant node should mean “no person is morphologically specified”. This would of course be wrong: 3rd person is often (sometimes even exclusively, as in English) morphologically marked and in alternative to 1st and 2nd. Moreover,

9. Despite their name, in order to trigger movement N-features should be common to a wide and rather heterogeneous collection of categories: nouns (for bare plural/mass arguments), pronouns, sentential subjects, VPs and prepositional subjects (given {that he left/to leave/leaving Rome} was good, under the bed is a good place to hide). All things being equal, a solution in terms of φ-features seems less ad hoc.

On the interpretability of φ-features 

while the absence of the individuation node seems to be a property of whole languages, the absence of the participant node would apply to specific forms within a paradigm (e.g., to he, but not you, I). Further empirical investigation casts more doubts on Harley & Ritter’s solution. Examining a wider class of pronominal elements (clitics, reflexives, pronouns of formal address), Bianchi (2005) shows that the details of the “3rd person = no person” idea are actually not straightforward to spell out, and moves on to propose pers may indeed be the only feature which is crucial for the licensing of Nominative case. More generally, it seems to me that the idea that the default value for pers always amounts to lack of pers mostly reflects the absence of a principled distinction between the two notions of default discussed in Section 1.1, and between the lack of a feature and the lack of a feature value, which can be used to implement it. Consider again the motivations for seeing 3rd as a default. CP and VP subjects always trigger 3rd person agreement (and singular number, see Section 5.2). Bare nominal subjects, where no D is visible, are also always 3rd person, despite the fact that N and CP can coexist with a non-3rd-person pronoun, in the constructions: (14) [Noi {linguisti/che siamo linguisti}] amiamo [we1plur {linguists/that are linguists}] like1plur

gli alberi cespugliosi. the trees bushy

(14) can be taken as a sign that Ns and CPs are not themselves marked with pers. When an overt D is present in the local domain they take whatever person D specifies; in bare nouns, no Pers specification comes from D and bare nominal or CP subjects receive “3rd person” via exceptional default, just as CPs or other non-DP subjects receive masculine agreement (or neuter, in Russian). This is of course quite a different thing from saying that all third person elements lack pers. Once two types of defaults are recognized, we could say that that certain pronouns are marked as pers = 3rd (so that the -s verbal affix in English can be sensitive to the presence of this value, rather than to the absence of a feature), while Ns, CPs and VPs do not come with a pers feature. As discussed in Section 1.1, this solution makes a difference for disjunctive feature specification. Longobardi (p.c.) observes that regarding 3rd as absence of person would make it impossible to distinguish between pronouns that can take 1st, 2nd and 3rd person antecedents (e.g., Russian sebie) and pronouns that can just take 1st and 2nd ones. Some differences between person and other features exist, but they are not unexpected given the double notion of default. (14) contrasts with (15)/(16), where gender/ number mismatches in DP-internal agreement lead to a crash: here the fact that the masculine/singular associated with morphological affixes are “normal defaults”

 Roberto Zamparelli

(i.e., unmarked values for gender and number) does not entail that they can be overridden by alternative, marked specifications.10 (15) *La ragazzo thefem boymasc (16) a. *Onesing boysplur (was with me.) b. *Two [boy and girl] (were together.)

The situation illustrated in (16)/(15) cannot be replicated with person simply because there is no overt person marker in anything but pronouns. In turn, (16)/(15) contrast with (17), where masculine emerges out of gender mismatch among the conjuncts. The difference is probably that there is no agreement between conjuncts; due to the mismatch no gender is assigned to the coordination as a whole. As a result, the DP is in the same situation as a subject CP/VP, and the exceptional default resolution rule applies, deriving masculine. In other languages the resolution strategies can of course be quite different: in Ikalanga, a Bantu language, a reciprocal verb cannot take a non-human coordinated subject (18a), and must resort to a comitative construction (Letsholo & Safir 2007). I will return to this case in Section 5.3. (17) [Il ragazzo e la ragazza] erano {alti/*alte}. [the boy and the girl] were {tallmasc/tallfem} (18) a. *Shumba ne vubu dza-ka-bon-an-a. lions10 and hippos10 sa10-past-see-recip-fv desired meaning: ‘the lions and the hippopotamus saw each other’ b. Shumba dza-ka-bon-an-a ne vubu. Lions10 sa10-past-see-recip-fv with hippos10 lit: ‘lions saw each other with hippos’

In conclusion, there seem to be grounds to distinguish between a genuine 3rd value for the Pers feature and the possibility of assigning pers = 3rd to categories which lack any person specification, in a way analogous to default gender assignment to non-gender marked categories. Notice, however, that this forces us to abandon the idea that default values are computed and added before Merge (see the end of Section 1.1). In this solution, the difference between normal and exception defaults was invisible to Agree, but this fails to explain (14). The exceptional

10. Notice that (16) is semantically fine: in most theories of plurality one is semantically compatible with a plurality (and anyway 1.0 is: we have, on average, 1.0 customer*(s) per minute), and a NP-conjunction denotes a set of pairs (witness a boy and girl loved each other, discussed in Heycock & Zamparelli 2005). The problem of (16) is purely feature mismatch.

On the interpretability of φ-features 

default resolution rule must be used after other possibilities have been tried, and this apparently includes agreement with a feature-marked category, as in (14). One solution is, of course, to mark the difference between values which come from the lexicon and values which have been filled by defaults. Ultimately, a modification of the Agree system might turn out to be a more elegant solution, but I will not try to spell it out in this paper. 4. Gender There is a wide consensus that gender is a property of nominal roots, possibly mediated via the selection of a word marker (e.g., Spanish -o- in perr-o-s “dogsMasc-Plur”, see Harris 1991). Italian and Spanish have a default gender, which is masculine.11 Gender can then be transmitted to adjectives and determiners, bearing their own word-markers. Gender markings also appear on pronouns and deictics, even in the absence of a noun. The question now is whether gender is an interpretable feature, and if so, in what sense. I will consider first the role of gender on nouns, then discuss the role of gender in null-nominal constructions. 4.1 The Gender of Nouns In Romance, “masculine” and “feminine” grammatical genders take their names from the biological sex of canonical nouns referring to men, women and to male/ female animals, while neuter tends to apply to inanimates.12 However, as is well known, the domain of grammatical gender is much wider than that of biological sex: most feminine or masculine nouns denote objects without any connection with biological sex; in this case the assignment of the noun to a gender class is syncronically arbitrary. The strongest connection between grammar and biology obtains with nouns which have a double possibility for gender: when these nouns refer to an animal with two sexes (i.e., they are +animate, assuming that for linguistic purpose every animate is taken to have a biological sex, modulo plants), the grammatical masculine will be used for the male and the feminine for the female (19). (19) a. bambinomasc ‘male child’ vs. bambinaFem ‘female child’ b. gattomasc ‘tomcat’ vs. gattafem ‘female cat’ 11. Once again, the main reason to consider masculine the default is that citations, VP or CP subjects, which are not listed in the lexicon as having a particular gender, trigger masculine agreement. 12. If the canonical nouns for man and woman share the same gender the opposition is between a “common” and a “neuter” gender, as in various Scandinavian languages.

 Roberto Zamparelli

In these cases the masculine is often used also to refer to the entity irrespective of gender, but when the name of a species has a single form for the two sexes, this can be masculine (e.g., (20)) or feminine (21). (20) a. alcemasc ‘moose’ b. lucciomasc ‘pike’

(male or female) (male or female)

(21) a. tigrefem, ‘tiger’ b. ocafem ‘goose

(male or female) (male or female)

Similarly, there are cases of human roles which are only masculine or feminine, but can apply to humans of either sexes. (22) a. guidafem ‘guide’ b. genitoremasc ‘parent’

(male or female) (male or female13)

There are then nouns with a double gender possibility which regularly mark distinctions only loosely connected with biological sex, like the tree/fruit distinction in (23) (not all fruits follow this pattern) and others where the distinction expressed by gender is idiosyncratic and unpredictable ((24) and (25)) (23) a. melomasc ‘apple tree’ vs. melafem ‘apple fruit’ b. aranciomasc ‘orange tree’ vs. aranciafem ‘orange fruit’ (24) a. tavolomasc ‘table’ vs. tavolafem ‘table set to eat’, b. legnomasc ‘wood (material)’ vs. legnafem ‘wood for burning’ c. buco masc ‘hole’ vs. bucafem ‘hole in the ground’ (25) a. frontemasc ‘war front’ vs. frontefem ‘forehead’ b. moralemasc ‘morale’ (psych.) vs. moralefem ‘moral’ (of a story)

Sometimes, as in (25), these noun fall in the two-way declension (singular -e, plural -i), which neutralizes gender distinctions, so no morphological information on the word reveals gender (these cases can be mapped onto Harris’ (1991) “Outer Core” nouns). Finally, there are cases with no (syncronic) semantic relation between homophonous masculine/feminine pairs: (26) a. contemasc ‘earl’ vs. contafem ‘count’ b. partomasc ‘delivery’ vs. partefem ‘part’

In a few cases, masculine nouns can refer to canonically female animates (see the Italian sopranomasc, contraltomasc). In the other direction, the German and Dutch

13. Genitricefem exists, but in present-day Italian it has the more formal biological meaning of “female who has generated X”.

On the interpretability of φ-features 

names for “girl” are grammatically neuter (madchen, medje) due to the presence of a frozen neuter diminutive suffix. This brief survey shows that while there is a connection between biological and grammatical gender, this is certainly not the one-to-one correspondence one finds in languages like Tamil or Kannada, where male human beings are masculine, female ones, feminine and all the rest is neuter (Corbett 1991, 2.1). In these languages there can be little doubt that gend is a semantic feature; in Romance, things are different. Noticing that grammatical gender is mapped onto biological sex in a subset of the roots that refer to animals or humans, we can abstract the feature +animate (modulo plants) and formulate (27) for Italian: (27) Gender Generalization: When the gend feature combines with a +animate N root which can be either masculine or feminine, masc /fem Ns will denote biologically male/female living beings, respectively.

This applies to roots like gatt- “cat”, but not lucc(i)- “pike”. The question now is: is this biology-grammar connection sufficient reason to assume that the feature gend should be represented at the C-I with a meaning of its own? I believe that the answer must be negative. The reason is that grammatical gender does not add any information which is not present in the lexical semantics of the root it helps to select. Essentially, we would get the same empirical coverage by saying that gend is a non-interpretable feature which can distinguish between roots. Some of these roots differ semantically in the dimension of biological sex, others in other dimensions (cf. the neuter associated with diminutives in German, or the feminine of the abstraction affixes -heit, -keit, -ung, -schaft). In (27), which is based on the items which come closest to a regular semantic mapping, the correspondence between gend and biology requires another feature value, ±animate and an appropriate root meaning. In most other cases, sending to C-I the value of the gend feature itself would be not just redundant but misleading (as in volpefem for a male fox): from a statistical standpoint, the cases of gender covered by (27) are a tiny minority. What is missing is point (b) of the definition of “interpretable”: the substantial independence of the feature meaning from the lexical meaning of roots. It is important here to keep apart the notions “interpretable” and “distinctive”. There is little doubt that gend is a distinctive feature: the roots tavol- and bambinare disambiguated only when the information about their grammatical gender is retrieved from the word markers. But once the correct root is selected, denoting, say, a table set to eat or a biologically male child, gender has exhausted its function, much as the feature [±voiced], after helping us to select pull rather than bull can be dispensed with at C-I. This would of course be true even if certain phonemes had a tendency to be associated with certain semantic class (as in the English initial /tw/

 Roberto Zamparelli

in twist, twirl, tweak, twiddle, twine, twinge, twitch).14 While it is possible that such subregularities might play a more or less subconscious role in the comprehension and creation of lexical items, they should not be part of “core” grammar, which should however certainly include the systematic syntax-to-semantics mapping of number or person. 4.2 Gender in the absence of overt nouns It is of course possible to take a completely different stand, and assume that gend does have a semantics related to biological sex but that this semantics is ‘suspended’ (i.e., the feature is not interpreted) in connection with particular stems, like those in (20)–(22).15 As we have seen in (8)–(10), cases where lexical information overrides the meaning of features are far from unknown; however, applying this strategy to gender seems to me to turn the exception into a rule, since the cases where gend would be interpretable are a small proportion of those which realize this feature (even setting adjectives aside). One could envision a more general system in which the interpretability of features (i.e., whether they are interpreted at all, not just the way in which they are) is triggered on and off depending on other features, but such a system would not be any less problematic for a movement system based on Agree which also adopts (3).16 Despite these shortcomings, the possibility that Gender is interpretable in general should not be dismissed too easily. One way to investigate it for Italian is to study the meaning of gender markings which appear on pronouns and determiners, in the absence of a nominal root which might condition the meaning. In these cases, it seems initially plausible that gender might play the role of restricting the range of objects the gender marked element can denote. However, this conclusion would

14. An interesting comparison is provided by declension classes. In Italian, the morpheme cont- is ambiguous between a root in the first declension (conto/conti “bill”) and one in the second (conte/conti “earl”), both masculine. Here the declension class disambiguates, yet as far as I am aware nobody has ever proposed that declension classes should be part of the information present at C-I. 15. A system of this sort has been suggested to me by Orin Percus (p.c.). 16. In a diachronic perspective, the comparison between Tamil and Italian raises the interesting question of whether the interpretability of a feature should hold cross-linguistically. If not, when a language like Tamil evolves, genderwise, into a language like Italian and further on to a quasirandom association between stems and genders, there seems to be a point where its gend feature loses its original semantic content. Given the biconditional in (3) it seems that at this point variations in the pattern of movement could be expected. These cases would be interesting to study, if they exist at all.

On the interpretability of φ-features 

be too hasty, since these cases actually support the idea that the semantic effect of gender is determined by the semantic content of the elements it is associated with. Consider first the case of tonic personal pronouns (Italian lui/lei, etc.). These pronouns are expressions of the feature +animate (and +definite), so it is not surprising that when gend associates with them, it behaves exactly as in the association with a two-gender nominal root like gatt- “cat”, and gets mapped onto biological sex. In English, where there is no evidence at all that gender is a grammatical feature, the connection with biological gender is part of the lexical meanings of he and she (though quite possibly not of it, they).17 The connection between gender-marked pronouns and biological sex has been modeled as a presupposition (see Heim & Kratzer 1998; Schlenker 2003) (see (28)). (28) a. [Every director]i [ti likes heri mother] b. Assertion: For all x such that ⟦director⟧(x) ⟦tx likes herx mother⟧ c. Presupposition: For all x such that [[director]](x), x is female.

In Italian, a language with gender agreement, the situation is different. Consider first cases where a linguistic antecedent is present. (29a) and (29b) differ in whether the pronoun agrees with its binder/antecedent guidafem “guide” or not. Both are grammatical, but for most speakers I have asked, the masculine, non-agreeing version (29a) carries a presupposition that all the guides are males (as in English), while the female, agreeing version (29b) allows for the guides to be male or female.18 (29) a.

La/Ogni guidafem sosteneva che con luimasc nessuno si era mai perso. the/every guide claimed that with him nobody had ever got lost

b. La/Ogni guidafem sosteneva che con leifem nessuno si era mai perso. the/every guide claimed that with her nobody had ever got lost

This contrast shows that the gender on a bound personal pronoun can either obtain the semantics of biological gender from the feature +animate of the pronoun (English-style, ignoring the gender-to-semantics mapping proposed by its binder, contra Kratzer 2008) or derive the mapping from the binder/antecedent.

17. What I am claiming is thus that Italian personal pronouns can be considered pure bundles of grammatical features, while English he/him and she/her have a residual lexical meaning related to biological sex, which also allows for exceptions like the she for ships. In Italian, essomasc “it”, sometimes considered a neuter element, actually behaves, English-style, as a word that refers to inanimates. 18. For the other speakers I consulted the feminine actually presupposes that all the guides are females. For these speakers, gender is always interpreted “at the pronoun”, English-style. One reviewer points out that in Modern Greek, non agreeing cases equivalent to (29)b block sloppy identity readings in VP ellipsis. I have no account for this fact.

 Roberto Zamparelli

In (29b), guida, a feminine root which does not map grammatical gender onto biological gender allows leifem to refer to male guides too.19 In (29a) the non-agreeing pronoun is not entirely independent from the features of the binder. This is shown by the fact that in (30) the (marked) +animate feature on the pronoun is not able to override a –animate from the binder. Lui is marginal with a masculine –animate binder from the class illustrated in (24).20 (30)

Ogni tavolomasc aveva una sedia tra {?luimasc /essomasc /*leifem} every table had a chair between {him /it /her} e la parete. and the wall

Let us now turn to pronouns which take their referent from the non-linguistic context in which they are uttered. Contextually-controlled pronouns are important in a discussion of gender, since, with the grammatical gender on the antecedent out of the linguistic context they really seem to offer the best opportunity to observe the semantic contribution of gender, if it exists, in “pure” form. With this goal in mind, we need to exclude from consideration 1st and 2nd person pronouns and tonic 3rd person pronouns, which are all +animate. As we have just seen, in this case the pronoun is not really free from a linguistic antecedent, since the feature +animate behaves as if the gender feature was associated with a nominal root of the gatt- type, enforcing the gender-biology mapping in (27). To find more informative examples of null-nominal, gender marked pronouns one has to turn to the 3rd person object clitics (French le/la/les/les, Italian lo/la/li/le in (31)), but also gender-marked quantifiers like ognun-o/a in (32) and demonstratives (33), which do not necessarily apply to animates. (31) a.

Carlo lo / la vide. Carlo CLmasc / CLfem saw ‘Carlo saw it/him/her’

b. Guardale bene! look_at_themfem well 19. The choice is also subject to pragmatic constraints. Since soprano are almost always females, a masculine agreement seems harder. (i)

Ogni soprano pensavo che il pubblico applaudisse per ?lui/lei. every sopranomasc thought that the audience clapped for him/her.

Notice that both (i) and (29) run against the tendency to use masculine as a general form to cover both males and females. 20. The fact that lui is not completely impossible is probably due to the fact that the use of esso “it” is vanishing from present-day spoken Italian.

On the interpretability of φ-features 

(32) a.

C’ erano dieci casse, e ognuna conteneva qualcosa. there were ten boxesfem, and each onefem had something inside.

b. Avevano dieci guide, ma qualcuna era in disaccordo. they_had ten guidesfem, and somefem was in disagreement (33) Guarda quelli /queste! look_at thosemasc /thesefem!

What is interesting in these examples – and should make us wary – is that the contribution of gender is essentially metalinguistic. Gender does not give any direct information about an object, but merely signals that a possible nominal description of that object is grammatically feminine or masculine.21 This is very different from, for instance, the contribution of number or person, which tells us something about the domain of discourse or the conversation participants. Suppose, however, that this type of metalinguistic information is admissible. Masculine clitics or deictics can apply only to objects whose linguistic realization is masculine, so it seems that the use of a particular gender should (minimally) have the effect of restricting the possible referents of a pronoun, demonstrative or quantifier. Saying aprila “open itfem” in front of a closed door signals that admissible descriptions for the object are portafem “door”, cosafem “thing”, possibly barrierafem “barrier”, but not usciomasc “door”, portonemasc “building door”, oggettomasc “object”, etc. Since suffi ciently abstract descriptive nouns of either genders are not hard to find it should always be possible, if hard-pushed, to adapt either masculine or feminine to pretty much any object. In actual linguistic use, however, things do not work this way. The first thing to observe is that gender is non-restrictive. For instance, (34) (said in the context of a psycholinguistic experiment with a distractor) is completely impossible, even if the use of gender should make it clear and unambiguous that the hearer should look at the pear and ignore the pineapple. (34) [looking at a screen where a pear (‘perafem’) and a pineapple (‘ananasmasc’) appear briefly] *Guardala e Look at itfem and

ignoralo. ignore itmasc

One could object that a feature on a personal pronoun might not be restrictive just because pronouns do not have a restrictive scope anyway. However, gender does

21. One problem is that gender applies to a possible name for the antecedent, but sometimes more than one description exists, as with cosomasc “thing (concrete)” vs. cosafem “thing (concrete or abstract)”.

 Roberto Zamparelli

not work any better as a restrictor of quantifiers and of one-pronominals. (35) cannot mean “every one/some of the grammatically feminine nouns (i.e., forks and bowls) had been polished”.22 (35) Sul tavolo c’ erano three forchettefem, due zuppierefem, due on_the table there were 3 forksfem, two bowlsfem, two cucchiaimasc ed un coltellomasc. {Ognunafem /qualcunafem} era stata lucidata. spoonsmasc and 1 knifemasc. {Everyonefem /Somefem} had been polished. (36) [same context as above] ??Ne ho presa una. CL I_have taken onefem Expected but unnatural meaning: ‘I took one of the forks and bowls’

Next, the expression in (37), in the context given, cannot be an instruction to look at the one object in the scene which is both yellow and expressed by a masculine noun. The adjective is restrictive, the gender on the adjective is not. (37) [Observing a picture with a yellow pear (‘perafem’) and a yellow pineapple (‘ananasmasc’)] Guarda quello giallo. look at thatmasc yellowmasc ‘Look at the yellow one’

Now imagine that the screen displays a non-figurative knot of lines and colors, something for which I cannot find an interpretation or a name. In these circumstances (38) is marginal for some speakers, especially in the feminine version (which in principle should latch on to descriptions such as cosafem, or figurafem). The possibility of using a pronoun improves acceptability to the extent I can name the object with a plausible category name. (38) Guardala /Guardalo attentamente. look at itfem /look at itmasc carefully

Taken together, these observations suggest that the search for null-nominal constructions where the semantic value of gender can emerge in “pure” form might be fruitless. Gender on contextually-controlled pronouns does not act as a device to restrict the search space of pronouns or quantifiers, in contrast with the

22. This reading is possible if the feminine nouns are amassed as the end of the multiple conjunction, either because of proximity effect, or because in this position the selected nouns can form a constituent (in a binary branching view of coordination). Either way, this effect has nothing to do with gender.

On the interpretability of φ-features 

information provided by person, definiteness, proximity and maybe number.23 Gender does of course carry grammatical information, but such information is metalinguistic (and in a sense, non local) and thus irrelevant for our purposes. One might object that the non-restrictive behavior of gend is unsurprising given the idea that this feature is presuppositional (see (28)). Notice, however, that gend is restrictive when mapped onto biological sex, (so in (39) the quantifier ranges over male or female cats). This is exactly what one expects from an account in which gend selects a root, this root has a (possibly biological) meaning, and this meaning is restrictive. (39) Ogni {gatto /gatta} mangiava. every {catmasc /catfem} was eating

In conclusion, in this section I have argued that the gender-to-semantics marking of pronouns is obtained from the noun stem. If these ideas are on the right track, gender (in the languages I have considered) is never autonomously interpretable. 5. Number While there is little disagreement on the idea that number is interpretable on DPs, many unresolved issues remain. What is the source of number, in a functional projection view of the DP? Is there a single feature for number, or more than one, and in the second case, which one(s) are relevant for subject-verb number agreement? Can the number feature(s) entirely determine verbal agreement? Heycock and Zamparelli (2005) (henceforth HZ05) propose that at least two number features for number are necessary within DPs. One, ±plur, stands for morphological plurality, and is normally contributed by the plural affix. The other, ±latt, signals presence or absence of what could be called “denotational plurality” – whether the denotation of a nominal is structured as a semilattice or not. Count plurals and masses are both +latt, but their plurality is formed by distinct semantic operations, so that only pluralities formed from count nouns have atomic elements. The latt feature has a syntactic reflex: only +latt nominals can appear without a determiner after verbs and prepositions in continental Romance and Germanic languages. On the other hand, only +plur nominals trigger plural agreement on determiners and adjectives. The source of the two features is also 23. The restrictive role of proximity and person is self-evident from take this, not that, or I want you, not him. The restrictiveness of number is harder to test, since when contrasting a plural with a singular antecedent, the use of a plural pronoun can easily extend to the totality of objects, singulars included.

 Roberto Zamparelli

different: plur comes from the noun, latt, from an abstract functional head above the NP proper, called Pl, assuming the structure in (40). (40) a. [DP Det [NumP Num [PlP Pl [NP (Modifiers) Noun (Modifiers) (Compl)]]]] b. [DP Those [NumP few [PlP Pl [NP linguistic papers]]]]

After Agree, latt and plur have spread to every element within the DP, obtaining the following distribution for mass, count and pluralia tantum: (41)

a b c d

water+latt, −plur car−latt, −plur cars+latt, +plur scissors−latt, +plur

collection of water parts with no atomic elements a set of cars a collection of car sets a set of scissors

The system set up in HZ05 completely abandons the idea that unvalued features cannot be interpretable. Nouns are valued for plur and unvalued for latt; the functional head Pl is valued for latt and unvalued for plur, yet both change their denotation depending on the values they receive from each other via agree, generating the denotations indicated informally in (41).24 With this choice of features, I now turn to the nature of the agreement that appears in Romance on finite and non finite V and predicative adjectives (collectively, “Pred agreement”).25 (42) Imasc,plural bambinimasc,plural sonoplural rimastimasc,plural contentimasc,plural the children are been glad ‘The children have been glad’

For subject-verb agreement, there are three theoretical possibilities: either the agreement is strictly syntactic (as a pure Agree-based system would lead us to

24. The situation is analogous to the cases discussed in Pesetsky & Torrego (2007) for sentences: t, just like plur, is morphologically marked at V, but interpretable, though unvalued, at the Tns (Tense, formerly Infl) node. 25. I set aside the case of predicate nominals, where the apparent agreement between the predicate and its subject breaks down for gender (except in the case of “profession nouns” such as dottore “doctor”, attrice “actress”, which do not require an article as predicates; see Zamparelli 2005 and the references cited there) and number, where mismatches are possible in the case of collectively interpreted predicates (ia). (i) a.

[Carla e Maria]fem,plur sono unmasc,sing grossomasc,sing problemamasc,sing. Carla and Maria are a big problem

b. Carla e Maria sono {bambine/*una bambina/*attrice}. Carla and Maria are {children/a child/actress} I will take the problem of una bambina “a child” in (ib) to be falsity, not ungrammaticality.

On the interpretability of φ-features 

expect); or it is semantics-based, or both. Let us consider the various possibilities in turn. 5.1 Pure syntactic agreement In a pure syntactic system with two number features, the first issue to resolve is which one should play a role for verbal agreement. The answer is that it must be the more “syntactic” one, plur, since both plurals and singular mass nouns are +latt, but the verb is plural only with the former.26 Since I know of no morphological verbal affix common to singular mass and plural count nouns, I will concentrate on the role of plur. In first approximation, the role of this feature in verbal agreement could be stated as in Rule (43): (43) Syntactic Agreement: a verbal predicate is plural iff its subject is +plur.

Counter examples to the idea that agreement is syntactic abound, but in my opinion many do not undermine the mechanism of verbal agreement itself, but simply point to the need for a less naïve view of the elements plur might be associated with. A first set of cases concerns the possibility of having singular agreement with strings of elements that contain plural nominals: e.g., (44). (44) [King prawns cooked in chili salt and pepper] was very much better, a simple dish succulently executed. (Biber et al. 1999: 187)

In my view, such cases simply mean that (as pointed out in the Di Sciullo and Williams 1987 for examples like a don’t-look-at-me-that-way expression) sometimes whole strings can be reanalyzed as (temporary) lexical items (“named entities”), which receive a singular interpretation by default. Common noun interpretations for phrases can also be observed in e.g. (45) Mr. “I would never leave you” is back.

Technically, this could be accomplished by positing a covert semantic operator from phrases to named entities/titles. Other cases of unexpected singular agreement arise with metonymy (46) and measure phrases (47). (46) a. The hash browns at table nine are/*is getting cold. (Pollard & Sag 1994) b. The hash browns at table nine is/*are getting angry.

26. This does not exclude the possibility that, in principle, the verb might be sensitive to latt as well, adapting its denotation in response to the latt value it receives. Bianchi (2006) makes use of HZ05’s PlP at the VP, to obtain event pluralization. Semantically, however, it is perfectly possible for the verbal pluralizer to be always active, in contrast with the nominal one.

 Roberto Zamparelli

(47) a. [Five pounds] is/*are a lot of money. b. [Two drops] deodorizes/*deodorize anything in your house. c. [Fifteen dollars] in a week is/*are not much.

Again, these counter examples are not problematic if one is willing to allow in syntax “silent nouns” like the ones proposed in Kayne (2005): (48) a. [MP five pounds] AMOUNT is not much. b. [DP the [ hash browns] PLATE]… c. [DP the [ hash browns] RECIPIENT]…

Evidence for this solution comes from Italian, where the agreement is optionally plural or singular. If singular is chosen, gender must be the default masculine, not the feminine gender of sterline “pounds”. (49) [Cinque sterlinefem,plur] {è moltomasc,sing /sono moltefem,plur /*è moltafem,plur} [5 pounds] {is much /are many /is much}.

Another large class of apparent agreement mismatches can be found with the internal or the external nominal in pseudo-partitives (50) (Selkirk 1977) and the “kind” construction (51). (50) A number of objections were raised. (51) This type of tigers are almost extinct.

Here, too, the solution is to assume that the highest nouns, number and kind are not the heads of the DP, but modifiers in a spec position (at least in one of the possible structures; see Wilkinson 1995; Zamparelli 1998). As Carlson (1977) observes, there are dialects of English where the determiner agrees with the internal noun: (52) Those [kind] of cars

In conclusion, what these apparent counter examples show is that (43) can be maintained if revised as (53): (53) S yntactic Agreement (revised): a verbal predicate is plural iff the (overt or covert) head of its subject is +plur.

Where named entities created by conversion operators (e.g., (44)) are marked -plur only by application of the exceptional default value assignment rule, and the head of complex nominals can be an abstract or internal. 5.2 Coordination cases The most serious class of counter examples for a syntax-only agreement model is the case of coordinated subjects, which shows that the plural agreement on the verb cannot always be due to features from N, be they overt or covert.

On the interpretability of φ-features 

(54) [Carlo−PLUR e Marta-PLUR] sono+PLUR felici+PLUR [Carlo and Maria] are happy

In an attempt to save a syntactic solution, one could assume that [+plur] comes from and itself. A range of uses where and does not trigger plurality shows that this solution is not viable. First, in English, the conjunctions of two singular NPs under one D triggers singular or plural verb agreement (55a and b), depending on whether the conjunction denotes a set of single individuals who have both properties expressed by the NPs, or a set of pairs, each member of which has one of the properties. Following HZ05 I will call these the joint and the split readings. (55) a. [DP This/A [NP friend and colleague]of mine] arrives today. b. [DP This/A [NP boy and girl]] were in love with each other.

Joint Split

Crucially, even in the split reading, the determiner is obligatorily singular ((55b) vs. (16b)), showing that D, subject to strict internal agreement, follows the plur status of the Ns, while verb agreement seems to go with the semantic status of the whole DP. (55b) would be extremely difficult to explain if the required +plur feature was generated by and.27 Similarly, British English allows certain group nouns to trigger plural agreement on the verb, but these nouns still require a singular determiner. (56) {This/*These} committee are to meet tomorrow.

Even when coordination applies at the DP level plural verb agreement is not always triggered. Singular agreement is possible (or even forced) with conjunctions of quantificational determiners like those in (57a and b) in Italian. (57) a.

[Ogni ministro] ed [ogni sottosegretario] ha/?hanno firmato. [every minister] and [every secretary] has/?have signed

b. [Nessun docente] e [nessuno studente] ha/??hanno firmato. [no teacher] and [no student] has/??have signed

Even the conjunction of proper names or definites does not trigger plural agreement when one of the two conjuncts is identical to (or a subpart of) the other: (58) a. John and only John is allowed in here. b. This bomber and its cargo probably weighs over a hundred tons.

(Corbett 1994: 58) (Biber et al. 1999: 180)

27. Positing an underlying structure [This [[boy and girl] PAIR]] plus the assumption that PAIR behaves like committee in (56) would reduce the problem of (55) to that of (56), with the additional problem that (55) also holds in American English, while (56) is typical of British English. See HZ05 for arguments against an ellipsis analysis for (55) and related constructions.

 Roberto Zamparelli

Finally, as well known, no plural agreement arises from the conjunctions of sentential subjects (59), (61) or infinitive/participial subjects (60). (59) [[Che sia partito così tardi] e [che sia tornato [[That he left so late] and [that he has returned

subito]] immediately]]

è/*sono strano. is/*are strange

(60) [Partire presto] e [tornare tardi] è/*sono tipico/*i di Mario. [leaveINF early] and [returnINF late] is/are typical of Mario Cf. ‘Leaving early and returning late is/*are typical of Jack’ (61) The fact(??s) that [John left early] and [Mary returned so late] (puzzled me).

In the other direction (a VP which becomes plural with a formally singular subject) the Bantu comitative example seen above in (18b) is problematic for a syntactic route to plurality, since the elements that should form the plural argument of the reciprocal do not share the subject position. Taken together, these observations falsify the hypothesis that a syntactic feature for plurality might come from the conjunction operator. Rather, the semantics of the elements that are conjoined plays a crucial role. 5.3 Semantic agreement A completely different possibility is that Subject-Pred agreement is entirely semantic. Since verbs are not plural with mass subjects, the notion of “plurality” relevant here requires a semilattice denotation built over atomic elements (masses will either have no atomic elements in their denotations, as in Link 1983, or will have “atoms” which are not directly accessible, as in Chierchia 1998). (62) S emantic Agreement: Pred takes +plur agreement iff its subject denotes a proper plurality (a lattice with more than one atom)

A pure semantic approach to agreement has a wider empirical coverage than a pure syntactic one. (55) follows straightforwardly, since the (a) case will denote an individual, the (b) case a pair. (58) is also not a problem, since the denotations of the two conjuncts will collapse into a single element. The problem of plural with group nouns in British English requires a semantic operator that maps a group onto the plurality of its members: (63) This[OP 〈〈 gt 〉 〈〈 et 〉 t〉〉 committee] have met.

British English

This solution is not open for the Ikalanga comitatives in (18)b, but a semantic theory of how comitative constructions are interpreted (so that Russell wrote

On the interpretability of φ-features 

“Principia Mathematica” with Whitehead entails Russell and Whitehead wrote “Principia Mathematica” (together)) would have immediate positive consequence for the puzzle in (18)b. Overall, these cases require further study. Returning to Romance, (59) and (60) follow if we assume that the denotations of CPs and nominalized infinitives (be they propositions, facts or events) do not form distinct atomic elements when conjoined, but combine to yield a larger proposition, fact or event, behaving essentially as the conjunction of abstract mass nominals at sub-DP level. If this is on the right track, the singular agreement of (61) has to be put in relation with the singular agreement in (64) (NP conjunction), and the plural agreement of Spanish (65) (with an article selecting the CP) is related to the plural in (66). The structure of (59)–(60) might thus be [DP D [CP and CP]], with D empty. (64) La sua [pazienza e determinazione] era/?erano nota a tutti. thesing his [patience and determination] was/?were known to everybody ‘His [patience and determination] was known to everybody’ (65) [El che haya claudicato] y [el che se retire] no me [the that he_has backed] and [the that CL he_withdraws] not me importan. botherplur

‘That he has backed down and that he is withdrawing doesn’t bother me.’

(66) [La sua pazienza] e [la sua determinazione] ?è/sono proverbiale. [the his patience] and [the his determination] ?is/are well-known. ‘His patience and his determination ?is/are well-known’

More problematic are conjunctions of quantificational DPs like (57), where we could tentatively hypothesize an LF where the two quantifiers have raised across the board from a DP internal position (67), to quantify over the individuals contained in the pairs.28 (67) [DP Ognii [[NP ti ministro] and [NP ti sottosegretario]]] ‘λP∀x such that x is a member of the pair formed by one minister and one subsecretary, P(x)’

The challenging cases for a semantics-only approach are, of course, those that came most naturally in a syntactic approach. One is dualia tantum like forbici “scissors” in (68).

28. HZ05 gives evidence for the DP-internal origin of singular quantifiers. In Italian qualche “some”sing can appear after an indefinite determiner, while every can notoriously appear after a possessor phrase in English (John’s every wish, see Abney (1987).

 Roberto Zamparelli

(68) Le forbicifem,plural sono apertefem,plural the scissors are open.

These cases might still be regarded as cases of genuine plurality, if we take scissors to denotes pairs of scissor-halves, with an invisible dual operator PAIR-OF associated to the lexical entry. (69) a. scissors = PAIR-OF(scissor-halves) b. two scissors = two(PL(PAIR-OF(scissor-halves))) PL = pluralization operator

A more serious challenge to the semantics-only hypothesis comes from cases where the plurality on the predicate does not seem to correspond to any real multiplicity of atoms. Apart from the Latin or Greek plural place names, exemplified in (8), problems arise with mass nouns which seem to be syntactically plural, like English goods, groceries, greens, politics, etc.; It. viveriplur, vettovaglieplur “food supplies” viscereplur “bowels”. Although it may be difficult to draw a line between scissors and these cases, the latter examples seem to behave as mass nouns in not making available atomic parts that can be counted via cardinality modifiers (70) (except by selecting alternative meanings: three groceries = three grocery shops; two goods = two kinds of goods, etc.)29 (70) {molte/*tre} vettovaglie /viscere {many/three} food_supplies/bowels

However, they have the morphology of plural count nouns. Since singular mass nouns induce singular predicate agreement, plural ones ought to do the same. But they do not: (71) a.

I pochi viveriplur sono stati divisiplural the few food supplies have been divided

A different problem is raised by rational numbers, all requiring +plur on the predicate. (72) a.

In media, ogni mese 1.0 domandeplur,fem vengonoplural accolteplural,fem. on average, every month 1.0 applications are accepted

b. Questo mese, zero domande sono state accolte this month, zero applications haveplural been accepted.

29. A related problem, which will not be considered here, is the frequent observation that in some languages that have neuter gender, plural neuters tends to trigger singular agreement. This is sometimes cast in the form of a constraint against the combination of plural and neuter feature values (see Corbett 2000), but it could also be due to a “massifying” effect of neuter gender, and thus be reduced to (66). For lack of space I will not address this problem here.

On the interpretability of φ-features 

The problem, in this case, is not so much agreement, but a theory on what denotation should be appropriate for fractions of the objects normally denoted by count nouns. Yet, a requirement for such a theory would be, presumably, that 1.0 applications be coextensional with one application, so a plural V is unexpected. Finally, a semantic account of agreement has to face the challenge of the use of plural pronouns for singular formal address (the vous2P-plur or voi2P-plur for 2nd person singular of French and Southern Italian; see (10) above).30 Facts of this sort suggest that the best empirical fit might be a hybrid syntax/ semantics approach to agreement (as also proposed in Kim 2003). (73) Hybrid Agreement: Pred agreement is singular if the subject denotes a single atomic value (variable or constant) and carries [−plur] feature specification. In all other cases, Pred agrees in the plural.

We can make some sense of this conjunctive definition by observing that syntactic singularity and atomic denotations are, respectively, the unmarked notions for the syntax and semantics of quantity. Plural marking on verbs, then, can be seen as a departure from the “perfect” unmarked situation (singular), triggered by the presence of syntactic or semantic markedness on the subject. This leads to the following reinterpretation of (73): (74) Markedness-Spreading Agreement Rule: Pred takes the marked number value whenever its subject takes a marked value for syntactic or semantic number.

Where any application of the pluralizer (in the system of HZ05) creates a semantically marked number. A mixed syntax/semantics rule like (74) or (73) is, of course, unexpected in an Agree driven system (including that of Pesetsky & Torrego 2007), where verbal agreement depends on the presence of purely syntactic features, but it might

30. Wechsler (2004) proposes an ingenious solution to this problem: vous/nous are actually not plural forms of je/tu, but different persons (1a, 1s, 2a, 2s). Finite V agrees with them in person, not number. Non finite predicative material, which normally agrees in number with 3pers element, must in this case resort to semantic agreement. This proposal, which does not undermine my claim that part of agreement is semantic, faces two problems. One is why verb agreement with the vous-person has the same aspect as genuine plural agreement, and not something else. A second, empirical problem (also noted by Bianchi 2005) is that authorial “we” (and the “we” in plurale maiestatis) patterns like vous in French, but like “scissors” in Italian. (79) a. b.

Noi siamo sempre stat-i/*-o fedel-i/*-e alla grammatica generativa. we aux always beenplur/*sing faithfulplur/*sing to generative grammar Nous avons toujours été loyal envers la grammaire générative. we aux.1pl always been loyal.sing toward the grammar generative

Such cross-linguistic differences seem to require a finer classification.

 Roberto Zamparelli

eventually be the only choice. It is of course possible to stipulate that “multiple atom denotation” corresponds to a novel syntactic feature and that this is what triggers agreement, but this choice must present empirical evidence that the new feature behaves as other grammatical features are known to behave, and not as a pure restatement of a semantic fact. I can think of no syntactic reason apart from agreement to distinguish between the subjects of (75a) and (75b).31 (75) a. [John and only John] has/*have taken a decision. b. [John and only Mary] (i.e., not Sue, Olga, Sherry and Lue) *has/have taken a decision.

6. Conclusions and further work The conclusion that can be drawn from this review of the φ-features is that the link between valuation and interpretability proposed in Chomsky (2000), (2001) is at best highly idealized. The idea that valued features are always interpretable seems to break down in the case of gend. In the domain of number (num/plur), V and other predicates seem able to acquire a value from syntax or semantics. Last but not least, the management of defaults discussed in connection with the pers feature might require modifications of the Agree system. This set of mostly negative conclusions should naturally be followed, on a more constructive side, by work to establish a new and better connection between meaning and feature values. The system set up in Pesetsky and Torrego (2007) is a step in the right direction, but possibly more radical changes are needed. At the empirical level, an interesting question I hope to address in future research is whether the discussion on the status of gender extends to languages with a wider set of nominal classes, like the Bantu languages. Assuming that Bantu classes can be treated on a par with gender markings on nouns, with prefixes bearing only number information (Carstens 1993), we can ask whether the provision of a form of classification (albeit often semantically opaque) over the domain of nominals can be regarded as a form of ‘interpretability’, and if so, why. I also hope to be able to

31. Of course, it is possible to hold the view that all linguistic semantics can be represented by means of syntax-like features (as a reviewer suggests). I do not find this solution convincing. A feature systems to represent meanings should have the expressive power of second order logic, at least. Moreover, it would raise the question of how the value of these features is set as the semantic derivation unfolds: essentially, we would trade the problem of how verbal agreement can interface with semantics for the problem of how these new features interface with denotational semantics. Finally, if semantic features of this sort exist, we should ask why internal agreement seems to be largely insensitive to them (cf. the ungrammaticality of (16b)).

On the interpretability of φ-features 

test the behavior of Bantu languages with null-nominal constructions and gender markers, to see whether their behavior differs from that of a language like Italian.

References Abney, S.P. 1987. The English Noun Phrase in its Sentential Aspect. Ph.D. dissertation, MIT. Adger, D. 2003. Core Syntax: A Minimalist Approach. Oxford: OUP. Adger, D. & Harbour, D. 2003. The syntax and syncretisms of the person case constraint. In MIT Working Papers in Linguistics, K. Hiraiwa & J. Sabbagh (Eds). Cambridge MA: MIT. Anagnostopoulou, E. 2003. The Syntax of Ditransitives. Evidence from Clitics. Berlin: Mouton de Gruyter. Asudeh, A. & Toivonen, I. 2006. Symptomatic imperfections. Journal of Linguistics 42(2): 395–422. Babby, L.H. 1988. Noun phrase internal case agreement in Russian. In M. Barlow & C. Ferguson (Eds). Barlow, M. & Ferguson, C.A. (Eds).1988. Agreement in Natural Language: Approaches, Theories, Descriptions. Stanford CA: CSLI. Benveniste, É. 1946[1966]. Structure des relations de personne dans le verbe. Problèmes de linguistique générale I: 225–257. Bianchi, V. 2003. On finiteness as logophoric anchoring. In Temps et point de vue/Tense and Point of View, J. Guéron & L. Tasmovski (Eds), 213–246. Paris: Université Paris X, Nanterre. Bianchi, V. 2005, February. The person asymmetry: Underspecification of person and number features? Universiy of Siena, handout of a talk at “Underspecification in morphology and syntax”, Cologne. Bianchi, V. 2006. Number agreement and event pluralization: A case study. In Phases of Interpretation, M. Frascarelli (Ed.). Berlin: Mouton de Gruyter. Biber, D., Johansson, S., Leech, G., Conrad, S. & Finegan, E. (1999). Longman Grammar of Spoken and Written English. London: Longman. Borer, H. 2004. Structuring Sense: In Name Only. Oxford: OUP. Brody, M. 1997. Perfect chains. In Elements of Grammar, L. Haegeman (Ed.), 139–167. Dordrecht: Kluwer. Carlson, G.N. 1977. Reference to Kinds in English. Ph.D. dissertation, University of Massachusetts at Amherst. Carstens, V. 1993. On nominal morphology and DP structure. In S.A. Mchombo (Ed.). Chierchia, G. 1998. Plurality of mass nouns and the notion of “semantic parameter”. In Events and Grammar, S. Rothstein (Ed.), 53–104. Dordrecht: Kluwer. Chomsky, N. 1995. The Minimalist Program. Cambridge MA: The MIT Press. Chomsky, N. 2000. Minimalist inquiries: The framework. In Step by Step, R. Martin, D. Michaels & J. Uriagereka (Eds). Cambridge MA: The MIT Press. Chomsky, N. 2001. Derivation by phase. In Ken Hale: A Life in Language, M. Kenstowicz (Ed.), 1–52. Cambridge MA: The MIT Press. Chomsky, N. 2004. Beyond explanatory adequacy. In Structures and Beyond, A. Belletti (Ed.), 104–131. Oxford: OUP. Corbett. 2000. Number. Cambridge: CUP. Corbett, G. & Fraser, N. 1993. Network morphology: A DATR account of Russian nominal inflection. Journal of Linguistics 29: 113–142.

 Roberto Zamparelli Corbett, G.G. 1991. Gender. Cambridge: CUP. Corbett, G.G. 1994. Agreement. In Encyclopedia of Language and Linguistics, 54–60. Oxford: Pergamon. Corbett, G.G. 2006. Agreement [Cambridge Textbooks in Linguistics]. Cambridge: CUP. Delfitto D. in press. A case-study on grammar, meaning and linguistic diversity: the nature of person features. In P.M. Bertinetto (Ed.) Atti del Convegno Annuale (2006) della Società Italiana di Glottologia. Roma: Il Calamo. Deprez, V. 2005. Morphological number, semantic number and bare nouns. Lingua 115(6): 857–883. Di Sciullo, A.-M. & Williams, E. 1987. On the Definition of Word. Cambridge MA: The MIT Press. Fox, D. 1999. Economy and Semantic Interpretation. Cambridge MA: The MIT Press. Giusti, G. 1992. La Sintassi dei Sintagmi Nominali Quantificati: Uno studio comparativo. Ph.D. dissertation, University of Venice. Giusti, G. 1999. The functional structure of noun phrases: A bare phrase structure approach. University of Venice Working Papers in Linguistics 9(1–2). Greenberg, J. 1966. Language Universals, with Special reference to Feature Hierarchies. The Hague: Mouton. Harley, H. & Ritter, E. 2002. Structuring the bundle: A universal morphosyntactic feature geometry. In Pronouns – Grammar and Representation, H.J. Simon & H. Wiese (Eds), 23–39. Amsterdam: John Benjamins. Harris, J. 1991. The exponence of gender in Spanish. Linguistic Inquiry 22(1): 27–62. Heim, I. & Kratzer, A. 1998. Semantics in Generative Grammar. Oxford: Blackwell. Heycock, C. & Zamparelli, R. 2005. Friends and colleagues: Plurality, coordination, and the structure of DP. Natural Language Semantics 13: 201–270. Hornstein, N., Nunes, J. & Grohmann, K.K. 2005. Understanding Minimalism: An introduction to Minimalist Syntax. Cambridge: CUP. Kayne, R.S. 2005. Movement and Silence. Oxford: OUP. Kim, J.-B. 2003. Hybrid agreement in English. In The Role of Agreement in Natural Language: TLS 5 Proceedings, W.E. Griffin (Ed.), 59–72. Texas Linguistics Forum 53. Kiparsky, P. 1982. Lexical phonology and morphology. In Linguistics in the Morning Calm, I.-S. Yang (Ed.). Seoul: Hanshin. Kratzer, A. 2008. Building a Pronoun. Fake Indexicals as Windows into the Properties of Bound Variable Pronouns. MS, University of Massachusetts at Amherst, to appear in Linguistic Inquiry. Lapointe, S.G. 1988. Toward a unified theory of agreement. In M. Barlow & C. Ferguson (Eds), 67–88. Lehmann, C. 1988. On the function of agreement. In M. Barlow & Ferguson, (Eds) 55–65. Letsholo, R.M. & Safir, K. 2007. Ikalanga anaphora sketch. (Vers. 1.1; Ms., URL: http://www.africananaphora.rutgers.edu/Ikalanga01.1.pdf). Link, G. 1983. The logical analysis of plurals and mass terms: A lattice-theoretical approach. In Meaning, Use, and the Interpretation of Language, R. Bäuerle, C. Schwarze & A. von Stechow (Eds), 302–323. Berlin: de Gruyter. Longobardi, G. 2006. Reference to individuals, person, and the variety of mapping parameters. MS, to appear in Essays on Nominal Determination, A. Klinge & H. Müller (Eds), John Benjamins. Matushansky, O. 2006. Head movement in linguistic theory. Linguistic Inquiry 37(1): 69–109.

On the interpretability of φ-features 

Mchombo, S.A. (Ed.). 1993. Theoretical Aspects of Bantu grammar. Stanford CA: Leland Stanford Junior University. Pesetsky, D. & Torrego, E. 2007. The syntax of valuation and the interpretability of features. In Phrasal and Clausal Architecture: Syntactic Derivation and Interpretation, S. Karimi, V. Samiian & W. Wilkins (Eds). Amsterdam: John Benjamins. Pollard, C. & Sag, I.A. 1994. Head-Driven Phrase Structure Grammar. Chicago IL: University of Chicago Press & CSLI Publications. Schlenker, P. 2003. A plea for monsters. Linguistics and Philosophy 26: 29–120. Selkirk, E. 1977. Some remarks on noun phrase structure. In Formal Syntax, P. Culicover, T. Wasow, & A. Akmajan (Eds), 285–316. San Diego SA: Academic Press. Sells, P. 1985. Lectures on Contemporary Syntactic Theories. An introduction to GovernmentBinding theory, Generalized Phrase-Structure Grammar and Lexical-Functional Grammar. Stanford CA: CSLI. Watanabe, A. 2000. Feature copying and binding: Evidence from compementizer agreement and switch reference. Syntax 3: 159–181. Wechsler, S. 2004. Number as person. In Empirical Issues in Syntax and Semantics 5 (on-line Proceedings of the Fifth Syntax And Semantics Conference In Paris), O. Bonami & P. Cabredo Hofherr (Eds), 255–274. Wilkinson, K. 1995. The semantics of the common noun kind. In The Generic Book, Ch. 10, G. Carlson & F. Pelletier (Eds), 383–397. Chicago IL: University of Chicago Press. Zamparelli, R. 1998. A theory of Kinds, Partitives and OF/Z Possessives. In Possessors, Predicates and Movement in the Determiner Phrase, A. Alexiadou & C. Wilder (Eds), 259–301. Amsterdam: John Benjamins. Zamparelli, R. 2005. Bare predicate nominals in romance languages. MS, to appear in Essays on Nominal Determination, A. Klinge & H. Mueller (Eds). Amsterdam: John Benjamins.

Agreement and concord in nominal expressions1 Giuliana Giusti

Università Ca’ Foscari Venezia

This paper claims that feature sharing should be analyzed as the result of at least two different processes, which are named here Agreement and Concord, and inquires how these two processes are manifested inside nominal expressions (NEs). Agreement is the transfer of the Person features of the possessor (the “subject” of the NE) onto some functional head (parallel to subject Agreement in the clause) with the effect that (Genitive) Case is assigned. On the contrary, Concord is the transfer of Number, Word Class, and Case specifications from a functional head onto a modifier, which is first-merged as a Specifier of that functional head. The claim is that, quite differently from Agreement, Concord arises from the merger of a modifier, underspecified for uninterpretable features, in the specifier of a functional head, carrying a copy of those features. In other words, Concord is directly enhanced by the Spec-Head configuration; it does not involve merger of a probe which targets a goal and, as a consequence, never triggers (overt or covert) movement. This proposal can dispense with a number of otherwise unmotivated movements and can derive the different properties of these two kinds of feature sharing phenomena. The argument is supported by observing macro-parallelisms across Bantu and Romance languages, in particular Swahili and Xhosa on the one hand and Romanian and Italian on the other hand.

1. Aims and structure of the paper Feature sharing is a pervasive property of natural languages which has long been considered quite puzzling. Being the source of redundancy, it is a crucial issue in the minimalist program which aims to reduce language to a conceptually necessary system (cf. Chomsky 2005). In this paper, I claim that, contrary to what is currently assumed in the literature, feature sharing is a non-homogeneous phenomenon and . I thank the audience of the Bantu-Romance Connection for criticism and support. I also thank Josef Bayer, Petra Bernardini, Guglielmo Cinque, Peter Cole, Carlo Geraci, Lena Dal Pozzo, Liliane Haegeman, Gabriella Hermon, Megan Rae, and Iulia Zegrean for comments on a previous version. I am indebted to the anonymous reviewers for constructive criticism. All errors are obviously mine.

 Giuliana Giusti

must be differentiated in at least two notions: Agreement and Concord. I limit my inquiry to how these two notions are manifested inside nominal expressions (NEs):2 Agreement is the transfer of the Person features of the possessor (the “subject” of the NE) onto some functional head (parallel to subject Agreement in the clause), with the effect that (Genitive) Case is assigned; by contrast, Concord is the transfer of Number, Word Class, and Case specifications of the NE to adjectival modifiers.3 In the minimalist program, feature transfer is reduced to checking and deleting uninterpretable features. I follow mainstream literature in assuming that the Agreement relation arises from the operation Agree, namely merger of a formal uninterpretable feature (a probe) in the extended projection of N, and is checked against a constituent (the goal), in a lower specifier, which contains the interpretable counterpart of those features. This triggers movement (copy or re-merge) of the relevant features to obtain a Spec-Head configuration with the probe, resulting in covert movement (only the features move) or overt movement (the moved feature pied-pipes the whole constituent). Here I claim that, quite differently from Agreement, Concord arises from the first-merger of a modifier with uninterpretable features. In other words, Concord is directly enhanced by the Spec-Head configuration, it does not involve merger of a probe targeting a goal, and never triggers overt or covert movement. The proposal outlined here is inspired by Grimshaw’s (1991) notion of “extended projection” as a chain of heads sharing all features. It complies with minimalist assumptions that formal features are interpreted once but have copies which serve syntactic licensing. The notion of “head” used here is compatible with Matushansky’s (2006: 70) recent definition of head as “a syntactically indivisible bundle of formal features”. I will try (with difficulty) to avoid the discussion of two currently competing analyses of the syntax of NE: namely the traditional headmovement and the more recent phrasal movement initiated by Kayne (1994) and developed by Shlonsky (2004); Cinque (2003, 2005, 2007) and Laenzlinger (2005),

. Mila Dimitrova-Vulchanova (p.c.) suggests the term “nominal expression” or NE to avoid ambiguity with the label DP. NE is parallel to “clause” (which is no longer a label in the tree), in that it refers to the entire nominal constituent with no commitment to the actual label of the highest projection. . I use the terms “agreement” or “concord” (in lower case) as basically synonyms referring to the pre-theoretical notion of feature sharing. I will use “Agreement” and “Concord” (with initial capital letter) to refer to the two syntactic relations I am studying. “Agree” is instead the rule of grammar that enhances the Agreement relation, involving a probe and a goal and often resulting in movement. Agree is the probe.

Agreement and concord in nominal expression 

among many others. For reasons of simplicity, I develop the proposal in a headmovement approach but the theoretical point at issue here is independent. Along with the rest of the contributions in this volume, the empirical domain of the paper is provided by the two unrelated families of Romance and Bantu. As for all natural languages, micro-variation is found in strictly related languages, while macro-parallelisms can be drawn between two languages belonging to unrelated families. The latter approach is less frequently entertained than the former, but I think it may provide unbiased perspectives in long-standing problems.4 My contribution seeks to offer a novel perspective into some recently studied phenomena in Bantu and Romance, whilst at the same time making the theoretical point of a differentiation between Agreement and Concord. Among Romance languages, I concentrate on Romanian, which presents interesting parallels with Swahili (mainly taken from Carsten’s work mentioned in the references), and Xhosa (taken from Visser 2006). The richness of concord on possessors and modifiers is studied as parallel to prefixes in Swahili and Xhosa. The richness of the determiner system in Romanian (and Spanish) is shown to be parallel to pre-prefix spreading and multiple demonstratives in Xhosa. As for all the work in the generative enterprise, the ambition is that the general principles discussed here hold for all natural languages, subject to different parameters which derive all and only possible variation. Within the limits of a paper, I try to give an idea of possible parameters. Section 2 introduces the general proposal. Section 3 argues that, parallel to what happens in the clause, the syntactic operation Agree merges a probe which targets the Person features of an argument of the NE. There I also observe that in both Romance and Bantu, Person features are weak and do not generally piedpipe the whole (Genitive) argument to the specifiers of the probe, unless the goal is exclusively formed by the probed features (namely, in the case of a pronoun). Section 4 supports the distinction between Agreement and Concord by showing that they can co-occur on one and the same element, namely the genitival article in Romanian and connectors in Bantu. Section 5 further specifies the notion of Concord and discusses its application to pre-prefix spreading in Bantu and double definiteness in Romanian. Section 6 draws some conclusions.

. I want to thank the organizers of the workshop for having required the comparison between Romance & Bantu, which forced me to start approaching this immense empirical domain. Due to lack of time and of native speakers, the Bantu data are all taken from previous literature.

 Giuliana Giusti

2. Two different structural relations Romance and Bantu languages are particularly rich in both manifestations of feature sharing: the transfer of the features of an argument onto the predicate, as in (1a) and in all sentences of (2); and the transfer of the features of a head (usually a noun) onto its specifiers (usually adjectives), as in (1b) and (2b–c): (1) a.

Le ragazze mangiano il pesce. the girls.fem.pl eat.3p.pl the fish.masc.sg ‘The girls eat fish.’

(Italian)

b. Le belle amiche italiane di Maria. the.fem.pl beautiful.fem.pl friends.fem.pl Italian.fem.pl of Mary ‘Mary’s beautiful Italian friends.’ (2) a.

Wa-toto wa-nakula mkate. cl2.child cl2.eat fish ‘The children eat fish.’

b. M-toto m-dogo a-mefika. cl1.child cl1.little cl1.arrived ‘The little child arrived.’ c.

(Swahili, Barlow & Ferguson 1988: 3)

(Swahili, Katamba 2003: 111)

Ki-kapu ki-dogo ki-mefika. cl7.basket cl7.little cl7.arrived ‘The little basket arrived.’

This pervasive phenomenon contradicts general principles of economy at both interfaces. In fact, it is not clear why some features must redundantly appear on more than one element, and how the interpretive component can recognize this redundancy as such and, moreover, tolerate it. This is why feature sharing and feature spreading have always been the subjects of much debate. In the principles-and-parameters theory, agreement was assumed to take place in dedicated projections (AgrP) in the extended projections of clauses and nominal expressions (cf. Belletti 2001 for an overview and a defense of this kind of approach). In the minimalist program (Chomsky 1995 and much following work), which eliminates X-bar theory and labels, a proliferation of AgrPs with various specifications (AgrObj, AgrSubj, etc.) is problematic and various proposals have been put forth to derive agreement from independent principles (cf. Hornstein, et al. 2005: ch. 9, for a discussion). In previous literature, “agreement” and “concord” are traditionally synonyms (Barlow & Ferguson 1988: 1). The core notion of Agreement is subject/predicate agreement, and feature sharing inside the nominal expression has often been reduced to current theories for subjectpredicate agreement. For example, Bosque & Picallo (1996) (for Spanish) extend

Agreement and concord in nominal expression 

the AgrP-formalism of the principles-and-parameters framework; Carstens (2000) (for Romance & Bantu) applies the Move-F approach in the minimalist framework to concord between noun and adjectives (and also prepositions in Bantu). In this paper, I challenge this unification approach and claim that feature sharing is related to (at least) two different structural relations, namely “subject of ” and “modifier of ”, which I label Agreement and Concord respectively. Chomsky (2000, 2001, 2004) proposes that “Agree” is an operation of the computational system, independent of Move, which involves a probe (a head with uninterpretable features) and a goal (a constituent in its c-commanding domain with matching interpretable features). The Agree operation does not therefore necessarily involve a Spec-Head configuration and can be satisfied in situ as in (3a):

(3) a.

TP Spec

T' Tprobe [uφ]

vP Spec NEgoal [iφ]

Agree

v' ...

Agree applies to delete the uninterpretable Person features of T against the interpretable Person features of the highest argument NEgoal in the vP. This operation affects the argument by assigning it Nominative Case. Movement of the goal to SpecTP is subject to parametrization. The probed features can in fact either move alone (covertly) or pied-pipe the entire constituent (Hornstein et al. 2005: ch. 9), as in (3b):

TP

(3) b. Spec NE[iφ]

T' T° [uφ]

Move

vP NE[iφ]

In section 3, I apply the same process to NEs, but let us first address adjectival concord in (4). I adopt Cinque’s (1994) proposal that APs are directly merged into functional specifiers, as reformulated in a minimalist fashion by Giusti (2002). I find no reason either to assume the presence of a probe with an uninterpretable

 Giuliana Giusti

feature targeting one or more APs, or to merge APs in a low position and later move them to higher specifiers. In (4a), the head F is an interpretable copy of the φ-features of the NE (Number, Gender and Case) which move (with or without N, according to the language) up to D. FP is only created to allow for the AP to be merged in a specifier: 5

(4) a.

F2' F2° iφ

FP1 Spec APuφ

F1' F1°iφ

NP

Concord

N°iφ

Differently from Agreement, Concord does not take place in two steps. It simply holds of any Spec-Head configuration. If a possessive AP is first merged as SpecNP in which it receives a θ-role from N, Concord can also take place, as in (4b):

(4) b.

NP Spec PossAPuφ

N'

N°iφ Concord

Other crucial differences between Agreement and Concord are the following: (i) Agreement is obligatory in clauses (see the Extended Projection Principle) but optional in NEs; (ii) Concord (at least in the languages under consideration) . Another possibility would be to propose that adjectives are adjoined to NP or nP (Norbert Corver p.c.). This would dispense with merger of the functional head F0, later labeled Conc0, and therefore be more “economical”. It would also distinguish Concord from Agreement since Concord would occur between the adjunct and the head. However, I envisage two problems in this line of approach (proposed by Ritter 1991; Valois 2006 and reference mentioned there): (i) It must assume that adjunction must obey the adjectival hierarchy, contrary to the traditional assumption that adjunction is more free than merger into Spec; (ii) It must still provide at least one position for intermediate N-movement (which accounts for one prenominal and one postnominal adjective), thereby losing the economy of intermediate functional head only hosting uninterpretable features. I do believe that these problems may be overcome in the minimalist framework. However, for reasons of relevance and space, I will not attempt to do it here.

Agreement and concord in nominal expression 

only takes place in NEs but not in clauses; (iii) Agreement is instantiated by the semantic relation “subject of ” while Concord is related to modification. As for (i), it is reasonable to assume that the uninterpretable Person features of T must intersect with Time reference in order to fix the truth value of the proposition. It is less clear what Agreement is doing inside an NE. The answer could come from the observation that the reference of a possessor NE clearly interacts with the reference of the main NE. If we assume that the referential value of an NE containing a possessor requires a compositional interpretation of the reference of the two NEs, we straightforwardly derive the optionality of Agreement in NEs. In fact, an NE can have reference without having a possessor. But, if a possessor is present, its referential value must interact with the reference of the whole NE. So, a (provisional) reason why a possessor NE may be the goal of a probe in NE is that it must reach the position in which it is interpreted compositionally with the referential value of the main NE.6 As for (ii), differently from Agree, Concord is required by the inflectional properties of modifiers such as adjectives, which come from the lexicon with uninterpretable features. In other words, Concord is required by the inflectional morphology of the modifier and not of N. As for (iii), the notion “subject of ” is unique. Agreement must therefore occur only once in the relevant portion of structure (cf. the “freezing effect” discussed in Rizzi 2006). On the contrary, Concord can be iterated and appears not only on modifiers in the intermediate layer, but also on determiners, quantifiers (which belong to the topmost layer), and even on arguments (which belong to the lowest layer). This difference can be explained by the same relations advocated above: there is one and only one subject position, while modifiers are optional and can be more than one.

. Evidence for this is the well-known phenomenon of definiteness spreading, in which the (in)definiteness of the possessor influences the interpretation of the whole NE, as exemplified in (i), taken from Giusti (2002). The definite NE la segretaria di un onorevole refers to an indefinite (even if possibly unique) person; this is confirmed by the subjunctive mode on the verb which implies non-existence:

(i) Scommetto che non troverai mai [ne [la segretaria [pp di un onorevole]] che sia disposta a testimoniare contro di lui]. I bet you’ll never find the secretary of a congressman who be.subj willing to witness against him. In (i), it is clearly the indefinite nature of the PP which is responsible for the indefinite interpretation of the NE. Giusti (2002) takes this as evidence for covert movement of the possessor to SpecDP (cf. the discussion of (5b) below). This is in line with an analysis in terms of Agree with no movement. A covert probe in the high portion of the structure targets the Person feature of the possessor, which moves silently to the specifier of the probe.

 Giuliana Giusti

3. Agree and EPP in the NE Following the seminal work by Abney (1987), Ritter (1988) and much research after them (cf. among many others, Cornilescu 1995; Giusti 1996; Cardinaletti & Starke 1999; Bernstein 2001), an NE is assumed to display three main layers (5b) parallel to clausal structure (5a). In both kinds of syntactic objects, the three layers can be “split”, resulting in much larger structures:7 (5) a. [Clause CP-layer [IP-layer [VP-shell]]] b. [NE DP-layer [FP-layer [NP-shell]]]

Among the parallels, one can list the following: (i) both clausal and nominal structure establish semantic/thematic relations in the lexical layer (for the concept of VP-shell cf. Larson 1988; Carstens 2000 for its extension to the NP-shell); (ii) morphosyntactic requirements involving feature sharing are realized in an intermediate functional layer (IP, FP) in which the subject of the clause is also (re)merged; (iii) the peripheral layer (CP, DP) has the double function of hosting the elements which are valued for propositional/referential interpretation while at the same time satisfying the syntactic requirements of the superordinate structure. Agreement and Concord principally involve the intermediate layer, with obvious relations to the adjacent layers. In particular, Concord copies the (Number and Gender) features of the head noun in the extended projection of the NE, up to the DP layer. Agreement is triggered by uninterpretable Person features merged at the top of the intermediate layer and checked against the interpretable Person features present in the highest argument first merged in the NP-shell (or nP). The probe triggering Agreement is labeled Agree; the nominal expression in which it is projected is labeled NE, and the nominal expression targeted by Agree is labeled NEgoal. I maintain that in NE, Agree targets the Person features of the highest argument (NEgoal) merged in the NP-layer (nP, according to Carstens 2000), even if in many Romance and Bantu languages this does not have any overt counterpart in the inflectional morphology of NE (contrary to what happens in Ugro-Finnic and Turkish).8 In this section, I show that in Romance and Bantu, Agree triggers movement of NEgoal only if the latter exclusively consists of Person features. This results in the raising of pronominal possessives only. Possessors with lexical content stay low and are preceded by a P. Section 3.1 presents the different word orders arising in

. Notice that the label FP in (5b) is provisional. Other labels that have been used for FP in (5b) are AgrP or NumP. In the course of the paper, it will be revised into series of Conc(ord)P. . Cf. Szabolcsi (1994) for Hungarian; Kornfilt (1997: 185–6) for Turkish; Dal Pozzo (2008) for Finnish.

Agreement and concord in nominal expression 

Romance and Bantu NEs with full and pronominal NEgoals. Section 3.2 claims that these orders arise from the weak nature of the Person feature, which cannot piedpipe a full NE. Section 3.3 compares Romance and Bantu with Hebrew, which may present a possible challenge to this proposal. 3.1 The ordering of pronominal and full-NE possessors In the last three decades, research in DP structure9 has formulated the following empirical generalizations on the positioning of the possessor in Romance and Bantu:

(6) a. Full Genitive NEgoals appear embedded into a PP and located after postnominal adjectives; b. The order of these Genitive PPs is rather free in object-referring NEs; c. Possessive adjectives and possessive pronouns usually appear preceding prenominal adjectives; d. No more than one possAP is allowed e. A strict hierarchy is observed even in object-referring nominals as regards θ-role assignment to a possAP.

The embedding into a PP (6a) and the consequent freedom of order indicated in (6b) shows that a possessive PP is not strictly bound to a structural position. A dedicated unique position singles out possessives with pronominal reference as located in a specifier immediately c-commanded by D0 (6c)–(6d). The restriction operated by the hierarchy (6e) suggests that this unique position is derived by A-movement, (which does not apply to PPs). These observations, as well as the relevant data and literature, are rather well-known. For this reason, I limit the discussion to a brief overview of representative data. In (7a), the two argument PPs must appear after the postnominal AP sbiadita. In (7b), the possessive AP sua must precede the prenominal AP vecchia. In (7c), we observe that no other possessive AP can appear elsewhere, even if a postnominal position would in principle be possible (as in (10)–(11) below). This suggests the uniqueness of the Agreement relation (only one probe can be merged) as opposed to trivial uniqueness of the dedicated structural position: (7) a.

la vecchia fotografia sbiadita di Gina di Mario the old picture faded of Gina of Mario ‘Gina’s old faded picture of Mario’/‘Mario’s old faded picture of Gina’

b. la {sua} vecchia {*sua} fotografia sbiadita di Mario the (her) old (*her) picture faded of Mario ‘her old faded picture of Mario’/*“Mario’s old faded picture of her’

. Cinque (1980); Giorgi & Longobardi (1991); Carstens (1991) among the many others.

 Giuliana Giusti

c. *la mia {tua} fotografia {tua} / la mia fotografia di te the my (your) picture (your) / the my picture of you ‘my picture of you’

The glosses in (7a) show that there is some freedom in the relative order of PPs, contrary to what is found with possessive adjectives in (7b). The position of the possessive to the left of prenominal APs suggests movement of the possessive AP. The obedience to the thematic hierarchy (possessor > agent > theme) argues for merger of the nominal arguments in a hierarchical fashion in NP (or nP). Mutatis mutandis, Swahili behaves like Italian. According to Carstens (1991: ch. 3), all the properties in (6) also hold for Swahili. The only difference in word order is due to Swahili N-to-D movement: (8) a.

picha mpya ya Amira ya Hasan 9picture 9new 9of Amira 9of Hasan

[ N – ap – PP – PP] Carstens (1991: 100 (39a)) ‘Amira’s new picture of Hasan’/‘Hasan’s new picture of Amira’

b. picha {yake} mpya {??yake} ya Hasan 9picture 9her 9new 9her 9of Hasan

[ N – poss.pron – ap – PP] Carstens (1991: 101 (41a/b) ‘her new picture of Hasan’/‘*Hasan’s new picture of her’

c. *uharibifu wake wao / wao wake 14destruction 14his 14their / 14their 14his [N – poss.pron – poss.pron] Carstens (1991: 86 (11e))

Full Genitive PPs follow the AP in (8a), suggesting that they are in the NP layer or at least lower than the SpecFP where the AP is merged. The order of Genitive PPs is rather free, as is apparent from the gloss of (8a). Possessive pronouns precede prenominal APs, as in (8b), suggesting that they are also moved to this position. The possessive pronouns must respect the hierarchy, as the gloss of (8b) suggests. No two possessive Genitive pronouns can appear, as shown in (8c). The generalizations in (6) can be captured in (9), where the high functional head F, filled with the probe agree, targets the FF features of NEgoal (bottom arrow). If, and only if, NEgoal is a pronoun, there is movement of NEgoal to SpecFP; otherwise only the features FFi move covertly (upper arrow):

(9) [NE D [FP FFi [F°AGREE] [FP AP …. [NP [NEgoal FF]i ... N]]]]

The proposal in (9) raises two questions: (i) why are pronominal and adjectival possessives moved out of the lexical layer to the high portion of the intermediate

Agreement and concord in nominal expression 

layer? (ii) why can full PPs not be moved in the languages under consideration? The rest of this section deals with these two questions. I first argue that it is the FF of NEgoal to be probed. In Italian and Swahili this feature is not strong enough to pied-pipe the rest of the NEgoal (with or without a PP). Then, I turn to languages like Hebrew and Romanian where movement of a Genitive NEgoal appears to be possible, showing that neither N-movement to D (which is present in Swahili but also in Romanian and Hebrew) nor concord between the NE and the P head of its argument (which is a property that Romanian shares with Swahili) can derive movement of NEgoal. I provisionally conclude that the strength of FF cannot be reduced to independent properties of nominal syntax and must be learned as a separate parameter. 3.2 Person features are weak in Bantu and Romance NEs To express pronominal possessors, Romance languages display a wide use of possessive APs (e.g., Italian mio, tuo, suo) which cannot be considered as Genitive forms of possessive pronouns (cf. di me, di te, di lui/di lei) for the simple reason that in some cases both elements are available. A plausible reason for movement of a possessive AP to the specifier of a functional head could prima facie be its adjectival nature. If the possessive AP is merged in NP, one may suppose that this position does not allow for adjectival Concord; one could assume that the possessive AP must move to a Specifier in which such concord is allowed. This would, however, lead to unwelcome empirical and theoretical results. In a minimalist perspective the null hypothesis is that concord is a relation between a Spec and a Head, without any further specification of the labels, as in (4) above. From the empirical point of view, it would be difficult to motivate why so-called relational adjectives, which receive a θ-role like possessives, must be postnominal (10), while possessive adjectives, which can be postnominal in the marked order, are prenominal only in the unmarked case (11): (10) a.

la vecchia opinione razzista italiana the outdated opinion racist Italian

b. *L’italiana vecchia opinione razzista the Italian outdated opinion racist ‘the outdated Italian racist opinion’ (11) a. ?la vecchia opinione razzista tua the outdated opinion racist your b. la tua vecchia opinione razzista the your old opinion racist ‘your outdated racist opinion’

 Giuliana Giusti

If adjectival concord can be satisfied in the postnominal position for relational adjectives, it should also be satisfied in the same position for possessive adjectives, as it probably is in (11a). A second piece of evidence against the reduction of movement of possessive APs to their adjectival nature is the fact that the 3rd Person Plural pronoun loro, which has no adjectival concord, has the same distribution as the other possessive APs (12a): (12) a.

la {loro} vecchia {*loro} fotografia sbiadita {?loro} the {their} old {*their} picture faded {?their} ‘their old faded picture’

b. la {loro} vecchia {*loro} opinione razzista {?loro} the {their} old {*their} opinion racist {?their} ‘their old racist opinion’

Finally, in formal, bureaucratic, or playful registers, personal pronouns may appear in the high position even if embedded in a diPP, while full DPs cannot. Example (13a) is taken from a political trial in 1821; (13b–c) is contemporary Italian. The ungrammatical counterparts with full genitive NEs are given in (14): (13) a. Allora i de Cristofaro scaricano i di loro schioppi contro Ramaglia (…) Then the de Cristofaro fire the of them rifles against Ramaglia ‘Then the de Cristofaros fire their rifles against Ramaglia.’ from: Political Trials – Envelope 48 File 2b – Archivio di Stato di Campobasso, 1821. http://www.ripamici.it/storia/storia.html (March 2007) b. con una nuora autoritaria e le di lei tre figlie with a bossy daughter-in-law and the of her three daughters ‘with a bossy daughter-in-law and her three daughters’ (http://www.pannostrale.it/scheda.php?compagnia = I+TEATRANTI (March 2007) c. Applausi scroscianti in sala e sorriso stellare sulle labbra del protagonista, mentre il di lui cane – di nome Pinocchio – zampetta giocoso sul palco del Teatro dell’Arte, e la di lui figlia – Teresa – abbozza un accenno di pianto fra le braccia di mamma Francesca, e la di lui band – Saturnino in primis – osserva l’intera scena …

the of him dog … the of him daughter … the of him band ‘his dog’ … ‘his daughter’ … ‘his band’

http://www.mybestlife.com/ita_anima/Jovanotti_Autobiografia_di_una_ festa_sito.htm (March 2007)

Agreement and concord in nominal expression 

(14) a. *le di sua nuora tre figlie the of his daughter-in-law three daughters b. *i dei de Cristofaro schioppi the of the De Cristofaro rifles c. *il di Jovanotti cane/la di Jovanotti figlia/*la di Jovanotti band the of Jovanotti dog/daughter/band

These facts suggest that what is responsible for raising the possAP is not its adjectival nature but its pronominal reference. In other words, movement of possAPs and pronouns is independent of Concord and is triggered by Agreement, parallel to what happens to clausal subjects. There is a straightforward parallel between Italian possAPs and Swahili possessive pronouns. They have Person features. The answer to our first question above (Why do only pronouns and possessive adjectives move?) is that they uniquely contain the FF targeted by AGREE. These features are not strong enough to piedpipe a whole NE, but if there is nothing to pied-pipe (apart from a P in the marked case) the unmarked choice is to realize the higher copy. If it is a matter of the weak/strong nature of FF, then we expect some variation across languages. For example Germanic languages obligatorily move both pronominal and DP possessors, but not PPs. Hebrew seems to leave open the possibility of moving both pronominal and DP possessors, or merging both pronominal and DP possessors into a PP. (cf. Borer 1984; Ritter 1991; Siloni 1997 a.o.). Among Romance languages, Romanian presents apparent similarities to Hebrew, as is reported in the next section. 3.3 Hebrew construct state and Romanian possessive constructions In the Hebrew free state (15) we have a šelPP both with a full DP and a pronoun (cf. Siloni 1997: 21–26). In the construct state (16) N moves to D, substituting for the definite article and displaying a reduced morphology. From this position, N licences (abstract) Genitive Case, given that the NEgoal (full NE or pronoun) is not introduced by a P:10 (15) a.

ha-bayit ha-gadol šel ha-iša the-house the-big of the-woman

b. ha-bayit ha-gadol šel-a the-house the-big of-him

. I thank Gabriella Hermon and Ur Shlonsky for providing the Hebrew data.

 Giuliana Giusti

(16) a.

beyt ha-iša ha-gadol house the-woman the-big

b. beyt-a ha-gadol house-her the-big ‘her big house’

The obvious parallelism with Italian and Swahili (the latter noticed by Carstens 1991) is that the Genitive position is in the same high specifier immediately following D0. However, while in Italian the D position is filled by an article and in Swahili by an N prefixed with a Concord morpheme, in the Hebrew construct state, D is filled by a reduced form of N, which is in complementary distribution with the prefixed article. The construct state can therefore be captured in (17a) by assuming that N in D becomes the probe. Being c-commanded by N+AGREE, FF become strong and pied-pipe NEgoal, as in (17a). But if D hosts the definite article as in (17b), the probe remains in F0. The FF of NEgoal move to SpecFP but do not pied-pipe the entire NEgoal which remains in situ:11 (17) a. [NE [D° N+AGREE] [FP [NEgoal FF]i [F° N+AGREE] [FP AP [NP [NEgoal FF]i ... N]]]]

b. [NE [D ha]-[FP FFi [F° N+AGREE] [FP AP [NP [šel [NEgoal FFi ]] ... N]]]]

I leave open the possibility of analyzing P (e.g., šel) as either a Case marker in NEgoal, as in (17b), or as a probe itself, along the lines of Kayne (2004). Since Kayne’s proposal hinges on phrasal movement, while I have taken the head-movement option here, I do not discuss the latter possibility. The fact that in Hebrew (15), N is constituent-initial cannot be the (only) reason for the strong nature of the FF. In fact, in Swahili (8) the head noun also moves to D without making FF any stronger. Even if it would be more interesting to derive the strength of features from a general property of inflectional morphology, I assume that the strength of Person is independent of N-movement.

. An alternative approach is the recent analysis of construct state provided by Cinque (2003) and Shlonsky (2004) in terms of phrasal movement. In this approach the “second” position of the genitive is derived by phrasal movement of the low constituent [N+[NEgoal]]. This analysis makes the construct state completely parallel to the Romanian case and therefore irrelevant for the supposed contrast in strength of FF discussed here.

Agreement and concord in nominal expression 

Another language in which FF are apparently strong in NEs is Romanian, a Romance language belonging to the Balkan Sprachbund.12 The Romanian possessive construction has also been related to Hebrew construct vs. free state (cf. Grosu 1988; Longobardi 1996; Dobrovie-Sorin 2000): (18) a.

casa fetei *(cea) frumoasă house-the girl-the.gen the nice

b. casa frumoasă a fetei house-the nice of-the.fem girl-the.gen (19) a.

casa sa house-the possap.3p.sg.fem

/ei (cea) frumoasă /pron.3p.sg.gen (the) nice

b. casa frumoasă a sa /a ei house-the nice of+the poss.ap.3p.sg.fem /pron.3p.sg.gen ‘the nice house of the girl’

There is a striking similarity and many differences between Romanian possessive constructions and the Hebrew construct and free state. In both languages, the preposition-less genitive must immediately follow a constituent-initial N realized in a specific morphological form, which is however different in the two languages: in Hebrew, it is the reduced form; in Romanian, it is the N inflected for the definite article. Dobrovie-Sorin (2000) enumerates other differences in detail, among which: (i) The constituent initial position of N is obligatory in Hebrew (as in Swahili) but not in Romanian, which can have the head noun in intermediate position, both in the case of the presence of an indefinite article (20a), and in the presence of a prenominal descriptive or evaluative AP (20b–c). Notice that only (20b) is discourse marked, the other two cases represent the only possible orders: (20) a.

o casă frumoasă a fetei a house nice of-the girl-the.gen ‘a nice house of the girl’

b. frumoasa casă a fetei nice-the house of-the girl-the.gen ‘the nice house of the girl’ c.

biata prietenă a fetei piteous-the friend of-the girl-the.gen ‘the poor friend of the girl’

. For a systematic presentation of the most relevant Balkan features cf. Mišeska Tomič (2006). For discussing the Romanian data in (18)–(21), I am indebted to Iulia Zegrean, Cristina Rasnoveanu, Raluca Ciobanu and Eliza Pintilie.

 Giuliana Giusti

In no cases of (20) is a preposition-less genitive allowed. But a pronominal possessor is possible in (21) after any adjective inflected for a definite article, irrespective of the adjectival (sa) vs. pronominal (ei) nature of the possessive, taking us back to the difference between pronominal and full DP genitives and to the similarity between possessive adjectives and pronouns, observed in 3.1. (21) a.

frumoasa sa/ ei/ *fetei casă nice-the poss ap.3p.sg. /pron.3p.sg.gen /girl-the.gen.fem.sg house ‘her nice house / *‘the nice house of the girl’

b. biata sa/ ei/ *fetei prietenă piteous-the possap.3p.sg/pron.3psg.gen/girl-the.gen.fem.sg friend ‘her poor friend/*‘the poor friend of the girl’

(ii) In Hebrew, the choice between construct and free state is apparently open, while Romanian requires a preposition-less genitive if no element intervenes between N+art and the genitive: (22) casa (*a) sa/ ei/ fetei house-the of-the possap.3p.sg/pron.3psg.gen /girl-the.gen.fem.sg

(iii) The presence of a preposition-less genitive disallows for the cooccurrence of a bare postnominal descriptive AP (23b), while a pronominal genitive does not (23a): (23) a.

casa sa/ei (cea) frumoasă house-the her nice ‘her nice house’

b. casa fetei *(cea) frumoasă house-the girl-the.gen nice ‘the girl’s nice house’

The adjectival article cel has several functions, which are briefly reviewed in (39)–(40) below. For the moment, it suffices to say that it can introduce APs functioning as reduced relative clauses (cf. Coene 1999; Cinque 2004). The obligatory presence of cea in (23b), therefore, suggests that the AP is not in the intermediate layer of NE and is inserted in a more complex structure which we disregard for space limitations. It seems that a full genitive NE following the head noun freezes the construction blocking the insertion of any further modifier. This can be captured in the framework sketched above by claiming that in Romanian the N inflected for the definite article can be the probe for NEgoal. But in this case, N and NEgoal remain frozen, in configuration (24). Being inflected with the definite article, N can project a DP which is required for the NE to act as an argument (cf. Longobardi 1994).

Agreement and concord in nominal expression 

There is no movement of NEgoal to SpecDP. This keeps the generalization that FF are weak in Romance and Swahili without exceptions: (24) [NE=DP N+art+AGREE [NP [NEgoal] N]]

The freezing effect observed in NEs reminds us of the “freezing” effect proposed by Rizzi (2006) for Subject agreement.13 Rizzi proposes that the subject requirement is a criterial relation and that criterial relations have a “freezing” property which ensures that, once a criterial feature is checked, no other criterial feature can be assigned or checked on the same element. I propose that possessor licensing, is also criterial in the sense that it must be checked once and only once. As a result of checking, Genitive Case is assigned to the possessor and both the probe and the goal remain frozen in that configuration. The possessor criterion is optional in NEs, while the subject criterion is obligatory in clauses for the independent nature of nominal vs. clausal reference discussed in section 1. 4. Agree vs. Concord Concord behaves quite differently from Agreement. It is found on more elements at the same time, and does not produce any freezing; on the contrary, the richer Concord is, the freer the word order. Furthermore, concord can combine with Agreement. A straightforward case is the possessive AP in Romance. In (10)–(14) above, I have argued that it is not Concord on the possAP which triggers the movement to the periphery of the intermediate level, but the targeted Person features of this possessive. It is, however, a fact that possAPs in Romance may concord for Number and Gender with N. I claim here that Concord is not a criterial feature and can therefore combine with Agreement. In Italian, a possAP displays concord for Number and Gender both in prenominal and postnominal positions in (25)–(26): (25) a. la mia cartella b. la cartella mia ‘my bag’

. There is one major difference however; namely the freezing in Rizzi’s view only regards the moved element and not the triggering head. This is because Rizzi observes cases of movement and not of Agree in place. For lack of space, I will leave the discussion of this difference for future research.

 Giuliana Giusti

(26) a. il mio libro b. il libro mio ‘my book’

The structural representation of the Agree and Concord combination is given in (32)–(33) below. What is interesting to observe here is the case of a full NEgoal, which undergoes Agreement without movement. It is a fact that Agreement targets a single element (multiple agreement being special cases to be derived by either concord or anaphoric binding). If Agreement and Concord can co-occur on one and the same element, and if Agreement occurs only once, then Concord must necessarily be different in nature. As for (17b) above, I do not dwell upon the nature of the preposition. It could either be an overt Case marker in the extended projection of NEgoal, or it could be a probe in the extended projection of NE, along the lines of Kayne (2004). In the former case, one could propose that we have Agreement without movement (27a). In the latter case, one must adopt Kayne’s proposal of P triggering movement to its Spec and then further moving to precede the NE (27b), with consequent phrasal movement to obtain the Romance order: (27)

a.

[NE ... [FP F+AGREE [FP N [NP [NEgoalP [DP]]N]]]]

b.

(i) (ii)

[NE [FP P+AGREE [ NP [nP [NEgoal DP] NP]]]]→ [NE [FP NP P [FP DP [ P+AGREE [FP NP [nP [NEgoal DP]NP ]]]]]]

The choice between (27a) and (27b) is independent of our proposal. For expository reasons I stick to (27a).14 What is crucial for our discussion here is that a marker of Agreement with NEgoal can display Concord with NE. This creates an unexpected parallel between most Bantu languages on the one hand and Romanian (alone in the Romance family) on the other. We have already seen examples of this fact in Bantu (8a), repeated here as (28a) and in Romanian (20) (equivalent to (28b)):

. The overt nature of the P (a probe or a marker for the checked Agree with a probe) is probably the ground for the lack of the freezing effect, as well as the violation of the hierarchy of possessivization noticed in (6) above. For reasons of space, I cannot develop this point here.

Agreement and concord in nominal expression 

(28) a.

picha mpya ya Amira 9picture 9new 9of Amira

[N – ap – PP] Carstens (1991: 100(39a)) ‘Amira’s new picture’

b. portretul nou al Mariei portrait-the.masc.sg. new. masc.sg A-the.masc.sg Maria.gen.fem.sg ‘Mary’s new portrait’

The status of the genitival element a+art in Romanian has been the topic of much research. Grosu (1988, 1994), argues for a P status. Cornilescu (1995, 2003) argues that it is a recursive D embedding the prossessor DP. D’Hulst, Coene, and Tasmowski (2000) argue that it is a functional N. A comparison with Swahili allows us to take a different perspective. If it is trivially observed that UG allows for a case marker to be underspecified for the nominal features of the selecting/modified noun, as is the case in most (maybe all) Bantu languages, then we may regard a, as a case marker. We can further observe this property not only in Romanian but also in Albanian (another Balkan language which is, however, not part of the Romance family):15 (28) c. portreti i ri i portrait-the.masc.sg. the.masc.sg new.m.sg the.masc.sg Maries Maria.gen.fem.sg

I propose that parallel to As, Case markers (Bantu connectors and Balkan genitival articles) can come from the lexicon as underspecified for nominal features. In both versions of (27) above, P must concord with NE. In (27a), it is merged in the specifier of a ConcP; in (27b), it is a Conc0. The next section will settle the question of how Concord is realized in the structure. I will turn to Concord on P in 5.1. 5. Concord In this section, I briefly formalize a proposal on the structure-building procedure assumed so far. The intermediate layer is the place for direct merge of attributive APs, which in many languages are underspecified for nominal features. I follow Cinque’s (1994, 1999) seminal idea that APs and adverbs appear in functional specifiers and are merged according to a universal hierarchy, which is not subject to . I thank Giuseppina Turano for her usual availability and in particular for providing the Albanian example.

 Giuliana Giusti

acquisition. I depart from Cinque’s proposal in two respects. Reformulating what I suggested in Giusti (2002), I propose that the hierarchy of adjectival modifiers does not reflect a rigid sequence of functional projections which are always present, but a principle which constrains the merger of modifiers. Furthermore, I assume that the functional heads in the inflectional layer are not labeled for the hierarchical features but are trivial copies of the φ-features of the NE (i.e., Number, Gender or Word Class, and Case). I follow Giorgi and Pianesi (1997), in assuming that functional features are ordered hierarchically by the Universal Ordering Constraint, and that the hierarchy is not violated if two or more ordered features are represented in one and the same head, as implied by the Feature Scattering Principle (Giorgi and Pianesi 1997: 14–16), which does not force each feature to head a projection: (29) a. Universal Ordering Constraint Features are ordered so that given F1 > F2, the checking of F1 precedes the checking of F2. b. Feature Scattering Principle Each feature can head a projection

Giorgi and Pianesi’s proposal can dispense with empty (or inert) functional heads and specifiers. A head is projected only if needed (e.g., to keep projecting an independently needed specifier); otherwise features can be bundled in one and the same head provided that this does not violate the hierarchy. This proposal captures the fact that As (or other elements such as Ps) can display concord for the same bundle of formal features, and not for separate dedicated features (such as Number, Gender, or Speaker-Orientation, Size, etc.). It also captures the observation, also made by Carstens (2001), that agreement in the clause results in sharing the features of the subject with the inflectional morphology of the verb (the lexical head in the clause), while nominal concord is quite the opposite, in that it consists of sharing the features of the lexical head N with its modifiers. Carstens (2001: 332, ex.(28)) unifies the two procedures by assuming that adjectival agreement is also triggered by targeting a lower element (AP or DP merged inside NP) and attracting it to its Spec. I take the opposite viewpoint. While for genitives we have evidence for a base and a derived position, there is no such evidence for APs. Furthermore, apart from possessive and relational APs, other APs do not saturate a θ-role assigned by N. Finally, in the perspective of a parallelism with the clause, APs can be compared to adverbs, which never A-move.16 I propose that feature sharing resulting from Concord is not obtained

. Of course adverbials, as well as adjectives, can move to comply with discourse requirements and A-bar move. This is the ground for Cinque (1999) to propose that the functional

Agreement and concord in nominal expression 

by movement but is the result of merger of a modifier in a SpecConcP. The SpecHead configuration transfers the interpretable features of Conc0 onto the functional structure of the AP in SpecConcP. An Italian example with two prenominal and one postnominal AP is given in (30):17 (30) a. b. D0 le iNum iGen F0 uNum uGen

le altre belle ragazze the.fem.pl other.fem.pl nice.fem.pl girls.fem.pl ‘the other nice Italian girls’

italiane Italian.fem.pl

DP=NE ConcP3 FP

Conc3′ AP Conc30 altre iNum iGen

F0 uNum uGen

ConcP2 FP

Conc2′ AP Conc20 belle ragazze iNum iGen F0 uNum uGen

ConcP1′ FP

Conc1′ AP Conc10 italiane ragazze iNum iGen

NP ragazze iNum iGen

In (30b), the lexical N ragazze, with its interpretable formal features of Num (plural) and Gen (feminine), moves out of NP (i.e., it is re-merged as Conc10). I take this to be due to the requirement that Number and Gender features reach the highest head. The interpretable features of Num and Gen of N in Conc0 are then remerged to F0 (a functional head in the extended projection of the A, as in (31a)). In Italian, N further moves to the left of the higher ConcP (to Conc20). If more than one adjectival modifier is present in the lexical array, merger of other

hierarchy is always projected. For the adjectival hierarchy cf. Cinque (1994); Laenzlinger (2005). My proposal here differs minimally in that I propose that the hierarchy of modifiers is parallel to the Universal Ordering Constraint of Giorgi and Pianesi and is not necessarily represented as separate (often inactive) projections in the structure. . I indicate ufeatures as uninterpretable features to be valued and ufeature as uninterpretable copies of interpretable features that have raised.

 Giuliana Giusti

APs proceeds in the same fashion, subject to the universal hierarchy. In (31b), the evaluative AP belle is inserted. The head N in Italian does not move any further but silent copies of its features keep moving, creating as many ConcPs as needed for the merging of the APs present in the lexical array (31c). When the last ConcP containing the hierarchically higher AP is merged, the highest projection is created to host the edge of the NE where the reference features of NE are merged (here DP). If these features are covert, interpretable Number and Gender are realized overtly on the definite article (31d): (31) a.

[ConcP1 [FP [ufem, upl] [ap italian-]] [Conc’1 ragazze[ifem,ipl] [NP ragazze[ifem, ipl]]] →

b. [ConcP2 [FP [ufem, upl] [ap bell-]] [Conc’2 ragazze[ifem, ipl] [ConcP1 [FP [ufem, upl] [ap italian-]] [Conc’1 ragazze[ifem, ipl] [NP ragazze[ifem, ipl] ]]]]] → c. [ConcP3 [FP [ufem, upl] [ap altr-]] [Conc’3 [ifem, ipl]] [ConcP2 [FP [ufem, upl] [ap bell-]] [Conc’2 ragazze+[ifem, ipl]] [ConcP1 [FP [ufem, upl] [ap italian-]] [Conc’1 ragazze[ifem, ipl]] [NP ragazze[ifem, ipl] ]]]]]]] → d. [DP FFj [D’ le[ifem, ipl] [ConcP3 [FP [ufem, upl] [ap altr-]] [Conc’3 [ifem, ipl] [ConcP2 [FP [ufem, upl] [ap bell-]] [Conc’2 ragazze+[ifem, ipl] [ConcP1 [FP [ufem, upl] [ap italian-]] [Conc’1 ragazze[ifem, ipl] [NP ragazze[ifem, ipl]]]]]]]]]]

In Giusti (1997, 2002), I proposed that the position relevant for the interpretation of the NE is not D0 but SpecDP, the left edge of NE. D0 is the head in which Case is assigned, and the different articles (definite/indefinite) are bundles of Case, Number, and Gender features, which licence an empty operator (as proposed in Campbell 1996). The present proposal is in the same spirit. Concord features are merged to build the extended projection of NE and to allow for what is merged in the Specifier (the edge of the NE) to copy the features of NE. Next I will claim that the reference features of NE merged at the left edge combine with the reference features of NEgoal merged in the immediately lower specifier. 5.1 Prefixes and genitival articles This proposal can straightforwardly handle feature transfer on Bantu connectors and Romanian genitival articles. Consider that Concord is feature transfer from a head onto its specifier. It does not take place in a dedicated projection. An NE merged in SpecNP is therefore expected to concord with N, if it comes from the lexicon with unvalued Num and Gen. In (32a), the Bantu connector ya agrees with the Class9

Agreement and concord in nominal expression 

N picha, parallel to the adjective mpya. In (32b–c) we observe the steps in the derivation: (32) a. picha mpya ya Amira 9picture 9new 9of Amira b. [ConcP1 [FP m[u9][ap pya]] [Conc’1 picha[i9]] [NP [NEgoal ya[u9] [DP Amira]] [N’ picha[i9]]]]] → c. [DP FFj+FFi [D’ picha[i9]+[AGREEFFi] [ConcP1 [FP m[u9][ap pya]] [Conc’1 picha[i9] [NP [NEgoal FFi ya[u9][DP Amira]]i [N0picha[i9]] ]]]]]

In (32b), the head noun picha is inserted with its interpretable Class9 feature. Then the NEgoal Amira is inserted in SpecNP. In this configuration, the highest functional head F0 of the NEgoal concords with N. Then, N is copied and remerged. The Class9 features of the covert copy become uninterpretable and the interpretable counterparts are still attached on picha. Then the AP pya is merged in SpecConcP1 and its functional head copies the Class9 features of N. The derivation continues in (32c) with N remerged for the second and last time, creating a SpecDP available as the edge of the NE and the derivation stops. The reference of NE is obtained from the composition of the Person features FFj of picha and the Person features FFi of Amira. No movement of the NE goal is triggered because FFi cannot pied-pipe it. Now observe the parallel case in Romanian. In (33b), the N portretul is first merged already inflected with the definite article. The possessor NE is inserted in SpecNP and in this configuration the functional element a copies the Number and Gender features of the head N, which moves out of NP and creates ConcP1. The extended projection of an adjective can then merge in SpecConcP1 and copy the features of N. In (33c), N remerges for the second time. The specifier can function as left edge and the projection can be completed by probing the Person features FFi of the possessor Maria and compose them with the Person features FFj of the NE portretul: (33) a.

portretul nou al Mariei portrait-the new p-the Mary.Gen

b. [ConcP1 [FPF0[uMASC, uSG]] [ap nou-]] [portretul[iMASC, iSG]] [NP [NEgoal [P0 al[uMASC, uSG]] [Mariei]] [N0portretul[iMASC, iSG]]]]]] c. [DP = NE FFj+FFi [D0portretul[iMASC, iSG]+agree] [ConcP1 [FPF0[uMASC, uSG]] [ap nou-]] [portret[iMASC, iSG]] [NP [PP = NEgoal [P0al[uMASC, uSG]] [Mariei]i] [N0portretul[iMASC, iSG]]]]]

I have assumed in (32c) and (33c) that the Person features of NE and the Person features of NEgoal combine at the left edge. This is done overtly in Germanic, where the possessor is in complementary distribution with determiners. For Romanian

 Giuliana Giusti

and Swahili, a possible alternative is to project the AGREE probe in a separate projection ConcP2 in (34), headed by a silent copy of N and its φ-features: (34) [NE FFj [D0N[iφ]] [ConP2FFi [Conc’ N[iφ]+agree [ConcP1 N[iφ] [NP [NEgoal al[uφ] [Mariei]i] [N’ N[iφ]]]]]

The option in (32c) and (33c) is more minimalistic than (34) for two reasons: on the one hand there is no overt evidence for (34) in which ConcP2 has a non-overt head and a non-overt specifier; on the other hand (32c) and (33c) are parallel to Germanic (35a)–(36a). However, when FF are realized in pronominal and adjectival possessives in Romance and Bantu, FFj and FFi do appear in separate specifiers, as in (35b)–(36b): (35) a. her other nice Italian friends b. le sue altre belle amiche italiane the her other nice friends Italian (36) a. [NE FFj +[NEgoal FFi [her]] [Conc’4 [ipl]] [ConcP3 [FP [ap other]] [Conc03 [ipl]] [ConcP2 [FP [upl] [ap nice]] [Conc02 +s[ipl]] [ConcP1 [FP [upl] [ap Italian]]i [Conc01 friends[ipl]] [NP [NEgoal her]i] friends[ipl] ]]]] b. [NE FFj [D0 le[ifem, ipl] [ConcP4 [FP FFi [ufem, upl] [ap sue]] [Conc04 [ifem, ipl]] [ConcP3 [FP [ufem, upl] [ap altr-]] [Conc03 [ifem, ipl]] [ConcP2 [FP [ufem, upl] [ap bell-]] [Conc02 amiche+[ifem, ipl]] [ConcP1 [FP [ufem, upl] [ap italian-]]i [Conc01 amiche[ifem, ipl]] [ [FPFFi [apsu-]] [NP amiche[ufem, upl] ]]]]

I assume that the hierarchy of reference features is NE>NEgoal, namely FFj >FFi, as represented in (24) and (36) above. In (36a) FFj and FFi are merged in the same specifier, but they are checked in the required order, according to the Universal Ordering Constraint in (29b). In (36b) each FF is in a different specifier according to this hierarchy. The question of why, in some languages it is the peripheral position in the intermediate layer to host the attracted agreeing possessor, while in others it is the peripheral position in the upper layer, remains open for the time being. The answer may lie in how compositional reference is obtained at LF (possibly the two features must combine in the edge at LF in all languages), and how this interacts with overt vs. covert movement in syntax. What is crucial here is that we can bundle features together (both in heads and in Specifiers) provided that they do not violate the relevant hierarchy. 5.2 Pre-prefixes and double definiteness With additional assumptions on the relation between Case morphology, articles and definiteness, (cf. Giusti 1997, 2002), the present proposal can be extended

Agreement and concord in nominal expression 

to the phenomena of pre-prefixes spreading in Bantu and double definiteness in Romanian. It is commonly accepted that pre-prefixes are related to definiteness in some way (cf. Katamba 2003: 107–108). It is interesting to remark that Katamba rejects an identification of the pre-prefix with the European definite article, on the ground that the pre-prefix can serve different functions such as indicating the pragmatic role of definiteness, specificity, or focus, and that its presence may also be syntactically driven (e.g., present in affirmative but not in negative clauses, spread on nominal modifiers, etc.). I consider Katamba’s observations as actually confirming the article function of the pre-prefix, in the sense that European articles and Bantu pre-prefixes can serve a variety of functions, being principally a syntactic device to license the DP-layer, as I claimed in Giusti (1997, 2002). As reported by Katamba (2003), the Bantu pre-prefix has been argued to derive from a pronoun (Bleek 1869: 150 for Xhosa S41) and may be absent or present, subject to a low level parameter (e.g., Nurse & Hinnebush (1993: 338–445)). In fact, despite its absence in Swahili, it is present in closely related languages. These properties are perfectly parallel to the European article. The former property is widespread: Romance and Germanic articles derive from either a personal pronoun or from a demonstrative (which itself can be used in a pronominal function and is often the base for the new formation of a personal pronoun).18 The latter property is reproduced in Slavic where Bulgarian has an article, while other languages do not.19 Surprisingly, the extensive system of concord in Bantu languages is also reminiscent of article doubling in Balkan languages. Finally, prefixes and pre-prefixes are acquired parallel to articles from a no-marking stage to a possible shadow vowel and finally to well-formed morphemes (Demuth 2003). Their nature of phonologically proclitic elements may explain their omission in early childhood speech as related to the impossibility of unstressed initial syllables in early speech (cf. much work by Demuth a.o. 1988, 2001, 2003 and its application to Italian by Gozzi 2004; Giusti and Gozzi 2006). Double definiteness is a diversified phenomenon which includes multiple occurrences of the same article (as in Hebrew above) and can be shown to be straightforwardly parallel to multiple occurrences of prefixes, as I have argued in 4.1, or pre-prefixes, as I claim here. Another case of double definiteness is the

. For a generative analysis of the development of the Germanic article cf. Giusti (1994), and of the Romance article cf. Giusti (2001). . Notice that the oldest attested Slavic language, Old Church Slavonic, being based on Old Bulgarian, already shows the presence of the developing article (cf. Dimitrova-Vulchanova & Vulchanov 2007) showing that this is not a very recent innovation.

 Giuliana Giusti

co-occurrence of a determiner (e.g., a demonstrative) with a definite article. The two phenomena often, but not always, go together. I think that once again the macro-parallelisms between Bantu and Romance can give a broader perspective to the issue. Let us start with pre-prefixes. In Xhosa, prefixes as well as pre-prefixes are copied on all modifiers (which are postnominal). Visser’s (2006) examples in (37) give the contrast between a definite and an indefinite NE, which support the hypothesis that pre-prefixes are related to definiteness. Here I use Visser’s system of glossing: (37) a.

Andiyifuni lokhwe ibomvu Andiyifuni lokhwe i-bomvu neg-is-agro[9]-want-neg dress[9] [9].red ‘I don’t want a(ny) red dress of a(ny) girl’.

yantombi. ya-ntombi gen[9]-girl

b.

Andiyifuni ilokhwe ebomvu yentombi. Andiyifuni i-lokhwe a+i-bomvu ya+i-ntombi neg-is-agro[9]-want-neg det+dress[9] det[9].red det.gen[9].girl ‘I don’t want the red dress of the girl’.

The NE in (37a) only displays Class prefixes and has indefinite interpretation. The NE in (37b) displays the same prefixes on all modifiers enriched by a further prefixal morpheme and has definite interpretation. Visser (2006) proposes that in cases such as (37b), the adjectival and genitival modifiers have a larger structure than they have in (37a), labeled DP.20 I see two problems in assuming that an adnominal AP is embedded into a DP: (i) It is a departure from Longobardi’s (1994) original proposal that the DP is the projection of argumenthood. The AP is clearly part of the direct modification of NE and not an argument itself. Furthermore the possessor is an argument, and as such it projects its own DP, but what agrees with NE is the preposition, which is clearly not an independent argument itself, and therefore cannot be topped by a further argumental projection; (ii) It could be envisaged that these adnominal DPs are appositive reduced relatives (something like “the dress, the red one, the one of the girl’). This is certainly possible in some marked cases. But appositive reduced relatives should be different from direct modification (also cf. recent work by Cinque 2007). By combining Giusti’s (2002) proposal that Case is the marker for argumenthood and is realized in the topmost head of the NE with the proposal that features

. This kind of proposal is admittedly the most intuitive one and has been proposed for adjectival articles in Balkan languages, including Romanian, by Cornilescu (1992).

Agreement and concord in nominal expression 

can be bundled, we can hypothesize that Number and Gender, bundled with (abstract or morphological) Case, are merged with N, as in (30)–(32). Since they are merged in a bundle, they reach the topmost head to check the hierarchically highest feature, namely Case. But moving through (and actually instantiating) all functional heads in the extended projection of the NE, they trigger agreement for the whole bundle of features (including Case) in the functional head F0 of their modifiers, which in these languages have two paradigms of concord: one for Partitive Case (indefinite interpretation) and one for Direct Case (definite interpretation). The structure for (37b) is (38), where the functional structure of the AP (bomvu) contains uninterpretable features a+i including Word Class and Case: ConcP2=NE

(38) Conc2° ilokhwe iClass Case

ConcP1 FP/AP 5 F/A° a(+i) bomvu uClass Case

Conc1' Conc1° ilokhwe iClass Case F/P° ya+(i) uClass Case

NP FP

N' DP N° ntombi ilokhwe iClass Case

The Romance parallel to pre-prefixes spreading is the adjectival article observed in (18a) and (23b) above. In (39a), we see that the Romanian FP/AP can be optionally preceded by cel. Notice that only APs with restrictive interpretation can be preceded by cel. A relational AP (which is non restrictive) cannot, as in (39b). Furthermore, the adjectival article is one possible means to realize nominal ellipsis, as in (40a–b) and to introduce a NE with a numeral AP (40c): (39) a.

tabloul (cel) frumos picture-the (CEL) nice ‘the nice picture’

b. invasia (*cea) italiană invasion-the CEA Italian ‘the Italian invasion’

 Giuliana Giusti

(40) a.

cel frumos (*tablou) CEL nice ‘the nice one’

b. (*tabloul) cel al Mariei CEL AL Maria.gen. c.

cei trei (băieţi) ‘the three boys’

In (39)–(40), we see that cel is optional in (39a), ungrammatical in (39b), obligatory in all cases of (40), and we remember that it was obligatory after a genitive phrase in (23b) but optional after a pronominal possessor in (23a) above. From this varied behavior, it is reasonable to single out at least three different functions (cf. Coene 1999; D’Hulst et al. 2000 for a detailed discussion). In (39), it introduces a restrictive FP/AP in postnominal position. In (40a), it cannot appear when the FP/ AP is in prenominal position, but it can licence an empty N. This is also the case in (40b), where a PP is present and N must be non-overt. Finally, cel can appear in D0 and co-occur with N in the presence of a numeral AP. Such variation in how overlapping bundles of features are realized is expected in the present framework. I follow Cinque (2007) in suggesting that cel is the head of reduced relative clause, indicated by the restrictive interpretation of the AP in (39a). In (40a–b), cel overtly realizes the same φ-features in D0, thereby licensing an empty N0. In (40c), it realizes F0 for numeral APs which are otherwise uninflected. For reasons of space, I leave the details of cel realization for future research. The contrast with Xhosa supports the hypothesis that in this language, the adjectival article is the result of mere Concord with the NE; while in Romanian, we see that the Concord with the NE is overtly realized in particular cases, when the functional structure of the AP is either richer than adnominal APs (as I assume reduced relatives are in (39a) and (40a–b)), or must be realized as a free morpheme due to the uninflected nature of the numeral as in (40c). We now leave the topic of adjectival articles and observe another instance of double definiteness also present both in Bantu and in Romance, namely the cooccurrence of demonstratives with article-like morphemes. Across languages, demonstratives are ambiguously analyzed as either modifiers (like APs) or functional heads (like articles). In the debate on whether languages with no article miss or project a DP, the presence of a demonstrative as an argument in favor of the assumption of a DP is often dismissed by the demonstrative-as-modifier hypothesis. This is also supported by the fact that demonstratives usually inflect like adjectives, may have a more liberal positioning than articles, pattern with other elements (adverbials, verbs, pronouns) to express deictic or discourse-anaphoric values, and

Agreement and concord in nominal expression 

can be used in isolation. On the opposite side of the debate, reasons to match demonstratives with articles are as follows: they obviously build a closed class, even in languages where there is a rich pattern of deixis; and in many languages, they are in complementary distribution with articles. In my view, both lines of reasoning are motivated and are arguably not as incompatible as usually claimed. An easy way out of this dichotomy would be to assume that the status of demonstratives (article-like functional head vs. adjectival-like modifier) is subject to variation, and must be learned in acquisition. But even in languages like Italian, in which it is always in complementary distribution with the article, the demonstrative is uttered in isolation much earlier than articles, on a par with adjectives, quantifiers and other lexical categories. Furthermore, the very co-occurrence vs. complementary distribution of the demonstrative cannot be a straightforward sign in the input as to the status of the demonstrative in a language (contra Bernstein 1997), since some languages provide children with both kinds of input for one and the same element. The possibility of a postnominal demonstrative and its position in the word order is a significant point of variation in Romance. French, Italian and Portuguese (41a–c) only have a prenominal demonstrative, while Romanian and Spanish have two possible positions for the demonstrative: the prenominal one, common to the other kin languages; and a postnominal one, which surfaces differently in the two languages: the Romanian demonstrative (first discussed by Dobrovie-Sorin 1987; Grosu 1988; Giusti 1994a, 2002, 2005) displays a second position immediately following N+art (42), while the Spanish demonstrative (cf. Bernstein 1991; Brugè 1996) appears to be the lowest in the sequence of postnominal APs (43): (41) a. b. c.

ce bel garçon questo bel ragazzo este bonito rapaz this nice boy

French Italian Portuguese

(42) a.

acest băiat frumos this boy nice

Romanian

b. băiatul acesta frumos boy-the this nice (43) a.

este chico hermoso this boy nice

b. el chico hermoso este the boy nice this

Spanish

 Giuliana Giusti

Another important issue is the position of a possible locative re-enforcer, which can help us establish the first merge position of the demonstrative, as observed by Brugé (1996). Parallel to the phenomenon of floating quantifiers in the VP-internal subject position (Sportiche 1988), the discontinuous position of the demonstrative and the locative re-enforcer is derived by moving the demonstrative from the first merge position (marked by the locative adverb) to a higher position triggered to check the referential features carried by the demonstrative: (44) a. b. c. d.

ce garçon [ce [ci]] questo ragazzo [questo [qui]] este rapaz [este [aqui]] este chico hermoso [este [de aqui]] this boy nice (of) here

French Italian Portuguese Spanish

The internal structure of the demonstrative is particularly interesting. It is not clear whether the Locative is a complement or a modifier of the demonstrative. What is clear is that it must match its distal features. In French, they are not specified on ce and the distal features are only expressed by the locative (cf. ce-ci (“this”), vs. ce-là (“that”)). In Italian, they must match the locative (cf. questo qui (“this here”) vs. *questo lì (this there), and quello lì (“that there”) vs. *quello qui (that here)). In Spanish, the presence of the all-purpose preposition de, further suggests that we are dealing with a complex structure. What forces movement of part of the complex structure to the left edge of the NE in some languages is the referential features they realize. If the reference of the nominal expression is checked in the left edge of the NE, as we hypothesized in (24) and (36) above, then the anaphoric or deictic features must be moved there. There are various reasons to assume that we are dealing with XP-movement and not with head movement. (i) First of all, head movement is local and the demonstrative appears to skip the entire intermediate layer. (ii) Head movement is reserved to the lexical head, while the demonstrative appears to skip the positions occupied by N. (iii) The demonstrative can co-occur with the article in other languages even when it is in prenominal position (as in modern Greek, cf. Horrocks & Stavrou 1987; Grohman & Panagiotidis 2005, among many others). In (45), I reformulate the proposal suggested by Brugè and Giusti (1996) and by much of their separate work. In (45a), the demonstrative starts in a predicative structure FP with the locative AdvP. The projection of the demonstrative contains reference features including Person, deictic/anaphoric value, etc. These are represented as FFj, as in (24) and (36) above. In Spanish, we observe the head of the predicative structure de; in the other languages we may think it is covert. In (45b), the FP is in a low ConcP to copy nominal features. In some languages, it is not necessary to move the whole DemP, but it suffices to move its referential features

Agreement and concord in nominal expression 

FFj. In order to licence this silent copy an article must be merged in the structure, as in Spanish (43b) above. In (45c), the demP is moved overtly to SpecDP, leaving the locative in place. In this case it is in complementary distribution with an article in D0 because the interpretable Gen and Num features and the abstract Case specification are recoverable from the uninterpretable copies which are present on the demonstrative which concords with D: (45) a. [FP [demP FFj + este] [F0 de] [AdvP aqui]] b. [DP=NE FFj [D0 D] [ConP [FP1 [demP FFj + dem] F0 [AdvP]] Conc0 [NP N0]]] c. [DP=NE [demP FFj .dem] D0 [ConP [FP1 demP F0 [AdvP]] Conc0 [NP N0]]]

To conclude, parallel to possessors, demonstratives are XPs related to both the internal layer and the highest specifier (the left edge) of the NE, which hosts the referential interpretation of NE. The proposal that the checking position of referential features is the left edge of the NE is forced by the observation that the demonstrative can only be a maximal projection. Such a hypothesis is even more necessary in Bantu, where N clearly moves to D0 and does not appear to interact with demonstrative movement. In Swahili, the demonstrative is generally postnominal (46a) but can also be prenominal with anaphoric (non-deictic) function (46b), according to Amidu (2006: 145): (46) ‘Juma Zaidi saw a big tree’ … a. … Juma Zaidi a-li-ka-a chini y-a mti u-le … J.Z. sat 9under part 9of 3tree 3that b. … Juma Zaidi a-li-ka-a chini y-a u-le mti … J.Z. sat 9under part 9of 3that 3tree

Amidu observes that in a context in which the referent has already been introduced, as is the case in the reported example, the anaphoric demonstrative can appear in either position. It seems, however, that the prenominal position is somehow losing the deictic function and is specializing for a purely anaphoric one. This can be taken as an initial grammaticalization of the demonstrative towards the formation of a definite article in the European sense, namely a free morpheme which encodes the last merge of interpretable Word Class features bundled with Case. The hypothesis that demonstratives are first merged in a low position and then moved to a higher position predicts that in some languages the two copies of the demonstrative be both realized. This is the case of Xhosa (47), as discussed by Visser (2006). The comparison with Romance suggests a revision of Visser’s analysis, which postulates multiple DP projections in NE, by simply assuming that for discourse reasons Xhosa can optionally realize the lower copy of the demonstrative, as in (47c):

 Giuliana Giusti

(47) a. b.

ló mtuana (lò) this child this èzì zinja (ézì) these dogs these

c.

[DP=NE demP [D0 D+N+Conc] [ConP [(demP)] N+Conc [NP N]]]

Still following Visser, a possible alternative to doubling is the postnominal position of a demonstrative with the following properties: (i) The postnominal forms are emphatic while the prenominal forms are unmarked; (ii) They are morphologically stronger than the prenominal forms (e.g., loo vs. lowo in (48a–b)); (iii) The strong form can also appear prenominally in (48c); (iv) The strong form in either position co-occurs with the pre-prefix on N (48b–c): (48) a.

loo mntu that person

b. umntu lowo pre-pref.person that c.

lowo umntu that pre-pref.person

The weak demonstrative in (48a) is in SpecDP as a regular demonstrative in Romance cf. (45c) and (47c). Parallel to Romance, economy reasons prevent merger of a pre-prefix on N0 to realize interpretable Class features on D0. The demonstrative in SpecDP concords for Class features and realizes them overtly. This is represented in (49a).21 The postnominal strong demonstrative in (48b) remains in its first merge position (probably due to a Focus feature). For this reason, in (49b) the head noun inflected with a pre-prefix moves to D whose Spec hosts covert definite FF, parallel to the merger of a definite article in Romance (45b): (49) a. [NE [demP dem+uclass] [D0 [iclass]] [NP N[iclass]]]] b. [NE FFj [D0 N+[iclass]] [ConP [demP FFj + [[dem+uclass]+Focus] [Conc0 [N+iclass]] [NP N0+[iclass]]]]

In this analysis, the strong prenominal demonstrative cannot be in the same structural position as the weak prenominal one. Since it requires the pre-prefix, it must be in the position of the postnominal demonstrative. But how can we motivate the different word orders? I propose that (48c) is obtained by splitting the DP-layer into multiple projections to host discourse features, and fronting the focused possessor

. Remember that for minimalist requirements we do not create vacuous projection. No ConcP appears in the structure because no AP is present in this case.

Agreement and concord in nominal expression 

to check emphatic features. This movement does not target SpecDP but a higher projection which is headed by a discourse feature (Foc0). For this reason the head noun is also merged with the pre-prefix, because the silent referential features of the demonstrative are in Spec DP and need licensing: (49) c. [FocP [[dem+uclass]+Foc] [Foc0] [DP FFj [D0 N+[iclass]] [ConP [demP FFj + [[dem+uclass]+Foc] [conc0 [N+iclass]] [NP N0+[iclass]]]

The possibility of introducing discourse features into the NE opens up a whole chapter of nominal syntax. For reasons of space this cannot be dealt with in the current paper. I refer the reader to what I have proposed for other marked orders in Romance and Balkan language in Giusti (1996, 2005, 2006). In this section, I supported the claim that Concord is different from Agreement. I have observed that Concord is feature transferring from the head onto the specifier. I have also taken the label Conc as a dummy for any projection created to allow for such relation to occur. However, I have proposed that Concord takes place in all kinds of projections including DP and NP. With additional assumptions motivated previously for Romance and other European languages, I have derived the distribution of pre-prefixes and double definiteness in Bantu.

6. Conclusions In this paper, I have made a theoretical point: Agreement and Concord are different kinds of relations. In particular I have assumed the minimalist notion of Agreement in NEs as happening in two steps (probe and move) and involving Person features, as in Chomsky (1995, 2000, 2001, 2004, and following work). As for Concord, I have proposed that it happens in what is traditionally called the Spec-Head Agreement Configuration and does not involve any movement. Furthermore, it involves Number, Gender or Class features, and Case. I have supported my claim by means of macro-parallelisms among Romance and Bantu languages, focusing on Italian, Romanian and Spanish in the former and on Swahili and Xhosa in the latter family. In the course of the discussion, I have hinted at many long standing problems dealt with by previous literature on the syntax of NEs, such as the trigger of possessor raising and the nature of possessor licensing, the relationship of feature sharing and the phenomenon of double definiteness. I hope to have shown that the macro-parallelisms offered by the unrelated Bantu and Romance languages can offer a novel perspective to analyze these apparently very different phenomena. Due to limits of space, I could not deal with many related issues. For example, it is not clear whether feature sharing in adjectival predication and reduced relative

 Giuliana Giusti

clauses or multiple agreement in verb morphology and in negative concord should be instances of Agreement, Concord or a third kind of feature transferring relation. I hope that future research can reduce these phenomena to a single Agreement relation combined with multiple occurrence of Concord.

References Abney, S. 1987. The English Noun Phrase in Its Sentential Aspect. Ph.D. dissertation, MIT. Amidu, A.A. 2006. Pronouns and Pronominalizations in Kiswahili Grammar. Köln: Rüdiger Köppe. Barlow, M. & Ferguson, C.A. 1988. Introduction. In Agreement in Natural Language, M. Barlow & C.A. Ferguson (Eds), 1–22. Stanford CA: CSLI. Belletti, A. 2001. Agreement projections. In The Handbook of Contemporary Syntactic Theory, M. Baltin & C. Collins (Eds), 483–510. Oxford: Blackwell. Bernstein, J.B. 1991. DPs in French and Walloon: Evidence for parametric variation in nominal head movement. Probus 3: 1–26. Bernstein, J.B. 1997. Demonstratives and reinforcers in Romance and Germanic languages. Lingua 102: 87–113. Bernstein, J.B. 2001. The DP hypothesis: Identifying clausal properties in the nominal domain. In The Handbook of Contemporary Syntactic Theory, M.R. Baltin & C. Collins (Eds), 536–561. Oxford: Blackwell. Bleek, W.H.I. 1869. A Comparative Grammar of South African Language, Part2: The Concord, Section 1: The Noun. Cape Town: J.C. Juta Trübner & Co. Borer, H. 1984. Parametric Syntax. Dordrecht: Foris. Bosque, I. & Picallo, C. 1996. Postnominal adjectives in Spanish DP. Journal of Linguistics 32(2): 349–385. Brugè, L. 1996. Demonstrative movement in Spanish: A comparative approach. University of Venice Working Papers in Linguistics 6(1): 1–53. Brugè, L. & Giusti, G. 1996. On demonstratives. GLOW Talk, Athens. Campbell, R. 1996. Specificity operators in SpecDP. Studia Linguistica 50(1):161–188. Cardinaletti, A. & Starke, M. 1999. The typology of structural deficiency. In Clitics in the Languages of Europe, H. van Riemsdijk, (Ed.), 145–235. Berlin: Mouton de Gruyter. Carstens, V. 1991. The Morphology and Syntax of Determiner Phrases in Kiswahili. Ph.D. dissertation, UCLA. Carstens, V. 2000. Concord in minimalist theory. Linguistic Inquiry. 31: 319–355. Carstens, V. 2001. Multiple agreement and case-deletion: Against φ-(in)completeness. Syntax 4(1): 147–163. Chomsky, N. 1995. The Minimalist program. Cambridge MA: The MIT Press. Chomsky, N. 2000. Minimalist inquiries: The framework. In Step by Step: Studies in Honor of Howard Lasnik, R. Martin, D. Michaels & J. Uriagereka (Eds), 89–155. Cambridge MA: The MIT Press. Chomsky, N. 2001. Derivation by phase. In Ken Hale: A Life in Language, M. Kenstowicz (Ed.), 1–52. Cambridge MA: The MIT Press. Chomsky, N. 2004. Beyond explanatory adequacy. In Structures and Beyond, A. Belletti (Ed.),

Agreement and concord in nominal expression 

104–31. Oxford: OUP. Chomsky, N. 2005. Three factors in language design. Linguistic Inquiry 36(1): 1–22. Cinque, G. 1980. On extraction from NP in Italian. Journal of Italian Linguistics 5: 47–99. Cinque, G. 1994. On the evidence for partial N-movement in the romance DP. In Paths Towards Universal Grammar, G. Cinque, J. Koster, J.-Y. Pollock, L. Rizzi & R. Zanuttini (Eds), 85–110. Washington DC: Georgetown University Press. Cinque, G. 1999. Adverbs and Functional Heads. Oxford: OUP. Cinque, G. 2003. On Greenberg’s universal 20 and the semitic DP. In Grammar in Focus Festschrift for Christer Platzack 18 november 2003, Vol. 2, L.-O. Delsing, C. Falk, G. Josefsson & H. Sigurðsson (Eds), 65–73. Lund: Wallin and Dalholm. Cinque, G. 2004. A phrasal movement analysis of the Romanian DP. In Studia Linguistica et Philologica in Honorem D. Irimia, A. Minuţ & E. Munteanu (Eds), 129–142. Editura Universităţii ‘A.I. Cuza’. Iaşi. Cinque, G. 2005. Deriving Greenberg’s universal 20 and its exceptions. Linguistic Inquiry 36(3): 315–332. Cinque, G. 2007. The dual source of adjectives and phrasal movement in the DP. MS. Ca’ Foscari University of Venice. Coene, M. 1999. Definite Null Nominals in Romanian and Spanish. A Generative Approach to the Internal Syntax of DP. Ph.D. dissertation, University of Antwerp. Cornilescu, A. 1992. Remarks on the determiner system of Romanian. Probus 4(3): 189–260. Cornilescu, A. 1995. Romanian genitive constructions.  In Advances in Roumanian Linguistics, G. Cinque & G. Giusti (Eds), 1–54. Amsterdam: John Benjamins. Cornilescu, A. 2003. Romanian genitive constructions revisited. Bucharest Working Papers in Linguistics 5(1): 104–128. Dal Pozzo, L. 2008. Reconsidering the Finnish possessive system. Annali di Ca’ Foscari. XLVI . Demuth, K. 1988. Noun classes and agreement in Sesotho acquisition. In Agreement in Natural Languages: Approaches, Theories, and Descriptions, M. Barlow & C.A. Ferguson (Eds), 305–321. Stanford CA: CLSI. Demuth, K. 2001. Prosodic constraints on morphological development. In Approaches to Bootstrapping, Vol. 2, J.Weissenborn & B. Hohle (Eds), 3–22. Amsterdam: John Benjamins. Demuth, K. 2003. The acquisition of Bantu languages. In The Bantu Languages [Routledge Language Family Series], D. Nurse & G. Philippson (Eds), 209–222. London: Routledge. D’Hulst, Y., Coene M. & Tasmowski L. 2000. Last resort strategies in DP: Article reduplication in Romanian and French. In Studies in Romanian Syntax, V. Motapanyane, (Ed.), 135–176. Leiden: Elsevier. Dimitrova-Vulchanova, M. & Vulchanov, V. 2007. An article evolving: The case of old Bulgarian. To appear in Proceedings of DIGS 8. D. Jonas & S. Anderson (Eds). Oxford: OUP. Dobrovie-Sorin, C. 1987. A propos de la structure nominal en Roumain. Rivista di Grammatica Generativa 12: 126–151. Dobrovie-Sorin, C. 2000. (In)definiteness spread: From Romanian genitives to Hebrew construct state nominals. In Comparative Studies in Romanian Syntax [North-Holland Linguistic Series 58], V. Motapanyane, (Ed.), 177–193. Elsevier: Amsterdam. Giorgi, A. & Longobardi, G. 1991. The Syntax of Noun Phrases. Cambridge: CUP. Giorgi, A. & Pianesi, F. 1997. Tense and Aspect. From Semantics to Morpho-Syntax. Oxford: OUP. Giusti, G. 1994a. Heads and modifiers among determiners. In Advances in Roumanian, G. Cinque & G. Giusti (Eds), 103–125. Amsterdam: John Benjamins.

 Giuliana Giusti Giusti, G. 1994b. Enclitic article and double definiteness: A comparative analysis of nominal structure in Romance and Germanic. The Linguistic Review 11: 241–255. Giusti, G. 1996. Is there a TopP and a FocP in the noun phrase? University of Venice Working Papers in Linguistics 6: 105–128. Giusti, G. 1997. The categorial status of determiners. In The New Comparative Syntax, L. Haegeman, (Ed.), 95–123. London: Longman. Giusti, G. 2001. The birth of a functional category: From Latin ILLE to the Romance article and personal pronoun. In Current Studies in Italian Syntax. Studies in Honor of Lorenzo Renzi, G. Cinque, Guglielmo & G. Salvi (Eds), 157–171. Dordrecht: North-Holland. Giusti, G. 2002. The functional structure of noun phrases: A bare phrase structure approach. In Functional Structure in DP and IP: The Cartography of Syntactic Structures, Vol. 1, G. Cinque (Ed.), 54–90. Oxford: OUP. Giusti, G. 2005. At the left periphery of the Romanian noun phrase. In On Space and Time in Language, M. Coene & L. Tasmowski (Eds), 23–49. Cluj: Clusium. Giusti, G. 2006. Parallels in clausal and nominal periphery. In Phases of Interpretation, M. Frascarelli (Ed.), 151–172. Berlin: Mouton de Gruyter. Giusti, G. & Gozzi, R. 2006. The acquisition of determiners. Evidence for the full competence hypothesis. In Language Acquisition and Development: Proceedings of GALA 2005, A. Belletti, E. Bennati, C. Chesi, E. Di Domenico & I. Ferrari (Eds), 232–238. Newcastle: Cambridge Scholars Press. Gozzi, R. 2004. The acquisition of determiners. A longitudinal study on an Italian child. BA thesis, Ca’ Foscari University of Venice. Grimshaw, J. 1991. Extended projections. MS, Brandeis. Grohmann, K.K. & Panagiotidis, P. 2005. An anti-locality approach to Greek demonstratives. In Contributions to the Thirtieth ‘Incontro di Grammatica Generativa’. Venice, February 26–28, 2004, L. Brugè et al. (Eds), 243–264. Venezia: Libreria editrice cafoscarina. Grosu, A. 1988. On the distribution of genitive phrases in Rumanian. Linguistics 26: 931–120. Grosu, A. 1994. Three Studies in Locality and Case. London: Routledge. Hornstein, N., Nunes J. & Grohmann, K.K. 2005. Understanding Minimalism [Cambridge Textbooks in Linguistics]. Cambridge: CUP. Horrocks, G. & Stavrou, M. 1987. Bounding theory and Greek syntax: Evidence of wh-movement in NP. Journal of Linguistics 23: 79–108. Katamba, F. 2003. Bantu nominal morphology. In The Bantu languages [Routledge Language Family Series], D. Nurse & G. Philippson (Eds), 103–120. London: Routledge. Kayne, R.S. 1994. The Antisymmetry in Syntax. Cambridge MA: The MIT Press. Kayne, R.S. 2004. Prepositions as probes. In Structures and Beyond, A. Belletti (Ed.), 192–212. Oxford: OUP. Kornfilt, J. 1997. Turkish [Routledge Language Family Series]. Routledge: London. Laenzlinger, C. 2005. French adjective ordering. Perspectives on DP-internal movement types. Lingua 115: 645–689. Larson, R. 1988. On the double object construction. Linguistic Inquiry 19: 335–391. Longobardi, G. 1994. Reference and proper names. Linguistic Inquiry 25(4): 609–665. Longobardi, G. 1996. On the typological unity of Indoeuropean and Semitic genitive case. In Studies in Afroasiatic Grammar, J. Lecarme, J. Löwenstamm & U. Shlonsky (Eds), 179–214. The Hague: HAG. Matushansky, O. 2006. Head movement in linguistic theory. Linguistic Inquiry 37(1): 69–109. Mišeska Tomič, O. 2006. Balkan Sprachbund Morphosyntactic Features. Dordrecht: Springer

Agreement and concord in nominal expression 

Nurse, D. & Hinnebush, T.J. 1993. Swahili and Sabaki: A Linguistic History. Berkeley CA: University of California Press. Ritter, E. 1988. A head-movement approach to construct state noun phrases. Linguistics 26: 909–929. Ritter, E. 1991. Two functional categories in noun phrases: Evidence from modern Hebrew. In Perspectives on Phrase Structures [Syntax and Semantics 25], S. Rothstein (Ed.), 37–62. San Diego CA: Academic Press. Rizzi, L. 2006. On the form of chains: Criterial positions and ECP effects. In Wh-Movement: Moving on, L. Cheng & N. Corver (Eds), Cambridge MA: The MIT Press. Shlonsky, U. 2004. The form of Semitic noun phrases. Lingua 114(12): 1465–1526. Siloni, T. 1997. Noun Phrases and Nominalizations. The Syntax of DP. Dordrect: Kluwer. Sportiche, D. 1988. A theory of floating quantifiers and its corollaries for constituent structure. Linguistic Inquiry 19(4): 425–449. Szabolcsi, A. 1994. The noun phrase. In The Structure of Hungarian [Syntax and Semantics 27], F. Kiefer & K.É. Kiss (Eds), 179–274. San Diego CA: Academic Press. Valois, D. 2006. Adjectives: Order within DP and attributive adjectives. In The Blackwell Companion to Syntax, Vol. 1, M. Everaert & H. van Riemsdijk (Eds), 61–82. Oxford: Blackwell. Visser, M. 2006. DP structure in Isixhosa: Definiteness properties and the occurrence of the noun class prefix. Handout. The Bantu-Romance Connection. LeEds, May 25–27, 2006.

A unified syntactic analysis of Italian and Luganda nouns Franca Ferrari-Bridgers

Department of Speech and Communication Studies, IONA College In this paper, I propose a unified syntactic analysis of Luganda and Italian simple nouns. I argue that Italian and Luganda nouns are formed in the syntax via merge and move operations. More specifically, I show that in both languages all nouns are formed via the merger of the nominalizer head [n] with a nominal stem [LP] yielding the nominal structure [nP [n [LP]]] and that syntactic movement is necessary in the noun formation process of Italian nouns to derive the correct morpheme order. In order to prove that the structure [nP [n [LP]]] is representative for both languages, I demonstrate that the nominalizer head [n] corresponds to both the Italian gender feature and the Luganda class feature and that, therefore, gender and class are the same feature. The data analysis in sections (2) and (3) of this paper supports the claim that gender and class are the same feature because of their identical inflectional and derivational functions. At the inflectional level, gender and class trigger VP and DP agreement and at the derivational level gender and class function as n-marked heads whose merger with an XP yields a noun.

1. Theoretical framework and proposal In the last ten years, many scholars have proposed a novel syntactic approach to word formation processes, showing that words, very much like sentences, can be formed in the syntax via Merge and Move operations: see Marantz (1997); Josefsson (1998); Kihm (2001); Alexiadou (2001); Lacarme (2001); Julien (2002); Pylkkänen (2002); Ferrari-Bridgers (2005), (2006); and Lowerman (2006). In contrast to a strict lexicalist approach to word formation, which considers words as lexical units impenetrable by the syntax (see Di Sciullo & Williams 1987: 47), the syntactic approach claims that words do not exist as morphological units in the Lexicon. Rather, they are the product of the merger of the lexical features

 Franca Ferrari-Bridgers

n/v/a, which are considered functional heads, with a lexical head (L), yielding either a noun, a verb or an adjective as represented in (1): (1) a. [nP [n [LP]]] b. [vP [v [LP]]] c. [aP [a [LP]]]

Though there are several interpretations in the literature with regard to the nature of the lexical head (L), (see Ferrari 2005 for a full discussion), most authors agree that (L) projects in the syntax as a lexical head separate from the functional feature categorizing it (see Picallo (1991); Marantz (1997); Ferrari (2005), among many others). Following Ferrari’s (2005: 30) quantitative data analysis, in this paper I assume that lexical heads (L) are stems, i.e., roots [√] categorically marked for one of the three functional features n/v/a. There are three types of stems: verbal stems [√+v], adjectival stems [√+a] and nominal stems [√+n]. Because stems are categorically marked, they are morphologically analyzable by the syntax and therefore they project in the syntax as independent LPs. Within this syntactic framework, in this paper I propose a comparative syntactic analysis of the Italian and the Luganda nominal system focusing in particular on the derivation of simple nouns.1 I argue that Italian and Luganda simple nouns are derived in the syntax via the merger of a nominalizer head [n] with an LP and that their underlying representation conforms to the structure given in (1a). In order to prove that the structure in (1a) is representative for both languages, I argue the following: (i) the nominalizer head [n] in (1a) corresponds to the Italian gender feature and to the Luganda class feature and, consequently, gender and class are the same feature; and (ii) syntactic movement is a necessary operation in Italian noun formation process to derive the correct output morpheme order. With regard to the claim that gender and class have a similar functional nature, Corbett (1991: 19) and Kihm (2001: 2) among many others point out that gender and class have the identical inflectional function of triggering VP and DP agreement. The morphological analysis of Italian and Luganda simple nouns proposed in section (2) of this paper shows that in these two languages, gender and class

. In morphological terms, there are three types of nouns: simple, derived and compound nouns. Simple nouns are formed with one nominal stem; derived nouns are formed via the merger of derivational morphemes with a nominal or a non-nominal stem; and compound nouns are formed by the merger of two nominal and/or non-nominal stems.

A unified syntactic analysis of Italian and Luganda nouns 

also have the identical inflectional function of triggering agreement. Furthermore, in section (3) of this paper, I provide evidence that gender and class also share an identical derivational function, i.e., they are nominalizer n-heads whose merger with an LP or an XP yields a noun. Though the derivational nature of gender and class has been previously observed by Ritter (1995) and Lacarme (2002) for the gender feature, and by Kihm (2001); Mufwene (1980); Myers (1990); Katamba (2003) and Schadeberg (2003) for the class feature, the data analysis presented in (3) unifies the two languages by showing that the derivational nature of gender and class depends on the fact that gender and class correspond to the functional feature [n]. In other words, contrary to what was previously assumed by Picallo (1991) and Myers (1987), gender and class are not two distinct functional heads external to the NP and selecting for a NP, but they are the functional feature [n], i.e., the head of the nP itself. With regard to the second claim relative to syntactic movement, in section (4) of this paper I show that structure (1a) alone is still insufficient to account for the fact that gender and class markers occupy symmetric positions with respect to the lexical head (L). In Luganda, in fact, the class markers are prefixed to an LP, whereas in Italian the gender markers have traditionally been assumed to be suffixed to the base as indicated in (2): (2) a.

mu-ntu class1-person ‘person’

b. cas-a house-feminine gender ‘house’

Under the assumption that languages are uniform at their core level of representation, I argue that the structure in (1a) is representative of all Luganda simple nouns as well as the default class of Italian nouns, i.e., masculine nouns. Through a reanalysis of Italian masculine noun morphology, I show that the nominalizer head [n] of masculine nouns has no overt morphological realization, given that masculine noun endings 〈o〉 and 〈e〉 are not morphemes but epenthetic segments added to the nP at PF. On the other hand, the merger of [n] with an LP alone is not sufficient to account for the derivation of Italian feminine nouns, whose n-feature is morphologically realized with the suffix 〈a〉. I show that the suffix position of the marker 〈a〉 depends on movement of (L) to the left of [n] and that it is not the effect of a parametric variation in the position of the feature [n].

 Franca Ferrari-Bridgers

2. I talian gender and Luganda class systems and the nature of the feature [n] In the first part of this section, I will briefly introduce the Italian and Luganda nominal systems (2.1, 2.2), showing that gender and class have the identical inflectional function of triggering agreement. In the next section, I argue that gender and class are n-marked features which have identical derivational functions in the noun formation process. Because gender and class have identical grammatical functions, I conclude that they are the same feature. 2.1 Italian nominal system In Italian, all nouns are marked either for masculine or for feminine gender. Gender is obligatory for purposes of DP and VP agreement as indicated in the following examples: (3) La bell-a cas-a ross-a è stat-a distrutt-a The-f beautiful-f house-f red-f has been-f destroyed-f ‘The beautiful red house has been destroyed.’

Looking at the endings of nouns, it is possible to establish five noun types: (i) masculine nouns ending in 〈o〉; (ii) feminine nouns ending in 〈a〉; (iii) masculine and feminine nouns ending in 〈e〉; (iv) recent loans; and (v) a residue class of nouns, all as represented in table (1): Table 1. Italian nominal system Type

Gender

sg.

Examples

〈o〉 〈a〉 Epenthetic Loans Residue

Mas. Fem. Mas. Fem. Mas. Mas. Fem.

-o -a -e -e 0 è,ì,ò à,ù,

tavolo “table” -i casa “house” -e fiore “flower” -i parete “wall” -i bar 0 caffè “coffee” 0 città “city”

pl.

Examples tavoli “tables” case “houses” fiori “flowers” pareti “walls” bar “bars” caffè “coffees” città “cities’

With regard to gender morphology, many Italian grammars have claimed the existence of a morphological correlation between noun endings and gender. More precisely, nouns ending in 〈o〉 have been associated with masculine gender, and nouns ending in 〈a〉 with feminine gender. Though these correlations have been questioned (Thornton 2001: 485), quantitatively they retain a certain amount of validity. Despite

A unified syntactic analysis of Italian and Luganda nouns 

the presence of a small number of exceptions, according to Ferrari’s (2005) Italian noun database containing over 4,000 nouns, masculine nouns ending in 〈o〉 are 32% of the total number of nouns in the database; and feminine nouns ending in 〈a〉 make up another 32%. Feminine nouns ending in 〈o〉 (as in (4a)) and masculine nouns ending in 〈a〉 (as in (4b)) are less than 1% of the nouns in the database. (4) a. moto ‘motorbike’, mano ‘hand’, metro ‘underground’ b. poema ‘poem’, sistema ‘system’, tema ‘theme’

Nouns ending in 〈e〉, recent loans, and residue nouns constitute three separate groups, whose gender feature is expressed not through morphemes but through DP or VP agreement as shown in (5a–b): (5) a.

il sole cald-o the-m sun hot-m ‘the hot sun’

b. il bar nuov-o the- bar new-m ‘the new bar’

With regard to nouns ending in 〈e〉, 〈e〉 is not a morpheme but an epenthetic segment added to the stem for syllabification purposes in order to avoid word-final coda consonants. The epenthetic nature of 〈e〉 is evident in its tendency to undergo deletion when it is preceded by a sonorant consonant (l, r, n, or m), i.e., the phenomenon of “troncamento” (truncation). The elision of this segment, however, does not deprive the noun of its gender information. Gender is still retrievable from DP agreement as illustrated in (6): (6) il sole levante/ il sol levante the-m sun rising/ the-m sun rising ‘the rising sun’

It is interesting to note that the deletion of the segment 〈e〉 does not take place when it functions as a feminine plural morpheme, as (7a) shows. Similarly, the feminine singular gender marker 〈a〉 does not undergo deletion as shown in (7b): (7) a.

Gianni mostra le suol-e / *suol rovinat-e delle sue scarpe Gianni shows the soles-f. pl. worn-out-f.pl. of his shoes ‘Gianni shows the worn-out soles of his shoes.’ (Nespor 1993: 227)

b. Gianni mostra la suol-a/ *suol rovinat-a della sua scarpa Gianni shows the sole-f. sg. worn-out-f.sg. of his shoe ‘Gianni shows the worn-out sole of his shoe.’

With regard to loans, the majority of loans of inanimate nouns are masculine by default. According to Thornton (2003c), there are few feminine loans whose gender is assigned

 Franca Ferrari-Bridgers

manually. According to Ferrari’s database (2005), loans account for 6% of the nouns in the database and less than 0.5% of these nouns are feminine. Loans of nouns denoting human beings, instead, undergo semantic gender assignment; i.e., they are feminine if they denote a female person and masculine if they denote a male person. Finally, nouns in the residue class are few in number and belong to closed classes of number-invariable nouns. It is possible to predict the gender of these nouns by looking at their final vowel. For instance, all nouns ending in 〈à〉, 〈tù〉 and 〈i〉 are feminine: e.g., città “city”, virtù “virtue”, crisi “crisis”; whereas nouns ending in 〈è〉, 〈ù〉 and 〈ì〉, which are for the most part loans, are masculine: e.g., caffè “coffee”, iglù “igloo”, colibrì “humming bird”. 2.2 Luganda As with many Bantu languages, Luganda is characterized by an elaborate noun class system which accounts for a large number of classes with each noun class marked for a specific set of agreement markers. According to Ashton et al. (1954); Welmers (1973) and Katamba (2003: 109), in Luganda there are 19 noun classes.2 This number is derived by looking at the number of noun class prefixes as well as by looking at the grammatical agreement markers occurring on nouns, adjectives, numerals, determiners and pronouns. Noun classes are usually numbered in pairs, e.g., class 1/2 and class 3/4. Each pair is characterized by the presence of two noun class prefixes, expressing identical class information, but differing in their number morphology. The first member of the pair is associated with the singular while the second element indicates the plural, e.g., Luganda cl.1/2: mu-ntu “person” – bantu “people”, as illustrated in table (2): As in the case of gender for Italian, class for Luganda is obligatory for the purposes of agreement. According to Maho’s (1999: 128–129) typological classification of Bantu noun class systems, Luganda belongs to the traditional formal type, i.e., type A. In Bantu Type A languages, all agreement markers (in Bantu terminology “concords”) are chosen on the basis of the class information inherent in the stem, disregarding the meaning of the stem. Noun class prefixes and concords simultaneously mark the same noun class. An example of the Luganda agreement system (from Ashton et al. 1953:23) is given in (8) below: (8) Aba-ana ba-no aba-satu aba Mukasa te-ba-leese kintu. cl1-children cl1-these cl1-three cl1-of Musaka not-cl1-brought anything ‘These three children of Mukasa have not brought anything.’ . There are 3 additional locative classes which for the moment I do not include in the discussion, as they are not pertinent to the present analysis.

A unified syntactic analysis of Italian and Luganda nouns 

Table 2. Luganda nominal system Class

Prefix

Singular

1 3 5 6a. 7 9 11 12 14 15 18 20

mu- mu- li-/ri-/e- ma- ki- 0/n- lu- ka- bu- ku- tu- gu-

mu-ntu (man) mu-ti (tree) ri-nnya (name) ma-zzi (water) ki-ntu (thing) n-te (cow) lu-naku (day) ka-ntu (small thing) bu-ntu (manhood) ku-lima (cultivating) tu-zzi (drop of water) gu-ntu (huge thing)

Class

Prefix

Plural

2 4 6

ba- mi- ma-

ba-ntu (people) mi-ti (trees) ma-nnya (names)

8 10 10 13

bi- 0/n- 0/n- bu-

bi-ntu (things) n-te (cows) n-naku (days) bu-ntu (small things)

22

ga-

ga-ntu (huge things)

From the Italian and Luganda data in (3), (5) and (8) and from the above observations, one can draw the conclusion that gender and class have the identical inflectional function of triggering agreement at the DP and VP/IP level.

3. Gender and class as n-marked Features As seen above, gender and class have an identical inflectional nature. In this section, I argue that gender and class also have an identical derivational nature. They are n-marked features whose merger with a nominal stem, e.g., [√+n], and non-nominal bases, e.g., [√+a], [√+v], VP, VoiceP, ApplicativeP and other types of XPs, generates a noun. The data analysis of Italian and Luganda derived nominals in sections (3.1, 3.2 and 3.3.) of this paper confirms the validity of my argument. 3.1 De-nominal noun formation De-nominal nouns are derived nouns formed via the merger of an n-marked derivational head with an n-stem. In Luganda all de-nominal nouns are formed via the merger of class morphemes with n-stems (see 9a–d), whereas in Italian de-nominal nouns are formed either via the merger of derivational morphemes such as 〈-eria〉, 〈-ile〉 with an n-stem, as in pizza 〈pizz-eri-a “pizza place” and fieno “hay” fienile “barn”, 〉 or via the merger of a gender marker with a nominal stem as in (9e–k):

 Franca Ferrari-Bridgers

(9)

a. b. c. d. e. f. g. h. k.

b-beere (cl.5) ‘breast’ > Ki-beere (cl.7) ‘udder’ ki-batu (cl.7) ‘palm of the hand’ > lu-batu (cl.11) ‘handful’ bu-ligo (cl.14) ‘filth’ > mu-ligo (cl.1) ‘dirty person’ ndiga (cl. 9) ‘sheep’ > ka-liga (cl.12) ‘lamb’ ferrament-o (mas) ‘iron tool’ > ferrament-a (fem) ‘hardware’ ranci-o (mas) ‘ration’ > ranci-a (fem) ‘the distribution of the ration’ buc-o (mas) ‘small hole’ > buc-a (fem) ‘pit’ or ‘big hole’ ciucci-o ‘pacifier’ (mas) > ciucci-a (fem) ‘breast’ mel-o (mas) ‘apple tree’ > mel-a (fem) ‘apple’

As the data in (9a–d) show, in Luganda the change of the class morpheme corresponds to the formation of a new noun. Similarly, in Italian (9e–k), the passage from one gender to another modifies the meaning of the stem, giving rise to a new noun.3 3.2 De-adjectival noun formation De-adjectival nouns are formed via the merger of a class or gender morpheme with an adjectival stem. (10)

a. b. c. d.

avu ‘poor’ > Bw-avu ‘poverty’ genge ‘leprous’ > Mu-genge ‘leper’ lungi ‘beautiful’ > bu-lungi ‘beauty’ > mu-lungi ‘beautiful person’ lebbroso ‘leprous’ > lebbros-o ‘leper male’, lebbros-a ‘leper female’

It is important to point out that, although in Italian the majority of de-adjectival nouns are formed via derivational heads such as 〈ezz-a〉, 〈ur-a〉, 〈izi-a〉 etc. as in “bello” (beautiful) > “bellezza” (beauty), de-adjectival nouns formed with gender markers 〈o〉 and 〈a〉 are used mainly to derive nouns denoting human beings. 3.3 De-verbal noun formation De-verbal nouns are formed via the merger of gender and class morphemes with either a verbal base or a constituent which in Italian and Luganda can range from a verb stem to a VP, a CausativeP or a VoiceP: (11) a.

Mu-bb-i cl.1-vstem-bb-i (from kubba ‘to steal’) ‘robber’

b. Bu-bb-i cl.14-vstem-bb-i ‘robbery’

. Though it has often been claimed in the literature that such gender alternation is unproductive for inanimate nouns and is limited to a few fossilized forms, the above data show that gender alternation as a noun formation device is in fact productive. Pairs such as ciuccio-ciuccia, for instance, are relatively new formations (1951). For more data and discussion about the productivity of such alternations see Ferrari (2005).

c.

A unified syntactic analysis of Italian and Luganda nouns 

Mu-ba-zz-i cl.1-vstem-ba-CAUSATIVE-i (from kubazza ‘to cause to do carpentry work’) ‘carpenter’

d. mu-zaal-is-a cl.1 vstem-CAUSATIVE-a (kuzaalisa ‘to assist at childbirth’) ‘midwife’ e.

s-som-er-o cl5 vstem-APPLICATIVE-o ( from kusom-er-a ‘to read for’) ‘school’

With regard to the data in (11), the nouns in (11c,d) are built on the Causative head 〈iz-is〉 and the noun in (11e) is built on the applicative head 〈er-a〉. According to Pylkkänen’s (2002) syntactic representation of Luganda’s IP, causative and applicative morphemes are heads which project independently from the VP respectively as CausativeP and ApplicativeP. It follows, therefore, that the class morpheme merges directly with a causative or applicative phrase to derive a noun. It is interesting to note that the Luganda de-verbal nouns given in (11) have different suffixal endings, e.g., [-i,-o,-a,-e,-u]. According to Myers (1991: 38–39), the final vowels on verbal stems are morphemes: the vowel 〈a〉 is an inflectional morpheme indicating an active verb form; and the remaining vocalic segments are derivational morphemes which signal the presence of an n-feature (or an afeature). More specifically, for Ashton et al. (1954: 371–379): (i) the suffix 〈i〉 is associated with the feature agent/instrument; (ii) the suffix 〈e〉 is associated with the feature patient; (iii) the suffix 〈o〉 is associated with the feature indicating “state” or is found on nouns which express “the result of an action”; and (iv) the marker 〈u〉 is found on adjectives or on de-adjectival nouns. Notwithstanding the foregoing, Ferrari (2005) points out several exceptions to these vowel-meaning correlations which make it difficult to assert the morphological and semantic values of these suffixes. For instance, in the case of the noun ending 〈i〉, there are many de-verbal nouns ending in 〈i〉 which indicate abstract concepts (see 11b). At the same time, 〈i〉 can also be found as an adjective ending or on de-adjectival nouns which do no necessarily express an agent (see 10c). A similar lack of a one-to-one relation between suffix ending and meaning is found for the suffix 〈o〉. De-verbal nouns ending in 〈o〉 are found across different noun classes and express: abstract concepts (e.g., bu-fumbo “marriage”, en-rooto “dream”); as well as things/objects (e.g., mu-biniko “cover”, ki-sumuluzo “key”); and agentivity (e.g., mu-logo “sorcerer”, mu-fumbiro “cook”). The same can be said for nouns ending in 〈a〉 which can be: agents (e.g., mu-yigiriza “teacher”, muwooza “tax collector” and midwife in (11d)); abstract nouns (e.g., bu-yinza “power”, mpisa “habit”); as well as objects/instruments (e.g., ki-kulumbaza “microscope”).

 Franca Ferrari-Bridgers

Finally, it is interesting to point out that, in Luganda, certain de-verbal nouns indicating agents can end in either 〈i〉 or 〈a〉 indiscriminately, e.g., mu-goba “driver”/ mu-gobi “driver”, mu-sika “heir”/musisi “heir”. From the above data, it is impossible to conclude that the vocalic endings associated with de-verbal nouns are morphemes expressing specific semantic and lexical information or playing clearly defined roles in derivational processes. Although the data show that the relation between noun endings and meaning/ function is not always systematic, at the present time there is not enough data to confirm or discount the hypothesis that final vocalic segments are types of derivational morphemes. I table that discussion for a future stage of my research. Moreover, given that the focus of the present paper is on the derivation of simple nouns, i.e., nouns whose LP is an n-stem, I will consider the final vocalic segment to be part of the n-stem rather than an independent head. With regard to Italian de-verbal nouns, one strategy for deriving de-verbal nouns is to add the feminine or the masculine gender marker to a verbal stem as in the following examples: (12)

a. b. c. d.

abbandonare comandare accumulare deliberare

‘to abandon’ > abbandono (N) ‘abandonment’ ‘to command’ > comando (N) ‘command’ ‘to accumulate’ > accumulo (N) ‘accumulation’ ‘to deliberate’ > delibera (N) ‘deliberation’

Evidence that the derivational process goes from a verbal stem to a noun and not the other way around is the presence of pairs of de-verbal nouns, one formed with a gender marker as in (13) and the other formed with de-verbal suffixes such as 〈mento〉 and 〈zione〉 as in the subsequent examples: (13) a. Delibera/deliberazione ‘deliberation’ b. Accumulo/accumulazione ‘accumulation’ c. Rettifica/rettificazione ‘rectification’

〈mento〉 and 〈zione〉 are derivational morphemes which select exclusively for a verbal stem. It follows that abbandon and comand are verbal stems. From the examples in (12) and (13), I propose that gender markers also function as derivational morphemes in the same way as 〈mento〉 and 〈zione〉. Finally, a second group of productive de-verbal nouns comprises nouns ending in 〈ito/ita〉 and 〈ato/ata〉 , which are built on a past participle base as shown in (14): (14) a.

Il cucit-o Past participle – masculine gender ‘the art of sewing’

b. una cucit-a Past participle – feminine gender ‘the single act of sewing’

A unified syntactic analysis of Italian and Luganda nouns 

According to Scalise (1992), in these nouns, the feminine gender marker 〈a〉 functions as a derivational morpheme. The same is also true for the masculine marker 〈o〉. Assuming, like Cinque (1999), that a past participle is built on VoiceP, it follows that the gender marker is merged with a VoiceP rather than with a simple v-stem. With regard to the derivation of de-verbal nouns, as demonstrated by Alexiadou 2001 and Ferrari-Bridgers (2006), the merger of the feature [n] with a VoiceP or ApplicativeP yields a noun whose underlying structure includes a large chunk of the vP-IP structure. For instance, using Cinque’s (1999) or Pylkkänen’s (2002) representation of the vP-IP structure, a de-verbal noun built on a VoiceP as in (14b) shows the following underlying structure: (15) [nP [n [VoiceP [AspP4 [CausP [ApplP [XP [[vP [VP [LP]]]]]]]]]]]

In summary, both the Italian and the Luganda data confirm the initial claim that gender and class morphemes function as derivational n-heads, whose merger with different types of bases forms a noun. Given it is through the presence of a gender/ class morpheme that an XP acquires a nominal reading, one can conclude that gender and class correspond to the nominalizer feature [n]. From the data and observations in sections (2) and (3), one can conclude that gender and class are the same feature at both the derivational and inflectional level. 4. The syntactic representation of simple nouns As stated above, I claim that both Italian and Luganda simple nouns are formed via the merger of a nominalizer head [n] with a nominal stem. The underlying representation for nouns in both languages is given in (16): (16) [nP [n [√+(n)]]]

Before explaining how the structure in (16) applies to both Italian and Luganda nouns, I outline briefly the syntactic framework underlying my analysis. 4.1 The Framework Syntactically, I assume that Kayne’s Linear Correspondence Axiom (LCA) (1994) applies at all levels of representation, i.e., at CP, IP, VP, DP and nP levels.

. In (19), the term AspectP comprises a set of aspectual projections identified by Cinque (1999) to be hierarchically organized as follows: Celerative aspect (ii) > Repetitive aspect (ii) > Frequentative aspect (ii) > Completive Aspect (ii) > CausativeP … XP.

 Franca Ferrari-Bridgers

Kayne (1994: 40–41) suggests that compound nouns and verbs such as “can opener” and “overturn” are the result of the adjoining of separate XPs followed by movement. In the same vein, in this paper I claim that simple nouns as represented in (16) are the result of adjoining two separate XPs and, like all other major constituents, are subject to the LCA. According to the LCA, specifiers are universally merged on the left of X’ and all movement is upward and leftward. Because specifiers are merged obligatorily and solely on the left of the head, Kayne (1994: 47) suggests that UG imposes a Spec-Head-Complement order which makes it unlikely to account for word order variations through a “directionality parameter”. This implies that many of the word order differences between languages or “sub-parts of languages” are not the by-product of a difference in their parametric setting, but the result of different combinations of movements. With regard to movement within the LCA, Kayne (1998, 2000, 2005); Koopman & Szabolcsi (2000) and Roeper (1999) consider syntactic movement leftward, overt and preferably of the phrasal type. Within this syntactic framework, I illustrate how the structure in (16) and the LCA accounts for the representation of Luganda and Italian nominals in (4.2); I discuss the modality of movement in simple noun formation processes in (4.3); and, finally, I rule out the possibility that the symmetric order of the feature [n] in Italian and Luganda is the result of a parametric variation in (4.4). 4.2 The analysis With regard to the specific derivation of nouns in Italian and Luganda, the SpecHead-Complement order predicts that, if the feature [n] is the head of the noun, then [n] selects LP as its complement. As a result, LP is found on the right of [n] as given in (16). It follows, therefore, that a final rightward position of the feature [n] with respect to LP, as in the case of Italian feminine nouns in (2b), is the result of movement. As previously established by the analysis in Section (3) above, the feature [n] corresponds to the gender/class feature. At the formal level, as also noted by Harley (2005: 12), the nominalizer [n] can be either phonologically null [ø] or phonologically realized. For instance, in English, Harley indicates that [n] is either null [ø] or realized as an n-marked derivational morpheme, e.g., 〈-ity〉, 〈-ment〉 and 〈-tion〉, etc. Similarly, in Luganda, [n] has different morphological realizations. The feature [n] is phonologically null for nouns of class 9 and for loans mostly belonging to class 9. The feature [n] is morphologically realized for nouns of all the other

A unified syntactic analysis of Italian and Luganda nouns 

classes, including loans found in class 5. Based on these observations, I propose the following syntactic representation of Luganda simple nouns in (17) below: (17)

a. b. c. d.

[nP [ø-cl.9 [bbululu-n]]] ‘blue’ [nP [ki-cl.7 [ntu-n]]] ‘thing’ [nP [b-cl.5 [beere-n]]] ‘breast’ [nP [mu-cl.1[ntu-n]]] ‘man’

The structure in (17) is the result of the merger of the head [n] with a nominal stem. The feature [n] surfaces on the left of its LP complement in agreement with the head-complement order imposed by the LCA. Move, therefore, does not intervene and the derivation of Luganda simple nouns is complete. Somewhat different and more complex is the case of Italian, for which I propose two different derivations. I argue that masculine nouns, represented in (18a–c), like Luganda simple nouns, are the result of the merger of the feature [n] with a nominal stem and do not require movement. The derivation of feminine nouns, on the other hand, includes syntactic movement of the LP as indicated in (19). (18) a. [nP [n-ø [tavol 〈o〉 -n]]] ‘table’ b. [nP [n-ø [bar-n]]] c. [nP [n-ø [fior 〈e〉 -n]]] ‘flower’ (19) a. [nP [n-a [cas-n]]] b. Movement of LP > [nP [cas-n]i [n-a [ti]]]

In accordance with the representations in (18a–c), I claim that the nominalizer head [n] is phonologically null, [ø], which means that the structure in (16) is able to account simultaneously for three types of masculine nouns: nouns ending in 〈o〉; nouns ending in 〈e〉; and loans. The null nature of [n] also provides insight as to why non-nominal parts of speech, such as adverbs and conjunctions, are masculine once they are nominalized although they do not show any masculine morphology; e.g., perchè “why” becomes il perchè “the-mas why”. Furthermore, the phonologically null nature of [n] predicts that the final vocalic segments 〈e〉 and 〈o〉 are not gender morphemes, but rather simple epenthetic segments added to the n-stem for syllabification purposes in order to avoid word-final coda consonants. We saw that this prediction is borne out for the vowel 〈e〉, whose epenthetic nature is exemplified in its tendency to undergo truncation as shown in (5) and (6) above. Following Cardinaletti & Repetti (2003: 7), I claim that the epenthetic nature of 〈o〉 is implicit in its morphological neutrality. The final vowel 〈o〉, in fact, can be considered as morphologically neutral, because it can be found in: (i) morphological salient positions, i.e., when 〈o〉 is suffixed to noun and adjective stems and

 Franca Ferrari-Bridgers

functions as a gender marker; and (ii) in morphological non-salient positions, i.e., as an ending for gerunds (cantando “singing”); adverbs (poco “a little”, molto “a lot”, tanto, “a lot”, presto “early”); quantifiers and indefinite pronouns/adjectives (uno “one”, qualcuno “someone”, altro “else”, nessuno “nobody”, tutto “all”, tanto “much”). According to Cardinaletti & Repetti (2003: 6), morphological saliency has been shown to be one of the factors which allow morphological neutral segments to be reanalyzed as inflectional morphemes. In Italian, for instance, stem final positions are considered morphologically salient, given that most of the morphological information is expressed through suffixation. This explains why, in loan words, the end vocalic segment 〈o〉 is often reanalyzed as masculine gender, e.g., mango > manghi, tango > tanghi. In addition to morphological neutrality, the epenthetic nature of 〈o〉 is also suggested by the fact that 〈o〉, as we saw for 〈e〉, is subject to deletion when preceded by (l, n, r, m), i.e., the phenomenon of “troncamento” as shown in (20): (20) a.

Che bel ciel-(o) sereno, che c’è stasera! ‘What a beautiful clear sky, there is tonight!’

b. L’oracol-(o) di dio non mente mai. ‘God’s oracle never lies.’

Nespor (1990) argues that “troncamento” is a lexical rule which applies mainly to verbs. Nominal and adjectival truncated forms are lexicalized items. However, according to Vogel (1983) and Meinschafter (2004), the rule of “troncamento” applies to all lexical categories. Meinschafter’s analysis of an extensive corpus of spoken Italian (300,000 words) shows that the reason why “troncamento” seems to operate more freely on verbs than on nouns and adjectives is linked to the fact that the prosodic domain of the final vowel deletion is the phonological phrase and not the intonational phrase, as previously assumed. Final vowel deletion is, therefore, subject to constraints on prosodic branchingness of the phonological phrase (Nespor & Vogel 1986), as she illustrated with the following two examples: (21) a.

Prosodic non-branching, final vowel deletion applies optionally [Per fare delle previsioni]f > per far delle previsioni

b. Prosodic branching, final vowel deletion does not apply [Per fare]f [delle previsioni]f > *[per far] [delle previsioni]

The impossibility of applying the rule at the boundaries of phonological phrases explains why verbs are more susceptible to “troncamento”. Nouns and adjectives, unlike verbs, often occur in the final position of a phonological phrase where they cannot easily undergo “troncamento”.

A unified syntactic analysis of Italian and Luganda nouns 

Finally, the examples in (22) show a group of masculine nouns which freely alternate between 〈o, e〉 or 〈o, e, ø〉 as their final endings, without a gender change. (22)

a. b. c. d. e. f.

forestier/forestiero/ forestiere ‘foreign’ ananas/ananasse/ananasso ‘pineapple’ corsier/corsiero/corsiere ‘charger’ destrier/destriero/destriere ‘war horse’ scialle/sciallo ‘shawl’ cece/cecio ‘chickpea’

Given the morphological neutrality of 〈o〉, its possibility of undergoing truncation and of being interchanged with 〈e〉 or 〈ø〉, I conclude that 〈o〉 has an epenthetic rather than a morphological nature. The epenthetic nature of 〈o〉 and 〈e〉 confirms my hypothesis of a phonologically null nature of the nominalizer [n] and validates my proposed representation for masculine nouns as given in (18a–c). Similarly to Luganda, the structure of simple masculine nouns in Italian does not presume syntactic movement. Once the feature [n] has been merged to the n-stem, the derivation is complete. The epenthetic vowels 〈e〉 and 〈o〉 are added to the structure outside the syntax, most likely at PF. With regard to feminine nouns, as shown in (19), the feature [n] is morphologically realized as 〈a〉. There is no doubt regarding the morphological nature of 〈a〉. Firstly, it cannot be deleted under truncation as illustrated in (7b) above, which suggests that 〈a〉 is a morpheme whose deletion causes a loss of information. Secondly, according to Ferrari (2005), Ferrari-Bridgers (2007) and Thornton (2001: 485, 2003a, 2003b), the correlation between feminine gender and the marker 〈a〉 is almost without exception from a quantitative standpoint. Finally, the morpheme 〈a〉 is used to express the features [+human,+female] in 99.9% of Italian nouns. In fact, there are only three counter-examples: moglie “wife”, vergine/vestale “virgin” and madre “mother”. Given the morphological nature of 〈a〉, it follows that 〈a〉 is added to the structure in the syntax. Notwithstanding the clear morphological nature of 〈a〉, the derivation of feminine nouns is not complete once the morpheme 〈a〉 has been merged into the structure as in (19a). According to the LCA, the final rightward position of [n] derives from leftward and upward movement of its complement. This suggests that movement is necessary to derive the correct output form for feminine nouns, where the head [n] sits on the right of its LP complement as represented in (19b). One of the most challenging questions is to determine whether the correct output order in feminine nouns is derived via phrasal movement or via head movement. In the next section, looking at diminutive formation, I will make an argument in favor of phrasal movement.

 Franca Ferrari-Bridgers

4.3 Phrasal movement versus head movement As stated above, the derivation of feminine nouns implies the presence of syntactic movement and, as suggested in (19b), I assume that movement involves the whole LP and not only the head L. The choice of phrasal movement over head movement inside a small syntactic unit such as simple nouns is based on the ability of LP, as a stem, to move independently from the feature selecting it, as I prove to be the case in the following analysis of Italian diminutives. In Italian, diminutives are intervening heads between the LP and the n-marker as shown in (23):5 (23) a.

cas-a house-gender ‘house’

b. cas-in-a house-diminutive-gender ‘little house’ c.

cas-ett-a house-diminutive-gender ‘little house’

Italian diminutive heads 〈-in-〉, 〈-ett-〉 are n-marked derivational heads which select for a whole nP rather than for a simple LP. The data in (23) show that Italian diminutives are not marked by their own gender and that an nP modified by a diminutive head retains its own gender. As gender in Italian is expressed via the feature [n], it follows that the diminutive head merges with a full nP as represented in (24): (24) [Dim (n)P [-in [nP [-a [LP [cas-]]]]]]

With regard to the syntactic derivation of nouns modified by diminutives, I propose a two step derivation process. Firstly, the diminutive n-head merges with the nP as represented in (25a, 26a) respectively for masculine and feminine nouns. The LP then moves to the spec of the diminutive projection to derive the correct order as illustrated in (25b, 26b).

. In Italian, adjectives can also be modified by the diminutive heads 〈in〉 and 〈ett〉. Semantically, however, the diminutive heads 〈in〉 and 〈ett〉 have different meanings depending on whether they are merged with an nP or with an aP. Contrary to the examples in (23) above, in the case of a merger with an adjective, for instance, the diminutive head 〈ett〉 forms an adjective with a negative connotation rather than a simple diminutive meaning, e.g., piccolletto does not mean “very small”, but it denotes in a negative way that a person is either too short, small or young. I interpret the meaning differences of diminutive heads used with nouns or with adjectives as an indication that these heads are not the same morphemes, but they are distinct lexical heads.

A unified syntactic analysis of Italian and Luganda nouns 

(25) a. [dimP [-in [nP [n -ø [LP alber-]]]]] b. [dimP [alber]i[-in [nP [n -ø [ti]]]]] (26) a. [dimP [-in [nP [n -a [LP cas-]]]]] b. [dimP [cas-]i[-in [nP [n -a [ti]]]]]

The structure in (24), however, cannot account for the derivation of feminine nouns if one assumes that the stem L moves via head movement. The movement of L as a head yields the wrong morpheme order as represented in (27): (27) a. [dimP [-in [nP [n-a [LP [cas-]]]]]] b. [dimP [-in [nP [n-casi-a [LP [ti]]]]]] c. *[dimP [[casi-a]y -in [nP [n-[ty][LP [ti]]]]]]

From the data in (27), one can conclude that XP movement is the only plausible type of movement able to derive diminutive formations.6 Given the similar nominal nature of feminine nouns and diminutive formation, I posit that there is no principled reason to assume the presence of two different types of movements, i.e., head movement for feminine noun formation and phrasal movement for diminutive formation. Therefore, I conclude that XP movement is also the type of movement used for the derivation of feminine nouns. 4.4 Movement vs. parametric variation The hypothesis that the final output order of morphemes in feminine nouns depends on movement and is not the result of a parametric difference in the position of [n] in Italian and in Luganda is based on the following observations. Firstly, as stated above, the structure proposed for nouns in (1a) accounts for the derivation of all Luganda nouns as well as for all masculine Italian nouns. Movement is, therefore, assumed to intervene in the derivation of feminine nouns only. Given that parameters are supposed to explain difference across languages, the hypothesis of a parametric setting seems theoretically implausible, as it would only account for the sub-class of feminine nouns. Secondly, if the position of [n] is a matter of parametric variation in the case of diminutive formation, then the diminutive head, which is n-marked, would also have to be merged on the right of the nP. Even discounting the fact that right merger is excluded by the LCA, if the diminutive is merged on the right of the nP, it would yield the wrong output form, e.g., *cas-a-in.

. The hypothesis that XP movement is the only plausible type of movement within diminutive formation processes is also compatible with a framework which assumes that diminutives are adjuncts attached below or above the number node of a noun (see Bachrach & Wagner 2007).

 Franca Ferrari-Bridgers

Finally, at the typological level, as Haegeman (1994: 19) suggests, each observed difference between two languages does not necessarily correspond to a single discrete parameter. Usually “the properties with respect to which languages vary tend to organize themselves in clusters which are stable across languages and which allows us to arrive at a typology of languages. If a language has a property X, it will also have the property Y and the property Z.” In the case of Italian and Luganda, the hypothesis of a parametric variation implies that the two languages should show other types of word order differences, such as occur with languages belonging to two different typological classes. But, according to Greenberg’s (1963) study on Universals and their relation with word order types, Italian and Luganda are part of the same class of languages labeled as type II. Both languages show VO order, have prepositions, and have Noun-Genitive and Noun-Adjective word order. The fact that Italian and Luganda are part of the same typological class weakens the hypothesis of the existence of a parametric variation as a means of explaining the differences within the [nP]. In other words, if the limited differences within the [nP] observed for Italian and Luganda were the result of a parametric difference, the two languages would display more word order differences. 5. Conclusion In this paper, I argue in favor of a unified syntactic analysis of Luganda and Italian simple nouns with the claim that nouns are formed via the merger of the nominalizer head [n] with a nominal stem, yielding [nP [n [LP]]]. My analysis shows that the nominalizer head [n] corresponds to both the Italian gender markers and the Luganda class markers. In reaching this conclusion, I further argue that the structure in (1a) is representative for all Luganda and masculine Italian nouns, whereas Italian feminine nouns are derived by movement. In addition, looking at diminutive formations, I argue that XP movement is the preferred type of movement in derived nominal formations. Consequently, I conclude, based on the data and the syntactic analysis, that by reinterpreting the gender/class feature as a nominalizer, it is possible to account uniformly for the derivation of nouns in two languages which belong to two distinct language families.

References Alexiadou, A. 2001. Functional Structure in Nominals. Amsterdam & Philadelphia: John Benjamins. Ashton, E.O. et al. 1954. A Luganda Grammar. London: Longmans, Green and Co. Bachrach, A. & Wagner, M. 2007. Syntactically driven cyclicity vs. output-output correspondence: The case of adjunction in diminutive morphology. In U. Penn Working Papers in Linguistics 10(1): 157–173.

A unified syntactic analysis of Italian and Luganda nouns 

Cardinaletti, A. & Repetti, L. 2003. Word-final epenthetic vowels and morphology in Romance. Paper presented at LSRL 2003. University of Indiana, Bloomington IN. Cinque, G. 1999. Adverbial and Functional Heads. Oxford: OUP. Corbett, G.G. 1991. Gender. Cambridge: CUP. Di Sciullo, A. & Williams, E. 1987. On the Definition of Word. Cambridge MA: The MIT Press. Ferrari-Bridgers, F. 2007. The predictability of gender in Italian. Lingua et Linguistica 1(1): 146–167. Ferrari-Bridgers, F. 2006. A syntactic analysis of Italian deverbal-nouns. In New Perspectives on Romance Linguistics, N. Chiyo & J.P.Y. Montreuil (Eds), 97–113. Amsterdam: John Benjamins. Ferrari, F. 2005. A Syntactic Analysis of the Italian and Luganda Nominal Systems: How Nouns Can Be Formed in the Syntax. Ph.D. dissertation, New York University. Greenberg, J. 1963. Some universal of grammar with particular reference to the order of meaningful elements. In Universals in Language, J.H. Greenberg (Ed.), 152–169. Cambridge MA: The MIT Press. Haegeman, L. 1994. Government and Binding Theory. Oxford: Blackwell. Harley, H. 2005. Bare phrase structure, acategorial roots, one-replacement and unaccusativity. In Harvard Working Papers on Linguistics 9, S. Gorbachov & A. Nevins (Eds), 45–63. Cambridge MA: Harvard University. Kayne, R.S. 1994. The Antisymmetry of Syntax. Cambridge MA: The MIT Press. Kayne, R.S. 2000. Parameters and Universals. Oxford: OUP. Kayne, R.S 2005. Movement and Silence. Oxford: OUP. Katamba, F. 2003. Bantu nominal morphology. In The Bantu Languages, D. Nurse & G. Philippson (Eds), 97–112. London: Routledge. Kihm, A. 2001. Noun class, gender, and the lexicon-syntax-morphology interfaces: A Comparative Study of Niger-Congo and Romance Languages. Ms, CNRS, Laboratoire de Linguistique Formelle, Univesite de Paris 7. Koopman, H. & Szabolcsi, A. 2000. Verbal Complexes. Cambridge MA: The MIT Press. Josefsson, G. 1998. Minimal Words in a Minimal Syntax. Amsterdam: John Benjamins. Julien, M. 2002. Syntactic Heads and Word Formation: A Study of Verbal Inflection. Ph.D. dissertation, Universitet i Tromsø. Lacarme, J. 2002. Gender polarity theoretical aspect of Somali nominal morphology. In Many Morphologies, P. Boucher & M. Plenat (Eds), 156–178. Sommerville MA: Cascadilla. Lowerman, J. 2006. On n, √ and types of nouns. Ms, Universite de Paris 7. Maho, J.F. 1999. A Comparative Study of Bantu Noun Classes. Gothenburg: Acta Universitaties Gothoburgensis. Marantz, A. 1997. No escape from syntax: Don’t try morphological analysis in the privacy of your own lexicon. U Penn Working Papers in Linguistics 4(2): 201–225. Meinschafter, J. 2004. Final vowel deletion in Italian and the notion of optional rule application. Paper presented at the Workshop on Variation and change in Phonology and Phonetics. University of Potsdam, Germany. Myers, S. 1990. Tone and the Structure of Words in Shona. New York NY: Garland Publishing. Mufwene, S. 1980. Bantu class prefixes: Inflectional or derivational? Papers from the Sixteenth Regional Meeting, 16: 246–258. Chicago IL: CLS. Nespor, M. 1990. Vowel deletion in Italian: The organization of the phonological component. The Linguistic Review 7: 375–398. Nespor, M. 1993. La Fonologia. Bologna: il Mulino. Nespor, M. & Vogel, I. 1986. Prosodic Structures. Dordrecht: Fortis.

 Franca Ferrari-Bridgers Picallo, C. 1991. Nominals and nominalization in Catalan. Probus 3(3): 279–316. Pylkkänen, L. 2002. Introducing Arguments, Ph.D. dissertation, MIT. Ritter, E. 1995. Where’s gender. Linguistic Inquiry 26(4): 795–803. Roeper, T. 1999. Leftward movement in morphology. MIT Working Papers in Linguistics 34: 35–66. Scalise, S. 1992. Compounding in Italian. Rivista di Linguistica 4: 175–198. Schadeberg, T. 2003. Derivation. In The Bantu Languages, D. Nurse & G. Philippson (Eds), 173–192. London: Routledge. Thornton, A.M. 2001. Some reflections on gender and inflectional class assignment in Italian. In Naturally! Linguistic Studies in Honor of Wolfang Ulrich Dressler Presented on the occasion of his 60th Birthday, C. Schaner-Wollens & J. Rennison (Eds), 479–487. Torino: Rosenber & Sellier. Thornton, A.M. 2003a. La rappresentazione dell’informazione morfologica nelle entrate lessicali. In Scritti di morfologia, in onore di Sergio Scalise in occasione del suo 60 compleanno, A. Bisetto, C. Iacobini & A.M. Thornton (Eds) 67–84. Roma: Caissa Italiana Editrice. Thornton, A.M. 2003b. L’assegnazione del genere in Italiano. In Atti del XXIII CILFR, F. SanchezMiret (Ed.), 467–481. Tübingen: Niemeyer. Thornton, A.M. 2003c. L’assegnazione del genere ai prestiti inglesi in Italiano. In Atti del covengno di Venezia 200, A. Sullam Calimani (Ed.). Firenze: Cesati. Vogel, Irene. 1993. Verbs in Italian morphology. Yearbook of Morphology 1993, 219–254. Welmers, W.E. 1973. African Language Structure. Berkeley CA: UCP.

part 3

Information structure

The fine structure of the topic field Mara Frascarelli

Università degli Studi di Roma Tre Based on interface evidence, this paper shows that the interpretation of discourse categories relies on syntactic conditions. Specifically, the existence of a systematic connection between formal properties and discourse functions shows that Topics are licensed in dedicated positions in the C-domain, which correspond to specific realizations at the PF-interface. Different types of Topics thus correspond to specific tonal events and show A’-properties which discard a movement approach. Finally, the interface properties of postverbal subjects are examined, showing that their position does not depend on the argument structure of the verb and is not necessarily different from the position of (referential) preverbal subjects. Though the present analysis is mainly concerned with Italian, a comparison with Bantu data is also proposed in order to show similarities, discuss (apparent) discrepancies and highlight implications for future research.

1. Discourse categories at the interface A large number of recent analyses have investigated the relationship between syntax and phonology (see among others, Selkirk 1984; Inkelas & Zec 1990), having the aim of identifying the principles of prosodic mapping on universal grounds. In this area of research, the analysis of Focus and Topic constituents plays a major role, showing the existence of a significant correlation between discourse grammar and PF-correlates (see Kenesei & Vogel 1990; Hayes & Lahiri 1991; Zubizarreta 1998). In particular, Frascarelli (2000) shows that prosodic phonology maps syntactic structure into Intonational Phrases (I) in which the informative part of the sentence is separated from what is “topical” information, as is shown in Table 1 (Frascarelli 2000: 83). The presence of an I-boundary between the Topic and the rest of the sentence is shown in Italian by phonological rules like Stress Retraction (a repair strategy avoiding a clash between two primary stresses; see Hayes 1989). This rule has no effect on a Topic (even though the following word bears initial stress, cf. (1)), while

 Mara Frascarelli

Table 1. Prosodic constituency in the presence of Focus and Topic constituents a. [YPTopic]I [XP[+F] Verb]I [YPTopic]I Initial Focus marks the left boundary of the sentential I. Non-focused material is topicalized and forms independent Is. b. [YPTopic]I [Verb XP[+F]]I [YPTopic]I In situ Focus marks the right boundary of the sentential I. Non-focused material is topicalized and forms independent Is. c. [YPTopic]I [XP VP ZP]I [YPTopic]I A Topic is never part of the sentential I.

it applies between a verb and the following Focus when a stress clash is at issue (cf. the realization of porterò in (2)):1 (1) [questa città]I [Càrlo non la sopporta]I this town Carlo not it stand-3sg ‘Carlo cannot stand this town.’ (2) [portèro Crlo]I [in quella città]I bring.fut.1sg Carlo in that city ‘I will bring Carlo, to that city.’

Similar conclusions can be drawn for prosodic constituency in Bantu languages. Kanerva (1990) argues that broad Focus sentences2 in Chichewa trigger a global restructuring including all the constituents within a single “Focal Phrase” (an I, in our terms). What is not focal is included in a separate I, as is shown by the phonological rule of preboundary lengthening: (3) a.

[anaményá nyumbá ndí mwáála]I subj.m-pst-hit 9house prep stick ‘He hit the house with a stick.’

b. [anaményá nyumba]I [ndí mwáála]I ‘He hit the house, with a stick.’

(= broad Focus)

(= object Focus)

A broad Focus sentence is argued to be contained within a single prosodic domain in SiSwati as well (see Thwala 2006) and, along the same lines, Costa and Kula . The list of abbreviations used in the glosses is the following: agr = agreement; antiagr = antiagreement; f = feminine; fm = Focus marker; fut = future tense; fv = finite verb; impers = impersonal pronoun; m = masculine; neg = negation; obj = object; perf = perfective; pl = plural; prn = pronoun; pst = past tense; sg = singular; sbj = subjunctive; subj = subject. . Broad Focus sentences carry “all new” information, whereas in narrow Focus constructions new information is restricted to a single constituent.

The fine structure of the topic field 

(this volume) show that a prosodic boundary is always present after a narrow Focus in Bemba, as is signalled in (4)–(6) by tonal variations: (4) [tù-kà-byáálá ínyànyé mwííbala màílò]I 1pl-fut-plant 9maize 16garden tomorrow ‘We will plant maize in the garden tomorrow.’

(= broad Focus)

(5) [tkbyl]I [ínyànyé mwííbala màílò]I ‘We will plant maize in the garden tomorrow.’

(= verb Focus)

(6) [tùkàbyáálá nyny]I [mwííbala màílò]I ‘We will plant maize, in the garden tomorrow.’

(= object Focus)

Given this pattern, in which nominal Foci always appear in postverbal position and the extraposed status of a Topic is invariably signalled by prosodic structure, Kula raises the question of whether syntax might be totally dispensed with in the interpretation of discourse categories. In this perspective, no syntactic position is needed to account for their properties, and relatively free word order can be assumed. Though a “division of labour” between syntax and phonology for interface interpretation may appear extremely advantageous, some evidence shows that this is not a possible solution. Topic and Focus constituents show different formal and scopal properties, which can only be explained if their interpretation is assumed to rely on syntactic conditions before the point of Spell-out. The following sections will propose data to support the claim that the prosodic structure illustrated in Table 1 – whose validity is supported by Bantu data – is not the result of a PF operation but, rather, the phonological output of a syntactic structure in which discourse categories are interpreted in specific functional projections in the C-domain. 2. Discourse categories and syntactic properties The notions of Focus and Topic have an acknowledged status in the literature and have been extensively studied in the last thirty years. Narrow Focus can be defined as the informative part of the sentence, insofar as it can add new information or have a contrastive value. Within the Generative framework of grammar, Focus has been analyzed in terms of a syntactic operator (see Chomsky 1977; Horvath 1995; Rizzi 1997), since it identifies an individual (a “variable”) within a set that is defined by the presupposition. A Topic, on the other hand, has no quantificational properties and does not play an identificational role. Rather, it is the constituent

 Mara Frascarelli

proposed as the subject of predication at a discourse level, also playing a crucial role in the building-up of the sentence (see Givón 1976; Lambrecht 1994; Frascarelli 2000). As is well known, these interpretive differences correspond to distinct formal properties so that, for instance, Topics (contrary to Foci) do not show WCO effects (cf. (7)). Moreover, Topics can be resumed by clitic pronouns (8a) and allow for multiple realizations (9a), while both options are excluded for Foci (cf. (8b) and (9b)):3 (7) a.

Giannii, suai madre loi ha sempre apprezzato. Gianni his mother him have.3sg always appreciated ‘(as for) Giannii, hisi mother always appreciated himi.’

b. *Giannii suai madre ha sempre apprezzato ti Giannii, hisi mother have.3sg always appreciated ti (8) a.

Questa gonna, *(l) ha comprata Maria. this skirt it have.3sg bought.f Maria ‘This skirt, Maria has bought it.’

b. *questa this (9) a.

gonnak skirt

(*l)k ho comprata tk. it have.1sg bought.f

A Leo, questo libro, glielo darò to Leo this book to.him-it give.fut.1sg ‘I will give this book to Leo.’

b. *A Leo darò questo libro to Leo give.fut.1sg this book

These asymmetries can hardly be accounted for in a PF approach to discourse functions since, according to standard assumptions, no phonological operation can have interpretive effects at LF. Moreover, the possibility of clitic resumption shows that a Topic is not merged in argument position, or a violation of the θ-criterion would be at issue.4 Hence, extraposition cannot be considered as a mere phonological operation. Interestingly, similar asymmetries can be found in Bantu languages as well. As is shown in Thwala (2006) for SiSwati (10) and in Costa and Kula (this volume) for Bemba (11), topicalization of an object triggers the presence of an object clitic (incorporated into the verbal head); clitic resumption is, on the other hand, excluded for Foci: . In line with Minimalist tenets, indices in examples simply indicate coreference relations and are not assumed as “objects” of syntactic computation. . Following Uriagereka (1994) and Kayne (1994), I assume that a clitic pronoun enters the computation as the D° head of some sister NP, where NP is an argument pro.

The fine structure of the topic field 

(10) a. topic Tjwala, bafana ba-to-bu-natsa. 14alcohol 2boys 2subj.m-fut-14obj.m-drink ‘The alcohol, the boys will drink it.’ b. focus Bafana ba-to-(*bu)-natsa tjwala. ‘The boys will drink alcohol.’ (11) a. topic tu-álíí-mu-lóndol-a Mutale. subj.m1pl-pst-om-find-fv Mutale ‘We found him, Mutale.’ b. focus tu-álíí-(*mu)-lóndol-á Mutale. ‘We found mutale.’

As far as subjects are concerned, they are usually resumed by a verbal prefix of the same class in Bantu languages, as is shown in Demuth (1990) for Sesotho (12a). However, in a thetic sentence (in which the subject is new information), the verb shows a prefix of the locative class (cf. hó in (12b)) – hence, it takes an expletive subject – and subject resumption is totally excluded (cf. (12c)): (12) a.

ba-shányáná bá-pálám-é li-pére 2-boys 2-ride-perf 10-horse ‘The boys are riding horses.’

b. hó-fihl-ílé li-pére 2-arrive-perf 10-horse ‘There arrived horses.’ c. *li-fihl-ílé li-pére 10-arrive-perf 10-horse

This is evidence that the interpretation of discourse categories relies on syntactic conditions both in Romance and in Bantu languages. In this line of analysis, the following two sections will be concerned with the interface properties of Topics in Italian; the Romance-Bantu connection will then be explored in sections 5 and 6.

3. Discourse categories, tonal events and the cartographic approach In a cartographic approach to discourse functions, the original CP-node (a recursive phrase targeted by different categories) has been reanalyzed as an array of functional projections, each dedicated to a specific function related to Information Structure. In this line of analysis, Focus constituents are interpreted in the so-called Focus Phrase (FocP), a non-iterable functional projection above IP, while Topics are licensed in the Topic Phrase (TopP), which is generally considered as a

 Mara Frascarelli

recursive projection (as is indicated by the “star” in (13); see Rizzi 19975), located on either side of the FocP: (13) [ForceP [TopP Topic* [FocP Focus [TopP Topic* [FinP [IP [VP …]]]]]]]

While a movement analysis (i.e., re-merge, according to the Copy Theory of the Minimalist framework) is generally acknowledged to account for the operator properties of a Focus, the properties of clitic-resumed Topics totally disprove the expectations of a movement approach (see discussion in section 5). Extrasentential Merge is therefore assumed for these types of Topics, in line with Cinque’s (1990) original proposal for CLLD.6 According to a cartographic approach, discourse functions are syntactically defined and their interpretation relies on different formal properties. For the purposes of this paper it is therefore important to consider the interpretation of Topics in more detail, in order to shed light on the relationship between discourse grammar and functional projections in the C-domain. Recent analyses have shown the existence of a systematic correlation between the discourse functions of Topics and their intonational properties, which is encoded in a strict hierarchy in the C-domain (Frascarelli & Hinterhölzl, henceforth F&H, 2007). In other words, not only are Topics not included in the prosodic domain of the sentence (cf. Table 1), they are also made prominent by different tonal events, according to their specific discourse function. In particular, three tonal events can be distinguished, as illustrated below.7 The complex L*+H tone is realized with Topics which establish “what the sentence is about” (see Reinhart 1981). More specifically, the aboutness quality is here inherently associated with a shift in the conversation. Hence, the so-called Aboutness-shift Topic is something that is either newly introduced or newly returned to. The example offered in (14) – in which a student (speaker B) is talking about

. “No interpretative problems arise in the case of a recursion of Top: nothing excludes that a comment may be articulated in turn as a topic-comment structure, so that topic phrases can undergo free recursion” (Rizzi 1997: 297). . Nevertheless, the nature and derivation of dislocated constituents is a much debated issue in the literature and a movement approach is also maintained for clitic-resumed Topics by some authors (see Shlonsky 1992; Rizzi 1997; Cecchetto 1999; Boeckx 2003). . According to the ToBi system (Pierrehumbert 1980), tunes are described as sequences of low (L) and high (H) tones (which determine the shape of the F0 contour). According to this framework, there are six different types of pitch accent: two simple tones – high (H*) and low (L*) – and four complex (bitonal) ones. In this perspective, all pitch accents render prominent the material with which they are associated, regardless of the specific tonal event.

The fine structure of the topic field 

her computer skills – illustrates the relevant discourse function and the tonal event associated with it:8 (14) A: il computer tu lo avevi già usato qualche volta? B: sì mi è capitato però era MacIntosh [. . .] ho sempre avuto un approccio di base non è che abbia mai fatto chissà cosa - però mi è capitato anche di farci un esame comunque insomma ecco un pochino - cioè come si accende [. . .] ecco per esempio il CD Rom invece non l’avevo mai usato A: ‘Did you ever use a computer before?’ B: ‘Yes, it happened, but it was a Mac […] I only had basic skills, I never did anything special with it – I also attended a course however, well, I can do a few things – I can turn it on […] well, for instance, I had never used a CD Rom.’ (14') il CD Romk invece non lk’ avevo mai usato. the CD Rom instead not it have.pst.1sg never used ‘I had never used a CD Rom.’

m il C D

Ro

in ve

ce

non l'a ve

vo

ma

i

u sa

to

Figure 1. Aboutness-shift Topic (L*+H).

As we can see from (14), the Topic proposed by the interviewer (speaker A) is il computer (located in the left-periphery and resumed by the clitic lo) and the student replies by talking about her computer skills. However, after a while, she wants to shift on a specific argument, namely the use of a CD Rom. This new Topic is signalled by a sharp rise in the F0 which is aligned with the tonic vowel in its full extension and reaches its peak on the syllabic coda (cf. Figure 1). The rest of the sentence (the so-called “Comment”) does not present any significant tonal event and shows a final falling contour (as is the general case for broad Focus sentences; see D’Imperio 2002; Frascarelli 2004a). . The analysis of Topics in Italian is based on a corpus of natural data (Bonvino 2006). To make discourse analysis clear, I provide part of the context preceding the sentence at issue. Then, the relevant sentence (underlined in the text) is repeated and examined in detail.

 Mara Frascarelli

An L* tone, on the other hand, is realized with Topics that refer to given information in the discourse. The relevant Familiar Topic is therefore a discourse-linked constituent (Pesetsky 1987) which is typically destressed (see Lambrecht 1994) and used for topic continuity or as an “afterthought”. Consider the following conversation: (15) A: Come impari le parole? le studi sui libri? e come le impari quando le studi davanti al computer? c’è qualcosa di diverso? B: allora in genere nei libri aspetto, cioè leggo tutto e aspetto di capirla – magari nel computer se so che c’è il vocabolario sono portata a cercarla subito A: il fatto che ci sono delle immagini non ti aiuta a capire la parola maggiormente che in un testo? B: sì quello sicuramente cioè basta che mi trattengo un po’ e probabilmente nemmeno ho bisogno di cercarla la parola. A: ‘How do you learn words? Do you study them in books? And how do you learn them when you study on computers? Is there anything different?’ B: ‘Well, in general, when I study in books I wait, that is to say, I read throughout the text and try to understand the meaning – with on-line dictionaries, I’d rather check it immediately.’ A: ‘Do images not help you to understand the meaning of a word faster than reading a text?’ B: ‘Yes, of course – in fact, if I refrain from immediate checking, probably, I do not even need to search the word.’

As is made clear by the context, the direct object la parola represents an element of continuity in the discourse. It is introduced by the interviewer, resumed by several (clitic) pronouns and finally repeated by the student, as a right-hand Topic: (15') nemmeno ho bisogno di cercarlak la parolak. not even need.1sg of to search.it the word ‘I do not even need to search the word.’

ne

mme no

so ho bi

gno

di

cer

car

la

la

pa ro la

Figure 2. (Right-hand) Familiar Topic (L*).

The fine structure of the topic field 

In this case, the tonic vowel of the right-hand Topic la paròla is realized with a low tone which, crucially, must not be considered as a physiological consequence of its final location. Indeed, Familiar Topics are low also when they are located in the left periphery of the sentence. Consider the following example, in which questa attività – a Familiar object-Topic resumed by the clitic la – is realized with a low tone in preverbal position (cf. Figure 3): (16) B: si doveva fare quella attività di riconoscere delle parole – delle lettere – che avevano un dato suono – io lì ho fatto un macello infatti l’ho fatto più di una volta c’era la correzione e sì anche la correzione non l’ho trovata tanto chiara. A: va bè adesso io francamente questa attività in particolare non me la ricordo. B: ‘We were supposed to do an exercise to recognize words – letters – which were associated to some specific sound – I did a mess there, in fact I did it more than once – there was also a key for self-correction but it was not very clear as well.’ A: ‘Well, now, frankly, I cannot remember that exercise in particular.’ (16') questa attivitàk in particolare non me this exercise in particular not to.me ‘I cannot remember that exercise in particular.’

questa

a

tti vi tà

in par

ti co la re

non me la

lak ricordo it remember.1sg

ri

cor

do

Figure 3. (Left-hand) Familiar Topic (L*).

Finally, an H* tone is associated with Topics which induce alternatives in the discourse, which have no impact on the Focus value of the sentence and which create oppositional pairs with respect to other Topics. Following Kuno (1976) and Büring (1999), F&H (2007) define this type of Topic as Contrastive. As an illustration, consider the following text, in which speaker B proposes an opposition between two Topics (francese and inglese, respectively). Each of them is marked by

 Mara Frascarelli

a high pitch and followed by a broad Focus sentence (the Comment) expressing the informative part of the relevant contrast: (17) A: come mai hai fatto due lingue, cioè, inglese e francese? B: francese l’ho fatto alle medie per tre anni con una professoressa con cui mi sono trovata benissimo [. . .] – con l’inglese mi son trovata sempre a disagio A: ‘Why did you study two languages, namely English and French?’ B: ‘I have studied French at school for three years with a professor that I liked a lot [. . .] (while) with English I never felt at ease.’ (17') francesek lk’ho fatto alle medie per French it have.1sg done at.the school for tre anni [. . .] con l’inglese mi son three years with the English me be.1sg trovata sempre a disagio found.f always uneasy ‘I have studied French at school for three years […] with English I never felt at ease.’

c fran

e s e l ho

gle l'in fa a me per tre a tt lle die

nni

con

se

ta mi sontrova

sem pre

a

disa gi

Figures 4a–b. Contrastive Topics (H*).

As we can see, differently from the Aboutness-shift Topic (Figure 1), the rising contour of Contrastive Topics is aligned with the pre-tonic syllable, while the tonic vowel marks the highest part of the relevant tonal event. It is interesting to note that an H* tone also marks Contrastive Foci (see Frascarelli 2004a). The interpretation of contrastive elements, however, is never ambiguous: Contrastive Topics are clitic-resumed, they do not identify an individual with the exclusion of other options and have no impact on the Focus value of the sentence. The following examples illustrate: (18) a.

contrastive topic Il libro, l’ ho dato a Mario e il giornale, a Leo the book it have.1sg given to Mario and the newspaper to Leo

‘The book (ContrTop1), I gave it to Mario (Comment) and the newspaper (ContrTop2) to Leo (elliptical Comment).’

The fine structure of the topic field 

b. contrastive focus

Il Libro ho dato a Mario (… non il giornale) the book have.1sg given to Mario (… not the newspaper) ‘It is the book (ContrFoc) that I gave Mario (Presupposition).’

These observations suggest that the interpretation of Contrast – like the other discourse categories – relies on a number of formal properties, which are present at the moment of Spell-out. The notion of Contrast is therefore not an inherent property of either Focus or Topic constituents, but a functional feature which is interpreted in a dedicated position in the C-domain, where either a Topic can be realized or a Focus can move. This predicts that only one constituent can be contrasted in a single clause (as is borne out in F&H’s 2007 investigation).

4. The fine structure of the topic field The analysis of spoken corpora shows that, in multiple realizations, the three different types of Topics illustrated above are located in the C-domain according to a specific order. This means that not any kind of Topic can be found in any position. Specifically, the Aboutness-shift Topic is located in the highest Topic projection in the C-domain (higher than Focus and Contrastive elements)9 and cannot be iterated, while Familiar Topics occupy the lowest TopP projection (right above the FinP) and can be recursive.10,11 Finally, Contrastive Topics are located in an intermediate position between Shift and Familiarity. F&H (2007) therefore refute . It is important to remember that the hierarchy proposed in (19) is only concerned with clitic-resumed Topics. Hence, the Aboutness-shift Topic is not – and should not be confused with – a Hanging Topic, whose formal and discourse properties are totally different (for a detailed discussion, see Frascarelli 2007). The types of Topics considered are all located below ForceP because of distributional evidence: they follow complementizers like che (“that”) which, according to Rizzi (1997), are merged in Force°. . FamP seems to be the only Topic projection that allows for multiple realizations in the same clause. This is consistent with the fact that different elements can be part of background information (and used for topic continuity), while only one Topic can be used with a shifting function in a single clause or be contrasted with other Topics (in different clauses). . Benincà & Poletto (2004) argue against a low Topic projection (below the Focus Phrase) and suggest that such a position corresponds to a secondary Focus. The analysis of data, however, offers clear evidence that Familiar Topics are located lower than Contrastive Foci and whconstituents. Consider, for instance, the following sentence (taken from the corpus), in which the presence of clitic resumption and the low tone exclude a Focus analysis for questa ricetta:

 Mara Frascarelli

a “free recursion analysis” for Topics and propose a refinement of the hierarchy given in (13), in which dedicated functional projections are visible for specific interpretations at the interface levels: (19) [ForceP

[ShiftP [+aboutness] L*+H

[ContrP [FocP H*

[FamP* L*

[FinP

[IP ……]]]]]]]

Hence, if different constituents are realized in the left periphery of the sentence, they are expected to be produced in the order given in (19) and marked with the relevant tonal events.12 This prediction is borne out by data. Consider, for instance, the context given in (20) below. As is clear, inglese is part of background information, while the pronoun io is used to propose a shift. As expected, the latter is a rising (L*+H) Topic while the former is low-toned: (20) la lingua inglese attraverso i programmi sul computer diciamo non l’avevo mai fatta [. . .] comunque l’inglese risultava anche facendolo da solo più interessante [. . .] io inglese non -premetto- non l’avevo mai fatto. ‘I had never studied English through computer programs […] and through self-learning English appeared more interesting to me […] I admit that I had never studied English before.’

(i)

A chi questa ricettak (L*) lak dovrei consegnare? To whom this receipt it should.1sg to give ‘To whom should I give this receipt?’

Moreover, the existence of a low Topic position can account for the syntax and interface properties of right-dislocated constituents (see discussion below). Of course, this does not exclude the possibility that secondary Foci can also be located in the low C-domain (though no occurrence could be found in the corpus examined). For further data and details, see F&H (2007) and Frascarelli (2007). . Interestingly, Abels & Miriungi (2005) show that fronting of an object in Kitharaka is always associated with Contrast and, if a subject precedes, the latter is necessarily treated as a Topic:

(i) Maria i-nyomba a-ra-k-ir-e Maria fm-9house 1subj.m-prn-build-perf-fv ‘As for Maria, she bought a house (… not a flat).’ This interpretation cannot be explained if we assume that discourse categories are interpreted by morpho-phonological features, without the intervention of syntax. On the other hand, it is completely expected if we assume the hierarchy given in (19). For a comprehensive comparison, future research should be dedicated to the intonational properties of Topic constituents in Bantu languages.

The fine structure of the topic field 

(20’) Io inglesek non – premetto – non lk’ avevo I English not admit.1sg not it have.pst.1sg mai fatto. never done ‘I admit that I had never studied English before.’

i o

i

n gle

se

pre me non

tto

non l'ha ve vo

mai

fa

tto

Figure 5. Aboutness-shift and Familiar Topic in the left periphery.

As is clear from (19) above, the existence of a “right periphery” is excluded. According to the present theory, right-hand Topics are Familiar Topics and their final position is derived through IP-inversion to a functional projection in the C-domain which is higher than FamP (see Frascarelli 2000, 2004b). Following Poletto & Pollock’s (2004) suggestion, this position can be identified in the Ground Phrase (GP) projection (see F&H 2007 for details and Cardinaletti 2001, 2002 for a similar analysis): (21)

[ForceP [ShiftP [+aboutness] [ContrP [GP [FocP [FamP  [IP …………] ]]]]]]

Apart from theoretical considerations discarding rightward operations (see Kayne 1994), this analysis can account for the formal and discourse properties of righthand Topics. From an intonational point of view, it provides an explanation for the fact that Familiar Topics are low-toned in both peripheries. Syntactically, it accounts for their scopal properties: as we can see in (22b) a right-hand Topic is not in the scope of the quantified indirect object, thus excluding a distributive reading. (22) a.

Darò ad ogni autore il suo libro (distr./indiv.) give.fut.1sg to every author the his book ‘I will give every author his book.’

 Mara Frascarelli

b. Lo darò ad ogni autore, il suo libro (indiv. only) it give.fut.1sg to every author the his book ‘I will give it to every author, his book.’

Finally, it accounts for the fact that right-hand Topics�� cannot be used to introduce a referent in the discourse. Indeed, the inverted IP obtains scope over the Familiar Topic, yielding the prediction that Contrastive or Shifting Topics cannot be located in the right periphery of the sentence, as is exactly the case. Compare (23) with the unacceptable (24): (23) Io ho due fratelli: Luigij loj vedo spesso, I have.1sg two brothers Luigi him see.1sg often mentre Mariok non lok vedo mai. while Mario not him see.1sg never ‘I have two brothers: Luigi, who I see often, and Mario, who I never see.’ (24) *Io ho due fratelli: loj vedo spesso Luigij, mentre non lok vedo mai, Mariok.

Once more, these asymmetries can be hardly explained in a non-configurational account of discourse categories. Finally, it is worth mentioning an aspect concerning the left periphery of embedded clauses. It has long been acknowledged that peripheral Topics are only marginally acceptable in embedded clauses (see Kuno 1976; Lambrecht 1994; De Cat 2002). The study of corpora, however, shows that this restriction only applies on a specific type of Topic, namely, on rising Topics. Hence, Shifting and Contrastive Topics are always located in matrix C-domains, while Familiar (low-toned) Topics can appear in any clausal type (matrix, subordinate, adverbial and relative clauses), in both peripheries. The present cartographic approach can provide an explanation for this distinction. As is discussed in recent works, there is a connection between new information and the illocutionary Force of matrix clauses (see Haegeman 2002; Giusti 2006). It is therefore plausible to conclude that ShiftP (in which a new Topic is proposed) and ContrP (in which Contrast is encoded) are not available in embedded C-domains, in which only Familiar information can be expressed. The implications and the comparative issues of this proposal will be the subject of future research. In the light of the evidence discussed so far, the next sections will be dedicated to the analysis of subjects in SV and VS constructions, showing that subjects occupy different positions, associated with specific formal properties and discourse functions. Interface evidence will also show that preverbal and postverbal subjects do not necessarily occupy different positions.

The fine structure of the topic field 

5. Topics, preverbal subjects and minimality effects As is extensively argued in Frascarelli (2000, 2004b, 2007) scopal properties, antireconstruction and minimality effects show that clitic-resumed Topics are merged in the C-domain according to a Locality requirement. Evidence is provided by contrasts like the following (for further examples and discussion, see cited references): (25) Leok non sa [CP chii ha visto una foto Leo not know.3sg who have.3sg seen a picture di se stessoi/*k] of himself ‘Leo does not know who saw a picture of himself.’ (26) La propria*i/k foto, Leok non sa chii l’ the his own picture Leo not know.3sg who it ha vista have.3sg seen.f ‘The picture of himself, Leo does not know who saw it.’

As we can see, while in (25) the anaphora se stesso can only be interpreted as coreferent with the wh-constituent chi (consistent with the fact that it is located within the same local domain), the anaphora interpretation in (26) shows that the Topic (la propria foto) is interpreted in a position that is c-commanded by the DP Leo, but not by chi. This is to be expected in a theory – such as the one I assume – in which Topics do not move from an IP-internal position, but are merged in the local C-domain selecting the IP in which the coreferent pro(noun) is located:13 (27) [la propria*i/k foto]j Leok non sa [〈la propria*i/k foto〉j [chii [lj ’ha vista [DP tcl [NP proj]]]]]

In other words, Topics can move from TopP to TopP14 (as in (27)) and be reconstructed for interpretation in their local C-domain, but no copy can be found in the VP, in which clitics are merged. This analysis also explains the absence of minimality effects between Topics and syntactic operators (like Foci or wh-constituents): (28) [TopP questo libroi [FocP a chi [IP pro loi darai 〈a chi〉 ]]]? this book to whom it give.fut.2sg ‘To whom will you give this book?’

. The relevant Locality requirement can be understood as a condition on the identification of pro (for details, see Frascarelli 2004b, 2007). . This is consistent with the fact that languages have a dislocation property, connected with scopal and discourse requirements (see Chomsky 2001).

 Mara Frascarelli

In a Bantu-Romance connection perspective, it is interesting to notice the absence of minimality effects in contexts similar to (28) in languages like Kinande. Consider the following example (from Schneider-Zioga 2000): (29) a.

Iyɔndi yɔ Yosefu akaβʊla [εkihi nga-ky’ who that Yosefu wonder.3sg what if-obj.agr

a-kalangira]? subj.agr-see.3sg

‘Who does Yosefu wonder if he sees what?’ [lit.: ‘Who does Yosefu wonder what if (he) sees it?’]

As we can see, the wh-subject-Topic Iyɔndi and the wh-object εkihi do not interfere with each other. This follows from the fact that preverbal subjects originate in a TopP position. Hence – as is argued by the author – what appears to be a case of crossing actually involves two independent instances of extraction: (29) b. Iyɔndik yɔ Yosefu akaβʊla [TopP tk [FocP εkihij nga-ky’ [IP a- kalangira tj]]]?

Minimality effects, however, arise when the wh-object is subject to long movement, as in the following case (adapted from Schneider-Zioga’s (20a)): (30) a. *εkihi kyɔ Yosefu akaβʊla [Iyɔndi nga-y’ what that Yosefu wonder.3sg who if-subj.agr

u-kalangira]? antiagr-see.3sg

‘*What does Yosefu wonder who if sees?’

Notice that in this case the verb kalangira does not show the canonical subject agreement (a–), but a verbal prefix of “antiagreement” (u–) which appears when the subject operator moves from an IP-internal position (antiagreement effects are also typical of Cushitic languages in similar contexts, see Frascarelli and Puglielli 2007). This means that the wh-subject in (30a) is not a Topic and, moving from the subordinate Spec,VP, it determines minimality effects with the wh-object (since they are both operators): (30) b. *εkihij kyɔ Yosefu akaβʊla [FocP Iyɔndik nga–ky’ [IP tk u– kalangira tj]]?

We can therefore conclude that sentences like (30a) do not invalidate the claim that in Bantu languages (as in Italian) clitic-resumed Topics are different types of A’-dependencies with respect to wh-constituents (i.e., only the latter are operators); rather it shows that two positions must be posited for the generation of

The fine structure of the topic field 

subjects: one in the C-domain (as some kind of Topic) and one in an A-position (in the VP).15 This assumption is not a shortcoming of the theory of grammar. On the contrary, since functional projections in the C-domain encode discourse-semantic features which correspond to specific formal properties, such an option is expected to obtain distinct interpretations. In this line of analysis, we can predict that the subject-wh in (29a) does not imply a genuine wh-question but, rather, represents a Contrastive Topic, while the subject-wh in (30a) is a genuine informationseeking indefinite. In other words, we expect that an appropriate answer to (29a) requires (at least) a pair of sentences in which two oppositional subject-Topics (corresponding to iyɔndi) are opposed with respect to some new information (corresponding to the wh-object). An indirect confirmation of this hypothesis comes from Italian data: a sentence similar to (29a) is present in the corpus and the “subjects” in the relevant answer are intonationally marked as Contrastive Topics.16 This hypothesis is also supported by the fact that indefinite subjects (like genuine wh-subjects) always induce obligatory antiagreement in Kinande (cf. (31a)) and that quantified subjectTopics are only allowed if they can receive a specific interpretation (cf. (31b)), (both adapted from Schneider-Zioga 2000): (31) a.

si-ha-li mundo oyo *a-/u-kayenda. neg-there-be person that *subj.agr-/antiagr-leave.pst ‘Nobody left.’

b. oβoli mwana a-lagira Marya. every child subj.agr-see.pst Marya ‘Every child (of a given set) saw Marya.’

We can therefore conclude that in Bantu languages referential/specific preverbal subjects are generated in the C-domain as Topics, while indefinite/non specific preverbal subjects are merged in an A-position and move to Spec,IP as grammatical subjects. This analysis represents an important issue for the Bantu-Romance . Interestingly, this is exactly the analysis proposed in Bresnan & Mchombo (1987) for Chichewa; in Demuth & Harford (1999) for comparative Bantu; and in Frascarelli (2007) for Italian (see discussion below). .

The relevant case is given below:

(i) A: Un ottimo lavoro, ma vorrei sapere chi ha fatto cosa. ‘Quite a good piece of work, however I’d like to know who did what.’ B: Io ho curato la comparazione dei dati e Marco la parte teorica. ‘I dealt with the comparison of data and Marco wrote the theoretical part.’

 Mara Frascarelli

connection since this is also the conclusion drawn in Frascarelli (2007) for Italian preverbal subjects: while referential DPs are merged as Topics and identify a pro in argument position, non-specific subjects are IP-internal constituents.17 The A/A’ status of subjects is shown, among other things, by intonation, scope and binding properties (see cited reference for details and discussion). 6. The interface interpretation of postverbal subjects Postverbal subjects in a “free inversion language” like Italian (see a.m.o., Rizzi 1982; Jaeggli & Safir 1986) have been generally considered as located in a position structurally different from that of preverbal subjects. In particular, after Burzio’s (1986) influential work, the VS order has been considered as the unmarked option for unaccusative verbs and a split has been assumed with respect to unergative verbs. According to this proposal, the postverbal subject of a verb like arrivare (“to arrive”) is located within the VP (the position of “external merge” in Minimalist terms, cf. (32b)), while the postverbal subject of a verb like telefonare (“to call”) is right-adjoined after movement (cf. (33b)): (32) a.

E’ arrivato Leo. ‘Leo has arrived.’

b. [VP [e] V NP] (33) a.

Ha telefonato Leo. ‘Leo has called.’

b. [VP ti V] NPi

Apart from independent considerations (see Kayne 1994), a right-adjunction analysis and, in general, a distinction between unaccusative and unergative verbs to account for the postverbal position of subjects can be shown to be totally inadequate in interpretive terms. Interface analysis shows that (unfocused) postverbal subjects are invariably associated with two types of intonational contours, which are independent of the argument structure of the verb, as well as of the TAM features associated with it.18 The semantic features of the subject itself (e.g., definiteness or quantification) are also immaterial in

. In particular, Frascarelli (2007) assumes a non-pronominal account for the Agr head and proposes that a Topic-antecedent is the only relation responsible for the identification (i.e., copying of ϕ-features through Agree) of a referential pro in argument position, discarding all instances of “accidental” coreference. . The analysis provided in this section (and summarized in Table 2) is based on the study of 134 third person (unfocused) subjects, taken from 340 clauses.

The fine structure of the topic field 

this respect. Interface interpretation, on the other hand, is consistently connected with the discourse functions of subjects. Consider the following table: Table 2. The tonal realisation of subjects

Tonal event

SV

VS

a. b. c.

H+L* L* L*+H

0 67 39

7 21 0

As we can see, postverbal subjects can either mark the final part of a downgrading contour (H+L*) or be realized with an L* tone. Note that the low tone is available both for preverbal and postverbal subjects, while the latter cannot be associated with a rising tone (as expected, cf. section 4) and preverbal subjects are never included in a downgrading contour.19 When the tonic vowel of a postverbal subject marks the falling part of a downgrading contour, it is included in a broad Focus sentence (i.e., it is part of the Comment). Neither breaks nor F0 resetting can be found between the verb and the subject (which belongs to the sentential I). Consider (34), in which the postverbal subject l’idea is clearly part of new information: (34) lei ha incominciato a far pratica a scrivere le notizie/questa è andata come me cioè senza in realtà sapere cosa volevamo – cioè ci attirava l’idea. ‘She started to practice, to write the news, she went there like me, that is, without really knowing what we wanted – that is to say, the idea was appealing to us.’ (34') cioè ci attirava l’ idea. that is us attract.pst.3sg the idea ‘That is to say, the idea was appealing to us.’

cioè

cia

tti

ra

va l'i de

a

Figure 6. Postverbal subject in a broad Focus sentence (I).

. For an interface analysis of preverbal subjects and further details on Table 2, see Frascarelli (2007), Frascarelli & Trecci (forthcoming).

 Mara Frascarelli

The same pattern can be found when the subject is a fully referential DP, with an agent θ-role in a transitive construction, as is shown in Figure 7 below: (35) A: non ti sembra esagerata quella pila di CD? B: me l’ha regalata Riccardo A: ‘Don’t you think that pile of CDs is an exaggeration?’ B: ‘Riccardo gave it to me as a present.’ (35’) me l’ha regalata Riccardo. to.me it have.3sg given.f as a present Riccardo ‘Riccardo gave it to me as a present.’

me

l'ha

re

ga

la

ta

Ri

ccar do

Figure 7. Postverbal subject in a broad Focus sentence (II).

On the other hand, when postverbal subjects are realized with a low tone, they are produced after the speaker’s baseline level (i.e., the minimum value of F0) has been reached and they are normally preceded by a rhythmic break. Consider (36) below: (36) c’è un programma proprio per il vocabolario e vanno a temi però non mi ricordo ora e io ci ho dato un’occhiata perché c’erano dei titoli carini insomma c’erano tutta una serie di sentimenti per esempio e io non ho mai saputo come si dicevano ‘sti sentimenti in inglese.

‘There is a program exactly to learn words, they are arranged by subjects, however I can’t remember now and I gave it a look because there were some nice titles, that is to say, a number of feelings were listed, for instance, and I never learnt how to say these feelings in English.’ (36') non ho mai saputo come si dicevano not have.1sg never known how impers tell.pst.3pl ‘sti sentimenti in inglese. these feelings in English ‘I never learnt how to say these feelings in English.’

The fine structure of the topic field 

e io non mai sa pu to come sidi ce vanc ho

sti

sentimen

ti in in gle se

Figure 8. Low-toned postverbal “subject”

As is clear from the context and the tonal properties of the relevant constituent, ‘sti sentimenti qualifies as a Familiar right-hand Topic: like la parola in (15), it is an element repeated as an afterthought and its intonational curve is totally flat. The following constituent (in inglese) is also a Familiar Topic, characterized by the same formal properties (remember that FamP is a recursive projection; cf. note 10). In the corpus examined, postverbal subjects are consistently associated with these two intonational patterns which, as we have seen, correspond to specific discourse functions. Crucially, the analysis of data shows that the relevant PFrealizations are also associated with specific syntactic properties. Low-toned postverbal “subjects”20 can be coreferent with a c-commanding DP. Consider the following example, in which the speaker is talking about getting lost on the road to the Roman Ring: (37) mi aveva detto alla fine troverai la strada, il Raccordo – finisce sul Raccordo – e invece non finiva sul raccordo, cioè quella portava a Fiumicino quella strada lì. ‘He told me at the end you will find the road, the Ring – it gets to the Ring – but it did not lead to the Ring, that is to say, that road led to Fiumicino.’ (37') quellak portava a Fiumicino [quella strada lì]k. that one lead.pst.3sg to Fiumicino that road there ‘That road led to Fiumicino.’ [lit.: That one led to Fiumicino, that road] . For simplicity, I will continue to define them as “subjects”, using quotation marks to indicate a non-conventional use of the relevant term (i.e., meaning that they are Topics connected with the subject position).

 Mara Frascarelli

As is clear from the context, the speaker is justifying her mistake: a friend gave her instructions to get to the Ring, but the road she took was not the right one. That road (as opposed to the one she should have taken) led to Fiumicino:

que lla por

ta

va a

Fiu

mi

ci no

quella

stra da

lì

Figure 9. Coreference between a Topic and a postverbal “subject”.

The tonal events illustrated in Figure 9 show the discourse functions associated with the relevant constituents. The demonstrative pronoun quella forms a prosodic unit in initial position and is realized with an H* tone, hence it qualifies as a Contrastive Topic. On the other hand, the postverbal DP quella strada lì is low-toned and produced after the baseline level has been reached. It is therefore a right-hand Topic, merged in an A’-position (i.e., FamP).21 This accounts for the binding possibility in (37). This possibility, on the other hand, is excluded for postverbal subjects which are part of a downgrading (broad Focus) contour, showing that they are located in a different position: (38) a. *Quellak , ci attirava l’ideak that one us attract.pst.3sg the idea

(cf. (34))

b. *luik , me l’ha regalata Riccardok he to.me it have.3sg given.f Riccardo

(cf. (35))

In this respect, it is also interesting to note that referential “subjects” in postverbal and preverbal position show the same binding possibilities. As is shown in (39) below, the subject-contained R-expression Leo can be coreferent with the object-clitic lo. The relevant coreference is also allowed for a low-toned “subject”

. Fiumicìno, on the other hand, is the narrow Focus of the sentence and, as such, it is marked by a pitch. On focused subjects, see Frascarelli (2004a).

The fine structure of the topic field 

(cf. (40a)), while it is excluded for a postverbal subject that is included in a downgrading contour (40b): (39) [il fratello di Leok] lok ha visto alla stazione. the brother of Leo him have.3sg seen at-the station ‘Leok’s brother saw himk at the station.’ (40) a. lok ha visto alla stazione, [il fratello di Leok] (= L*) b. *lok ha visto alla stazione [il fratello di Leok] (= H+L*)

This is evidence that the DP Leo in (40b) is in the c-commanding domain of the object clitic lo, while this is not the case for Leo in (40a). This shows, on the one hand, that both preverbal and postverbal “subjects” have A’-properties and, on the other, that postverbal subjects included in a downgrading contour are located in a low A-position. Since the Spec,IP position must be filled for independent (EPP) requirements (see Chomsky 1981, 2001), we can account for these positions as follows:

a. postverbal subjects included in a downgrading contour are located in a low A-position (presumably an edge position in the vP-phase, see Chomsky 2005) and connected with an expletive subject (pro) which satisfies the Subject Criterion (see Rizzi 2006). The relevant final position is derived through VP remnant movement.22

b. postverbal low-toned “subjects” are right-hand Familiar Topics, located in an A’-position. In this case a thematic pro satisfies the Subject Criterion and is connected with the relevant Topic. The postverbal position is derived through IP remnant-movement (cf. (21) above).23

. VP remnant movement accounts for the post-participial position of subjects with an agent θ-role, which are externally merged in Spec,vP (cf. (35): me l’ha regalata Riccardo). . Bantu languages like Chichewa show a perfect parallel with Italian data, supporting the present analysis. Consider the following sentences (adapted from Bresnan & Kanerva 1989):

(i) Maseru basadi ba-ile kajeno Maseru 2women 2agr-go.perf today ‘The women, they went to Maseru today.’ (ii) Maseru ba-ile kajeno, basadi ‘They went to Maseru today, the women.’ (iii) Maseru ho-ile basadi kajeno Maseru 17agr-go.perf 2women today ‘To Maseru there went the women today.’

(iv) *Maseru ho-ile kajeno basadi

 Mara Frascarelli

Additional evidence in favour of this analysis comes from floating quantifiers in Italian. As is known, they are traditionally assumed to be possible only after subject movement to Spec,IP. It is therefore crucial to observe that quantifiers can also float with a right-dislocated (low-toned) “subject”. Let us consider the VOS sentences in (41a–c):24 (41) Dovevano partire con quel treno … ‘They should have left with that train …. a.

… ma hanno perso il biglietto, [tutti i miei amici]L* but have.3pl lost the ticket all the my friends … but all my friends have lost the ticket.’

b. … ma hanno tutti perso il biglietto, i miei amici c.

… ma hanno perso tutti il biglietto, i miei amici

On the contrary, floating quantifiers are not allowed when the postverbal subject is part of a broad Focus sentence. Let us consider sentences (42a–c): (42) Dovevamo partire con quel treno… ‘We should have left with that train…. a. …ma [hanno perso il biglietto tutti i miei amici]H+L* but have.3pl lost the ticket all the my friends … but all my friends have lost the ticket.’ b. *… ma hanno tutti perso il biglietto i miei amici c.

*… ma hanno perso tutti il biglietto i miei amici

In the light of the present proposal, the relevant data suggests a refinement for the traditional account of floating quantifiers: the crucial requirement does not seem to rely on subject movement to Spec,IP, but rather on the presence of a nonexpletive subject in Spec,IP. As we can see, preverbal and postverbal subject-Topics must be resumed by a verbal prefix (i)–(ii). On the other hand, in the case of locative inversion the locative DP is interpreted as the Topic of the sentence, the verb shows an expletive type of agreement and the subject remains in a low position in the vP-phase (iii). The vP-internal position is confirmed by the fact that the inverted subject cannot follow an adverbial expression like “today” (cf. (iv)). . The analysis of natural data shows that the informational distribution given in (41a) for the VOS order is the most common in Italian (namely, with a topicalized subject and a focused object or VP). On the other hand, the VOS order is never associated with a focused subject (when the object is “given”, it is always realized as a clitic) or with a contrastive focalization of the verb (options that are claimed as possible in Cardinaletti 2001, 2002). Finally, a broad Focus reading could be also found for the VOS order, with destressing of the object (as is illustrated in (42)).

The fine structure of the topic field 

VS constructions are also extensively present in Bantu languages, either to realize thetic sentences or signal the extraposed status of the inverted subject as “given” information. Next to syntactic evidence (see Sesotho in (12a–c) above), the application of phonological rules provides substantial support to the analysis offered in this section for Italian, strengthening the Bantu-Romance informationstructural connection. For reasons of space, the illustration of the relevant data will be limited to Chichewa, in which the rule of tone retraction (a tonal change concerning the penultimate stressed syllable within a prosodic domain) shows that right-dislocated subjects are preceded by a boundary. Consider the following sentences (from Bresnan & Mchombo 1987): (43) Ndikufúná kutí [áná ánga a-pitiriz-é phúnziro] want.1sg that children my subj.m-continue-sbj lesson ‘I want my children to continue the lesson.’ (44) Ndikufúná kutí [áná ánga a-li-pitirĭz-e] want.1sg that children my subj.m-obj.m-continue-sbj phúnziro lesson ‘I want my children to continue it, the lesson.’

As we can see in (43), the subjunctive suffix –é is marked by a high tone when it is followed by an object in argument position. On the other hand, when it is followed by an object-Topic (resumed by li) tonal retraction applies (cf. (44)). This means that Topics do not belong to the Intonational Phrase containing the verb (as in Italian; cf. Table 1). It is therefore significant to notice that tonal retraction on the verb also applies in VS constructions, showing that subjects form an independent prosodic domain: (45) Ndikufúná kutí [a- pitirĭz-e] áná ánga want.1sg that subj.m-continue-sbj children my ‘I want my children to continue.’ [lit.: I want that (they) continue, my children]

This analysis provides a unified account for preverbal and postverbal (referential) “subjects” as extrasentential constituents which are merged in a Topic position in the left periphery of the sentence (in line with recent works; see, among others, Longobardi 2002; Frascarelli 2007) and provides interface evidence for their A’-properties. Indeed, not only do preverbal and postverbal “subjects” share syntactic properties, they are also realized with the same type of intonational

 Mara Frascarelli

contour. Compare ‘sti sentimenti (Figure 8) with loro and i gladiatori in Figure 10 below: (46) a.

C’era una grande pedana coperta di sabbia e su questa, lorok, i gladiatorik lottavano.

‘There was a big stage covered with sand and on this, they, the gladiators, would fight.’

e su

que

s

ta lo ro

i

gla dia to

ri

lo

tta va no

Figure 10. Preverbal low-toned “subjects”.

As we can see, after a locative PP-Topic (su questa), two coreferential low-toned “subjects” follow, first as a pronoun and then as a full DP. This coreference would be necessarily excluded for constituents sitting in an A-position, while it is perfectly consistent with a Topic analysis. Specifically, the relevant constituents are merged in the FamP position which, as we know, can be iterated. As a matter of fact, if one of the two “subjects” in (43) is realized in an A-position (hence, pronounced within a downgrading contour) the relevant coreference is totally excluded: (46) b. *C’era una grande pedana coperta di sabbia e su questa, [lorok]L*, [lottavano i gladiatorik]H+L*

Since clitic resumption is a major feature of the type of Topics analyzed in this paper, the final section will deal with a short discussion on the so-called “clitic-optionality” for right-hand Topics in Italian.

7. Clitic-resumed Topics vs. “marginalized” objects While clitic-resumption is obligatory for left-peripheral DO Topics in Italian (cf. (47a)), it seems to be optional for right-hand Topics (cf. (47b)):

(47) a.

The fine structure of the topic field 

Il dolce, *(lo) faccio io. the cake it make.1sg I ‘I will make the cake.’

b. (Lo) faccio io, il dolce.

Given the present analysis, in which Topics are merged in the C-domain and connected with a clitic/pro, while right-hand topicalization is derived through IPinversion, optionality cannot be maintained for clitics. The lack of clitic resumption must therefore be considered as the signal that a different construction is at issue in the relevant case. This has been extensively discussed in recent works, in which non clitic-resumed dislocated constituents have been analyzed in terms of marginalized objects (see Antinucci & Cinque 1977; Frascarelli 2000; Cardinaletti 2001, 2002).25 In particular, Frascarelli (2000) proposes that non clitic-resumed dislocated objects – contrary to Topics proper – are generated within IP and scramble to an A’position above the VP (which is then subject to remnant movement). This analysis predicts that marginalized objects are reconstructed within the VP for interpretation, contrary to clitic-resumed Topics. This prediction is borne out by data. Consider the following sentences (in which the DO il suo vicino di tavolo is not part of the Comment): (48) a.

Mariak ha presentato [ad ogni ospite]j, Maria have.3sg introduced to every guest

[il suok/j vicino di tavolo] the her/his table-mate

b. Mariak lo ha presentato [ad ogni ospite]j, Maria him have.3sg introduced to every guest

[il suok/*j vicino di tavolo] the her/*his table-mate

‘Maria has introduced his/her table-mate to every guest.’

As we can see, in the absence of clitic resumption (48a) the possessive suo can be coreferent either with the indirect object (ad ogni ospite) or with the subject (Maria), while no ambiguity arises for the clitic-resumed Topic in (48b). This is evidence

. Notice that marginalized objects are typically preceded by a narrow (not necessarily contrastive) Focus constituent, as is the case of the inverted subject (io) in (47b). “Marginalization” has thus been interpreted as a sort of “defocalization” process, obtaining the most embedded position the Focus constituent (i.e., the case of unmarked stress assignment).

 Mara Frascarelli

that the marginalized object in (48a) is merged within the VP, while the DO-Topic in (48b) is not in the scope of ogni: it is merged in extrasentential position and the only possible binder is the preverbal “subject” (merged in a Topic position that is higher than the GP projection targeted by IP-inversion). Additional evidence for a movement analysis of marginalized objects is provided by the fact that they are clause-bound (49b), while clitic-resumed Topics are not (49a): (49) a.

Non lok mangi perché stai male, il dolcek? not it eat.2sg because be.2sg sick the cake ‘You are not eating the cake because you feel sick?’

b. *Non mangi perché stai male, il dolce?

To conclude, optionality for clitic-resumption must be excluded: non-clitic resumed objects are merged within IP and then scramble, while Topics are merged in the C-domain. Two different constructions are therefore at issue, with different syntactic properties and different discourse functions.26 Intonational properties provide conclusive support to the present analysis. Let us consider (50) and (51) below, in which questions are realized. In the Roman variety examined, wh-questions have a falling contour, while yes-no questions are marked by a typical H boundary tone. It is therefore important to notice that a right-hand Topic in a wh-question is realized after the falling contour is concluded, at a baseline level (cf. Figure 11), while the final syllable of a marginalized object marks the highest point of the rising tone in a yes-no question (cf. Figure 12): (50) Ma de- chi but of- who

ce to.us

lek them

fa le do.3sg the

domandek? questions?

‘But- who is going to ask us the questions?’

. In this respect notice that the afterthought function requires Topicalization proper, excluding marginalized objects: (i) A: So che hai incontrato Luigi ieri. ‘I learnt that you met Luigi yesterday.’ B.

Sì, e *( l)’ ho trovato proprio bene, Luigi yes and him have.1sg found really well Luigi

‘Yes, and he is doing very well.’

The fine structure of the topic field 

ce

chi ma

de-

le fa

le

do

man

de

Figure 11. Right-hand Topic in a wh-question.

(51) Trovi positiva questa cosa? find.2sg positive this thing ‘Do you consider this thing as positive?’

tro vi

po

si ti

va que

sa

sta co

Figure 12. Marginalized object in a yes-no question.

The relevant contrast shows, once more, that different syntactic constructions are provided a different interpretation at the PF-interface, supporting the existence of a crucial connection between syntax, discourse grammar and prosodic structure.27 . Also notice that the constituent positiva in Figure 12 is marked by an H pitch forming a “hat contour” on the tonic vowel – consistent with the fact that it is the Focus of the sentence – and that the marginalized object is preceded by a short pause showing the presence of a prosodic boundary. Due to the rarity of this construction in the corpus examined, it was not possible to

 Mara Frascarelli

8. Conclusions The data analyzed in this paper have shown that Italian and a number of Bantu languages share significant syntactic and prosodic properties concerning the structural partitioning of discourse grammar categories. In particular, evidence has been provided that clitic-resumed Topics are merged in extrasentential position and consistently mapped into independent prosodic domains. Interface analysis has also shown the necessity of positing two Merge positions for subjects: one within the VP and one in the C-domain (as a Topic). The latter is typically the case of specific/referential DPs. The Bantu-Romance connection thus represents an important contribution towards a deeper understanding of information structure in an interface perspective. In this respect, a closer investigation on the tonal properties of Topics in comparative Bantu (along the lines proposed in F&H 2007) can represent an important step for a more comprehensive comparison in future research.

References Abels, K. & Muriungi, P. 2005. The focus particle in Kitharaka. Paper presented at the conference on Focus in African Languages. Berlin, October 2005. Antinucci, F. & Cinque, G. 1977. Sull’ordine delle parole in italiano: L’emarginazione. Studi di Grammatica Italiana 6: 121–146. Benincà, P. & Poletto, C. 2004. Topic, focus, and V2. Defining the CP sublayers. In The Cartography of Syntactic Structures, Vol. 2: The Structure of CP and IP, L. Rizzi (Ed.), 52–75. Oxford: OUP. Boeckx, C. 2003. Islands and Chains. Amsterdam: John Benjamins. Bonvino, E. 2006. Le sujet postverbal en Italien parlé: Syntaxe, zones et intonation [Biblioteque des Faits des Langues]. Paris: Ophrys. Bresnan, J. & Mchombo, S. 1987. Topic, pronoun and agreement in Chichewa. Language 63: 741–782. Bresnan, J. & Kanerva, J. 1989. Locative inversion in Chichewa. Linguistic Inquiry 20: 1–50. Büring, D. 1999. Topic. In Focus – Linguistic Cognitive and Computational Perspectives, P. Bosch & R. van der Sandt (Eds), 142–165. Cambridge: CUP. Burzio, L. 1986. Italian Syntax. A Government-Binding Approach. Dordrecht: Reidel.

measure the entity of this boundary on a scale allowing for a fruitful comparison with rightdislocated constituents. Of course, the contrast provided by Figures 11 and 12 clearly shows that marginalized objects – contrary to Topics – do not form independent I-domains and it is feasible to suppose that the relevant break corresponds to an Intermediate Phrase boundary of the type discussed for English prosody (as an anonymous reviewer suggests). However, a definite answer in this direction needs further research.

The fine structure of the topic field 

Cardinaletti, A. 2001. A second thought on emarginazione. Destressing vs. right dislocation. In Current Studies in Italian Syntax. Essays Offered to Lorenzo Renzi, G. Cinque & G. Salvi (Eds), 117–135. Oxford: Elsevier. Cardinaletti, A. 2002. Against optional and zero clitics. Right dislocation vs. marginalization. Studia Linguistica 56: 29–57. Cecchetto, C. 1999. A comparative analysis of left and right dislocation in Romance. Studia Linguistica 53: 40–67. Chomsky, N. 1977. On wh-movement. In Formal Syntax, P.W. Culicover, T. Wasow & A. Akmajian (Eds), 71–132. San Diego CA: Academic Press. Chomsky, N. 1981. Lectures on Government and Binding. Dordrecht: Foris. Chomsky, N. 2001. Beyond explanatory adequacy. In Structures and Beyond, Vol. 3: The Cartography of Syntactic Structures, A. Belletti (Ed.), 104–131. Oxford: OUP. Chomsky, N. 2005. On phases. Ms, MIT. Cinque, G. 1990. Types of A’-dependencies. Cambridge MA: The MIT Press. De Cat, C. 2002. French Dislocation. Ph.D. dissertation, University of York. Demuth, K. 1990. Locatives, impersonals and expletives in Sesotho. The Linguistic Review 7: 233–249. Demuth, K. & Harford, C. 1999. Verb raising and subject inversion in comparative Bantu. Journal of African Languages and Linguistics 20: 41–61. D’Imperio, M. 2002. Italian intonation: An overview and some questions. Probus 14: 37–69. Frascarelli, M. 2000. The Syntax-Phonology Interface in Focus and Topic Constructions in Italian. Dordrecht: Kluwer. Frascarelli, M. 2004a. L’interpretazione del focus e la portata degli operatori sintattici. In Il Parlato Italiano. Atti del Convegno Nazionale, F. Albano Leoni, F. Cutugno, M. Pettorino & R. Savy (Eds), B06. Napoli: M. D’Auria Editore, CIRASS. Frascarelli, M. 2004b. Dislocation, clitic resumption and minimality: A comparative analysis of left and right topic constructions in Italian. In Romance Languages and Linguistic Theory 2002, R. Bok-Bennema, B. Hollebrandse, B. Kampers-Manhe & P. Sleeman (Eds), 99–118. Amsterdam: John Benjamins. Frascarelli, M. 2007. Subject, Topics and the interpretation of referential pro. An interface approach to the linking of (null) pronouns. Natural Language and Linguistic Theory 25, 4: 691–734. Frascarelli, M. & Hinterhölzl, R. 2007. Types of topics in German and Italian. In On Information Structure, Meaning and Form, S. Winkler & K. Schwabe (Eds), 87–116. Amsterdam: John Benjamins. Frascarelli, M. & Puglielli, A. 2007. Focus marking and Universal Grammar. In Cushitic and Omotic languages: Proceedings of the 4th International Conference on Cushitic and Omotic Languages, A. Amha, M. Mous & G. Savà (Eds), 119–134. Cologne: Rüdiger Köppe. Frascarelli, M. & Trecci, A. Forthcoming. Subjects in a pro-drop language. Syntactic roles, discourse categories and the interpretation of pro. In Proceedings of the International Congress on La Comunicazione Parlata, A. Leoni, F. Cutugno, R. Savy (Eds). Roma: Bulzoni Editore. Giusti, G. 2006. Parallels in clausal and nominal periphery. In Phases of Interpretation, M. Frascarelli (Ed.), 163–184. Berlin: Mouton de Gruyter. Givón, T. 1976. Topic, pronoun and grammatical agreement. In Subject and Topic, C.N. Li (Ed.), 151–181. San Diego CA: Academic Press. Haegeman, L. 2002. Anchoring to speaker, adverbial clauses and the structure of CP. Georgetown University Working Papers in Theoretical Linguistics 2: 117–180.

 Mara Frascarelli Hayes, B. 1989. The prosodic hierarchy in meter. In Phonetics and Phonology. Rhythm and Meter 1, P. Kiparsky & G. Youmans (Eds), 201–260. San Diego CA: Academic Press. Hayes, B. & Lahiri, A. 1991. Bengali intonational phonology. Natural Language and Linguistic Theory 9: 47–96. Horvath, J. 1995. Structural Focus, structural case and the notion of feature assignment. In Discourse Configurational Languages, É.K. Kiss (Ed.), 176–206. Cambridge: CUP. Inkelas, S. & Zec, D. (Eds). 1990. The Phonology-Syntax Connection. Chicago IL: The University of Chicago Press. Jaeggli, O. & Safir, K. (Eds). 1989. The Null Subject Parameter. Dordrecht: Kluwer. Kanerva, J.M. 1990. Focusing on phonological phrases in Chichewa. In The Phonology-Syntax Connection, S. Inkelas & D. Zec (Eds), 145–161. Chicago IL: The University of Chicago Press. Kayne, R. 1994. The Antisymmetry of Syntax. Cambridge MA: The MIT Press. Kenesei, I. & Vogel, I. 1990. Focus and phonological structure. MS, University of Cambridge & University of Budapest. Kuno, S. 1976. Subject, theme, and the speaker’s empathy. A re-examination of relativization phenomena. In Subject and Topic, C.N. Li (Ed.), 417–444. San Diego CA: Academic Press. Lambrecht, K. 1994. Information Structure and Sentence Form: Topic, Focus and the Mental Representations of Discourse Referents. Cambridge: CUP. Longobardi, G. 2002. Postverbal subjects and the mapping hypothesis. Linguistic Inquiry 31: 691–702. Pesetsky, D. 1987. Wh-in-situ: Movement and unselective binding. In The Representation of (In)Definiteness, E. Reuland & A. ter Meulen (Eds), 98–129. Cambridge MA: The MIT Press. Pierrehumbert, J. 1980. The Phonology and Phonetics of English Intonation. Ph.D. dissertation, MIT. Poletto, C. & Pollock, J.-Y. 2004. On the left periphery of some Romance wh-questions. In The Structure of CP and IP. Vol. 2: The Cartography of Syntactic Structures, L. Rizzi (Ed.), 251–296. Oxford: OUP. Reinhart, T. 1981. Pragmatics and linguistics: An analysis of sentence topics. Philosophica 27: 53–94. Rizzi, L. 1997. The fine structure of the left periphery. In Elements of Grammar. Handbook in Generative Syntax, L. Haegeman (Ed.), 281–337. Dordrecht: Kluwer. Rizzi, Luigi. 1982. Issues in Italian Syntax. Dordrecht: Foris. Rizzi, L. 2006. On the form of chains: Criterial positions and ECP effects. In Wh Movement: Moving On, L. Cheng & N. Corver (Eds), 97–133. Cambridge MA: The MIT Press. Schneider-Zioga, P. 2000. Anti-agreement and the fine structure of the left Periphery. University of California Irvine Working Papers in Linguistics, Vol. 6. Selkirk, E. 1984. Phonology and Syntax: The Relation between Sound and Structure. Cambridge MA: The MIT Press. Shlonsky, U. 1992. Resumptive pronouns as a last resort. Linguistic Inquiry 23: 443–368. Thwala, Nhanhla. 2006. Aspects of the syntax of focus in Siswati. Talk presented at the BantuRomance Connection workshop (Leeds 2006). Uriagereka, J. 1994. Aspects of the syntax of clitic placement in Western Romance. Linguistic Inquiry 26: 79–123. Zubizarreta, M.L. 1998. Prosody, Focus and Word Order. Cambridge MA: The MIT Press.

Focus at the interface Evidence from Romance and Bantu João Costa & Nancy C. Kula

Universidade Nova de Lisboa/University of Essex Word order variation in Romance and Bantu has been related to information structure portrayed in the different discourse functions of the sentential elements involved. Based on the distribution of new information focus in Romance and Bantu, this paper argues that discourse functions need not be directly encoded in syntax. The position defended here is that syntax generates all possible structures which are filtered out at the interface with the phonological component. The prosodic phrasing of these structures is what indicates focused constituents occurring in positions of prominence. The paramount significance of prosody for the determination of focus is particularly illustrated in those cases both in Romance and Bantu where, for syntactic reasons, change in word order is restricted but prosodic effects still accompany focus. The proposed interface approach to focus accounts for the variation in focus strategies and the intimate relation of focus to prosody which is attested in Romance and Bantu. Crucially, the evidence from Romance and Bantu is complementary. The Romance data provide the necessary syntactic evidence for not positing a designated focus position in the syntactic hierarchy for focus, whereas the Bantu data show that prosodic effects may emerge in varying ways, providing evidence for not linking a single syntactic position to a given prosodic effect. The paper in this respect highlights the similarities and differences of the role of prosody in indicating focus in stress versus tone languages.

1. Introduction This paper investigates how central the role of syntax is in explaining word order and prosodic facts related to information structure. It is argued that cross-linguistic and language internal variation across Romance and Bantu favor an interface view of these phenomena, demoting some of the explanation from the syntactic component. Rather than presenting new data, we gather empirical evidence from both Romance and Bantu, and show how the comparison between these two language

 João Costa & Nancy C. Kula

groups strengthens the argument for an interface view of information structure. The paper is organized as follows: Section 2 summarizes the debate on the role the syntactic component plays in explaining cases of word order and prosodic variation related to new information focus, and sets the goals of the paper. In section 3, we present data from Romance showing that word order variation facts are a by-product of what syntax can generate, and that marked stress appears as a last-resort strategy whenever the syntax, for reasons independent of information structure, is not able to generate a given output. Section 4 presents data showing the properties of focused constituents in Bantu, providing evidence that a syntactic approach fails to explain the correlation between prosody and focus, and the variation in prosody attested in focus constructions. In section 5, we make clear how the comparison between Romance and Bantu contributes to a better understanding of the interaction between syntax, prosody and discourse. Finally, in section 6, the main conclusions of the paper are presented. 2. Syntactic approaches to focus vs. interface views The relation between information structure and syntax has been subject to a long debate over the last decade. It is consensual that word order variation in many languages reflects the discourse function of the elements of the sentence. An example of this so-called discourse configurationality is attested by data such as those illustrated by the variable position of the subject in European Portuguese: (1) A: O que é que o João fez? ‘What did João do?’ B: O João tossiu. ‘João coughed.’

#Tossiu o João. ‘Coughed João.’

(2) A: Quem tossiu? ‘Who coughed?’ B: Tossiu o João. ‘Coughed João.’

#O João tossiu. ‘João coughed.’

Focus at the interface: Evidence from Romance and Bantu 

As shown in (1) and (2), the position of the subject appears to depend on its information status: if it yields new information, or is the focus of the sentence, it occurs postverbally; if it conveys given information, as in (1), it can be moved to the sentence-initial position, where it gets nominative case and/or checks the EPP feature. This type of word order variation is well documented for Romance languages, such as European Portuguese (Costa 1996), Spanish (Zubizarreta 1998) and Italian (Belletti 2002). As discussed by most authors who have worked on this topic, the word order variation facts are not to be considered without taking into consideration the intonational properties of these contexts. Whenever the subject occurs in a low position, it bears the sentence nuclear stress, which gives it the appropriate prominence in order to be interpreted as focus. Independently of the variation in the analysis of this type of phenomena, all descriptions converge on the following generalizations: a. Word order variation may reflect different information structures; b. Information focused constituents occur at the rightward periphery where they are assigned sentence nuclear stress. A general descriptive conclusion which can be drawn from these generalizations is that any account of facts of this type must take into account syntax, prosody and information structure. The debate alluded to above has to do with different analyses of the interaction between these components in explaining these facts. In short, two main lines of enquiry have developed. One type of approach codifies focus (and other discourse-related categories) as a syntactic primitive, instantiated as a feature (e.g., Horvath 1986) or as a functional category (e.g., Belletti 2002). Under this type of approach, the prosodic effects are either disregarded or taken as a secondary consequence of the syntactic configuration. The latter view is, for instance, endorsed by followers of Cinque’s (1993) syntactic approach to sentence nuclear stress assignment. If we combine these two views, namely the codification of focus as a syntactic primitive with a syntactic algorithm for sentence stress assignment, we end up with a syntacticocentric view of this type of phenomena. An alternative approach to a syntactic analysis of discourse configurationality proposes that discourse notions like new-information focus are not encoded in syntax as functional categories. According to this view (defended, for instance, in Reinhart 1995; Szendrői 2001; and Costa 1998, 2004) the syntactic component is blind to this type of information and its role is solely to generate structures that may be filtered out at the interface with the phonological component. The latter will read the syntactic information and map it onto prosodic categories, as claimed in Nespor and Vogel (1986).

 João Costa & Nancy C. Kula

In order to distinguish between the two types of approach, let us consider a well-known example of a syntactic operation that interacts with prosodic factors: heavy NP shift, illustrated in (3): (3) a. *I read yesterday the book. b. I read yesterday all the books my friend had recommended.

A view advocating the integration of prosodic information in the syntactic component might claim that the rightward movement of the DP object is driven by prosody, since the condition for it to operate is the weight and prosodic branching of the DP object (see, for instance, Frota & Vigário 2002).1 The less syntacticocentric view would claim that syntax generates shifted and non-shifted DP objects, and at PF, shifted light DPs are ruled out, because of prosodic constraints. It is not clear how empirical argumentation on the properties of Heavy NP shift enables us to distinguish between the two approaches, so further evidence from other domains is needed in order to make an explicit comparison between the two types of analyses. We contend that a closer scrutiny of word order variation and prosodic realization of focus across Romance and Bantu provides the necessary empirical domain to address this issue. The goal of this paper is, therefore, to present data from Romance and Bantu showing that there is reason to believe that syntax does not necessarily encode focus as a primitive, and that the prosodic manifestation of focus is not derived in syntax. It will be argued that the different sources of evidence provided by the Romance and the Bantu data are complementary, leading to a more robust argumentation in favor of the interface view. 3. Focus in Romance 3.1 Focus as a syntactic primitive? As mentioned in the previous section, it is a robust fact that null-subject Romance languages allow subject-verb inversion, and that this word order is used whenever the subject is focused. Two competing analyses of inversion have been put forward in recent literature. Costa (1996) and Zubizarreta (1998), among others, propose that the inverted subject is in Spec,VP, while Belletti (2002) suggests that the low

. For the sake of the argument, it is not relevant to know whether heavy NP shift really involves rightward movement or some other type of derivation. Therefore, we are adhering to the most traditional analysis of this construction.

Focus at the interface: Evidence from Romance and Bantu 

subject sits in the specifier of a low focus-related functional category. The two partial representations are given in (4):2 (4) a. [TP V [VP Subj tV … b. [TP V [FocP Subj [ tsubj tV …

Under Belletti’s approach the focus interpretation of the inverted subject is accounted for straightforwardly, since the motivation for the low position is for it to occupy a designated functional projection responsible for the discourse interpretation. The Spec,VP approach does not account for the interpretation directly. Costa (2004) contends that syntax of null-subject languages may generate SV and VS orders. Whenever a VS order reaches the phonological component, the sentence is mapped onto an Intonational Phrase where the rightmost element is assigned the main prominence of the sentence (Frota 1998). Since focused elements must be salient, this conspiracy between syntax and prosody accounts for the focus interpretation of the subject. At this point, it is necessary to make clear how discourse interacts with the other components. According to the view advocated here, syntax is blind to the information status of the sentence constituents. As such, the low position of the subject is not a consequence of its informational status, but rather an option made legitimate by the parameter setting of null- subject languages. In a similar way, nuclear sentence stress assignment only operates by assigning prominence to the rightmost constituent in the Intonation Phrase. At this stage, in a null-subject language, there are two potential outputs being generated, requiring a filtering operation to apply. The proposal we are making is that discourse plays this filtering role. Discourse requirements mean that the focused constituent of a sentence must be prominent. Since prominence has been assigned by sentence stress assignment, an output in which the focused constituent is in the rightmost position may be selected. As such, discourse interacts with the computational component acting as a device selecting or filtering out convergent derivations. There are three domains in which the two types of analyses make different predictions. In order to understand how they can be compared, it is important to emphasize that encoding focus as a syntactic primitive predicts that focalization is a matter solved in the syntactic component. Accordingly, whenever a given constituent is marked as focus, it is a good candidate to move to a Focus Phrase. Consequently, it is expected that all focused subjects in languages which display

. An alternative to this analysis is proposed in Samek-Lodovici (2006), who argues that the right-peripheral focused constituents are right-dislocated. For arguments against generalizing this view for all cases of sentence-final information focus, see Costa (2004).

 João Costa & Nancy C. Kula

this type of discourse configurationality may occupy this position. Under the Spec,VP approach, it is expected that, provided the syntactic licensing requirements for subjects in Spec,VP are met, any subject can sit in this position. However, if such requirements are not met, it will not be able to surface there, independently of its focal nature. In other words, the main difference between the two approaches lies in the relevance of focus for determining the position of the subject in the syntactic component. 3.2 Subject-verb inversion cross-linguistically Costa and Figueiredo Silva (2006), following many authors, show that European and Brazilian Portuguese do not differ significantly as far as discourse-configurationality is concerned. The two varieties of Portuguese display a wide array of constructions for which discourse notions such as topic and focus are relevant (e.g., VP-ellipsis, null objects, topicalization). Importantly, the general Romance tendency to place focused constituents at the right periphery is observed in Brazilian Portuguese, and is attested, for instance, in the order of complements of ditransitive verbs, as illustrated in (5–8) (from Costa & Figueiredo Silva 2006):

(5) Brazilian Portuguese:

A: O que o João deu pra Maria? ‘What (did) João give to Maria?’ B: O João deu pra Maria um CD. ‘João gave Maria a CD.’

(6) Brazilian Portuguese:

A: Pra quem o João deu o CD? ‘To whom (did) João give the CD?’ B: O João deu o CD pra Maria. ‘João gave the CD to Maria.’

(7) European Portuguese:

A: O que é que o João deu à Maria? ‘What (did) João give to Maria?’ B: O João deu à Maria um CD. ‘João gave Maria a CD.’

(8) European Portuguese:

A: A quem é que o João deu o CD? ‘To whom (did) João give the CD?’ B: O João deu o CD à Maria. ‘João gave the CD to Maria.’

Focus at the interface: Evidence from Romance and Bantu 

Since these data show that Brazilian Portuguese displays the regular tendency for aligning focused constituents with the right periphery, and since, according to Belletti (2002), focus at the right periphery emerges because of the availability of the low Focus Phrase, it is legitimate to suppose that Brazilian Portuguese also has a low Focus Phrase in its inventory of functional categories. However, a problem arises when the behaviour of focused subjects with transitive and intransitive verbs is compared in these two languages. As shown in (9), in Brazilian Portuguese, unlike what we described above for European Portuguese, focused subjects do not invert. Instead, they occur preverbally and bear a marked stress: (9) A: Quem comeu o bolo? ‘Who ate the cake?’ B: a. Comeu o João. ate João

(EP/*BP)

a'. O JOÃO comeu. João ate

(#EP/BP)

Since it cannot be claimed that Brazilian Portuguese is not a discourse configurational language, there are no grounds on which to say that the difference between the two languages derives from some macro-parametric difference in the relation between word order and discourse. There is also no reasonable explanation to consider that the low FocP is not projected only when the subject of a specific subclass of verbs is involved. Let us then consider how the view which does not encode focus as a syntactic primitive handles this set of data. Remember that according to this view, a subject can be inverted for purely morpho-syntactic reasons. Its focus interpretation derives from the assignment of prosodic prominence at the interface with PF. Thus, from this view, what triggers the low position for the subject is not its informational status, and this is the crucial difference. A well-known difference between European and Brazilian Portuguese is that the latter is losing null subjects and some of the constructions related to this property (see Duarte 1995; Figueiredo Silva 1996, among others). One of the constructions related to the null subject property which has been lost is subject-verb inversion. This explains, on syntactic grounds, that the subject is not able to sit in a low position, which is completely independent of its informational status. In accordance with the type of analysis argued for above, the syntax of Brazilian Portuguese, unlike what happens in European Portuguese, is not able to generate a VS order. So, the only legitimate syntactic output that can be sent to PF is an SV order. The SV sentence is mapped onto an Intonational Phrase, but the prominence at the right periphery will not be

 João Costa & Nancy C. Kula

in accordance with discourse requirements. A marked stress is then used as a last-resort strategy, as defended in Reinhart (1995, 1999).3 As suggested in Costa and Figueiredo Silva (2006), this context is a good testing ground for comparing the two approaches to focus and inversion. The one claiming that focus is a syntactic primitive does not provide a clear explanation for the language internal and cross-linguistic variation just presented, because it predicts that all focused constituents will be able to surface at the designated functional category which is responsible for their interpretation. 3.3 (Focus-)Inversion is sensitive to locality A further argument in favour of the interface view comes from cases in which a focused subject is found inverted under an infinitival form, as described in Costa (2004). It is proposed that null-subject languages allow inversion because subject case licensing can be established under Move or Agree (Chomsky 2001), as illustrated in (10): (10) a. [IP S [CASE] I[CASE] [vP t t b. [IP I[CASE] [vP S [CASE] t

Evidence for licensing under Agree comes from the fact, illustrated under (11), that a phase boundary forbids the licensing of inversion: (11) a.

Querem ler todos os alunos esse livro. Want-3pl read all the students that book

b. *Recusaram ler todos os alunos esse livro. Refused-3pl read all the students that book c.

todos os alunos recusaram ler esse livro. All the students refused to read that book

d. *Negaram ler todos os alunos esse livro. Denied to read all the students that book e.

todos os alunos negaram ler esse livro. All the students denied to read that book

Costa (2004) shows, based on transparency effects, that the difference between querer “want”, and recusar “refuse” or negar “deny” is that the latter two project an embedded CP, blocking the Agree relation between the matrix verb and the embedded subject.

. For evidence from acquisition for the last-resort nature of this type of marked stress in focus marking, see Szendrői (2003) and Costa & Szendrői (2006).

Focus at the interface: Evidence from Romance and Bantu 

The relevance of this set of data to the debate is the following: if the condition for a subject to appear in a low position is its information status and the requirement that it appear in a focus projection, where it is assigned a specific interpretation, such a position should be available in the embedded domain, independently of there being a CP boundary between the matrix and the embedded position. Moreover, the existence of such a boundary does not preclude focus assignment, as shown in (11c) and (11e). Instead, the focus subject is marked as focus with a heavy stress, in a way reminiscent of what was observed for Brazilian Portuguese. Alternatively, if the low embedded position is just a position where the subject can be licensed under Agree by the matrix T head, it follows that this type of licensing is sensitive to a phase boundary. Such sensitivity crucially depends on purely syntactic matters, and the focus interpretation for the embedded subject is an independent matter. Summing up, this specific context reinforces the idea defended above for the comparison between European and Brazilian Portuguese that discourse configurationality is parasitic on the independent availability of a given syntactic configuration. Accordingly, the encoding of focus is not a purely syntactic matter. This case also presents evidence for the last resort strategy for marking focus with heavy stress, in line with Reinhart’s (1995) proposal, which was shown to work under the view that sentences are mapped onto intonational phrases, with assignment of default prominence and, if necessary, further repair. 3.4 Possessives in Portuguese and Italian In Cardinalletti (1998), it is shown that pre-nominal and post-nominal possessives are XPs in Italian, and that post-nominal possessives are used in definite contexts, only if they are focused. Data in (12) are drawn from Cardinaletti (1998: 19–20): (12) a. *la sua casa, non tua the her house, not yours b. la casa sua, non tua the house hers, not yours ‘Her house, not yours.’

These data are interesting, since they confirm, in the nominal domain, the general Romance tendency to place focused constituents at the right periphery. Comparing the possessive system of Italian with that of European Portuguese, Miguel (2001) and Castro and Costa (2003) provide evidence, based on modification, prosodic reduction and coordination facts, to argue that Portuguese pre-nominal possessives are heads. Assuming these results, we obtain a categorial difference between pre-nominal possessives in the two languages: in Italian, they are XPs, while in European Portuguese, they are heads.

 João Costa & Nancy C. Kula

Focalization of the possessive is one of the areas in which the two languages diverge. Unlike in Italian, in European Portuguese, a focalized possessive is not post-nominal (Castro & Costa 2003). Instead, it bears heavy stress: (13) a.

Esse é o meu problema, não o teu. that is the my problem, not the yours

b. *Esse é o problema meu, não o teu. that is the problem my ‘That is my problem, not yours.’

This is similar to what is found in focalization involving heads in the sentence. When we compare the behavior of arguments and adjuncts with the behavior of verbs in the Romance languages, it is possible to observe that XPs (arguments and adjuncts) may display different positions, reflecting different information structures, while Xºs, like verbs, have a much more rigid distribution. In line with what we proposed above, this follows from syntactic constraints: XPs may be licensed in several positions, depending on the language at stake, as was exemplified with the two positions for licensing subjects in European Portuguese, while Xº movement is categorical, occurring across-the-board. For example, if a language has V-to-T movement, the verb always moves to T, and there is no case in which it is licensed in-situ. Bearing this difference between Xºs and XPs in mind, we can understand the different focus strategies used for possessives in European Portuguese and in Italian. The difference follows from the categorial status of possessives in the two languages. Assuming that pre-nominal possessives are heads in Portuguese, they do not enter into configurations typical of XPs, which permit the post-nominal order. In Italian, because of their XP-status, possessives may surface in two different configurations. This case is interesting, since it reveals that the position of the possessive and its relation with its discourse function derives solely from its categorial status. Notice that it cannot obviously be argued that Portuguese does not resort to word order strategies for encoding focused constituents. This is simply not an available option for heads, since they are licensed in the syntax in a more categorical way. If focus were encoded directly into the syntactic structure of the DP, as proposed for instance by Roca (2006), we would expect possessives to surface in a similar position (with a possible variation between specifier and head position) independently of their categorial status. 3.5 Partial conclusion The data discussed in this section provide evidence for not encoding focus directly in syntax, but for a more articulated theory of focus in which the legitimate

Focus at the interface: Evidence from Romance and Bantu 

outputs of syntax are mapped onto the prosodic and discourse components. The facts described also enable a better understanding of the distribution of focus in Romance, departing from a simplistic view according to which one would say that this discourse function is marked by word order alone. It could nevertheless still be argued that an alternative view provides evidence for treating the tendency for focus to appear rightmost as a consequence of a direct encoding of discourse into the computational component. Cinque (1993) proposes that sentence nuclear stress is assigned within syntax to the most embedded constituent. If a constituent is marked as focus in syntax, it must stay in the position where it receives sentence stress. According to this view, syntax is capable of accounting for the position, the interpretation, and the stress of focused constituents. The evidence to be presented from Bantu languages in the next section, where no notion of sentence stress can be upheld, will help to clarify this matter and dismiss this alternative proposal. 4. Focus in Bantu Focused constituents can occur in in-situ, post-verbal, initial and final positions in Bantu, depending on the language (see Marten 2007 for a review of focus positions in Bantu). In addition, cleft constructions are an alternative way of marking focus which is used by all Bantu languages. As has been shown for Romance, Bantu languages also provide negative evidence for treating focus as encoded in syntax. We will look at two issues in this respect: focus in the so-called immediate after verb position (IAV) as established by Watters (1979) where focus involves change in word order; and in-situ focus where no change in word order is seen but, rather, prosodic phrasing identifies focused constituents. In contrast to the Romance data discussed above, the Bantu data will make no reference to the notion of sentential stress and accompanying nuclear stress rules, as the cases investigated in detail involve tone languages where stress plays no role. The idea of prominence is therefore also, pending further investigation of Bantu intonation, not to be understood in the classical sense but can only be assumed in as far as we assume focus constituents are in some sense prominent in relation to the other non-focused constituents in a sentence.4 It will be shown that focus is . The issue of prominence is not an easy one in Bantu. Sentential accent has been related to penultimate lengthening attested in most southern Bantu languages but issues of culminativity remain contentious. See Zerbian (2006) and references therein for some discussion.

 João Costa & Nancy C. Kula

best analyzed as indicated by prosodic phrasing with the focus of a sentence occupying the right-edge of a prosodic phrase, at least for the cases at hand. Like Romance languages, Bantu languages have also been argued, on the basis of data such as (14) from Chichewa (Bresnan & Mchombo 1987), which can be replicated in many other Bantu languages, to be at least partially discourse configurational (see for example Baker 2003; Morimoto 2000).5 (14) Chichewa a. Njûchi zi-ná-wá-lum-a 10.bees sm10-past-om2-bite-fv ‘The bees bit the hunters.’

b. c. d. e. f.

Zináwáluma alenje njûchi. Alenje zináwáluma njûchi. Zináwáluma njûchi alenje. Njûchi alenje zináwáluma. Alenje njûchi zináwáluma.

a-lenje. 2-hunters

(SVO)

(VOS) (OVS) (VSO) (SOV) (OSV)

Although these data are discussed in relation to the distribution of object and subject clitics in Bresnan and Mchombo (1987), namely that the different word orders for the same expression in (14) are licensed by the presence of pronominal clitics on the verb, the different word orders are also generally understood to involve information structure and the licensing of different discourse functions (Marten 2007). If such variation in word order is licensed by the syntax of Bantu and accompanied by discourse functions then it could be argued that a Focus Phrase which satisfies the discourse functions of focus constituents also exists and attracts focus to occur in the head of this functional projection. Data showing focus in an immediate after verb position, which we discuss in the next section, could be such a case. 4.1 Focus in immediate after verb (IAV) position Since Watters (1979) and Hyman and Watters (1984) the Bantu literature has made reference to an IAV position in some Bantu languages. The data on which this position is based is owed to the Western Grassfields Bantu language, Aghem, where constituents in focus consistently appear in the position immediately following the

. The following less common abbreviations are used in the Bantu data in the remainder of this paper: d = determiner; disj = disjoint form marker; es = expletive subject; fv = final vowel (in verb forms); loc = locative; om = object marker; sm = subject marker; p1 = past of today; p4 = most remote past; and qm = question marker. Numbers in glosses, unless used for person and number, indicate noun class.

Focus at the interface: Evidence from Romance and Bantu 

verb as the data in (15) show. Aghem and all Bantu examples to follow have canonical SVO word order: (15) a. t´-i-bvu – t`i--b`i-ghà mf¥ z`G nέ bέ kf¢. dogs two p1 eat today fufu d ‘The two dogs ate fufu today.’ Ø

Ø

b. à mf¢ z`G t´-i-bvu – t`i--b`i-ghà bέ kf¢ ES p1 eat dogs two fufu d ‘The two dogs ate fufu today.’ Ø

c.

nέ. today

t´-i-bvu –´ t`i--b`i-gha mf¥ z`G z´-in bέ kf¢. dogs two p1 eat when fufu d ‘When did the two dogs eat fufu?’ Ø

d. à mf¢ z`G ndúghf¢ bέ kf¢ ES p1 eat who fufu d ‘Who ate the fufu today?’ Ø

Ø

nέ à? today qm

In (15a) we see that the adverb appears in the IAV position under focus, as does the subject in (15b), as well as the WH-words in (15c–d). Based on these and similar facts in other Bantu languages a number of analyses including Sabel and Zeller (2006), Aboh (2007), Ndayiragidje (1999) and van der Wal (2006), have argued for a low focus position along the lines of Belletti (2002). The idea is that a low FocP projection is available between the inflectional domain and the verb, and that focused constituents get their focal interpretation by moving to the specifier of this position. We would like to challenge this position, at least for a sub-set of Bantu languages for which it has also been argued that an IAV position exists. Specifically, we would like to argue that there is no FocP or structurally defined position for the IAV in syntax, but rather that what turns out to be IAV focus stems from a sub-set of interrelated factors, in particular the so-called conjoint-disjoint distinction and prosodic phrasing. Before we come to this, however, let us first review some of the facts that may be considered amenable to an IAV focus position analysis. In both Makhuwa (van der Wal 2006) and Bemba there is reason to consider an IAV position as available in these languages. Consider the data from Makhuwa (van der Wal 2006: 239–241) where the focused constituent is in the IAV position in (16), and the Bemba case in (17) where it seems as though the focused constituent moves to the IAV position: (16) a.

ni-m-váhá enuní maátsi. 1pl-pres-give 10birds 6water ‘We give the birds water.’

b. ni-m-váhá maatsí enúni. 1pl-pres-give 6water 10birds ‘We give the birds water.’

 João Costa & Nancy C. Kula

c.

o-hiy-alé esheení wameétsa? 2sg-leave-perf what 16table ‘What did you leave on the table?’

d. o-hiy-alé wameétsá eshéeni? 2sg-leave-perf 16table what ‘What did you leave on the table?’ e.

o-m-vah-alé páni ekanéta? 2sg-1om-give-perf who 9pen ‘Who did you give a pen?’

In (16a) the canonical order between direct and indirect object contrasts with (16b) where the focused direct object occurs prior to the indirect object in the IAV position. (16c) and (16e) are constituent WH-questions where the WH-words occur in the IAV position in contrast to in-situ. Note though that (16d) with the WH-word in-situ is also possible, although it is less preferred. On the surface then, it seems that Makhuwa can focus constituents in the IAV position. Consider further the case of Bemba in (17) where the same sentence can attest different word orders where the constituent in IAV (convincingly in 17d–e) is in focus: (17) Bemba a. tù-kà-byáálà ínyànjé mwííbala màílò. 1plsm-fut-plant 9maize 16garden tomorrow ‘We will plant maize in the garden tomorrow.’

(broad/VP focus)

b. tùkàbyáálá ínyànjé mwííbàlà màílò. ‘We will plant maize in the garden tomorrow.’

(verb focus)

tùkàbyáálà ínyànjè mwííbàlà màílò. ‘We will plant maize in the garden tomorrow.’

(object NP focus)

c.

d. tùkàbyáálà mwííbàlà ínyànjé màílò. ‘We will plant maize in the garden tomorrow.’ e.

tùkàbyáálà màílò ínyjànjé mwííbàlà. ‘We will plant maize in the garden tomorrow.’

(locative focus) (adverbial focus)

(17a) presents the canonical order of the sentence and also broad VP focus. (17b) indicates verb focus despite having constituents following the verb, an issue to do with the difference in tone on the verb in contrast to the other examples. We return to this presently. (17c) shows an object focused in IAV but is not conclusive as this would also be its canonical position. In contrast, (17d) and (17e) show that the locative and the adverb appear in the IAV position in order to induce different scopes of focus for the same sentence.

Focus at the interface: Evidence from Romance and Bantu 

Based on the data in (16–17) it seems as though the low focus-related functional category may be a viable option. However, closer examination of the distribution of focus in these languages warrants a different perspective. An important fact about the data in (16–17) is that focus in IAV is intertwined with the so-called “conjoint-disjoint distinction”. The conjoint-disjoint distinction (an observation owed to Meeussen, 1959 for Kirundi) is a distinction found mainly in verb forms of particular tenses which indicates whether a verb is final in its clause or not. Although this distinction is generally found in verbs it has also been argued to be attested in nouns (see Sharman & Meeusen 1955; Kula 2007). The following are the oft-cited examples from Creissels (1996) from Tswana: (18) Conjoint-disjoint in Tswana a. kè tlàà bíná lí nná. isg. tns dance and 1sg. ‘I too dance/am dancing.’ b. kè tlàà bínà lé èné. isg. tns dance and 1sg. ‘I am dancing with him.’ c.

kè à mó-rátà. sm1sg. disj om1-like ‘I like him.’

disjoint (dj)

conjoint (cj)

dj

d. kè ràtà Mphó. sm1sg like m. ‘I like Mphó.’

cj

kè à mó-rátà Mphó. ‘I like him, Mphó.’

dj

e.

The conjoint-disjoint distinction can either have a segmental morpheme marker as in (18c–e) or be purely tonal as in (18a,b). Creissels’ (1996) observation of the distinction is that when the disjoint form of the verb is used, the verb is final in its clause; any following information is discourse old. In contrast, conjoint verb forms indicate that the verb is not final and there is following, new information. In (18a) the disjoint form of the verb is marked by two high tones on the verb while in (18b) the conjoint form is marked by the high-low tone pattern of the verb. In this case the verb cannot be final and new information given by him must follow the verb. In examples (18c,d) the distinction is marked by the presence of the low toned morpheme -à- in the disjoint form (18c) while its absence indicates the conjoint form in (18d). (18c) presupposes that the person to whom the pronoun him refers has already been introduced in the discourse so that its use here signals old information, hence the use of the disjoint form. In contrast, use of the proper

 João Costa & Nancy C. Kula

name Mpho in (18d) signals that it is discourse new and the conjoint form of the verb must be used. If Mpho occurs as an afterthought topic, the disjoint form of the verb must be used as shown in (18e). Note that, while the description of the conjoint-disjoint distinction (as observed by Meeussen, 1959) as indicating whether a verb is final or not in its clause is uncontentious, this is not the case with the direct relation of focus to the distinction. Thus, Buell (2005, 2006) argues against such a direct relation showing that in Zulu conjoint forms can also be found with presupposed information such as resumptives. Consider the Zulu data in (19a,b) where the disjoint form is indicated by the presence of the morpheme ya- and the conjoint form by its absence. In these cases (19a,b) the focus facts are the expected but (19c) on the other hand, where the conjoint form is used with a following resumptive pronoun, poses a problem: (19) Zulu a. abafana ba-ya-cul-a. 2boys 2sm-disj-sing-fv ‘The boys are singing.’ b. abafana ba-cul-a ingoma. 2boys 2 sm-sing- fv 9song ‘The boys are singing a song.’ c.

y-ingomai e-ngi-m-cul-el-e yonai. cop-9song rel- sm1sg -1om-sing- appl- perf 9prn ‘It’s a song that I sang (it) for him.’

dj

cj

cj

In (19c) the direct object is doubled with the resumptive independent pronoun yona but the conjoint form of the verb is still used. This is unexpected if the conjoint form directly encodes a following constituent as focused.6 Based on this and other cases where sentences with neutral focus also use the conjoint form, Buell (2006) rejects an analysis that directly links the conjoint-disjoint distinction to focus, at least for Zulu, although a weak correlation may be made in that no constituent following a disjoint form can be in focus. He rather proposes that the distinction is

. A reviewer points out that the sentence in (21c) could be analyzed as involving two clauses with yona and the noun it refers to not being in the same clause and in this case yona could be construed as not being a resumptive pronoun. Without being drawn into the long debate on the definition of resumptive pronouns, this would still be problematic for the idea of using the conjoint verb form here, as pronouns can never be focused in object position in Zulu. In this case a conjoint form would be followed by a constituent which cannot be focused; hence Buell’s point on the lack of consistency in the argument that conjoint forms unambiguously mark following focus.

Focus at the interface: Evidence from Romance and Bantu 

sensitive to constituency, in particular AgrSP, and solely indicates whether a verb form is final or not within this constituent. Any focal interpretations are treated as the by-product of a range of interpretations available to non-topicalized elements. Further, in Buell (2007) he argues that focus is essentially in-situ and adheres to focus induced extraposition where no other elements may appear in the VP when a focused element also occurs there. We return to these issues presently but we are basically in agreement with this analysis but veering towards prosody as the indicator of this “extraposition”. Note though that, interestingly, Zulu presents a slight reversal of the facts as attested so far, as although constituents in IAV are not systematically focused after a conjoint verb form, constituents in IAV are systematically not focused with disjoint forms.7 For both the Bemba and Makhuwa data presented, IAV focus only occurs in conjoint forms; even in these cases, in-situ focus, as already seen in the Makhuwa constituent question example in (16d), is marginally acceptable. Furthermore, in the Bemba case we have investigated and presented here, the conjoint-disjoint distinction interacts with prosodic phrasing so that disjoint forms have a phonological phrase boundary following the verb while conjoint forms do not. To highlight the role of prosody in these cases we will only concentrate on cases where the conjoint-disjoint is tonally marked or accompanied by specific prosodic effects. As we have also argued for Romance above we contend with the idea that the different word orders attested in the cases of IAV focus are not a result of movement to a focus position but are rather owing to a free and blind generation of syntax, concurring with the well established fact of relatively free word order in Bantu (cf. Bresnan & Mchombo 1987). The different word orders are then mapped onto a prosodic phonology whose prosodic phrasing is aided by the (tonally marked) conjoint-disjoint distinction. Thus, we are in agreement with Hyman (1999), who shows that tone can only be indirectly related to focus. Furthermore, given the observations of Buell (2006), we do not want to directly encode focus as induced by the conjoint-disjoint distinction per-se but rather argue that the conjoint-disjoint distinction, particularly when it is tonally marked, provides cues for prosodic phrasing which help to assign a focused constituent within or at a phonological phrase edge. This implies that other factors may influence prosodic phrasing choices and the conjoint-disjoint distinction is only one of these. What is crucial then is that prosodic phrasing provides enough information to identify focus constituents as falling, in this instance, at the right edge of a phonological phrase.

. This is reminiscent of Ndayiragidje’s (1999) treatment of the conjoint-disjoint forms in Kirundi which is marked by -ra in the disjoint form. He treats -ra, which can never signal narrow focus on a following constituent, as an anti-focus marker.

 João Costa & Nancy C. Kula

We now look in more detail at the relation between conjoint forms and prosodic phrasing, on the one hand, and focus, on the other. 4.1.1 Conjoint-disjoint forms and focus Within the Bantu languages, there is a disparity in IAV focus between languages with the conjoint-disjoint distinction and those without it. One explanation for this distinction on the verb is that it provides information for the hearer on whether a following constituent is salient for the interpretation of the foregoing discourse. Thus, as already pointed out, conjoint tone is indicative of new information while constituents following disjoint tone are discourse old.8 This is further illustrated by the impossibility of having conjoint tone when an object concord/pronoun is incorporated in the verb form.9 (20) Bemba a. tu-álée-lóndól-a Mutale. sm1pl-p4prog-find-fv Mutale ‘We were finding Mutale.’

cj

b. tu-áléé-mú-lóndól-á Mutale. sm1pl-p4prog-1om-find-fv Mutale ‘We were finding him Mutale.’

dj

c. *tu-áléé-mú-lóndól-a Mutale. sm1pl-p4prog-1om-find-fv Mutale ‘We were finding him Mutale.’

cj

As seen in (20c) the use of conjoint tone with the object marker mu- makes the sentence ungrammatical because the object NP Mutale is discourse old; disjoint tone must be used instead as in (20b). This is in contrast to (20a) where the object marker is absent and Mutale is new information; the conjoint tone is used. Having a closer look at the data in (16–17) we notice that they portray this conjoint-disjoint distinction. In Bemba, disjoint forms are indicated by having high tone run through the verb form to the final syllable, while no such tone spread in seen in disjoint forms. Thus, in (17b) where the verb is in focus and hence final in . For ease of reference we will refer to the tonal marking of the conjoint-disjoint distinction as involving conjoint tone and disjoint tone, respectively. . The conjoint-disjoint distinction is attested in Bemba in almost all tenses. The tenses which exclusively use tone are the three progressive pasts, the simple negative past 1, the present positive and negative, the three progressive futures, the inceptive & completive immediate future (future 1), the indicative and the subjunctive tenses. The tonal distinction is indicated by a rule which spreads the last structural high on the verb to the final vowel of the disjoint verb form (cf. Sharman 1956; Givón 1975).

Focus at the interface: Evidence from Romance and Bantu 

its clause, disjoint tone is seen on the verb form indicating that following constituents fall outside the verb’s clause. In contrast to this, conjoint tone is seen in (17c–e) where the following constituent is incorporated into the verb’s clause and is hence focused. In Makhuwa (16), on the other hand, the morphological marking of the distinction (indicated on disjoint forms) goes alongside a tonal distinction on the constituent following the conjoint form. A process of tonal lowering is applied to the focused constituent, as shown in the differing tone pattern of maátsi “water” as LHL when it is not in focus, but LLH when it is, and preceded by a conjoint verb form. In both cases, there is a strong co-relation between focused constituents and the conjoint form of the verb even though the marking of the conjoint and disjoint forms is language specific. The presence of the conjoint-disjoint distinction is attested to varying degrees in different Bantu languages.10 An interesting fact with respect to IAV focus is that those languages that lack the conjoint-disjoint distinction generally do not exhibit IAV focus and if they do, it tends to be accompanied with definitive prosodic effects. Consider in this respect Swahili and Tumbuka, which have no conjoint-disjoint distinction. In Swahili, which is non-tonal, focus is final and no IAV focus is seen, as the examples in (21) (Ashton 1942; Krifka 1985) show: (21) Swahili a. zi-jaz-e ma-ji ndoo. 10om-fill-subj 6water 10bucket

‘Fill the buckets (not the tin cans) with water.’ *‘Fill the buckets with the water.’

b. zi-jaz-e ndoo maji. 10om-fill-subj 10bucket 6water

‘Fill the buckets with water (not milk).’ *‘Fill the buckets with water.’

In Tumbuka (Downing 2006) on the other hand, focus can be either initial (not a cleft) as in (22a,b) or can be in the IAV position (22c,d). In both cases, however, prosodic effects are seen, namely; prosodic boundary narrowing creating a phonological phrase boundary after the focused constituent. In addition to this, a constituent following the focus is noticeably downstepped (indicated by !) in the

. According to Nurse (2006: 193) this is mainly the Savannah languages of zones D60, M40, M50, M60, P20–30, K21, S20–30 and S40–50 of the Bantu language area, following the classification of Guthrie (1967–71).

 João Costa & Nancy C. Kula

IAV case. Data in (22) are presented with parentheses indicating phonological phrasing: (22) Tumbuka a. (ngóoma) (ti-zamu-limilíra namchéero). maize we-will-plant tomorrow ‘We will plant maize tomorrow.’ b. (pa-mupáanda) (zi-ka-dúka mbúuzi). 16loc-3wall 10- tns-jump 10goat ‘The goats jumped over the wall.’ c.

(ba-ka-pása báana) ! (ma-búuku) 2sm-tns-give 2child 6book ‘They gave (to the) children, books.’

d. (ba-ka-pása ma-búuku) ! (báana) 2sm-tns-give 6book 2child ‘They gave books to the children.’

The cue for phonological phrase boundaries in Tumbuka (as in a number of other Bantu languages) is penultimate lengthening. Thus, in (22a,b) the penult of the focused constituent is lengthened, indicating that a phonological phrase boundary must immediately follow. The same effect is seen for focus in the IAV position, in addition to which the following constituent is downstepped; i.e., high tone on this constituent is produced at a lower pitch. We therefore see that in the absence of the conjoint-disjoint distinction in Tumbuka some other prosodic effect is at work to indicate IAV focus.11 The variable occurrence of focus in final and initial position in (21–22), and also possibly in-situ focus in Tumbuka, is not expected under a focal functional projection analysis. The prediction in this case would be that every focus occurs in the same position, at least in the same language. Indeed, analyses which try to maintain such a position have been proposed, involving a focal functional projection above the verb in the C-domain (see e.g., Thwala 2006), which then entails remnant movement to achieve IAV focus. Such analyses, however, crucially fail to relate the conjoint-disjoint facts or other prosodic effects to IAV focus. For Bemba and Makhuwa, we have seen in (16–17) that IAV focus is accompanied by prosodic effects in the form of conjoint tone in the former case and tonal lowering on the focal element in the latter. To complete the picture, let us consider two cases where the conjoint-disjoint distinction is undergoing decay. . Downing (2006) points out that above all phonological phrasing is the strongest cue for focus in Tumbuka where in-situ focus is also possible and seen to trigger specific prosody. Her argument is based on the fact that culminative sentence accent is not consistently assigned to the focused constituent in Tumbuka.

Focus at the interface: Evidence from Romance and Bantu 

In Duala and Basaa (Nurse 2006) only remnants of the conjoint-disjoint forms can be seen, dealt with under a tonal process called Metatony. Metatony is the spread of a high tone to a following tone-bearing unit. In the languages concerned here a verb-final vowel becomes high when it is followed by a complement. Disregarding metatony for the moment, the resulting effect is that verbs which are final in their clause are tonally distinguished from those that are not. The focus facts are identical to those seen with conjoint and disjoint forms: (23) Duala a. bitó bá-manda. woman 2sm-buy ‘women buy.’ b. bító bá-mandá mabato. woman 2sm-buy 6basket ‘women buy baskets.’ (24) Basaa a. a bínuŋul. ‘he sold’ b. a bínuŋúl bísεl. ‘he sold baskets.’

(
Thus, in Duala and Basaa, where a tonal distinction with respect to a following complement can still be seen, we have IAV focus as opposed to initial focus, pointing to the fact that the tonal effects, rather than the conjoint-disjoint distinction which is here under decay, are the indicator of focus (via prosodic structure). To sum up, focus effects in IAV coincide with the conjoint-disjoint distinction because the distinction is associated with prosodic effects. Languages lacking the distinction show either an absence of focus in IAV position, preferring to have focus on the right or left periphery, or IAV focus only if it is accompanied by extra prosodic effects. We would like to suggest that this correlation points to a prosodic marking of focus and further suggest that this can quite easily be extended to focus in initial and final position since the emphasis is on prosodic phrasing and not the conjoint-disjoint distinction itself. We consider some further evidence for disfavoring a syntactic analysis of focus in the next sub-section, before we explore how a prosodic characterization of focus can be formalized. 4.1.2 Further evidence against a syntactic focus position in Bantu A few other analyses in the Bantu literature also argue against a FocP for Bantu. Marten (2007) proposes an analysis in Dynamic Syntax which does not assign syntactic positions to topic and focus but rather suggests that focus interpretation results from an interaction between structural underspecification and context.

 João Costa & Nancy C. Kula

As already pointed out, Buell (2005, 2007) assumes an analysis where focus is essentially in-situ with a VP based focus constituent inducing extraposition of all other elements in the VP. Hyman and Polinsky (2007) also argue against a low FocP analysis of the Aghem data illustrating IAV focus, as briefly presented in (15) above. Their proposal is that a focus operator in CP unselectively binds the lowest XP in the clause which then emerges as focused. While these are all viable syntactic options, we opt for an account that also incorporates the associated prosodic effects. Needless to say, the prosodic effects could well be achieved in one of the ways above, but we leave this to a future occasion. We instead focus on the arguments which have been presented against a FocP. Firstly, for IAV focus, not all constituents which appear in this position are always in focus, as would be predicted by the presence of a FocP. As we have seen in Makhuwa and Bemba, this may only be the case where a conjoint verb form is present. Since the conjoint form is only attested in some tenses per language, in tenses where the distinction is absent, constituents in the IAV will not be in focus. Secondly, the presence of a low FocP requires the verb to precede the focused element in the specifier of FocP above the VP which would require some form of V-to-T movement which, as Buell (2005) argues for Zulu on morpho-syntactic grounds, is poorly motivated. Thirdly, some of the Bantu languages can focus constituents in multiple positions. As seen in Tumbuka, focus can be both initial and in IAV. Similarly, in Bemba, question words may be focused in IAV or in-situ. An analysis postulating a low focus position would then have to further explain how and when movement to this position occurs, and why it appears to be optional in Tumbuka and Bemba. As seen in the Tumbuka case, the different prosodic effects accompanying the attested focus positions directly support a prosodic analysis. Finally, the availability of multiple constituent questions remains unaccounted for if only one FocP is available. Consider the multiple question forms in (25) below. Zulu data are taken from Buell (2007): (25) Zulu a. u-zo-nika bani ini? 2s-fut-give 1who 9what ‘Who will you give what?’ Bemba b. ú-ká-shìtìshà bààní ínshì? 3sg-fut-buy.caus 2who what ‘Who will you sell what?’

c.

Focus at the interface: Evidence from Romance and Bantu 

ú-ká-shìtìshà ínshì bàànì? 3sg-fut-buy.caus what 2who ‘What will you sell to who?’

The Bemba examples show that both the subject and object WH-phrases appear after the verb, but that their respective order is free. This seems to indicate that their order is not determined by fixed structural positions, but rather that the syntax of multiple WH-questions is syntactically less restricted, and thus that focus is not tied to a specific designated syntactic projection. We now consider an outline of a prosodic account of focus. 4.2 A prosodic account of focus Returning to the Bemba facts presented in (17), we contend that syntax generates the four different word orders which get mapped into different prosodic structures at PF. Phonological phrasing in Bemba is broadly indicated by a downstepped high, usually accompanied by a pause. A final phonological phrase is indicated by a final low tone. This prevents a high tone doubling rule (which doubles a high tone onto a following vowel) from applying, in cases where its conditions are met. There is a rising tone on the final vowel of the verb in the conjoint form, so that the absence of a phonological phrase boundary is indicated by the rise. In addition, a segmental effect such as vowel fusion between adjacent constituents is also indicative of the absence of a phonological phrase boundary whenever the conditions for this are met (as in example 26c).12 Consider the phrasing of the Bemba data in (26) below: (26) Bemba a. (tù-kà-byáálà ínyànjé mwííbàla! màílò)φ. 1plsm-fut-plant 9maize 16garden tomorrow ‘We will plant maize in the garden tomorrow.’

(broad/VP focus)

b. (tùkàbyáálá!)φ (ínyànjé mwííbàla! màílò)φ. ‘We will plant maize in the garden tomorrow.’

(verb focus)

(tùkàbyáálèényànjè)φ (mwííbàla! màílò)φ. ‘We will plant maize in the garden tomorrow.’

(object NP focus)

d. (tùkàbyáála! mwííbàlà)φ (ínyànjé màílò)φ. ‘We will plant maize in the garden tomorrow.’

(locative PP focus)

(tùkàbyáála! màílò)φ (ínyjànjé mwííbàlà)φ. ‘We will plant maize in the garden tomorrow.’

(adverbial focus)

c.

e.

. The full characterization of phonological phrasing and intonation in Bemba is still under investigation but our current state of knowledge is sufficient for the present exposition.

 João Costa & Nancy C. Kula

The generalization to be drawn from this phrasing is that the focused constituent occurs at the right edge of a phonological phrase, both for the disjoint form in (26b), where the verb is in focus, and the conjoint forms in (26c–e). Also of significance is the fact that the tonal pattern of (26a), where a broad focus interpretation holds despite having a conjoint form of the verb, differs from (26c–e) tonally, showing that there is no phonological phrase break between any of the constituents following the verb. Thus we have a case here where phonological phrasing overrides the conjoint form.13 An analysis which identifies the relevant right edge of the phonological phrase is easily formalized using interface and Alignment constraints in Optimality Theory, for example, along the lines of Truckenbrodt (1999) and Kula (2007). The salient point is, however, that under a syntax which generates all word orders, this prosodic phrasing acts as a filter on which orders are actually possible and which discourse functions occur with them; i.e., it identifies which constituent is focused by reference to a phonological phrase edge. Note that we do not actually allude to a notion of prominence, as it still remains to be seen what relevance this has for tone as opposed to stress languages. Unlike some of the Bantu languages discussed here, Bemba has no penultimate lengthening which may support any notion of sentence accent. The prosodic effects which highlight or help identify focused constituents are in this sense carried solely by the phonological phrasing. This prosodic phrase edge effect can presumably also be extended to Bantu languages with initial and final focus, as seen in Tumbuka and Swahili respectively, with the difference simply being that in these cases the conjoint-disjoint distinction plays no role in designating the relevant prosodic phrase edges. We now turn to a case where no change in word order is seen but different constituents within the same sentence can nonetheless be focused. 4.3 In-situ focus In Chichewa, Kanerva (1990: 98) convincingly argues that focus is prosodically marked, and that the focus of different constituents in a sentence coincides with different phonological phrasings. Consider in this respect (27): (27) Chichewa a. (a-na-ményá nyumbá ndí mwáála)φ sm-tns-hit 9house prep rock ‘She/he hit the house with a rock.’

(broad/VP focus)

. This could be an answer, at least for Bemba, to one of Buell’s (2006) concerns that sentences with neutral focus also use the conjoint form; their prosodic phrasing crucially differs from conjoint forms where following constituents are in focus.

Focus at the interface: Evidence from Romance and Bantu 

b. (a-na-méenya)φ (nyuúmba)φ (ndí mwáála)φ. ‘She hit the house with a rock.’ c.

(a-na-ményá nyuúmba)φ (ndí mwáála)φ. ‘She hit the house with a rock.’

d. (a-na-ményá nyumbá ndí mwáála)φ. ‘She hit the house with a rock.’

(verb focus) (object NP focus) (oblique PP focus)

Phonological phrase boundaries are marked by penultimate lengthening and tonal retraction in Chichewa, as (27) illustrates. Tonal retraction ensures that the end of a phonological phrase does not coincide with a high tone, seen here, for example, in the change of the final tone of anamenya when it is phrase-final. The phonological phrase restructuring reveals that the right edge of a phonological phrase coincides with a focused constituent. We assume that on some principled grounds the leftmost phonological phrase is the relevant one for focus in (27b,c). (27a,d), which have the same phrasing and same tonal patterns, can either involve broad VP focus or focus of the rightmost constituent as in (27d). The relevance of these data to the current discussion is that under a prosodic analysis, prosodic phrasing, just as in the conjoint cases, lends support to in-situ focus so that a focused constituent occurs at the edge of a relevant phonological phrase. In a syntactic analysis, on the other hand, these data would be surprising, given that Chichewa has illustrated word order flexibility in (14). In this case, if a syntactic position was available for focus we would expect the focus to occur there and nowhere else. Under our view prosodic phrasing assigned to the canonical word order at PF would yield the outputs in (27). Thus, these data are best served by an analysis that does not codify focus in syntax. The foregoing discussion on Bantu has tried to establish that prosody is the most robust cue for focus. Thus, while focus has been related to change in word order and the conjoint-disjoint distinction in IAV focus, thereby suggesting a possible (low) focus projection, it has been shown that in all these cases there are prosodic effects which accompany such movement manifested either on the verb or the noun. Furthermore, in languages where the conjoint-disjoint distinction is not available, constituents can be focused in initial and final position, with IAV also an option, as long as the relevant prosodic effects are present. It has been shown that the positions for focus are varied, occurring both at the right and left periphery, in clause-medial position (the IAV) and virtually in any position (in-situ). What remain consistent in all these cases are the prosodic effects, albeit manifested differently, which accompany focus. Thus we contend that syntax generates all the possible word orders allowed in a language and prosodic phrasing assigned at PF aids the identification of focus. Let us now have a look at how these Bantu facts relate to the Romance facts already discussed and how they both conspire to support an interface view of focus.

 João Costa & Nancy C. Kula

5. The Romance-Bantu conspiracy Remember that at the end of section 3 we raised the hypothesis, following Cinque (1993), that syntax alone is capable of accounting for the prosodic, positional and interpretive effects of focused constituents. It was not clear, based on the Romance data alone, how to discard this hypothesis. Based on the evidence from Bantu, it is now possible to readdress this issue. The Romance facts showed that there is not a designated functional position for focus. Instead, it was argued that focus tends to appear where the sentence nuclear stress falls, and, whenever, for purely syntactic reasons, this is impossible, repair strategies occur post-syntactically in order for prominence to be assigned. These facts may be summarized as follows: a. There is no designated information focus position in Romance; b. The fact that information focus appears on the right periphery is a tendency, contingent on facts that have little or nothing to do with discourse, and not a categorical fact; c. Focus is always marked prosodically in Romance. These conclusions are important for the discussion of Bantu languages. The common observation that in Bantu languages, as in Romance, word order appears to be sensitive to discourse functions, calls for a common explanation. If one assumes that in Romance this type of sensitivity is to be explained by encoding discourse functions into the syntactic component, this analysis should extend to Bantu. The evidence put forward in section 3 casts doubt on this line of inquiry for Romance, and offers an alternative approach based on an interface view of syntax, prosody and discourse. If this type of interface view holds for Romance, it should also hold for Bantu. In fact, in section 4, we showed that in Bantu there is also clear evidence to say that there is no necessary one-to-one correlation between word order and information structure. Crucially, the Bantu facts also show that focus can be indicated by various forms of prosodic effects from which we conclude similarly to Romance that: a. There is no designated information focus position in Bantu; b. The fact that information focus has been strongly linked to an IAV position is contingent on prosodic effects which may be replicated in other positions and cannot be a categorical fact; c. Focus is always related to prosody in Bantu.14 . One possible counter-example to this is Northern Sotho for which Zerbian (2006) argues

Focus at the interface: Evidence from Romance and Bantu 

This evidence is important for comparing the interface view with an analysis in which position, prosody and interpretation are all accounted for within the syntactic component. Summing up, the data from Romance made it possible to question the evidence for a one-to-one relationship between word order and information structure, which turned out to be important for the evaluation of the Bantu data. On the other hand, the array of prosodic facts across Bantu languages made it possible to show that prominence assignment cannot be accounted for within syntax in terms of sentence nuclear stress, as has been proposed for Romance. Together, the evidence coming from the two language groups shows that there is a wider variation, both on the prosodic and on the syntactic side, than one might suppose, but the two sets of evidence conspire in favour of the interface view of focus, according to which focus is not a syntactic primitive, and prosody and discourse act post-syntactically. 6. Conclusions We have argued throughout this paper that the distribution of information focus is best understood if focus is not considered a syntactic primitive. The evidence for this claim came from the distribution of focus across Romance and Bantu. It was argued that cross-linguistic and language internal variation in the distribution of focus is accounted for if one takes into consideration the structures independently made available by syntax, solely based on syntactic primitives, and their mapping onto prosodic structure. It is often mentioned that Romance and Bantu have in common the fact that word order reflects information structure. We hope to have provided evidence for a more precise formulation of this similarity. What the two language families have in common is a syntax with a flexibility, due to independent syntactic reasons, which makes it possible to map different word orders onto prosodic structures where the focused constituents are assigned different types of prominence. However, the two

that there is no consistent prosodic cue for focus because syntactic requirements override prosodic requirements but the prosodic shape of an utterance remains constant; penultimate lengthening remains final. This scenario is, however, not unexpected in our proposal since we claim that discourse functions do not trigger movement but only act as a filter to structures independently licensed by syntax. Thus, if syntax allows a constituent to occur in clause final position its focus status will coincide with the accentual cues of the clause final position. What is crucially different from Romance, as Zerbian points out, is that no default or last resort strategy is applied to the focus constituent when it does not fall in final position.

 João Costa & Nancy C. Kula

language families differ in their realisation of ‘prominence’, with Romance generally opting for coinciding with stress as determined by nuclear stress rules and Bantu opting for prosodic phrasing as the main indicator of the position of focus. Rather than providing evidence for encoding discourse notions in the syntactic component, a closer look at the similarities and differences between the two language families strengthens an interface approach to focus and word order.

References Aboh, E.O. 2007. Leftward versus rightward focus: The Kwa-Bantu conspiracy. SOAS Working Paper in Linguistics 15: Bantu in Bloomsbury, Special issue on Bantu Linguistics, N.C. Kula & L. Marten (Eds), 81–104. London: SOAS. Ashton, E.O. 1942. Swahili Grammar. London: Longman. Baker, M.C. 2003. Agreement, dislocation and partial configurationality. In Formal Approaches to Function in Grammar: In Honor of Eloise Jelinek, A. Carnie, H. Harley & M.-A. Willie (Eds), 107–134. Amsterdam: John Benjamins. Belletti, A. 2002. Aspects of the low IP area. In The Structure of IP and CP. The Cartography of Syntactic Structures 2, L. Rizzi (Ed.), 16–51. Oxford: OUP. Bresnan, J. & Mchombo, S.A. 1987. Topic, pronoun, and agreement in Chichewa. Language 63: 741–782. Buell, L. 2005. Issues in Zulu Verbal Morphosyntax. Ph.D. dissertation, University of California at Los Angeles. Buell, L. 2006. The Zulu conjoint/disjoint verb alternation: Focus or constituency? ZAS Papers in Linguistics 43: Papers in Bantu Grammar and Description, L.J. Downing, L. Marten & S. Zerbian (Eds), 8–30. Berlin: ZAS. Buell, L. 2007. Evaluating the immediate postverbal position as a focus position in Zulu. Ms, Leiden University. Cardinaletti, A. 1998. On the deficient/strong opposition in possessive systems. In Possessors, Predicates, and Movement in the Determiner Phrase, A. Alexiadou & C. Wider (Eds), 17–53. Amsterdam: John Benjamins. Castro, A. & Costa, J. 2003. Weak forms as Xº: Prenominal possessives and preverbal adverbs in Portuguese. In Romance Linguistics: Theory and Acquisition, A.T. Pérez-Leroux & Y. Roberge (Eds), 95–110. Amsterdam: John Benjamins. Chomsky, N. 2001. Derivation by phase. In Ken Hale. A Life in language, M. Kenstowicz (Ed.), 1–52. Cambridge MA: The MIT Press. Cinque, G. 1993. A null theory of phrase and compound stress. Linguistic Inquiry 24: 239–298. Costa, J. 1996. Positions for subjects in European Portuguese. In Proceedings of WCCFL XV, B. Agabayani et al. (Eds), 49–63. Stanford CA: CSLI. Costa, J. 1998. Word Order Variation. A Constraint-based Approach. Ph.D. dissertation, Leiden University. Costa, J. 2004. Subject Positions and Interfaces. The Case of European Portuguese. Berlin: Mouton de Gruyter. Costa, J. & Figueiredo Silva, M.C. 2006. On the (in)dependence relation between syntax and pragmatics. In The Architecture of Focus, V. Molnar & S. Winkler (Eds), 83–104. Berlin: Mouton de Gruyter.

Focus at the interface: Evidence from Romance and Bantu 

Costa, J. & Szendrői, K. 2006. Acquisition of focus marking in European Portuguese: Evidence for a unified approach to focus. In The Acquisition of Syntax in Romance Languages, V. Torrens & L. Escobar (Eds), 319–329. Amsterdam: John Benjamins. Criessels, D. 1996. Conjunctive and disjunctive verb forms in Setswana. South African Journal of African Languages 16: 109–115. Downing, L.J. 2006. The prosody and syntax of focus in Chitumbuka. ZAS papers in Linguistics 43: Papers in Bantu Grammar and Description, L.J. Downing, L. Marten & S. Zerbian, (Eds), 55–79. Berlin: ZAS. Duarte, M.E. 1995. A Perda do Princípio Evite Pronome no Português do Brasil. Ph.D. dissertation, Universidade Estadual de Campinas. Figueiredo Silva, M.C. 1996. A posição do sujeito em português brasileiro – frases finitas e infinitivas. Campinas: Editora da Unicamp Frota, S. 1998. Prosody and focus in European Portuguese. Ph.D. dissertation, University of Lisbon. Frota, S. & Vigário, M. 2002. Efeitos de peso no Português Europeu. In Saberes no Tempo. Homenagem a Maria Henriqueta Costa Campos, M.H. Mateus & C.N. Correia (Eds), 315–333. Lisboa: Colibri. Givón, T. 1975. Focus and scope of assertion. Some Bantu evidence. Studies in African Linguistics 6: 185–205. Guthrie, M. 1967–71. Comparative Bantu, 4 Vols. Farnborough: Gregg. Horvath, J. 1986. focus in the Theory of Grammar and the Syntax of Hungarian. Dordrecht: Foris. Hyman, L.M. 1999. The interaction between focus and tone in Bantu. In The Grammar of Focus, G. Rebuschi & L. Tuller (Eds), 151–177. Amsterdam: John Benjamins. Hyman, L.M. & Polinsky, M. 2007. Focus in Aghem. Ms, University of California at Berkeley & Harvard University. Hyman, L.M. & Watters, J.R. 1984. Auxiliary focus. Studies in African Linguistics 15: 233–273. Kanerva, J. 1990. Focus and Phrasing in Chichewa Phonology. Ph.D. dissertation, Stanford University CA. Krifka, M. 1985. Word order and word order change in Swahili. Kiswahili, Journal of the Institute of Kiswahili Research 52: 13–72. Kula, N.C. 2007. Effects of phonological phrasing on syntactic structure. The Linguistic Review 24: 201–231. Marten, L. 2007. Focus strategies and the incremental development of semantic representations: Evidence from Bantu. In Focus strategies in African languages: The interaction of focus and grammar in Niger-Congo and Afro-Asiatic, E.O. Aboh, K. Hartmann & M. Zimmermann (Eds), 113–135. Berlin: Mouton de Gruyter. Meeussen, A.E. 1959. Essaie de grammaire Rundi [Annales du Musée Royal du Congo Belge, Série Sciences Humaines 24]. Tevuren: Royal Museum of Central Africa. Miguel, M. 2001. Para uma tipologia dos possessivos. Actas do XVII Encontro Nacional da Associação Portuguesa de Linguística, 287–300. Lisbon: APL. Morimoto, Y. 2000. Discourse Configurationality in Bantu Morphosyntax. Ph.D. dissertation, University of Stanford. Ndayiragije, J. 1999. Checking economy. Linguistic Inquiry 30: 399–444. Nespor, M. & Vogel, I. 1986. Prosodic Phonology. Dordrecht: Foris. Nurse, D. 2006. Focus in Bantu. ZAS Papers in Linguistics 43: Papers in Bantu Grammar and Description, L.J. Downing, L. Marten & S. Zerbian (Eds), 189–207. Berlin: ZAS. Reinhart, T. 1995. Interface strategies. Ms, Utrecht University.

 João Costa & Nancy C. Kula Reinhart, T. 1999. The processing cost of reference-set computation: Guess patterns in acquisition. UiLOTS Working Papers in Linguistics. Roca, F. 2006. Demonstrative systems in Spanish and in Romance: Some remarks on diachronic and dialectal variation. Paper presented at Workshop “Romania Nova”, Maceió- Brazil. Sabel, J. & Zeller, J. 2006. Wh-question formation in Nguni. In Selected Proceedings of the 35th Annual Conference on African Linguistics, J. Mugane et al. (Eds), 271–283. Somerville MA: Cascadilla. Samek-Lodovici, V. 2006. When right dislocation meets the left-periphery. A unified analysis of Italian non-final focus. Lingua 116: 836–873. Sharman, J.C. 1956. The tabulation of tenses in Bantu languages. Bemba: Northern Rhodesia. Africa 26: 29–46. Sharman, J.C. & Meeussen, A.E. 1955. The representation of structural tones, with special reference to the tonal behaviour of the verb, in Bemba, Northern Rhodesia. Africa 25: 393– 404. Szendrői, K. 2001. Focus and the Syntax-phonology Interface. Ph.D. dissertation, University College London. Szendrői, K. 2003. Narrow and wide focus interpretation in the acquisition of only sentences. Paper presented at GALA 2003, Utrecht University. Thwala, N. 2006. Aspects of the syntax of focus in SiSwati. Ms, SOAS. Truckenbrodt, H. 1999. On the relation between syntactic phrases and phonological phrases. Linguistic Inquiry 30: 219–255. Van der Wal, J. 2006. The disjoint verb form and an empty immediate after verb position in Makhuwa. ZAS Papers in Linguistics 43: Papers in Bantu Grammar and Description, L.J. Downing, L. Marten & S. Zerbian (Eds), 233–256. Berlin: ZAS. Watters, J.R. 1979. Focus in Aghem. In Aghem Grammatical Structure [Southern California Occasional Papers in Linguistics 7], L.M. Hyman (Ed.). Los Angeles CA: University of California, Department of Linguistics. Zerbian, S. 2006. Expression of Information Structure in the Bantu Language Northern Sotho. Ph.D. dissertation, Humboldt University. Zubizarreta, M.L. 1998. Focus, Prosody and Word Order. Cambridge MA: The MIT Press.

Agreement in thetic VS sentences in Bantu and Romance* Jenneke van der Wal

Leiden University Centre for Linguistics

Both Bantu and Romance languages use a V(erb) S(ubject) construction to express thetic (“out-of-the-blue”) sentences. Two types of languages can be distinguished within these language families, with respect to the verbal agreement in a thetic VS sentence: in type 1 the verb has default agreement, whereas in type 2 the verb agrees with the postverbal subject. In the Bantu languages these two types also display a difference in the use of conjoint and disjoint verb forms. Collins (2004), Carstens (2005), and Baker (2008) have previously analyzed such agreement and word order phenomena. These accounts, attributing the differences between types 1 and 2 to parameter settings of the Agree system, do not offer a satisfactory explanation. This paper proposes that the difference is due to the status of the agreement of the verb, which is pronominal in type 1 languages and purely grammatical in type 2. Arguments for this analysis are found in Case, Binding Theory and information structure. The focus in this paper is on the Romance languages French and Italian, and on the Bantu languages Sesotho and Makhuwa.

1. Introduction In the literature concerning Information Structure a distinction has often been made between categorical and thetic statements. The categorical is a twofold judgement, stating the existence of an entity and then predicating something on it. A thetic statement, on the other hand, is an unstructured judgement expressing only the recognition (or rejection) of an event or a state. Sasse (1987) uses (1) and (2) as typical examples of these two types of statements. The categorical judgement in (2) first names the entity John and then predicates on him that he is intelligent.

* I want to thank my informants Ali Pwanale, Joaquim Nazario, Momade Ossumane, Adelino

Raposo and Sualehe Molde. I thank for their helpful input the members of our Bantu Team (Lisa Cheng, Thilo Schadeberg, Leston Buell and Kristina Riedel), Mélanie Jouitteau, Anna Cardinaletti, the two anonymous reviewers and the participants in the ESF exploratory workshop ‘The Bantu Romance Connection’ in Leeds, 26/27 May 2006. The main ideas in this paper have been presented in a poster in that workshop.

 Jenneke van der Wal

The thetic judgement in (1) does not involve the independent recognition of some entity, but simply recognizes the state or situation of “raining”. (1) It is raining. statement (2) John entity

is intelligent. statement

Lambrecht (1994, 2000) takes thetic sentences to be “topicless”, because of their need to be paradigmatically distinguishable from categorical statements (Lambrecht 2000). He takes a topic-comment articulation as the unmarked state of affairs. Lambrecht calls this the Predicate Focus (PF) construction, which can be distinguished from Argument Focus (AF, in other studies sometimes referred to as “narrow focus”) and Sentence Focus (SF, alternatively called “all-new utterance” or “thetic sentence”). In a PF sentence the subject is usually the topic. In a SF or thetic sentence, however, both the subject and the predicate are in focus. In order to avoid the default reading of the subject as the topic of the sentence (as in a categorical or PF sentence) the subject must be “detopicalized”. One strategy for detopicalization is to use a Verb-Subject order. I show in this paper that languages in both the Bantu and the Romance language families make use of this strategy. There is a difference, however, in the agreement on the verb in thetic VS sentences, which is found in both language families. Bantu and Romance languages can be divided into two types according to their agreement. In type 1 languages the verb has default agreement, whereas in type 2 languages the verb agrees with the postverbal subject. The grammatical system of conjoint and disjoint verb forms and the phonological phrasing of the Bantu languages also behave differently in thetic VS sentences. These data are discussed in section 2. In section 3 the accounts of Collins (2004), Carstens (2005) and Baker (2008) are examined. They attribute the difference in agreement patterns to a different parameter setting for agreement. In type 1, agreement would require internal merge (move), and the verb (or T) agrees with the element which moves to its specifier. In type 2 the operations Move and Agree can be applied separately and agreement is dependent on Case checking. These analyses are shown to not fully account for all the agreement facts. An alternative explored in section 4 is to investigate the characteristics of the verbal agreement markers. Bresnan and Mchombo (1987) argue that these agreement markers can be either pronominal or purely grammatical. I propose that the differences between the two types in agreement as well as conjoint/disjoint verb forms are a result of the pronominal or grammatical status of the agreement marker, which is pronominal in type 1 non-agreeing languages and grammatical in type 2 agreeing languages. Section 5 evaluates the analyses and forms a conclusion.

Agreement in thetic VS sentences in Bantu and Romance 

2. Expression of theticity 2.1 Detopicalization As mentioned in the introduction, the subject of a thetic sentence must be detopicalized in order not to be interpreted as a topic. The principle of detopicalizing the subject is defined by Lambrecht (2000: 624) as follows: “SF marking involves cancellation of those prosodic and/or morphosyntactic subject properties which are associated with the role of subjects as topic expressions in PF sentences”. Topical subjects are ideally preverbal, pronominal and unaccented. Thus, subjects in thetic sentences are typically expressed with (some of) the opposite features: they are full nouns, accented and in a postverbal position. Lambrecht captures this in the Principle of subject–object neutralization: “In a SF construction, the subject tends to be grammatically coded with some or all of the prosodic and/or morphosyntactic features associated with the focal object in the corresponding categorical construction” (p. 626). Languages with a more rigid syntax and a flexible focus structure will place a focus marker or move the sentence accent (so-called “prosodic inversion”, Lambrecht 2000), rather than moving a constituent, whereas those with a rigid focus structure and a much more flexible syntax will rearrange the word order instead of shifting the accent (Van Valin 1999). Examples of the first language type are English and Ewe. In English (3a) the stress has been shifted to the subject and the word order remains as in a categorical sentence (3a’). Ewe detopicalizes the subject by placing a focus marker (3b); the word order is left unchanged. The second type is exemplified for Italian (3c) and Sesotho (3d),1 which use the SV order in categorical sentences (3c’) and the VS order in thetics.2 The capitals in the examples indicate the main sentence accent.

. The language Sesotho (Guthrie’s S.30) is spoken in Lesotho, South Africa and Botswana by approximately 4.9 million people (Ethnologue). . One other way to construct a thetic sentence is a so-called “split structure” (Sasse 1996) or “bi-clausal presentational construction” (Lambrecht 1994). The first clause has an existential marker (usually “be” or “have”) followed by the entity to be presented and the second is a relative clause adding information on the introduced referent. The relative clause can also be left out to form an existential or presentational thetic sentence.

(i)

Once there was a wizard who was very wise and rich. (Lambrecht 1994)

Makhuwa (ii)

Y-aá-háa-vo enámá e-motsá e-n-aátsím-íyá ncóco. 9-past-be-loc 9.animal 9-one 9-pres-call-pass 3.impala ‘There was an animal which is called impala’.

These are used in many languages, often in addition to a VS or subject-accented construction,

 Jenneke van der Wal

(3) a. My CAR broke down. (‘What happened?’) a'. My car broke DOWN. (‘What happened to your car?’) Ewe (Ameka, to appear) b. Deví-á-wó-é gba ze-a. child-def-pl-foc break pot-def ‘The children broke the pot.’ (‘What happened?’) Italian (Lambrecht 1994) c.

Mi si è rotta la MACCHINA. (‘What happened?’) me se is broken the car

c’. La mia macchina si è ROTTA. (‘What happened to your car?’) the my car se is broken Sesotho (Demuth 1990)3 d. Hó-lisá ba-shányána. 17-herd 2-boys ‘There are boys herding.’

2.2 Romance and Bantu thetics In both the Romance and the Bantu language families VS constructions can be used to express a thetic sentence, as exemplified in (4)–(7). Most languages in these families require the verb to be intransitive, with some languages having an even stronger requirement that the verb be unaccusative (others allow both unaccusatives and unergatives). The examples below are from both language families and show an unaccusative verb in the (a) examples and, if possible, an unergative in the (b) examples (see section 2.4 for more information on the markings “cj” and “dj”).4 to form an existential or presentational construction. They are primarily used for one type of thetic sentence: the entity-central type. The other type is the event-central thetic sentence, on which the current paper focuses. For more information on various functions and constructions of thetic sentences, see Sasse (1996). . Abbreviations and symbols used in this paper: 1/2/3 etc (noun classes), 1sg/pl (1st person singular/plural), acc (accusative), asp (aspect), caus (causative), cj (conjoint), cls (subject clitic), dem (demonstrative), dj (disjoint), ds (dummy subject), dur (durative), fem (feminine), fut (future tense), fv (final vowel), h (high tone), impf (imperfective tense), int (intended meaning), l (low tone), masc (masculine), neg (negative), nom (nominative), om (object marker), pass (passive), past (past tense), perf (perfective tense), pers (persistive), pp (past participle), pres (present tense), p2 (past tense), redupl (reduplication), refl (reflexive), rel (relative), rem (remote tense), t (tense), |tt| (retroflex voiceless stop), | (right boundary of phonological phrase). Liaison is indicated by an apostrophe, high tones are indicated by an acute accent (on or before the element), low tones are unmarked. . The typical Northern Italian dialects pattern with French, and these also allow the VS

Agreement in thetic VS sentences in Bantu and Romance 

Sesotho (Demuth 1990, adapted) (4) a. CJ Hó-fihl-ílé li-pé:re. 17-arrive-perf 10-horses ‘There arrived horses.’ b. CJ Hó-lóha bo-nkhó:no. 17-weave 2b-grandmother ‘There are grandmothers weaving.’ French (Lambrecht 2000) (5) a. Il est (*sont) venu trois femmes. it is (*are) come:pp:masc.sg three women ‘There came three women.’ b. Il tombe une goute. it falls a:fem drop:fem ‘A DROP is falling.’ Makhuwa5 (6) a. DJ O-hoó-khwá mwanámwáne. 1-perf.dj-die 1a.child ‘There has died a child.’

construction with unergative verbs, as in (i). Since the information structure has not yet been much discussed in the syntactic accounts of the Italian dialects (as is visible, for example, in the lack of an idiomatic translation of (i)), I use French as an example of the Romance type 1 in this paper.

Ciàno d’ Enza (Manzini & Savoia 2002) (i) De d’la a 'dorm i pu'te. in there cls sleeps the children Note also that it has been claimed for Italian and its dialects (e.g., Brandi & Cordin 1989) that ‘free inversion’ is possible, i.e., with all types of verbs and no restriction in definiteness. However, the information structure differs with the type of verb and definiteness of the inverted subject. This paper only considers the thetic VS constructions, not those with a contrastive or other narrow focus on the subject as in (ii). Standard Italian (Brandi & Cordin 1989) (ii)

Parlerá Mario, non Lucio. speak.fut Mario, neg Lucio ‘Mario will speak, not Lucio.’

5. The language Makhuwa (Guthrie’s P. 30) is spoken in the north of Mozambique by approximately 5 million people (Sebastian Floor, p.c.). All Makhuwa examples in this paper are from my own fieldwork conducted in 2005 and 2006 on the variant Enahara, which is spoken on and around Ilha de Moçambique.

 Jenneke van der Wal

b. DJ A-náá-ttónyá maátsi. 6-pres.dj-drip 6.water ‘There is water leaking out.’ Italian (Lambrecht 1994) (7) a.

E arrivato GIOVANNI. is arrived Giovanni ‘GIOVANNI arrived.’

b. Ha telefonato GIOVANNI. has called Giovanni ‘GIOVANNI called.’

2.3 Two types of agreement As can be seen in (4)–(7) the subject agreement on the verb varies in the languages examined. This is true even within the Romance and the Bantu language families. The languages can be divided into two types, according to their agreement in thetic sentences:6 (8) Type 1: the agreement on the verb is a default or dummy agreement; Type 2: the verb agrees with the postverbal subject.

The Romance language French and the Bantu language Sesotho are both of type 1. The verb in French cannot have the gender/number features of the subject, see (9a). The auxiliary agrees in number with the preverbal dummy element il, which is singular, and not with the postverbal logical subject trois femmes, which is plural. The past participle cannot be marked for either feminine or plural, but has to take its default masculine singular form; compare with the SV order in (9b) where the auxiliary and the past participle do agree in number and gender. French (Lambrecht 2000) (9) a.

Il est venu trois femmes. 3sg.masc is come.masc.sg three women ‘There came three women.’

. One of the reviewers raised the question of whether there exist languages which allow both types of agreement. So far I have not come across any language which has a choice in agreement. What must be kept in mind is that this paper considers VS constructions of the eventcentral type, excluding existential structures with the copulas ‘be’ or ‘have’, as partly described in endnote 2. According to Schoorlemmer (2007), in an existential construction with ‘have’ as the copula, the presented entity gets accusative case and can therefore not agree with the copula (unlike constructions with ‘be’ as the copula). These existential ‘have’ constructions and the thetic VS sentences discussed in this paper can behave differently with respect to agreement; and languages frequently have more than one way to express different thetic sentences. See for more information on functions and constructions of thetic sentences Sasse (1996).

Agreement in thetic VS sentences in Bantu and Romance 

b. Les femmes sont venues. the.pl women are come.fem.pl ‘The women have come.’ c.

*Il sont venues trois femmes. 3sg.masc are come.fem.pl three women int: ‘There came three women’.

In general, Bantu languages have a rich noun class system and rich morphology on the verb. One of the morphemes on the verb is the subject marker or prefix, which agrees in noun class with the subject in categorical sentences. This is exemplified for Sesotho (type 1) in (10a), where the subject bashanyana ‘boys’ is in class 2 and the subject marker on the verb agrees in class 2. In the thetic example in (10b) the verb takes a subject prefix in noun class 17, which is a locative class. There cannot be agreement with the postverbal logical subject, which is in noun class 10 (10c). Sesotho (Demuth 1990) (10) a. CJ Ba-shányáná bá-pálám-é li-pére. 2-boys 2-ride-perf 10-horse ‘The boys are riding horses.’ b. CJ Hó-fihl-ílé li-pé:re. 17-arrive-perf 10-horses ‘There arrived horses.’ c. CJ *Li-fihl-ílé li-pé:re. 10-arrive-perf 10-horses

Type 2 languages are also found in both the Romance and the Bantu language families. In contrast to the French examples, in Italian the auxiliary as well as the past participle agrees in number with the subject, in both categorical (11a) and thetic sentences (11b). Italian (Graziano Savá, p.c.) (11) a.

Tre ragazze sono arrivate. three girl.3pl be.3pl arrive.pp.3pl ‘Three girls have arrived.’

b. Sono (*è) arrivate tre ragazze. are (*is) arrive.pp.3pl three girl.3pl ‘There arrived three girls.’

In the Bantu language Makhuwa (type 2) the subject prefix on the verb agrees in noun class with the subject in categorical as well as thetic sentences. The preverbal (12a) or postverbal (12b) subject is in class 3, and so is the agreement marker on the verb (o‑).

 Jenneke van der Wal

Makhuwa (12) a. DJ Ntháná o-náá-ki-weréya. 3.back 3-pres.dj-1sg-hurt ‘My back HURTS.’ b. DJ O-náá-ki-weréyá nthána. 3-pres.dj-1sg-hurt 3.back ‘My BACK hurts.’

2.4 Conjoint-Disjoint distinction The two Bantu languages in types 1 and 2 (Sesotho and Makhuwa) exhibit another distinguishing property apart from the agreement facts. In order to describe this property, some basics on the grammatical system of some southern Bantu languages must first be introduced. The inflection of verbs in several languages, including Makhuwa and Sesotho, has pairs of conjugational categories (“tenses”) which refer to the same TAM semantics (Buell 2005), but differ in their “linkage” with what follows the verb. These verb forms are referred to as conjoint (cj) and disjoint (dj). They are segmentally marked by different TAM markers and indicate a difference in information structure. Their difference in phrase-final distribution is the most easily detectable characteristic. If the element directly following the verb is in a (focus) domain, the verb has its cj form; if this is not the case, or if the verb is phrase-final, the verb appears in its dj form (cf. Creissels 1996, Buell 2006). Both Makhuwa and Sesotho use a prefix with -a- to mark the dj form of the present tense (13a), (14a), which is absent in the cj form (13b), (14b). Makhuwa (13) a. DJ O-náá-thípa. 1-pres.dj-dig ‘She is digging.’ b. CJ O-n-thípá 1-pres.cj-dig ‘She digs a hole.’

nlittí. 5.hole

Sesotho (Demuth 1990) (14) a. DJ Ba-shányáná bá-á-li-fé:pa. 2-boys 2-pres.dj-10-feed ‘The boys are feeding them.’ b. CJ Ba-shányáná bá-fepá 2-boys 2-pres.cj.feed ‘The boys are feeding horses.’

li-pé:re. 10-horses

Agreement in thetic VS sentences in Bantu and Romance 

The two forms are also marked tonally in Makhuwa: a different tonal pattern is applied to the element directly following the cj verb form (15a) (cf. Stucky 1979), a process referred to as Predicative Lowering (PL; Schadeberg & Mucanheia 2000; van der Wal 2006b). This process consists of the loss of the first underlying H tone of the noun stem, and the possible addition of a boundary tone to indicate the right boundary of some prosodic phrase. More information on its use will follow in section 4.2. Thus the tonal pattern of ekuwo “cloth” is LHL in citation form (15), but LLH (immediately) after a cj verb form (15a). The element following a dj verb form has not undergone Predicative Lowering and its tonal pattern is as in citation form: LHL (15b).7 Makhuwa (15) citation: ekúwo ‘cloth’ (LHL) a. CJ Ki-n-kátthá ekuwó. 1sg-pres-wash 10.clothes ‘I wash clothes.’

(LLH)

b. DJ Ki-náá-káttha ekúwo. 1sg-pres.dj-wash 10.clothes ‘I wash clothes.’

(LHL)

In Sesotho the cj/dj distinction is not marked by a different tonal pattern on the object, but it is visible in the phonological phrasing. The right boundary of a phonological phrase in Sesotho is designated by the lengthening of the penultimate syllable.8 A verb in its cj form is phrased together with the following element, as can be seen in the penultimate lengthening which is only present on the object (16a) (Demuth 1990). The dj verb form is in a phonological phrase by itself, as is the postverbal element: both verb and object have penultimate lengthening (16b). The interpretation also differs, as indicated by the translation. Sesotho (Demuth 1990, adapted) (16) a. CJ Ba-shányáná bá-fepá li-pé:re|. 2-boys 2.pres-feed 10-horses ‘The boys are feeding horses.’

. The difference in interpretation between the cj and dj verb form with an object is very hard to define. When asked to give a context, informants immediately came up with a context for the cj form, namely as an answer to ‘what did you pound?’. A context for the dj form was very hard or impossible to find, even though it was easy enough to produce the dj form. If any explanation was given, it was that the dj sentence is uttered as a way of giving plain information or assertion of the proposition. More elaborate data are needed for further research. . Another indication for the boundaries of phonological phrases in Sesotho is a tonal change, which is described in Demuth (1990).

 Jenneke van der Wal

b. DJ Ba-shányáná bá-á-li-fé:pa| li-pé:re|. 2-boys 2-pres.dj-10-feed 10-horses ‘The boys are feeding them, the horses.’

The behaviour of cj/dj verb forms is rather similar for the two languages in SVO sentences, but differs in thetic constructions: Sesotho uses a cj form and phrases the postverbal logical subject with the verb (17), whereas Makhuwa uses a dj form, and the tonal pattern of the postverbal subject has not undergone Predicative Lowering (18). Sesotho (Demuth 1990, adapted) (17) CJ Hó-fihl-ílé li-pé:re|. 17-arrive-perf 10-horses ‘There arrived horses.’ Makhuwa (18) DJ O-hoó-khwá mwanámwáne (PL: *mwanamwáne). 1-perf.dj-die 1.child ‘There died a child.’ (as an announcement on the news)

The difference in the use of the cj or dj verb form and the phonological marking related to it correlates with the two types of agreement on the verb. The question is how these phenomena are related, and how to account for blocking the agreement with the logical subject in type 1 and/or allowing the agreement in type 2 languages. 3. Possible analyses So far it has been shown that the Romance language family as well as the Bantu language family distinguishes two types of agreement in thetic VS sentences. In some Bantu languages this difference is also visible in the relation between the verb and the postverbal logical subject, as observed in the use of a conjoint or disjoint verb form and the corresponding phonological phrasing. In languages of type 1 the verb does not agree with the postverbal logical subject, but instead has default agreement. The Sesotho verb appears in its conjoint form and the logical subject is phrased with the verb. In type 2 languages the verb agrees with the postverbal subject and in Makhuwa the verb has a disjoint form and prosody. In order to consider possible analyses of these phenomena, I first examine the position of the logical subject. The subject in a categorical sentence is assumed to be in specTP, or even specTopP. This preverbal position is associated with identifiable referents (Lambrecht 2000), which have a topic status. Since the

Agreement in thetic VS sentences in Bantu and Romance 

subject in a thetic sentence lacks this property, the specTP (let alone specTopP) position is pragmatically not appropriate for the inverted subject. There is no motivation in information structure for the subject to move from its original position. One word order argument for Bantu can be found in the position of the verb. I assume, following (among others) Myers (1990), Julien (2002), Kinyalolo (2003, cited in Carstens 2005), and Buell (2005), that the verb in Bantu starts out as a root and only moves in the lower part of the VP to incorporate the derivational extensions. It then terminates in a functional projection lower than T. It does not headmove further to T, and the inflectional prefixes on the verb represent functional heads spelled out in their base positions. (19) [tp [T [inflection …[fp Vi[vp S [ti O]]]]]]

If the subject were moved to specTP, the verb would need to move to a position even higher than that to derive a Verb-Subject order. It is unclear to which position this movement would be (indicated by XP in (20)) and what would trigger the movement. Furthermore, because the verb is not a complex head in T, but contains several in-situ functional heads and a stem, it cannot head-move to the position above specTP. The movement of the verb would thus have to be remnant movement. However, the part moved will be the part under the T' node (20), which is not a maximal projection.9 There are thus several arguments against an analysis with the subject in specTP in a VS construction. (20) [xp [t’ [V]]i [X [tp S ti]]]

Indeed, on the basis of word order, agreement and binding properties it has been argued for various languages that the inverted subject is in a VP-internal position (Bobaljik & Jonas 1996; Déprez 1990; Belletti 1999; Samek-Lodovici 1996; Alexiadou & Anagnostopoulou 1999, 2001; Demuth & Harford 1999, among many others). Having established the position of the subject of a thetic VS sentence as lower than T, there are two general ways to analyze the different agreement patterns in thetic VS sentences.10 The first contains the analyses of Carstens (2005) and Collins (2004) on the one hand and Baker (2008) on the other hand. These analyses attribute the agreement patterns to a different parameter setting in terms of features (phi, EPP) or direction of agreement (upward/downward). The second approach proposes that the difference is connected to the status of

.

Thanks to Lisa Cheng for pointing this out to me.

. This could be specVP or specvP, depending on whether the verb is unaccusative or unergative. Belletti 1999 introduces a low FocP for the inverted subject.

 Jenneke van der Wal

the agreement marker, which can be pronominal or grammatical (cf. Bresnan & Mchombo 1987). 3.1 Agree/uφ requires Move/EPP Collins (2004) and Carstens (2005) start out with the agreement properties in Chomsky’s (2001) Minimalist framework. In this framework a head (prototypically T) with unvalued, uninterpretable features (henceforth uF) probes down in the structure. It searches and finds the closest c-commanded element that matches these features: the goal. When the search is successful, probe and goal agree, which means that the uF of the probe are valued. The goal is claimed to also have uF (case, wh) which can also be valued in the Agree relation. If, in addition, the probe has an EPP feature, the agreed-with goal will be re-merged (moved) in the specifier of the probing head. According to Chomsky (2001) all movement is dependent on an Agree relation. Not all Agree relations, however, are dependent on movement, as is apparent in English expletive constructions (21). The verb/T here agrees with the goal (the subject) without moving it to its specifier, that position being filled by the expletive “there”. (21) There is a problem in the White House. There are problems in the White House.

Collins (2004) argues that for Bantu languages it is impossible to treat EPP and Agree separately from each other (unlike in English). He thus proposes the Agreement parameter for Bantu languages, where Agree requires internal Merge (Move). (22) Agreement Parameter (minimally adapted)11: Let Agree (X, YP), where X contains the probe (uninterpretable phi-features) and YP contains the goal, then X has an EPP feature that is satisfied by YP

Carstens (2005: 266) formulates the dependency of Agree and Move as the Featurelinking parameter: uninterpretable phi-features (henceforth uφ) have EPP as a subfeature in Bantu. Both authors thus claim that agreement in Bantu languages is so closely tied to movement that it is impossible for a head with uF to agree with an element that stays lower in the structure. Carstens supports her claim with data on three constructions in Bantu languages where agreement takes place. Firstly, . Instead of EPP, Collins uses the abbreviation OCC here, which stands for ‘occurrence’. He argues that the term is more or less the minimalist equivalent to EPP, but better since it is more general than EPP (which for him only applies to specIP). Since so far no linguist has been able to explain the content of the feature, I find its naming uninteresting, and since EPP is more commonly used in minimalist work I will use the abbreviation EPP in this paper.

Agreement in thetic VS sentences in Bantu and Romance 

there is dummy agreement in the VS structures, as exemplified above. Secondly, there is agreement with the complementizer. Carstens (2005) argues that when C has uφ features, it also has EPP. So when C probes down to search for a whelement, for example, it will agree with the wh-element and consequently move it to specCP, as can be seen in (23b). The verb in (23b) does not agree with the logical subject in class 2, as it does in (23a), but rather with the question word biki, which is in class 8.12 Kilega (Carstens 2005) (23) a.

Bábo bíkulu b-á-kás-ílé mwámí bikí mu-mwílo? 2.dem 2.women 2-T-give-perf 1.chief 8.what 18-3.village ‘What did those women give the chief in the village?’

b. Bikí bi-á-kás-ílé bábo bíkulu mwám mu-mwílo? 8.what 8-T-give-perf 2.dem 2.women 1.chief 18-3.village ‘What did those women give the chief in the village?’

Thirdly, there is agreement with preposed locatives. In (24a) the verb agrees with the subject, which accordingly moves to the specifier position in Carstens’/Collins’ theory. However, in the case where it is not the logical subject but another element which occupies the preverbal position (presumably for reasons of information structure), it is this other element with which the verb agrees.13 In (24b) the locative is preposed and the verbal agreement is with the locative (class 17) and not with the logical subject (class 2). Chichewa (Bresnan & Kanerva 1989) (24) a.

A-lendô-wo a-na-bwérá ku-mu-dzi. 2-visitors-2.dem 2-rec.past-come 17-3-village ‘Those visitors came to the village’.

b. Ku-mu-dzi ku-na-bwérá a-lendô-wo. 17–3-village 17-rec.past-come 2-visitors-2.dem ‘To the village came those visitors’.

These tests show different results when applied to the Bantu language Makhuwa. As is clear from examples in the previous section, Makhuwa allows agreement with the verb in thetic VS constructions, where the subject does not move to specTP. Carstens’s second example concerning complementizer agreement cannot be properly compared to Makhuwa, since the language uses a cleft with relative .

See Sabel & Zeller (2005) for an account of in-situ versus ex-situ wh-questions in Nguni.

. I have been unable to find a motivation for the type 1 VS construction in Carstens (2005) or Collins (2004).

 Jenneke van der Wal

verb form when preposing wh words. The third argument, however, does have a counterpart in Makhuwa. When preposing a locative (25b, class 16), the agreement on the verb is still with the subject (class 2), which is now in a postverbal position. Agreement in Makhuwa can thus be applied without moving the goal to the probe’s specifier. Makhuwa (25) a.

Aléttó a-náá-phíyá wakisírwa. 2.guests 2-pres.dj-arrive 16.island ‘The guests arrive on the island’.

b. Wakisírwá a-náá-phíyá alétto / *wa-náá-phíyá. 16.island 2-pres.dj-arrive 2.guests / 16-pres.dj-arrive ‘On the island arrive guests’.

From these Makhuwa examples it can be concluded that the relation between uφ and EPP, formulated as the Feature-linking parameter (Carstens 2005), or the Agreement Parameter (Collins 2004) does not hold for all Bantu languages. Since “the Bantu languages” are a heterogeneous group of some 550 languages (cf. Nurse & Philippson 2003), it is probably more appropriate to propose parameters for individual languages, or patterns in languages. The Agreement Parameter (Collins 2004) or the dependency of Agree and EPP (Carstens 2005) could then apply to the Bantu languages of type 1 only. A separate account should be given for the agreement pattern in the languages of type 2. If in these languages the verb can indeed agree with the subject without moving it to the specifier position of TP, is there anything else correlating with this agreement? Carstens (2005) proposes that whereas in Bantu languages (type 1) uφ have an EPP feature, in Indo-European languages uφ have Case as a subfeature. The EPP is believed to be separate from the agreement relation in IndoEuropean (IE) and case-checking is believed to be separate in Bantu. If the relation Agree-EPP holds not for “Bantu”, but for languages of type 1 (as I have just suggested), the relation Agree-Case might in a parallel manner hold for languages of type 2 (and not just IE). An argument for the absence of an Agree-Case relation in type 1 is found in the multiple subject agreement on complex tenses or auxiliary structures in Bantu, as in (26). The subject mikoko “sheep” (class 4) triggers noun class agreement on ‘to be’ and two aspectual phrases, which is impossible in a language like English, which has agreement only on the auxiliary (27). Kilega (Kinyalolo 1991) (26) Mikoko z-á-bézágá zí-se z-á-sínz-ua. 4.sheep 4-T-be 4-about.to 4-T-slaughter-pass ‘Sheep were about to be slaughtered’.

Agreement in thetic VS sentences in Bantu and Romance 

(27) He is crying. *He is crie-s/crying-s.

If Case were connected to Agree in Kilega as in English, the Case features would be deleted in the first instance of agreement. The subject would no longer be active and would thus be invisible for a probing head. The fact that it can agree more than once is taken as an argument to claim that the subject’s case features are syntactically active until the next strong phase (Carstens 2003). Since agreement is more strongly connected to Case-deletion in IE than in Bantu/type 1, Case in IE is immediately deleted at Agree and hence no multiple agreement can take place. If Carstens’ Agree-Case relation held for type 2 languages, Makhuwa (as a type 2 language) should behave identically to IE and have Case deletion at Agree. Thus it should not allow multiple agreement in complex tenses. As can be seen in (28), Makhuwa does in fact have multiple agreement in complex tenses. This demonstrates that Agreement in Makhuwa (type 2) is not related to Case either. See Pires-Prata (1960) for more information on Makhuwa complex tenses. Makhuwa (28) Ki-háana ki-khum-áka. 1sg-have 1sg-go.out-dur ‘I have to go out.’ (29) Álé aa-khálá aa-vélávela. 2.dem 2.perf-stay 2.perf-be.stuck ‘They were trapped.’

In summary, the Agreement Parameter (Collins 2004) or the relation Agree-EPP (Carstens 2005) works for languages of type 1, but the relation Agree-Case cannot be generalized to all type 2 languages. This approach thus does not provide motivation for the agreement pattern observed in type 2 languages, and although it does present a technical way to analyze the agreement in type 1, it does not offer us insight into the deeper motivation behind the surface agreement patterns. In the next section Baker’s (2008) approach is considered, which also takes into account agreement in languages like Makhuwa which do not connect agreement to either Case or EPP. 3.2 Direction of agreement vs. case dependency Baker (2003) has related agreement to EPP and Case as well. He claims that in IE languages uφ is a property of Case features, while in Bantu it is a subfeature of EPP. Baker (2008) captures these facts in two parameters on agreement on a functional head F.

 Jenneke van der Wal

(30) The Direction of Agreement Parameter F agrees with DP/NP only if DP/NP asymmetrically c-commands F. (31) The Case-Dependency of Agreement Parameter F agrees with DP/NP only if F values the Case feature of DP/NP or vice versa.

According to Baker, each language (family) is set to “yes” or “no” for each parameter. For Bantu languages (though not all of them) the parameters are set “yesno”: agreement is only sensitive to the position of the element it agrees with; it is independent of Case. In Indo-European languages agreement is case-sensitive, regardless of the direction of agreement (“no-yes”). In other words: Bantu agrees with whichever element is in the specifier of the probe, whereas IE agrees with whichever element it checks the case of. The different agreement patterns in thetic VS sentences may be explained by these different parameter settings. The reason that Baker introduces these parameters is that they are needed to restrict the possibilities of agreement after he has extended Chomsky’s principles of syntactic agreement. The first principle is that the probe must c-command the goal (Chomsky 2000, 2001). This is a constraint on the direction in which a probing head searches: it can only probe down. Baker (2008) claims that a head F can agree with its goal XP not only if F c-commands XP, but also if XP c-commands F. This claim leads to considerable freedom with respect to probing: a head can probe upward as well as downward. Since languages generally do not allow optional agreement with the one or the other probe (upward or downward), there must be restrictions. These are the parameters Baker (2008) proposes. When two parameters (Direction and Case) can be set for two values (yes and no), the result are four different types of languages. Baker (2008) indeed finds example languages of these four types in his 100 language sample from the World Atlas of Language Structures (WALS). With the data from the previous section the conclusion can be drawn that the settings for Makhuwa are “no-no”: agreement in Makhuwa is not dependent on movement or a spec-head relation, as can be seen in structures where the verb/T agrees with the postverbal subject, such as thetic sentences (6), and sentences with preposed locatives (25). Neither is agreement dependent on Case, because Makhuwa can have multiple subject agreement in complex tenses (28). The “no-no” languages pose a problem to Baker’s theory, because in this theory agreement in these languages is not restricted by any principle or parameter. Agreement could thus randomly be upward or downward. For example, in the case of a preposed locative the verb could agree with the subject or with the preposed locative (locative inversion). This is not what we find, because agreement in Makhuwa is regular and predictable. Baker admits in a footnote that his description leaves open how the grammar determines which element T agrees with.

Agreement in thetic VS sentences in Bantu and Romance 

In conclusion, the approaches which attribute the different agreement patterns of types 1 and 2 to the relations in the agreement mechanism do not offer a satisfactory explanation for the agreement in type 2, which is sensitive neither to EPP/direction nor to Case. If the Agree mechanism is unable to shed light on the difference, there must be a more fundamental distinction which drives the difference in agreement. 4. Pronominal vs. grammatical agreement marker This section presents an alternative analysis of the different agreement patterns in thetic VS sentences. It proposes a difference in the status of the agreement marker of type 1 and type 2: type 1 languages have pronominal agreement, whereas in languages of type 2 the agreement is purely grammatical. 4.1 Sesotho (type 1) pronominal agreement The first hint at the pronominal nature of the agreement in type 1 is the agreement in French thetic VS sentences, where the auxiliary agrees in number with the pronoun il. French (Lambrecht 2000) (32) Il est venu trois femmes. 3sg.masc is come.masc.sg three women ‘There came three women’.

Such a free pronoun is not used in most Bantu languages, since these are pro-drop languages. Instead, the subject agreement marker can function as a pronoun or as agreement only. Bresnan and Mchombo (1987) claim that the subject prefix can ambiguously be used for grammatical and anaphoric agreement (at least in Chichewa). In Bresnan and Mchombo (1987) grammatical agreement is defined as “a verbal affix redundantly expressing the person/number/class of an NP bearing an argument relation to the verb”. In anaphoric/pronominal agreement the agreement prefix is an incorporated pronominal argument of the verb, and possible coreferential NPs are non-arguments. Pronominal agreement, as used in this paper, is equivalent to an analysis in which the agreement prefix agrees with an empty category pro. Proposing that the agreement marker ho- in Sesotho thetics is a pronoun has certain implications, each of which will be discussed below. Firstly, a pronoun can have Case. Bantu languages do not mark case overtly, but if, in Sesotho thetics, the default agreement marker is a pronoun, it has nominative case (as in French il),

 Jenneke van der Wal

which enables the verb to mark the logical subject as its complement. At this point Lambrecht’s (2000) principle of subject-object neutralization (section 2.1) is of importance: in Sesotho the postverbal subject in a thetic sentence indeed has certain properties associated with the focal object in a categorical sentence. As shown in (17), repeated here as (33), the postverbal logical subject is in one phonological phrase with the verb, indicated by the penultimate lengthening on the logical subject only. The verb does not have penultimate lengthening and is thus not phrased separately. Sesotho (Demuth 1990, adapted) (33) CJ Hó-fihl-ílé li-pé:re|. 17-arrive-perf 10-horses ‘There arrived horses.’

A second implication of analyzing the agreement marker as a pronoun is that it cannot be coreferential with the postverbal logical subject. It cannot agree with this subject, since the two are in the same clause and agreement would lead to a violation of principle B. According to this principle in standard Binding Theory pronouns must be free in their local clause, which is not the case in (34) where the pronoun and the full noun are both in the same phrase. As a result the example is ungrammatical. Sesotho (34) CJ *Li-fihl-ílé li-pé:re|. 10-arrive-perf 10-horses int: ‘There arrived horses.’

In order to avoid such a violation the language could either dislocate the logical subject NP, so that the pronoun and the full NP are no longer in the same phrase, or avoid agreement with the postverbal logical subject. The first strategy could not be applied, since it would render a different reading: a dislocated NP is interpreted as a topic. In Northern Sotho (closely related to Southern Sotho/Sesotho) the subject can be right-dislocated with an afterthought reading, as indicated by the translation in (35). The verb accordingly takes its disjoint form and the penultimate lengthening shows that the verb and the subject are in two different phonological phrases. Northern Sotho (Zerbian 2006) (35) DJ Ó-a-só:ma| mo:-nna|. 1-pres.dj-work 1-man ‘He is working, the man.’

Since in a thetic sentence the subject must be detopicalized (see section 1), dislocation is no alternative to the principle B violation. Instead, as already mentioned,

Agreement in thetic VS sentences in Bantu and Romance 

the agreement on the verb is in a different noun class than the postverbal logical subject. This class 17 marker can be considered default agreement or a locative pronoun. The last option finds an argument in the comparison with locative inversion, which is used to express thetic sentences in the Bantu language Chichewa (Bresnan & Kanerva 1989), for example. In such a construction a locative element is preposed, the logical subject stays in a postverbal position and the verb agrees with the preverbal locative. Chichewa (Bresnan & Kanerva 1989) (36) Ku-mu-dzi ku-na-bwérá a-lendô-wo. 17-3-village 17-past-come 2-visitors-dem ‘To the village came those visitors.’ (37) M-mi-têngo mw-a-khal-a a-nyani. 18-4-tree 18-perf-sit 2-baboons ‘In the trees are sitting baboons.’

Demuth (1990) shows that “locative inversion” in Sesotho, at least synchronically, is in fact an impersonal presentational construction with the optional addition of a locative adverb in preverbal position.14 In Chichewa the verbal agreement with the locative can be analyzed as grammatical agreement, but since in Sesotho the preposed locative is not required, there is no argument-bearing NP to agree with and hence the agreement cannot possibly be grammatical agreement. Thus there is direct and indirect evidence for an analysis of the Sesotho agreement marker as a pronoun in thetic sentences and evidence against an analysis as a purely grammatical agreement marker. 4.2 Makhuwa (type 2) grammatical agreement In thetic VS sentences the verbal agreement marker in Makhuwa is argued to be grammatical instead of pronominal. Whereas a pronominal agreement marker can have case, as in type 1 languages like Sesotho, a grammatical agreement marker cannot. The verb in languages of type 2 can thus only check nominative case on the postverbal subject, not on the pronominal subject marker. There is proof that . Note also that the agreement in Sesotho thetic sentences is in the locative class 17, which has become the default in this language. Other languages may have a different default agreement, as for example Kirundi, which takes a class 16 agreement marker in an inverse construction with focus on the subject. Kirundi (Ndayiragije 1999) (i)

Ha-á-nyôye amatá abâna. 16-past-drink.perf milk children ‘Children (not parents) drank milk’.

 Jenneke van der Wal

the postverbal subject in type 2 indeed has nominative case, as can be seen in (38). In Latin the verb agrees with the pre- or postverbal subject and the subject has nominative case in the thetic SV (38a) as well as the categorical VS sentence (38b). Latin (Lambrecht 2000: 635) (38) a.

Taurus mugit. bull.nom.sg bellow.3sg.pres ‘The bull is bellowing.’

b. Mugit taurus. bellow.3sg.pres bull.nom.sg ‘There is a bull bellowing.’

Since Latin is a dead language, we cannot be completely sure of the thetic interpretation, but Rodríguez-Mondoñedo (2005) shows for Spanish (type 2) that the subject in VS event-central thetics can at least not be accusative. In a so-called split structure (39a; see also endnote 2) the presented entity, the object, can be replaced by an accusative clitic (39b), but this is impossible in an inverted subject construction (40). Thus the postverbal subject in type 2 is not object-like, as is the case with logical subjects in languages of type 1. Spanish (Rodríguez-Mondoñedo 2005) (39) a. Había unos niños en el parque. have some kids in the park ‘there were some kids in the park’ b. Los había. masc.pl.acc have (40) a.

Llovieron insultos sobre el árbitro. rained.pl insults over the referee ‘The referee was pelted with insults.’

b. *Los llovieron. cl.acc rained.pl

A second implication mentioned for type 1 languages was the impossibility of coreference between the pronominal agreement marker and the postverbal logical subject, consistent with principle B. As I propose to analyze the agreement marker in type 2 as purely grammatical agreement, this principle would not be violated, making coreference between the agreement marker and the postverbal subject possible. If the agreement marker and the subject are thus allowed to occur within the same phrase in type 2, the conjoint verb form may be expected to occur in the Bantu thetic sentences. However, in thetic sentences in Makhuwa the disjoint form is used. There are general and language-specific reasons why the cj verb form is not used in Makhuwa thetic sentences.

Agreement in thetic VS sentences in Bantu and Romance 

The first (general) reason is that the use of the cj form requires a certain relation between the verb and the following element. Apparently it is not sufficient for the verb to simply have an element in a postverbal position, but this element needs to be in a specific domain or have specific characteristics to allow the verb to appear in its cj verb form. There are two possible hypotheses as to why the relation between the verb and the postverbal subject in a type 2 language does not satisfy the requirements for the use of a cj form. Firstly, there is some evidence that the use of the cj form requires the presence of an element in a certain domain (Buell 2006) or in the immediate postverbal position (Van der Wal 2006a). In a VS construction the postverbal subject indeed follows the verb, but (according to this hypothesis) it would be in a position other than immediately postverbal, not in the right domain. Secondly, the element following a cj verb form gets an exclusive interpretation in Makhuwa. Since the postverbal thetic subject is to be interpreted only as nontopical, and not as exclusive, there is a clash in interpretation if the cj form would be used. Another reason why the conjoint verb form cannot be used in thetics in Makhuwa specifically is that it would render a different syntactic construction. If the conjoint form were used in a VS construction, it would be interpreted as a pseudocleft. The “conjoint” verb form is translated as a headless relative and the postverbal logical subject as a nominal predicate. I will first explain the form of the relative verb and then give more information on nominal predication in Makhuwa in order to see how the interpretation as a pseudocleft comes about. In Makhuwa relative clauses the cj/dj distinction is inapplicable, but the relative verb is formally equal to the cj verb form (compare (41b) and (41c)) (see Katupha 1983, van der Wal in preparation). No special relative morphology is used. As illustrated in the subject relative in (42b,c), a headless relative is formed by omitting the head noun. Makhuwa (41) a. DJ Nlópwáná oo-thípa. 1.man 1.perf.dj-dig ‘The man dug.’ b. CJ Nlópwáná o-thip-alé nlittí. 1.man 1-dig-perf 5.hole ‘The man dug a hole.’ c. REL Nlópwáná o-thip-alé. 1.man 1-dig-perf ‘The man who dug.’

 Jenneke van der Wal

(42) a. DJ Mwanámwáné o-hoó-khwa. 1.child 1-perf.dj-die ‘A/the child died.’ b. REL Mwanámwáné o-khwa-alé o-rí owáani. 1.child 1-die-perf 1-be 17.home ‘The child who died is at home.’ c. REL O-khwa-alé o-rí owáani. 1-die-perf 1-be 17.home ‘The one who died is at home.’

The tonal process called Predicative Lowering, as also explained in section 2.4, is applied to the object after a conjoint form. However, it is also used to change a noun into a nominal predicate, hence the name (43; van der Wal 2006b). (43) mwanámwáne ‘child’ (LHHL) mwanamwáne ‘it is a child’ (LLHL)

Considering these properties of relativization and predication in Makhuwa, the combination of a verb which resembles a conjoint form and a following (tonally lowered) subject is interpreted as a pseudocleft, and not as a thetic sentence. This is confirmed by the context given for the phrases in (44). Whereas (44a) is used as an announcement, out of the blue, (44b) is used as an answer to the question “who is it that died?”( i.e., we already know that someone died). (44) a. DJ O-hoó-khwá mwanámwáne. 1-perf.dj-die 1.child ‘There died a child.’ b. “CJ” O-khw-aalé mwanamwáne. 1-die-perf 1.child ‘The one who died is a/the child.’

The Makhuwa disjoint VS structure could alternatively be analyzed as pronominal agreement with a right-dislocated subject, and a comparison with the Northern Sotho example in (35) is easily made. However, the environments in which the Makhuwa (disjoint) VS sentences are used (first lines of stories, out of the blue) argue against this. Example (45) is the first line of a story, clearly thetic, and a VS construction is used. (45) Nihúkú ni-motsa ohíyú w-aanú-mwááryá mweéri. 5.day 5-one 14.night 3-rem.pers-shine 3.moon ‘One night the moon was shining.’

We hardly find any right dislocation in the environments where dislocation of the subject would seem most appropriate. Instead, demonstratives are used.

Agreement in thetic VS sentences in Bantu and Romance 

For example, at the end of an anecdote about a child: the child is playing, then he is run over by a car, he is lying there, he is taken to hospital and he died, that child. The Makhuwa story would end as in (46). (46) Ni ólé mwanámwán’ oolé o-hoó-khwa. and 1.dem 1.child 1.dem 1-perf.dj-die ‘And (then) that child died.’

The few attestations of subject right dislocation found in thirteen Makhuwa stories also use demonstratives, either as a free pronoun (47) or to modify the right-dislocated noun (48). (47) Ophéélá o-´m-pwésh-ak-átsá óle, namárókolo. 15.want 15-1-hit-dur-pl 1.dem, 1a.hare ‘Wanting to hit him, Hare.’ (48) Álé aa-pácérá w-ií-hímya-ká-tsá akúnyá ale. 2.dem 2.perf.dj-begin 15-refl-say-dur-pl 2.white 2.dem ‘They began to identify themselves, those Portuguese.’

An even more convincing argument against an analysis of right-dislocation is the use of a VS construction with an indefinite subject. From the Romance examples in (49) it is clear that indefinites cannot be right-dislocated.15 Since indefinites do occur in Makhuwa VS constructions, it follows that these postverbal subjects cannot be dislocated (51). Italian (Linda Badan, p.c.) (49) *Non è venuto, un ragazzo. neg is come.pp, a boy int: ‘He didn’t come, a boy.’ French (Larsson 1979) (50) *Je crois qu’ il vient d’ entrer par la porte, un monsieur. 1sg believe.1sg that he come.3sg of enter through the door, a gentleman int: ‘I believe that he has (just) entered through the door, a man.’

. Larsson (1979: 22): “Une des conséquences de ces conditions contextuelles est l’impossibilité d’avoir une phrase nominale indéfinie à référence spécifique dans la position disloquée” (one of the consequences of these contextual conditions is the impossibility of having an indefinite specific phrase in a dislocated position). Delais-Roussarie, Doetjes & Sleeman (2004: 522) mention that indefinites can in fact be right-dislocated, but only when they have a generic interpretation, or when they are textually/situationally evoked or inferred. This makes their status in information structure more topic-like (Erteschik-Shir 2007), which is an interpretation avoided in thetic sentences.

 Jenneke van der Wal

Makhuwa (51) a. Ohoó-wá ńtthu. 1.perf.dj-come 1.person ‘There came someone.’ b. Kha-wa-álé ńtthu. neg.1-come-perf 1.person ‘No-one came.’

Thus there is clear evidence that the Makhuwa VS construction expresses theticity, and that the construction cannot be analyzed as right dislocation. Yet another alternative would be to use the same structure as Sotho, with a conjoint verb form and dummy/locative agreement. It seems, however, that locative nouns in Makhuwa generally have the status of adverb instead of argument. It may be for this reason that locatives cannot function as pronouns in VS constructions. The locative cannot trigger agreement in locative inversion either (52). That the locative is not a true object is shown in (53), where subjectivization of the locative is impossible.16 (52) a.

Aléttó a-náá-phíyá wakisírwa. 2.guests 2-pres.dj-arrive 16.island ‘The guests arrive on the island.’

b. Wakisírwá a-náá-phíyá alétto. 16.island 2-pres.dj-arrive 2.guests ‘On the island arrive guests.’ c. *Wakisírwá wa-náá-phíyá alétto. 16.island 16-pres.dj-arrive 2.guests int: ‘On the island arrive guests.’ (53) a.

.

Ki-núú-hélá ephepélé mpoótíli-ni. 1sg-perf.pers-put 9.fly 18.jar-loc ‘I put the fly in the jar.’

Locatives can in fact be arguments in Makhuwa, as subjects in an impersonal passive.

Makhuwa (i)

Oparásá woo-cáw-él-íya. 17.fortress 17.perf.dj-run-appl-pass

(ii)

Mparásá moo-cáw-él-íya. 18.fortress 18.perf.dj-run-appl-pass ‘The fortress was run to.’

Agreement in thetic VS sentences in Bantu and Romance 

b. Ephepélé e-núú-hél-íyá mpoótíli-ni. 9.fly 9-perf.pers-put-pass 18.jar-loc ‘The fly was put in the jar.’ c. *Mpoótílí-ní n-núú-hél-íyá ephepéle. 18.jar-loc 18-perf.pers-put-pass 9.fly int: ‘In the jar was put a fly.’

Another indication for the non-argument status of the locative in Makhuwa is the use of Predicative Lowering (see also sections 2.4 and 4.2). Objects following a cj verb form undergo this tonal change (54a), whereas temporal (54b), manner (54c) and locative (54d,e) adverbs do not. (54) a.

Ki-m-wéhá muupattétthe. 1sg-pres-see 3.beehive ‘I see a/the beehive.’

(non-PL: *muúpáttétthe)

b. Ki-n-thítá

nańnáanová. 1sg-pres-pound right.now ‘I am pounding right now.’

c.

O-phwany-alé woowáákhuvéya. 1-find-perf quickly ‘She found (it) quickly.’

(PL: *woowaakhuvéya)

d. Ki-n-rúpá wa-khaámá-ni. 1sg-pres-sleep 16-bed-loc ‘I sleep in a bed.’

(PL: *wakhaamaní)

e.

(PL: #nannaanová)

Ki-n-rówá o-pharéya. 1sg-pres-go 17-beach ‘I go to the beach.’

(PL: *ophareyá)

In summary, the evidence regarding Case and coreference under principle B in thetic VS sentences suggests an analysis of the verbal subject agreement marker as grammatical agreement in languages of type 2. Furthermore, the interpretation as a pseudocleft and the limited use of right-dislocation suggest that an analysis of the subject marker in type 2 as pronominal is incorrect. Finally, a locative agreement marker as in Sesotho (type 1) is not an alternative either in Makhuwa (type 2). 4.3 Remarks on an analysis of agreement in Italian dialects Manzini & Savoia (2002) observe the same two types of languages (agreeing and non-agreeing) in the Southern and Northern Italian dialects. In their account of the different agreement in VS constructions, they propose that there is a set of

 Jenneke van der Wal

nominal inflections in addition to T (a head D, with the features Number, Person and predicative N). They propose furthermore that these features (always) attract an aspectual feature. This aspectual feature is already associated with the logical subject and its features, when it is attracted by T and then by D. The difference between the two language types is a lexicalisation choice in the agreement between T, D and this aspectual feature: in the typical Northern dialects (my type 1) only the predicative N-feature is shared; in the typical Southern dialects and Standard Italian (my type 2) all features are shared (Number, Person, Gender). This technical account does not seem completely incompatible with the proposal put forward in this paper, but it does not take into consideration the status of the postverbal logical subject. It remains unclear whether there are differences in the subject in the two types of languages/dialects, in terms of case or (phonological or syntactic) phrasing. However, a further examination of these dialects in the light of the language types 1 and 2 looks interesting. 5. Conclusion I have shown above that thetic sentences can be expressed by a construction using a Verb-Subject order in both Romance and Bantu languages. The Bantu and Romance language families thus show a similarity in expressing theticity. Moreover, both language families show two different types of agreement on the verb, which can be either default (type 1) or agreement with the postverbal subject (type 2). Accounts by Collins (2004), Carstens (2005) and Baker (2008) have attributed this difference to the mechanism of agreement. Their parameters on agreement, however, do not provide a motivation for all the differences in agreement observed in thetic VS sentences. Instead, this paper argues that there is a fundamental difference in the system, which is a difference in the status of the agreement marker. In languages of type 1 this agreement marker is pronominal, while it is purely grammatical in type 2 languages. Arguments for this analysis are found in Case, Binding Theory and information structure. More research needs to be conducted on the status of the agreement in Romance and Bantu languages to better understand the motivation for these different agreement patterns. In general the debate on clitics and affixes and their status as pronouns or agreement markers deserves more attention, while the typology of agreement would benefit from more new data as well as comparison with other languages and language families. It is important in this research to always take into account the information structure in the sentence, by looking at the context and the prosody, for example, in addition to the syntactic facts and word order.

Agreement in thetic VS sentences in Bantu and Romance 

References Alexiadou, A. & Anagnostopoulou, E. 1999. EPP without Spec, IP. In Specifiers: Minimalist Approaches, D. Adger, S. Pintzuk, B. Plunkett & G. Tsoulas (Eds), 93–109. Oxford: OUP. Alexiadou, A. & Anagnostopoulou, E. 2001. The subject-in-situ generalization and the role of case in driving computations. Linguistic Inquiry 32: 193–231. Ameka, F.K. To appear. Information packaging in Kwa: Micro-variation and typology. In Studies in Kwa comparative syntax, E. Aboh & J. Essegbey (Eds). Baker, M. 2003. Agreement, dislocation and partial configurationality. In Formal Approaches to Function in Grammar, A. Carnie, H. Harley, E. Jelinek & M. Willie (Eds), 107–132. Amsterdam: John Benjamins. Baker, M. 2008. The Syntax of Agreement and Concord. Cambridge: CUP. Belletti, A. 1999. ‘Inversion’ as focalization and related questions. Catalan Working Papers in Linguistics 7: 9–45. Bobaljik, J. & Jonas, D. 1996. Subject positions and the roles of TP. Linguistic Inquiry 27: 195–236. Brandi, L. & Cordin, P. 1989. Subject clitics in two Italian dialects: A Perspective on the null subject parameter. Journal of Italian Linguistics 8: 1–24. Bresnan, J. & Kanerva, J. 1989. Locative inversion in Chichewa: A case study of factorization in grammar. Linguistic Inquiry 20: 1–50. Bresnan, J. & Mchombo, S. 1987. Topic, pronoun and agreement in Chichewa. Language 63: 741–782. Buell, L.C. 2005. Issues in Zulu Verbal Morphosyntax. Ph.D. dissertation, UCLA. Buell, L.C. 2006. The Zulu conjoint/disjoint verb alternation: Focus or constituency? ZASPiL 43: 9–30. Carstens, V. 2003. Rethinking complementizer deletion: Agree with a case-checked goal. Linguistic Inquiry 34: 393–412. Carstens, V. 2005. Agree and EPP in Bantu. Natural Language & Linguistic Theory 23: 219–279. Chomsky, N. 2000. Minimalist inquiries: The framework. In Step by Step, R. Martin et al. (Eds), 89–155. Cambridge MA: The MIT Press. Chomsky, N. 2001. Derivation by phase. In Ken Hale: A Life in Language, M. Kenstowicz (Ed.), 1–52. Cambridge MA: The MIT Press. Collins, C. 2004. The agreement parameter. In Triggers, A. Breitbarth & H. van Riemsdijk (Eds), 115–136. Berlin: Mouton de Gruyter. Creissels, D. 1996. Conjunctive and disjunctive verb forms in Setswana. South African Journal of African Languages 16(4): 109–115. Delais-Roussarie, E., Doetjes, J. & Sleeman, P. 2004. Dislocation. In Handbook of French semantics, F. Corblin & H. de Swart (Eds), 501–528. Stanford CA: CSLI. Demuth, K. 1990. Locatives, impersonals and expletives in Sesotho. The Linguistic Review 7: 233–249. Dordrecht: Foris. Demuth, K. & Harford, C. 1999. Verb raising and subject inversion in comparative Bantu. Journal of African Languages and Linguistics 20(1): 41–61. Déprez, V. 1990. Two ways of moving the verb in French. MIT working papers in linguistics 13: 47–85. Erteschik-Shir, N. 2007. Information Structure. Oxford: OUP.

 Jenneke van der Wal Ethnologue at www.ethnologue.com (as consulted on 15–08–2007). Haspelmath, M., Dryer, M.S., Gil, D. & Comrie, B. (Eds). 2005. The World Atlas of Language Structures (WALS). Oxford: OUP. Katupha, J.M.M. 1983. A Preliminary Description of Sentence Structure in the e-Sáaka Dialect of e-Mákhuwa. MA thesis, SOAS, London. Lambrecht, K. 1994. Information Structure and Sentence Form. Cambridge: CUP. Lambrecht, K. 2000. When subjects behave like objects. Studies in Language 24: 611–682. Larsson, E. 1979. La dislocation en français. Lund: CWK Gleerup. Manzini, M.R. & Savoia, L.M. 2002. Parameters of subject inflection in Italian dialects. In Subjects, Expletives and the EPP, P. Svenonius (Ed.), 157–200. Oxford: OUP. Ndayiragije, J. 1999. Checking economy. Linguistic Inquiry 30: 399–444. Nurse, D. & Philippson, G. (Eds.). 2003. The Bantu Languages. London: Routledge. Pires Prata, Pe. A. 1960. Gramática da língua Macua e seus dialectos. Cucujães: Sociedade Missionária Portuguesa. Rodríguez-Mondoñedo, M. 2005. Case and agreement in Spanish existentials (and beyond). Ms, University of Connecticut. Sabel, J. & Zeller, J. 2005. Wh-question formation in Nguni. Oxford University Working Papers in Linguistics, Philology & Phonetics 10: 169–186. Samek-Lodovici, V. 1996. Constraints on Subjects. An Optimality Theoretic Analysis. Ph.D. dissertation, Rutgers University. Sasse, H.-J. 1987. The thetic-categorical distinction revisited. Linguistics 25: 511–580. Sasse, H.-J. 1996. Theticity. In Arbeitspapiere 27. Cologne: Institut für Sprachwissenschaft der Universität zu Köln. Schadeberg, T.C. & Mucanheia, F.U. 2000. A grammar of Ekoti. Cologne: Rüdiger Köppe. Schoorlemmer, E. 2007. Agree and existential constructions. In Romance Linguistics 2006: Selected papers from the 36th Linguistic Symposium on Romance Languages (LSRL), J. Camacho, N. Flores-Ferrán, L. Sánchez, V. Déprez & M. José Cabrera (Eds), 269–282. Amsterdam: John Benjamins. Stucky, S. 1979. The interaction of tone and focus in Makua. Journal of African Languages and Linguistics 1: 189–198. Van der Wal, G.J. 2006a. The disjoint verb form and an empty Immediate After Verb position in Makhuwa. ZASPiL 43: 233–256. Van der Wal, G.J. 2006b. Predicative tone lowering in Makhuwa. In Linguistics in the Netherlands, J.M. van de Weijer & B.Los (Eds.), 224–236. Amsterdam: John Benjamins. Van der Wal, G.J. In preparation. The Makhuwa non-subject relative as a participial modifier. Van Valin, R. 1999. A typology of the interaction of focus structure and syntax. In Typology and the Theory of Language: From Description to Explanation, E. Raxilina & J. Testelec (Eds). Moscow. Zerbian, S. 2006. Expression of Information Structure in Northern Sotho. Ph.D. dissertation, Humboldt University, Berlin.

Index of languages

A Aghem, 304, 305, 314 B Basaa, 313 Bemba, 263, 264, 305, 306, 309, 310, 312, 314–16 C Chichewa, 5, 6, 35, 97–102, 149, 150, 262, 277, 283, 285, 304, 316–17, 335, 341 Chishona, 111, 112–18, 122–26 D Duala, 313 E English, 151, 161, 177, 181, 183, 187–94, 250, 325, 334, 336, 337 Ewe, 325–26 F Faroese, 145–46 French, 22, 42, 44, 45–47, 73–75, 78, 84–97, 102–06, 112–19, 159, 195, 229, 230, 326–29 Old French, 90–92 G German, 145, 146 Greek, 173, 194, 230 H Haya, 45, 47, 71, 78 Hebrew, 213–16

I Ikalanga, 178, 192 Isizulu, 112–18, 120, 121, 123, 124–26 Italian, 22, 41–79, 104, 105, 112–19, 135, 147, 148, 173, 177–88, 190–95, 211–14, 221, 222, 229, 230, 240–46, 248–56, 261–64, 267–75, 278–84, 286–88, 301, 302, 325–29, 345, 347, 348 Northern dialects, 43, 62, 326, 327, 347, 348 K Kannada, 181 Kikuyu, 142 Kilega, 335–37 Kinyarwanda, 24, 45, 69, 101 Kirundi, 24, 309, 341 Kitharaka, 272 L Latin, 20–22, 173, 194, 342 Luganda, 240, 241, 244–51, 255, 256 M Makhuwa, 305, 306, 309, 311, 312, 314, 325, 327, 329–32, 335–38, 341–47 N Northern Sotho, see Sotho (Northern) Nsenga, 25 O Otjiherero, 24, 25, 27–37

P Portuguese, 229, 230 European Portuguese, 294, 295, 298–302 Brazilian Portuguese, 298–301 R Romanian, 213–16, 218, 219, 222, 223, 225–29 Rundi, see Kirundi S Sesotho, 265, 325–32, 339–41 siSwati, 7, 23–24 Sotho (Nothern), 318, 319, 340 Spanish, 5–7, 10–19, 22, 43, 47–48, 103, 104, 132–33, 135, 136, 145, 147, 151, 153, 160, 162, 179, 193, 229–31, 233, 295, 342 Swahili, 24, 105, 133, 134, 136–47, 152, 154–58, 160, 204, 210, 211, 213, 214, 217, 224, 225, 231, 311, 316 T Tamil, 181, 182 Tshiluba, 45, 69, 74, 106 Tswana, 24, 307 Tumbuka, 23, 26, 311, 312, 316 X Xhosa, 71, 226, 231, 232 Z Zulu, see Isizulu

General index

A *adjunction, 23, 25, 26, 28, 29, 37 local, 20–22 late, 16, 17, 19, 25 A-position, 277, 283, 286 A’-position, 282, 283, 287 A-bar move, 88, 220 aboutness-shift topic see topic adjectival article see article agglutination, 111–13 Agree relation, 161–63, 168, 170, 172, 173, 178, 179, 188, 202, 205, 300, 301, 334–37 agreement, 6, 8, 36, 42, 103, 131, 137, 149–52, 160–62, 167–75, 187, 188–96, 202–09, 217–19, 324, 328–30, 333–48 see also concord default, 332, 339, 341 grammatical, 244, 339, 341–47 marker, 6, 23, 36–38, 149, 244, 339–48 pronominal, 339–41 semantic, 68, 192–96 subject, 8, 33, 103, 104, 160–62, 201, 202, 217, 276, 277, 328–30, 336, 338, 339 anaphora, 6, 7, 12–16, 18, 19, 98, 101, 230, 231, 275, 339 article, 153, 213–16, 222–32 adjectival, 216, 227, 228 genitival, 219, 222–24 B backgrounding, 5, 18, 19 broad focus see focus C categorical statement, 323–25, 329, 332, 340, 342

C-domain, 266, 271–77, 287, 288, 312 hierarchy, 266 class marker, 64, 99 clitic, 22–29, 36–38, 41–79, 84–85, 89–90, 97–99, 102–06 consonantal, 63–66 doubling, 5–7, 13, 14 morphologically complex, 64–66, 71 object, 9, 24, 25, 67 clitic-resumed topic see topic clitic left dislocation (CLLD) see dislocation concord, 160–63, 202–08, 211–14, 217–33, 310 see also agreement conjoint-disjoint, 307–13, 330–32 context dependence, 13–19 contrast, 5, 19, 263, 269–71, 274, 277, 282 see also topic, focus contrastive topic see topic coreference, 264, 275, 281, 282, 286, 287, 339–42 consonantal clitic see clitic cross-linguistic variation, 25, 195, 292, 300 D default agreement see agreement default features, 170–73, 241 person, 105, 175–78 gender, 178, 179, 190 number, 171 demonstrative, 154–59, 184, 185, 225, 226, 228–33 detopicalization see topic discourse, 90–92, 215, 297, 298–303, 307, 308, 310, 318, 319 see also topic, focus categories, 261–72, 274

configurationality, 294, 295, 298, 299, 301, 304 functions, 261, 264–67, 274, 279, 281, 282, 288, 294, 304, 316, 318, 319 dislocation, 16, 18, 19, 272, 284, 285, 287, 340, 344–47 Clitic Left Dislocation, 16 Hanging Topic Left Dislocation, 18, 19 Dynamic Syntax (DS), 6, 8–22, 30, 32, 35, 313 E enclisis, 42, 49–58, 60, 61, 64, 70, 73, 74 epenthetic vowel, 62–64 EPP, 208, 283, 295, 333–37 F familiar topic see topic features, 85, 104–06, 119, 135, 145, 160–162, 167–97, 201–11, 217, 232–33, 240–41, 271, 295, 328, 333–38, 348 gender, 64, 66–68, 131–49, 160–62, 179–87, 208, 217, 220–23, 227, 242–253, 328, 348 number, 66–70, 72, 74, 77, 78, 118, 135–40, 160, 187–96, 208, 217, 220–23, 227, 328, 329, 339, 348 person, 66–68, 70–74, 77–78, 118, 175–79, 208, 211–13, 223, 339, 348 word class, 202, 220, 227, 231 feature spreading see spreading focus, 5, 25, 114–15, 261–67, 271, 293–321, 324, 325 broad, 262, 263, 270, 279, 280, 282, 284 narrow, 262, 263, 324

 General index G gender see features genitival article see article goal, 161, 162,168, 202, 203, 205, 207, 217, 334, 336, 338 grammatical agreement see agreement H hanging topic left dislocation (HTLD) see dislocation I impersonal si see si indefinite, 193, 207, 215, 222, 226, 227, 252, 277, 345 inflection see verbal inflection interface, 167, 168, 170, 171, 261–63, 272, 278–89, 293–96, 300, 316–20 IP-inversion, 273, 274, 287, 288 L LCA, 42, 58–60, 249–51, 253, 255 late *adjunction see *adjunction left periphery, 4–9, 23–27, 113, 115, 123, 269, 272–74, 285, 313 local *adjunction see *adjunction local scrambling see scrambling local underspecification see underspecification locative, 8, 25, 26, 44, 46–48, 51–53, 63, 66, 67, 72–74, 88, 89, 101, 102, 149–51, 230, 231, 265, 286, 306, 315, 338, 341, 346, 347 inversion, 27, 29, 30, 33–37, 284, 335, 336, 338, 341, 346 subject markers, 25 long-distance scrambling see scrambling M marginalized objects, 287–90 markedness, 195 Merge, 153–58, 162, 202, 239–40, 266, 334 Minimalism, 161, 167–169, 202, 204, 205, 211, 266, 334

morphological fusion, 112, 113, 118, 125, 126 morphologically complex clitic see clitic N narrow focus see focus noun class, 29, 118, 120, 132–49, 244, 247, 329, 336 noun phrase, 22, 27, 28, 97, 98, 154, 155, 159, 188, 210, 211, 241, 296, 339–41 nominal expression, 202–33 noun-raising, 152–55 null nominals, 150–51, 182–87, 250, 251 number see features O object clitic see clitic object marker, 7, 8, 23–27, 29, 32–34, 36, 83–106 object-subject inversion, 5 Optimality Theory, 316 P passive, 8, 30–37 periphery see right periphery, left periphery person see features Person-Case constraint, 77–79 postverbal subject see subject possessive, 301, 302 adjective, 209–213, 216, 217 DPs, 88, 89 pronoun, 209–11, 213, 216 preposed locative, see locative inversion (pre-)prefix spreading see spreading preverbal subject see subject Principle B, 340, 342, 347 probe, 161, 162, 168, 202, 203, 205, 207–11, 214, 216–18, 224, 334–36, 338 proclisis, 42, 43, 49–58, 60, 61, 63–65, 69, 70, 73, 75, 76, 104, 225 pro-drop, 12, 14, 18, 27, 339 prominence, 266, 295, 297, 299, 301, 303, 316, 318, 319

pronominal agreement see agreement pronoun, 7, 12–14, 45–47, 60, 61, 66, 67, 70–77, 98, 120, 121, 152, 153, 177–79, 339–41 doubling, 7, 19 weak, 46, 47, 90–95 prosody, 121–23, 261, 262, 295–97, 314–19, 325 prosodic domain, 252, 262, 266, 285 prosodic phrasing, 309–13, 316–17 R raising see noun-raising reflexive si see si right periphery 4–9, 17, 19, 23–27, 273, 274, 298, 299, 301, 313, 318 S scrambling, 7, 9, 19–22, 27, 36, 37 local, 7, 20 long-distance, 7, 9 short-distance, 9 semantic agreement see agreement short-distance scrambling see scrambling si impersonal, 48, 51, 54, 56, 57, 59, 63, 75–77 reflexive, 48, 51, 54, 63, 67, 74–78 split-INFL, 119–23, 125 spreading definiteness, 207 feature, 188, 204 (pre-)prefix, 225, 227 tone, 310, 313 s-sonorization, 61, 63 STAY, 111, 113, 126 structural underspecification see underspecification subject 5, 6, 8, 9, 30–37, 43, 207, 324, 325, 342–45, 347, 348 marker, 23–29, 31, 34, 98, 329, 341 postverbal, 35, 278–84, 328, 332, 340, 341–45

General index  preverbal, 18, 19, 275–79 syntactic, 9, 37 subject agreement see agreement syntactic subject see subject T tense marker, 28–30, 34, 69 theticity, 265, 323–32, 339–47 tone, 263, 265–71, 279–88, 303, 306, 307, 309–13, 315–17, 331, 332 tone spreading see spreading Topic, 98, 262–66, 275, 276, 285, 286 aboutness-shift, 266, 267, 271, 273

clitic-resumed, 264, 266–71, 275, 276, 286–89 contrastive, 269–71, 274, 277, 282 detopicalization, 324–26, 340 familiar, 268, 269, 271, 273, 274, 281, 283 U underspecification, 6–9, 12, 15, 16, 36, 37, 148, 168, 169, 219, 313 local, 4, 9, 19–22 structural, 6–9, 15, 19–22, 26, 313

unfixed nodes, 6, 7, 15–37 Universal, 20, 153, 154 V V2 90, 91, 112–15 verbal inflection, 115–23, 125, 126 W weak pronoun see pronoun WH-extraction, 111–18, 123–25 word class see features word order, 6, 19–22, 45–48, 71–79, 94, 151–60, 209–11, 217, 229, 232, 250, 256, 263, 293–96, 299, 302–05, 309, 315–19, 325, 333

Linguistik Aktuell/Linguistics Today A complete list of titles in this series can be found on the publishers’ website, www.benjamins.com 137 Pfau, Roland: Grammar as Processor. A distributed morphology account of spontaneous speech errors. Expected January 2009 136 Kandybowicz, Jason: The Grammar of Repetition. Nupe grammar at the syntax–phonology interface. xiii, 159 pp. + index. Expected November 2008 135 Lewis, William D., Simin Karimi, Heidi Harley and Scott O. Farrar (eds.): Time and Again. Theoretical Perspectives on Formal Linguistics. In honor of D. Terrence Langendoen. xiv, 261 pp. + index. Expected November 2008 134 Armon-Lotem, Sharon, Gabi Danon and Susan Rothstein (eds.): Current Issues in Generative Hebrew Linguistics. vi, 388 pp. + index. Expected November 2008 133 MacDonald, Jonathan E.: The Syntactic Nature of Inner Aspect. A minimalist perspective. xv, 234 pp. + index. Expected November 2008 132 Biberauer, Theresa (ed.): The Limits of Syntactic Variation. 2008. vii, 521 pp. 131 De Cat, Cécile and Katherine Demuth (eds.): The Bantu–Romance Connection. A comparative investigation of verbal agreement, DPs, and information structure. 2008. xix, 355 pp. 130 Kallulli, Dalina and Liliane Tasmowski (eds.): Clitic Doubling in the Balkan Languages. ix, 431 pp. + index. Expected October 2008 129 Sturgeon, Anne: The Left Periphery. The interaction of syntax, pragmatics and prosody in Czech. xi, 139 pp. + index. Expected November 2008 128 Taleghani, Azita H.: Modality, Aspect and Negation in Persian. 2008. ix, 183 pp. 127 Durrleman-Tame, Stephanie: The Syntax of Jamaican Creole. A cartographic perspective. 2008. xii, 190 pp. 126 Schäfer, Florian: The Syntax of (Anti-)Causatives. External arguments in change-of-state contexts. 2008. xi, 324 pp. 125 Rothstein, Björn: The Perfect Time Span. On the present perfect in German, Swedish and English. 2008. xi, 171 pp. 124 Ihsane, Tabea: The Layered DP. Form and meaning of French indefinites. 2008. ix, 260 pp. 123 Stoyanova, Marina: Unique Focus. Languages without multiple wh-questions. 2008. xi, 184 pp. 122 Oosterhof, Albert: The Semantics of Generics in Dutch and Related Languages. 2008. xviii, 286 pp. 121 Tungseth, Mai Ellin: Verbal Prepositions and Argument Structure. Path, place and possession in Norwegian. 2008. ix, 187 pp. 120 Asbury, Anna, Jakub Dotlačil, Berit Gehrke and Rick Nouwen (eds.): Syntax and Semantics of Spatial P. 2008. vi, 416 pp. 119 Fortuny, Jordi: The Emergence of Order in Syntax. 2008. viii, 211 pp. 118 Jäger, Agnes: History of German Negation. 2008. ix, 350 pp. 117 Haugen, Jason D.: Morphology at the Interfaces. Reduplication and Noun Incorporation in Uto-Aztecan. 2008. xv, 257 pp. 116 Endo, Yoshio: Locality and Information Structure. A cartographic approach to Japanese. 2007. x, 235 pp. 115 Putnam, Michael T.: Scrambling and the Survive Principle. 2007. x, 216 pp. 114 Lee-Schoenfeld, Vera: Beyond Coherence. The syntax of opacity in German. 2007. viii, 206 pp. 113 Eythórsson, Thórhallur (ed.): Grammatical Change and Linguistic Theory. The Rosendal papers. 2008. vi, 441 pp. 112 Axel, Katrin: Studies on Old High German Syntax. Left sentence periphery, verb placement and verbsecond. 2007. xii, 364 pp. 111 Eguren, Luis and Olga Fernández Soriano (eds.): Coreference, Modality, and Focus. Studies on the syntax–semantics interface. 2007. xii, 239 pp. 110 Rothstein, Susan (ed.): Theoretical and Crosslinguistic Approaches to the Semantics of Aspect. 2008. viii, 453 pp. 109 Chocano, Gema: Narrow Syntax and Phonological Form. Scrambling in the Germanic languages. 2007. x, 333 pp.

The Bantu-Romance Connection: A comparative investigation of verbal agreement, DPs, and information structure (Linguistik Aktuell Linguistics Today)

The Bantu-Romance connection: a comparative investigation of verbal agreement, DPs, and information structure

Argument Structure (Linguistik Aktuell Linguistics Today)

The Structure of Stative Verbs (Linguistik Aktuell Linguistics Today)

Agreement Systems (Linguistik Aktuell Linguistics Today, Volume 92)

Advances in Comparative Germanic Syntax (Linguistik Aktuell Linguistics Today)

Experimental Pragmatics Semantics (Linguistik Aktuell Linguistics Today)

Multiple Wh-Fronting (Linguistik Aktuell Linguistics Today)

Parentheticals (Linguistik Aktuell Linguistics Today, 106)

Phase Theory (Linguistik Aktuell Linguistics Today)

Causatives in Minimalism (Linguistik Aktuell Linguistics Today)

Cyclical Change (Linguistik Aktuell Linguistics Today)

The Rise of Agreement: A Formal Approach to the Syntax And Grammaticalization of Verbal Inflection (Linguistik Aktuell Linguistics Today)

Verbal Prepositions and Argument Structure: path, place and possession in Norwegian (Linguistik Aktuell Linguistics Today)

Locality and Information Structure: A cartographic approach to Japanese (Linguistik Aktuell Linguistics Today, Volume 116)

Verbal Complement Clauses: A Minimalist Study of Direct Perception Constructions (Linguistik Aktuell Linguistics Today

Syntatic Aspects of Topic and Comment (Linguistik Aktuell Linguistics Today)

On Information Structure, Meaning and Form: Generalizations across Languages (Linguistik Aktuell Linguistics Today)

The Derivation of Anaphoric Relations (Linguistik Aktuell Linguistics Today)

The Emergence of Order in Syntax (Linguistik Aktuell Linguistics Today)

The Copy Theory of Movement (Linguistik Aktuell Linguistics Today)

Adverbials and the Phase Model (Linguistik Aktuell Linguistics Today, 177)

Current Issues in Generative Hebrew Linguistics (Linguistik Aktuell Linguistics Today)

The Grammar of Focus (Linguistik Aktuell Linguistics Today)

The Grammar of Focus (Linguistik Aktuell Linguistics Today)

Scrambling and the Survive Principle (Linguistik Aktuell Linguistics Today)

The Order of Prepositional Phrases in the Structure of the Clause (Linguistik Aktuell Linguistics Today)

Modality and Its Interaction with the Verbal System (Linguistik Aktuell Linguistics Today, LA 47)

Balkan Syntax and Semantics (Linguistik Aktuell Linguistics Today)

Modality and Its Interaction with the Verbal System (Linguistik Aktuell Linguistics Today, LA 47)

Perspectives on Negation and Polarity Items (Linguistik Aktuell Linguistics Today)

The Bantu-Romance Connection: A comparative investigation of verbal agreement, DPs, and information structure (Linguistik Aktuell Linguistics Today)

The Bantu-Romance connection: a comparative investigation of verbal agreement, DPs, and information structure

Argument Structure (Linguistik Aktuell Linguistics Today)

The Structure of Stative Verbs (Linguistik Aktuell Linguistics Today)

Agreement Systems (Linguistik Aktuell Linguistics Today, Volume 92)

Advances in Comparative Germanic Syntax (Linguistik Aktuell Linguistics Today)

Experimental Pragmatics Semantics (Linguistik Aktuell Linguistics Today)

Multiple Wh-Fronting (Linguistik Aktuell Linguistics Today)

Parentheticals (Linguistik Aktuell Linguistics Today, 106)

Phase Theory (Linguistik Aktuell Linguistics Today)

Causatives in Minimalism (Linguistik Aktuell Linguistics Today)

Cyclical Change (Linguistik Aktuell Linguistics Today)

The Rise of Agreement: A Formal Approach to the Syntax And Grammaticalization of Verbal Inflection (Linguistik Aktuell Linguistics Today)

Verbal Prepositions and Argument Structure: path, place and possession in Norwegian (Linguistik Aktuell Linguistics Today)

Locality and Information Structure: A cartographic approach to Japanese (Linguistik Aktuell Linguistics Today, Volume 116)

Verbal Complement Clauses: A Minimalist Study of Direct Perception Constructions (Linguistik Aktuell Linguistics Today

Syntatic Aspects of Topic and Comment (Linguistik Aktuell Linguistics Today)

On Information Structure, Meaning and Form: Generalizations across Languages (Linguistik Aktuell Linguistics Today)

The Derivation of Anaphoric Relations (Linguistik Aktuell Linguistics Today)

The Emergence of Order in Syntax (Linguistik Aktuell Linguistics Today)

The Copy Theory of Movement (Linguistik Aktuell Linguistics Today)

Adverbials and the Phase Model (Linguistik Aktuell Linguistics Today, 177)

Current Issues in Generative Hebrew Linguistics (Linguistik Aktuell Linguistics Today)

The Grammar of Focus (Linguistik Aktuell Linguistics Today)

The Grammar of Focus (Linguistik Aktuell Linguistics Today)

Scrambling and the Survive Principle (Linguistik Aktuell Linguistics Today)

The Order of Prepositional Phrases in the Structure of the Clause (Linguistik Aktuell Linguistics Today)

Modality and Its Interaction with the Verbal System (Linguistik Aktuell Linguistics Today, LA 47)

Balkan Syntax and Semantics (Linguistik Aktuell Linguistics Today)

Modality and Its Interaction with the Verbal System (Linguistik Aktuell Linguistics Today, LA 47)

Perspectives on Negation and Polarity Items (Linguistik Aktuell Linguistics Today)

Recommend Documents