Locality Domains in the Spanish Determiner Phrase (Studies in Natural Language and Linguistic Theory)

LOCALITY DOMAINS IN THE SPANISH DETERMINER PHRASE

Studies in Natural Language and Linguistic Theory VOLUME 79 Managing Editors Marcel den Dikken, City University of New York Liliane Haegeman, University of Ghent, Belgium Joan Maling, Brandeis University Editorial Board Guglielmo Cinque, University of Venice Carol Georgopoulos, University of Utah Jane Grimshaw, Rutgers University Michael Kenstowicz, Massachusetts Institute of Technology Hilda Koopman, University of California, Los Angeles Howard Lasnik, University of Maryland Alec Marantz, Massachusetts Institute of Technology John J. McCarthy, University of Massachusetts, Amherst Ian Roberts, University of Cambridge

For further volumes: http://www.springer.com/series/6559

LOCALITY DOMAINS IN THE SPANISH DETERMINER PHRASE by

M. EMMA TICIO University of Syracuse, NY, USA

13

M. Emma Ticio Dept. of Languages, Lit &Ling, 340 HBC Syracus University Syracuse, NY 13244-1160 USA [email protected]

ISBN 978-90-481-3397-0 e-ISBN 978-90-481-3398-7 DOI 10.1007/978-90-481-3398-7 Springer Dordrecht Heidelberg London New York Library of Congress Control Number: 2009940142 # Springer ScienceþBusiness Media 2010 No part of this work may be reproduced, stored in a retrieval system, or transmitted in any form or by any means, electronic, mechanical, photocopying, microfilming, recording or otherwise, without written permission from the Publisher, with the exception of any material supplied specifically for the purpose of being entered and executed on a computer system, for exclusive use by the purchaser of the work. Printed on acid-free paper Springer is part of Springer Science+Business Media (www.springer.com)

Preface

This book deals with the syntax of nominal expressions from a generative point of view. The book concentrates mainly on empirical data taken from Spanish, although the study also discusses properties shown by other languages, to draw conclusions of theoretical interest. Although the properties of nominal expressions in Romance languages have been extensively studied in the literature, there has not yet been a book that examines theoretically all aspects of the structure of nominal expressions in Spanish. Aspects of Spanish nominal expressions have been profusely studied but with a focus on particular phenomena in the nominal domain. For instance, Spanish data on extraction out of nominal expressions has not been very well represented in the literature. Most authors who discuss Spanish do not relate the Spanish extraction data to the other main phenomena in the Spanish nominal expressions and the basic works on extraction out of Spanish nominal expressions has not been published. The publication of this research monograph meets then two needs: first, to provide an empirical description of the properties of Spanish nominal expressions and to disseminate the descriptive generalizations, preliminary results and problems of previous research among the researchers and second, to give an explanation for the different phenomena discussed in the Spanish literature under a uniform account. The major contribution of the book is the proposal of a single explanation for several phenomena that have not been analyzed under a single basic account previously. Specifically, the book explores how economy notions interact with a number of functional categories, with the length and type of movements allowed and with the existence of domains within nominal expressions. Additionally, some previously unobserved data regarding ellipsis and cliticization in Spanish nominal expressions are discussed. Therefore this monograph can be seen as the reference point for further research in Spanish linguistics and it is my hope that the ideas presented here stimulate further research on these and related topics. This book has grown out of my doctoral dissertation completed at the University of Connecticut in May 2003. Some of the ideas presented in this book developed while I was a graduate student at this institution and I would like to express my warmest thanks to all the members in the department, v

vi

Preface

especially to my dissertation committee members, for the stimulating work environment and for their constant help and guidance. As a member of the University of Houston, I have also benefited from the interaction with my colleagues and students there, who have contributed with judgments and suggestions to this book. While working on this material, I have had fruitful discussions and exchanges with many linguists. I do not think I can possibly reproduce the entire list here but I thank all of them for their direct or indirect contribution to this book. Particular thanks are due to two anonymous SNLT reviewers for patiently wading through previous versions of the manuscripts and for their comments. Different parts of the material in this book were presented in presentations at events at the University of Connecticut, University of Houston, University of Arizona, The Ohio State University, University of New Mexico, University of Sonora, University of Minnesota, The Pennsylvania State University and University of Western Ontario. Audiences at these places have always offered very helpful suggestions and stimulating questions, and I thank them for that. Lastly and most importantly, I would like to thank my wonderful family, Aaron, Daniel and Sara, for putting up with me during the writing of this book. This book is dedicated to them. Research on this book was partially possible thanks to a New Faculty Research Grant that the University of Houston awarded me during the summer of 2005. New York

M. Emma Ticio

Contents

1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.1 Scope of the Book . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.2 On the Parallelism Between Clauses and Nominal Phrases . . . 1.2.1 Nouns and Verbs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.2.2 The Challenges . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.3 Theoretical Framework . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.3.1 Foundations on Minimalism . . . . . . . . . . . . . . . . . . . . . 1.3.2 Locality in the Minimalism Framework . . . . . . . . . . . . 1.4 Overview of the Book. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Notes. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1 1 2 2 5 8 8 11 14 19

2. On the Structure of Nominal Expressions . . . . . . . . . . . . . . . . . . . . . 2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.2 Basic Properties of Spanish Nominal Expressions. . . . . . . . . . 2.3 Structural Relations Within Nominal Constructions . . . . . . . 2.3.1 Structural Hierarchy Within Nominal Expressions . . . 2.3.2 Two Types of PPs in Spanish Nominal Expressions. . . 2.4 The Structure of Nominal Expressions in Spanish . . . . . . . . . 2.4.1 Theoretical Assumptions . . . . . . . . . . . . . . . . . . . . . . . . 2.4.2 The Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.4.3 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Appendix: Genitive Pronouns . . . . . . . . . . . . . . . . . . . . . . . . . Notes. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

23 23 27 29 30 38 49 50 55 71 72 74

3. Specificity, Definiteness and the Definite Article in Spanish Nominal Expressions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.2 Fundamentals: Definiteness and Specificity. . . . . . . . . . . . . . . 3.2.1 The Distinction Definite vs. Indefinite . . . . . . . . . . . . . 3.2.2 Specificity in Indefinite Nominal Expressions. . . . . . . . 3.2.3 Specificity in Definite Nominal Expressions . . . . . . . . . 3.2.4 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

83 83 86 86 88 90 93

vii

viii

Contents

3.3

3.4

3.5

Extraction, Definiteness and Specificity. . . . . . . . . . . . . . . . . . 3.3.1 The Specificity Effect. . . . . . . . . . . . . . . . . . . . . . . . . . . 3.3.2 Extraction Out of Specific Nominal Expressions . . . . . 3.3.3 Extraction Out of Nominal Expressions with the Definite Article . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.3.4 Extraction Out of Specific Indefinite Nominal Expressions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.3.5 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A Syntactic Analysis. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.4.1 Basic Assumptions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.4.2 Extraction Out of Specific Definite Spanish DPs . . . . . 3.4.3 Explaining the Blocking Effects Out of Specific Indefinite Nominal Expressions . . . . . . . . . . . . . . . . . . 3.4.4 On the Spanish Definite Article. . . . . . . . . . . . . . . . . . . 3.4.5 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Summary and Further Research . . . . . . . . . . . . . . . . . . . . . . . 3.5.1 Main Claims of the Chapter . . . . . . . . . . . . . . . . . . . . . 3.5.2 Further Research . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Notes. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

93 94 96 97 97 99 100 100 103 105 107 115 115 115 116 117

4. Adjective Placement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.2 Description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.2.1 Linear Order . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.2.2 The Relation Between Meaning and Linear Position . . 4.2.3 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.3 Towards an Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.3.1 Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.3.2 Evidence from the Syntactic Behavior of the Different Groups of Adjectives . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.3.3 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.4 Previous and Further Lines of Research . . . . . . . . . . . . . . . . . 4.4.1 Problems for a Uniform Analysis for Spanish Adjectives 4.4.2 The Role of Noun Movement in the Analyses of Romance Adjectives . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.4.3 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.5 Summary of the Chapter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Notes. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

121 121 124 124 127 132 134 134

152 155 155 158

5. Nominal Ellipsis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.2 Nominal Ellipsis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.2.1 Description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.2.2 Classifying the Data . . . . . . . . . . . . . . . . . . . . . . . . . . .

163 163 165 165 172

143 146 146 147

Contents

5.3

5.4

5.5

ix

Putting Together the Analysis . . . . . . . . . . . . . . . . . . . . . . . . . 5.3.1 General NP Ellipsis . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.3.2 On the Restrictions of the Definite Article in NP Ellipsis Contexts. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.3.3 Additional Evidence . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.3.4 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Agreement and Focus in Ellipsis Processes . . . . . . . . . . . . . . . 5.4.1 The Relevance of Strong Agreement in Nominal Ellipsis 5.4.2 The Relevance of Focus in Nominal Ellipsis. . . . . . . . . 5.4.3 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Summary of the Chapter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Notes. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

176 176 184 186 188 188 189 194 196 197 197

6. Conclusions and Further Research . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.2 Major Findings of the Book. . . . . . . . . . . . . . . . . . . . . . . . . . . 6.3 Further Research . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

203 203 203 206

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

207

Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

215

Chapter 1

Introduction

1.1 Scope of the Book This book analyzes the syntactic structure of nominal expressions in Spanish and some related languages within a minimalist framework (Chomsky (1995) and subsequent work). More concretely, I provide an analysis for the structure of arguments and adjuncts in nominal expressions and for the possibilities for extraction of these elements. With this goal in mind, the book considers crucial phenomena in the nominal domain such as extraction out of nominal phrases, ellipsis in nominal phrases and modification of nominal phrases. The book adopts the so-called Determiner Phrase hypothesis (cf. Abney’s (1987), among others). Thus, I assume throughout the book that the internal structure of nominal phrases resembles the internal structure of clauses in its functional configuration. Following this line of research, I show that the division of clause structure into three domains (cf. Rizzi (1997), Platzack (2001), Grohmann (2003), among many others) can be extended to account for the structure and properties of Spanish nominal phrases. Hence, nominal phrases in this book have a lexical domain, where all the lexical items enter the derivation; an agreement domain, where the basic agreement relations between lexical and functional categories are established; and a discourse domain that interacts with the discourse. The adoption of this tripartite structure for nominal phrases and the locality and anti-locality constraints on movement proposed in the literature enable me to explain the peculiar extraction properties of Spanish nominal phrases as the result of locality violations. Phenomena such as the properties of attributive adjectives, partial cliticization and nominal elision in Spanish nominal phrases are used to test the empirical adequacy of the analysis. Therefore, this book offers a single explanation for a number of phenomena that have not yet been analyzed under a unified account. Although I am aware that there are many answers left open here, I hope the data and ideas presented in this book provide an overview of the current research on nominal expressions in Spanish and stimulate further research on these and related topics in Spanish and some other languages. M.E. Ticio, Locality Domains in the Spanish Determiner Phrase, Studies in Natural Language and Linguistic Theory 79, DOI 10.1007/978-90-481-3398-7_1, Ó Springer ScienceþBusiness Media B.V. 2010

1

2

1

Introduction

This first chapter is a general introduction to the research on the field and the book and is organized as follows. Section 1.2 is a very brief introduction to the main lines of research on the nominal domain and to their challenges. For ease of exposition, I first describe the line of research that argues for a parallelism between the clause and the nominal phrase in Section 1.2.1 and the main points that will be used in the book in Section 1.2.2. In Section 1.3, I introduce the proposals that will be the foundation for the analysis pursued in this book. I start this section with a subsection, (1.3.1), which describes the fundamental notions proposed in the minimalist framework to establish the general framework that will be used in the rest of the book. This is a very basic introduction to the theoretical framework, which can be skipped by those readers familiar with the theoretical models described here. The second part of this section (1.3.2), focuses on some of the core notions of my analysis: the locality and anti-locality conditions on movement. Therefore, this section details the proposals that will be the basis for the analysis of this book, since one of the main goals of the book is to show that proposals such as anti-locality can be carried over to the analysis of the nominal expressions in Spanish. Finally, the chapter ends in Section 1.4, where I outline the main contents of the book. I briefly explain the issues I will be dealing with in Chapters 2, 3, 4, 5, and 6 and advance some of the solutions I will be providing.

1.2 On the Parallelism Between Clauses and Nominal Phrases The structure of nominal phrases has been extensively studied in the generative framework.1 There are several studies on this issue, which cover the topic from different points of view. This section does not intend to offer a comprehensive discussion but rather aims to describe the main lines of research that argue for a parallelism between the clause and the nominal phrase and to point to the departures of this study from previous approaches. The section starts with a discussion on the lexical parallelism between nouns and verbs and links it to the line of research that considers the existence of a functional parallelism between noun and verb (Section 1.2.1). After the short introduction, I focus on the challenges for this line of thought; namely, I discuss a few problems for the parallelism between nouns and verbs, such as the postulation of functional categories or the existence of noun movement (Section 1.2.2).

1.2.1 Nouns and Verbs Lees (1960) was the first author to describe the similarities between sentences and nominal phrases in generative grammar. Lees observed that sentences and nominal phrases are similar in their distribution. As the examples in (1) illustrate, sentences and nominal phrases, which are considered arguments, can

1.2 On the Parallelism Between Clauses and Nominal Phrases

3

appear as subjects, as in (1a) and objects, as in (1b) and undergo passivization, as in (1c). (1) a. John surprised me That John came surprised me b. I know John I know that John came c. John1 was known t1 by many linguists That John came1 was known t1 by many linguists (Abney 1987: 23, (17)) Thus, Lees’s observation was that it is not only the case that some verbs and their corresponding deverbal nouns share some basic semantic properties but also that the contexts in which any sentence and any noun phrase can appear are closely related. Ten years after Lees’s study, Chomsky (1970) pointed out a possible explanation of the syntactic parallelism between these two categories: the structure related to a verb is the same as the structure related to a noun due to the fact that verbs and nouns share some lexical properties. Therefore, the obvious semantic relation between (2a) and (2b) should be captured as the result of the verb and the corresponding deverbal noun sharing some semantic properties, which produces nominal phrases and clauses having similar structures. (2) a. John loves Mary b. John’s love for Mary Chomsky’s claim is that the point of departure for several studies on nominal expressions argues for a parallelism between nominal phrases and clauses. Several explanations have been provided for this parallelism: the initial studies on this topic postulate that the parallelism between verbal and nominal phrases is due to the argumental structure of nouns being projected in a similar fashion as the argumental structure of verbs. In other words: nouns and verbs share their lexical representations as far as their thematic properties are concerned. This is what is known as the configurational hypothesis and is represented by works such as Anderson (1979), Cinque (1980), Torrego (1987), Giorgi and Longobardi (1991), Picallo (1991), Valois (1991), Martı´ n (1995), among many others. Under the configurational approach, the theta-grid of nouns parallels the theta-grid of verbs and the differences appearing on the surface are due to the intervention of different modules of the grammar. Hence, returning to the examples in (2) above, the theta-grid of love, noun or verb, is the same: it selects for an agent, John and a theme, Mary. The different structures are derived by the different ways in which case and agreement are established in the nominal and clausal domain. Research under this approach has produced invaluable contributions to the study of noun phrase syntax. For instance, Grimshaw (1990) distinguished two types of nouns regarding their lexical representation: event nouns, which denote events or processes and behave as verbs assigning

4

1

Introduction

theta roles and result nouns, which do not assign theta roles and do not denote events.2 A second line of research started with Abney’s (1987) Determiner Phrase hypothesis. According to this proposal, the traditional structure for noun phrases (NPs) in generative grammar, represented in (3), was neither empirically or theoretically accurate. (3)

NP

N’

Determiner the

N book

One of the problems noted for the structure in (3) was how to represent the range of prenominal elements with a single specifier position provided. That is, NPs containing more than one element preceding the noun, such as those two men, cannot receive a straightforward representation in the structure (3). Abney’s claim is that noun phrases (NPs) are embedded under a higher functional projection (Determiner Phrase, DP), a syntactic structure that can be easily correlated with the higher functional projections (Complementizer Phrase (CP) or Inflection Phrase (IP), depending on the analysis) that dominate the Verbal Phrase (VP). This assumption enables Abney to explain several semantic and syntactic parallelisms between nominal phrases and clauses and to accommodate in the standard X’ theory the full range of prenominal elements in structures such as (4): (4)

DP D’ D the

NP book

Under this approach it is assumed that DP is the extended projection of N, as the CP is the extended projection of V. Both D and C turn their complements into arguments and both seem to contain a number of associated functional categories that mediate between the lexical head (N or V) and the highest functional category (D or C). The presence of these functional categories provides us with the extra positions needed to represent all the prenominal elements in the DP. The existence of functional categories inside DPs has been corroborated in numerous analyses since the emergence of the DP hypothesis. Szabolcsi (1983), for instance, presented evidence in support of a richer structure for DPs in Hungarian. The idea that DPs contain intermediate functional projections between determiner and noun has been further confirmed by Ritter

1.2 On the Parallelism Between Clauses and Nominal Phrases

5

(1988a, b, 1991) regarding Hebrew. In Romance linguistics, a good number of proposals have also argued for the presence of different functional categories within DP to account for the properties of Romance DPs. For instance, Picallo (1991) argued for the existence of gender phrase to accommodate the order of arguments with respect to the noun in Catalan DPs; Bernstein (1993, 2001a) claimed that a similar functional category, the Word Marker Phrase, could be useful to explain data regarding nominal ellipsis in Romance languages.3 From a theoretical point of view, the empirical evidence for the existence of DP reinforces the parallelism between nouns and verbs since it shows that the internal structure of nominal phrases resembles the internal structure of clauses in its functional configuration.

1.2.2 The Challenges The proposals described above face challenges regarding the notion of a parallelism between nouns and verbs and regarding some of their fundamental assumptions, such as the functional categories postulated and the existence of noun movement. This subsection examines some of these challenges and positions the book with respect to these major issues in the field. A natural question to consider here is to which extent a parallelism between nouns and verbs is postulated. Although there are several striking properties that point to a parallel structure between nouns and verbs, such as the similarity in the selection and structural properties discussed in Section 1.2.1, there are still some differences between the two categories that need to be considered. Just to mention one difference, it has often been asserted that nouns, contrary to verbs, take arguments only optionally (with most nouns being typically used without any arguments). This distinctive property of nouns remains largely unexplained in the literature and casts doubts on the proposed parallelism of nouns and verbs. Obviously, the main reason to pursue the noun verb parallelism is to fulfill some type of economy considerations in our theory of grammar. Since Abney’s (1987) pioneering approach, the goal of this line of research has been to simplify the theory by unifying the treatment given to nouns and verbs under the X’ theory. Although a substantial number of problems are still to be addressed theoretically and empirically, the results achieved so far confirm that the line of research focusing on the parallelism between nouns and verbs is the most promising line of research. This and the fact that this book deals with the nominal expressions internal structure, which resembles the internal hierarchy observed in the sentence domain, lead me to follow the existence of a parallelism between nouns and verbs as the null hypothesis. Nevertheless, as will be pointed out in several sections of the book, the parallelism is not complete.

6

1

Introduction

Alongside the main question on how deep the parallelism between nouns and verbs should be, the development of the theoretical framework has posed some questions to some basic assumptions of the line of research discussed. I elaborate here on the role of functional categories and head movement in the minimalist framework and on how the current trends in the field regarding those issues affect the research on nominal constructions. An important question still under debate in the minimalist framework is the final number and type of functional categories in the clausal and the nominal domains. Since Chomsky (1995), there has been an increased interest in this issue. The main question here is how to motivate the postulation of functional projections in a theoretical framework that has economy notions as its main force. Chomsky (1995) for instance argues against the postulation of functional categories such as agreement in the clausal domain, due to the fact that they seem to contain only formal features that must be deleted during the derivation. Hence, the function of Agr is to provide a structural configuration in which formal features are checked and would remain invisible after those features are checked out prior to spell out. The tendency in the minimalist framework should be then to propose only those functional categories that are semantically well motivated. This position is not only motivated from a theoretical point of view but also empirically. As Truswell (2006, forthcoming) discusses, even the proposal to have an enormous number of functional categories to host different attributive adjectives is not able to generate enough specifier positions for all the attributive adjectives in the different languages. This is because there is not such a rigid hierarchy according to the data in Truswell (2006, forthcoming). Therefore, an analysis based on the existence of multiple and different functional categories within the nominal domain does not have enough empirical coverage; and it will not be economical at all. Taking the previous discussion into consideration, in this book, I return to the basics and will restrict the number of functional categories available in the nominal domain to those that are strictly necessary to maintain the basic tripartite structure proposed for the sentential domain (cf. Rizzi (1997), Platzack (2001), Grohmann (2003), among many others on this issue). That is, following the parallelism with the tripartite sentence structure regarding this point, I will be assuming three functional categories as essential: a functional category that links the discourse to the sentence or the nominal phrase, CP or DP; a functional category where the relevant (meaningful) agreement relations can be established, IP or AgrP; and a third functional category that conveys the agentivity notion, vP or nP. These functional categories have been extensively and independently justified and will only enter in the derivation as needed. Furthermore, in an attempt to avoid the proliferation of functional categories in the nominal domain, I will take the most restrictive point of view and the insertion of any additional functional category in this basic structure will be avoided. In addition to the above mentioned functional categories, only those functional categories fully justified from an empirical point and with certain

1.2 On the Parallelism Between Clauses and Nominal Phrases

7

semantic content can constitute part of the structure in a given numeration; and only those functional categories needed to keep the parallelism with the tripartite structure in the sentence are necessary. Research under the DP hypothesis also led to proposals based on the parallelism between sentence and nominal expression that go far beyond the type of functional structure involved. A look at the word order variation within semitic nominal expressions (Szabolcsi (1983), Ritter (1988a, b) and Siloni (1997), among many others) triggered the first proposals for movement of the noun to D. The two possible genitive constructions in languages such as Hebrew, construct state and free state, were then explained via the resort to noun movement.4 From a typological point of view, the differences between Romance and Germanic nominal expressions regarding their internal word order provided new empirical motivation for the noun movement (N movement) proposal.5 This way, it was assumed that adjectives in the nominal domain and adverbs in the clausal domain should follow parallel analyses (cf. Cinque (1995) and others). Under this view, similarly to what happens in the clausal domain, the head movement of the main lexical category should be ultimately responsible for the different relative surface word orders in the relevant construction. However, there has recently been a line of research (cf. Boeckx and Stjepanovic (2001) and many others) that has argued against the existence of head movement, due mainly to its non cyclical character. In the nominal domain, Lamarche (1991), Alexiadou (2001b), Ticio (2003), Shlonsky (2004), among many others, postulate that N movement is not empirically motivated for Romance languages and suffers in addition from conceptual problems. While the evidence against the possibility of having N movement will be discussed in Chapter 4 as part of the analysis of adjectives in Spanish, let me just advance one of the most compelling arguments against N movement here. As Lamarche (1991) notes, an analysis that derives the difference between Romance and Germanic languages regarding the relative position of adjectives with respect to the N as the result of N movement cannot account for the relative ordering of postnominal Romance adjectives with respect to the N shown in (5). (5) a. un coche blanco oxidado a car white rusty b. a rusty white car The examples in (5) show that the relative ordering of postnominal Romance adjectives with respect to N is the mirror image of the English order. Under the type of analysis that assumes that N movement accounts for the postnominal position of adjectives in Romance, the only difference between Romance and Germanic languages is the position of the noun. The order of the adjectives relative to one another should be the same in both types of language and the order in (5) remains unexplained. These criticisms to the head movement proposals will be taken into consideration in this book and will be one of the reasons why this book does not make

8

1

Introduction

use of this mechanism to derive the different surface word orders available in the Spanish nominal expression. This brief discussion has highlighted some of the ideas developed in this book. To summarize: this book assumes a parallelism between clausal and nominal domains and derives this parallelism from the similar basic functional configuration associated with nouns and verbs and from general principles and constraints of grammar applying to both nouns and verbs. The next section introduces the proposals that will be the foundation for the analysis pursued in this book. The section starts with a brief introduction to the fundamentals in minimalism (Section 1.3.1) and later on, details some of the core notions of my analysis: the locality and anti-locality conditions on movement (Section 1.3.2).

1.3 Theoretical Framework 1.3.1 Foundations on Minimalism The central notion of the minimalist framework is economy. This notion is guiding the design of a theory of grammar that tries to achieve descriptive and empirical adequacy with the minimal number of theoretical mechanisms possible. Under minimalism, the component of the human mind/brain dedicated to language, the language faculty, is optimal in its design. The superficial inadequacies observed (i.e., cross-linguistic differences and non-economical solutions in a given derivation) are due to the properties of the two performance systems, the articulatory-perceptual system (A-P) and the conceptual-intentional system (I-C). These two performance systems interact with the language faculty via two levels of linguistic representation (i.e., interface levels): Phonetic Form (PF) at the articulatory-perceptual interface and Logical Form (LF) at the conceptualintentional interface. Thus, restrictions on linguistically well-formed objects are formulated as restrictions on the two interfaces necessary for the language faculty: PF and LF. A basic representation6 of the minimalist model of grammar is given below: (6) Lexicon

Numeration

Spell Out

PF == A-P system LF == C-I system

Once a numeration (i.e., a collection of lexical items) is selected from the lexicon, the computational system, which yields the derivation from numeration to the LF interface, generates each linguistic expression via two operations: merge and move. The operation merge takes two elements from the relevant numeration and assembles them. The operation move is actually a combination of two operations: merge and copy. To move an element, the computational

1.3 Theoretical Framework

9

system creates a copy of the given element and merges it to the top of the structure created so far. Once a significant amount of structure is built, it is shipped to the two interfaces. The different points at which the information in the structure is shipped to the interfaces are known as spell out.7 After spell out applies, the information contained in the structure affected by spell out is inaccessible to the computational system. After spell out has been applied at a particular point of the derivation, the derivation follows unconnected paths. On the one hand, the path to the A-P component results in a PF representation. During this part of the derivation various processes are applied. These include the linearization of the structure; morphological processes, such as last resort insertions; and stylistic processes, such as optional rearrangements of linear orders. On the other hand, the path to the I-C component results in a LF representation. During this part of the derivation, the syntactic operations applied to the derivation do not have an effect on the phonological result of the derivation but only on the interpretation of the structure at the semantic component. Once the derivation reaches the two representational levels, PF and LF, a linguistic object is obtained. If it meets the conditions at the interfaces (bare output conditions), it converges and then is considered a well-formed linguistic object. If it does not meet the bare output conditions, the derivation crashes and an ill-formed linguistic object is obtained. To clarify the workings of the model, let me illustrate it with a particular derivation. Let us assume that we are deriving the sentence in (7) below: (7) John sleeps The numeration selected from the lexicon to derive (7) is shown in (8):8 (8) {John, sleeps, I, v, C} From this point on, the computational system will combine the elements in the numeration using the only two operations available to the system: merge and move. As noted before, the operation merge takes two elements from the relevant numeration and assembles them. Every time that merge takes place, a relevant feature of the two lexical items involved in the operation is checked off. Via successive applications of merge, we obtain the structure in (9): (9)

I’ I

vP

v’

John v

sleeps

10

1

Introduction

At the point of the derivation represented in (9), the computational system has already inserted the nominal expression John into the derived object. However, in order to derive the sentence above, I’ needs to combine with a nominal element to check its case feature. The computational system must resort to the second operation allowed in the system, move, to accomplish this. To move an element, the computational system creates a copy of the given element and merges it at the top of the structure created so far. The resulting structure is shown in (10): (10)

IP Johni

I’ 0

vP

I

Johni

v’ v

sleeps

The computational system will combine all the remaining elements in the numeration, until it obtains a structure similar to the one given below: (11)

CP C

IP Johni

I’ vP

I Johni

v’ v

sleeps

Once the structure in (11) is built, spell out ships the structure to the two interfaces,9 where it is interpreted and evaluated with respect to the conditions at the interfaces (bare output conditions). Some processes must take place in the structure given in (11) in order to obtain a well-formed linguistic object. For instance, to ensure that only one of the occurrences of John is pronounced a deletion process of copies must take place at the interface. If the resulting object does not meet the bare output

1.3 Theoretical Framework

11

conditions, the derivation crashes and an ill-formed linguistic object is obtained.

1.3.2 Locality in the Minimalism Framework Now that I have outlined the basic schema of the minimalist model, I will focus on a notion extensively used in this book. I am referring to the locality conditions that regulate the syntactic dependencies that elements can establish in a derivation. More concretely, this book is mainly concerned with the general locality conditions on movement and how they affect the nominal structure and properties. Dependencies between two positions are not free but they are subject to locality conditions. Consider the examples below. The grammatical (12a) has wh-movement out of a complement; while the ungrammatical (12b) shows a wh-movement out of a clause that is contained inside a complement. (12) a. Whati did John claim that Mary said ti? b. *Whati did John make the claim that Mary said ti? Putting aside the irrelevant details, the examples in (12) illustrate that an element within a complex NP cannot establish a relation with an element outside the complex NP; it seems to be that the movement is too long. This is one of the constructions that triggered the first locality constraints on movement, which date back to the seventies and was initially known as the complex NP island phenomenon and explained via the subjacency condition10 (cf. Ross (1967), Chomsky (1973) and subsequent work on the topic). Under this approach, the type of movement that the wh-object would have to undergo in (12b) in order for the sentence to be grammatical is considered too long according to the locality conditions on movement. Locality constraints have been one of the main contributions of generative approaches to syntax. Research on locality conditions has been very successful and locality conditions on movement have enormous empirical coverage. Consider the examples below, where the different surface word orders seem to trigger the different grammaticality status of the sentences: (13) a. Mary asked who read what. b. *Mary asked what who read. The basic generalization illustrated in (13) appears to be that while a whsubject can take the clause-initial position in multiple wh-questions, as in (13a), non-subject wh-items cannot raise to clause-initial position leaving a subject wh-item in situ, as in (13b). The ungrammaticality of (13) is usually explained by assuming the existence of locality conditions that are imposed on all relevant dependencies and delimit the upper bound of a well-formed dependency between two positions created by movement. In other words, the type of

12

1

Introduction

movement that the wh-object would have to undergo in (13a) in order for the sentence to be grammatical is considered too long according to the locality conditions on movement. However, this is only so given that there is another shorter movement option; namely, the movement of the wh-subject to the clause-initial position, which is closer to the sentence initial position. This is a case of Relativized Minimality11 (Cf. Rizzi (1990) and subsequent work on the issue), which explains that dependencies between two positions are affected not only by the (structural) distance but also by the elements intervening in the structure. The different locality conditions on movement proposed in earlier theoretical frameworks12 have been restated under the minimalist framework. Thus, locality conditions, such as subjacency, Relativized Minimality and others, are regulated now by general economy principles,13 such as the Minimize Chain Links Principle (Chomsky and Lasnik 1993) or the Minimal Link Condition (Chomsky 1995), which require each movement to be as short as possible. Taking into consideration the value of economy notions in our theory of grammar, Manzini (1994) suggests that locality conditions, such as subjacency, Relativized Minimality and Huang’s (1982) Condition on Extraction Domain, can be derived from general notions such as domain of a head and minimal domain of a head (represented below as (X)) if they are built into the definition of a more general locality, formulated as in (14). (14) Locality For all i, let Ai be in (Xi). Given a dependency (Ai, . . ., An), for all i, (Xi) (Manzini (1994:483)) and (Xi+1) are adjacent. What the definition in (14) states is that two elements next to one another in a dependency must be minimally distant in the tree, where minimum distance means adjacency of minimal domains. In other words: elements must move to the immediately next domain. Manzini (1994) proves that this locality condition is able to derive straightforwardly the empirical evidence supporting wellknown locality conditions such as the Head Movement Constraint (Travis 1984), Relativized Minimality (Rizzi 1990) and others. In addition, this locality condition can provide a principle explanation for some data with parasitic gaps, which were previously unaccounted for. Furthermore, the question whether locality could also have a lower bound has been formulated recently by authors such as Bosˇ kovic´ (1994), Bosˇ kovic´ (1997b), Saito and Murasugi (1999) and Grohmann (2003), among others. The proposed answer is that dependencies, such as movement, must also cover a minimum distance in order to be legitimate. That is, there is an anti-locality condition that dependencies must satisfy. Let me illustrate this line of research with an example. Consider the data in (15): (15) a. *John likes b. John likes himself

1.3 Theoretical Framework

13

According to recent work on thematic relations and checking theory (Bosˇ kovic´ (1994), Bosˇ kovic´ and Takahashi (1998), Hornstein (1999), among others), theta roles are formal features that need to be checked in an appropriate configuration. These recent proposals allow us to reformulate early generative analyses, such as Lees and Klima’s (1963), which considered local anaphors to be the result of the movement of an argument and of the application of a reflexivization rule to the initial trace of the argument. In other words, if movement into theta positions is allowed, a possible derivation of (15b) (schematized in (16)) could follow the insights of Lees and Klima’s (1963) proposal: (16)

[VP likes John] [VP John likes John] [VP John likes himself]

Movement of John Reflexivization rule

Combining this type of analysis with the standard assumptions regarding deletion of copies, namely, that copies are identical and that only one of the copies survives deletion at the PF interface, the ungrammaticality of (15a) is unexpected. Assuming that there is nothing in the theory of grammar that prevents movement of arguments into theta positions, as argued by Bosˇ kovic´ (1994), Bosˇ kovic´ and Takahashi (1998), Hornstein (1999), among others, John could undergo overt movement from the theme to the agent position. As usual, the head of a chain, namely, the higher copy of John, is pronounced in PF. Grohmann (2003), following Bosˇ kovic´ (1994), approaches this problem by assuming that there must be an independent constraint on the theory of grammar that prevents (14a) and allows (14b). Grohmann (2003) assumes, as originally proposed by Bosˇ kovic´ (1994), that the relevant constraint is related to the length of the movement that the argument underwent. In other words, the movement of John in (16a), illustrated in (17), is too short, which prevents the deletion of the lower copy to avoid violating anti-locality (i.e., the ban on movement that is too short; see the discussion below). (17)

vP v’

AG v

VP

V

TH

The main hypothesis is that, as John moves internally within a clausal internal domain (i.e., within the y-domain, see below), the movement is too short and the sentence can be grammatical only if the lower copy of John is spelled out as a different phonological realization of John at PF; namely, as an anaphor.14 Grohmann (2003) discusses different cases of ill-formed movement in the clausal domain. His observations on the length and type of movements

14

1

Introduction

disallowed in the clausal domain lead him to argue for the existence of different clausal internal domains, where anti-locality applies. Crucially, these internal clausal domains correspond to the original tripartite sentence structure prior to the proliferation of functional categories (cf. Pollock (1989), Rizzi (1997), among many others). More concretely, this tripartite structure assumes that the clause is divided into three domains: a thematic domain, which contains the predicate and its arguments; an agreement domain, where arguments can receive case and phi-features; and a discourse domain, where discourse information is encoded. Each of these domains has a specific role in the clause structure, can possibly contain a number of functional categories and constitutes a prolific domain. Each prolific domain forms a part of the derivation where the PF and LF components evaluate the derivation.15 Then, Grohmann (2003) reinterprets previous proposals on the existence of domains in the sentence as local domains for movement. A movement16 within any of these domains is too short and constitutes a violation of anti-locality. This violation of anti-locality would cause ungrammaticality, since the same element would appear twice in the same prolific domain. Anti-locality violations can be grammatical only when the duplication of the element is avoided by a last resort procedure that yields a drastic effect on the output17, such as the realization of the copy with a different PF matrix. Note, however, that realizing the copy with a different PF matrix is not always a possibility, since there are not anaphoric elements that can realize all the copies of all the possible movements in the different languages. In those cases, movement must be ruled out in order to preserve the grammaticality of the particular construction in a given language. Such a conception allows us to rule out ungrammatical cases, which otherwise would have to invoke a number of additional conditions, mainly in the form of criteria and filters.18 In this book, I provide empirical support for recent proposals on locality conditions, such as Manzini’s (1994) Locality and Grohmann’s (2003) AntiLocality, from the syntactic properties of Spanish nominal expressions. The book will illustrate that the major syntactic properties observed in Spanish nominal expressions can receive a straightforward explanation if we assume that phrasal movements cannot cross more than one maximal projection (cf. Manzini’s (1994) Locality) and that phrases cannot move within the boundaries of a particular domain (cf. Grohmann’s (2003) Anti-Locality). The next section details the organization of the book and advances the main issues dealt with in each chapter.

1.4 Overview of the Book This book is divided into five chapters, besides the present one. Chapter 2 and 3 define the basic structure of Spanish nominal expressions used throughout the book and Chapter 4 and 5 test this structure with further data. The final

1.4 Overview of the Book

15

chapter, Chapter 6, summarizes the main points discussed in the book and points to new lines of research on nominal expressions. In this section I provide an overview of the main topics discussed in each chapter and the conclusions reached. Chapter 2 focuses on the structural relations among the elements within nominal expressions, paying special attention to the prepositional phrases (PPs) within the nominal expression and to their possibilities of extraction. In order to do so, the chapter introduces some other general background about Spanish nominal expressions. First, the chapter starts with a description of the types of arguments and modifiers allowed in these constructions (namely, adjectives, PPs and genitive pronouns). The study of the main properties of arguments and modifiers in Spanish nominal expressions will lead me to argue for the existence of two groups of PPs in Spanish: syntactic arguments and syntactic adjuncts. Crucially, it will be shown that many of the PPs that were previously classified as adjuncts behave as arguments. This classification is supported by the (im)possibility of extraction out of Spanish nominal expressions. Furthermore, this chapter proves that there is a structural hierarchy of argument PPs that is represented in the syntax proper. This hierarchy is supported by binding and extraction facts and is not affected by linear word order. The analysis of these data enables me to determine the basic design of Spanish nominal constructions and to explain the structural relations available in them. The structure introduced in Chapter 2 is based on two main assumptions: first, I pursue Abney’s (1987) DP hypothesis and assume that the internal structure of nominal expressions resembles the internal structure of clauses in its functional configuration. Furthermore, following Grohmann’s (2003) interpretation of the division of the clause structure into three domains and its extension to the nominal domain proposed in Grohmann and Haegeman (2002), I assume that the nominal domain is split into three domains: a thematic domain, which contains the noun and its arguments; an agreement domain, where the arguments can receive case and phi-features; and a discourse domain, where the discourse information is encoded. Moreover, based on the results from the earlier analyses of Spanish nominal expressions, I claim that phrasal movements within them are extremely local. That is, I propose that phrasal movements within nominal expressions cannot cross more than one maximal projection. This assumption will be combined with the anti-locality hypothesis (cf. Bosˇ kovic´ (1994), Bosˇ kovic´ (1997), Grohmann (2003), among others). More specifically, I will assume Grohmann’s (2003) implementation of the anti-locality hypothesis, which postulates that phrases cannot move within the boundaries of any of the three prolific domains he establishes in the clause. The analysis of nominal expressions in Chapter 2 then derives the basic properties of certain phrasal constituents (i.e., PPs) within the nominal expression and predicts well observed generalizations in Spanish. For instance, this analysis is able to derive straightforwardly that the only PP that can be extracted from a nominal expression containing more than one PP is

16

1

Introduction

the one that occupies the highest position in the thematic hierarchy (possessor>agent>theme). In Chapter 3, the focus is shifted to the highest part of the structure proposed in the book. The chapter discusses the location available to the different types of determiners depending on their meaning and the interaction between the type of determiner (and the semantics it introduces) and the overall syntactic properties of the structure. Regarding this issue, the main claim of this chapter is that the presence or absence of the functional category DP in the structure will affect the movement possibilities in the structure, due to its interaction with the locality conditions on movement analyzed in Chapter 2. Consequently, this chapter also tests the basic structure given in Chapter 2. In fact, it examines the explanation for the extraction possibilities out of Spanish nominal expressions with some previously unobserved examples that escape the basic patterns of extraction of wh-elements in Spanish nominal expressions. More concretely, I investigate in this chapter the properties of wh-extraction from nominal expressions headed by the definite article in Spanish and compare them to the cases of so-called specificity effects in the literature. Surprisingly, extraction out of nominal expressions headed by the definite article is banned only with certain PP arguments; while extraction out of Spanish specific nominal expressions is always forbidden. This difference is attributed in this chapter to the features contained in the category D in the structure and to the peculiar properties of the definite article in Spanish. I will assume that there are two features that are involved in the relevant contrasts: definiteness and specificity. Notions such as definiteness (or indefiniteness) and specificity (or lack of it) play a fundamental role in the discussion and analysis of Chapter 3 and there are brief introductions along the chapter to familiarize the reader with the main questions pertaining definiteness and specificity and with the different ways of representing these notions in Spanish. The claim to this respect is that the definite article in Spanish can have both features (definiteness and specificity), which will produce the blocking effects, or just the definiteness feature, which will result in the possibility of extracting an argument out of the structure. Furthermore, the presence or absence of these features corresponds with the projection (or not) of the functional category DP in the structure, which is in turn responsible for the specificity of the nominal expression and for the blocking effects referred to as specificity effects. Then, to the extent that it is correct, the main claim put forward in this chapter is coherent with recent attempts to divide further semantic notions such as definiteness (cf. Lyons (1999), Julien (2005), among many others). Chapters 2 and 3 show that a considerable amount of Spanish data can be accounted for in a principled way if we assume the basic structure detailed in this part of the book. The reminder of the book will be devoted to provide additional empirical confirmation of my model of nominal constructions by examining the properties of attributive adjectives and nominal ellipsis in Spanish.

1.4 Overview of the Book

17

In Chapter 4, I analyze the syntactic and semantic properties of Spanish attributive adjectives. More concretely, I investigate the position of Spanish attributive adjectives within the nominal construction, their internal structure and their order relative to the noun and to other adjectives and constituents in the nominal expressions. The contribution of this chapter to the open discussions in the field is mainly based on the examination of the Spanish facts and on how they interact with the overall properties of Spanish nominal expressions. The results of this investigation strongly suggest that the different types of adjectives enter in different positions in the derivation. Therefore, I will be arguing for the need of a nonuniform theory of adjectives in the nominal expressions. More concretely, I will be claiming in Chapter 3 that some adjectives are generated as adjuncts while some other adjectives are generated as specifiers, depending on their semantic properties. Thus, the hypothesis is that the different syntactic and semantic properties of adjectives are derived from their different generation sites. In other words, my analysis of attributive adjectives in Spanish nominal expressions is based on the idea that the position at which attributive adjectives enter the derivation, which depends on their semantic properties, determines the syntactic properties of adjectives and the nominal expression containing them. I divide Spanish adjectives into three major groups: one group of adjectives is placed as the specifier of NP, the second group of adjectives appears as adjuncts to NP and the third group behaves as PP arguments, appearing in the position assigned to these elements in Chapter 2. More concretely, I argue that prenominal attributive adjectives (i.e., adjectives that do not combine directly with the noun that they modify and must move to get their interpretation) are generated as the specifier of NP, while postnominal attributive adjectives (i.e., adjectives that combine directly with the N that they modify and remain in situ in LF) are adjoined to NP. Moreover, relational adjectives, which behave as PP arguments, will receive a treatment similar to PP arguments. The adoption of the multiple generation sites for adjectives in the analysis presented in Chapter 4 explains a number of facts and properties of adjectives without postulating the application of syntactic mechanisms, such as N or NP movement, to derive the properties of adjectives in this language. Additionally, the analysis pursued in this chapter contributes to the general discussion on linguistic theory by providing some additional empirical evidence for the need to have a distinction between adjuncts and specifiers in the theory of grammar. As for the cross-linguistic debate, this chapter simply offers some speculations on how my analysis can be expanded to some other languages. This way, taking the predictions of the analysis regarding the relative height of the different adjectives seriously, I show that the general pattern (partially) holds in different types of languages and that the differences in linear order can be attributed to a difference in the phonological realization of adjuncts, which can be spelled out preceding or following the structure they are adjoined. This analysis then argues against N movement as an explanation for the crosslinguistic differences found and points to a new direction to explore.

18

1

Introduction

Chapter 5 examines the properties of nominal elision in Spanish and their relation to other elements within the nominal expression, to classify the phenomenon and to be able to provide an analysis that extends the empirical coverage of the current analysis of nominal expressions in Spanish. In doing so, some of the fundamental open questions regarding ellipsis in natural languages will be addressed. For instance, the question that pertains to the nature of the omitted elements in ellipsis data; that is, the structure (or lack of it) of the ellipsis site. Syntactic analyses of ellipsis processes have developed in two main directions. One view argues that there is no syntactic structure in the ellipsis site. That is, this line of research argues that the computational system generates empty categories in place of the elided material. After the derivation is sent to the interfaces, the meaning is recovered from context at the LF interface. A second line of thought claims that there is a full internal structure in the ellipsis site at some point of the derivation. The fundamental idea under this approach is that the computational system derives a full structure in place of the elided material and sends it to both interfaces. At the LF interface, the structure is interpreted. At PF, an operation that deletes the redundant phonetic and phonological material creates the ellipsis site. Under this view, ellipsis is a PF phenomenon. A second question discussed in this chapter is the question regarding the type of gap created by the ellipsis process in the nominal construction. Some authors, Bernstein (1993) and Valois (1991), among others, suggested that only maximal categories were affected by the ellipsis process (i.e., NP ellipsis, XP ellipsis), while some others, Jackendoff (1971) and Valois (1991), among others, postulated that the omission was affecting smaller structures in the nominal constructions (hence, N-ellipsis, head ellipsis, or N’ ellipsis). In Chapter 5 I provide some answers to those questions. First, I will provide strong evidence for the hypothesis that argues for having syntactic structure in the ellipsis site. The main reason to argue for an analysis of ellipsis as a PF phenomenon will come from the interaction between nominal ellipsis properties and extraction facts discussed in Chapter 2 and 3. As will become clear in Chapter 5, the only successful account for nominal ellipsis in Spanish is the one that allows some remnant elements move outside of their maximal projections prior to the application of the ellipsis operation at PF. Second, I compare the different cases of omission and conclude that only an approach assuming that nominal ellipsis is XP ellipsis (i.e., ellipsis of maximal categories) will be able to account for all properties discussed in the chapter. Therefore, the data and discussion in this chapter will also provide strong empirical evidence to the line of research that claims that nominal ellipsis is XP ellipsis. Furthermore, in this chapter I will also elaborate on the role that notions such as agreement and focus have in ellipsis processes. As we will see, most of the previous analyses of ellipsis have assumed that agreement has a licensing role in this construction. In this chapter, I claim that the differences between nominal ellipsis in expressions with or without the definite article in Spanish are

Notes

19

proof that agreement cannot be the only licensor for the construction. As for the role of focus in the construction, the movement restrictions detailed in Chapter 2 defy an analysis of NP ellipsis based exclusively on focus as a licensor via movement. In a nutshell, in Chapter 5, I claim that Spanish NP elision is a PF deletion phenomenon, which is applied after some stylistic PF operations. In particular, I maintain that elements generated within NP may move outside of the NP prior to the application in PF of the ellipsis operation. I also argue that in addition to NP ellipsis, Spanish has a surface anaphora that should be treated in terms of LF copying. Finally, to account for the restrictions of nominal ellipsis in definite DPs, I follow previous analyses, such as Brucart and Gra`cia (1986) and Raposo (1999) and assume that there is a syntactic constraint involved in these cases, which implies the boundaries of phases. The final chapter of the book, Chapter 6, is a summary of the main findings. Here I review the main conclusions I have reached through the book and point to the main directions for further research.

Notes 1

2

3

4

The most recent survey on the syntax of nominal expression in the generative framework appears in Alexiadou et al. (2007). The reader should consult this book for a detailed discussion on the main research areas in nominal expressions and for data from different languages. This brief overview on the line of research linking verbs and their deverbal nominals can be completed with the discussion on the argumental properties of deverbal nominals in Alexiadou et al. (2007: 477–546). The interested reader can find there a detailed summary of many of the proposals mentioned in the text, such as Chomsky (1970), Giorgi and Longobardi (1991), Grimshaw (1990), along with extensive discussion on some other recent proposals to account for the properties of deverbal nouns, such as some approaches following the distributed morphology approach (cf. Halle and Marantz (1993)). The data illustrating the different approaches come from a range of sample languages including English, Dutch, Turkish, Greek, Hebrew and Russian, among others. The reader is also referred to Ormazabal (1991), Zamparelli (2000), Valois (1991), Bernstein (1993), Sa´nchez (1996), Malle´n (2001) and Uriagereka (2001), among many others, for different proposals that make use of functional categories within Romance noun phrases. The authors mentioned above explain different phenomena within Romance noun phrases by assuming the existence of different functional categories: Ormazabal (1991) argues for the existence of KompP in Spanish to account for certain extraction data; Zamparelli (2000) analyzes the different semantic properties of noun phrases via the introduction of several functional categories such as, for instance, kind phrase; Valois (1991), Bernstein (1993) and Uriagereka (2001) assume the existence of NumP to derive the different positions of attributive adjectives in Spanish; and Sa´nchez (1996) argues for the existence of predicate phrase and person-agreement phrase to account for attributive adjectives placement in Spanish. In addition, for a general overview of the history of the functional categories in the nominal expression in the generative framework, see Alexiadou et al. (2007: 227–281). Cf. Alexiadou et al. (2007: 247–253) for a summary of these proposals to analyze construct state constructions.

20 5

6 7

8

9

10

11

12 13

14

15

16

17

18

1

Introduction

Cf. Alexiadou et al. (2007: 253–255), Longobardi (1994) and subsequent work and Chapter 4 in this book for a more thorough consideration of noun movement as a device to explain the cross- linguistic differences between Romance and Germanic nominal expressions. Cf. Chomsky (1995) for a detailed description of the model. The determination of when spell out applies is an open issue. This issue is not relevant for the general purposes of this study. For ease of exposition, following Chomsky (2001), I assume that spell out only applies when a phase is obtained. Also, following Chomsky’s (2001) hypothesis, I assume that each phase is determined by a subarray LA1 of the selected lexical items contained in the numeration and that a subarray LA1 should contain exactly one item that labels the resulting phase. According to Chomsky (2001), only CP and vP are phases. The number of phases has been extended in the literature (cf. Bosˇ kovic´ (2002), Bosˇ kovic´ (2005), Roehrs (2002), among others) to cover DP and PP. This book assumes that DPs and PPs also constitute phases; that is, DPs and PPs are points at which spell out is applied. For expository reasons, I will not adopt here the split-IP hypothesis. See Pollock (1989) and Chomsky (1995) on this issue. Assuming that vP is also a phase, spell out should apply when the structure reaches vP. For ease of exposition, I ignore this issue. Subjacency can be defined as follows: a single instance of movement can cross at most one bounding node, where the bounding nodes are S and NP (Lasnik and Uriagereka 1988: 21, (58)). Relativized Minimality can be defined in the following way: a movement operation cannot involve X and Y over a Z that is relevantly identical to Y in the configuration . . .X. . .Z. . .Y. . . if Z c-commands Y (Hornstein et al. 2006). Cf. Chomsky and Lasnik (1993) for a summary and discussion of this issue. The success obtained by general economy principles, such as Minimize Chain Links Principle or the Minimal Link Condition, to explain facts previously explained by locality constraints such as Relativized Minimality, is not complete, as a reviewer makes me note. Given the introductory character of this part of the book, I put aside those cases. Note that the passive will involve movement from TH to spec, IP, thus, not movement within the thematic domain. Hence, the trace is not spelled out as an anaphor. Note that Grohmann (2003) differentiates between this evaluation process and the operation spell out. According to Grohmann (2003), spell out is an operation that can be applied multiple times in a derivation and that, following Uriagereka’s (1999) idea of multiple spell out, morphologizes the object at the interfaces. This morphologization causes the object to be inaccessible to further syntactic operations. However, Grohmann’s (2003) idea of evaluation consists in making available some information contained in a prolific domain to LF and PF. This evaluation process, which does not cause the derivational object (i.e., the prolific domain) to be inaccessible to further syntactic operations, does not preclude the existence and application of a spell out operation at a later stage of the derivation. Grohmann (2003) argues that the only phrasal movement possible in overt syntax is a substitution movement (i.e., movement to a spec position) and that adjuncts must be the result of base-generation (i.e., introduced by direct merge). This claim is derived from the properties of bare phrase structure. See Grohmann (2003) for details and below for a more detailed discussion of the difference between specifiers and adjuncts. I follow Grohmann (2003) and assume that adjuncts cannot be the result of movement in overt syntax. According to Grohmann (2003), a drastic effect in the output is the overt realization of the relevant copy with a different PF matrix. Grohmann (2003) shows that the anti-locality condition also accounts for the impossibility of moving wh-elements to topic positions and for the impossibility of elements receiving two case features. For additional evidence for the ban on movement that is too short see Bosˇ kovic´ (1994), Bosˇ kovic´ (1997), Saito and Murasugi (1999), Ishii (1999), Abels (2001)

Notes

21

and Bosˇ kovic´ (2002). These authors show that the ban on movement that is too short can account for the impossibility of short-subject topicalization and short-zero subject relativization in English (Bosˇ kovic´ 1997), certain extractions out of nominal expressions (Bosˇ kovic´ 2002), the immobility of IP (Abels 2001) and the that-trace effect (Ishii 1999). I therefore assume that positing the ban on movement that is too short is empirically well motivated and adopt Grohmann’s (2003) version of this ban.

Chapter 2

On the Structure of Nominal Expressions

2.1 Introduction This chapter details the main properties of nominal expressions1 in Spanish and presents a syntactic structure to account for them.2 The study of the main properties of Spanish nominal expressions will lead us to argue for the existence of two groups of prepositional phrases (PPs) in Spanish: syntactic arguments and syntactic adjuncts. Crucially, it will be shown that many of the PPs that were previously classified as adjuncts behave as arguments. This classification is supported by the (im)possibility of extraction out of Spanish nominal expressions. A second important claim of this chapter is that there is a structural hierarchy of argument PPs that is represented in the syntax proper. This hierarchy is supported by binding and extraction facts and is not affected by linear word order. Before I can give the arguments for the two main points of the chapter, I will introduce some other general background about Spanish nominal expressions. First, the chapter starts with an overview of the types of modifiers allowed in these constructions (namely, adjectives, prepositional phrases and genitive pronouns). This basic introduction to Spanish nominal expressions is intended for readers with no previous knowledge of Spanish syntax and sets the stage for the rest of the chapter. After this brief overview, the chapter explores in detail the different structural relations among PPs appearing within nominal expressions, to provide support for the existence of an internal hierarchy of PP arguments in syntax proper. The descriptive part of the chapter concludes with a discussion on the possibilities of extraction out of Spanish nominal expressions, which supports the need for a new division of PPs within nominal expressions attending solely to their syntactic properties. The main goal of the description in this chapter is to achieve a deeper understanding of the internal dependencies allowed within the Spanish nominal expressions, which will help us to design their basic structure. Attending to the results attained in the descriptive part of this chapter and the supporting evidence for them, the structure of the Spanish nominal M.E. Ticio, Locality Domains in the Spanish Determiner Phrase, Studies in Natural Language and Linguistic Theory 79, DOI 10.1007/978-90-481-3398-7_2, Ó Springer ScienceþBusiness Media B.V. 2010

23

24

2 On the Structure of Nominal Expressions

expressions must contain at least the elements and dependencies in the schematic structure in (1) (1)

XP

Adjunct PP

XP

X’

Possessor

YP

X

Y’

Agent

Y

Object

The rough structure in (1) shows the main dependencies established among PPs in the Spanish nominal expressions. That is, there is a syntactic hierarchy among argumental PPs (possessors higher than agents and agents higher than objects) and a division of PPs into argumental (represented by the possessor, agent and object in the structure) and non-argumental PPs (represented by adjunct PP in the structure) according to their syntactic properties. The structure of nominal expressions proposed in this book follows this preliminary structure and fully develops it following well established assumptions in current research. This way, the proposal divides the nominal expressions structure into three basic domains (i.e., discourse, agreement and thematic related domains) and accounts for the different internal dependencies among the different constituents of nominal expressions by a combination of general economy principles of the grammar and of the properties of the different generation sites of constituents. The assumptions mentioned above give rise to a basic structure of nominal expressions, cf. (2), which generates in different domains the different constituents allowed within Spanish nominal expressions according to their feature composition.

2.1 Introduction

25

(2)

FocP Foc’ ω-domain(=discourse domain) Foc

DP D’

D

AgrP ϕ-domain(=agreement domain) Agr’

Agr

POSSESSOR

nP

n’

AGENT θ-domain(=lexical domain)

n

NP

NP

ADJUNCTS

| N’ N

OBJECT

Then, this chapter will be postulating the existence of a parallelism between the internal structure of nominal expressions, represented in (2) and the internal structure of clauses with respect to their functional configuration. In particular, it follows the division of clause structure into three domains (Chomsky (1986), Rizzi (1997), Platzack (2001), Grohmann (2003), among others) and extends it to the nominal structure (see also Grohmann and Haegeman (2002) for a similar approach). The structure of nominal expressions is then split into three domains: a (thematic) y-domain, which contains the noun and its arguments; an (agreement) j-domain, where the arguments can receive case and phifeatures; and a (discourse) o-domain, where the discourse information is encoded. Chapter 2 is then organized as follows. The chapter is divided into two main parts: a descriptive part (Sections 2.2 and 2.3) and a theoretical part (Section 2.4). Section 2.2 is an overview of the basic properties of Spanish nominal expressions. I discuss here basic facts regarding the types and number

26

2 On the Structure of Nominal Expressions

of constituents appearing by the noun and their interactions. More concretely, I show that we can have adjectives, possessive pronouns and prepositional phrases within the Spanish nominal expressions. I also briefly describe in Section 2.2 the basic linear word order of constituents in Spanish nominal expressions and point to some PF restrictions as the cause of the different linear word orders available. As will be shown, full PP constituents cannot occupy the prenominal position in Spanish, while genitive pronouns and adjectives can. The examination of different types of facts will ultimately lead us to the conclusion that this linear word order restriction is the result of the different prosodic properties between full PPs and genitive pronouns and adjectives. In Section 2.3, I focus on one of the different types of elements allowed in the Spanish nominal expressions; namely, on PPs. First, I investigate the c-command relations among PPs within Spanish nominal expressions. The goal of Section 2.3.1 is to determine the relative location of PPs in the nominal structure. The analysis of the c-command relations shows the existence of a thematic hierarchy among the nominal PP arguments holding in the syntax proper, which is not affected by the linear word order. Section 2.3.2 examines the different types of PP modifiers in Spanish nominal expressions and classifies them as syntactic arguments or syntactic adjuncts depending on their behavior with respect to some syntactic operations, such as wh-extraction from nominal expressions. The goal of this section is to determine the influence that the different generation sites have in the properties of PPs within Spanish nominal expressions. The existence of a unified PP argumental group and a unified PP adjunct group enables us to improve upon our model of the Spanish nominal expressions and to explain basic properties, such as the unavailability of movement for adjuncts, which were left unexplained in previous analyses. Therefore, the conclusions of the first part of this chapter (Sections 2.2, 2.3) will state the major descriptive generalizations to be explained by the theoretical system. The goal of Section 2.4 is precisely to explain how the basic structure for Spanish nominal expressions postulated (see (2) above) combines with some other components of our grammar to generate the set of data exposed herein. The proposed structure maintains an internal division similar to the one postulated in recent research for the clause structure (cf. Chomsky (1986), Rizzi (1997), Platzack (2001), Grohmann (2003), among others) and enters the different elements according to their feature composition. Furthermore, following the evidence from the earlier analyses of Spanish nominal expressions (Torrego (1987) and Ormazabal (1991), among others), I claim that phrasal movements within these constructions are extremely local. That is, I propose that phrasal movements within nominal expressions cannot cross more than one maximal projection (cf. Manzini (1994), Fox and Lasnik (2003)). This assumption will be combined with the anti-locality hypothesis (cf. Bosˇ kovic´ (1994), Bosˇ kovic´ (1997), Grohmann (2003), among others). Thus, I will be also assuming the anti-locality hypothesis (as stated in Grohmann (2003)), which postulates that phrases cannot move within the boundaries of a particular domain.

2.2 Basic Properties of Spanish Nominal Expressions

27

The chapter ends with a brief appendix where some particular properties of genitive pronouns are discussed. The combination of the results obtained in the different sections of this chapter and of the general assumptions of our theory of grammar is conducive to an analysis that has as its base the structure of nominal expressions advanced in (2). This structure enables us to explain all the descriptive generalizations established in the first part of this chapter as the consequence of the types of movement allowed in the Spanish nominal expressions and the feature composition of the elements projected within the nominal expressions.

2.2 Basic Properties of Spanish Nominal Expressions Nominal expressions in Spanish can host several types of modifying phrases such as adjectives, genitive pronouns3 and PPs, as illustrated in (3a–c). Contrarily to what happens in some other languages, full (non-pronominal) nominal expressions modifiers cannot appear within another nominal (3)

a. El alto chico/ el chico alto The tall boy / the boy tall b. Su chico/ El chico suyo his/her boy/ the boy of her/his c. El chico de Marı´ a The boy of Mary

adjective genitive pronoun PP

expression. It is impossible to say something like (4) with any linear word order. (4)

*Marı´ a el chico/ *El chico Marı´ a/ *el Marı´ a chico ‘Mary’s boy’

In Spanish, all modifiers4 of the noun other than genitive pronouns must be expressed in a PP, which must appear postnominally. Thus, one of the most obvious properties of Spanish nominal expressions is that no full PP or NP modifier can occupy the prenominal position. As illustrated in (3a) and (3b), adjectives and genitive pronouns are an exception to this linear word order restriction and both types of elements can appear pre or postnominally. Accordingly, it is possible to say su chico (his/her boy) with the genitive pronoun appearing prenominally and un alto chico (a tall boy) with the adjective preceding the noun. The pre or postnominal position of the adjective is closely related to its meaning and syntactic properties and it will be extensively discussed in Chapter 4. I will show in that chapter a number of tests that prove that prenominal adjectives are generated in a different position than their

28

2 On the Structure of Nominal Expressions

postnominal counterparts. Specifically, I will defend that prenominal adjectives are generated in a specifier position while postnominal adjectives occupy adjunct positions. The interested reader can go to that chapter for a full discussion of these issues. As for genitive pronouns, their behavior cannot be equated with the behavior of all full PPs in the Spanish nominal expressions.5 This is because the meanings available to genitive pronouns in Spanish seem to be reduced to the possessor, agent or object, as illustrated in (5). (5)

Compramos (we)bought

su foto her/his picture (possessor)/(agent)/(theme).

The underlined nominal expression in (5) shows that the genitive pronoun su (his/her) can be understood exclusively as the owner, author or theme of the picture. Therefore, it seems to be more accurate to relate genitive pronouns with a subkind of full PPs to determine their exact structure: the PPs that will be referred to later as PP arguments (see Section 2.3.2 for discussion on PP arguments). Therefore, the behavior and properties of genitive pronouns set them apart from the rest of elements available in the Spanish nominal expressions and will be discussed in the appendix to this chapter for ease of exposition. I now focus on the PPs that can accompany the noun (henceforth, N) in Spanish. Most PP modifiers in a nominal expression generally show up marked with the preposition de (of). Thus, the example in (6) is in three ways ambiguous: (6)

El libro de Marı´ a (agent/theme/poss) the book of Marı´ a ‘The book by Marı´ a/ The book about Marı´ a/ Marı´ a’s book’

As shown by the English translation, the modifier introduced by de (of) in (6) can be understood as the agent, theme or possessor6 of the N libro (book). Although de (of) is the most common P in nominal modifiers, other Ps can also introduce nominal modifiers in Spanish: (7)

´ de Roma por los ba´rbaros a. La destruccion the destruction of Rome by the Barbarians. b. El camino hacia Burgos the path towards Burgos c. La historia sobre Chomsky the tale on/about Chomsky

The example in (7a) shows that agents can appear as complements to the P por (by), (7b) illustrates that locatives can appear as complements to P other than de

2.3 Structural Relations Within Nominal Constructions

29

(of) and (7c) illustrates that some themes can be expressed via the P sobre (on/ about). Note that all the examples in (7) are also acceptable with P de (of): ´ de Roma de los ba´rbaros (8) a. La destruccion the destruction of Rome of the Barbarians. b. El camino de Burgos the path of Burgos c. La historia de Chomsky the tale of Chomsky Another interesting property of Spanish nominal expressions is also related to the linear order in which the different PPs appear within a nominal expression.7 The descriptive generalization stated in Ormazabal (1991) is that, when more than one PP appears to modify the N, these elements can show up in almost any order to the right of the N head: (9) a. Pedro conocı´ a [el retrato [de las Meninas]obj [de Vela´zquez]ag] Pedro knew the portrait of the Meninas of Velazquez b. Pedro conocı´ a [el retrato [de Vela´zquez]ag[de las Meninas]obj ] Pedro knew the portrait of Velazquez of the Meninas (Ormazabal (1991: 23)) The above examples illustrate the apparently free linear order that PP modifiers have in Spanish nominal expressions. In this chapter (section 3.1.3) I will claim that this free linear order can be analyzed as the result of a stylistic rule (i.e., a rule that changes the order of constituents at PF without affecting the syntactic hierarchy or LF interpretation). In the next section (3), I will provide some evidence for this claim. The evidence comes from the interaction between linear order and syntactic operations, such as binding and supports the idea of a stylistic rule that must operate after all overt syntactic operations have been applied. Now that I have introduced the basic properties of Spanish nominal expressions, I will focus on the PP constituents and examine how the most common types of PPs can relate to each other in Spanish nominal expressions.

2.3 Structural Relations Within Nominal Constructions Any syntactic analysis starts by determining the structural dependencies among the elements involved in the construction to be analyzed. This section deals precisely with this issue. First, in Section 2.3.1, I point to the syntactically vacuous character of the preposition de (of) in Spanish and provide evidence for this idea by examining some binding and quantification constructions in Spanish nominal expressions. Furthermore, I take into consideration the role of linear order into the previously mentioned phenomena

30

2 On the Structure of Nominal Expressions

and show that the syntactic hierarchy proposed in this section is not affected by superficial word order differences, since it holds in the proper syntax. The second part of this Section 2.3.2 analyzes the properties of the different types of PPs from a syntactic point of view. I will argue here for a division of PPs into arguments and adjuncts, according to their behavior regarding the basic patterns of extraction of wh-elements in Spanish nominal expressions. Furthermore, the extraction facts will provide some additional support for the idea that the thematic hierarchy holds in syntax by showing that some other syntactic operations, such as wh-movement, are also sensitive to the previously mentioned thematic hierarchy.

2.3.1 Structural Hierarchy Within Nominal Expressions The first part of this Section 2.3.1 investigates several phenomena, such as licensing of anaphors and pronouns and quantifier-variable readings, which can be used as tools in determining the basic structure of Spanish nominal expressions. More specifically, I explore the c-command relations among the elements that form these constructions to illustrate the vacuous syntactic character of the preposition de (of) in Spanish. Based on this data, I also argue that the syntactic hierarchy of argument PPs follows a thematic hierarchy (i.e., possessors are structurally higher than agents and agents are structurally higher than themes) within the Spanish nominal expressions, which holds in the proper syntax, since it is not affected by the different linear word order. Finally, I provide a summary that compiles the information in the subsection regarding the syntactic hierarchy established among the PPs in the Spanish nominal expression. 2.3.1.1 On the Different Types of Prepositions As is well-known, some of the morphological elements in languages do not correspond to their a priori syntactic classification. For instance, Romance linguistics has claimed repeatedly that the preposition a (to) in its non-locative use is an accusative and specificity marker (see Chapter 3 for a discussion of P a (to) as a specificity marker) and does not project a full PP on its own. The evidence for this last claim comes from the c-command possibilities of nominal expressions appearing with the non-locative a (to). Consider the obligatory control structure (cf. Williams (1980), Hornstein (1999) on control structures) in (10), where a syntactic relation is established between the object of the verb prohibir (forbid) and the subject of the verb leer (read). (10)

Juan prohibio´ a Marı´ a leer libros de Chomsky John forbade to Mary to read books by Chomsky ‘John forbade Mary to read books by Chomsky’

2.3 Structural Relations Within Nominal Constructions

31

Example (10) can be only understood as meaning that Marı´a is the subject of the verb leer (read); that is, the grammaticality of (10) shows that an object can c-command out of its apparent PP to establish a coreference relation with the null subject of the verb in the subordinate clause. Note that this is unexpected under standard definitions of c-command, since the PP should block this relation.8 Therefore, this type of example proves that nominal expressions headed by the non-locative a (to) do not behave as PPs but as nominal expressions, from a structural point of view. In other words, the examples in (10) show that the non-locative a (to) in languages such as Spanish must be considered a dummy P, not projecting a PP. Keeping the previous discussion in mind, I now return to the nominal domain. As shown in Section 2.2, the P de (of) is the most common P in Spanish nominal expressions and can express a variety of semantic notions. Interestingly, some authors, see Martı´ n (1995) on this issue, have claimed that de (of) is also a dummy P in Spanish. Their claim is supported by examples such as (11) where a binding relation is established between two PPs. (11)

´ [de sı´ mismoi]Theme [de Juani]Agent La descripcion the description of himself of Juan ‘Juan’s description of himself’

As (11) illustrates, the referential expression Juan can c-command outside of the PP in which it appears. This is the reason why the appearance of the anaphor sı´ mismo (himself) does not induce ungrammaticality in (11), since it is bound in its minimal domain, the nominal expression, by the referential expression Juan. Hence, the Spanish P de (of) also behaves as a dummy P, since it allows the referential expression to bind (i.e., to c-command) outside of the PP in which it appears. In contrast, note that other prepositions do not allow binding out of their PPs, as the example in (12) illustrates. (12)

´ [de sı´ mismo i] Theme [por Juani] Por-Agent *La descripcion the description of John by himself ‘Himself’s description of John’

The example in (12) expresses the same meaning as (11) but with a crucial change: the agent is now introduced by the P por (by). Note that, surprisingly, the referential expression Juan is not able to bind (i.e., to c-command) outside of the PP headed by the P por (by) in which it appears. Following the discussion above, the conclusion is then that Ps other than de (of) do not allow binding out of their PPs. Note that this seems to be also the case with some other Ps in some other Romance languages. As Giorgi and Longobardi (1991) have shown in the examples below, some nominal arguments in Italian introduced by di (of) do not behave as PPs either, since c-command out of these PPs is possible.

32

(13)

2 On the Structure of Nominal Expressions

L’opinione di se stessoi di Giannii e` troppo lusinghiera the-opinion of himself of Gianni is too flattering ‘Gianni’s opinion about himself is too flattering’

Then, the grammaticality of the example in (13) illustrates that the referential expression Gianni can c-command outside of the PP in which it appears. This is the reason why the appearance of the anaphor se stesso (himself) does not induce ungrammaticality in (13), since it is bound in its minimal domain, the nominal expression, by the referential expression Gianni.9 Now that I have established that de (of) is a dummy P in Spanish, the rest of this subsection will be devoted to explore the relationships available among the different types of PPs in the nominal expressions. In Section 2.3.1.2 I explore the main types of evidence to support the syntactic hierarchy in the Spanish nominal expression; namely, binding and quantification constructions. 2.3.1.2 On a Structural Asymmetry in Spanish Nominal Expressions Given that languages such as Spanish can have multiple PPs within a particular nominal expression, the question arises as how to determine the structural relations among them. As known, there is a structural asymmetry in the clause among the arguments, with subjects higher than objects for instance. There are many types of evidence for this clausal asymmetry in the literature but the most notorious one is the evidence that comes from binding and quantifier-variable facts. Let us consider the examples in (14), where we have an anaphor in the subject or object position, which is coreferent with a referential expression in the subject or object position. (14)

a. Johni loves himselfi b. *Himselfi loves Johni

The explanation for the different grammaticality judgments follows straightforwardly from any standard version of the binding theory. That is, assuming the (simplified) definitions of principles A, B and C in (15), the grammaticality of (14a) can be explained as the fulfillment of principles A and C, while the ungrammaticality of (14b) involves a violation of those two principles. As for (14a), the situation is that the referential expression is free and the anaphor is bound in its local domain. The ungrammaticality of (14b) is due to the lack of available binder for the anaphor and binding to the referential expression. (15)

Principle A: anaphors cannot be bound in their local domain Principle B: a pronoun cannot be bound in its local domain Principle C: referential expressions must be free

Crucially, the notion of binding involves the c-command possibilities in a structure, providing us with an accurate test to determine hierarchical

2.3 Structural Relations Within Nominal Constructions

33

structure. This way, the examples in (14) are also illustrating that subjects are hierarchically higher than objects in the syntax proper.10 This subsection’s goal is to determine the hierarchical structure of the Spanish nominal expression. To do so, I will be using binding and quantifier-variable constructions to identify the relative position of PPs within nominal expressions in the syntactic component. Binding As previously mentioned (cf. Section 2.3.1.1), the P de (of) is a dummy P in Spanish. One of the tests that authors, such as Martı´ n (1995), have put forward to demonstrate this claim involves the examination of binding relations inside the nominal expression. The relevant examples are below, where a binding relation is established between two PPs. ´ [de sı´ mismoi]Theme [de Juani]Agent es demasiado halagadora (16) a. La opinion the opinion of himself of Juan is too flattering ‘Juan’s opinion about himself is too flattering’ ´ [de e´li] Theme [de Juani]Agent es demasiado halagadora b. * La opinion the opinion of him of Juan is too flattering ‘Juan’s opinion about him is too flattering’ The examples in (16) illustrate that the Spanish P de (of) behaves as a dummy P, since it allows the referential expression to bind outside of the PP in which it appears. The contrasts on the grammaticality of the examples in (16) illustrate that the referential expression Juan can c-command outside of the PP in which it appears. This is the reason why the appearance of the anaphor sı´ mismo (himself) does not induce ungrammaticality in (16a), since it is bound in its local domain, the nominal expression, by the referential expression Juan. On the contrary, the presence of the pronoun e´l (he) in (16b) causes the ungrammaticality of the sentence, as the pronoun is bound within its local domain, the nominal expression, causing a principle B violation (see (15) above). In addition, the examples in (16) also help us to understand the relationships between PPs in the nominal expressions. In other words, examples as the ones in (16) illustrate that agents must be in a higher structural position than themes in the nominal structure. Interestingly, the binding possibilities of nominal arguments are not affected by the linear order in which they occur: (17)

´ [de sı´ mismoi]Theme [de Juani]Agent a. La descripcion the description of himself of Juan ‘Juan’s description of himself’ ´ [de Juani]Agent [de sı´ mismo i]Theme b. La descripcion the description of Juan of himself ‘Juan’s description of himself’

34

2 On the Structure of Nominal Expressions

The examples in (17) illustrate that the anaphor sı´ mismo (himself) can be bound by the referential expression Juan expressing the agent or possessor in the nominal expressions. Note that the alteration of the surface linear order does not affect the grammaticality status of the sentence in (17b). Hence, it seems that the different surface linear orders do not interfere with the binding possibilities of nominal arguments in Romance languages. In Section 2.3.1.3 I discuss in more detail the surface order possibilities in Spanish nominal expressions. Consider now the following examples with binding relations between agent PPs and theme PPs: (18)

´ a. La llamada telefonica [de Juan i]Agent [a sı´ mismo i]Theme the call phone of Juan to himself ‘Juan’s phone call to himself’ ´ b. La llamada telefonica [a sı´ mismo i]Theme [de Juan i]Agent the call phone to himself of Juan ‘Juan’s phone call to himself’ ´ c. *La llamada telefonica [de sı´ mismo i]Agent [a Juan i]Theme the call phone of himself to Juan *‘Himself’s phone call to Juan’ ´ d. *La llamada telefonica [a Juan i]Theme [de sı´ mismo i]Agent the call phone to Juan of himself *‘Himself’s phone call to Juan’

The grammaticality of (18a) and (18b) shows that an element expressing the agent in the nominal expression can also bind the anaphor introduced by the P a (to), independently of the linear order in which the anaphor appears. Furthermore, the ungrammaticality of (18c) and (18d) shows that there is an internal hierarchy among the nominal arguments: it is not possible to have an anaphor expressing the agent when its antecedent is the theme. Furthermore, note that, as previously mentioned, other prepositions do not allow binding out of their PPs, as the examples in (19) illustrate. ´ [de Juan i] Theme [por sı´ mismo i] Por-Agent (19) a. *La destruccion the destruction of John by himself ‘Himself’s destruction of John’ ´ [por sı´ mismo i] Por-Agent [de Juan i] Theme b. *La destruccion the destruction by himself of John ‘Himself’s destruction of John’ ´ [por Juani] Por-Agent [de sı´ mismoi] Theme c. *La destruccion the destruction byJohn of himself ‘John’s destruction of himself’ ´ [de sı´ mismoi] Theme [por Juani] Por-Agent d. *La destruccion the destruction of himself by John ‘John’s destruction of himself’

2.3 Structural Relations Within Nominal Constructions

35

As shown by the grammaticality status of the sentences in (19), binding is not possible from an agent por (by) phrase into a de (of) theme, or vice versa. Note that the different word orders do not improve the grammaticality status of the relevant sentences. Therefore, it seems to be a special property of the prepositions a (to) and de (of) that they allow their complement to establish a structural relation with some other constituents in the sentence. To summarize: the previous discussion has shown supporting evidence for the claim that PPs that express the agent, c-command the elements that express the theme but not the other way around.11 Quantifier Variable Another phenomenon (cf. Martı´ n (1995)) that provides evidence for the structural asymmetries among the elements expressing agents, possessors and themes in Spanish nominal expressions is the distribution of pronouns bound by a quantifier. (20)

a. El cuadro [de cada coleccionistai]Poss [de sui artista favorito]Agent the picture of each collector of his painter favorite ‘Each collector’s picture of his favorite painter’ b. El cuadro [de sui artista favorito]Agent [de cada coleccionistai]Poss the picture of his painter favorite of each collector ‘Each collector’s picture of his favorite painter’ c. *El cuadro [de sui coleccionistai]Poss [de cada artista favoritoi]Agent the picture of his collector of each painter favorite *‘His collector’s picture of each favorite painter’ d. *El cuadro [de cada artista favoritoi]Agent[de sui coleccionista]Poss the picture of each painter favorite of his collector *‘His collector’s picture of each favorite painter’

In order to get a bound pronoun reading, a pronoun must be c-commanded by the quantified expression that is its antecedent. Therefore, the contrasts in grammaticality in (20) are pointing to the conclusion that elements bearing the possessor role are able to c-command elements bearing the agent role. Furthermore, note that the grammaticality of (20a) and (20b) and the ungrammaticality of (20c) and (20d) illustrate again that the linear order in which the elements surface does not affect the structural relations established in nominal expressions. As shown in the examples in (21), a c-command relation can also be established between the element bearing the possessor role and the element expressing the theme of the nominal expression, with the possessor c-commanding the theme. (21) a. El cuadro [de cada coleccionistai]Poss [de sui ciudad favorita]Theme the picture of each collector of his city favorite ‘Each collector’s picture of his favorite city’

36

2 On the Structure of Nominal Expressions

b. El cuadro [de sui ciudad favorita]Theme [de cada coleccionistai]Poss the picture of his city favorite of each collector ‘Each collector’s picture of his favorite city’ c. * El cuadro [de sui coleccionista] Poss [de cada ciudad favoritai] Theme the picture of his collector of each city favorite *‘Its collector’s picture of each city’ d. * El cuadro [de cada ciudad favoritai] Theme [de sui coleccionista] Poss the picture of each city favorite of his collector *‘Its collector’s picture of each city’ Assuming that binding applies at LF12 and is dependent on the c-command hierarchy (cf. Lasnik and Uriagereka (1988), among others for discussion on the topic), the data examined in the last two sections show that the agent ccommands the theme and the possessor c-commands the agent and the theme in proper syntax. Hence, the arguments within Romance nominal expressions follow the thematic hierarchy Possessor> Agent > Object at the level in which c-command is relevant. Moreover, I have shown that there is no interaction between the linear order in which these elements appear and all the truly syntactic operations examined above. The next subsection discusses the surface order possibilities in the Spanish nominal expressions, the main claim is that the different word orders attested do not intervene in the structure of the nominal expressions because they are the result of post-syntactic operations in the PF component.

2.3.1.3 Surface Linear Order As illustrated in Section 2.2, one of the main characteristics of Spanish nominal expressions is that the linear order in which PPs appear is apparently free. The following example shows all the linear order possibilities of a nominal expression containing three PP modifiers; namely, an object, an agent and a possessor. (22)

a. El libro de lingu¨ı´ stica de Chomsky de Juan The book of linguistics of Chomsky of Juan b. El libro de lingu¨ı´ stica de Juan de Chomsky The book of linguistics of Juan of Chomsky c. El libro de Chomsky de Juan de lingu¨ı´ stica The book of Chomsky of Juan of linguistics d. El libro de lingu¨ı´ stica de Chomsky de Juan The book of linguistics of Chomsky of Juan e. El libro de Chomsky de lingu¨ı´ stica de Juan The book of Chomsky of linguistics of Juan f. El libro de Juan de lingu¨ı´ stica de Chomsky The book of Juan of linguistics of Chomsky

2.3 Structural Relations Within Nominal Constructions

37

All the variants in (22) have the same meaning and are grammatical. Recall that I have shown in the last subsections that there is no interaction between the linear order in which these elements appear and all the truly syntactic operations examined above. I have included below a few illustrative examples, corresponding to cases of binding, (23a-b) and quantifier variable, (23c-f). (23)

a. La pelı´ cula [de sı´ mismoi]Theme [de Juani]Agent The film of himself of Juan La pelı´ cula [de Juani]Agent [de sı´ mismo i]Theme The film of Juan of himself ‘Juan’s film of himself’ b. *La pelı´ cula [de sı´ mismoi]Ag. [sobre Juani] Theme/ The film of himself about Juan *La pelı´ cula [sobre Juani] Theme[de sı´ mismoi]Ag. The film about Juan of himself *‘Himself’s film about Juan’ c.. La foto [de cada nin˜oi]Poss [de sui mascota]Theme the picture of each boy of his pet ‘Each boy’s picture of his pet’ d.. La foto [de sui mascota]Theme[de cada nin˜oi]Poss the picture of his pet of each boy ‘Each boy’s picture of his pet’ e. * La foto [de sui nin˜o] Poss [de cada mascotai] Theme the picture of his boy of each pet *‘Its pet’s picture of each boy’ f. * El cuadro [de cada mascotai] Theme [de sui nin˜o] Poss the picture of each pet of his boy *‘Its pet’s picture of each boy’

The examples in (23a-b) illustrate that the anaphor sı´ mismo (himself) can be bound by the referential expression Juan expressing the agent in the nominal expressions. Note that the alteration of the surface linear order does not affect the grammaticality status of the sentence in (23b). The examples from (23c) to (23f) illustrate a similar situation regarding the relation between a quantifier cada (each) and the pronoun it binds. Again, the different linear orders shown, for instance in the cases in (23e–f), do not change the grammaticality status of the examples. Hence, it seems that we can conclude that the different surface linear orders do not interfere with syntactic processes. The previous discussion illustrated the lack of effect of the different surface word orders of nominal expressions in syntactic processes, such as binding. It follows then that the various surface word orders (in which possessors, agents and themes can occur in any order) must be the result of postsyntactic operations in the PF component. Hence, I will be assuming here that the different surface linear orders of PPs in Spanish nominal expressions are due

38

2 On the Structure of Nominal Expressions

to stylistic movements at PF and that this has no effect on other truly syntactic (or LF) operations (such as binding).13 The next section studies in detail the different types of PPs in the nominal expression and their behavior regarding extraction. The study of the behavior of wh-elements in Spanish nominal expressions provides us with additional evidence for the conclusions reached so far. 2.3.1.4 Conclusions The data examined in this subsection provide strong empirical support for the claim that the agent c-commands the theme and the possessor c-commands the agent and the theme in proper syntax in Spanish nominal expressions. Hence, the arguments within Romance nominal expressions follow the thematic hierarchy Possessor> Agent > Object at the level in which c-command is relevant. Moreover, I have shown that there is no interaction between the linear order in which these elements appear and all the truly syntactic operations examined above. It follows then that the various surface word orders (in which possessors, agents and themes can occur in any order) must be due to stylistic movements at PF that have no effect on other truly syntactic (or LF) operations (such as binding). The next subsection focuses on the different kinds of PPs in nominal expressions and studies the basic patterns of extraction of wh-elements in Spanish nominal expressions. The study of the behavior of wh-elements in Spanish nominal expressions provides us with additional evidence to determine the argumental properties of PPs appearing in Spanish nominal expressions.

2.3.2 Two Types of PPs in Spanish Nominal Expressions This section examines the different types of PPs within the Spanish nominal expression that have been previously classified as PP adjuncts or PP adverbials (cf. Ormazabal (1991: 5–6)) due to their semantic contribution to the construction. I focus on the extraction possibilities of PPs and show that, in fact, there is a subclass of PP modifiers that do not express the possessor, agent or object/theme in Spanish nominal expressions that pattern like arguments with respect to their possibilities for extraction. Therefore, I classify the PP modifiers in Spanish nominal expressions as syntactic arguments or syntactic adjuncts depending on their behavior with respect to wh-extraction from nominal expressions and other syntactic phenomena. The results of this section will then demonstrate that some PPs that were previously thought to be adjuncts are actually syntactic arguments (cf. Ormazabal (1991: 5–6)). This is crucial in order to determine the possible or impossible structural relations that the different types of elements within the Spanish nominal expression can establish. In other words: the existence of a unified PP argumental group and a unified PP

2.3 Structural Relations Within Nominal Constructions

39

adjunct group enables us to better specify the structure of the Spanish nominal expressions. This will ultimately allow us to explain basic properties, such as the unavailability of movement for adjuncts and the lack of interaction between argument and adjunct PPs, which were left unexplained in previous analyses.

2.3.2.1 Extraction Facts This subsection analyzes the basic patterns of extraction of wh-elements in Spanish nominal expressions. The goal of this subsection is to present some additional support for the idea that the thematic hierarchy (i.e., possessors are structurally higher than agents and agents are structurally higher than themes) holds in the syntax component of the grammar by showing that syntactic operations, such as wh-movement, are sensitive to the previously mentioned thematic hierarchy.14 Furthermore, the subsection will establish a series of properties that characterize the so-called argument PPs (ARG PPs); that is, the PP modifiers that express the possessor, agent or object/theme roles in Spanish nominal expressions. These properties will be used to classify some other PPs modifiers as ARG PPs or adjunct PPs in Section 2.3.2.2. The possibilities of extraction out of Romance nominal expressions depend on the type of PP present in the nominal expression.15 This way, as Ormazabal (1991) noted for Spanish and Cinque (1980) and Giorgi & Longobardi (1991), among many others, for other Romance languages, ARG PPs can be generally extracted out of a Spanish nominal expression if there is no other PP present in the structure. The examples below illustrate that objects,16 agents and possessors can be interrogated when they are the only ARG PP in the nominal expression. a. He leı´ do [varios libros [de fı´ sica] obj] (I)have read several books of physics ¿De que´ has leı´ do [varios libros [tobj]]?17 of what (you)have read several books b. He leı´ do [varios libros [de Cervantes] agent] (I)have read several books of Cervantes ¿De quie´n has leı´ do [varios libros [tagent]]? of whom (you)have read several books c. He leı´ do [varios libros [de Ana] poss] (I)have read several books of Ana ¿De quie´n has leı´ do [varios libros [tposs]]? of whom (you)have read several books Œ

Œ

Œ

(24)

Note that, as illustrated in (24), the presence of another ARG PP may affect the extraction possibilities of ARG PPs. In Spanish, as Ormazabal (1991) notes, when two or more PP arguments are present in a nominal expression, the extraction possibilities of arguments change. For instance, as the examples in (25) show,18 the presence of a possessor blocks the extraction of the object or agent.

40

(25)

2 On the Structure of Nominal Expressions

a. He leı´ do [varios libros [de Cervantes]ag [de Juan]poss] (I)have read several books [of Cervantes]ag [of Juan]poss b. *¿[De quie´n] has leı´ do [varios libros tag [de Juan]poss]? of whom (you) have read several books tag [of Juan]poss c. He leı´ do [varios libros [de fı´ sica]obj [de Juan]poss] (I)have read several books [of physics]obj [of Juan]poss d. *¿[De que´] has leı´ do [varios libros tobj [de Juan]poss]? of what (you)have read several books tobj [of Juan]poss

Hence, no agent or object can be extracted if a possessor is present in the nominal expression. Note, however, that the presence of an agent or an object in the structure has no effect on the extraction of a possessor from the nominal expression:19 (26) a. ¿[De quie´n] has leı´ do [varios libros [de fı´ sica]obj tposs]? of whom (you)have read several books [of physics]obj tposs b. ¿[De quie´n] has leı´ do [varios libros [de Cervantes]ag tposs]? Of whom (you)have read several books [of Cervantes]ag tposs] The conclusion is then that possessors must occupy a syntactic position that interferes with the path that agents and objects must follow in their extraction out of nominal expressions but not the other way around. This conclusion and the standard assumption that XP movement proceeds via specifiers positions, points to a representation of the nominal expressions on the line of (27), where I have omitted any nonrelevant details at this point. (27)

XP

X’

Possessor

YP

X

Y’

Agent

Y

Object

What (27) represents is that possessors must be placed in a specifier position higher than the position in which objects or agents appear in the

2.3 Structural Relations Within Nominal Constructions

41

structure. Besides, keeping in mind the results obtained with the binding and quantifier-variable data in the previous sections, we can add that agents must also appear in a higher position than objects in the structure. There is also additional evidence for the relative position of objects and agents coming from the extraction facts. As the examples below illustrate, object extraction is blocked by the presence of an agent, as in (28b) but the presence of an object has no effect on the extraction of an agent, as in (28c): (28) a. Conozco [varias traducciones [de La Celestina]obj (I)know [several translations of La Celestina [de escritores importantes]ag] of writers important] b. */??¿[De que´ obra] conoces [varias traducciones tobj of what work (you) know [several translations tobj [de escritores importantes]ag]? of writers important] c. ¿[De quie´n] conoces [varias traducciones [de La Celestina]obj tag]? of whom (you)know several translations [of La Celestina]obj tag Hence, it seems that agents must occupy a syntactic position that interferes with the path of objects in their extraction out of nominal expressions. This position corresponds a priori to a specifier position, as shown in (27) above. However, it is not true that all agents must occupy a specifier position in the path of object movement, since, as Ormazabal (1991) notes, if the agent is introduced by the P por (by), objects can be extracted.20 This is shown in the following examples: (29) a. ¿De que´ tema ´ tobj? has criticado una investigacion of what topic (you)have criticized a investigation por los americanos ? by the Americans ´ tobj b. */?? ¿De que´ tema has criticado una investigacion of what topic (you)have criticized a investigation de los americanos? of the Americans As the grammaticality of (29a) illustrates, only agents introduced by de (of) interfere with the movement of objects, while por (by) agents do not affect extractability of objects.21 Thus, the descriptive generalization, stated in Torrego (1987) and Ormazabal (1991), among others, is that the arguments within Romance nominal expressions (agents, possessors and objects) display a hierarchical relation, with possessors higher than agents and agents higher than objects.

42

2 On the Structure of Nominal Expressions

To summarize the results so far: data with extraction out of nominal expressions also show that arguments within Romance nominal expressions (agents, possessors and objects) display a hierarchical relation, with possessors higher than agents and agents higher than objects. This generalization is consistent with the results obtained with respect to other syntactic phenomena, such as binding of pronouns and anaphors and quantified variable readings of pronouns. The schematic structure in (27), repeated below for convenience, is confirmed by all the data accumulated so far. (30)

XP

X’

Possessor

YP

X

Y’

Agent

Y

Object

However, this section also pointed to a few data that cannot be accounted for with this structure; namely, the behavior of agents introduced by por (by). The next section (Section 2.3.2.2) deals with the behavior of some other PP modifiers in Spanish nominal expressions that have not been considered ARG PPs. That is, I examine PP modifiers that represent roles other than possessor, agent and theme (or that have been previously assumed to be adverbial or adjunct PPs). The results of the next section will provide us with new tools to determine if a PP is an ARG PP or not and will shed light on the status of por (by)-agents, which will not be considered ARG PPs.

2.3.2.2 Types of PPs Up to here I have discussed exclusively PP modifiers that express notions such as agent, possessor or theme/object. In this section, I turn to some other types of PPs within the Spanish nominal expressions, the ones previously classified as PP adjuncts or PP adverbials (cf. Ormazabal (1991)) due to their semantic contribution to the construction. Surprisingly, the results of this

2.3 Structural Relations Within Nominal Constructions

43

section show that some PPs that were previously thought to be adjuncts are actually syntactic arguments, as they pattern like PP arguments with respect to their possibilities for extraction. Therefore, I propose a new classification of PP modifiers in Spanish nominal expressions as syntactic arguments or syntactic adjuncts depending on their behavior with respect to wh-extraction from nominal expressions and other syntactic phenomena. The existence of a clear division between a unified PP argumental group and a unified PP adjunct group is crucial in order to determine the possible or impossible structural relations that the different types of elements within the Spanish nominal expressions can establish and enables us to better specify the structure of the Spanish nominal expressions. The conclusions of this section will ultimately allow us to explain some unexplained basic properties of Spanish nominal expressions, such as the unavailability of movement for adjuncts and the lack of interaction between argument and adjunct PPs. I start with some clarifications regarding the terminology used. In this book, I use the terms argument and adjunct from a strictly syntactic point of view.22 That is, assuming that arguments in general share some syntactic properties, for instance they can be extracted from wh-islands, I classify as an argument any maximal projection that behaves this way, independently of the type of theta role (if any) that it receives23 and the type of N that heads the maximal projection where it occurs. Spanish nominal expressions can contain adjuncts or adverbial-like PPs (cf. Ormazabal (1991)). Some examples are given in (31): (31) a. Compramos [una mesa de madera] (we)bought a table of wood ‘We bought a wood table’ b. Conocimos [varias chicas de Madrid] (we) met several girls of Madrid ‘We met several girls from Madrid’ c. Compramos [varios libros sin tı´ tulo/ sin pastas] (we)bought several books without title/ without covers d. Rompimos [el regalo para Ana] (we)broke the gift for Ana ´ por los americanos] e. Escuchamos [una descripcion (we)listened a description by the Americans The underlined PPs express very different semantic notions. For instance, the PP in (31a) expresses the material the table is made of, the PP in (31b) expresses the origin of the girls, the PPs in (31c) refer to a part of the book, the PP in (31d) tells us about the beneficiary of the present and the PP in (31e) expresses an agent introduced with the P por (by). An exhaustive study of the properties of adverbial-like elements in Spanish nominal expressions does not exist, to the best of my knowledge. In what

44

2 On the Structure of Nominal Expressions

follows I show that there are two groups of adverbial-like elements in Spanish nominal expressions, depending on their syntactic behavior. The two groups are illustrated in (32) and (33), respectively. (32) Group A- Arguments a. Vimos [varios despachos de la Universidad] (we)saw several offices of the University b. Conocimos [varias chicas de Madrid] (we)met several girls of Madrid c. Compramos [varios libros {sin tı´ tulo/pastas/ con dibujos}] (we)bought several books {without title/ cover/ with pictures} d. Vimos [un reportaje sobre Chomsky] (we)saw a documentary about Chomsky (33) Group B-Adjuncts a. Rompimos [el regalo para Ana] (we) broke the gift for Ana b. Escuchamos [el evangelio segu´n San Pedro] (we)listened the Gospel by San Pedro c. Cogimos [la carretera hacia Burgos] (we)took the road towards Burgos ´ por los americanos] d. Escuchamos [una descripcion (we)listened a description by the Americans The PPs in the examples in (32) behave as syntactic arguments, while those in the examples in (33) behave as syntactic adjuncts. The second group (group B) is composed of PP headed by the Ps en (in), por (by), para (for), segu´n (according to) and hacia (towards); while the first group (group A) contains PPs headed by the Ps de (part-whole), de (locative), sin (without), con (with), de (temporal), de (material) and sobre (about).24 There are several arguments in favor of the distinction between group A and group B. The first argument is that a group A PP can occur between N and ARG PP, in the event that the nominal expression contains a head N and an ARG PP. This illustrates that the group A PP exhibits the same freedom in its surface linear order as ARG PPs, (34). As for the group B PPs, they cannot occur between the head N and an ARG PP,25 when the ARG PP expresses the notion of object or theme, (35). (34) a. Visitamos {el despacho de Ana de la Universidad (we)visited {the office of Ana of the University/ / el despacho de la Universidad de Ana} the office of the University of Ana } b. Conocimos {varias amigas de Madrid de Juan (we)met {several friends of Madrid of Juan

2.3 Structural Relations Within Nominal Constructions

45

/ varias amigas de Juan de Madrid} / several friends of Juan of Madrid} c. Compramos varios libros de Chomsky sin tı´ tulo/sin pastas (we)bought several books of Chomsky without title/without covers c’. Compramos varios libros sin tı´ tulo/sin pastas de Chomsky (35) a. Rompimos { el regalo de cumplean˜os para Ana / (we)broke { the gift of birthday for Ana / *el regalo para Ana de cumplean˜os}26 the gift for Ana of birthday} ´ del b. Escuchamos {la descripcion libro por los americanos / (we)listened {the description of+the book by the Americans / ´ por los americanos del libro} *la descripcion the description by the americans of+the book} ´ de Barreras segu´n Chomsky / c. Escuchamos {la explicacion (we)listened {the explanation of Barriers according to Chomsky/ ´ segu´n * la explicacion Chomsky de Barreras} the explanation according to Chomsky of Barriers} Thus, (34a) is an example of ARG PP (possessor) de Ana (of Ana) and group A PP (part-whole) de la universidad (of the university), which shows that both surface orders are possible. Similarly, the examples in (34b) with ARG PP (possessor) de Juan (of Juan) and group A PP (origin) de Madrid (of Madrid) and (34c) with ARG PP (possessor/object/agent) de Chomsky (of Chomsky) and group A PP (circumstance) sin pastas (without covers) illustrate that both linear orders are allowed. In contrast, the examples in (35) that show ARG PP (object/ theme) de cumplean˜os (of birthday) and group A PPs beneficiary para Ana (for Ana), por (by)-agent por los americanos (by the Americans) and opinion/point of view segu´n Chomsky (according to Chomsky) have a fixed surface order. The differences in grammaticality between (34) and (35) suggest a different generation site for elements in the two groups of adjuncts, since they exhibit different properties. A second argument to distinguish the members of the two groups is related to the behavior of wh-elements. Previous research (cf. Chomsky (1986) and Culicover and Rochemont (1992), among others) has claimed that nominal expressions in languages such as English allow extraction of argument wh-phrases, as in (36a) but disallow extraction of adjunct wh-phrases, as in (36b-c): (36) a. [Who]i do you like [a picture of ti]? b. *[Which table]i did you like [NP a book [PP on ti]]? c. *[On which table]i did you like [NP a book ti]? Note that the grammaticality contrast between (36a) and (36b-c) is due to the different structural status of the PPs involved in each case.27

46

2 On the Structure of Nominal Expressions

Let us return to Spanish now. As shown in the previous section, Spanish ARG PPs can be extracted from non specific nominal expressions if they are the only modifier of the nominal expression. Taking the previous discussion into consideration, one should expect the impossibility of extraction out of nominal expressions of the so-called PP adjuncts. However, this is not always the case. The PP elements in group A behave as ARG PPs, while the PP elements in group B behave differently. I reproduce some of the examples below:28 (37)

Group A PPs extracted - grammatical ´ a. ¿De donde visteis varias chicas? of where (you)saw several girls b. ¿Sin que´ visteis varios libros? without what (you)saw several books c. ¿Sobre quie´n visteis un reportaje? about whom (you)saw a documentary

(38)

Group B PPs extracted- ungrammatical a. *¿para quie´n visteis varios regalos? For whom (you)saw several gifts ´ b. * ¿Por quie´n visteis una descripcion? by whom (you)saw a description c. *¿Segu´n quie´n visteis el evangelio? by whom (you)saw the Gospel

For the purposes of this section, the relevance of the contrasts in (38) is that true adjuncts cannot be extracted out of nominal expressions in languages with Ds such as Spanish, which gives us a test to differentiate between the two groups of PP adverbial-like elements in Spanish and support their division into two groups.29 Finally, the PP adverbial-like elements in group A behave as ARG PPs in that they can be extracted out of wh-islands. Consider first the examples in (39), which illustrate that ARG PPs can escape wh-islands: (39) a. ¿[De que´]obj {no sabes/ te preguntas} quie´n compro´ un libro tobj? of what {Neg (you)know/(you)wonder} who bought a book b. ¿[De quie´n]arg {no sabes/ te preguntas} quie´n compro´ una mesa tag? of whom {Neg (you)know/(you)wonder} who bought a table c. ¿[De que´ coleccionista]poss {no sabes/ te preguntas} quie´n compro´ una of which collector Neg (you)know/(you)wonder} who bought a mesa tposs? table tposs The examples in (39) show that ARG PPs can be extracted out of a nominal expression in the object position within a wh-island. Given that

2.3 Structural Relations Within Nominal Constructions

47

adjunct PPs cannot be extracted in general, we expect the ungrammaticality of the examples in (40), since adjunct PPs cannot escape wh-islands either. (40) a. *¿De cua´ndo {no sabes/ te preguntas} quie´n leyo´ varias peliculas? of when {Neg (you)know/(you)wonder} who read several movies b. *¿Segu´n quie´n {no sabes/ te preguntas} quie´n escucho´ According-to whom {Neg (you)know/(you)wonder} who listened varios evangelios? several Gospels Crucially, there is a group of adjuncts (group A PPs) that patterns differently again. Consider in this respect the examples in (41). The grammaticality status of these examples illustrate that extraction of group A PPs out of a wh-island does not produce ungrammaticality. (41) a. ¿De donde ´ {no sabes/ te preguntas} quie´n conocio´ varias chicas? of where {Neg (you)know/(you)wonder} who met several girls b. ¿De que´ {no sabes/ te preguntas} quie´n compro´ una mesa? of what {Neg (you)know/(you)wonder} who bought a table c. ¿Sobre quie´n {no sabes/ te preguntas} quie´n vio un reportaje? about whom {Neg (you)know/(you)wonder} who saw a documentary The examples above then show that the behavior of PPs in group A, traditionally considered non-argumental, is again similar to the behavior of ARG PPs. To summarize: the previously unobserved facts described in this section support the division of PP adverbial-like elements within nominal expressions into two groups. PPs belonging to group B behave as syntactic adjuncts, as they cannot be extracted from nominal expressions. Group A shows a remarkable parallelism with ARG PPs within nominal expressions: they can be extracted from nominal expressions, even when the nominal expression appears in a syntactic island and they show free ordering with respect to ARG PPs. Therefore, I conclude that the PP elements included in the group A (PPA) appear in an argumental position. The rest of this subsection is devoted to the investigation of the behavior of PPA with respect to wh-movement. Now that I have justified the classification of PP adjunct elements within nominal expressions into two groups and have established the parallelism between PPA and argumental elements, I turn to the interaction between the two groups of PP elements and argumental PPs with respect to wh-movement. According to Ormazabal (1991), all PP adverbial elements in nominal expressions are inert with respect to the extractability of any argumental element from the nominal expression; that is, they do not block argument extraction out of the construction:

48

2 On the Structure of Nominal Expressions

(42) a. ¿[De que´ pelı´ cula]obj has ´ tobj [de visto [una version 1986]adv]? of which movie (you)have seen a version from 1986 ´ de esa pelı´ cula tag b. ¿[De que´ director]ag has visto [una version of which director (you)have seen a version of that movie [de 1986]adv]? of 1986 ´ de esa pelı´ cula c. ¿[De que´ productora] has visto [una version of which producer (you)have seen a version of that movie de Almodovar tposs [de 1986]adv]? of Almodovar tposs [of 1986]adv] ´ tobj [segu´n (43) a. ¿[De que´ tema]obj vimos [una explicacion Chomsky]adv]? of which topic (we)saw an explanation according-to Chomsky ´ b. ¿[De quie´n]ag has visto [una explicacion de localidad tag of whom (we)have seen an explanation of locality [segu´n Chomsky]adv]? according-to Chomsky ´ de localidad c. ¿[De que´ lingu¨ista]poss has visto [una explicacion of what linguist (we)have seen an explanation of locality [segu´n Chomsky]adv tposs]? according-to Chomsky The examples in (42)–(43) illustrate the interaction between argumental PPs and each of the groups of adverbial-like PPs, respectively. The grammaticality of the examples above verifies the claim that PP adverbial elements do not affect extraction of arguments out of nominal expressions. More precisely, the data in (42) and (43) show that an object (42a–43a), an agent (42b–43b) and a possessor (42c–43c) can be extracted out of a nominal expression when a PP adverbial is present. As for the extraction of adverbial-like PPs, we need to focus on the behavior of PPA, since only this group of adverbial-like PPs can be extracted out of nominal expressions. (44) a. Han robado [varios incunables [de la Edad Media]adv [de Juan]poss] (they)have stolen several incunables of the Middle Age of Juan b. *¿[De que´ periodo] han robado [varios incunables tadv [de Juan]poss]? of which period (they)have stolen several incunables of Juan c. *¿[De que´ periodo] han robado [varias pinturas tadv [de Picasso]ag]? of which period (they)have stolen several paintings of Picasso d. ¿[De que´ e´poca] han robado [varias pinturas [de santos]obj tadv]? of which period (they)have stolen several paintings of saints The ungrammaticality of (44b–c) illustrates that the presence of an agent or a possessor blocks the extraction of PPA. This contrasts with the data in (44d), where we can observe that PPA extraction is allowed when an object is

2.4 The Structure of Nominal Expressions in Spanish

49

present in the nominal expression. In other words, possessors and agents occupy a syntactic position that interferes with the path of objects and PPA in their extraction out of nominal expressions. The conclusion of this section is then that there are two different groups of PP adverbial-like elements within nominal expressions. One of them (group B) consists in PP adverbial-like elements that do not allow extraction out of nominal expressions and that do not exhibit properties of ARG PPs, such as the possibility of appearance in different surface positions. The second group (group A or PPA) contains PP adverbial-like elements that show a completely parallel behavior to ARG PPs, in particular, object PPs: they can be extracted out of nominal expressions and out of wh-islands, they are freely ordered with respect to other arguments and their extraction is blocked by the presence of certain ARG PPs, namely, possessors and agents.

2.3.2.3 Conclusions Section 2.3.2 examined mainly data with extraction out of nominal expressions. The extraction facts confirm that ARG PPs within Spanish nominal expressions (agents, possessors and objects) display a hierarchical relation, with possessors higher than agents and agents higher than objects. As noticed, this generalization is consistent with the results obtained in Section 2.3.1 with respect to other syntactic phenomena, such as binding of pronouns and anaphors and quantified variable readings of pronouns. The extraction facts supported the main claim of this section, which is that there are two different groups of PP adverbial-like elements within nominal expressions. According to their extraction possibilities, two groups were established: one of them (group B) consists in PP adverbial-like elements that do not allow extraction out of nominal expressions and that do not exhibit properties of ARG PPs, such as the possibility of appearance in different surface positions. The second group (group A or PPA) contains PP adverbial-like elements that show a completely parallel behavior to ARG PPs, in particular, object PPs. This second group of PPs (group A or PPA) can be extracted out of nominal expressions and out of wh-islands and is subject to the same extraction restrictions of object PPs; namely, it is blocked by the presence of possessors and agents.

2.4 The Structure of Nominal Expressions in Spanish This section presents a syntactic analysis of nominal expressions that accounts for the descriptive generalizations established above; namely, the existence of a hierarchical relation among PP arguments in proper syntax and the existence of a group of PP that behaves as syntactic adjuncts.

50

2 On the Structure of Nominal Expressions

This section is divided into three major subsections. In Section 2.4.1, I present the basic theoretical assumptions in the analysis. More concretely, I introduce some locality and anti-locality notions that I will use in my analysis and review some underlying assumptions, such as the possibility of movement into a theta position. In Section 2.4.2, I describe the analysis itself and explain how this type of analysis is able to capture the descriptive generalizations stated in the first part of this chapter. Finally, the section ends with a summary (Section 2.4.3) that outlines the basic points of the analysis.

2.4.1 Theoretical Assumptions The analysis developed in this chapter assumes standard notions in the minimalist framework (Chomsky (1995) and subsequent work), such as general economy notions and takes advantage of some recent proposals, such as the possibility of movement into theta positions. In this subsection I lay out the two main assumptions adopted in my analysis of the basic structure of Spanish nominal expressions: the locality and anti-locality conditions operating in the Spanish nominal expression and the overall design of the Spanish nominal expression, influenced by the division of the clause structure into three domains. Before explaining these basic assumptions, however, it is important to review previous proposals in the clausal domain regarding these issues. Thus, I now refer briefly to the clausal domain. The elimination of Deep Structure (DS) in the minimalist model leads us to reconsider the existence of the principles that were applied at that level of representation. This way, principles that were standardly assumed, such as the theta criterion, are questioned. Some authors have taken seriously the consequences of the lack of DS in the system and derived the theta criterion effects from some other mechanisms of our theory of grammar (cf. Bosˇ kovic´ (1994) on this issue). Bosˇ kovic´ notes that the most important consequence of the theta criterion is that it ensures the ban on substitution into theta positions. That is, the existence of the theta criterion provides us with a straighforward explanation for the ungrammaticality of the following example. (45) *John believes to like Mary Assuming the existence of the theta criterion, the ungrammaticality of (45) comes from the fact that John, which is generated as the subject of like, would end up with two theta roles; namely, the subject theta roles of both believe and like. This is because the theta criterion makes it impossible for one argument to receive more than one theta role.30 Bosˇ kovic´ argues that the ban on substitution into theta positions can be derived without reference to the theta criterion or any additional mechanism. Under his approach, our grammar

2.4 The Structure of Nominal Expressions in Spanish

51

would predict sentences such as (45) as ungrammatical because they would violate general economy constraints in the grammar. This way, the ungrammaticality of (45) is due to the fact that John would receive the nominative case assignment in the subordinate clause and would not have any motivation to undergo a new movement in a framework where movement is severely restricted. Then, the final position of John as the subject of the higher clause causes a violation of the last resort condition. To see the relevance of the previous discussion, let me concentrate on the overall design of the clause. The idea that the overall design of the clause contains three core domains has been proposed implicitly or explicitly since Chomsky (1986). The three domains of the clause can be described as follows: a thematic domain, which contains the predicate and its arguments; an agreement domain, where arguments can receive case and phi-features; and a discourse domain, where discourse information is encoded. However, since the split CP hypothesis proposed in Rizzi (1997) or the tripartite division of the sentence in Platzack (2001), many authors have postulated that those three domains interact with some other principles of our theory of grammar to provide us with explanation for some long-standing puzzles. For instance, Grohmann (2003) reinterprets previous proposals on the existence of domains in the sentence as local domains for movement. According to Grohmann (2003), the division of clause structure into three domains can be combined with the proposals of Bosˇ kovic´ (1994), Bosˇ kovic´ (1997b) and Saito and Murasugi (1999) regarding locality conditions. Grohmann’s (2003) implementation of those proposals supports the division of the clause structure into three domains and the existence of a lower-bound on locality, which Grohmann (2003) refers to as anti-locality. The existence of anti-locality is an improvement of our theory since it allows us to explain grammaticality contrasts such as (46) below without resorting to any additional principle of our theory. (46) a. *John likes b. John likes himself As shown above, recent work on thematic relations and checking theory (Bosˇ kovic´ (1994), Bosˇ kovic´ and Takahashi (1998), Hornstein (1999), among others), argues that theta roles are formal features that need to be checked in an appropriate configuration. Assuming that these recent proposals are on the right track, they allow us to reformulate early generative analyses, such as Lees and Klima’s (1963), that considered local anaphors to be the result of movement of an argument and application of a reflexivization rule to the initial trace of the argument. In other words, if movement into theta positions is allowed, a possible derivation of (46b) (schematized in (47)) could follow the insights of Lees and Klima’s (1963) proposal:

52 (47)

2 On the Structure of Nominal Expressions [VP likes John] [VP John likes John] [VP John likes himself]

Movement of John Reflexivization Rule

Combining this type of analysis with the standard assumptions regarding deletion of copies, namely, that copies are identical and that only one of the copies survives deletion at the PF interface, the ungrammaticality of (46a) is unexpected. Assuming that there is nothing in the theory of grammar that prevents movement of arguments into theta positions, as argued by Bosˇ kovic´ (1994), Bosˇ kovic´ and Takahashi (1998), Hornstein (1999), among others, John could undergo overt movement from the theme to the agent position. As usual, the head of a chain, namely, the higher copy of John, is pronounced in PF.31 Grohmann (2003), following Bosˇ kovic´ (1994), approaches this problem by assuming that there must be an independent constraint in the theory of grammar that prevents (46a) and allows (46b). Grohmann (2003) assumes, as originally proposed by Bosˇ kovic´ (1994), that the relevant constraint is related to the length of the movement of the argument. In other words, the movement of John in (46a), illustrated in (48), is too short, which prevents the deletion of the lower copy to avoid violating anti-locality (i.e., the ban on movement that is too short; see the discussion below).

(48)

Grohmann (2003) discusses different cases of ill-formed movement in the clausal domain. His observations on the length and type of movements disallowed in the clausal domain lead him to support the claim of the existence of different clausal internal domains (cf. Rizzi (1997) and others on this issue). Grohmann (2003) argues that anti-locality applies at the internal domains. Each of these domains constitutes a prolific domain. Each prolific domain forms a part of the derivation where the PF and LF components evaluate the derivation.32

2.4 The Structure of Nominal Expressions in Spanish

53

The formal definition of prolific domain and the description of the three prolific domains in the clausal structure specified by Grohmann (2003) are given below: (49) Prolific Domain Let a Prolific Domain PD be a contextually defined part of the computational system, (i) Each PD spells out its context information and (ii) Spell Out feeds the PF and LF interface levels. (Grohmann (2003:75, his (31)) (50) Context Information Context information is determined by context values: i. |Y|: ranges over thematic relations; ii. |F|: ranges over agreement properties; iii. |O|: ranges over discourse information. (Grohmann (2003:75, his (32)) A movement33 within any of these domains (i.e., within the y-domain, the jdomain or the o-domain) is too short and constitutes a violation of antilocality. This violation of anti-locality would cause ungrammaticality, since the same element would appear twice in the same prolific domain. Antilocality violations can be grammatical only in case the duplicity of the element is avoided by a last resort procedure that yields a drastic effect on the output.34 The means by which anti-locality is applied in these prolific domains are summarized in the condition on domain exclusivity formalized in (51): (51) Condition on Domain Exclusivity (CDE) For a given Prolific Domain PD, and object O in the phrase-marker must receive an exclusive interpretation at the interfaces, unless duplicity of O yields a drastic effect on the output of that PD. (Grohmann (2003:78, his (35)) Returning to the contrasts in (40), repeated below, the combination of Grohmann’s (2003) assumptions allows us to explain the appearance of the anaphor in (52b) as the fulfillment of the condition on domain exclusivity above. (52) a. *John likes b. John likes himself As John moves internally within the y-domain, the movement is too short and the sentence can be grammatical only if the lower copy of John is spelled out as a different phonological realization of John at PF.

54

2 On the Structure of Nominal Expressions

Through the study of anti-locality effects in the three prolific domains of the clause,35 Grohmann (2003) concludes that the presence of a number of resumptive pronouns is due to the fact that the second (initial) copy of a given element is spelled out as a resumptive pronoun to overcome an antilocality violation. Let us return now to the nominal domain. It has been mentioned previously that, following Abney’s (1987) DP hypothesis, it is generally assumed that the internal structure of nominal constructions resembles the internal structure of clauses in the richness of its functional configuration. Therefore, it seems to be natural to base the analysis developed for the nominal domain on current developments regarding the CP structure that have been proved advantageous. This is precisely my goal in this book. I will be exploring recent research in the clausal domain, assuming the division of the nominal structure into three domains to provide a more economical explanation to properties of Spanish nominal constructions that have defied analysis so far. In this chapter, I follow the assumptions in Grohmann (2003) and I pursue the extension of this proposal presented in Grohmann and Haegeman (2002). Grohmann and Haegeman (2002) combine Grohmann’s (2003) proposal with the clausal DP hypothesis that assimilates the structure of the nominal layer to that of the clausal layer. Grohmann and Haegeman (2002) propose an analysis of the prenominal possessor doubling construction in Germanic, illustrated in (53): In order to explain the appearance of the resumptive pronoun in (53), Grohmann and Haegeman (2002) argue for the existence of three prolific domains within the nominal construction, indicated in (54) and adopt the (53)

Marie euren boek Marie her book ‘Marie’s book’

(West Flemish)

analysis in (55) for cases such as (53), where Marie is moving from the spec of possessor phrase to the spec of AgrP: (54) DPoD> AgrPjD > NPyD (55) [DP Spec D0 [AgrP Spec Agr0 Marie

[PossP

Spec

! euren

Poss0 [NP

N0]]]]

boek

According to Grohmann and Haegeman’s (2002) analysis of Germanic nominal expressions, the locus of checking/licensing possession is PossP, which is included in the j-domain. Furthermore, possessors must at least rise to AgrP to check possessive Case/Agr, which is contained in the j-domain. This movement

2.4 The Structure of Nominal Expressions in Spanish

55

produces an anti-locality effect. To save the derivation, the lower copy of the possessor must be spelled out as the resumptive pronoun euren (her).36 The proposal shows that it is possible to obtain accurate empirical coverage via a parallelism between the tripartite clausal structure and nominal structure with respect to the notion of prolific domains and anti-locality. Crucially for our purposes, the possibility of resumption is not generally available. It is not only the case that all languages cannot use resumption in the same constructions, as is evidenced by the impossibility of having constructions equivalent to (53) in languages such as English or Spanish; but also resumptive pronouns as a strategy are extremely rare. This particular property of resumptive pronouns has led many authors, such as Shlonsky (1992) or Hornstein (2001), to treat resumptive pronouns as last resort strategies, being licit only when the derivation is not going to converge without the resumptive pronoun. The repercussions that the previous discussion on resumptive pronouns has for the proposal of this book are important and will determine the analysis considered in detail in the next section. In other words, the main idea is that if resumptive pronouns are not available to solve cases such as (53) in a language, that particular language will not have any strategy to avoid the anti-locality effect and will have to ban the construction completely because of the impossibility to have the derivation to converge. The next subsection applies the notions summarized in this subsection to a new analysis of Spanish nominal expressions. First, I introduce the basic structure I will be arguing for in the rest of the book and describe each of its components. After that, I develop the analysis itself and explain how this type of analysis is able to capture the descriptive generalizations stated in the first part of this chapter. Finally, I will derive all the relevant constructions introduced in the previous sections and provide new supporting evidence for this analysis.

2.4.2 The Analysis 2.4.2.1 The Basic Structure of Spanish Nominal Expressions As illustrated in the first part of this chapter, any analysis of Spanish nominal expressions must account for two main generalizations that can be extracted from the description above. First, the ARG PPs within Romance nominal expressions (agents, possessors and objects/PPA) maintain a hierarchical relation. The data from licensing of pronouns and anaphors, quantified bound pronouns and wh-extraction show that possessors are structurally higher than agents, objects and PPB and agents are structurally higher than objects and PPA. Second, there are two different groups of PP adverbial-like elements within Spanish nominal expressions: group B behaves as syntactic adjuncts, while

56

2 On the Structure of Nominal Expressions

group A, PPA, shows the same behavior as argumental PPs, concretely they behave as objects. In Chapter 1, I observed that, since Abney’s (1987) DP hypothesis, a good number of proposals have argued for the presence of different functional categories within nominal expressions. It is not within the scope of the present discussion to review in detail the rich literature on the functional structure associated to the noun phrase.37 However, a word on some previous proposals is needed here to introduce the choices of functional categories present in my analysis. Szabolcsi (1994) proposes the existence of an agreement (Agr) projection in the nominal expressions of the same nature as the one found in clauses, AgrP. This AgrP allows her to explain the agreement relation overtly expressed between the possessor and the noun in Hungarian as the subject-verb agreement established in clauses. Alongside the previous proposal, Ritter’s (1991) analysis of construct state and free state constructions in Hebrew leads her to postulate the existence of a Number Phrase (NumP). This intermediate position between D and N is where the N checks its number features. Note that this is also a sub type of agreement. Some other authors, such as Valois (1991), Picallo (1991) or Bernstein (1993), among many others, have found more empirical evidence to adopt the functional category NumP as part of the nominal structure.38 Some languages, such as Spanish or Catalan, have an extended agreement in their nominal constructions, which includes the category gender. Picallo (1991) argued for the existence of gender phrase to accommodate the order of arguments with respect to the noun in Catalan nominal expressions. This proposal is adopted with certain changes by authors such as Bernstein (1993). Bernstein, who follows Harris’s (1991) hypothesis that gender is more a word marker (cf. also Ouhalla (2005) on this issue, among others) proposes that a slightly different functional category, word marker phrase, is necessary to explain NP-ellipsis data in Romance languages. Furthermore, the existence of a genitive phrase or possessor phrase has been defended by authors such as Cardinaletti (1998), who analyzes in detail the different possessive types and argues for the existence of more than one category to host them. Having the clause as a structural model, some other functional categories have been proposed in the literature. For instance, some authors have proposed the existence of an outer nP, headed by an abstract light noun (n), which will correspond to the outer vP shell in the verbal domain, proposed in Larson (1988), Hale and Keyser (1993) and others. Following this line of thought, authors such as Radford (2000) propose that this functional category should host PP agents in the nominal domain. Lastly, some other proposals in the literature regarding the functional categories in the nominal expressions have followed Rizzi’s (1997) proposals for the clausal structure. According to Rizzi’s (1997) proposal, CP must be decomposed into a sequence of functional projections, containing functional

2.4 The Structure of Nominal Expressions in Spanish

57

categories such as ForceP, FinP, FocusP (FocP) and Topic Phrase (TopP). Focusing on Romance languages, Giusti (1993) also assumed that elements such as demonstratives could appear in a focus position. Some other authors, such as Bernstein (1997, 2001), also noted that some elements typically occupy the right periphery of the nominal expression. More concretely, Bernstein claims the existence of a FocP that hosts a series of contrastively focused elements, such as possessive adjectives or demonstratives in Spanish and other Romance languages. In the same spirit Aboh (2002) and others have also proposed the existence of TopP in the nominal domain, obtaining empirical evidence from the specificity effects of languages such as Gungbe. The common ground in all these proposals is to provide some projections in the highest part of the nominal structure where the interface with discourse can be located, since nominal constructions seem to display discursive properties comparable to the ones displayed in the clause. This proliferation of functional categories in the nominal domain contrasts with the overall goal of economy in the minimalist framework since it produces structures with functional categories that do not seem to have a role in a particular derivation. A possible solution to this in a theoretical framework that emphasizes economy is to put restrictions on the features associated with functional categories in order to limit an unnecessary proliferation of functional categories in our theory of grammar. With this goal in mind and taking into consideration the empirical motivation for the different functional categories postulated so far in the literature, I assume a structure of nominal expressions containing only the minimal number of functional categories needed to keep the parallelism between the clausal and nominal domain regarding the existence of three domains of movement (i.e, the aim is to keep the parallelism in the tripartite structure but not in the individual functional categories that appear in the clause). That is, following the parallelism with the tripartite sentence structure regarding this point, I will be assuming three functional categories as essential: a functional category that links the discourse to the sentence or the nominal phrase, DP or some other category in its place; a functional category where the relevant (meaningful) agreement relations can be established, AgrP; and a third functional category that conveys the agentivity notion, nP. These functional categories have been extensively and independently justified and will only enter in the derivation as needed. Furthermore, in an attempt to avoid the proliferation of functional categories in the nominal domain, I will take the most restrictive point of view and the insertion of any additional functional category in this basic structure will be avoided. Only those functional categories fully justified from an empirical point will constitute part of the structure in a given numeration and only functional categories needed to keep the parallelism with the tripartite structure in the sentence are necessary. Therefore, the basic structure I propose for Spanish nominal expressions is as follows:39

58

2 On the Structure of Nominal Expressions

(56)

FocP Foc’ Foc

ω-domain

DP D’

AgrP

D

POSS

Agr’ Agr

ϕ-domain

nP n’

n

AGENT θ-domain

NP

N

OBJ

The structure in (56) shows the three prolific domains within the nominal structure without massive proliferation of functional structure. There is a y-domain, where the theta relations are established; a j-domain containing AgrP, where agreement properties are licensed; and a o-domain, where the discourse information is established. In addition, note that the structure in this chapter displays its specifiers to the right. This directionality of projection follows the treatment of specifiers in previous analyses of extraction out of Spanish nominal expressions (cf. Torrego (1987) and Ormazabal (1991), among others). In principle, this type of approach aimed to capture the postnominal position of PP elements in the Spanish nominal expression, something that was derived by assuming this directionality of projection in the first proposals on the issue. More recent proposals on the structure of the Spanish nominal expressions (cf. Sa´nchez (1996) or Malle´n (2001), for instance) generate the specifiers in the left-hand side and accept some type of noun movement to derive the postnominal surface word order of PPs within the nominal expression. Given that, in this book (see Chapter 4 for an extended discussion on the problems of noun movement), I propose to abandon the notion of noun movement as a device to derive the internal order of nominal expressions in Spanish.

2.4 The Structure of Nominal Expressions in Spanish

59

I maintain in my structures these previous assumptions with respect to the relative placement of specifiers in Spanish nominal phrases. A question that arises at this point is whether this type of structure is incompatible with the Linear Correspondence Axiom, LCA (Kayne 1994). There are several possibilities to make these structures compatible with the LCA. One possibility is to argue that the linear order is achieved through rightward movements in the stylistic component (i.e., PF). Under this approach to linear order, there are stylistic movements that adjoin the PP arguments to different positions in order to obtain the surface linear order. These stylistic movements follow the application of the LCA (cf. Chomsky (1995)) to obtain the surface linear order. A different approach, more in line with Kayne (1994), would move the N-Obj constituent leftwards in front of the agent and then move the constituent N-Obj-Agent leftwards in front of the possessor (i.e., a type of snowballing movement). Both approaches give the desired surface linear order and are compatible with the LCA. I will not pursue this issue here since it is not clear what would motivate these movements and instantiating them would raise several technical problems that go beyond the scope of this chapter. Nevertheless, the reader should bear in mind that the structures in this book are not necessarily inconsistent with the LCA. In addition, an alternative line of analysis has been explored in Ticio (2009), where I examine in detail the relation between hierarchical representation and phonological order within Spanish nominal expressions. The resulting analysis proposes a free generation analysis of PP arguments in narrow syntax, since I crucially assume that linear order takes place at the PF interface (cf. Chomsky (2001)) and not during syntax (contra Kayne (1994)). Then, the proposal is that it is the insertion of the higher functional categories in the course of the derivation that probe exclusively certain PP arguments, due to their different feature composition, which creates the desired hierarchical representation during the derivation. Thus, the different possible surface word orders are the consequence of chain resolution at the phonological interface, which is needed for linearization purposes and is constrained by language-specific phonological properties. Given the exploratory character of Ticio’s (2009) proposals, I maintain the option based on the different directionality of projection in the structures of this book for expository purposes. Let us return now to the structure in (56). At this point, a more refined explanation on the type and properties of the functional categories chosen for this structure is in order. The presence of (small) noun Phrase (nP) reinforces the parallelism between clausal and nominal domains. As discussed above, nP is the locus of agentivity, that is, it hosts agents and, as I will show later, it is only projected when an agent (i.e., a de (of)-agent or an adjective denoting an agent) is present. There have been different proposals in the literature providing empirical evidence for this functional category. For instance, Radford (2000) has also argued that a shell structure (nP-NP) is necessary to account for some

60

2 On the Structure of Nominal Expressions

argumental nominal expressions in English.40 Similarly, Carstens (2000) has proposed the presence of this functional category to derive the agreement relations within nominal expressions. For the purposes of this book, nP in the nominal domain is the counterpart of vP in the clausal domain. Similarly, as seen above, the presence of Focus Phrase (FocP) mimics the structure of clauses and follows recent proposals (Rizzi (1997), among others) concerning the possibility of splitting the CP-layer. Along the same line, Grohmann and Haegeman (2002), Haegeman (2004) and Aboh (2004), among others, argue for the presence of different functional categories within the odomain of the nominal domain. For instance, TopP in the nominal domain serves as the host of nominal left dislocation in languages such as West Flemish41 according to Haegeman. As shown before, similar proposals have been assumed for Romance languages (cf. Giusti (1993), Bernstein (1997, 2001a), and others) and share the goal to provide a projection in the highest part of the nominal structure where the interface with discourse can be established. In the analysis in this chapter, FocP is primarily used to host fronted whelements; therefore, FocP is the equivalent of Ormazabal’s (1991) K(om)P and it serves as the escape-hatch for the elements extracted out of nominal expressions. Moreover, as for the other functional categories, FocP42 is projected just in case it is required to check off a feature in the structure. As for Agreement Phrase (AgrP), a more detailed justification is needed. Recall that Chomsky (1995) claims that agreement functional categories should not be present in the clause structure as individual syntactic positions. This is due to the feature content of this functional category. Hence, the features in Agr are marking just the syntactic relation between the predicate and its subject and do not play any role in LF (i.e., they do not receive a meaning). In the nominal domain, though, the features hosted in Agr do play a role in LF, since, for instance, the number feature will contribute importantly to the interpretation of the nominal expression. It seems then that there is a departure from the parallelism between the functional composition of nominal expressions and clauses, since Agr seems to be required for nominal expressions but not for clauses (cf. Belletti (2001) for a detailed treatment of agreement projections). As seen above, some other categories related to agreement processes have been proposed in the literature. For instance, AgrP has been abandoned for gender phrase, number phrase, genitive phrase, or a combination of all of them. However, not everyone assumed all these functional categories. For instance, Ritter (1993) argued against the existence of gender phrase and postulated that gender is a parameterized feature. Taking into consideration the criticism exposed in the literature and the goal of economy that guides the current research, I group under a single AgrP functional category any of the agreement-related functional categories proposed in earlier analyses. That is, I replace NumP, GenP, PossP and others, by a more general AgrP, where all the agreement-based relations are established.43 Note that this category must always be present in the structure, which makes it different from the other functional categories seen, because it contains

2.4 The Structure of Nominal Expressions in Spanish

61

(at least) the number feature in the nominal expression. As for the motivation to generate possessors in the specifier of AgrP, this analysis assumes that possessors occupy this position as a reflex of its subject like properties.44 The presence of the possessor as the specifier of the functional category Agr has also been proposed recently in Giusti (2008), where it is assumed that the only relation of agreement in the nominal expression is precisely the one between the possessor and the functional category Agr, being the other feature sharing relations established among the different elements of the nominal expression the result of more local relation (a relation between a head and its specifier) termed concord.45 Finally, although it will be the main topic of the next chapter, let me briefly discuss the treatment of determiners I adopt here. I assume (following Abney (1987), Bernstein (1993), Zamparelli (2000), among many others) that not all determiners are generated in D. Thus, following Milsark’s (1977) division of Ds, I assume that only the presence of a strong determiner46 triggers the projection of DP and that weak determiners are generated in a lower projection, which I identify here as AgrP.47 The analysis in this book also acknowledges the two main ideas that have guided previous analyses with respect to extraction out of Spanish nominal expressions (cf. Torrego (1987), Ormazabal (1991), and others). First, previous analyses have adopted the idea of a very local movement in these constructions. More precisely, most analyses have assumed that movement of wh-elements out of nominal expressions requires intermediate landing sites within the nominal expression. Second, earlier analyses have focused on the difference between specifier and non specifier positions to account for the asymmetries presented between agents, possessors, objects and PPA. The current proposal follows these previous insights and implements them in the current framework to achieve the greatest empirical coverage possible. As for the locality relations operating in nominal expressions, I follow the recent implementations of the anti-locality hypothesis (cf. Grohmann (2003) for instance). Hence, movements within prolific domains can be grammatical only in the case the duplicity of the element moved is avoided by a last resort procedure that yields a drastic effect on the output. In short: elements within nominal expressions can only move from a prolific domain to a prolific domain, unless there is compelling evidence to the contrary at the interface. Since the latter scenario will not arise in the cases I am discussing because Spanish does not have resumptive pronouns in nominal expressions, for my purposes all movement within a prolific domain is disallowed in nominal expressions in Spanish, due to the impossibility to have the derivation to converge. Apart from the anti-locality hypothesis, I follow the intuition of previous analyses regarding the extremely local character of movement in Romance nominal expressions and I assume that phrasal movement within nominal expressions crosses only one maximal projection in each step. This restriction

62

2 On the Structure of Nominal Expressions

on the maximal length of movement follows some general conditions on movement, such as the minimize chain links principle of Chomsky and Lasnik (1993) or the locality principle introduced by Manzini (1994). The two principles aim to restrict possible movement operations by stating that an element must move the shortest distance. The minimize chain links principle regulates the length of the links of a chain to derive the notion of government. Several authors have pursued this line of research (cf. Manzini (1994), Fox and Lasnik (2003), among others). Manzini (1994) takes as her departure point this line of research and proposes that movement must involve two adjacent minimal domains. According to Manzini (1994), ‘the minimal domain of a head X consists of all and only the elements that are immediately contained by, and do not immediately contain, a projection of X’ (Manzini (1994: 482)); that is, the minimal domain of XP will contain its spec, X itself, the complement of X and the elements adjoined to XP. Moreover, Manzini (1994) considers two minimal domains (A and B) adjacent to one another if there is no member of a third domain (C) that contains A but not B, or vice versa. In short, Manzini’s (1994) locality principle requires movements from one maximal projection to the next maximal projection.48 To summarize: movement within nominal expressions is highly restricted. It cannot take place further than one maximal projection and the movement must be from one prolific domain to another prolific domain. The following subsections detail the interactions between the structural and movement constraints proposed to explain the different possibilities of extraction out of Spanish nominal expressions depending on the type of PPs that appear in the construction.

2.4.2.2 Extraction Out of Non Specific Spanish Nominal Expressions As shown in Section 2.2, the extraction out of Spanish nominal expressions headed by non-definite Ds, such as varios (several), un (a) or alguno (some), is restricted by the hierarchical relation that the arguments (agents, possessors and objects) maintain within nominal expressions, with possessors49 higher than agents and agents higher than objects. Under the assumptions discussed in Section 2.4.2.1, the explanation for the blocking effects in extraction out of Spanish non specific nominal expressions is straightforward: the presence of a possessor in the specifier of AgrP blocks the extraction of any element, since the wh-element cannot cross more than one maximal projection in its movement to the specifier of FocP. The derivation of the ungrammatical (57) is shown in (58): (57) *¿[De quie´n] has leı´ do [varios libros tag [de Juan]poss]? of whom (you)have read several books tag [of Juan]poss

2.4 The Structure of Nominal Expressions in Spanish

63

(58)

The derivation of the grammatical (59a), schematized in (59b), shows that the movement of the agent when a possessor is not present meets the requirements postulated for movements within nominal expressions.50 Each of the movements in (59b) crosses only one maximal projection and respects the antilocality hypothesis. (59) a. ¿[De quie´n] ag has leı´ do [varios libros tag]? of whom (you)have read several books tag b.

Then, the movement of the agent to the specifier of AgrP enables this element to be placed in a position from which it can be moved out of the nominal expression.

64

2 On the Structure of Nominal Expressions

Note that movement of objects follows the same restrictions. The descriptive generalization states that objects cannot be extracted from non specific nominal expressions whenever a possessor or agent is present. The explanation for the blocking effects of possessors on the movement of objects is completely parallel to the one described for the blocking effects of possessors on the movement of agents: the presence of a possessor in the spec of AgrP forces the object to skip the spec of AgrP as an intermediate landing site and this produces a violation on the conditions of locality established for nominal expressions. Let us pay attention now to the blocking effect of agents on the movement of objects. Under the analysis presented so far, the presence of nP would block the extraction of an object out of a non specific nominal expression due to a locality violation.

(60) a. *¿De que´ has leı´ do [varios libros [tobj] de Cervantes]? of what (you)have read several books of Cervantes b. FocP Foc’ Foc

De quéobj ϕ – domain

AgrP

Agr’ Agr varios

De quéobj nP

n’

de Cervantesag NP

n

N libros

θ – domain

De quéobj

This is because the movement of the object from its initial position to the spec of AgrP position would cross more than one maximal projection, namely, the object crosses nP. Therefore, in order for an object to be able to move, nP must not be present in the structure. Note that it will not be sufficient for nP to be present but its specifier empty. That is, it is crucial for my analysis that movement cannot cross more than one XP and the presence of nP, even

2.4 The Structure of Nominal Expressions in Spanish

65

with an empty specifier, will cause the movement to be too long. In addition, a derivation where the object moves to the specifier of nP is ruled out since it would produce a domain-internal movement, which is blocked by anti-locality. The derivation of the grammatical (61a) is then (61b): (61) a. ¿De que´ has leı´ do [varios libros [tobj]]? of what (you)have read several books b. FocP Foc’ Foc

de quéobj AgrP

θ-domain

NP N libros

ϕ-domain

de quéobj

Agr’

Agr varios

ω-domain

de quéobj

The evidence for the lack of nP projection when a de (of)-agent is absent comes from the different behavior of por (by)-agents and de (of)-agents with respect to extraction. As shown in Section 2.3, de (of)-agents block the extraction of objects out of non specific nominal expressions while por (by)-agents do not affect extraction of objects. The contrast is repeated below: ´ tobj (62) a. ¿De que´ tema has criticado una investigacion of what topic (you)have criticized a investigation por los americanos? by the Americans ´ tobj b. */?? ¿De que´ tema has criticado una investigacion of what topic (you)have criticized a investigation de los americanos? of the Americans As commonly assumed, por (by)-agents are not dependent on the theta-grid that predicates (Ns or Vs) project and it is the P por (by) that introduces the relevant ‘agentivity’ relation. That is why they typically appear in passivizations, where the agentive projection is not present. The contrasts in (62) thus show that the introduction of agentivity with a por (by)-agent does not trigger the presence of nP in non specific nominal expressions. Therefore, the extraction of objects is not blocked when a por (by)-agent is present since there

66

2 On the Structure of Nominal Expressions

is no nP in the relevant structure and the movement of the object respects the locality conditions established. Note that Bosˇ kovic´ (1997) shows a similar pattern with vPs, the counterpart of nP in the clausal domain. Bosˇ kovic´ (1997) discusses data such as: (63) a. b. c. d.

*[AgrO [vP Johni [VP ti wagered [IP Peterj to be tj smart]]]] [AgrO [VP John believes [IP Peterj to be tj smart]]] Peter was wagered (by Mary) to be smart Peter was believed to be smart

Given the generalization that agentive verbs cannot exceptionally case mark in general (cf. Pesetsky (1995)), the crucial difference between (63a) and (63b) is that verbs such as wager assign the agent theta-role to their subject, while verbs such as believe do not assign an agent theta-role. Bosˇ kovic´ (1997a) captures the difference by the presence or absence of vP: vP is only projected in (63a). Assuming that the embedded subject must reach the spec of AgrOP to get its case, Bosˇ kovic´ (1997a) then argues that the ungrammaticality of (63a) is due to the presence of vP, which blocks the required movement of the embedded subject to the spec of AgrOP. His analysis finds support in cases such as (63c), where, although we can have an agent introduced by the P by, the sentence is grammatical since the V has been passivized. Given Bosˇ kovic´’s (1997a) analysis, it follows then that an agentive by-phrase does not induce projection of vP; the contrast between (63a) and (63c) thus parallels the contrast between (62a) and (62b), which can be interpreted as providing additional evidence for the current analysis.51 Given the discussion above, the ungrammaticality of (64b) shows that agentive adjectives are introduced in the spec of nP. ´ americana de este tema (64) a. Criticaron una investigacion (they)criticized an investigation American of that topic ´ americana? b. * ¿De que´ tema criticaron una investigacion of what topic (they)criticized an investigation American The ungrammaticality of (64b) is due to the introduction of agentivity with the adjective americana (American), which triggers the presence of nP in the structure and blocks the extraction of the object.52 Note that in (64) the surface order must be relational adjective-object, the order object-relational adjective being unacceptable. A priori, this is unexpected under the current analysis, which places adjectives in the specifier position of nP. The explanation for this surface order comes from the special character of the combinations N-relational adjective. Rainer and Varela (1991) argue that the combinations N-relational As display characteristics that allow us to classify them as improper compounds. Assuming that N-relational As are in the first stage towards the creation of lexical compounds, a prerequisite for which PF adjacency between the noun and the relational adjective, the word

2.4 The Structure of Nominal Expressions in Spanish

67

order found in the examples in (64) is not surprising: the stylistic movements discussed earlier can create both orders, relational adjective-object and objectrelational adjective. However, the lexical requirements of the relational A would rule out the order object-relational adjective as a possible option. Nothing has been said so far regarding the placement of adjuncts in the structure. Remember that a division of adjuncts into two classes was established in the previous sections. I assume that adjuncts belonging to group B are generated as the adjuncts of NP while adjuncts belonging to group A (PPB) occupy the object position. The proposed structure is then (65): (65)

FocP

Foc’ ω-domain Foc

DP D’

D

AgrP

Agr’

Agr

POSS

φ-domain

nP

n’ n

AGENT θ-domain

NP NP

ADJUNCTS

N’

N

OBJ/PPB

The structure in (65) shows that adjuncts that cannot be extracted and do not affect movement of any other element in the nominal expressions (group B in Section 2.3 above) are attached to NP. Since PPA can be extracted and behave as objects regarding their extraction possibilities, they are considered syntactic complements in this book. If we have more than one object and PPA in the same structure, the NP would be recursive, as shown schematically in (66).

68

2 On the Structure of Nominal Expressions

(66) a. El libro {sin pastas de fı´ sica/ de fı´ sica sin pastas} The book {without covers of physics/ of physics without covers} b. . . ...[NP [ N’ libro [NP sin pastas [N’ N de fı´ sica]]]] Note that the locality conditions are not going to be violated since both elements belong technically to the same maximal projection; namely, NP.53 Finally, this type of structure will also explain that arguments can be extracted over adjuncts. Crucially, I will be assuming that specifiers and adjuncts are different entities of grammar, contra Kayne (1994) and work done following this line of research. This is the traditional line of thought in the generative grammar and has been safeguarded by authors such as Duffield (1999) and pursued recently in Uriagereka (1999), Chomsky (2001), Gallego (2007), among many others. Furthermore and following the logic in the minimalist program (cf. Grohmann (2003) on this issue), a derived assumption is that adjuncts are only basegenerated positions and, following Chomsky (2001), adjuncts are not in the search domain of the element that acts as a probe. Therefore, although in principle it might look like adjuncts are structurally higher in the representation given than the arguments, what the structure shows is that adjuncts are in another dimension. That is why they cannot freely interact with arguments and ultimately cannot block argument extraction. To conclude: the proposed analysis explains the core data regarding extraction out of Spanish nominal expressions and makes important predictions regarding the type of movements and constructions allowed in these expressions. In the remaining chapters and sections, I will show that these predictions are borne out in some properties of Spanish nominal expressions, such as the distribution of adjectives and nominal ellipsis and also in some previously unobserved data regarding partial cliticization. 2.4.2.3 On Some Predictions of the Analysis Let me finish the description of my analysis of Spanish nominal expressions by addressing an important prediction of my analysis borne out by some previously unobserved contrasts. The current analysis has shown that the locality conditions on movement explain the impossibility of moving an object out of a nominal expression whenever an agent or possessor is present in overt syntax. Therefore, the analysis predicts that we cannot have an object and a possessor or an object and an agent within the same nominal expression whenever the object must move out of the NP in overt syntax. There is a construction in Spanish that can help to test this prediction. Consider the data in (67): (67) Vimos el cuadro de las Meninas ! Lo vimos (we)saw the picture of the Meninas CL (we)saw The sentence in (67) illustrates a standard cliticization construction where we have cliticized the object, el cuadro de las Meninas (the portrait of the

2.4 The Structure of Nominal Expressions in Spanish

69

Meninas), with the clitic pronoun lo (it-[acc, masc, sing]). A priori, this type of cliticization can affect different parts of the nominal expression; the sentences in (68) illustrate different cases of partial cliticization where we have left stranded different arguments of the N. (68) a. Compramos varias traducciones [de ese escritor]ag (we)bought several translations of this writer y tu´ las compraste [de aquel polı´ tico]ag and you CL bought of that politician b. He leı´ do varios capı´ tulos [de Don Quijote]obj (I)have read several chapters of Don Quijote y tu´ los has leı´ do [de La Celestina]obj and you CL have read of La Celestina Note that the object in (68b) occupies the complement position of N. Thus, the examples in (68) seem to pose a problem for the standard assumption that clitics substitute for maximal projections. In other words, an analysis that claims that partial cliticization targets the NP must explain why the object can be left stranded after partial cliticization in the data in (68b). There are two possible ways out of this problem. First, one could assume that cliticization can target NPs or N. This analysis does not seem to be motivated by any other data, as the typical situation is that clitics substitute for the entire NP. Second, we could maintain the standard assumption that clitics substitute for NPs and explain the grammaticality of cases such as (68b) as the result of some intervening factor. I explore the second possibility here. Assuming that the most plausible analysis is that partial cliticization affects maximal projections, the grammaticality of (68b) can be explained by stating that, when cliticization applies, objects are not in their original positions, which allows them to be stranded. For the moment, let us assume that, in partial cliticization cases, the clitic Lo is base-generated adjoined to the V, following the hypothesis (cf. Burzio (1986), among others) that claims that the surface position of clitics is not the result of their movement. Moreover, let us assume that the clitic is associated with an empty NP in the position in which the nominal expression is found. A schematic representation of this analysis is given in (69): (69) [VP Vendimos [DP los [NP libros] [de Cervantes]ag] (we)sold the books of Cervantes y[VP Losi-compramos [DP [NP ei ] [de Chomsky]ag] and CL (we)bought e of Chomsky Assuming that partial cliticization involves LF copying of the antecedent NP into the empty NP in (69), the grammaticality of (68a) is expected. To derive the grammaticality of data such as (68b) we must assume that the object is located outside of NP when cliticization applies, as (70) illustrates:

70

2 On the Structure of Nominal Expressions

(70) [VP Vendimos [DP los [NP libros] [de fı´ sica]obj ] (we)sold the books of physics y[VP Losi-compramos [DP [NP ei ] [de quı´ micaobj] and CL (we)bought e of chemistry If objects can be located in a higher position (that is, outside of the empty NP) in cases such as (68b), the analysis of partial cliticization as LF copying can be maintained. Note that, in order to LF copy the antecedent NP into the empty category in the second conjunct, the antecedent cannot contain the object in the complement position. Two possible analyses can be implemented at this point. First, it may be the case that the object in the antecedent nominal expression is generated outside of NP. If so, LF copying can proceed directly and no overt movement is required. The second analysis proposes that the object is base-generated in the complement position in the antecedent NP. There must then be an overt movement of the object,54 to ensure that the antecedent NP and the empty NP are equal and LF copying can be applied.55 As we will see shortly, the Spanish data in this section point to the second analysis discussed above as the correct one. Remember that the analysis of this chapter has shown that we cannot have objects and agents or possessors present at the same time when the object must move out of the NP in overt syntax. Keeping the previous discussion in mind, let us consider the following previously unobserved data in (71) and (72). The data in (71a) and (72a) illustrate again that partial cliticization can affect different elements in the nominal expression. Interestingly, the data in (71b-b’) and (72b-b’), where we have an object and a possessor, as in (71b), or an object and an agent, as in (72b), as the remnants, are ungrammatical. (71) a. Compramos una casa de Juan y tu´ la compraste e de Ana (we)bought a house of Juan and you CL bought e of Ana b. *Compramos una casa de Juan con ventanas azules (we)bought a house of Juan with windows blue y tu´ la compraste e de Ana con piscina and you CL bought e of Ana with swimming pool b’. *Compramos [una casa [con ventanas azules] obj [de Juan]poss ]] (we)bought [ a house [with windows blue]obj [of Juan]poss ]] y tu´ la compraste [e [con piscina]obj [de Ana]poss ]]] and you CL bought [e [with swimming pool] [of Ana]poss ]]] (72) a. Compramos varias traducciones [de ese escritor]ag (we)bought several translations of this writer y tu´ las compraste e [de aquel polı´ tico]ag and you CL bought e of that politician b. */?? Compramos varias traducciones [de Don Quijote]obj (we)bought several translations of Don Quijote

2.4 The Structure of Nominal Expressions in Spanish

71

[de ese escritor]ag y tu´ las compraste [de La Celestina]obj of this writer and you CL bought of La Celestina [de aquel politico]ag of that politician b’. *Compramos [varias traducciones [de Don Quijote]obj (we)bought [several translations [of Don Quijote]obj [de ese escritor]ag]] y tu´ las compraste [e [de aquel polı´ tico]ag [of this writer]ag ]] and you CL bought [e [of that politician]ag [de La Celestina]obj]] [of La Celestina]obj]] Crucially, the ungrammaticality of (71b-b’) and (72b-b’) points to the second analysis discussed above as the correct one. That is, the object is base- generated in the complement position in the antecedent NP in all the cases discussed so far and there must then be overt movement of the object to ensure that the antecedent NP and the empty NP are equal so that LF copying may apply. The ungrammaticality of (71b-b’) and (72b-b’) follows from the impossibility of moving the object overtly in the antecedent NP, because the presence of an agent or a possessor blocks overt syntactic movement of objects.56 In other words, following the line of analysis presented so far, (71b-b’) and (72b-b’) are ungrammatical because overt syntactic movement of the object in the antecedent NP violates the locality conditions established in this chapter. Moreover, the data in (73) show that the appearance of a possessor and an agent is allowed, as expected under the current analysis. (73) Compramos varias traducciones [de ese escritor]ag [de Ana]poss y (we)bought several translations of this writer of Ana and tu´ las compraste [de Miguel]poss [de aquel polı´ tico]ag you CL bought of Miguel of that politician The grammaticality of (73) is then predicted, as the antecedent NP contains only the N and no movement outside of the antecedent NP is required for LF copying purposes. To conclude: the proposed analysis explains the core data regarding extraction out of Spanish nominal expressions and makes predictions regarding the type of movements allowed. As shown, these predictions are borne out in some previously unobserved data regarding partial cliticization. In the following chapters, I show that the proposed analysis can also account for the behavior of adjectives and nominal elision in Spanish nominal expressions.

2.4.3 Conclusions This chapter has defined the basic structure of Spanish nominal expressions. To do so, I have analyzed the structural relations among the elements within

72

2 On the Structure of Nominal Expressions

nominal expressions and the possibilities for extraction from nominal expressions. I have also compared the current analysis with earlier proposals regarding these phenomena. The analysis of Spanish nominal expressions developed in this chapter has pursued a parallelism between clausal and nominal expressions. The analysis has adopted some of the latest developments regarding CP structure (i.e., the division of clause structure into three domains and its interpretation in Grohmann (2003)) and has extended them to account for the properties of Spanish nominal expressions (following the line of research opened by Grohmann and Haegeman (2002)). The resulting analysis enables us to explain the full paradigm regarding the basic patterns of extraction observed in Spanish nominal expressions from the locality conditions on movement that elements within nominal expressions must satisfy. In the following chapters, I show that the proposed analysis can also account for further restrictions in the extraction data and for the behavior of attributive adjectives and nominal elision in Spanish.

Appendix: Genitive Pronouns57 This chapter has analyzed the different types of PP modifiers in the Spanish nominal expressions; I will examine now the properties of genitive pronouns in Spanish nominal expressions. Previous studies on this topic (Cinque (1980) for Italian, among others) maintain that genitive pronouns in Romance nominal expressions are the result of pronominalizing full PPs in the relevant structures. However, note that this is not always possible. Pronoun modifiers cannot have the same range of meanings as the PP modifiers. In fact, many PP modifiers, such as the ones in (75), cannot be pronominalized in Spanish. (75) a. ¿De que´ has comprado el libro tobj? of what (you)have bought the book b. *Su (= de fı´ sica) libro its (= of physics) book c. Un cuadro del siglo XV A picture of-the century 15th d. *Su cuadro58 Its (=of-the 15th century) picture More concretely, the meanings available to genitive pronouns in Spanish seem to be reduced to the possessor, agent or object. (76)

Su foto Her/his picture (owner)/(author)/(theme).

Therefore, we cannot equate the behavior of all full PPs with the behavior of genitive pronouns. It seems to be more accurate to relate genitive pronouns

Appendix: Genitive Pronouns

73

with a subkind of full PPs to determine their exact structure: the PPs that were referred to as PP arguments in the previous sections. A question that might cause concern at this point is what the difference between pronouns and full nominal expressions is. This difference seems to be such that one can occur prenominally but the other must occur within a postnominal PP. The answer I will explore here is based on the different case properties of pronouns versus full nominal expressions. In Spanish and other languages pronouns are able to show morphological genitive case, while nominal expressions need the presence or absence of a preposition to express case. (77) illustrates some differences between personal pronouns and full PPs regarding the expression of case. (77) a. Ella lee/ Ana lee She reads/Ana reads b. La veo/ veo a Ana her (I)see / (I)see to-Ana c. Su libro/ El libro de Ana Her (of-Ana) book/ The book of Ana The examples in (77) show that nominative (77a), accusative (77b) and genitive (77c) case can be expressed via morphological changes in the personal pronoun; while the examples with the nominal expression Ana resort to the addition of prepositions to express the different cases, (77b) and (77c). If we combine this structural difference (DP/NP for genitive pronouns versus PP for non-nominative full nominal expressions) with the different prosodic properties of a monosyllabic DP and a multisyllabic PP, we can account for the word order differences discussed. More concretely, it seems that no independent prosodic element can interfere between D and N in Spanish. The proposal is then that the positioning of a genitive pronoun prenominally does not break any prosodic group, while the presence of a genitive PP between the N and the D would break the intonational group. This approach seems to be on the right track in many ways if we consider that the only elements that can intervene between a D and an N are prenominal attributive adjectives. Note that these elements have some properties that seem to point to the fact that prenominal adjectives can form a prosodic unit with the D and N, as this proposal predicts. Consider the following examples: (78) a. El orgulloso padre The proud father b. *El orgulloso de su hijo padre The proud of this son father c. El padre orgulloso The father proud d. El padre orgulloso de su hijo The father proud of his son

74

2 On the Structure of Nominal Expressions

(79) a. ?La supuesta u´nica antigua amiga de mi madre (...) The supposed only old friend of my mother (Demonte (1999)) b. El sombrero rojo redondo tejido de Marı´ a The hat red round weaved of Mary The examples in (78) and (79) illustrate well-known properties of prenominal adjectives in Spanish; namely, they cannot be complemented (cf. (78b) vs. (78d)) and they cannot be stacked (cf. (79a) vs. (79b)). Note that these properties (and some others, see Chapter 4 for an extensive discussion of attributive adjectives) receive a straightforward explanation if we assume that there is a PF constraint in Spanish that prevents independent prosodic constituents to interfere between D and N.59 Assuming this line of explanation, the conclusion is then that the resulting surface order restrictions of full PPs are also derived from the interaction of the morpho-syntactic properties of pronouns and full nominal expressions and from the interaction of those with general PF constraints in Spanish.

Notes 1

2

3

4 5

Through the following chapters I use the terms nominal expressions, nominal phrases and nominal constructions as general terms covering determiner phrases (DPs) and nominal phrases. I only use the term DP for those nominal constructions that have a structure with a DP as a functional category. Leaving aside the properties of the determiners themselves; this will be analyzed in detail in the next chapter. See note 6 below and the appendix at the end of this chapter for a discussion and references of genitive pronouns. At this stage of the discussion, I use the term modifier to cover both arguments and adjuncts. Genitive pronouns in Spanish have some other peculiar properties. Consider the examples in:

(i) a. mi casa my house b. mis casas (*mi casas) my-pl house-pl The possessor in both nominal expressions in (i) is corresponding to a first person singular. The plural mark that the genitive pronoun displays in (ib) corresponds to the plurality denoted by the noun casa (houses). Thus, the genitive pronoun in Spanish refers to the possessor but agrees with the possessum, unlike regular genitives in other languages, such as English. 6

I do not examine in this book the differences between inalienable and alienable possessors in Spanish in this book. See Vergnaud and Zubizarreta (1992:596) for a definition of the terms and an examination of some inalienable constructions in Romance and Alexiadou et al. (2007: 613-616) for a discussion on some properties that point to a different argumental status between alienable and inalienable possessors.

Notes 7

8

75

Note that this is not true for all Romance languages. As Kolliakou (1999:746) notes, dephrases in French cannot be freely ordered and there is a tendency to have the possessor following the agent in the linear word order. A standard c-command definition is in (i) (i) Node A c-commands node B if every branching node dominating A also dominates B and neither A nor B dominates the other. (Carnie (2002: 82)) According to this definition, a referential expression contained in a PP cannot c-command out of the PP because the first branching nodes dominating the nominal expressions are the PP ones.

9

Apart from Giorgi and Longobardi (1991), many other authors have shown these structural relations to hold in different languages. Cf. Aoun (1985) for similar data in French and Alexiadou et al. (2007: 559) for a discussion of binding relations within English nominal expressions to determine the most plausible position for possessors. See also Longobardi (2001:562–577) on the properties of arguments of the head noun in languages such as Italian, English and German and for a discussion on the probably universal principles and parametric variation of the argument structure of nominal expressions. 10 I assume here that binding indicates c-command relations in the syntax. As will be shown in this chapter, the lack of effect of surface word order in the c-command relations and the behavior of extraction data strongly support this claim. 11 Note that this seems to be also the case with some other Ps in some other Romance languages, since Giorgi and Longobardi (1991) have shown in the examples below, some nominal arguments in Italian introduced by di (of) and non-locative a (to) do not behave as PPs; rather di (of) and non-locative a (to) must be considered dummy Ps, since c-command out of these PPs is possible. (i) a. L’opinione di se stessoi di Giannii e` troppo lusinghiera the-opinion of himself of Gianni is too flattering ‘Gianni’s opinion about himself is too flattering’ b. *L’opinione di luii di Giannii e` troppo lusinghiera the-opinion of him of Gianni is too flattering ‘Gianni’s opinion about him is too flattering’ 12

13

14

Note that there is some research (see Williams 1994, among others, to this respect) that claims that binding applies to a presyntactic argument structure. This line of thought is also consistent with the results reached in this chapter. As mentioned above, a complete different approach is explored in Ticio (2009). In Ticio (2009), I take seriously the claim that there is no order in syntax proper and argue for a free generation analysis of nominal expressions in which the different PP arguments enter the derivation in a free derivational order. The different feature composition of each PP and a general mechanism of chain resolution at the phonological interface account for the differences in hierarchical and surface orders. Given the exploratory status of Ticio’s (2009) proposal, it will not be adopted in this book. Most of the research on the topic assumes that there is a correlation between the possibilities for extraction out of nominal expressions in Romance and the existence of the thematic hierarchy described in the text. A different perspective has been put forward by studies such as Kolliakou (1999). Kolliakou (1999) examines French extraction data and proposes that the thematic hierarchy cannot explain the peculiar extraction patterns in the nominal expression of this language. Her approach is based on the consideration that the inherent semantic properties of de-phrases are the cause for the blocking effects in French. Kolliakou (1999)’s analysis divides the de-phrases in French in two groups (namely, individual or referential and property or non-referential) and assumes her nominal denotation hypothesis, reproduced in (ii):

76

2 On the Structure of Nominal Expressions (ii) Nominal denotation hypothesis: a noun head can take at most one individual dephrase and at most one property de-phrase. Kolliakou (1999:736, (24)) The extraction facts in French are then explained as the result of the impossibility of extracting a property de-phrase due to the lack of variable-binding mechanisms for this type of de-phrase. One of the biggest problems to extend this type of approach to Spanish lies in the nominal denotation hypothesis (NDH). Note that the NDH does not predict the situation in Spanish nominal phrases, since, as shown in the text, Spanish can have more than two de-phrases per noun. Taking into consideration these empirical issues, I will not consider this type of approach in the text and my approach will be based on the thematic hierarchy. However, note that some other aspects of Kolliakou’s (1999) proposal find empirical evidence in this book. For instance, the specificity facts discussed in the next chapter will receive a solution in the spirit of Kolliakou’s (1999) proposal, although our implementation is substantially different.

15

16

17

18 19

As we will see in the following chapter, they also depend on the specificity of the nominal structure. As one of the series editors makes me note the notion of ‘object’ does not really correspond with non deverbal nouns such as libro (book) and it would be more appropriate to label this type of occurrence as a ‘theme’. On a related topic, only complex event Ns, which possibly inherit their thematic role relations from their verbal origin, can be accurately said to have ‘traditional thematic roles’. Cf. Alexiadou et al. (2007: 555–556) discussion on this issue. For ease of exposition, I maintain the label ‘object’ in the examples. In this descriptive part of the chapter, I mark with t, which stands for trace, the original position of the element moved. Most examples in this subsection come (or are adapted) from Ormazabal (1991). Note that the grammaticality of the examples is not affected by the discourse linked (or dlinked) character of the wh-elements involved. The ungrammaticality of the following examples, with d-linked wh-elements (i.e. wh-elements that clearly refer to a referent already within the universe of discourse, cf. Pesetsky (1987)), illustrate the point: (i) a. *¿[De que´ autor] has leı´ do [varios libros tag [de Juan]poss]? of what author (you) have read several books tag [of Juan]poss b. *¿[De que´ tema] has leı´ do [varios libros tobj [de Juan]poss]? of what topic (you)have read several books tobj [of Juan]poss

20

The same observation has been reported in the French literature on extraction (cf. Godard (1992), Kolliakou (1999), among others). A representative contrast is below: (i)

a. l’interpre´tation [de la neuvie`me]Theme [de Karajan]Agent the interpetation of the Ninth of Karajan b. *son Theme interpre´tation [de Karajan] Agent

3. sg interpretation of Karajan c. l’interpre´tation [de la neuvie`me] Theme [par Karajan] Agent the interpetation of the Ninth by Karajan d. son Theme interpre´tation [par Karajan] Agent 3. sg interpretation by Karajan (adapted from Kolliakou (1999: 715, (3) & 731, (16)) The possessivization asymmetries in (1b-d) are showing that extraction is possible in the presence of a par (by) agent.

Notes 21

22

23 24

77

Por (by)-agents cannot be extracted, as the example in (i) shows: (i) * ¿Por quie´n has criticado varias invasiones [de Irak]obj tpor-ag ? by whom (you)have criticized several invasions of Iraq The behavior of por (by)-agents is examined in the following section. Contra authors such as Grimshaw (1990), I will not use semantic notions to differentiate between ‘PPs arguments’ and ‘PPs adjuncts’. According to Grimshaw’s (1990) typology of ´ (destruction), take arguments. Grimshaw Ns, only complex event Ns, such as destruccion relates this assumption to one property of complex event Ns that differentiates them from other types of Ns that do not take arguments: complex event Ns project a hidden eventive variable to be saturated by the argument. I will discuss the possible mechanisms for thematic role assignment in the next section. Note that there is a semantic relation underlying the PP adverbial-like elements belonging to group A. More concretely, the PP adverbial-like elements in group A correspond to the part-whole/material relation and to the locative or provenance relation. Hornstein, Rosen and Uriagereka (1999) refer to the part-whole/material relation as integral, cf. (ia) and to the locative relation as existentia’, cf. (ib).

(i) a. My Saab has a Ford Engine b. Canada has ten provinces According to these authors, integral and existential relations behave similarly regarding certain phenomena such as preposition stranding and agreement properties. This similarity between Hornstein, Rosen and Uriagereka’s (1999) classification and my classification points to an interesting topic of research, which I leave open for further research. 25

26

Note that this fixed word order (noun B-PPs A-PPs) calls for a different treatment of arguments and adjuncts at PF. This is possible if we assume that in general arguments and adjuncts are spelled out differently and that we cannot insert an adjunct freely at PF either. See Uriagereka (1999), Chomsky (2001) and Gallego (2007) for discussion on the recent treatment of adjuncts. Note that the examples in the text only contain object ARG PP preceding the group B PP. It seems that group B can alternate order with poss and agent, as examples such as (i) illustrate: (i) a. Rompimos { el regalo de Juan para Ana / *el regalo para Ana de Juan } (we)broke { the gift of Juan for Ana / the gift for Ana of Juan} b. Compramos {el libro de Chomsky de 1995/ de 1995 de Chomsky} (we) bought {the book of Chomsky of 1995/ of 1995 of Chomsky}

This evidence does not necessarily undermine the proposal in this chapter since adjunct PPs are adjoined to the NP and possessors and agents occupy a higher structural position than objects. 27

The grammaticality contrast between (31a) and (31b-c) in the text also follows a generalization given in Uriagereka (1988). According to Uriagereka (1988), who takes into consideration cross- linguistic data, adjunct extraction out of nominal expressions is impossible in languages with phonetically realized D. One relevant language is SerboCroatian, which according to Stjepanovic´ (1998) and Bosˇ kovic´ (2002) has no D and allows extraction of adjuncts out of nominal expressions, as in (i): (i) a. Petar je procˇitao knjige sa ove police Peter is read books from this shelf ‘Peter read books from this shelf’ b. Sa koje police je Petar procˇitao knjige?

78

2 On the Structure of Nominal Expressions From which shelf is Peter read books * ‘From which shelf did Peter read books?’ (Ochi (2000: 564))

28

In principle, as Juan Uriagereka (p.c.) notes, the examples in the text could be analyzed as involving small clause secondary predication. Under this analysis, the extraction of the PP elements would not be relevant for our purposes, since the PP would not belong to the nominal expression. There is, however, evidence against such an analysis of the examples in the text. Word order with secondary predication in Spanish is rather free; that is, the predicate can precede or follow the nominal expression, as the examples in (i) illustrate:

(i) a. Tomaron {a los estudiantes por vagos/ por vagos a los estudiantes} (they)took {to the students by lazy /by lazy to the students} ‘They considered the students lazy’ b. Considero {a Juan vago / vago a Juan} (I)consider {to John lazy /lazy to John} ‘I consider John lazy’ However, as the examples in (ii) show, cases such as the examples in the text do not display this alternation: (ii) a. Conocimos {varias chicas de Madrid /* de Madrid varias chicas} (we)met {several girls of Madrid/ of Madrid several girls} b. Vimos {varios libros sin tı´ tulo /*sin tı´ tulo varios libros} (we)saw {several books without title/ without title several books} I therefore assume that the relevant examples in the text are cases of PP adverbial-like elements within nominal expressions. 29

Interestingly, Cattell (1979) observes that in cases where the semantic relation part-whole is involved such as (i), the extraction of an adjunct out of a nominal expression seems to be allowed in English as well: (i) Which car do you like [the gears in ti]? The grammaticality of cases such as (i) points to a similarity between the cases I classify as PPs in group A and certain adjuncts in English. This point is left for further research.

30

Theta criterion: (i) a. every theta role must be assigned to some argument. b. every argument must be assigned some theta role (Bosˇ kovic´ (1994: 242, his (11)))

31

32

As an anonymous reviewer points out, a valid question is in what sense the two copies are identical given that they have two different theta roles. Note that this is a question for all the analyses that claim that movement into theta positions is an option. The answer to this question is not within the scope of this book and is left for further research. Note that Grohmann (2003) differentiates between this evaluation process and the operation spell out. According to Grohmann (2003), spell out is an operation that can be applied multiple times in a derivation and that, following Uriagereka’s (1999) idea of multiple spell out, ‘morphologizes’ the object at the interfaces. This ‘morphologization’ causes the object to be inaccessible to further syntactic operations. However, Grohmann’s (2003) idea of evaluation consists in making available some information contained in a prolific domain to LF and PF. This evaluation process, which does not cause the derivational object (i.e., the prolific domain) to be inaccessible to further syntactic operations, does not preclude the existence and application of a spell out operation at a later stage of the derivation.

Notes

33

34

35

36

37

38

40

41

79

Throughout the chapter, I retain this difference between ‘evaluation’ and spell out in my analysis of Spanish nominal expressions and assume that spell out does not need to apply every time we complete a prolific domain but only when we have a phase. Grohmann (2003) argues that the only phrasal movement possible in overt syntax is a substitution movement (i.e., movement to a spec position) and that adjuncts must be the result of base-generation (i.e., introduced by direct merge). This claim is derived from the properties of bare phrase structure. See Grohmann (2003) for details and below for a more detailed discussion of the difference between specifiers and adjuncts. I follow Grohmann (2003) and assume that adjuncts cannot be the result of movement in overt syntax. According to Grohmann (2003), a drastic effect in the output is the overt realization of the relevant copy with a different PF matrix. Grohmann (2003) shows that the anti-locality condition also accounts for the impossibility of moving wh-elements to topic positions and for the impossibility of elements receiving two case features. For additional evidence for the ban on movement that is too short see Bosˇ kovic´ (1994), Bosˇ kovic´ (1997b), Saito and Murasugi (1999), Ishii (1999), Abels (2001) and Bosˇ kovic´ (2002). These authors show that the ban on movement that is too short can account for the impossibility of short-subject topicalization and short-zero subject relativization in English (Bosˇ kovic´ 1997b), certain extractions out of nominal expressions (Bosˇ kovic´ 2002), the immobility of IP (Abels 2001) and the that-trace effect (Ishii 1999). I therefore assume that positing the ban on movement that is too short is empirically well motivated and adopt Grohmann’s (2003) version of this ban. Cf. Grohmann (2003) for different types of movements that are banned under his approach, such as the impossibility of moving wh-elements to topic positions and the impossibility of elements receiving two case features. Note that movements such as passives are allowed because the movement is sufficiently long (from the base-generated object position to spec of TP) for not producing a violation of anti-locality. The strikethrough elements indicate the copies left behind after movement, which were represented in previous frameworks as t (traces). See Alexiadou et al. (2007: 227–281) for a detailed discussion on the different functional categories proposed for the nominal expressions and their cross-linguistic empirical motivation. Cf. the discussion on interpretable and non interpretable features in the functional categories in the nominal expressions in Alexiadou et al. (2007:229–246). Cf. Alexiadou et al. (2007: 561-562) for a discussion of the problems that some properties of complex event nominals raise to Radford’s (2000) proposal and for a different use of the nP category as the host of possessors. Grohmann and Haegeman’s (2002) example of nominal left dislocation in West Flemish is given below: (i) Verhofstadt den dienen zen fouten Verhofstadt the that-Masc his mistakes ‘Verhofstadt’s mistakes’ (Grohmann & Haegeman (2002: 12))

42

43

44

45

For more information on the relation between focus phrase and wh-movement, the reader can consult Hagstrom (1998), Bosˇ kovic´ (1998) and Bosˇ kovic´ (2002), among others. Note that under current minimalist assumptions agreement relations do not need to involve (overt or covert) movement and can be established as long distance relations. See Alexiadou et al. (2007: 570–576) for a proposal of different positions for possessors in terms of derived or non-derived possessors, which takes into consideration their crosslinguistic behavior. Although I have pointed to one similarity between Giusti’s (2008) analysis and the proposal in this chapter, there are several incompatibilities between the two analyses. Maybe the most prominent difference is the one related to the use of functional categories

80

46

47

48

49

50

51

52

2 On the Structure of Nominal Expressions in Giusti’s (2008) proposal. For her, each concord relation implies the existence of a functional category that mediates between the N and the element that shares features with it. This leads to the existence of a number of functional categories in the nominal expression, something that goes against the spirit of my proposal. Milsark (1977) showed that determiners can be divided into two classes, strong and weak, depending on their syntactic behavior. This division of determiners has been extensively used in subsequent literature on DPs (cf. Zamparelli (2000) for an extensive summary on this topic). Note that the analysis developed in this chapter is also consistent with proposals that claim that weak Ds move (cf. Herburger (2000), Roehrs (2002), among others, on D movement). Nothing in my analysis depends on this assumption. The adoption of Manzini’s (1994) locality condition for Grohmann’s (2003) proposal in the clausal domain raises several problems, as movements in the clausal domain appear to be able to cross more than one maximal projection. A possible way to accommodate Manzini’s (1994) locality condition in Grohmann’s (2003) tripartite structure is to assume the possibility of multiple specifiers in the clausal domain (on this issue see Chomsky (1995), who allows this option rather freely for at least some projections in the clausal domain). This option would allow very short movements in the clausal domain, which would respect Manzini’s (1994) locality. I leave a more detailed explanation of movement within the clausal domain for future research. Nothing has been said so far regarding the mechanisms used to assign theta-roles. Recent work on thematic relations and checking theory (Bosˇ kovic´ (1994), Bosˇ kovic´ and Takahashi (1998), Hornstein (1999), among others) assumes that theta roles are formal features that need to be checked in an appropriate configuration. Under this type of approach, arguments do not need to be generated in the position where they receive or check off their theta features, as movement into theta positions is allowed. The analysis proposed in this chapter assumes a thematic domain, which contains the predicate and its arguments and where the theta relations must be in principle established. Arguments such as objects and agents, which appear merged in the theta domain, will check their features via merge. A valid question arises with respect to the possessor argument, which appears as the specifier of AgrP in the structures of the text. There are many analyses of possessors (cf. Alexiadou et al. (2007: 547–616) for an extensive discussion of different analyses of possessors and the different properties of these elements in languages such as English, Italian, French, Hungarian, Greek, Swedisch, Dutch, just to mention a few). One possibility is to treat possessors as derived arguments and to propose that possessors undergo an internal movement. If this is the case, we could move the possessor from a position internal to the thematic domain (possibly from the highest specifier within the recursive nP) to the specifier of AgrP, where it can receive its case and phi-features. Another possible analysis is the one illustrated in the text, where no movement of the possessor is proposed and an alternative analysis of possessors is needed. Although both options are compatible with the analysis presented in this chapter, I will keep generating possessors in the spec of AgrP for expository purposes. Following Grohmann (2003), I assume here that the only possible movement in overt syntax is movement to a specifier position. Adjunction can only be the result of basegeneration (see Grohmann (2003) for relevant discussion). Note that the wager-class verbs are the exception in the CP domain and the presence of vP does not usually produce any kind of restriction of movement. As noted previously, movements in the clausal domain appear to be able to cross more than one maximal projection. Technically, this difference between the nominal and the clausal domain can be captured via the possibility of having multiple specifiers in the clausal domain or by proposing the need for more local relations within the nominal domain. This point needs more careful study and it is left for future research at this point. For a detailed description of the treatment of adjectives under this analysis, see Chapter 4.

Notes 53

54

55

56

57

58

81

The possibilities of having more than one PPA and the impossibility of having more than two objects will be related to semantic reasons. That is, we cannot have two arguments with the same syntactic and semantic relation, unless they are conjoined. The issue of where this object moves is not clear to me at this point. However, a possibility to explore is that the spec of Agr must be filled in overt syntax. If no agent or possessor appears in the structure, the object must move to the spec of Agr, presumably to check the EPP feature of Agr. The trace of the object in the antecedent NP can serve as the trace of the object in the second NP. Similarly, the theta role assignment will be parallel in the two NPs. Crucially, note that the movement of the object must be syntactic in partial cliticization contexts. A stylistic movement of the object in partial cliticization contexts cannot account for the ungrammaticality of the examples in the text, as stylistic movements do not affect LF. As we noted before, what we call here genitive pronouns differ from English genitive pronouns in that they agree with the possessum, see footnote 5 above. For a typology of possessive pronouns see Alexiadou et al. (2007: 568–570) A reviewer wonders whether there is not a restriction to animacy that causes the ungrammaticality of (75b) and (75d). Although there is a tendency to have animate elements in this type of context, Spanish also has examples such as (i) where the element possessivized is not animated: (i) a. La causa de la enfermedad era clara The cause of the disease was clear b. Su causa era clara its cause was clear

59

Note that English differs from Spanish in this respect, since it allows modification of prenominal adjectives and stacking, along with full NPs (genitive possessors) preceding the N.

Chapter 3

Specificity, Definiteness and the Definite Article in Spanish Nominal Expressions

3.1 Introduction Chapter 2 studied the types of arguments and modifiers allowed in nominal expressions1 in Spanish. The focus was on the main properties of the two groups of prepositional phrases (PPs) appearing within nominal expressions; namely, argumental and adjunct PPs. The study of these PPs and their relationships led to a deeper understanding of the internal dependencies allowed within the Spanish nominal expressions, which also helped us to design their basic structure repeated in (1).

(1)

FocP Foc’ ω-domain

Foc

DP D’ AgrP

D

ϕ-domain

Agr’

Agr

POSS

nP n’

n

AGENT θ-domain

NP N

OBJ

M.E. Ticio, Locality Domains in the Spanish Determiner Phrase, Studies in Natural Language and Linguistic Theory 79, DOI 10.1007/978-90-481-3398-7_3, Ó Springer ScienceþBusiness Media B.V. 2010

83

84

3 Specificity, Definiteness and the Definite Article in Spanish Nominal Expressions

The structure in (1) accommodated the two main claims of the last chapter: there is a structural hierarchy of argument PPs (possessor higher than agent and agent higher than object) that is represented in the syntax proper and there is a new division of PPs within nominal expressions attending solely to their syntactic properties. The different syntactic properties of the different types of PPs (such as, for instance, their possibilities of extraction out of nominal expressions) were derived from their relative hierarchical position and from the limitations on the length of movement out of nominal expressions. Having those main goals in mind in Chapter 2, I did not elaborate a detailed explanation regarding the highest part of the structure and all the relevant examples contained non specific indefinite determiners for ease of exposition. In this chapter, the focus is shifted to the highest part of the structure. The chapter discusses the location of the different types of determiners available depending on their meaning and the interaction between the type of determiner (and the semantics it introduces) and the overall syntactic properties of the structure. The main claim of this chapter is that the presence or absence of the functional category Determiner Phrase (DP) in the structure will affect the movement possibilities in the structure, due to its interaction with the locality conditions on movement analyzed previously. Hence, this chapter tests the basic structure and the explanation for the extraction possibilities out of Spanish nominal expressions with some previously unobserved examples that escape the basic patterns of extraction of wh-elements in Spanish nominal expressions. More concretely, I investigate in this chapter the properties of wh-extraction from nominal expressions headed by the definite article in Spanish and compare them to the cases of so called specificity effects in the literature. Surprisingly, extraction out of nominal expressions headed by the definite article is banned only with certain PP arguments, while extraction out of Spanish specific nominal expressions is always banned. Consider (2) and (3) as representative constructions. a. *¿De que´ autor has leı´ do [los libros tag]? of which author (you)have read the books tag b. *¿De quie´n has visto [las fotos tposs]? of whom (you)have seen the photos tposs c. ¿De que´ cantante has visto [las fotos tobj]? of which singer (you)have seen the photos tobj (3) a. *¿De que´ autor has leı´ do [estos libros tag]? of which author (you)have read the books tag b. *¿De quie´n has visto [estas fotos tposs]? of whom (you)have seen the photos tposs c. *¿De que´ cantante has visto [estas fotos tobj]? of which singer (you)have seen the photos tobj

(2)

This difference between (2) and (3) is attributed in this chapter to the features contained in the category D in the structure and to the peculiar

3.1 Introduction

85

properties of the definite article in Spanish. I will assume that there are two features that are involved in the relevant contrasts: definiteness and specificity. The definite article in Spanish can have both features and that will produce the blocking effects, or just the definiteness feature, which will result in the possibility of extracting an argument out of the structure. Then, to the extent that it is correct, the main claim put forward in this chapter is coherent with recent attempts to divide further semantic notions such as definiteness (cf. Lyons (1999), Julien (2005), among many others). Furthermore, the presence or absence of these features corresponds with the projection (or not) of the functional category DP in the structure, which is in turn responsible for the specificity of the nominal expression and for the blocking effects referred to as specificity effects. The interaction of DP with the locality conditions on movement will then explain the extraction possibilities in the structure as violations of the locality conditions analyzed in Chapter 2. The general structure of the chapter is as follows. Section 3.2 introduces some general background on the main notions discussed here. First, in Section 3.2.1, the distinction definite / indefinite is presented. After that, Section 3.2.2. explores the classification of indefinites into specific and non specific. The importance of specificity in definite nominal expressions is discussed in Section 3.2.3. Section 3.3 describes the peculiar behavior of Spanish nominal expressions headed by the definite article with respect to extraction. In Section 3.3.1, I describe the so called specificity effects in Spanish and English. After that (Sections 3.3.2 and 3.3.4), I study the behavior of specific definite and indefinite nominal expressions in Spanish regarding extraction. In Section 3.3.3, I compare the behavior of specific definite and indefinite nominal expressions in Spanish regarding extraction with the behavior of definite nominal expressions in Spanish regarding extraction to illustrate that a priori these are different phenomena in Spanish. Section 3.4 presents a syntactic analysis for the different types of nominal expressions introduced in the chapter. The section starts by lying out the basic assumptions of the analysis and continues by applying the basic analysis to the relevant cases. In Section 3.4.4, the section examines in some detail the properties of nominal expressions headed by the definite article in Spanish. This section will then establish the main features of the definite article in Spanish, namely [+/- definite], [+/- specific]. In this section, I demonstrate that the structure and locality conditions described in Chapter 2, combined with the special properties of definite articles in Spanish, can resolve not only the puzzle with the extraction possibilities out of nominal expressions headed by the definite article but also the specificity effect in Spanish. More concretely, I propose that the definite article can be generated in two different positions: as the head of Agr when it behaves as a weak determiner and allows extraction out of the nominal expression and as the head of D when it behaves as a strong determiner and disallows the extraction out of the DP. In other words, it is the projection of DP that prevents the extraction out of specific DPs and certain types of definite nominal expressions in Spanish. Finally, in Section 3.4.5 I have compiled some additional evidence for the analysis.

86

3 Specificity, Definiteness and the Definite Article in Spanish Nominal Expressions

The chapter closes with Section 3.5.1, which summarizes the main findings of the previous discussion in Section 3.5.1 and uncovers some of the questions and problems of the analysis introduced and some ideas to explore in order to solve them (Section 3.5.2).

3.2 Fundamentals: Definiteness and Specificity Notions such as definiteness (or indefiniteness) and specificity (or lack of) play a fundamental role in the discussion and analysis of this chapter. Due to the importance of these two notions, this section is devoted to providing some general background on them. This section then aims to familiarize the reader with the main questions concerning definiteness and specificity and with the different ways of representing these notions in Spanish, with special attention to their markers inside nominal expressions.2 In the first part of this Section 3.2.1, I describe what definiteness and indefiniteness are and offer some examples illustrating how they are represented crosslinguistically and in Spanish. After that, I focus on the indefinite nominal expressions to introduce the notion of specificity (Section 3.2.2). I introduce later different instances of specific and unspecific indefinite nominal expressions in Spanish and discuss the leading theories to explain them in the literature. Finally, in Section 3.2.3, I turn to the definite nominal expressions to explore the consequences of the existence of the specificity notion in these type of constructions and provide some examples of specific definite nominal expressions in Spanish. The section concludes with a brief summary (Section 3.2.4), where the main conclusions of the section are stated.

3.2.1 The Distinction Definite vs. Indefinite Nominal expressions are generally considered definite or indefinite. Informally, a definite nominal expression makes reference to an entity with which both speaker and hearer are familiar. Let us consider (4) (4) I have the book When the speaker utters the sentence in (4) s/he knows the referent for the nominal expression the book and assumes that the hearer is also able to determine the referent for that expression; that is, there is a sense in which the speaker and hearer are familiar with the individual denoted by the definite nominal phrase. This is not the situation when an indefinite nominal expression is used, as in (5) (5) I have a book A book refers also to a clearly identifiable individual in the speaker’s mind (for instance, the speaker can be referring to the particular book s/he holds in

3.2 Fundamentals: Definiteness and Specificity

87

his/her hands while s/he utters the sentence in (5)) but the speaker does not expect the hearer to be able to identify the particular referent of that expression. From the speaker’s point of view, this is not relevant in this case. Definite and indefinite nominal expressions are frequently marked crosslinguistically by the presence of lexical items such as the or a; namely, by the use of definite/indefinite articles in the traditional grammar nomenclature. Spanish also instantiates the definiteness vs. indefiniteness dichotomy with definite and indefinite articles, as in (6). (6)

a. El nin˜o jugaba en el parque The boy played in the park ‘The boy was playing in the park’ b. Un nin˜o jugaba en el parque A boy played in the park ‘A boy was playing in the park’

The difference in meaning between (6a) and (6b) is then determined by the definiteness of the underlined nominal expression. In the case (6a) the referent of the nominal expression is a familiar, identifiable boy; in the case (6b), the referent of the nominal expression is not familiar, it is implied that the hearer is not expected to identify it. Up to this point, definite nominal expressions have been presented as those nominal expressions that contain a definite article and indefinite nominal expressions as those nominal expressions with an indefinite article. However, there are several other kinds of nominal expressions that indicate that the referent is identifiable or not. In fact, all nominal expressions are marked as definite or indefinite expressions in a given language. Traditionally, it has been assumed that names, pronouns and definite descriptions are definite nominal expressions; while some numerals and quantifiers have been classified as indefinite nominal expressions. There are some environments that have been used to determine whether a nominal expression is a definite nominal expression or not. I illustrate one of them below. (7)

a. There is/are {a/several/many/few/one} book(s) on the table b. *There is {the/all/every/that} book on the table

Existential constructions, as the ones illustrated in (7), force the presence of non definite nominal expressions, while ban the appearance of definite nominal expressions. This is what is referred to as the definiteness effect (Milsark (1974), Barwise and Cooper (1980), Keenan (1987), Belletti (1988), among many others). Keeping in mind the previous intuitive discussion, we could assume that a definite expression refers to a familiar entity while an indefinite expression does not need to. This preliminary conclusion falls short when trying to explain the behavior of some definite nominal expressions. Consider (8) (8) The president of Ghana is visiting tomorrow (Lyons (1999):3, example (10))

88

3 Specificity, Definiteness and the Definite Article in Spanish Nominal Expressions

In (8) the hearer does not have to be familiar with the person who occupies the presidency of Ghana to understand the referent of the definite nominal expression. Examples such as (8) have led to the conclusion that the notion of familiarity is not adequate to describe the concept of definiteness. Although there is no complete agreement on the exact notions implied in the semantics of definite nominal expressions,3 the relevant literature agrees that a definite expression refers to a unique object, i.e., to an object that can be identified as the only one that is denoted by the expression in the relevant discourse. This is basically a Fregean analysis, which assumes that definite elements carry some sort of presupposition, while indefinite elements do not. This presupposition involves notions of uniqueness, familiarity or identifiability. This way when the speaker uses the president of Ghana in example (8) the presupposition is that the hearer’s knowledge of the world will allow him/her to determine the unique object to which the speaker is referring. Taking this discussion of definiteness and indefiniteness as our point of reference, this chapter shows that the definite article in Spanish has the possibility to encode some other notions, such as specificity. In the next Section 3.2.2, I focus first on indefinite nominal expressions to describe the notion of specificity. The discussion will establish the existence of two groups of indefinite nominal expressions and their most accepted properties, before introducing how these two groups of indefinite nominal expressions are represented in Spanish.

3.2.2 Specificity in Indefinite Nominal Expressions Natural languages typically encode the notion of specificity in their nominal expressions. In this section, I first illustrate the expression of specificity in nominal expressions headed by indefinite determiners. After that, I introduce a few syntactic contexts that force the specific reading of indefinite nominal expressions in Spanish and comment on the leading theories to explain these facts. In principle, the concept of specificity is used to differentiate between the two possible readings, cf. (9b–c), of the underlined nominal expression in examples such as (9a). (9)

a. A student in Syntax 1 cheated on the exam b. His name is John c. We are all trying to figure out who it was (Fodor & Sag (1982))

The specific reading of a student in syntax 1 in (9a) seems to link that expression to another expression that the speaker has in mind, while the non specific interpretation in (9b) does not link the nominal expression a student in syntax 1 to any previous expression. Note that the notion of specificity seems to be close to the notion of definiteness we discussed in the previous section, since with both notions the speaker

3.2 Fundamentals: Definiteness and Specificity

89

identifies a particular referent for his/her nominal expression. The difference between the concept of specificity and definiteness seems then hard to grasp. According to some authors (cf. Ionin (2003) recently), the crucial difference between the two features is that the notion of definiteness is a shared state of knowledge between speaker and hearer, while the notion of specificity is knowledge only held by the speaker. Another significant difference between the two notions is the notion of uniqueness: specific indefinite nominal expressions do not refer to the unique individual in the universe of discourse that holds the relevant property; while definite nominal expressions do. This way, the specific reading of a student in syntax 1 in (9a) identifies a particular student (i.e., John) but does not imply that there are no other students of syntax 1 in the discourse. On the contrary, the use of the definite article in the same construction (namely, the student in syntax 1 not only identifies the particular student (i.e., John) but also denotes that he is the only one who has that property in the relevant discourse. Note that, although specificity has been traditionally related to the category D, there are different mechanisms that can force a specific or non specific reading of a particular nominal expression. Let me elaborate on this by considering the following examples in Spanish. (10)

a. Busco un libro en el que se analiza el modo Search-(1.sg a book in which REF analyze-IND the mood en las oraciones de relativo [+Specific] in the clauses of relative b. Busco un libro en el que analice el modo Search-(1.sg a book in which that REF analyze-SUB the mood en las oraciones de relativo [-Specific] in the clauses of relative ‘I’m looking for a book in which the mood in relative clauses is analyzed’

The examples in (10) illustrate that specificity can be expressed through the mood chosen in the relative clauses in Spanish (cf. Leonetti (1999), von Heusinger and Kaiser (2003), among others). The difference between (10a) and (10b) is that the speaker has a book in mind when s/he utters (10a); while s/he is uncertain of the particular book s/he is looking for in (10b), that is, the existence of the book is not implied in the non specific reading. Another mechanism that makes specific readings more salient in Spanish is the use of the preposition a (to) preceding the object (i.e., what is known as differential object marking). In the relevant literature on the topic (cf. Laca (1987), Masullo (1992), Torrego (1998), Leonetti (1999), among many others), it is common to find that the P a (to) in Spanish is a specificity marker in object position.4 Consider the examples in (11), where the indefinite nominal expression una chica (a girl) is preceded by the P a (to), (11b), or appears without the P a (to), (11a).

90

(11)

3 Specificity, Definiteness and the Definite Article in Spanish Nominal Expressions

a. Vimos una chica (we)saw a girl ‘we saw a girl’ (non specific) b. Vimos a una chica (we)saw to a girl ‘we saw a girl’ (specific)

In the first case, the indefinite nominal expression refers to any individual that can satisfy the predicate girl, while in the second example, when the indefinite is preceded by a (to), the nominal expression refers to a particular girl that the speaker has in mind. This way, the presence of the preposition a (to) seems to trigger the specificity reading in indefinite nominal expressions.5 There have been two main approaches to the notion of specificity: some authors think that it is some sort of a grammatical feature, while some others claim that it is the result of a pragmatic effect. For those who support the latter view (cf. Leonetti (2004), among others) the crucial point is that the notion of specificity lies on the speaker’s intention. On the contrary, for the competing line of research (cf. Campbell (1996), Karimi (1999), and many others), specificity is a grammatical feature, encoded differently depending on the particular language. The hypothesis in this chapter is that specificity is encoded in the functional category D; in other words, the claim is that specificity is one of the interpretable features that justify the existence of the functional category D in our theory of grammar. Now that I have introduced the notion of specificity and illustrated it with some examples of indefinite nominal expressions in Spanish, I will explore specificity in definite nominal expressions in Section 3.2.3.

3.2.3 Specificity in Definite Nominal Expressions This section explores the relation between specificity and definite nominal expressions. I will show that it is not always the case that the definite article and other definite determiners express both definiteness and specificity. I will also compile some evidence in this section for the claim that some definite nominal expressions are non specific, before analyzing their behavior in the following sections. Let me start the discussion by considering the examples in (12), parallel to the examples in (10) in the previous section. Remember that these type of examples illustrated that specificity can be expressed through the mood chosen in the relative clauses in Spanish (cf. Leonetti (1999), von Heusinger and Kaiser (2003), among others). Crucially, note that this contrast is also observed with [+ definite] nominal expressions (cf. Leonetti (1999: 865)), as illustrated in (12). (12)

a. Busco el libro en el que se analiza el modo Search-1.sg the book in which REF analyze-IND the mood

3.2 Fundamentals: Definiteness and Specificity

91

en las oraciones de relativo [+Specific] in the clauses of relative b. Busco el libro en el que se analice el modo Search-1.sg the book in which REF analyze-SUB the mood en las oraciones de relativo [-Specific] in the clauses of relative ‘I’m looking for the book in which the mood in relative clauses is analyzed’ Both sentences in (12) have the definite article in Spanish. Therefore, these sentences show that it is the specificity feature (and not definiteness) that determines the use of mood in relative sentences. Note that the difference between (12a) and (12b) is that the speaker has a book in mind when s/he utters (12a); while s/he is uncertain of the particular book s/he is looking for in (12b), that is, the existence of the book is not implied in the non specific reading. This way, the examples in (12) also illustrate that there are definite nominal expressions that are non specific. Alternatively, there are also contexts in Spanish when the presence/absence of a category depends exclusively on the definiteness feature and not on the specificity feature. According to Fernandez Soriano (1999), among others, clitic doubling with direct objects primarily depends on the category definiteness and not on the category specificity. The examples in (13) show that, in Rı´ o de la Plata Spanish, object clitic doubling is only allowed when the object is [+ definite], (13a–b) vs. (13c). (13)

a. La veo a Marı´ a Cl-Acc see-(1.sg Maria ‘I see Maria’ b. La veo a la mujer6 Cl-Acc see-(1.sg the woman ‘I see the woman’ c. *La veo a una mujer Cl-Acc see-(1.sg a woman

To sum up, there is evidence that definiteness and specificity are different features that determine different phenomena, even if both of them have been typically associated to the D category. Note that this observation is not new. Since Donnellan (1966), many authors have pointed to the existence of (at least) two uses of definite nominal expressions: a referential use of definite expressions and an attributive one. Both readings can be found in Donellan’s classical example in (14). (14) Smith’s murderer must be insane. In the referential reading Smith’s murderer has an identity. He is for instance Smith’s cousin. In the attributive reading Smith’s murderer has not yet been identified. Anyone can be the murderer. Hence, the referential use of definite nominal expressions corresponds to their specific use, while the attributive use equals to the non specific use of definite nominal expressions.

92

3 Specificity, Definiteness and the Definite Article in Spanish Nominal Expressions

As seen so far, there is enough supporting evidence to postulate a division of definite nominal expressions into two groups: specific definite nominal expressions and non specific definite nominal expressions. The difference in meaning between these two groups is shown in the following examples (taken from Ionin (2003:87–90). (15) [+Definite, +Specific] Definite, Wide scope, Speaker knowledge. At a bookstore Chris: Well, I’ve bought everything that I wanted. Are you ready to go? Mike: Almost. Can you please wait a few minutes? I want to talk to the owner of this bookstore. She is my old friend. (16) [+Definite, -Specific] Definite, Narrow scope, No speaker knowledge. At a supermarket Sales clerk: May I help you, sir? Customer: Yes! I’m very angry. I bought some meat from this store, but it is completely spoiled! I want to talk to the owner of this store. I don’t know who he is, but I want to see him right now! The relevant difference between (15) and (16) is that the speaker has a person in mind when s/he utters (15); while s/he is uncertain of the particular person s/he is talking about in (16), that is, the owner of the store has not been identified yet. Recent research ((cf. Ionin (2003), Julien (2005), among others) has explored this difference as a means to account for some phenomena that remained without satisfactory explanation. For instance, Ionin (2006) applies this difference to provide an explanation of the referential and non referential uses of demonstratives such as this in English, while Julien (2005) explains the syntactic and semantic properties of the Scandinavian definite nominal expressions by assuming that they are composed of different features that appear in different structural positions. Therefore, the assumption that there are different uses of definite nominal expressions regarding their specificity and that the different features encoded by definite nominal expressions can appear in different locations of our structure finds support in the cross-linguistic data. To sum up, this section has explored the notion of specificity and its relation with the notion of definiteness. I have described what a specific definite nominal expression is and provided some examples of how Spanish represents this type of nominal expression. One of the main claims of this section is that there exists the possibility of having non specific definite nominal expressions in Spanish. The consequence of this claim is that definite nominal expressions have two distinct features that can be isolated in the adequate contexts. In the rest of the chapter, I make use of the different feature composition of determiners to account for the syntactic properties, such as the possibilities of extraction, of the different types of nominal expressions in Spanish.

3.3 Extraction, Definiteness and Specificity

93

3.2.4 Conclusions Two crucial notions were discussed in this section: definiteness and specificity. Apart from defining and illustrating both notions, this section has detailed some of the ways that Spanish expresses these notions. Furthermore, a few different syntactic strategies to favor one of the readings over the other were explained, paying special attention to the role of mood and the P a (to) as specificity markers. One of the main claims of this section is that there exists the possibility of having non specific definite nominal expressions in Spanish. In the rest of the chapter, I make use of the different feature composition of every kind of determiner to account for the syntactic properties of the different types of nominal expressions in Spanish; more specifically, I will be focusing on how the different determiners affect the possibilities of extraction.

3.3 Extraction, Definiteness and Specificity This section focuses on the behavior of [+/-definite] and [+/- specific] nominal expressions regarding extraction facts. In Chapter 2 we illustrated that nominal expressions in Spanish allow their extraction out of nominal expressions if they were the unique element in the structure, cf. (17). (17)

a. Vimos varias copias del libro. (we)saw several copies of the book. ¿De que´obj visteis vosotros varias copias? Of what (you)saw several copies? b. Vimos varias copias de Juan. (we)saw several copies of Juan. ¿De quie´nposs/agent visteis vosotros varias copias? Of whom (you)saw several copies?

The examples in (17) illustrate that objects, agents and possessors are able to move out of the nominal expression varias copias de ___ (several copies of ____) whenever they are the only PP in the nominal expression. Consider now (18)–(19): (18)

a. Vimos las copias del libro. ¿De que´obj visteis vosotros las copias? (we)saw the copies of the book. Of what (you)saw the copies? b. *Vimos las copias de Juan. (we)saw the copies of Juan. ¿De quie´nposs/agent visteis vosotros las copias? Of whom (you)saw the copies?

(19)

a. *Vimos estas copias del libro. ¿De que´obj visteis vosotros estas copias? (we)saw these copies of the book. Of what (you)saw these copies?

94

3 Specificity, Definiteness and the Definite Article in Spanish Nominal Expressions

b. *Vimos estas copias de Juan. (we)saw these copies of Juan. ¿De quie´nposs/agent visteis vosotros estas copias? Of whom (you)saw these copies? Surprisingly, the examples in (18) and (19) do not follow the generalizations put forward in Chapter 2, since extraction out of the nominal expressions in these examples is banned in most cases. Therefore, the analysis offered in Chapter 2 will not be able to account for the new set of data. This chapter examines the extraction facts with non specific nominal expressions in Spanish and concludes that the properties of the determiner are the cause for the atypical extraction facts seen in these nominal expressions. This section focuses on detailing the extraction possibilities with different kinds of nominal phrases. I will show that the extraction of PPs out of nominal expressions headed by determiners others than non specific indefinites has a more restricted pattern. More concretely, I illustrate that specific definite nominal expressions do not allow extraction out of the nominal expression, Section 3.3.2, that extraction out of nominal expressions headed by the definite article is severely restricted, Section 3.3.3 and the behavior of specific indefinites, Section 3.3.4. The data presented in this section will constitute the empirical base of the analysis introduced in the next section.

3.3.1 The Specificity Effect This section is devoted to establishing the basis of comparison for some of the phenomena that will be discussed later. Hence the section discusses the most studied restriction on extraction out of nominal phrases: the specificity effect. The specificity effect, as described in Chomsky (1986), Fiengo and Higginbotham (1980), among others, is displayed by data such as (20): (20)

a. Whoi did you see pictures/a picture of ti? b. * Whoi did you see the/these pictures of ti?

As the contrast in (20) shows, wh-movement out of specific nominal expressions (i.e., nominal expressions headed by the definite article or a demonstrative) is impossible in English. The generalization commonly accepted in the literature is that wh-movement out of specific nominal expressions is excluded because of the presence of a specific determiner. There are two main lines of research to provide an explanation for data such as (20). On the one hand, there is a line of research that claims that the semantic properties of the structure prevent the extraction to happen (cf. Erteschik-Shir (1973) as one of the first accounts of this type). For instance, Davies and Dubinsky (2003) develop an analysis of extraction out of nominal expressions following this line of research and conclude that the type of noun and the referentiality properties of the nominal expression play a greater role than the syntactic structure of the

3.3 Extraction, Definiteness and Specificity

95

construction. They illustrate their claim with examples such as (21), in which the noun examination can be a process, (21a), or a result, (21b).7 (21)

a. Which patient did the surgeon forget to observe the examination of ? b. *What subject was the student hoping to pass the examination of ? Davies & Dubinsky (2003: 17)

Interestingly, extraction out of the process interpretation of the noun examination is allowed even in cases where the definite article is present in the nominal expression. The grammaticality status of (21b) shows that extraction is banned out of the result reading of the same noun. This type of generalization does not hold in Spanish. As shown through the examples of Chapters 2 and 3, Spanish permits extraction out of the result and concrete nominal expressions.8 On the other hand, it has been assumed that the semantic properties of the determiner lead to a different syntactic configuration of the nominal expression, which prevents the extraction out of its domain. This general approach, which will be the one adopted in this chapter, has received abundant supporting evidence. For instance, one popular explanation (Chomsky (1986), Fiengo and Higginbotham (1980), Torrego (1987), among others) is that the determiner blocks the relation between the wh-operator and its trace, which behaves as a variable and cannot be bound. This way the nominal expression becomes a syntactic island for movement. Following this line of research, and taking as the departure point for his analysis certain structural similarities between the clausal and the nominal domain, Campbell (1996) claims that there is a subject predicate structure in the nominal domain, which corresponds structurally to a small clause selected by an operator in the specific Determiner Phrases (DPs). Then the impossibility of extraction out of specific DPs is derived by the presence of an operator in the specifier of the DP, which binds the variable in the subject position of the small clause and prevents the movement out of the nominal expression via its specifier. A similar instantiation of the last line of research (cf. Karimi (1999), among others) argues that the specificity effect cannot be explained solely on the basis of the semantic properties of the determiner or the DP. Building her theory upon modern Persian data, Karimi (1999) shows that a specific nominal expression is subject to the specificity effect only if the specifier of the construction is lexically filled; it is only in this case that the nominal expression can be considered an island. Recently Bosˇ kovic´ (2005) has also proposed that certain extraction effects must be derived from the different structural properties of nominal expressions. His proposal takes into consideration Uriagereka’s (1988: 113) observation that only languages without overt determiners allow certain types of extraction to derive the Left Branch extraction (LBE) effects (i.e., the impossibility of extracting an element in the left branch of the structure) cross-linguistically. Bosˇ kovic´ (2005) concludes that there are two types of languages with respect to the type of nominal structure they project and he proposes this as an argument for his parameterized DP hypothesis, which proposes that nominal expressions have

96

3 Specificity, Definiteness and the Definite Article in Spanish Nominal Expressions

different structures depending on the type of language.9 More concretely, languages without articles allow certain extractions because their nominal expressions are bare NPs (the N-over-A structure), while languages with articles do not allow some movements out of the nominal expression because their nominal expressions are DPs (the A-over-N structure). The analysis developed in this chapter follows this line of research and proposes that we can find structural variation in the same language, since nominal expressions do not have obligatorily the functional category DP. This Section 3.3.1, has been a brief overview of the so called specificity effect, which was illustrated in English and other languages. In the next Section 3.3.2, I explore this phenomenon and introduce the relevant cases, in Spanish. The conclusion of the next section will set the stage for the conclusion of section 3: definite nominal expressions do not need to have the same properties as specific nominal expressions in Spanish. This book adopts a purely syntactic explanation of the specificity effect. I will argue that the presence of the feature [+ specific] forces the projection of the functional category D, which in turn produces violations of locality because the movements out of nominal expressions become too long.

3.3.2 Extraction Out of Specific Nominal Expressions Spanish also has cases of the so-called specificity effects in the literature and the properties of wh-extraction from Spanish specific nominal expressions are very different from the properties of wh-extraction out of non specific nominal expressions in this language. Let us consider the examples in (22): (22)

´ a. *¿De que´ fotografo has visto [estas fotos tag]? of which photographer (you)have seen these photos tag b. *¿De que´ coleccionista has visto [estas fotos tposs]? of which collector (you)have seen these photos tposs c. *¿De que´ cantante salieron publicadas [estas fotos tobj]? of which singer were published these photos tobj

The ungrammaticality of the data in (22), where the nominal expression appears with demonstratives, shows that Spanish has analogous cases to the ones shown in (20) for English. Furthermore, the grammaticality status of the examples in (22) illustrates that, as in English, specificity effects in Spanish do not discriminate among agents, possessors and objects. The extraction of any of them causes ungrammaticality.10 To sum up: Spanish has specificity effects, which are similar to the ones discussed in the relevant literature in other languages. In the next section, I compare the behavior of the specific nominal expressions and the behavior of nominal expressions with the definite article, to unveil the extraction properties of nominal expressions headed by the definite article in Spanish. I will show that

3.3 Extraction, Definiteness and Specificity

97

the specific effects do not hold with all nominal expressions headed by the definite article. The results of this examination will lead me to argue for the existence of a dummy or weak definite article in Romance languages, which does not have the full set of features typically associated to the definite article.

3.3.3 Extraction Out of Nominal Expressions with the Definite Article The extraction of arguments out of nominal expressions headed by the definite article follows a very peculiar pattern in Spanish. As the data in (23) illustrate, extraction out of a definite nominal expressions is severely restricted. (23)

a. *¿De que´ autor has leı´ do [los libros tag]? of which author (you)have read the books tag b. *¿De quie´n has visto [las fotos tposs]? of whom (you)have seen the photos tposs c. ¿De que´ cantante salieron publicadas [las fotos tobj]? of which singer were published the photos tobj

(23a) and (23b) show that agent or possessor extraction out of a definite nominal expression is banned; (23c) illustrates that extraction of objects out of definite nominal expressions is allowed.11 Therefore, it seems that the grammaticality of examples such as (23c) argues against a uniform analysis of all the examples, based, for instance, on the specificity effect. This is because, if the ungrammaticality of (23a) and (23b) was due to the fact that the nominal expression is a specific nominal expression and bans extraction of its constituents, the grammaticality of (23c) would remain without an explanation. In this chapter, I will argue that the properties of the definite article in Romance are the key to obtain an answer for the non uniform blocking effects of definite nominal expressions. More concretely, I will propose that the existence of a weak definite article in Spanish allows for some extraction possibilities out of the definite nominal expressions of this language, cf. Section 3.4 for a detailed analysis.

3.3.4 Extraction Out of Specific Indefinite Nominal Expressions Chapter 2 described the extraction possibilities out of indefinite nominal expressions in Spanish. In that chapter, examples similar to the ones in (24) illustrated that extraction out of indefinite nominal expressions was possible when there was no more than one argument in the nominal structure. (24)

a. Vimos varios libros de lingu¨ı´ stica. (we)saw several books of linguistics. ¿De que´obj visteis vosotros varios libros? Of what (you)saw several books?

98

3 Specificity, Definiteness and the Definite Article in Spanish Nominal Expressions

b. Vimos varios libros de Juan. (we)saw several books of Juan. ¿De quie´nposs/agent visteis vosotros varios libros? Of whom (you)saw several books? In this chapter, Section 3.2.2, it was shown that indefinite nominal expressions can be divided into two groups, specific or non specific, depending on the presence (or not) of a specificity feature. Furthermore, Sections 3.3.1 and 3.3.2 have shown evidence linking the presence of the [+ specific] feature to the impossibility of extraction out of the nominal expression. A natural question at this point is how the specific indefinite nominal expressions behave with respect to extraction in Spanish. This section provides an answer for this empirical question, showing that extraction is not possible out of specific indefinite nominal expressions. This generalization is then consistent with the general extraction pattern discussed in the chapter so far and concludes the empirical discussion of this part of the chapter. There are certain ways to favor the specific reading of indefinite nominal expressions. First, the presence of elements such as cierto (certain) or en particular (in particular) forces the specific reading. Consider (25): (25)

a. Juan quiere comprar un cierto libro de lingu¨ı´ stica de Chomsky Juan wants to-buy a certain book of linguistics of Chomsky b. Juan quiere comprar un libro de lingu¨ı´ stica de Chomsky en particular Juan wants to-buy a certain book of linguistics of Chomsky c. *¿De que´ quiere Juan comprar un cierto libro/ un libro en particular? Of what wants Juan to-buy a certain book/ a book in particular d. *¿De quie´n quiere Juan comprar un cierto libro/ un libro en particular? Of whom wants Juan to-buy a certain book/ a book in particular

In both (25a) and (25b), the speaker intends to refer to a unique individual x; for instance, Juan wants to buy Barriers. That is, the indefinite nominal expressions in these examples are specific. Interestingly, neither the object nor the agent/possessor can be extracted out of the relevant nominal expression, (25c) and (25d), when they are the only PP argument in the nominal structure. As seen in Section 3.2.2, another mechanism that makes specific readings more salient in Spanish is the use of the preposition a (to) preceding the object. In other words, by using the P a (to) in (26a) the nominal expression dos estudiantes de lingu¨ı´stica (two students of linguistics) has a particular set of two individuals as its referent (i.e, John and Mary). On the contrary, the nominal expression dos estudiantes de lingu¨ı´stica (two students of linguistics) in (26b) refers to any two students that are studying linguistics and the speaker does not refer to a particular set of individuals. The extraction possibilities seem to be affected by the presence of the P a too, cf. (26c) and (26d). (26)

a. Han visto a dos estudiantes de lingu¨ı´ stica (they) have seen to two students of linguistics

3.3 Extraction, Definiteness and Specificity

99

b. Han visto dos estudiantes de lingu¨ı´ stica (they) have seen two students of linguistics c. ¿De que´ han visto dos estudiantes? Of what (they) have seen two students d.??/* ¿De que´ han visto a dos estudiantes?12 Of what (they) have seen to two students In (26c), it is illustrated that a non specific indefinite nominal expression allows for the extraction of its PP argument; however, the ungrammatical status of (26d) shows that this is not the case with specific indefinite nominal expressions. To summarize: this section has presented some unobserved data that show that extraction is not possible out of specific indefinite nominal expressions. The data and conclusion in this section confirm previous conclusions in the chapter that point to the feature [+ specific] as the relevant feature in order to determine if a nominal expression allows for the extraction of its argumental PPs.

3.3.5 Conclusions Section 3.3 has revealed a series of surprising facts regarding extraction out Spanish nominal expressions. Table 3.1 is a graphic summary of the section. Table 3.1 shows that there are different extraction possibilities depending on the type of determiner that heads the nominal expression and (a priori) on the semantics denoted by the nominal expression. This way, non specific indefinite nominal expressions allow for extraction of any of their argumental PPs, as seen in Chapter 2, if they are the only element in the structure. On the contrary, specific definite and specific indefinite nominal expressions do not permit extraction out of them. Finally, nominal expressions headed by the definite article seem to follow a distinctive pattern and tolerate extraction of object PPs but ban extraction of possessor and agent PPs. In this section we have then found enough empirical evidence to link the analysis of extraction out of nominal expressions to the presence or absence of specificity in the expression. The next section proposes an analysis that takes seriously this relationship and proposes that the presence of the interpretative feature [+ specific] in a nominal expression triggers the projection of the functional category DP. The

Table 3.1 Extraction out of nominal expressions containing one argumental PP13 Definite Indefinite Extracted element Def art non-def art Spec non-spec Object PP Agent PP Poss PP

OK NO NO

NO NO NO

NO NO NO

OK OK OK

100

3 Specificity, Definiteness and the Definite Article in Spanish Nominal Expressions

presence/absence of the functional category DP and the general locality conditions already introduced in Chapter 2 will be enough to derive the extraction properties of the different types of nominal expressions in Table 3.1.

3.4 A Syntactic Analysis This section details a syntactic analysis for the nominal expressions and the generalizations in Table 3.1 above. The analysis assumes that the possibility of extracting out of a nominal expression is due to the lack of DP projection in that particular case. More concretely, I will argue that only strong determiners project the functional category DP, while weak determiners are generated in a lower functional category, Agr. The combination of this assumption with the locality conditions operating in the Spanish nominal expressions enable us to derive their extraction possibilities. The section starts by lying out the basic assumptions of the analysis (Section 3.4.1). Here, I summarize the main assumptions of the analysis for Spanish nominal expressions introduced in Chapter 2 and detail how the different kinds of determiners are generated. The section continues by applying the basic analysis to the relevant cases illustrated above. In Section 3.4.2, the section puts forward the analysis for specific definite nominal expressions and in Section 3.4.3 we test it with the data of specific indefinite nominal expressions. In Section 3.4.4, the section examines with some detail the properties of definite nominal expressions in Spanish. First, I introduce the behavior of the definite article and the previous approaches to analyze its atypical behavior. This section will then establish the main features of the definite article in Spanish, namely [+/- definite], [+/- specific]. In this section, I demonstrate that the structure and locality conditions described in Chapter 2, combined with the special properties of definite articles in Spanish, can resolve not only the puzzle with the extraction possibilities out of definite nominal expressions but also the specificity effect in Spanish. More concretely, I propose that the definite article can be generated in two different positions: as the head of Agr when it behaves as a weak determiner and allows extraction out of the nominal expression; as the head of D when it behaves as a strong determiner and disallows the extraction out of the DP. Finally, the last part of this Section 3.4.5 brings out some additional evidence for the type of analysis presented.

3.4.1 Basic Assumptions The analysis developed in this book assumes a basic structure of Spanish nominal expressions that is divided into three major domains, mimicking recent developments of the clausal domain. Each of the domains provides a derivational space in which elements can establish certain kind of relationships. For

3.4 A Syntactic Analysis

101

instance, the most embedded domain is the theta-domain and is devoted to satisfy the basic selectional requirements of the N predicate; the second domain is the agreement domain, where the agreement relations needed can be established. Finally, the least embedded domain is the discourse domain, which links the structure to the rest of the discourse. The different domains contain a limited number of functional categories (see Chapter 2 for a detailed description of the different functional categories in the structure), as shown in the basic structure proposed for the relevant constructions, which is repeated in (27). (27)

FocP Foc’ ω-domain Foc

DP D’

D

AgrP ϕ-domain Agr’

Agr

POSS

nP n’

n

AGENT θ-domain

NP

N

OBJ

I also assumed the division of determiners into two classes. Milsark (1977) showed that determiners can be divided into two classes, strong and weak, depending on their syntactic behavior. A diagnostic for classifying quantificational expressions as either strong or weak draws upon the observation that existential there-sentences may contain certain noun phrases (i.e., those considered weak) following the copula but some other noun phrases (i.e., those considered strong) are ungrammatical in the same linguistic environment, as shown in examples (28a) and (28b). (28) a. There is a man in the garden. b. *There is the/every man in the garden. Example (28) is then ungrammatical because a defining characteristic of strong determiners is that they carry a presupposition of existence that makes strong determiners incompatible with existential there constructions (cf. Heim (1982), among others).

102

3 Specificity, Definiteness and the Definite Article in Spanish Nominal Expressions

This division of determiners has been extensively used in subsequent literature on nominal expressions (cf. Zamparelli (2000) for an extensive summary on this topic) and was the base for proposing (following Abney (1987), Bernstein (1993), Zamparelli (2000), among many others) that not all determiners were generated in D. Thus, following the goal of restricting the number of functional categories, I assume that only the presence of a strong determiner triggers the projection of DP and that weak determiners appear generated in a lower projection, which I identify here as AgrP.14 The examples in (29a–b) provide examples of every category of nominal expression and (29c–d) shows the representation under the current analysis (where irrelevant details have been omitted for ease of exposition). (29) a. Compramos una casa/ varias casas (we)bought a house/several houses b. Compramos esta casa/ cada casa (we) bought this house/every house AgrP c. c. Agr’ Agr una

NP casa DP

d. D’

AgrP

D esta Agr’ Agr

NP casa

Note that this proposal is consistent with previous research that assumes a correspondence between the two kinds of nominal expressions, argumental and non argumental expressions and two different syntactic structures for nominal expressions. For instance, this proposal would be coherent with Longobardi (1994), who claims that only DPs are referential expressions, while NPs are non referential expressions.15

3.4 A Syntactic Analysis

103

Another line of research that is compatible with the analysis put forward in this chapter is the one that relates specific readings of nominal expressions to the discursive status of the nominal expression (cf. Enc (1991), Diesing (1992), and more recently Leonetti (2004) and Lima (2006), among many others). Although these proposals vary regarding the exact mechanism used to analyze the relevant constructions, all of them have in common that they claim that there is a relation between specificity and topicality. Under the current analysis this relation is possible because the current analysis associates the specificity feature with the projection of a functional category, DP, in the discourse domain of the nominal expression.16 Finally, the last assumption made in Chapter 2 in order to explain the typical patterns of extraction out of indefinite nominal expressions was that movement within nominal expressions is highly restricted. It cannot take place further than one maximal projection and movement must be from one domain to another. The few independently needed assumptions summarized above make it possible to explain the basic pattern of extraction out of non specific nominal expressions in Spanish, as well as the intricate relations among the different types of PP elements within the nominal expression. In this section, I will show that this analysis is also able to accommodate the (in principle) atypical behavior regarding extraction out of definite and specific nominal expressions.

3.4.2 Extraction Out of Specific Definite Spanish DPs Let us first reexamine the specificity effects observed in Spanish. In Section 3.3 we illustrated that the specificity effect, as described in Chomsky (1986), Fiengo and Higginbotham (1980), among others, exist in Spanish. The descriptive generalization is then that extraction out of specific nominal expressions is generally banned. Note that the determiners (henceforth Ds) that head a specific nominal expression correspond to the Ds that are typically considered strong Ds in the literature: (30) a. *¿De que´ autor has leı´ do [{todos los/ estos} libros tag]? of which author (you)have read all the/ these books tag a’. *¿De que´ autor has leı´ do [{cada/todo} libro tag]? of which author (you)have read each/every book tag b. *¿De quie´n has visto [{todas las/estas} fotos tposs]? of whom (you)have seen all the/these photos tposs b’. *¿De quie´n has visto [{cada/todo} foto tposs]? of whom (you)have seen each/every photo tposs c. *¿De que´ cantante salieron publicadas [{todas las/ estas} fotos tobj]? of which singer were published all the/these photos tobj c’. *¿De que´ cantante salieron publicadas [{cada/todo} foto tobj]? of which singer were published each/every photo tobj

104

3 Specificity, Definiteness and the Definite Article in Spanish Nominal Expressions

According to the conclusions reached so far, the abstract structure of a specific nominal expression under my analysis is (31): (31)

FocP ω-domain Foc’ Foc

DP D’

D

AgrP

cada Agr’

ϕ-domain

POSS

Agr NP N

θ-domain

OBJ

As (31) illustrates, the presence of a strong D, such as cada (each) in examples such as (30), forces DP to be projected in the structure.17 Note that the presence of DP in the structure derives straightforwardly the impossibility of extraction out of specific DPs: any movement from an element in the spec of AgrP to the spec of FocP (cf. (32)) will not meet the locality conditions on movement, since this movement would cross more than one maximal projection; that is, the wh-element moves from the spec of AgrP to the spec of FocP. (32)

FocP POSS

Foc’ Foc

ω−domain

DP

D’ ϕ−domain

AgrP

D cada (. . .)

POSS

Furthermore, due to anti-locality, the extracted element cannot land in the spec of DP, since that would necessarily involve a second movement within the

3.4 A Syntactic Analysis

105

o-domain to escape out of the FocP. This movement is illustrated in (33), where I disregard the lower part of the structure: (33)

In short: the presence of the feature [+ specific] in strong Ds causes them to be generated in a DP. Then, the projection of the DP category in specific DPs causes the wh-movements of modifiers within DP element to violate the locality conditions on movement that elements within DPs must satisfy. In the next sections I extend this analysis to the nominal expressions containing specific indefinite Ds and definite articles. First, in Section 3.4.3, I deal with the representation of specific indefinite nominal expressions and after that, I analyze the surprising behavior of the definite article in Spanish nominal expressions with respect to extraction.

3.4.3 Explaining the Blocking Effects Out of Specific Indefinite Nominal Expressions In Section 3.3.4 it was illustrated that indefinite nominal expressions can block the extraction of their PP arguments when they are used as specific. The examples below show that the extraction of PP arguments out of indefinite nominal expressions preceded by the P a (i.e, specific indefinites) is banned, while the extraction out of those indefinite constructions in their non specific reading is allowed. (34)

a. Juan busca a un estudiante de lingu¨ı´ stica Juan looks for to a student of linguistics b. Juan busca un estudiante de lingu¨ı´ stica Juan looks for to a student of linguistics c. ¿??/*De que´ busca Juan a un estudiante of what looks for Juan to a student d. ¿De que´ busca Juan un estudiante?

specific Non specific

106

3 Specificity, Definiteness and the Definite Article in Spanish Nominal Expressions

The previously unobserved data led us to conclude that it was not possible to extract elements out of indefinite nominal constructions in their specific use. The analysis detailed in the Section 3.4.2 can be easily extended to the examples above if we consider that specific is a feature host in the D position, then the structure for the specific nominal expression in (34a) would be as shown in (35), where the specificity of the nominal expression causes the projection of the functional category D. (35)

FocP ω-domain Foc’ Foc

DP D’ AgrP

D a

POSS de Juan

Agr’ Agr un N Estudiante

ϕ-domain

ϕ-domain NP OBJ de lingüística

The structure in (35) shows that the indefinite article is still generated in Agr but the presence of the feature [+ specific], which is spelled out as the P a (to) in this case,18 is causing the projection of the D category. The blocking effects can then be straightforwardly explained under the current approach as a sub case of the general specificity effects derived in Section 3.4.2. In other words, under the current approach, the projection of the DP category in the specific use of indefinite nominal construction causes the wh-movements of PP arguments to violate the locality conditions on movement. Although the examples in (34) have what has been considered an explicit specific marker, the P a (to), indefinite nominal expressions can be specific and block extraction of their argumental PPs without displaying a marker for the D [+ specific] head. In those cases, this type of analysis assumes that the DP is projected and a non explicit [+ specific] D head present in the structure. Note that this line of analysis is fully compatible with some other independently proposed explanations to derive the specific readings of indefinite nominal expressions. For instance, Bleam (2006) accounts for

3.4 A Syntactic Analysis

107

the differential object marking in Spanish by assuming that there is a direct mapping between the presence of the P a (to) and the semantic type of the nominal expression. This way, specific nominal expressions, which are preceded by the P a (to), are individuals (type <e>) or generalized quantifiers (type <<e,t>t>). On the contrary, those indefinite nominal expressions that do not receive a specific reading are considered property-denoting nominals, which are type <e,t>. Bleam (2006) claims that the semantic difference is encoded as a structural difference and that only specific indefinite expressions are DPs, while non specific indefinite expressions are NPs or NumP. The proposal put forward in this chapter shares this main line of research and provides it with some important empirical support. Now that the basic analysis of specific and non specific nominal expressions has been developed, the chapter returns to the examination of the most atypical extraction facts seen so far: the correspondence of the extraction facts out of nominal expressions headed by the definite article.

3.4.4 On the Spanish Definite Article This chapter has compiled enough evidence to support an analysis of specific and non specific nominal expressions in terms of DP versus non DP constructions. The analysis developed is still not able to provide an account for the extraction possibilities out of nominal expressions headed by the definite article in Spanish. In previous sections, it was shown that the relevant descriptive generalizations to explain this are that only objects can undergo extraction out of nominal expressions headed by the definite article in Spanish. In the first part of this Section, 3.4.4.1, I introduce previous insights about the properties of definite articles in Spanish, which consider that there are two types of definite article in this language; namely, a weak definite article and a strong definite article. Subsequently, in Section 3.4.4.2, I explore those proposals and develop an analysis of extraction out of definite nominal expressions that is based on them. More concretely, I claim that the weak definite article in Spanish lacks the [+ specific] feature and allows extraction out of its nominal expression because it does not project a DP category;19 on the contrary, the strong definite article has a [+ specific] feature, which causes the nominal expression to be a DP and bans extraction. After that (Section 3.4.4.3), I demonstrate that previous attempts to explain the definite article contrasts in Spanish cannot account for all the facts and make inaccurate predictions, because they took into consideration only the properties of the different types of arguments (subject vs. objects) but did not contemplate the different properties of the definite article.

108

3 Specificity, Definiteness and the Definite Article in Spanish Nominal Expressions

Finally, in Section 3.4.4.4, I present some additional evidence for the analysis in this section. This additional evidence comes from the interaction of the definite article with the scope of some quantifiers and from some unobserved properties of generic nominal expressions in Spanish.

3.4.4.1 Previous Approaches to the Definite Article Behavior There is a long research line (cf. Torrego (1987), Ormazabal (1991), Vergnaud and Zubizarreta (1992), Longobardi (1994), Gue´ron (2003), among others) that is based on the assumption that there are two versions of the definite article in Spanish. Thus, the existence of two different definite articles in Romance has been proposed to account for the different properties of definite articles depending on the context.20 The presence of a weak version of the definite article is responsible for the grammaticality of data such as the ones exemplified in (36): (36)

a. El Pepe vino ayer the Pepe came yesterday b. ¡Las cervezas que te bebiste anoche! the beers that CL (you)drank last night ‘The many beers you drank last night!’ c. Considero a Juan el mejor amigo (I)Consider Juan the best friend

The examples in (36) illustrate some atypical uses of the definite article in Romance. (36a) shows that the definite article can appear with proper Ns; this has been viewed by authors such as Longobardi (1994) as an indication of the existence of a weak definite article (i.e., an expletive definite article) in Romance languages. In the same way, the quantificational use of the definite article in (36b) is taken as evidence in Torrego (1987) for the existence of a weak definite article in Spanish. Finally, the example in (36c) illustrates the special properties of the definite article when it is heading a superlative construction. Note that in all these uses of the definite article it behaves as a weak determiner; that is, as a non specific determiner. Moreover, consider now the examples in (37). The different readings available in (37) show another difference between specific nominal expressions and nominal expressions with the definite article in Spanish: (37)

a. Juan vio una foto de todo el mundo Juan saw a picture of everybody b. Juan vio la foto de todo el mundo Juan saw the picture of everybody c. Juan vio esta foto de todo el mundo Juan saw this picture of everybody

narrow/wide narrow/wide

narrow/*wide

3.4 A Syntactic Analysis

109

The sentence in (37a) is ambiguous. It can mean Juan saw one picture of a group (narrow scope reading) or Juan saw several pictures (wide scope reading). Since May (1977), a syntactic movement (quantifier raising) in LF of the quantified element todo el mundo (everybody) has been assumed to derive these two readings. Note that the sentence in (37b) shows the same ambiguity, while the sentence in (37c) is not ambiguous: it only has the narrow scope reading. Thus, assuming that the wide scope reading of todo el mundo (everybody) in (37a) and (37b) is due to LF movement of the quantified element todo el mundo (everybody) outside of the object NP, we can conclude that Spanish nominal expressions with the definite article allow for extraction of their arguments, while Spanish specific nominal expressions do not. The contrasts illustrated above between the data with extraction out of nominal expressions headed by the definite article and the data with extraction out of specific nominal expressions also lead to the conclusion that a priori we are dealing with two separate phenomena in Spanish. In addition, as Tellier and Valois (1995) note, the existence of two different definite articles in Romance provides an explanation for the lack of specificity effects in French and Italian nominal expressions headed by the definite article. According to these authors, French and Italian allow extraction out of those constructions. The relevant example is given in (38): (38) De quel auteur as-tu lu le livre? of which author have-you read the book Note that, as (38) shows, the possibility of extraction out of nominal expressions headed by the definite article in French is more general than in Spanish, as we can extract agents, de quel auteur (of which author). Furthermore, it seems that the extraction possibilities out of these constructions in French would be similar to the extraction possibilities out of non specific nominal expressions in Spanish as illustrated in Chapter 2. Leaving the technical details for next section, it seems that a plausible conclusion for the previous discussion is that there are two definite articles in Spanish: a weak definite article that behaves similarly to non specific Ds in Spanish and allows extraction out of the nominal expression; and a strong definite article that behaves as a specific D and disallows extraction. Assuming this division of Ds in Spanish (including the existence of two different definite articles), I argue in the next section that the possibility of extracting an object out of a nominal expression headed by the definite article is due to the lack of DP projection in that particular case. Bowers (1987) reaches a similar conclusion for the French data discussed above. This author argues that the difference between noun phrases that allow extraction and those that do not is explained in terms of barrierhood: noun phrases that allow extraction are NPs, while noun phrases that do not allow extraction are DPs.

110

3 Specificity, Definiteness and the Definite Article in Spanish Nominal Expressions

3.4.4.2 Extraction Out of Nominal Expressions Headed by the Definite Article Let us return now to the possibilities of extraction out of definite article nominal expressions in Spanish. The line of analysis developed previously leads me to propose that the definite article does not project DP in cases where extraction out of these constructions is possible. The data are repeated below: (39) a. *¿De que´ autor has leı´ do [los libros tag]? of which author (you)have read the books tag b. *¿De quie´n has visto [las fotos de ese monte tposs]? of whom (you)have seen the photos of that mount tposs c. ¿De que´ cantante salieron publicadas [las fotos tobj]? of which singer were published the photos tobj The ungrammaticality of examples such as (39a–b) is the result of the presence of a DP projection headed by the definite article, while the grammaticality of example (39c) can be explained by assuming that no DP projection is introduced in the relevant structure. More concretely, I propose that the definite article can be generated in two different positions: as the head of Agr when it behaves as a weak determiner and allows extraction out of the nominal expression; as the head of D when it behaves as a strong determiner and disallows the extraction out of the DP. In other words, it is again the projection of DP that prevents the extraction out of specific DPs and of certain types of definite DPs in Spanish. Additional evidence for this syntactic account comes from the data in (40), which only have the specific interpretation. (40) a. *¿De que´ autor has leı´ do [los tres libros tag]? of which author (you)have read the three books tag b. *¿De quie´n has visto [las tres fotos de ese monte tposs]? of whom (you)have seen the three photos of that mount tposs c. *¿De que´ cantante salieron publicadas las tres fotos tobj? of which singer were published the three photos tobj The data in (40) show that in cases where we need to place the definite article as the head of the DP projection, since no other position is available to host this element in the structure (the numeral filling the position where weak articles are located), no argument can be extracted. The different structures for the data in (39c) above and (40c) are illustrated in (41a) and (41b), respectively, with the strong article in D and the weak article in Agr.

3.4 A Syntactic Analysis

(41) a.

111 ω-domain

FocP de qué cantante

Foc’ Foc

AgrP ϕ-domain de qué cantante

Agr’

Agr las

NP

θ-domain

de qué cantante

N fotos

ω-domain

FocP

b. Foc’

de qué cantante

DP

Foc D’ D los

AgrP ϕ-domain Agr’

Agr tres N libros

de qué cantante

NP θ-domain de qué cantante

The existence of two definite articles in Spanish also accounts for the contrast in extraction illustrated in (42): gustan las novelas t agent? (42) a. ¿[De que´ tipo de autor]agent te of which type of author CL (you)like the novels b. *¿[De que´ autor]agent te gustan las novelas t agent? of which author CL (you)like the novels The nominal expression in (42a) is a generic one; that is, it refers to a type or class of novels, not to a particular group of novels. Note that generic nominal expressions, which are typically headed by the definite article in Spanish, allow the extraction of agents. This property of generic nominal expressions can be

112

3 Specificity, Definiteness and the Definite Article in Spanish Nominal Expressions

easily explained under my analysis, since, as commonly assumed (cf. Torrego (1987), Ormazabal (1991), among others) the definite article introduced in generic nominal expressions is the weak version of the definite article. Therefore, the presence of the definite article in (24a) does not trigger the introduction of DP in the structure, which allows for the extraction of elements out of generic definite nominal expressions. Furthermore, the assumption of the existence of a weak definite article predicts that it will appear in cases where the definite article is required exclusively due to syntactic reasons. One of these contexts is the superlative constructions in Spanish where the definite article is needed to build the superlative construction. Note that extraction out of superlative constructions is possible, as predicted: (43)

¿De que´ autor has leı´ do los libros ma´s representativos? Of which author (you)have read the books most representative

Again, this example is taken as additional evidence for the existence of a weak version of the definite article in some of its uses. Up to now, I have provided enough evidence to conclude that this type of analysis is empirically accurate. The main difficulty of the current analysis is to explain the impossibility of extracting agents or possessors out of nominal expressions with definite articles and the possibility of extraction of objects in the same contexts. In other words, the question to answer is why the presence of an agent or a possessor turns the nominal expression into a specific construction. Although this is an open question at this point, I will discuss some possible answers in Section 3.5 below.

3.4.4.3 Previous Proposals There is a competing analysis to explain the atypical extraction possibilities of nominal expressions with the definite article in Spanish. Some authors (cf. Torrego (1987), Ormazabal (1991)) have argued that the relevant property to explain the contrasts is related to the properties of the different types of arguments in the nominal expression. For instance, under Ormazabal’s (1991) approach, we can explain the difference between extraction of subjects (ungrammatical) and extraction of objects (grammatical) from definite nominal expressions as the result of the status of the initial trace that is created: The crucial fact is that object traces can be licensed by the N, via head government, since N governs its complement. However, the initial trace of a subject is not properly licensed, since N cannot govern its specifier. Then, as the initial trace of a subject is not properly licensed, it produces a violation of the ECP and the ungrammaticality of the sentence. Note that this type of analysis of extraction out of definite nominal expressions does not provide an explanation for the different behavior of specific nominal expressions in this language. This is because nouns and their ability to

3.4 A Syntactic Analysis

113

govern the traces of their arguments should be identical in definite and specific nominal expressions and the prediction would be that both types of DPs should have the same grammaticality status in the relevant extraction cases. As shown previously, this prediction is not borne out. Therefore, it seems necessary to focus on the different properties of the determiners that head the construction to achieve an analysis empirically adequate in this case. Moreover, although the exact reason why the presence of an object prevents the projection of a DP is still uncertain, there is additional evidence to conclude that the possibility of extracting an object out of a definite nominal expression is due to the lack of DP projection when an object appears. In the following section, I pay attention to some phenomena that support the above claim.

3.4.4.4 Additional Evidence In this section I have developed an analysis for the extraction properties of nominal expressions containing the definite article in Spanish. This analysis is consistent with the general analysis put forward in previous sections to account for the different possibilities of extraction between specific and non specific nominal expressions and with previous independent analyses to account for some properties of Romance definite article that postulated the existence of weak and strong definite articles. In the remaining part of this section I support the current analysis of the definite article with some additional evidence. This additional evidence comes from the interaction of the definite article with the scope of some quantifiers and from some unobserved properties of generic nominal expressions in Spanish.

Interaction with Quantifiers There is additional confirmation of the same phenomenon, that is, of the lack of the DP projection in definite nominal expressions with objects. As noted previously, one of the differences between the specificity effect and definite effects is the different readings available in sentences such as (44): (44) a. Juan vio la foto de todo el mundo Juan saw the picture of everybody b. Juan vio esta foto de todo el mundo Juan saw this picture of everybody As noted above, the sentence in (44a) is ambiguous between a wide or narrow scope reading of the quantified element, while (44b) is not. This is related to whether QR of the determiner todo el mundo (everybody) is possible. Recall that a nominal expression modifier headed by the P de (of) is ambiguous in three ways in Spanish. Thus, the element de todo el mundo (of

114

3 Specificity, Definiteness and the Definite Article in Spanish Nominal Expressions

everybody) in (44a) can be interpreted as the agent, possessor or object of the N foto (picture). Consider now the following sentence: (45)

#Vimos el perro [de todo el mundo]poss (we)saw the dog of everybody

In (45), the interpretation of de todo el mundo as a possessor is forced by the lexical meaning of the N perro (dog), which does not accept object or agent PPs. In this case, the sentence is pragmatically odd since the only reading we get is we saw the unique dog that everybody owned. Importantly, the reading we saw several dogs (namely, Ana’s dog, Peter’s dog, etc.), where the universal determiner scopes over the definite article is not possible. These scope facts illustrate that a definite nominal expression that contains a possessor projects a DP layer and, therefore, QR of the universal determiner is not possible. On the contrary, a definite DP containing an object allows QR of the universal determiner, as definite nominal expressions with objects may appear without the DP layer.

A-Marking and Definite Article As seen before, the presence of the P a (to) forces the specific reading in Spanish indefinite nominal expressions. Although the presence of the P a (to) is obligatory with definite nominal expressions, there are a few exceptions to this rule: nominal expressions referring to [- human] entities do not require this marker and nominal expressions with generic readings can appear in general without a (to). We have already discussed the generic readings of definite nominal expressions above (Section 3.4.4.2) and shown that they allow their PPs to be extracted out of the generic nominal expression. Let us consider now the examples below, where the presence of the P a (to) is illustrating that the N mascotas (pets) can be considered [+ human], as in (46b). (46) a. Vieron las mascotas de Pepe poss (they)saw the pets of Pepe b. Vieron a las mascotas de Pepe poss (they)saw to the pets of Pepe That is, Spanish speakers can personify Pepe’s pets and make use of a to mark the specificity and the [+ human] character of the expression. For our purposes, the crucial fact is that, for those speakers who accept both sentences in (46),21 the possibilities of extraction out of these nominal expressions are higher in the version without the P a (to), as illustrated in (47). (47)

a. ¿De quie´n vieron las mascotas? Of whom (they)saw the pets b. ?? ¿De quien vieron a las mascotas? Of whom (they)saw to the pets

3.5 Summary and Further Research

115

Obviously the differences are very subtle and speakers do not always agree with the judgments. To the extent that there is a difference between the two possibilities, this type of example would also reinforce the correlation between presence of specificity feature and ban on extraction.

3.4.5 Conclusions In a nutshell, several phenomena point to the conclusion regarding the definite effect (that is, the impossibility of extracting some elements out of nominal expressions with definite articles) we reach in this chapter. The main cause of the definite effect with agents and possessors is that the definite article projects a DP in those cases because it is [+ specific]. The lack of definite effects with objects is due to the fact that the definite article used in these cases is not the strong definite article and it does not project a DP. The similarities regarding the extraction possibilities between specific nominal expressions and nominal expressions with definite articles in Spanish lie in the fact that, in the relevant cases of both types of nominal expressions, a DP projection is needed due to the feature composition of the strong definite articles and specific determiners. Hence, both types of phenomena receive the same explanation.

3.5 Summary and Further Research 3.5.1 Main Claims of the Chapter This chapter has applied the basic approach to the structure of Spanish nominal expressions proposed in the previous chapter to some data with specific and definite nominal expressions that had not received satisfactory explanation so far. To do so, I have analyzed the possibilities for extraction from specific and definite nominal expressions and I have argued that the differences between non specific and specific nominal expressions with respect to extraction are the result of the presence or absence of the DP projection in the structure. Therefore, the treatment of determiners given in this book is as follows: weak Ds are generated in Agr, strong Ds are generated in D and definite articles vary their base-generated position depending on their strength, namely, weak definite article are non specific and they are generated in Agr, while strong definite articles are specific and they are generated in D. Hence, the analysis given to the blocking effects with definite-headed nominal expressions makes them parallel to regular specificity effects, which can be explained without any additional assumption in the basic analysis. Furthermore, this type of analysis is consistent with previous proposals (cf. for instance Enc (1991), Diesing (1992), Leonetti

116

3 Specificity, Definiteness and the Definite Article in Spanish Nominal Expressions

(2004)), among many others) that have postulated a link between specificity and discursive component in the nominal expression.

3.5.2 Further Research So far, I have shown strong empirical evidence to conclude that the possibility of extracting an object out of a definite nominal expression is due to the lack of DP projection when an object appears. The main difficulty for this analysis is to explain the impossibility of extracting agents or possessors out of nominal expressions with definite articles and the possibility of extraction of objects in the same contexts. Although much research remains to be done on this issue, let me share here some thoughts on the issue. Two paths seem to me worth exploring: first, the relation between possession/agentivity and partitive nominal expressions and, second, the structural difference between specifier and complement. It has been noted that partitive constructions must be specific (Bruge´ (2000), for instance). This is because they are always linked to an already determined set of entities, which constitutes the domain of quantification for the D. For instance, in Spanish, they always appear preceded by the preposition a (to), (48) (48)

a. Buscamos a muchas de esas secretarias (we)look-for to many of those secretaries b. *Buscamos muchas de esas secretarias (we)look-for many of those secretaries

Hence, examples such as (48) show that, although partitive constructions are morphologically indefinite nominal expressions, semantically they are always referring to a specific set of entities, a part of the domain of quantification denoted in the expression. A possible hypothesis to explore is to what extent the semantics involved in possessors and agents, which seems to be close to a partwhole relationship, forces the specificity of the nominal expression headed by the definite article and, as a result, the projection of DP and the blocking effects of the construction. As for the second possibility to explore, I would like to take into consideration the different configuration between noun-object/theme and noun-possessor/agent. For instance, in the first configuration (head-complement), a recategorization process can take place, while this is not possible with arguments placed in specifier positions, such as possessors and agents. A possible hypothesis to explore is whether these structural differences can bring some repercussions with respect to the specificity of the nominal expression. At this point, however, I can only offer some speculations. The investigation of these hypotheses, along with deeper investigation of the properties of some nominal expressions, such as the ones involved in partitive constructions, is left for further research.

Notes

117

Notes 1

2

3

4

5 6

I use the terms nominal expression and nominal construction as general terms, while I reserve the term determiner phrase (DP) for those nominal expressions that project that particular functional category. There are markers and constructions not contained within the nominal expression that have been repeatedly related with the expression of specificity and definiteness in Spanish (cf. Leonetti (1999), von Heusinger and Kaiser (2003), among others). As illustrated in Section 3.2, specificity can be expressed through the verbal mood in Spanish or through the presence of the preposition a (to) preceding a nominal expression. See Leonetti (1999: 865–870) for a discussion of different contexts linked to specific interpretations of nominal expressions. For an exhaustive account of definite nominal expressions and their semantics, the reader can consult Lyons (1999). See Leonetti (2004) for a more detailed discussion on the possibility of considering a (to) as a specificity marker in Spanish. This is a very simplified version of what the facts are. See Leonetti (2004) on this issue. In standard Spanish clitic doubling is only possible with pronominal objects, as in (i) but not with full nominal expressions, such as the examples in the text, which are ungrammatical in that variety of Spanish. (i) La veo a ella Cl-Acc see-(1sg her ‘I see her’

Note however that pronouns are definite expressions, higher in the definiteness scale than nominal expressions headed by the definite article and the relevant category in standard Spanish involves definiteness too. 7 The reader can consult Grimshaw (1990) on the fundamental differences and classification of nouns according to their interpretation and on the syntactic repercussions of this division. In addition, Alexiadou et al. (2007: 495–546) summarizes Grimshaw’s account of deverbal nouns and describes some more recent approaches to event nominals, such as Alexiadou’s (2001a) and Borer’s (2005) purely syntactic analyses of nominalizations. 8 As discussed in the previous chapter, another proposal that links the semantic properties of the nominal expression with its extraction possibilities appears in Kolliakou (1999). 9 Several questions arise at this point regarding the universality of DP. I leave them aside in this book. Longobardi (1994), Chierchia (1998) and others have developed theories to account for the different interpretations of nominal expressions with or without determiners in different types of languages. In addition to the previous references, the reader can find an overview of the recent research on the existence (or not) of DP in different types of languages and the main challenges for that kind of proposal in Alexiadou et al. (2007: 159–225). 10 Additionally, note that extraction of the group of PP elements considered traditionally as adjuncts that show a completely parallel behavior to argumental PPs (termed PPAs in Chapter 2) is also affected by the specificity effect, which makes this group parallel to argumental PPs: (i) a. *¿De que´ paı´ s conoces [estas ciudades tadv]? of which country (you)know these cities tadv b. *¿Sobre quie´n viste [estos reportajes tadv]? about whom (you) saw these documentaries tadv c. *¿Sin que´ viste [estos libros tadv]? without what (you)saw these books tadv

118 11

3 Specificity, Definiteness and the Definite Article in Spanish Nominal Expressions

Furthermore, as the examples in (i) illustrate, the possibilities of extraction of some other PP elements are not affected by the presence of the definite article in Spanish. That is, the group of PP elements considered traditionally as adjuncts that show a completely parallel behavior to argumental PPs (termed PPAs in Chapter 2) can be extracted out of non specific nominal expressions and definite nominal expressions; while the group of PP elements that do not behave as argumental PPs cannot be extracted out of those constructions, cf. (id). The examples below illustrate this observation: (i) a. ¿De que´ paı´ s conoces [las ciudades tadv]? of which country (you) know the cities tadv b. ¿Sobre quie´n viste el reportaje tadv? about whom (you)saw the documentary c. ¿Sin que´ viste el libro tadv? without what (you)saw the book d. *¿Segu´n quie´n escuchamos el evangelio tadv? by who (we)listened the Gospel

12

13

14

15

16

17

Crucially, the grammaticality of (ia–c) gives us additional evidence for the assimilation of PPAs and argumental PPs. More concretely, (ia–c) shows that PPAs behave as objects, since, as shown, only objects can be extracted out of Spanish nominal expressions headed by the definite article. Some informants do not consider the examples with extraction out of an indefinite nominal expression preceded by the P a (to) fully ungrammatical but just deviant. However, all of the speakers consulted have a sharp distinction between the examples with and without a (to). For our purposes, the existence of this contrast is enough evidence. Note that all the discussion in this chapter refers exclusively to those nominal expressions that contain only one argumental PP, since they illustrate the simplest case. For a discussion on when it is possible to extract PPs out of nominal expressions containing more than one PP, the reader should consult the detailed explanation in Chapter 2. Note that the analysis developed in this chapter is also consistent with proposals that claim that weak Ds move (cf. Herburger (2000), Roehrs (2002), among others, on D movement). Nothing in my analysis depends on this assumption. Longobardi’s (1994) proposal argues that the presence of D is needed to have a nominal argument. The argumental value of D is a very debated issue in the current research, for a detailed overview of the topic see Alexiadou et al. (2007: Chapter 2) Note that another possibility would be to place the specificity feature in a topic projection. However, the cross-linguistic research on nominal topics has shown that there is a wide variation on the position in which languages place this functional category (cf. DimitrovaVulchanova & Giusti (1998: 347–356)’s discussion on the different positions of topic in Albanian and Bulgarian, for instance). This fact and the lack of research on nominal topics in Spanish nominal expressions lead me to place the specificity feature as part of the D projection. The exact position that strong Ds such as demonstratives occupy in the structure depends on the analysis of demonstratives we adopt. The most traditional analyses, such as Abney’s (1987) proposal, assume that demonstratives are heads of D. Analyses such as Guasti (1997), Panagiotidis (2000), Panagiotidis (2000), Bruge´ (2002)) and Grohmann & Panagiotidis (2005), among others, propose two positions for demonstratives. According to this approach, Spanish demonstratives could be base-generated in the specifier of a lower category and they optionally rise to the spec of DP to check off its features. Note that both types of analyses are fully compatible with the proposal in the chapter since the presence of a demonstrative will trigger the presence of DP under both approaches, which will interfere with the extraction facts discussed in this chapter.

Notes 18

119

Note that it is not always the case that the P a (to) can be considered a specificity marker. For instance, as Leonetti (2004) notes, the P a (to) must appear with some verbs, independently of the semantic properties of the nominal expression used, (i). (i) Juan castigaba *(a) {algunos estudiantes/Marı´ a} John punished to {some students/Marı´ a

19

20

21

Although I link the presence of the weak definite article with the absence of specificity in the expression and I assume that the weak definite article in Spanish is [+ definite, specific] in the relevant extraction examples in the chapter, see the discussion in Alexiadou et al. ( 2007: 185-188) with respect to the interpretation of this weak (or expletive) definite article in some other contexts such as generic nominal expressions or preceding proper nouns, where the definite meaning seems redundant or inexistent. Although in the text, I restrict the discussion to cases of atypical properties of definite articles to Romance languages, there are some other languages that use their definite article to mark unusual interpretations. One such language is Greek, which displays the phenomenon known as ‘Determiner Spreading’ and uses the presence of the definite article in adjectives of the nominal expressions to mark the different semantic properties of adjectives. For a discussion of determiner spreading see Alexiadou et al. (2007: 364–377). See also Alexiadou and Wilder (1998) for a particular approach to this phenomenon. This is only possible with this type of example. Examples with nominal expressions [-human] headed by the definite article cannot have the P a, as (i) shows. ´ (i) Vimos (*a) la pelı´ cula de Almodovar ´ (we)saw to the movie of Almodovar

Chapter 4

Adjective Placement

4.1 Introduction One of the most studied topics in Romance linguistics is the adjective within nominal expressions. The syntactic and semantic properties of the adjectives have been studied plentifully, along with their position within the nominal construction, their internal structure and their order relative to the noun and to other adjectives and constituents in the nominal expressions. Given the number of proposals on the topic, this chapter must have a narrow scope from the onset and does not aim to review all of the properties and proposals discussed in the relevant literature1 but to analyze the main properties of Spanish adjectives so that they can receive a uniform analysis, which is consistent to (and can be incorporated into) the basic structure of Spanish nominal expressions proposed in Chapter 2. The investigation of the syntactic and semantic properties of adjectives and their relation with other elements within Spanish nominal expressions will lead me to deal with some important issues in Romance linguistics. Although we will return to these issues as the chapter unfolds, let me briefly summarize here the main questions explored in this chapter. One of the most salient properties of Romance languages it that they allow their adjectives to precede or to follow the noun in a nominal expression. The different position (pre or postnominal) corresponds to different semantic interpretations of the nominal expression. This property is illustrated in (1) with examples from Spanish. (1) a. Una cierta noticia A certain news b. Una noticia cierta A news true The examples in (1) show that the same adjective cierta (certain/sure) can appear preceding or following the noun and the remarkable differences of meaning in the nominal expressions. These differences in linear word order have caused the development of two lines of research on the topic. First, there is M.E. Ticio, Locality Domains in the Spanish Determiner Phrase, Studies in Natural Language and Linguistic Theory 79, DOI 10.1007/978-90-481-3398-7_4, Ó Springer ScienceþBusiness Media B.V. 2010

121

122

4 Adjective Placement

a line of thought (cf. Cinque (1995), Scott (2002), or Laenzlinger (2005), among many others, just to mention some recent proposals) that postulates that prenominal and postnominal adjectives have a common base-generated position and that the different orders are derived via syntactic operations. Among the syntactic operations proposed to derive the different orders are N movement, predicate fronting and NP movement. According to these approaches, the derivation for the cases in (1) could be schematized as in (2): (2)

a. [Una noticia [cierta]] b. [Una [cierta]i noticia [ti]] c. [Una [cierta] noticia] d. [Una [noticia]i cierta [ti]] e. [Una [cierta] [noticia]NP] f. [Una [noticia]NPi cierta [ti]]

predicate fronting of the adjective N movement NP movement

(2a) and (2b) show the basic analysis for the approach that assumes that the adjective is merged in a postnominal position and is fronted to the prenominal position. (2c) and (2d) show that, following the N movement approach, the first structure generated contains the prenominal adjective and that the raising of the N derives the postnominal order. Finally, (2e-f) gives a rough outline of a variation on the N movement approach, where the entire NP is moved to an upper position to obtain the postnominal adjective order. A second line of research (cf. Valois 1991, Lamarche 1991, among many others, just to mention a few proposals based mainly on French data) has put forward the hypothesis that the different linear orders obey to different generation possibilities and that the two linear orders are therefore not related derivationally. Under this type of approach, prenominal and postnominal adjectives are merged in different positions due to their different syntactic and semantic properties. The possible generation sites for each adjective type vary from proposal to proposal (see Section 4.4 of this chapter for a discussion on the problems of these proposals to give explanation for the properties of Spanish adjectives and Alexiadou et al. (2007: 326–392) for a detailed survey of the different approaches on adjectives). Both lines of research have dealt with the cross-linguistic differences regarding the placement of adjectives in the nominal expression. The N movement approach has been embraced as the most empirically supported proposal from a cross-linguistic point of view. That is, the fact that Romance languages differ from Germanic languages in the relative order between nouns and adjectives in their respective nominal expressions seems to be more amenable to receive an explanation under the terms of the N movement analysis. This explanation is represented in (3) (3) a. D—Adj—N b. D––Ni––Adj—ei

Germanic Languages ! No N movement language Romance Languages ! N movement language

Hence, according to this approach, languages can be parametrized with respect to the N movement. As will be shown with more detail through the

4.1 Introduction

123

chapter (see Section 4.4), the postulation of N movement is not enough to derive neither the language particular properties of adjectives in Spanish nor the crosslinguistic variation. Lamarche (1991) noted originally that sentences such as (4) in French pose problems to the N movement approach. (4) a. une e´norme maison magnifique A big house beautiful ‘A beautiful big house’ b. [une e´norme [maison]i magnifique [ti]] Under an N movement approach, the relative order of adjectives in (4) is unexpected. This is because under that approach the prenominal French adjectives would be derived from the underlying English adjectival order by raising the N maison (house), as schematized in (4b). However, as the translation shows the English order does not match the French order. The contribution of this chapter to the previous discussion is mainly based on the examination of the Spanish facts and on how they interact with the overall properties of Spanish nominal expressions. The results of this investigation strongly suggest that the different types of adjectives enter in different positions in the derivation. More concretely, I will be arguing in this chapter that some adjectives are merged as adjuncts while some other adjectives are merged as specifiers, depending on their semantic properties. Thus, the different syntactic and semantic properties of adjectives are derived from this basic generation difference. Furthermore, the analysis shows that there is not a need to resort to the application of syntactic mechanisms, such as the N or NP movement, to derive the properties of adjectives in this language. As for the cross-linguistic debate, this chapter shows that my analysis can be expanded to some other languages with some surprising results. Taking the predictions of the analysis regarding the relative height of the different adjectives seriously, I show that the general pattern holds in different types of languages and that the differences in linear order can be attributed to a difference in the phonological realization of adjuncts, which can be spelled out preceding or following the structure they are adjoined. This analysis then argues against N movement as an explanation for the cross-linguistic differences found and points to a new direction to explore. This chapter is organized as follows. Section 4.2 is a general description of the properties of Spanish adjectives. Here I examine the main syntactic properties of Spanish adjectives and classify them according to their main semantic properties. The conclusions of this section will provide the major descriptive generalizations explained in the rest of the chapter. Section 4.3 develops my analysis of adjectives in Spanish. The hypothesis I explore will be based on the existence of two possible syntactic structures for nominal phrases in Spanish. The crucial difference between the two proposed structures is the position in which adjectives enter the derivation, which depends on their semantic properties. One group of adjectives will be placed as the specifier of NP, while the second group will appear as adjuncts to NP. More

124

4 Adjective Placement

concretely, I will be assuming that prenominal adjectives are merged as the specifier of NP, while postnominal adjectives are adjoined to NP. Moreover, relational adjectives, which behave as PP arguments, will receive a different treatment. This division of adjectives will enable us to account for different syntactic properties of nominal phrases with different types of adjectives in Spanish, attending to the semantic properties of adjectives. In Section 4.4, I briefly review the predictions of major competing analyses proposed in the literature with respect to the properties of Spanish adjectives (Section 4.4.1) and discuss the need of N movement in our grammar as a mechanism to derive the cross-linguistic variation in the placement of adjectives in nominal expressions, (Section 4.4.2). The conclusion of this section is that my analysis is superior to the competing analyses since it has greater empirical coverage and does not require additional operations to be assumed in our theory, such as the N or NP movement. Finally, Section 4.5 summarizes the findings of the chapter and points to a direction for further research. More concretely, I provide an outline of an account for the cross-linguistic differences in adjective placement in nominal expressions, which is mainly based on the differences in the phonological realization of adjuncts.

4.2 Description This section introduces the basic properties of Spanish adjectives. Although occasional references will be made to the properties of adjectives in some other languages, the description is heavily focused on Spanish for ease of exposition. The reader can consult some other references to obtain a detailed description of the properties of adjectives in a particular language (see, for instance, Cinque (1994) for Italian, Valois (1991) for French and Bernstein (1993) for a comparison of adjectives in Romance languages) or Chapter 1 of part III in Alexiadou et al. (2007: 284-394) for a presentation of cross-linguistic properties of adjectives in Romance, Germanic and some other languages, such as Greek, Gungbe and Fongbe. This descriptive section is divided into three subsections in order to describe the main properties of Spanish adjectives. Section 4.2.1 classifies Spanish adjectives with respect to their linear order possibilities. Section 4.2.2 examines the relation between the meaning and linear order of Spanish adjectives. More concretely, I study differences in the meanings of nominal expressions attributed to the presence of prenominal or postnominal adjectives. Finally, Section 4.2.3 provides a brief summary that states the main points of the description.

4.2.1 Linear Order Romance adjectives can appear prenominally or postnominally. This is one of the most important characteristics of adjectives in Romance languages and the

4.2 Description

125

crucial property that triggered much research on nominal expressions in Romance vs. Germanic languages. For examples and discussion of this particular property consult Alexiadou et al. (2007, 286–289), where the distributional patterns of adjectives in Romance and Germanic languages are illustrated. Attending to their linear position within the nominal expression, the situation found in Spanish can be described by assuming the existence of three differentiated groups of adjectives. There is a group of adjectives (group I) that can only appear in the postnominal position, as (5) illustrates: (5)

a. el libro corto the book short ‘the short book’ b. una zona industrial a zone industrial ‘an industrial zone’

vs.

a’. *el corto libro the short book b. *una industrial zona an industrial zone

There is a group of adjectives (group II) that can only appear in the prenominal position, as in (6): (6)

a. un mero accidente a mere accident ‘a mere accident’ b. el presunto asesino the alleged killer ‘the alleged killer’

a’. *un accidente mero an accident mere b’. *el asesino presunto the killer alleged

Finally, there is a group of adjectives (group III) that can appear pre or postnominally. The different positions of these adjectives result in different meanings of the nominal expressions that contain them. In some cases, subgroup 1, exemplified in (7), the change of meaning is predictable (cf. Section 4.2.2 for the different meanings of nominal expressions containing adjectives); in other cases, subgroup 2, illustrated in (8), the change of meaning is unpredictable. (7)

(8)

a. las olorosas flores the fragrant flowers ‘the fragrant flowers’ b. el oso peludo the bear furry ‘the furry bear’ a. una cierta cosa a sure thing ‘a particular thing’

a’. las flores olorosas the flowers fragrant b’. el peludo oso the furry bear b. una cosa cierta a thing sure ‘a sure thing’

126

4 Adjective Placement

b. un viejo hombre an old man ‘an antiquated man’

b’. un hombre viejo a man old ‘an old man’

Only a few adjectives compose subgroup 2. They are given below: (9) Nuevo Pobre Bueno Gran(de)

Prenominal different unfortunate considerable important

Postnominal unused without money good big (adapted from Luja´n 1980, 220)

Bernstein (1993), Sa´nchez (1996), Martı´ n (1995) and many others, note that the different surface position (prenominal or postnominal) in which adjectives appear determines the properties of adjectives in Spanish. More precisely, prenominal adjectives do not accept complements, as in (10a) and do not appear in elliptical nominal constructions, as in (11a). The opposite behavior is shown by postnominal adjectives, as examples (10b–11b) illustrate: (10) a. *el the b. el the

orgulloso de su hija padre proud of his daughter father padre orgulloso de su hija father proud of his daughter

(11) a. *Ayer vi a la verdadera terrorista y a la supuesta [e] yesterday (I)saw to the true terrorist and to the alleged [e] ‘Yesterday, I saw the true terrorist and the alleged one’ b. Ayer vi aquella casa azul y esta [e] verde yesterday (I)saw that house blue and this [e] green ‘Yesterday, I saw that blue house and this green one’ Another property that is often grouped with the ones discussed above (cf. Bernstein (1993, 2001), among others) concerns the relation between linear position and the ability to appear in predicative contexts. The generalization is that prenominal adjectives do not appear in predicative contexts, while postnominal adjectives do. Typical examples are given in (12): (12) a. *El accidente es mero the accident is mere b. El libro es corto the book is short Note that, although it seems to be true that prenominal adjectives in Spanish cannot appear in predicative constructions, it is not the case that all postnominal adjectives can appear in predicative constructions. The examples in (13)2 show this.

4.2 Description

(13) a. El diccionario me´dico the dictionary medical ‘ The medical dictionary’ b. El diccionario es me´dico the dictionary is medical

127

a’. *El me´dico diccionario the medical dictionary

Therefore, the above-mentioned property will not be taken into consideration in this book, since the ability of a particular adjective to occur in a predicative context seems to be related to the lexical semantics of the adjective, not only to its structural position.

4.2.2 The Relation Between Meaning and Linear Position In the previous section, I exemplified a group of adjectives (group III) that can appear pre or postnominally in Spanish. In this section, I focus on how the different positions of these adjectives result in different meanings of the nominal expressions that contain them. To do so, I provide here a semantic classification of basic semantic types of adjectives in Spanish, which follows Demonte’s (1999a) classification. After that, I turn to the relation between meaning and linear position of adjectives and discuss several syntactic properties of each group of adjectives. Although it is not the aim of this subsection to provide a detailed semantic classification of adjectives,3 I introduce a very intuitive semantic characterization of adjectives in Spanish. There are many possible semantic-based classifications of adjectives. I follow here Demonte’s (1999a) description of Spanish adjectives, although some of her insights will be left out of the discussion for the sake of clarity. Demonte (1999a) establishes a division of adjectives into three major groups: relational adjectives, qualitative adjectives and adverbial adjectives. I give an example of each group in (14): (14) a. Una revista americana a journal American b. Una persona rubia a person blonde c. Un futuro presidente a future president

(relational) (qualitative) (adverbial)

According to Demonte (1999a), Relational adjectives (R-As) denote a relation between two entities; namely, the entity denoted by the N they modify and the entity denoted by the N they derive from. In (14a) the meaning of the nominal expression could be paraphrased as a journal of things related to America but not as a journal of Americans. As for the Qualitative adjectives (Q-As), Demonte (1999a) argues that they may refer to a property assigned to the elements denoted by the N. That is,

128

4 Adjective Placement

the nominal expression in (14b) can be paraphrased as there is a person and that person has the property blonde. Finally, Demonte’s (1999a) Adverbial adjectives (A-As) do not assign properties to the elements denoted by the N. They modify either the intension of the N or the event denoted by the N. The adjective in (14c) modifies the intension of the N presidente (president) but not its extension. That is, this nominal expression does not refer to any entity in the actual world. According to Demonte (1999a), Q-As are subdivided into two groups: intensional Q-As and extensional Q-As. An illustrative example is given in (15): (15) a. Los geniales pintores the great painters b. Los pintores geniales the painters great Demonte (1999a) argues that prenominal Q-As correspond to Kamp’s (1975) intensional adjectives. The reason for this identity is that prenominal Q-As do not affect the extension of a modified N. In other words, if the extension of the N pintores (painters) is the set composed of Picasso, Miro´ and Dalı´ , the group denoted by the nominal expression in (15a) is the same exact set. Therefore, the extension of the N is not affected by the presence of the prenominal Q-A geniales (great). In contrast, postnominal Q-As do affect the extension of the modified N; therefore, they are extensional adjectives in Kamp’s (1975) nomenclature. That is, if the extension of the N pintores (painters) is the set composed of Picasso, Miro´ and Dalı´ , the group denoted by the nominal expression in (15b) can be composed only by a subset of this set, the one containing only the individuals that are in the denotation of the N and the denotation of the adjective. Thus, the postnominal Q-A geniales (great) in (15b) contributes to the denotation of the N; as a result, the extension of the N changes. After introducing Demonte’s (1999a) basic semantic types of adjectives in Spanish, I turn to the relation between meaning and linear position of adjectives. I start by discussing several syntactic properties of each group of adjectives mentioned above. R-As derive from Ns. R-As do not accept degree words or complements and only appear postnominally in Spanish nominal expressions, keeping strict adjacency with the N that they modify. (16) illustrates the behavior of R-As: (16) a. El sabor (*muy) mineral4 the taste (very) mineral ´ b. *Las fuerzas productivas para la nacion the forces productive for the nation ´ pu´blico c. *El plan de educacion the plan of education public As R-As only appear postnominally in Spanish nominal expressions and they have only one possible position in Spanish nominal expressions; the

4.2 Description

129

question on how their meaning affects the linear order is not relevant. However, there is a remarkable property of these adjectives: they tend to appear with deverbal Ns. Picallo (1991) studies the contexts in which these R-As occur in Catalan and concludes, following Giorgi and Longobardi (1991), that they typically appear in positions where they can receive a theta-role from the N.5 This explains the contrast below: (17) a. *La captura americana de Juan the capture American of Juan ´ americana de Irak b. La invasion the invasion American of Iraq The impossibility of having an R-A in (17a) is explained by the fact that the N captura (capture) is a passive N6; therefore, the R-A americana (American) cannot receive an external theta-role in this construction. The grammaticality of (17b) is due to the fact that the R-A can receive its external theta-role from the ´ (invasion). active N invasion The behavior of Qualitative As (Q-As) is more complicated. According to their linear order, they correspond to subgroup 1 of group III from Section 4.2.1. That is, Q-As are adjectives that can appear pre or postnominally. Each position results in a different meaning of the nominal expression that contains them. Crucially, the change of meaning is predictable: they give rise to the contrast restrictive/non-restrictive in definite nominal expressions and to the contrast specific/non-specific in indefinite nominal expressions.7 Let us pay attention first to the contrasts in meaning found in definite nominal expressions. I will start with the following example: (18) a. las olorosas flores (non-restrictive/intensional) the fragrant flowers b. las flores olorosas (restrictive/extensional) the flowers fragrant When an adjective of the subgroup A (i.e., a Q-A) appears prenominally, it has a non-restrictive interpretation; in this case, the Q-A is not delimiting a group of the objects denoted by the N but predicating of all of them. In other words, the example (18a) is interpreted with a meaning in which all the flowers that are denoted by the N have the property of being fragrant. On the contrary, when the same Q-A appears postnominally, the Q-A has a restrictive interpretation; in this position, the adjective delimits a subgroup of the objects denoted by the N. That is, the nominal expressions in (18b) refer to a particular subgroup of flowers: the fragrant ones. Another peculiarity of Q-As with definite nominal expressions is that prenominal Q-As can appear with proper Ns, while postnominal Q-As are not allowed in this context: (19) a. El genial Mozart the great Mozart

130

4 Adjective Placement

b. *El Mozart genial8 the Mozart great According to Zamparelli (2000) and Demonte (1999a), this is an indication that prenominal Q-As do not affect the extension of the N. Assuming with Longobardi (1994), among others, that the definite article that appears with proper nouns is semantically vacuous and that proper nouns are rigid designators9, the Q-A cannot affect the extension of the N. Therefore, Prenominal QAs emphasize an intrinsic quality of the entity denoted by the nominal expression; they are, in this sense, similar to appositive modifiers.10 As for indefinite nominal expressions, the presence of Q-As forces the contrast between specific/non-specific readings of the indefinite nominal expression: (20) a. Las cinco muchachas conocieron a un actor famoso [spec/non-spec] the five girls met to an actor famous b. Las cinco muchachas conocieron a un famoso actor [spec] the five girls met to a famous actor (20a) has two readings: there is a particular famous actor that each girl met or each girl met a different famous actor. However, (20b) has only one interpretation: there is a particular famous actor that each girl met. Bosque (2001) illustrates some of the properties that distinguish indefinite nominal expressions, depending on the linear position of the Q-A.11 For instance, indefinite quantifiers, such as cualquiera (any) and ninguno (no), which do not allow a specific reading due to their lexical meaning, are not compatible with prenominal adjectives: (21) a. */??Compra cualquier famosa novela (you)buy any famous novel b. Compra cualquier novela famosa (you)buy any novel famous c. *No he visto ningu´n interesante artı´ culo Neg (I)have seen no interesting paper d. No he visto ningu´n artı´ culo interesante Neg (I)have seen no paper interesting Another relevant contrast is related to the use of the so-called personal a (to). As discussed extensively in Chapter 3, this P, which can mark specific indefinites in Spanish (cf.Chapter 3 and the references cited there for a summary of personal/specific a (to) contexts), cannot be absent if the indefinite personal nominal contains a preposed Q-A: (22) a. Busco un actor famoso [non spec] (I)look-for an actor famous b. Busco a un actor famoso [spec] (I)look-for to an actor famous

4.2 Description

131

c. Busco a un famoso actor (I)look-for to a famous actor d. *Busco un famoso actor (I)look-for a famous actor Note that the difference between (22a) and (22b) lies exclusively in the specificity properties of the nominal expression. If the nominal expression is preceded by the P a (to), it will be specific; otherwise, it will be non specific. The presence of a prenominal Q-A seems to force the specific reading, as the ungrammaticality of (22d) illustrates. There are other syntactic contexts that force the non specific reading of indefinite nominal expressions. For instance, directive contexts are non specific,12 and indefinite nominal expressions with prenominal Q-As are not allowed: (23) a. Escribe (you)write b. *Escribe (you)write

una novela interesante. Te hara´s famoso a novel interesting. CL (you)will-be famous una interesante novela. Te hara´s famoso an interesting novel. CL (you)will-be famous

A similar situation is found in rhetorical questions. According to Bosque (2001), rhetorical questions license, in addition to the specific reading, the nonspecific reading of indefinites; in these contexts the presence of a prenominal Q-A in the indefinite nominal expression makes the rhetorical reading impossible (# stands for inadequate rhetorical reading): (24) ¿Cua´ndo me When CL /#interesante novela] / interesting novel] ‘When on earth have evenings?’

has regalado tu´ una [novela interesante have bought you an [novel interesting para leer por las noches? to read by the evenings you bought me an interesting novel to read in the

Another relevant test to isolate the readings of indefinite nominal expressions is the so-called donkey-sentences. In these contexts, indefinites are typically interpreted as non specific. The presence of a prenominal Q-A causes ungrammaticality: ´ que contrata a un actor famoso gana (25) a. Toda cadena de television Every channel of TV that hires to an actor famous wins dinero con e´l. money with him ‘Every TV channel which hires an actor famous, wins money after him’ ´ que contrata a un famoso actor b. ?? Toda cadena de television Every channel of TV that hires to a famous actor gana dinero con e´l wins money with him ‘Every TV channel which hires a famous actor, wins money after him’13

132

4 Adjective Placement

Finally, let me mention some syntactic properties of adverbial adjectives (A-As). A-As appear in prenominal position, as in (26a) and can accept degree modifiers, as in (26b): (26) a. *Las asesinas presuntas (las presuntas asesinas) the killers alleged b. Los extremadamente frecuentes viajes de Luis the extremely frequent trips of Luis

4.2.3 Conclusions The previous section described the major properties of adjectives in Spanish. The section focused on two phenomena: the linear order in which adjectives appear in Spanish and the relation between the different meanings of adjectives and the linear order they occupy in the nominal expression. I summarize the findings in the following table: (27) Relational-As Postnominal ‘argumental’

Qualitative As

PreN intensional (non-restrict) (spec)

Adverbial As

PostN extensional (restrictive) (non-spec)

PreN intensional/eventive

A generalization emerges from the description: postnominal adjectives are characterized by affecting the extension of the N, while prenominal adjectives do not modify the extension of the N. To illustrate the descriptive generalization above, let us consider the examples given in (28): (28) a. El agua mineral the water mineral (‘The mineral water’) b. La chica alta the girl tall (‘The tall girl’) b’. La alta chica the tall girl (‘The tall girl’) c. El futuro presidente the future president (‘The future president’)

(relational) (qualitative) (qualitative) (adverbial)

If we combine via predicate modification14 the adjectives in (28) with the N to which they modify, we only obtain the correct interpretation for (28b). For instance, the combination of the set denoted by agua (water) and the set denoted by mineral (mineral) give us the set composed of elements that are water and mineral. This is not the meaning of (28a). A rough paraphrasis of (28a) would be there is a unique entity x, such that x is water and x contains minerals.

4.2 Description

133

Note that a way out of this problem is to assume that the lexical meaning of R-As, such as mineral (mineral), is actually more complex than the lexical meaning of Q-As, such as alta (tall). Demonte’s (2000) analysis of the lexical meaning of adjectives in Spanish pursues this idea and, assuming the lexical semantics proposed in Pustejovsky (1995), derives the different meanings of the nominal expressions containing the R-A ele´ctrico (electric) in (29a) and (29b) from the different complex property denoted by the R-A in each case. (29) a. Tren ele´ctrico train electric b. Central ele´ctrica plant electric

(electric train) (power plant)

The intuitive meaning of (29a) is train powered by electricity, while the intuitive meaning of (29b) is plant that produces electricity. The difference of meanings is due to the fact that a different relation (i.e., a different property) mediates between the property electricity and the property denoted by each of the two Ns. For our purposes, this complex lexical entry of R-As is the result of their derivational history (cf. Section 3.1.2 on this issue) and allows us to assimilate them semantically to other types of modifiers within nominal expressions, namely, to PP modifiers. Returning to the examples in (28), note that predicate modification gives us the desired result in example (28b). The meaning of the nominal expression in (28b) is there is a unique individual x, such that x is a girl and x is tall. This is precisely the meaning of this nominal expression.15 On the contrary, predicate modification cannot derive the meaning of (24b’). Note that the interpretation of (28b’) obtained via predicate modification is the same as the one proposed for (28b). This is because predicate modification combines the same two properties, independently of the linear surface order in which they occur. The intuitive meaning of (28b’) is there is a unique individual x, such that x is a girl, who is tall. Thus, the Q-A seems to be a function that selects an individual over which the property denoted by the Q-A is applied. Hence, this indicates that the position of the Q-A must be outside of the nominal expression, the individual entity, at the point where the structure receives its meaning.16 Finally, the example in (28c) cannot be the result of combining the property denoted by the adjective futuro (future) and the property denoted by the N presidente (president) by predicate modification. First of all, note that it is hard to conceive a property such as being future. This fact has led to treating A-As (i.e., intensional adjectives) as functions from properties to properties. That is, in (28c) futuro (future) can be interpreted as a function that takes the property of being a president and gives us a different property, namely, the property of being a future president.17

134

4 Adjective Placement

To conclude: postnominal adjectives combine with the N via predicate modification to produce a new extension of the predicate over which the value of the D applies. On the contrary, prenominal adjectives do not denote properties that combine directly with the N via predicate modification but they denote functions and, in some cases, they must be interpreted outside of the scope of the D.

4.3 Towards an Analysis This section develops my analysis of adjectives in Spanish. The main goal of the analysis is to provide an account for the generalization noted in the second section of this chapter. Namely, postnominal adjectives are characterized by affecting the extension of the N (i.e., they are combined via predicate modification), while prenominal adjectives do not modify the extension of the N (i.e., they are functions). Following this generalization, the analysis proposes two different positions for prenominal and postnominal adjectives, which accounts for the different properties that these adjectives exhibit. This section is organized as follows. Section 4.3.1 presents the basic outline of the analysis and relates it to the analysis proposed in Chapter 2 for the structure of nominal expressions. My analysis of adjectives is based on the existence of two possible syntactic positions in which adjectives enter the derivation, depending on their semantic properties. I will argue that one group of adjectives will be placed as the specifier of NP, while the second group will appear as adjuncts to NP. More concretely, I will be assuming that prenominal adjectives are merged as the specifier of NP, while postnominal adjectives are adjoined to NP. Moreover, relational adjectives, which behave as PP arguments, will receive a different treatment. Section 4.3.2 provides evidence for the proposed analysis based on the different syntactic behavior of adjectives regarding NP ellipsis and partial cliticization. The section ends with a summary of the main points of the analysis and the conclusions reached in Section 4.3.3.

4.3.1 Analysis In Chapter 2, I proposed a structure for Spanish nominal expressions based on the parallelism between clauses and nominal expressions. The structure assumes the extension of the division of clause structure into three domains to the Spanish nominal expressions. Furthermore, I severely restricted the length of phrasal movements within nominal expressions. The combination of these assumptions enabled me to account for some syntactic properties of various PP elements within nominal expressions and for the interaction of PPs with the different types of Ds. The structure proposed previously is reproduced below:

4.3 Towards an Analysis

(30)

135

FocP Foc’ Foc

ω-domain DP

D’ D

AgrP φ-domain POSS

Agr’

Agr

nP n’

NP

n NP | N’ N

AGENT θ-domain

ADJUNCTS

OBJ/PPA

The analysis of adjectives in this study will be based on the existence of two major positions for these elements in Spanish nominal expressions. Therefore, I follow the line of research that postulates the existence of different positions for the different types of adjectives; namely, a non-uniform approach. In the next section (cf. (4)) I will consider with some detail some previous proposals that argue for a uniform position for adjectives in Spanish and will show that, apart from being a more costly option, they would not allow us to account for the different interpretations of adjectives in Spanish. Returning to my approach, the crucial difference between the two proposed structures is the position in which adjectives enter the derivation, depending on their semantic properties. One group of adjectives will be placed as the specifier of NP, while the second group of adjectives will appear as adjuncts to NP. More concretely, I will be assuming that prenominal adjectives are merged as the specifier of NP, while postnominal adjectives (except for relational adjectives, which receive a different treatment; see below) are adjoined to NP. This division of adjectives obeys the generalization discussed previously; namely, adjuncts cannot be extracted out of nominal expressions. Therefore, the group of adjectives that affect the extension of the N will be merged as adjuncts of NP, as they are interpreted within the scope of the D at LF; the group of adjectives that do not modify the extension of the N will be merged as the specifier of NP, a position from which some of them can move at LF to receive their

136

4 Adjective Placement

interpretation outside of the scope of the D. (cf. below for some details on how this covert movement proceeds) Note that I assume here a distinction between specifiers and adjuncts. Kayne (1994), among others, has questioned this distinction. In Kayne’s (1994) system, the elements usually considered to be specifiers have the properties of adjuncts. However, there have been a number of attempts to argue for the distinction specifier/adjunct. For instance, Duffield’s (1999) analysis of Maltese Arabic and Irish adjectives provides empirical evidence for maintaining the specifier/ adjunct distinction. More recently, Uriagereka (1999), Chomsky (2001) and Gallego (2007), among others, have developed theories that argue that the distinction between adjunct and argument is relevant enough to introduce a particular operation in charge of assembling adjuncts to the main structure; namely, pair-merge. According to these authors, the properties of adjuncts must be derived from their use of this different operation and cannot be assimilated to the properties of arguments in our theory of grammar. This study follows the line of research that postulates the existence of a distinction between specifiers and adjuncts in the theory of grammar. More concretely, I assume that adjuncts are only base-generated positions (i.e., the result of merge) and that there is only one specifier per maximal projection. Under these assumptions, the proposed structure is then (31)18 19: (31)

Finally, I assume that relational adjectives, which behave as PP arguments within nominal expressions (see below for discussion on the topic), occupy the positions I have determined for the different types of arguments in Chapter 2. Note that this proposal argues for a different structural height of adjectives, being in general postnominal (adjunct) adjectives structurally higher than

4.3 Towards an Analysis

137

pronominal (specifier) adjectives in the syntactic component of our grammar. This claim will be sustained in this chapter by the different behavior of adjectives in NP ellipsis and partial cliticization contexts. In what follows, I justify the different generation sites for adjectives in Spanish. I have divided the rest of this section in two subsections: the first one deals with the general distinction prenominal/postnominal and the second subsection deals with the different behavior of relational adjectives. 4.3.1.1 Prenominal Adjectives vs. Postnominal Adjectives The main property of prenominal adjectives under my analysis is that they cannot be combined by predicate modification with the N to which they modify. This property puts together very different types of adjectives. In this subsection, I discuss why those adjectives form a group. Consider the following Korean data from Kim (1997): (32) a. Ku ketaran the big b. Ketaran ku big the

namwu tree namwu tree

(Korean)

According to Kim (1997), there are two possible positions for Q-As in Korean and Japanese: a prenominal position, (32a) and a predeterminer position, (32b). These two positions correlate with the restrictive and non-restrictive readings of these adjectives. Namely, the example in (32a) is interpreted as a restrictive adjective, while the adjective in example (32b) only has a non-restrictive reading. Kim’s (1997) analysis of these facts leads him to conclude that Korean and Japanese non-restrictive adjectives must move overtly to the pre-DP position due to the presence of a strong NON-focus feature they need to check, while restrictive adjectives must stay in situ, as they do not carry that feature. Furthermore, Kim (1997) generalizes his analysis and proposes the existence of a parameter that divides languages into two types: languages that move nonrestrictive adjectives in the syntactic component of our grammar, such as Korean and Japanese and languages that do this movement in LF, such as English. My analysis of prenominal adjectives takes as its departure point Kim’s (1997) proposals and applies them to Romance languages. The difference between prenominal and postnominal Q-As in Spanish nominal expressions can be explained by assuming that prenominal Q-As in Spanish nominal expressions move at LF to a position higher than the D. Under my approach, prenominal Q-As in Spanish nominal expressions could move to a topic phrase, which will be the equivalent to the non-focus feature in Kim’s (1997) proposal. This would explain for instance the different interpretations of definite nominal expressions depending on the position of the Q-A.

138

4 Adjective Placement

Moreover, this line of approach will allow us to explain the contrast below. (33) a. Compramos varios famosos libros de lingu¨ı´ stica (we)bought several famous books of linguistics b. */??¿De que´ compramos varios famosos libros? of what (we)bought several famous books c. Compramos varios libros famosos de lingu¨ı´ stica (we)bought several books famous of linguistics d. ¿De que´ compramos varios libros famosos? of what (we)bought several books famous The contrasts above show that the presence of prenominal adjectives in constructions where an argument of the nominal expression has been extracted causes degradation in the grammaticality of the construction, while the presence of postnominal adjectives does not affect the grammaticality status of the sentences. Although more research is needed in this particular point and I leave accounting for the contrast in question to future research, the existence of this type of examples provides the current account with additional empirical evidence.20 In addition, if non-restrictive Q-As must move in Spanish, the most plausible position for these types of adjectives should be a specifier position, where they can be accessible to movement. On the contrary, the general properties of postnominal adjectives discussed above, such as the fact that they are interpreted under the scope of the determiner, show that postnominal adjectives cannot move; therefore, their placement as adjuncts is consistent with the observation that NP-adjuncts in general cannot move in DP languages such as Spanish (see Chapter 2 for a detailed discussion on PP adjuncts). Note that a movement operation for adjectives has been previously proposed in generative grammar. In previous analyses (Chomsky (1957) and Chomsky (1965), among many others), prenominal attributive adjectives were derived as the result of leftward movement of adjectives merged to the right of the N; in other words, the proposal was to analyze the prenominal adjectives as fronted postnominal adjectives. These types of analyses were abandoned due to an overgeneration problem (that is, all postnominal adjectives will be amenable to be fronted). Here I adopt from these analyses the possibility of having a movement operation for certain types of adjectives, which is triggered by their semantic properties.21 There is some evidence in favor of this approach. Bosque (2001) studies the different meanings resulting from the position of Q-As within indefinite nominal expressions. As shown above, prenominal Q-As force a specific reading in the indefinite nominal expressions, while postnominal Q-As allow for the non specific reading typical of indefinite nominal expressions. The two readings are repeated below for convenience: (34) a. Las cinco muchachas conocieron a un actor famoso [spec/non-spec] the five girls met to an actor famous a’. Readings: ‘There is a particular famous actor that each girl met’ ‘Each girl met a different famous actor’

4.3 Towards an Analysis

139

b. Las cinco muchachas conocieron a un famoso actor [spec] the five girls met to a famous actor b’. Readings: ‘There is a particular famous actor that each girl met’ *‘Each girl met a different famous actor’ Bosque’s (2001) analysis of the contrast in (34) assumes Heim’s (1982) analysis of indefinite nominal expressions. According to this analysis, non specific readings of indefinite nominal expressions are derived from the fact that they are interpreted as variables unselectively bound by generic operators and quantificational adverbs, or existentially closed by implicit quantifiers. As for the specific readings of indefinite nominal expressions, Bosque (2001) proposes that they are the result of the presence of an intermediate functional projection preventing the binding process of the variables by the operators. This analysis explains, according to Bosque (2001), why indefinite nominal expressions with prenominal Q-As are excluded from contexts where these operators are present: (35) a. Hace unos an˜os, un actor simpa´tico encontraba pronto trabajo ago some years an actor nice found soon job ‘Some years ago, a nice actor used to find a job soon’ b. *Hace unos an˜os, un simpa´tico actor encontraba pronto trabajo ago some years an nice actor found soon job The ungrammaticality of (35b) is due to the indefinite nominal expressions un simpatico actor (a nice actor) not being able to provide the variable needed; this results in vacuous quantification. According to Bosque (2001), the basic derivations for specific and non specific readings of indefinite nominal expressions are the following: (36) a. OPERATOR. . .[FP (famousi)[D un [DegP famousi[DegØ] [AP* actor [AP (famous)i]]]] b. OPERATOR. . .[D un [DegP famousi [Deg Ø] [AP* actor [AP (famous)i]]] The specific reading, represented in (36a), shows the adjective in the specifier position of a functional category (FP). The specific reading is due to the fact that the functional category (FP) intervenes between the operator and the variable to be bound in the NP. Prenominal adjectives must move to FP at LF to check a feature. As for the nature of this feature and the content of this FP, Bosque (2001) speculates that modal features (more specifically, epistemic modal features) could be hosted at FP. The representation in (36b) lacks the functional category and that explains why the operator can bind the variable in the NP and the resulting interpretation is the non specific one. For our purposes, the relevant difference between the prenominal and postnominal Q-As in indefinite nominal expressions is that prenominal Q-As must undergo movement out of the nominal expressions at LF. This point justifies

140

4 Adjective Placement

their generation as specifiers and enables us to group them with prenominal QAs in definite nominal expressions. The third type of adjectives that occur prenominally in Spanish are what I called above A-As. An example of these A-As is given in (37): (37) Vimos a la presunta ladrona (we)saw to the alleged thief Almost all syntactic analyses of adjectives (see next section for discussion of some representative analyses and Alexiadou et al. (2007: 326–394) for a more detailed discussion of the main approaches to the syntax of adjectives) merge these adjectives in a different position from other adjectives. Valois (1991), Bernstein (1993, 2001a), Martı´ n (1995) and Sa´nchez (1996), among others, note that A-As are interpreted in a different way than other adjectives. As I discussed above, these adjectives are functions that take a property and give back another property. For instance, the adjective presunta (alleged) is a function that takes the property of being a thief and gives back the property of being an alleged thief over which the D applies. Therefore, A-As are within the scope of the D, which differentiates these A-As from prenominal Q-As. However, there are some data that show that A-As can move within the nominal expression in which they appear. Consider the example in (34a) and the representations of the two readings of (38a) in (38b) and (38c): (38) a. That is John’s former house b. John’s & a former (house) c. former (John’s & a house)

N-modifying reading POSS-modifying reading (Larson & Cho (2000: (4)))

The reading in (38b) can be paraphrased as there is an object that John now possesses and that was once formerly a house, while the reading in (38c) can be paraphrased as there is a house that John possessed formerly. It appears that the two readings above can only be derived as the result of the different scopes that the A-As establish within nominal expressions. If this is the case, then the inclusion of A-As in the group of adjectives that need to move to receive their interpretation at LF seems to be justified. Therefore, I assume here that A-As are also merged in the specifier of NP and they may in principle move within nominal expressions. To summarize: although the adjectives that occupy the prenominal position in Spanish nominal expressions have different semantic properties, all of them seem to share two main properties: they cannot be combined with the N via predicate modification and they seem to be able to move from their basegenerated position in order to receive their interpretation at LF. The previous discussion leaves us with a uniform group of adjectives that appear in postnominal position in Spanish. This group is composed of adjectives that designate a property and that are interpreted in the scope of the D via predicate modification. As noted above, I assume that postnominal adjectives

4.3 Towards an Analysis

141

are base-merged as adjuncts to NP. This assumption finds evidence in the following two observations. First, as observed by Sa´nchez (1996), among others, the linear order of postnominal adjectives points to the conclusion that they modify NPs and not just Ns: (39) a. Marı´ a es una ladrona de joyas torpe Maria is a thief of jewels clumsy b. *Marı´ a es una ladrona torpe de joyas Maria is a thief clumsy of jewels (40) a. Ayer compre´ una casa de campo grande yesterday (I)bought a house of country big b. ??Ayer compre´ una casa grande de campo yesterday (I)bought a house big of country That is, the ungrammaticality of (39b) and (40b) indicates that the postnominal adjectives modify not only the N but also its complements, since they cannot precede the complement. Second, as noted in Chapter 2, adjuncts are the only elements that cannot be extracted out of nominal expressions. This is explained as a general property of languages that have articles (see discussion on this issue in Chapter 2); therefore, postnominal adjectives are candidates for that position, as they must be interpreted in situ. To conclude: there seems to be some evidence for the division of adjectives established: the different types of prenominal adjectives cannot be combined with the N via predicate modification and they seem to be able to move from their base-generated position in order to receive their interpretation at LF. On the contrary, postnominal adjectives in Spanish denote a property and are interpreted in the scope of the D via predicate modification.

4.3.1.2 Relational Adjectives As noted before, R-As seem to be argumental. The relevant examples are in (41): (41) a. Una novela policı´ aca a novel police (detective story) a.’ Una novela de policı´ as a novel of policemen ´ americana b. La invasion the invasion American ´ b’ La invasion de/por los americanos the invasion of/by the Americans The adjective in (41a) expresses the object of the N, while the adjective in (41b) expresses the agent of the N. As (41a’) and (41b’) illustrate, the examples are equivalent to the relevant nominal expressions with PP arguments.

142

4 Adjective Placement

Furthermore, as mentioned in Chapter 2, the appearance of an R-A denoting an agent blocks extraction of the theme. The examples are repeated below: ´ americana de este tema (42) a. Criticaron la investigacion (they)criticized the investigation American of that topic ´ americana? b.* ¿De que´ tema criticaron la investigacion of what topic (they)criticized the investigation American Therefore, the ungrammaticality of (42b) is due to the position in which the adjective americana (American) appears. By assuming that R-As are merged in their respective argumental position, we can explain that the introduction of agentivity with the adjective americana (American) triggers the presence of the functional category nP in the structure, which blocks the extraction of the object. As discussed in the previous chapter, the atypical surface order in which R-As must appear is the result of the strict adjacency requirement that the lexical properties of R-As impose on the syntactic output of the derivation (cf. Chapter 2 on this issue). In addition, since they are almost argumental in nature, their fixed order is easily derived as the result of the general constraints of Spanish, which does not allow prenominal arguments. An interesting question arises here. If it is correct that R-As behave as nominal expressions and occupy positions that were traditionally reserved to nominal expressions, how do they fulfill the requirements for those positions? In other words, is it the case that R-As receive thematic role and case? The answer to this question lies in the derivational history of R-As. As was noted before, R-As are derived from nominals and a possibility would be to assume (with Baker (2003) and Fa´bregas (2007), for instance) that these type of adjectives are, in fact, nouns that contain in their internal syntactic structure a semantically defective matrix of features that is spelled out as an adjectival affix.

4.3.1.3 Conclusion In this section I have developed an analysis of adjectives that divides them into two main groups: prenominal adjectives and postnominal adjectives. Prenominal adjectives, the group of adjectives that do not modify the extension of the N, have been merged as the specifier of NP, a position from which they can move in LF to receive their interpretation outside of the scope of the D. Postnominal adjectives, the group of adjectives that modifies the extension of the N, are adjuncts and they are interpreted in situ at LF. In addition, my analysis assumes that relational adjectives, which are always postnominal, are merged in their respective argumental position, since they behave as PP arguments. This analysis explains the different linear order possibilities of adjectives in Spanish nominal expressions as the reflex of the different structural positions in which adjectives are merged. Note that there is not a need to make use of explicit movement (for instance, N-movement) operations to derive the different linear

4.3 Towards an Analysis

143

positions of adjectives. In Section 4.4, I will elaborate on this point and show how this is an advantage of my analysis over some other previous analyses.

4.3.2 Evidence from the Syntactic Behavior of the Different Groups of Adjectives Additional evidence for the analysis outlined above also comes from the interaction of adjectives with arguments and adjuncts in nominal expressions. In this section I investigate the interaction of the different types of adjectives with respect to NP ellipsis and partial cliticization.

4.3.2.1 Adjectives and NP Ellipsis One of the most notorious differences between prenominal and postnominal adjectives is that only postnominal adjectives allow NP ellipsis in Spanish. The illustrative examples are in (43): (43) a. *Ayer vi a la verdadera terrorista y a la supuesta [e] yesterday (I)saw to the true terrorist and to the alleged [e] ‘Yesterday, I saw the true terrorist and the alleged one’ b. Ayer vi aquella casa azul y esta [e] verde yesterday (I)saw that house blue and this [e] green ‘Yesterday, I saw that blue house and this green one’ ´ americana c. El an˜o pasado vi la invasion the year last (I)saw the invasion American y este an˜o la [e] francesa and this year the [e] French ‘Last year, I saw the American invasion and this year the French one’ d. Saludamos al anterior presidente y tu´ saludaste (we)greeted to-the former president and you greeted al presente presidente to-the present president ‘We greeted the former president and you greeted the present president’ d.* Saludamos al anterior presidente y tu saludaste al presente [e] (we)greeted to-the former president and you greeted to-the present e The analysis above provides a simple explanation for the contrasts in (43). Assuming that NP ellipsis targets NPs,22 only elements that are base merged outside of the NP node can be left stranded after ellipsis has applied. Thus, as prenominal adjectives are specifiers of NP, they are contained in the ellipsis site and must undergo NP ellipsis. On the contrary, postnominal adjectives are not affected by NP ellipsis. Recall that postnominal Q-As are merged as adjuncts to NP and relational adjectives are merged in the spec of nP. None of these positions is contained in

144

4 Adjective Placement

the ellipsis site.23 Consequently, postnominal adjectives are predicted to be left stranded under our analysis. Graphically, the analysis can be represented as in (44), where the circle represents the element targeted by NP ellipsis: (44)

FocP Foc’ ω-domain Foc

DP D’

D

AgrP Agr’

Agr

ϕ-domain

POSS

nP n AGENT/R-As

n

θ-domain

NP NP

ADJUNCTS/Postnominal As

Prenominal As N’ N OBJ/PPB

Therefore, my analysis predicts the behavior of the different types of adjectives with respect to NP ellipsis in Spanish.24

4.3.2.2 Adjectives and Partial Cliticization Another piece of evidence for the analysis presented above comes from a set of previously unnoticed contrasts regarding partial cliticization. Recall that in partial cliticization data the clitic stands only for a part of the nominal expression. The relevant example is in (45), where some elements of the nominal expressions have been left stranded after the cliticization process. (45) a. Compramos los libros de Marı´ a y tu´ compraste los libros de Juan (we)bought the books of Maria and you bought the books of Juan ‘We bought Maria’s books and you bought Juan’s books’ b. Compramos los libros de Marı´ a y tu´ los compraste de Juan (we)bought the books of Maria and you CL bought of Juan ‘We bought Maria’s books and you bought Juan’s books’

4.3 Towards an Analysis

145

With this in mind, consider now the examples in (46) in which I illustrate the different behavior shown by the different groups of adjectives with respect to partial cliticization. (46) a. Compramos los libros rojos y tu´ compraste los libros azules (we)bought the books red and you bought the books blue ‘We bought the red books and you bought the blue books’ b. Compramos los libros rojos y tu´ los compraste e azules (we)bought the books red and you CL bought e blue ‘We bought the red books and you bought the blue ones’ c. Compramos los verdaderos manuscritos y tu´ compraste los (we)bought the true manuscripts and you bought the supuestos manuscritos alleged manuscripts ‘We bought the true manuscripts and you bought alleged manuscripts’ d. *Compramos los verdaderos manuscritos y tu´ los compraste we-bought the true manuscripts and you CL bought supuestos e alleged e ‘We bought the true manuscripts and you bought alleged ones’ e. Saludamos al anterior presidente y tu´ saludaste (we)greeted to-the former president and you greeted al actual presidente to-the current president ‘We greeted the former president and you greeted the current president’ f. *Saludamos al anterior presidente y tu´ lo saludaste actual [e] (we)greeted to-the former president and you CL greeted current e The data in (46) show that partial cliticization is not possible when it leaves a prenominal adjective stranded but it is allowed if the stranded adjective is postnominal. My analysis can straightforwardly account for the ungrammaticality of the data with prenominal adjectives: in order to leave a prenominal adjective stranded, partial cliticization should apply to N’. This is not possible because intermediate levels are invisible to syntactic operations (Chomsky (1995)). That is, the ungrammatical data with prenominal adjectives are explained due to the general impossibility of applying syntactic operations to intermediate one-bar levels. Moreover, assuming as before (see Chapter 2) that partial cliticization affects maximal projections, the analysis developed also accounts for the grammaticality of partial cliticization with postnominal Q-As. Similarly to the data with NP ellipsis, the grammaticality of the examples with postnominal adjectives is due to partial cliticization targeting the lower segment of the NP25; hence, postnominal Q-As can be stranded by partial cliticization. To sum up: the previously unobserved data regarding partial cliticization with adjectives can be accounted for under the proposed analysis.

146

4 Adjective Placement

4.3.3 Conclusions In this section I have developed an analysis that divides adjectives into two groups: prenominal adjectives and postnominal adjectives. Prenominal adjectives, the group of adjectives that do not modify the extension of the N, have been merged as the specifier of NP, a position from which they can move in LF to receive their interpretation outside of the scope of the D. Postnominal adjectives, the group of adjectives that modifies the extension of the N, are adjuncts and they are interpreted in situ at LF. Apart from explaining general properties of nominal expressions with adjectives, such as the different meanings they can have, this analysis can account for many syntactic properties, such as the different behavior of prenominal and postnominal adjectives in the contexts of NP-ellipsis and partial cliticization. There is another difference between prenominal and postnominal adjectives, which concerns extraction: (47) a. Han robado varios presuntos/famosos retratos [de Goya]ag (they)have stolen many alleged/ famous portraits of Goya b. ?? ¿ [De quie´n] han robado varios presuntos/famosos retratos tag? of whom (they)have stolen many alleged/ famoous portraits (48) a. Han robado varios retratos famosos/ importantes [de Goya]ag (they)have stolen many portraits famous/ important of Goya b. ¿De quie´n han robado varios retratos famosos/ importantes tag? Of whom (they)have stolen many portraits famous/ important The contrasts above show that the presence of prenominal adjectives in constructions where an argument of the nominal expression has been extracted causes ungrammaticality, while the presence of postnominal adjectives does not affect the grammaticality status of the sentences. I leave accounting for the contrast in question to future research.

4.4 Previous and Further Lines of Research The number of proposals on this topic make it impossible to review all of the analyses in this book and the goal in this section must be quite modest. I just briefly present some major lines of research in Romance adjectives, to offer a background against which we can evaluate the advantages of the analysis proposed in this chapter over previous analyses. To capture the properties of Romance adjectives, previous studies on the topic have either postulated the existence of one base-generated position for adjectives in Romance or the existence of more than one base-generated position for adjectives in Romance. The most usual analyses proposing just one base-generated position for Romance adjectives have also adopted the existence of noun movement (N-movement) in these languages. According to this line of research, the

4.4 Previous and Further Lines of Research

147

different word orders are then achieved through the combination of this N-movement and the base-generated positions in which Romance adjectives are proposed to appear. In Section 4.3, I argued for the existence of two main base-generated positions (as a specifier and as an adjunct) to derive the different types of Spanish adjectives. In Section 4.4.1 below, I review some other generation possibilities for adjectives, exemplify each of the options with a particular analysis and discuss how the Spanish adjective properties behave with respect to the proposal under discussion. Then, in Section 4.4.1, I examine the possibility of considering the Spanish adjectives as adjuncts, as specifiers or as the result of reducing relative clauses. The conclusions of this brief review of the main lines of analysis provide indirect evidence for proposals, such as the one in this chapter, which postulates multiple base-generated positions for adjectives. The second part of Sections 4.4, and 4.4.2, discusses the role that N-movement plays in all the previous analyses and the recent proposals that explore the possibilities of deriving adjectives under a NP-movement approach, to explain the relevant data in Spanish. The chapter concludes with Section 4.5 that summarizes the advantages of my proposal over the previous ones (i.e., it has better empirical coverage and does not make use of additional operations, such as N-movement) and shows that my analysis can be expanded to some other languages to explain the crosslinguistic differences in the location and properties of adjectives in nominal expressions.

4.4.1 Problems for a Uniform Analysis for Spanish Adjectives I have argued in this chapter (Section 4.3) for an analysis based on the existence of different base-generated positions (specifier/adjunct) for the different types of Spanish adjectives. This is obviously not the only theoretical possibility and the literature on Romance adjectives is abundant in examples of analyses that propose to generate adjectives in a particular position and resort to some device to determine the different linear orders (Cf. Alexiadou et al. (2007, 326–387) for a summary of the main lines of research to analyze the distribution of adjectives in different languages). In this section, I review some other generation possibilities for adjectives, namely I discuss the analyses that generate all adjectives as adjuncts, the analyses that propose that adjectives are specifiers and the ones that postulate a relation between relative clauses and adjectives. I exemplify each of the lines of research with a particular proposal and discuss the problems that the Spanish adjective properties pose to the proposal under discussion. Taking as their point of departure previous analyses of adverbs, such as Jackendoff ’s (1972) classification of adverbs for instance, a group of authors (cf. Valois (1991), Bernstein (1993, 2001a), among many others) analyze Romance adjectives assuming that they are the equivalent of adverbs in the

148

4 Adjective Placement

nominal domain. This type of analysis also assumes that there is N-to-D movement in Romance to explain the relative ordering of adjectives in these languages. One of the implementations of this line of research is Bernstein (1993, 2001a). Bernstein has proposed that, in general, adjectives are generated prenominally in Romance and noun movement across adjectives accounts for the unmarked surface order. The difference between the restrictive/nonrestrictive interpretations of some adjectives is due to the different adjunction sites: restrictive adjectives are adjoined to NP and non-restrictive adjectives are adjoined to an intermediate functional category named NumP. For instance, according to Bernstein (1993, 2001a), the derivations of non-restrictive and restrictive adjectives are as in (49a) and (49b), respectively: (49) a. non-restrictive adjective: un gran valle (a big valley)

a. non-restrictive adjective:un gran valle(a big valley) DP D un

NumP NumP

AP gran

Num NP vallei ti

b. restrictive adjective: un valle grande (a valley big) DP D un

NumP

Num vallei

NP

NP AP grande ti

Bernstein’s (1993) proposal is completed by assuming that some adjectives (strictly prenominal adjectives such as mero (mere)) are heads that project to AP and select an obligatory overt NP.26 (50) a. un mero accidente (a mere accident) DP D un

AP

A NumP mero Num NP accidentei t i

4.4 Previous and Further Lines of Research

149

This division of adjectives enables Bernstein (1993) to explain most of the properties of Romance adjectives. However, it is not clear how the two different adjunction sites proposed for adjectives within nominal expressions will explain the different interpretations (namely, restrictive/non-restrictive) that the nominal expressions exhibit. Note that in both cases the adjunction site will be within the domain of the D; therefore, the D will have scope over the two types of adjectives. In other words: a proposal that generates adjectives with non-restrictive readings within the scope of the D should be able to explain how the adjectives are interpreted outside of the scope of the D. Since the mechanism standardly assumed to be in charge of the different scopes is a movement operation in covert syntax, the base-generation position of non-restrictive adjectives must allow them to move out of that position. Assuming that elements in adjunct position cannot be extracted out of the nominal expression (cf. Chapter 2 regarding the impossibility of extracting adjuncts out of nominal expressions), the generation of non-restrictive adjectives in adjunct position does not seem a viable option for the relevant Spanish data. A second line of research explored profusely in the literature is the one developed by Cinque (1995). This approach proposes that adjectives are generated as the specs of different maximal projections inside nominal expressions, with the postnominal orders obtained via N-movement to different positions. The evidence for this analysis comes from the strict word order among adjectives, illustrated in (51); and from the limit in the number of adjectives present in the nominal expression, cf. (52): (51) a. Un masaje a massage

cardiaco largo cardiac long

vs.

*Un masaje largo cardiaco

(52) ??/*El coche japone´s azul deportivo the car Japanese blue sports (cf. El coche japone´s azul/ El coche japone´s deportivo) the car Japanese blue/ the car Japanese sports The existence of the word order effects in (51) can be explained as the consequence of the hierarchy of the functional projections assumed, which host the adjectives in their specifier positions. The data in (52) argue against an adjunction analysis of adjectives, since the number of adjuncts (contrary to the number of functional categories) allowed in a given structure is unlimited. Therefore, the overall properties of adjectives in Romance seem to be captured by the analysis of adjectives as specifiers of functional categories and have led many authors to follow this line of research, applying it to different languages (cf. Sa´nchez 1996, Gutie´rrez-Rexach and Malle´n 2002, Scott 2002, or Laenzlinger 2005, among many others). For our purposes (i.e., to explain the properties of Spanish nominal expressions), one of the most obvious problems with this type of approach when considering it for the Spanish data is the one relative to the extraction possibilities

150

4 Adjective Placement

of nominal expressions and the generation of adjectives in specifiier position. In other words: assuming that extraction out of nominal expressions takes place via specifiers, the main problem of this proposal to explain the Spanish adjectives data concerns extraction with adjectives. That is, Cinque’s (1995) account would predict blocking effects whenever adjectives are present in the construction. This prediction is not borne out, as in the previously unnoticed data given in (53) show: (53) a. Han robado varios presuntos/famosos retratos [de Goya]ag (they)have stolen many alleged/ famous portraits of Goya b. ?? ¿ [De quie´n] han robado varios presuntos/famosos retratos tag? of whom ( they)have stolen many alleged/ famous portraits c. Han robado varios retratos famosos/ importantes [de Goya]ag (they)have stolen many portraits famous/ important of Goya d. ¿De quie´n han robado varios retratos famosos/importantes tag? Of whom (they)have stolen many portraits famous/ important These contrasts show that the presence of prenominal adjectives in constructions where an argument of the nominal expression has been extracted causes ungrammaticality, while the presence of postnominal adjectives does not affect the grammaticality status of the sentences. Hence, the same generation site for both types of adjectives cannot explain the contrast in (53). Finally, there is another line of analyses (cf. recent examples of these proposals such as Kayne (1994), Cinque (2005), among others.) that needs to be mentioned. This type of analysis postulates that adjectives can be derived from restrictive relative clauses via a projection of the functional components of the relative clause and an ulterior movement of the category containing the adjective via predicate fronting. Let us consider this type of analysis as exemplified in Kayne (1994), one of the most recent modifications of this classical analysis. As noted, Kayne (1994) proposes that adjectives must be analyzed as reduced relative clauses. Therefore, the underlying intuition is that (54a) and (54b) are derivationally related. (54) a. the yellow book b. The book that is yellow The structure of a reduced relative clause contains the adjectives functioning as a predicate, as shown in (55). (55) the [CP [XP yellowj] [C0 [IP[book] [I0[ej] Therefore, under this approach, the D takes a CP as its complement (i.e, a relative clause), which contains the adjective and the N in a predicative relation. A second step in the derivation moves the adjective from the internal clausal position to the specifier of CP, resulting in the linear order D-A-N. The main problem with this type of approach is that it predicts that all adjectives will in principle have the same properties as those restrictive relative clauses they are derived from. This is not the case. Crucially, Spanish

4.4 Previous and Further Lines of Research

151

prenominal adjectives do not share properties with restrictive relative clauses but with appositive relative clauses. As the examples in (56) show, restrictive relative clauses and postnominal adjectives can undergo NP ellipsis, while nonrestrictive relative clauses and prenominal adjectives cannot. (56) a. Vimos varios osos peludos (we) saw several bears hairy a’. Vimos varios [e] peludos (we) saw several [e] hairy b. Vimos varios peludos osos (we) saw several hairy bears b’. *Vimos varios peludos [e] (we) saw several hairy [e] c. Vimos varios osos que eran peludos (we) saw several bears that (they)were hairy c’. Vimos varios [e] que eran peludos (we) saw several [e] that (they)were hairy d. Vimos varios osos, que eran peludos (we) saw several bears that (they)were hairy d’. *Vimos varios [e], que eran peludos (we) saw several [e] that (they)were hairy

(restrictive)

(non-restrictive)

Therefore, an account that derives the prenominal position of adjectives in Spanish from a base-generated restrictive relative clause plus a predicate fronting movement cannot explain the grammaticality status of the examples above, which shows a parallel behavior of postnominal adjectives and restrictive relative clauses and prenominal adjectives and non-restrictive relative clauses. Similarly, this type of account will not be able to explain the difference of meanings in the different positions. That is, this approach will predict (56a) and (56b) above to have the same exact meaning; more concretely, it will predict that in both cases the meaning would be the restrictive meaning. Obviously, this is an undesirable prediction and an additional mechanism is required to explain why some adjectives exhibit different syntactic and semantic properties in different linear positions. Furthermore, this type of approach treats all adjectives as a uniform group. This is not correct either. That is, as seen in Section 4.2, some Spanish adjectives are exclusively prenominal; others are exclusively postnominal, while there is a third group of adjectives that can be found in either position. The problem is then that the general type of derivation will predict that any relative clause can be reduced to an adjective and that that particular adjective could appear freely pre or postnominally. Since this is not the case, an additional restrictor to this analysis will be needed. To conclude: there seems to be enough evidence that prenominal and postnominal adjectives in Spanish are generated in different positions. This has been

152

4 Adjective Placement

the claim of several proposals that have argued that the properties of adjectives in Romance cannot be captured by assuming a unique base-generated position of these elements (cf. Demonte 1999b, Duffield 1999, Ticio 2003, Truswell 2006, among many others). In the next section, I will examine the role of nounmovement (N-movement) in the analyses of Romance adjectives.

4.4.2 The Role of Noun Movement in the Analyses of Romance Adjectives As illustrated in the previous sections, most analyses of adjectives in nominal expressions make use of noun-movement to derive either the different linear orders in a particular language or the different linear orders found crosslinguistically. This way, the different adjective positions (i.e., preceding or following the N in the surface order) in languages such as Spanish have been explained by raising the N across the prenominal adjectives in the literature, a movement that is not needed under my analysis.27 This may not be desirable in light of some arguments against the existence of N-movement. As mentioned previously in Section 4.1, Lamarche (1991) was the first to notice some arguments against the existence of N-movement in Romance. Afterwards, some other authors (cf. Bouchard 2002, Alexiadou 2001b, Laenzlinger 2000, 2005, Shlonsky 2004, Ticio 2003, Cinque 2005, among others) have shown more empirical motivation to abandon N-movement as the base of our explanation of the different linear orders (and meanings) of adjectives in nominal expressions. In this section I briefly review the most important problems raised to the N-movement approach; see the references above and Alexiadou et al. (2007, 338–347) for more discussion on the N-movement approach and its problems. Lamarche’s main criticism is based on the fact that the relative ordering of postnominal Romance adjectives with respect to the N is the mirror image of the English order. Consider the following examples: (57) a. un coche blanco oxidado a car white rusty b. a rusty white car c. una fruta naranja enorme a fruit orange enormous d. an enormous orange fruit As Lamarche (1991) notes, an analysis that derives the difference between Romance and Germanic languages regarding the relative position of adjectives with respect to the N as the result of N-movement cannot account for the linear orders shown in (58). More concretely, an analysis that assumes that N-movement accounts for the postnominal position of adjectives in Romance assumes the following two abstract representations for deriving

4.4 Previous and Further Lines of Research

153

the relative ordering of adjectives with respect to the N in Romance and Germanic languages: (58) a. Adj2—Adj1—N b. Ni––Adj2—Adj1—ei

Germanic Languages Romance Languages

Under this type of analysis, the only difference between Romance and Germanic languages is the possibility of having N-movement in Romance languages. The order of the adjectives relative to one another should be the same in both types of language. Therefore, as Lamarche (1991) concludes, the analyses that postulate N-movement to derive the linear order of adjectives relative to the N cannot account for the cross-linguistic differences in (57). A second argument against N-movement analyses is that postnominal adjectives do not appear preceding the complement of the N. This point is illustrated in (59): ´ (59) a. Los productores de petroleo independientes the producers of oil independents ‘The independent oil producers’ ´ b. *Los productores independientes de petroleo the producers independents of oil c. Los estudiantes de fı´ sica simpa´ticos The students of physics nice ‘The nice physics students’ d. *Los estudiantes simpa´ticos de fı´ sica The students nice of physics The contrast in (59) is difficult to explain under the N-movement analysis, since it appears that under this analysis we would expect that the adjective could appear between the N and its complement, as Lamarche (1991) notes. Another empirical argument in favor of abandoning the N-movement approach is based on the relative scope of adjectives and its relation with linear order. As Svenonius (1994, 450) notes, the scope relations between the two adjectives in the examples below create a problem for the N-movement analysis. (60) a. Pollo congelado troceado chicken frozen chopped ‘Chopped frozen chicken’ b. Pollo troceado congelado chicken chopped frozen ‘Frozen chopped chicken’ (Adapted from Svenonius (1994, 450, (17)) (60a) refers to chicken that was first frozen and then chopped, while (60b) refers to chicken that was first chopped and then frozen. Then the closest adjective to the N has the narrowest scope (that is, it has scope only over the N), while the outer adjective has scope over the adjective and the N.

154

4 Adjective Placement

The scope relations are unexpected under an N-movement account because they are inverted in English and Spanish and, as the schema in (61) illustrates, the different scope relations and linear word orders cannot be derived via N-movement. (61) a. N ADJ1 a’. ADJ2 b. N ADJ2 b’. ADJ1

ADJ2 Adj1 ADJ1 Adj2

N N

Spanish, ADJ 2 >Adj1, (60a) English, Adj2>Adj1 Spanish, ADJ 1 >Adj2, (60b) English, Adj1>Adj2

As shown in the schema above, the linear orders of the adjectives are the mirror image too, which defies a derivation based exclusively in N-movement. The most compelling evidence for the existence of N-movement in Romance languages comes from the behavior of the definite article in Romanian. Consider (62): (62) a. Ba`iati-ul sa`rac ti boy- the poor b. Acest sa`rac ba`iat this poor boy As the examples in (62) show, Romanian definite articles appear attached to the N, which suggest that the N reaches D0 in this language. However, note that, as pointed, for instance, in Bernstein (1993), the definite article in Romanian can be also attached to an adjective, as the example in (63) illustrates: (63) simplui fapt ca. . . mere-the fact that. . .

(Bernstein 1993, 89)

Therefore, the example in (63) undermines the hypothesis that N must raise to D in Romanian and leave the analyses that argue for N-movement in Romance without their strongest support. Moreover, note that the existence of examples such as (63) seems to point to a PF explanation of this phenomenon in Romanian as well. In other words: it can be the case that the definite article in Romanian has a clitic character too and triggers some type of last resort PF movement of the closest stressed head. Furthermore, as mentioned earlier, the availability of head movement in our theory of grammar is currently under debate. Although Matushansky (2006) and Pereltsvaig (2006) are arguing for the need of head movement in our theory of grammar, there is a line of research represented by Boeckx & Stjepanovic (2001) for instance that has tried to relocate head movement into the phonological component. Similarly, authors such as Alexiadou (2001a) and Shlonsky (2004), among others, have developed analyses that derived the alleged effects of N-movement from phrasal movement. Hence, given not only the questionable status of N-movement in Romance as the mechanism to derive the different linear orders and the properties of

4.5 Summary of the Chapter

155

adjectives and PP modifiers but also its questionable theoretical status, it seems to be a strength of my analysis that I do not have to rely on it.28

4.4.3 Conclusions Section 4.4 has pointed out the problems that the different properties of Spanish adjectives pose to the main lines of research explored to analyze the properties of adjectives in Romance. A few conclusions can be extracted from this review: first, it seems necessary to postulate more than one position for adjectives in Romance unless an additional mechanism of movement is postulated to derive the different word orders; and, second, adjectives that can scope out of the nominal expression must be merged in a position from which they can move, to be consistent with the movement restrictions in the nominal domain shown in previous chapters. The remaining of this chapter summarizes my analysis of adjectives in Spanish and speculates how it can be extended to some other languages.

4.5 Summary of the Chapter The analysis of adjectives presented in this chapter divides them into two main groups: prenominal adjectives and postnominal adjectives. Prenominal adjectives, the group of adjectives that do not modify the extension of the N, have been merged as the specifier of NP, a position from which they can move in LF to receive their interpretation outside of the scope of the D. Postnominal adjectives, the group of adjectives that modifies the extension of the N, are adjuncts and they are interpreted in situ at LF. In addition, my analysis assumes that relational adjectives, which are always postnominal, are merged in their respective argumental position, since they behave as PP arguments. Furthermore, the current analysis does not need to make use of N-movement to derive the different linear positions of adjectives in Spanish nominal expressions. So, the different adjective positions (i.e., preceding or following the N in the surface order), which have been explained by raising the N across the prenominal adjectives in the literature, can be explained under my analysis as the result of different generation sites for each type of adjective. Given the above assumptions, the analysis developed in this chapter explains a number of facts and properties of nominal expressions, such as adjective placement, differences in NP ellipsis and different meanings of different types of adjectives, which raised problems for previous approaches to the nominal expressions in Romance. Moreover, the current analysis also explains some previously unobserved facts, such as those concerning partial cliticization data with adjectives and draws several cross-linguistic generalizations that were not noted before.

156

4 Adjective Placement

Nevertheless, the analysis of adjectives put forward in this chapter seems to be focused on the empirical evidence extracted from the Spanish nominal phrase and a valid question at this point is how this analysis can account for the cross-linguistic facts previously described. Although this is certainly topic for further research and falls out of the scope of this chapter, let me provide here a few speculations explaining how to expand the empirical coverage of this type of analysis to other languages. My analysis of adjectives follows the idea of a direct correlation between position and interpretation of adjectives. This means that the position an adjective occupies in the nominal expression reflects the way the adjective is interpreted. This idea links my analysis with a line of research that proposes that the sequential order on adjective co-occurrences is structure dependent and identifies different positions associated with different classes of adjectives (cf. for instance Scott (2002) to this respect). In my analysis of Spanish adjectives, the different meanings correspond with pre and postnominal positions for those adjectives that can receive different interpretations and result in a difference on the height of the adjectives (i.e., prenominal adjectives being higher than postnominal adjectives). The prediction for some other languages is that the structural height should maintain. This prediction is borne out in some cases in English, as the examples in (64) illustrate. (64) a. b. c. d.

*a fake beautiful Picasso a beautiful fake Picasso *a former wonderful actress a wonderful former actress

Assuming that linear precedence implies structural dominance, the ungrammaticality of the examples in (64a) and (64c) can make us conclude that the adjectives beautiful and wonderful, which correspond to the postnominal position in Spanish, are higher than fake and former, prenominal adjectives in Spanish. There is, according to the previous examples, two prenominal positions for adjectives in languages such as English and they keep a similar structural hierarchy as the one seen in Spanish. Assuming that something along this line is on the right track, the difference between Spanish and English in this respect rests on the position in which we linearize the Q-A restrictive adjectives. This proposal is based on an idea that has been argued several times in the literature (cf. Chomsky (1995) for instance) and claims that adjuncts can be ordered freely (i.e., preceding or following) with respect to the category they are attached to. In other words, the proposal is that in Spanish we pronounce the adjunct following the NP, while in English the adjunct is pronounced preceding the NP. This working hypothesis could explain the linear precedence with respect to the adjectives occupying the specifier position in English, while maintaining the structural hierarchy and the similar semantic properties in both languages. A schema of the crosslinguistic variation is in (65)

4.5 Summary of the Chapter

(65) a. D–––Adj (adjunct) –––Adj (spec) ––– N b. D–––Adj (spec) –––N–––Adj (adjunct)

157

English Spanish

Note that this type of analysis is also able to explain the problem of relative scope of adjectives in Spanish and English, the data are repeated below. (66) Pollo congelado troceado chicken frozen chopped ‘Chopped frozen chicken’ (67) Pollo troceado congelado chicken chopped frozen ‘Frozen chopped chicken’ Recall that the scope relations are inverted in these examples: the first example refers to chicken that was first frozen and then chopped, while the second example refers to chicken that was first chopped and then frozen. Under the analysis outlined above, this is expected, as illustrated in (68). (68) a. D–––Adj1 (adjunct) –––Adj2 (adjunct)––– N b. D––– N–––Adj1 (adjunct)–––Adj2 (adjunct)

English Spanish

As seen in (80), the relative ordering between the two adjectives is maintained and the only difference is how both adjuncts as a group are linearized with respect to the NP in each particular language. Unfortunately, not all the data seem to support this type of analysis. In fact, although it seems that there is a relative ordering of adjectives within English nominal expressions and there is evidence for two positions for adjectives in English; the relative ordering does not always correspond to the expected order under my working hypothesis.29 Consider the examples in (69) in this respect: (69) a. a poor rich boy b. Un pobre nin˜o rico a poor boy rich As shown in the translation into Spanish (cf. (81b)), the linear order in English prenominal adjectives in (69a) is the opposite to the one predicted by our hierarchy. In these examples it is poor, which corresponds to the prenominal adjective in Spanish, the one that precedes the restrictive adjective.30 No definitive conclusion can be reached at this point regarding the possibility of the extension of my analysis to some other languages, since further research is still needed in these issues. Nevertheless, a hypothesis that emphasizes the role of the PF interface in the cross-linguistic variation of adjectives seems to be worthy of exploration. The main advantage of the analysis presented in this chapter is its empirical coverage. The current analysis not only explains a number of facts and properties of nominal expressions, such as adjective placement, differences in NP ellipsis and different meanings of different types of adjectives, which raised problems for previous approaches to the nominal expressions in Romance but

158

4 Adjective Placement

also it is able to explain some previously unobserved facts, such as those concerning partial cliticization data with adjectives and draws several crosslinguistic generalizations that were not noted before. In addition, the analysis is compatible with the explanation put forward in previous chapters for some other properties of nominal expressions in Spanish, which allows us to have a uniform analysis for the nominal expression in Spanish. Finally, to the extent that is correct this analysis provides empirical evidence for the need to have a distinction between adjuncts and specifiers in the theory of grammar and supports the line of research that proposes the non-existence of N-movement.

Notes 1

2

3

4

For an extensive survey of the recent research on adjectives and for a detailed discussion of the basic properties of adjectives in different languages the reader can check part III, Chapter 1 of Alexiadou et al. (2007: 284–394). The adjectives in (13) fall under those described below as relational adjectives, which will be discussed below (see examples in (16)). The reader can also find a detailed discussion on classifying adjectives in Romance languages in Alexiadou et al. (2007, 318–320, 330, and 390–393). For a survey on classical semantic classifications of adjectives (such as the distinctions attributive/predicative, intensional/extensional and intersective/non-intersective, among others) check Alexiadou et al (2007, 290–336). Note that there are cases of R-As that accept degree, such as (i): (i) Una revista muy americana a journal very American However, in this case the adjective americana is not considered an R-A in the literature but a Q-A.

5

6

Kayne (1984) and many other authors reach the same conclusion: a significant number of relational adjectives, the so-called referential adjectives, behave as arguments in the nominal domain. According to Cinque (1980), among many others, Ns such as capture are inherently passive. Passive Ns have only the object readings in cases such as the one given in (i): (i) La captura del soldado the capture of+the soldier Thus, the PP del soldado (of the soldier) cannot be interpreted as an agent in (i). Furthermore, if an agent appears in the nominal expression, it only can appear with the P por (by), as (ii) illustrates: (ii) a. La captura del soldado por los americanos. the capture of+the soldier by the Americans b. *La captura del soldado de los americanos. the capture of+the soldier by the Americans

7

According to Demonte (1999a), the overall contrast is intensional (prenominal) vs. extensional (postnominal).

Notes 8

159

Note that examples such as (15b) can be grammatical in contexts such as the one given below: (i) El Mozart genial era el hijo, no su padre. the Mozart genial was the son, not his father I think that in these cases, we are treating the proper N as a common noun, as we compare different individuals that are in the denotation of Mozart.

9

10

11 12

For a detailed study of the properties of proper Ns with adjectives, see Uriagereka (2001), who proposes a different approach to proper Ns and adjective combinations in Spanish. Many authors (Martı´ n 1995, Sa´nchez 1996, among others) have attempted to analyze appositive relative clauses and prenominal Q-As as related. Most examples and observations in this part of the chapter come from Bosque (2001). Bosque (2001, 22) claims that imperatives do not generally allow for specific indefinite expressions. This does not imply that we cannot have specific nominal expressions in directive contexts, (cf. Escribe El Quijote, (you) write El Quijote). However, in cases with indefinite expressions we force the specific reading through the presence of the preposition a (to), as in (i) below:

(i) a. Contrata a un escritor famoso (you)Hire to a writer famous b. (?) Contrata a un famoso escritor (you)Hire to a writer famous Note that in (ib) it is possible (although unnatural) to have a prenominal adjective and in both cases the indefinite is interpreted as specific. 13

14

The relevant data are not completely clear here. As Juan Uriagereka (p.c.) points out to me, there are donkey-sentences that are not ungrammatical when a prenominal adjective is inserted within the indefinite nominal expression. I leave this point open here, since it does not affect the analysis proposed in this chapter. Predicate Modification is defined as follows: (i) Predicate Modification (PM) If a is a branching node, fb; gg is the set of a’s daughters and kbk and kgk are both in D 5 e;t 4 , then kak ¼ l x 2 De kbkðxÞ ¼ kgkðxÞ ¼ 1. (Heim & Kratzer (1998:65))

15

The formal denotations for girl and tall are given in (i) and their combination via predicate modification is given in (ii): (i) jgirl’j ¼ l x: x is a girl jtall’j ¼ l x: x is a tall (ii) jgirl’jðjtall’jÞ ¼ l x: x is a girl and tall

16

Formally, a prenominal Q-A has the denotation in (i), where the prenominal Q-A is a function that takes an individual and returns the same individual with a presupposition. (i) jtall’j ¼ l x : x is a tall:x The formal denotation of (24b’) is represented in (ii), where the adjective tall takes the individual denoted by the girl: (ii) jtall’jðjthe girl’jÞ ¼ l x : x is a tall. x is the unique individual that is a girl.

17

The formal denotation of A-As such as future is represented in (i): (i) For any t 2 T :kfuture’kt ¼ l f 5 i; 5 e;t 4 4 :l x½fðtÞðxÞ ¼ 0&9t’ after t: fðt’ÞðxÞ ¼ 1

160

4 Adjective Placement

Combining the denotation in (i) with the denotation of president in (iia), we obtain the denotation in (iib) for future president. (ii) a. For any t 2 T :kpresident’kt ¼ l x:. x is
a president at t. b. For any t 2 T :kfuture’kt kpresident’kt ¼ l x:. x is not a president at t and there is a t’ after t such that x is a president at t’. 18

19

20

21

22 23

Note that, for expository purposes, I reproduce the structure containing the functional categories used in Chapter 2. This does not imply that all the functional categories must be present and equal in all the derivations, as shown in Chapter 2. As noted in Chapter 2, for expository reasons, I do not assume the LCA-compatible analyses in this chapter. I follow the standard analyses that argue that the surface orders are obtained by the directionality of projection of the different categories involved in the structure (cf. Torrego (1987) and Ormazabal (1991) for similar proposals) with agent and poss occupying right specs. Note that, in order to account for the order of adjectives under my proposal, it cannot be the case that all categories project the specifiers to the right, since prenominal adjectives would follow the N in this situation. To accommodate the data with adjectives, I assume here that functional and lexical categories differ in the directionality they project their specifiers. This proposal is in the spirit of Romano’s (1991) analysis of ASL. Romano (1991) examines the structure of verb sandwich constructions, tensed sentences and classifiers in ASL and concludes that ASL is head-initial for lexical categories and head-final for functional categories. Assuming that the directionality of projection can extend to specifiers, the claim here is that Spanish lexical categories within nominal expressions project their specifier to the right, while Spanish functional categories within nominal expressions project their specifier to the left. Crucially, following Huang (1982), I assume that some locality conditions, including some of those discussed in Chapter 2, do not hold in LF where the movement of the adjective would take place. More specifically, the problem of adjective-fronting type analyses is an overgeneration one. That is, this type of analysis overgenerates prenominal adjectives, since in principle all postnominal adjectives are amenable to be fronted. This is not the case, as was shown previously, because some postnominal adjectives cannot appear in the prenominal position. Let me explicitely note here that this complication does not arise with my analysis of adjectives for two reasons: first, my use of predicate fronting applies only to an already preselected group of adjectives (i.e., only to those adjectives displaying certain semantic properties). Second, the group of adjectives amenable to be moved is already in prenominal position, therefore, there is not movement across the noun. In other words, the classical adjective fronting analyses and my analysis only coincide in the type of operation used to obtain the final order of a subset of adjectives. See Chapter 4 for a detailed analysis of nominal elision in Spanish. This is not entirely true, since it might be the case of an R-A in the NP. For instance, if we have novela amorosa (romance novel), where the adjective amorosa (loving) is the object/ theme of the novel. These type of nominal constructions can leave stranded the R-A in nominal ellipsis contexts, as (i): (i) Pepe lee muchas novelas de aventuras y Juan lee demasiadas [e] amorosas Pepe reads many novels of adventures and Juan reads too-many [e] loving Note that in order to provide a uniform explanation for all ellipsis contexts, we must say that the R-A moves as the PP object would do, via a stylistic movement.

24

As an anonymous reviewer notes, a difficult case to explain here will be the possibilities of NP ellipsis with adjectives in German. German generally allows NP ellipsis with adjectives,

Notes

25

26

27

28

161

while it has all its adjectives in the prenominal position. Although a more detailed study of German adjectives is required for me to provide a definite answer on this issue, I will tentatively assume that German adjectives can scope out the nominal expression in some circumstances, which will allow having some prenominal adjectives in nominal ellipsis contexts. Note that the target of partial cliticization (i.e., the lower segment of NP) can be considered a maximal category even if the NP does not contain the adjunct. This approach is similar to the one developed by Sportiche (1988) to account for floating quantifier data (cf. also Bosˇ kovic´ (2002b)). According to Sportiche’s (1988) analysis of floating Qs, floating Qs are adjoined to the NP and are left stranded after the NP (its lower segment) undergoes movement. Bernstein (1993,105n49) acknowledges that there is a problem with the claim that adjective selects NP in the structure in (33). She proposes to adopt Grimshaw’s (1990) notion of extended projection to accommodate these data. Thus, in (33) it is possible to say that adjective selects an obligatory NP, since the functional projection NumP is, in this sense, considered a projection of the head N. My analysis derives the different orders by placing some adjectives in specifiers or adjuncts rather than raising N over some adjectives. See Cinque (2005) for a different solution to derive the nominal expressions internal order and Matushansky (2006) for an analysis of head movement. Recent research (cf. Aboh (1998), Shlonsky (2004) or Cinque (2005), among others) has derived the different linear word orders of adjectives from a common structural base via leftward movement of NP. See Alexiadou et al. (2007, 379–383) for a more detailed discussion of the proposal and its challenges. For our purposes, the relevance of this type of analysis is that it seems to be able to derive some data that could not be explained via N-movement approaches. For instance, the relative order of adjectives with respect to other elements in the nominal expressions, see (i) below, can be captured now. ´ (i) a. Los productores de petroleo independientes the producers of oil independents ‘The independent oil producers’ ´ b. *Los productores independientes de petroleo the producers independents of oil c. Los estudiantes de fı´ sica simpa´ticos The students of physics nice ‘The nice physics students’ d. *Los estudiantes simpa´ticos de fı´ sica The students nice of physics

The ungrammaticality status of (ib) and (id) can be predicted under this type of approach since the complement of the N moves along with the N and no adjective is be expected to intervene between the N and its complements. Similarly, some cross-linguistic differences regarding the relative ordering of adjectives, see example (iia), could receive now an explanation. This is because this type of analysis will make it possible to derive the order of the adjectives relative to one another by raising the NP to a higher specifier position and by pied piping the lower adjective in this movement, as illustrated in (iic). (ii) a. una fruta naranja enorme A fruit orange enormous ‘an enormous orange fruit’ b. Adj2—Adj1—NP c. [ NPi Adj1]-Adj2—ej—ei

Germanic Languages Romance Languages

162

4 Adjective Placement

As (iic) shows the NP will raise from its original position (i.e., the one that corresponds to the Germanic surface order) to the specifier preceding Adj1. In its second movement, NP collects Adj1 and places it in the specifier position preceding Adj2. As a result of pied piping, the Romance linear word order is obtained. Then, the NP movement type of analysis could solve the problems for analyses based on N-movement in the examples above. Nevertheless, this analysis has some problems as well. First, the most general problem is that there is no clear motivation for all the necessary movements to derive the desired linear orders. Second, a more specific problem is how to make this type of analysis, which is based on indiscriminated movements to specifier positions, compatible with the movement restrictions shown in Chapter 2 for languages such as Spanish. 29

30

See Truswell (2009) and the references indicated there, for extensive data that seem to go against the working proposal in the text. Note that this can be due to several factors. First, it is known that the relational adjective appears closer to the noun and they are lower than all other adjectives, as in (i). (i) a. *A French wonderful car b. a wonderful French car

Hence, in a poor rich boy, the adjective rich can be seen as a relational/classifying adjective and wonderful as an evaluative adjective. If this is on the right track, the position of rich can be explained under my hypothesis. Furthermore, as observed in the literature (cf. Alexiadou et al. (2007, 305–326) for an extensive summary on the topic), there are several factors that can affect the relative ordering of adjectives within nominal expressions. For instance, focus on the adjective can alter the grammaticality status of a particular ordering sequence, as illustrated in the examples below. (ii) a. The big grey cat b. *The grey big cat c. The GREY big cat, not the white big one. Alexiadou et al. (2007: 320)

Chapter 5

Nominal Ellipsis

5.1 Introduction The different types of elements that can appear in the Spanish nominal expression and their properties have been the central issue of inquiry in the previous chapters of this book. The results of this investigation have led to the proposal of a structure of nominal expressions that accounts for a number of syntactic properties so far. More concretely, general word order data, extraction data and modification data have been explained under the same analysis. In this chapter, I show that this analysis is also able to accommodate and predict the omission of linguistic material in the nominal expression. The purpose of this chapter is therefore to examine the types and properties of nominal elision in Spanish1 in order to incorporate them into the basic structure of Spanish nominal expressions proposed in this book (see Chapter 2). Thus, I will discuss the properties of nominal elision in Spanish and their relation to other elements within the nominal expression, to classify the phenomenon and to be able to provide an analysis that extends the empirical coverage of the current analysis of nominal expressions in Spanish developed in Chapter 2. Ellipsis constructions have raised some important questions in the theory of grammar. In this chapter I will contribute to the current debate on the nature and properties of ellipsis constructions by providing some answers to these questions, supported by evidence extracted from the analysis of nominal ellipsis constructions in Spanish. One of the fundamental open questions regarding ellipsis in natural languages pertains to the nature of the omitted elements in ellipsis data. Syntactic analyses of ellipsis processes have developed in two main directions. One view argues that there is no syntactic structure in the ellipsis site. That is, this line of research argues that the computational system generates empty categories in place of the elided material. After the derivation is sent to the interfaces, the meaning is recovered from context at the LF interface. M.E. Ticio, Locality Domains in the Spanish Determiner Phrase, Studies in Natural Language and Linguistic Theory 79, DOI 10.1007/978-90-481-3398-7_5, Ó Springer ScienceþBusiness Media B.V. 2010

163

164

5 Nominal Ellipsis

A second line of thought claims that there is full internal structure in the ellipsis site at some point of the derivation. The fundamental idea under this approach is that the computational system derives a full structure in place of the elided material and sends it to both interfaces. At the LF interface, the structure is interpreted. At PF, an operation that deletes the redundant phonetic and phonological material creates the ellipsis site. Under this view, ellipsis is a PF phenomenon. In this chapter I will provide strong evidence for this latter view of ellipsis. The main reason to argue for an analysis of ellipsis as a PF phenomenon will come from the interaction between nominal ellipsis properties and extraction facts discussed previously (cf. Chapters 2 and 3 in this respect). As will become clear in Section 5.3, the only successful account for nominal ellipsis in Spanish is the one that allows some remnant elements to move outside of their maximal projections prior to the application of the ellipsis operation at PF. Focusing on the nominal domain, research on nominal ellipsis has discussed extensively on the type of gap created by the ellipsis process in the nominal construction. Some authors, Bernstein (1993) and Valois (1991), among others, suggested that only maximal categories were affected by the ellipsis process (i.e., NP ellipsis, XP ellipsis), while some others, Jackendoff (1971) and Valois (1991), among others, postulated that the omission was affecting smaller structures in the nominal constructions (hence, N-ellipsis, head ellipsis, or N’ ellipsis). In this chapter I compare the different cases of omission and conclude that only an approach assuming that nominal ellipsis is XP ellipsis will be able to account for all properties discussed in the chapter. Therefore, the data and discussion in this chapter will also provide strong empirical evidence to the line of research that claims that nominal ellipsis is XP ellipsis. Furthermore, in this chapter I will also elaborate on the role that notions such as agreement and focus have in ellipsis processes. As we will see, most of the previous analyses of ellipsis have assumed that agreement has a licensing role in this construction. In this chapter, I claim that the differences between nominal ellipsis in expressions with or without the definite article in Spanish are proof that agreement cannot be the only licensor for the construction. As for the role of focus in the construction, the movement restrictions studied in Chapter 2 will defy an analysis of NP ellipsis based exclusively on focus as a licensor via movement. The structure of the chapter is as follows: Section 5.2 describes the major syntactic properties of nominal elision in Spanish.2 In Section 5.2.1, I briefly describe the contexts in which different nominal elements can be elided. In Section 5.2.2, I compare the phenomena illustrated in Section 5.2.1 with data containing ellipsis and gapping in the clausal domain and argue that the phenomena in Section 5.2.1 must be classified as cases of ellipsis of major projections within nominal expressions. Section 5.2 concludes with the major descriptive generalizations that any analysis of nominal elision in Spanish must explain. Namely, that nominal ellipsis

5.2 Nominal Ellipsis

165

applies generally in Spanish and that any modifier (except for prenominal adjectives) can be left stranded by the application of elision to nominal expressions headed with Ds others than the definite article. The atypical behavior of the definite article with respect to ellipsis will be attributed to their peculiar properties; more concretely, to its clitic-like character independently noted in the literature. Section 5.3 presents my analysis of elision in Spanish nominal expressions. The analysis is based on the structure presented in Chapter 2 and on the assumption of ellipsis as a PF phenomenon. Therefore, I propose that some maximal projections, such as NP, are deleted in PF. To account for the data with remnant elements, I assume that some remnant elements move outside of their maximal projections prior to the application of the ellipsis operation at PF. Finally, to account for the restrictions in nominal ellipsis in nominal expressions headed by the definite article, I follow previous analyses, such as Brucart and Gra`cia (1986) and Raposo (1999) and assume that there is a phonological constraint involved in these cases. In Section 5.4, I examine the relevance of agreement and focus in nominal ellipsis processes. To do so, I briefly review the main analyses proposed in the earlier literature on elision in Spanish nominal expressions that take these notions as crucial for their analyses and evaluate them with respect to the descriptive generalizations obtained in Section 5.2. This evaluation enables us to check the relevance of these notions to any analysis of nominal ellipsis in Spanish, to determine the adequacy of previous analyses and to point to possible extensions of the proposal offered in this chapter. The chapter ends with a summary of the findings and problems encountered by the proposed analysis.

5.2 Nominal Ellipsis 5.2.1 Description Spanish allows elision in nominal expressions, as the examples in (1) illustrate. The examples in (1) illustrate that we can elide the N in a nominal expression containing only the D and the N, unless the D is the definite article, as seen in (1e). (1) a. Compramos muchos libros y tu´ compraste algunos [e]3 (we)bought many books and you bought some [e] b. Compramos algunos libros y tu´ compraste demasiados [e] (we)bought some books and you bought too-many [e]. c. Compramos estos libros y tu´ compraste aque´llos [e] (we)bought these books and you bought those [e]

166

5 Nominal Ellipsis

d. Compramos dos libros y tu´ compraste uno [e] (we)bought two books and you bought one [e] e. *Compramos un libro y tu´ compraste el [e] (we)bought a book and you bought the [e] Similarly, the examples in (2) show that any of the modifiers contained within the nominal expression can be elided with the N in Spanish. Again, the only restriction concerns the presence of the definite article as the head of the nominal expression, as the ungrammaticality of (2e) shows. (2) a. Compramos muchos libros de fı´ sica y tu´ compraste algunos [e] (we)bought many books of physics and you bought some [e] b. Compramos estos libros con pastas azules y tu´ compraste (we)bought these books with blue covers and you bought aque´llos [e] those [e] c. Compramos bastantes libros para regalo y tu´ compraste uno [e] (we)bought several books to gift and you bought one [e] d. Compramos muchas reproducciones de las Meninas de Vela´zquez (we)bought several reproductions of the Meninas of Velazquez de este coleccionista y tu´ compraste algunas [e] of that collector and you bought some [e] e. *Compramos varios libros sobre Chomsky y tu´ compraste (we)bought several books about Chomsky and you bought el [e]4 the [e] This way, the examples in (2a) and (2d) show that any of the PP arguments (and any number of them) can be omitted in the ellipsis process; similarly, the grammaticality of (2b) and (2c) illustrate that no argumental PP modifiers can be also elided with the N in Spanish. As in (1e), the ungrammaticality of (2e) exemplifies that the only time elision is impossible is when it leaves the definite article followed by the ellipsis site.5 Apart from the data where nominal elision applies to the N and its modifiers, Spanish allows what I will call here partial nominal elision. Thus, any of the PP elements contained within the nominal expression can be left stranded after elision in Spanish. The data in (3) exemplify this claim: (3) a. Compramos muchos libros de fı´ sica y tu´ compraste algunos [e] (we)bought many books of physics and you bought some [e] de quı´ mica of chemistry b. Compramos estos libros con pastas azules y tu´ compraste aque´llos [e] (we)bought these books with covers blue and you bought those [e] con pastas rojas with covers red

5.2 Nominal Ellipsis

167

c. Compramos bastantes libros para regalo y tu´ compraste uno [e] (we)bought several books to gift and you bought one [e] para consulta to consult d. Compramos bastantes libros de Ana y tu´ compraste uno [e] (we)bought several books of Ana and you bought one [e] de Pepe of Pepe e. Compramos bastantes libros de Cervantes y tu´ uno [e] (we)bought several books of Cervantes and you one [e] de Shakespeare of Shakespeare f. *Compramos varios libros sobre Chomsky y tu´ compraste el [e] (we)bought several books about Chomsky and you bought the [e] sobre Postal about Postal g. Compramos varios libros de Chomsky y tu´ compraste el [e] (we)bought several books de Chomsky and you bought the [e] de Postal of Postal The grammaticality of the examples in (3a) and (3b) illustrate that an object or a PPA (cf. Chapter 2 for a discussion of PPA modifiers) can be the remnant of nominal elision. Similarly, the grammaticality of (3c) shows that an adjunct can be left behind in a nominal elision context. As for possessors and agents, the grammaticality of the examples in (3d) and (3e) illustrates that they can also be remnants of nominal elision. Finally, the contrast in the grammaticality in (3f) and (3g) refines the condition on the occurrence of the definite article in elision contexts. A nominal expression headed by the definite article can undergo nominal elision only if the P de (of) introduces the remnant element.6 These properties of ellipsis in nominal expressions with the definite article will be the topic of Section 5.3.1.2 below. I will suggest that although ellipsis in nominal expressions with the definite article follows the general pattern of nominal ellipsis in Spanish and its analysis, the clitic status of the definite article causes it to exhibit some differences with respect to other Ds in nominal ellipsis contexts. Returning to the types of elements that can appear as remnants, note that the situation changes when the nominal expression contains an adjective: (4) a. *Ayer vi a la verdadera terrorista y a la supuesta [e] yesterday (I)saw to the true terrorist and to the alleged [e] b. Ayer vi la casa azul y la [e] verde yesterday (I)saw the house blue and the [e] green c. Vimos bastantes novelas policı´ acas y tu´ viste (we)saw several novel police and you saw algunas [e] roma´nticas

168

5 Nominal Ellipsis

some [e] romantic d. Vimos bastantes descripciones americanas y tu´ viste (we)saw several description American and you saw algunas [e] francesas some [e] French As mentioned in Chapter 3, nominal elision can have postnominal adjectives as remnant elements, while prenominal adjectives cannot be left stranded after the elision process applies.7 Furthermore, the same regularity of patterns can be observed when we have more than one element as the remnant. The data below illustrate this point: (5) a. Compramos el libro de fı´ sicaobj de Marı´ aposs y tu´ compraste el [e] (we)bought the book of physics of Marı´ a and you bought the [e] de Juanposs de quı´ micaobj of Juan of chemistry ´ de Irakobj de los USag y tu´ viste la [e] b. Vimos la invasion (we)saw the invasion of Iraq of the US and you saw the [e] de Checheniaobj de Rusiaag of Chechnya of Russia (6) a. *Juan conocio´ al presunto asesino japone´s y Marı´ a al Juan met to+the alleged murderer Japanese and Marı´ a to+the verdadero [e] americano true [e] American b. Juan conocio´ al presunto asesino japone´s y Marı´ a al [e] Juan met to+the alleged murderer Japanese and Marı´ a to+the [e] americano American c. Juan conocio´ a la mujer alta simpa´tica y Peter conocio´ a la Juan met to the woman tall friendly and Peter met to the [e] antipa´tica [e] unfriendly d. Juan conocio´ a la mujer alta simpa´tica y Peter conocio´ a la [e] Juan met to the woman tall friendly and Peter met to the [e] bajita antipa´tica short unfriendly As the sentences in (5) and (6d) show, nominal elision can have two PP modifiers and two postnominal adjectives as its remnants. However, the presence of a prenominal adjective, as in (6a), causes ungrammaticality. The descriptive generalization is that any modifier (except for prenominal adjectives) can be left stranded by the application of nominal elision. Focusing on the D in the elliptical construction, there are some other properties to consider here. According to Bosque and Demonte’s (1999) description of the relevant data, the presence of a D is required in nominal elision data. This explains the ungrammaticality of (7):

5.2 Nominal Ellipsis

169

(7) a. *Compramos muchos libros de fı´ sica y tu´ compraste [e] (we)bought many books of physics and you bought [e] b. *Compramos estos libros con pastas azules y tu´ compraste [e] (we)bought many books with covers blue and you bought [e] Bosque and Demonte (1999) account for the ungrammaticality of the data in (7) by assuming that a D must appear phonetically realized in nominal elision contexts, due to its referential value. However, this claim is problematic. First, it is not clear that there can be a relation between reference, a semantic notion and phonetic content, given that semantics and phonetics do not interface in the standard conception of the grammar. Note also that, if we link semantic reference to phonetic content of a given element, we make strong cross-linguistic predictions, such as the prediction that languages that do not have overt Ds would not be able to refer. Moreover, putting aside the cross-linguistic predictions, this link between semantic reference and phonetic content would have problems explaining why null subjects in Spanish can refer. Furthermore, note that Spanish can have bare plurals in some restricted contexts, as in (8). Crucially, nominal ellipsis is allowed in these contexts, as (9) shown by: (8) a. Llegaron estudiantes Arrived-3ppl students (9) a. Llegaron estudiantes ignorantes pero no [e] maleducados Arrived students ignorant but not [e] rude b. Compramos librosde lingu¨ı´ stica y Pedro vendio´ [e] de matema´ticas we-bought books of linguistics and Peter sold [e] of mathematics Taking into consideration the data in (9), it cannot be the case that the presence of a phonetically realized D is required in all nominal elision contexts in Spanish. Consider now the following contrast, which will help us to understand the ungrammaticality of the data in (7) and to derive the associated generalization: (10) a. Compramos libros de lingu¨ı´ stica y Pedro vendio´ libros (we)bought books of linguistics and Peter sold books de matema´ticas of mathematics b. * Compramos libros de lingu¨ı´ stica y Pedro vendio´ [e] (we)bought books of linguistics and Peter sold [e] The ungrammaticality of (10b) shows that it is not possible to elide the entire nominal expression when a bare plural is the antecedent. This fact is related to the constraints on null complement anaphor studied in Depiante (2001).

170

5 Nominal Ellipsis

According to Depiante (2001), null complement anaphora (i.e., the cases in which the clausal complement of the V is null) is highly restricted in Spanish and Italian: (11) a. Juan quiere ir al cine y Marı´ a tambie´n quiere [e] Juan wants go to+the cinema and Marı´ a too wants [e] ‘Juan wants to go to the cinema, and Marı´ a also wants’ (Depiante (2001: 16)) b. *Juan espera ir al cine y Marı´ a tambie´n espera [e] Juan hopes go to+the cinema and Marı´ a too hopes [e] ‘Juan hopes to go to the cinema, and Marı´ a also hopes’ As the examples in (11) illustrate, only the clausal complement of some restructuring verbs, such as querer (want), can be null, while the clausal complement of other restructuring verbs, such as esperar (hope), cannot be null. I will assume here that the ungrammaticality of cases such as (7) and (10) above, where we cannot elide the entire nominal expression when a bare plural is the antecedent, is related to the general constraints on null complement anaphora in Spanish. That is, the above cases of nominal elision are the result of a constraint that forbids null complement anaphora in Spanish in norestructuring Vs contexts: (12) a. *Juan quiere a Marı´ a y Pepe tambie´n quiere [e] Juan wants to Marı´ a and Pepe too wants e ‘*Juan loves Mary and Pepe also loves’ b. *Juan espera a Marı´ a y Pepe tambie´n espera [e] Juan waits-for to Marı´ a, and Pepe too waits-for e ‘*Juan waits for Mary and Pepe also waits for’ To summarize the previous discussion, there is no need for the presence of a D phonetically realized in the cases of nominal ellipsis in Spanish. Two more descriptive generalizations are required at this point. First, as the examples in (13) illustrate, unstressed possessive pronouns cannot appear in Spanish nominal elision contexts. (13) a. Compramos tus libros de lingu¨ı´ stica pero no mis libros de astronomı´ a (we) bought your books of linguistics but not my books of astronomy ‘We bought your linguistics books but not my astronomy books’ b. *Compramos tus libros de lingu¨ı´ stica pero no mis [e] de astronomı´ a (we) bought your books of linguistics but not my [e] of astronomy c. Compramos tus libros de lingu¨ı´ stica pero no los [e] mı´ os (we) bought your books of linguistics but not the [e] mine de astronomı´ a of astronomy The ungrammaticality of (13b) with the unstressed possessive pronoun mi (my) and the grammaticality of the same example with the stressed possessive

5.2 Nominal Ellipsis

171

pronoun mı´o (mine) illustrate that unstressed possessives cannot occur in nominal elision contexts. Note that an explanation of this restriction based on the unstressed character of the pronoun cannot account for facts such as the ones in (14): (14) a. *Compramos tus cuatro libros de lingu¨ı´ stica pero no mis tres [e] (we) bought your four books of linguistics but not my three [e] de astronomı´ a of astronomy b. *Compramos tus horribles libros de lingu¨ı´ stica pero no mis (we) bought your horrible books of linguistics but not my adorados [e] de astronomı´ a lovely [e] of astronomy The examples in (14) show a stressed element between the unstressed possessive pronoun and the ellipsis site. Therefore, the ungrammaticality of the sentences above suggests that it is not the lack of a stressed element by the unstressed possessive pronoun that causes the impossibility of unstressed possessive pronouns and ellipsis contexts. Although more needs to be done in this issue, I tentatively suggest that the special derived character of unstressed possessive pronouns (i.e., they are the result of a movement) is the key to achieving an explanation for these facts.8 Finally, there is one restriction on the relation between the antecedent and the omitted element: the antecedent and the deleted N may differ in number but not in gender.9 (15) a. Juan llamo´ a la hermanaf,sg de Ana y Pepe llamo´ al hermanom, sg Juan phoned to the sister of Ana and Pepe phoned to-the brother a’. *Juan llamo´ a la hermanaf,sg de Ana y Pepe llamo´ al [e] Juan phoned to the sister of Ana and Pepe phoned to-the [e] de Miguel of Miguel b. Juan llamo´ a las hermanasf,pl de Ana y Pepe llamo´ a la hermanaf,sg Juan phoned to the sisters of Ana and Pepe phoned to the sister de Miguel of Miguel b’. Juan llamo´ a las hermanasf,pl de Ana y Pepe llamo´ a la [e] Juan phoned to the sisters of Ana and Pepe phoned to the [e] de Miguel of Miguel To sum up: this section has introduced the major contexts in which nominal ellipsis can appear in Spanish. More specifically, I have shown that nominal ellipsis is possible with any D except in certain cases with the definite article and unstressed possessive pronouns and that partial nominal ellipsis can leave any PP as a remnant element.

172

5 Nominal Ellipsis

5.2.2 Classifying the Data As seen briefly in the introduction to this chapter, the literature on nominal ellipsis has debated extensively the structural nature of the element omitted in nominal ellipsis constructions. This section explores the properties of nominal elision in Spanish and compares them to the properties of two well-known phenomena, VP ellipsis and V-Gapping. The study of the properties of nominal elision in Spanish helps us to classify the relevant constructions as nominal ellipsis cases, where we elide a maximal projection. The phenomena described in Section 5.2.1 have been considered as cases of NP ellipsis (or N’-Deletion) or N’-Gapping. Thus, examples such as (16a) are referred to as NP ellipsis by Bernstein (1993) and Valois (1991) and as N’Deletion by Jackendoff (1971), while examples such as (16b) are considered N’Gapping by Jackendoff (1971) and Valois (1991). (16) a. Tina saw Jack’s pictures and Nellie saw Dan’s [e] b. Jack liked Mary’s picture of Paris, and Bill liked Mona’s [e] of London Focusing now on the Romance literature, there have been some analyses based on the idea that the nominal ellipsis data in Spanish are cases of elision of the N. This is the type of proposal put forward in Brucart and Gra`cia’s (1986) analysis of nominal elision in Spanish and Italian. They propose that the N position is occupied by an empty category in the cases of elision of the N. According to these authors, the existence of a non-phonetically realized N is confirmed by the contrasts below: (17) a. Dammi lo sgabello, quello [e] alto give-me the stool, that [e] tall b. *Dammi lo sgabello, quell’alto (18) a. Quell’alto monte that tall mountain b. *Quello alto monte The contrast in grammaticality between (17a) and (17b) is due to the impossibility of contracting the determiner quello (that) and the adjective alto(tall) in Italian in contexts where the N has no phonological content. This contraction is typically obligatory in Italian, as the ungrammaticality of (18b) shows. In other words, in (17b) we cannot omit the vowel because there is an empty category that prevents it. Brucart and Gra`cia (1986) argue that the empty category that occupies the N position in nominal elision contexts is not pro but PRO.10 It is not clear how this type of analysis would account for cases such as (19), where the elements elided contain not only the N but also its modifiers: (19) Compramos muchos libros de fı´ sica y tu´ compraste algunos [e] (we)bought many books of physics and you bought some [e]

5.2 Nominal Ellipsis

173

Under this approach, PRO11 should stand for N in some cases and for NP or parts of NP in others. This is not a desired result, since it will force us to propose different types of PRO depending on the context and the overall analysis of nominal ellipsis in Spanish will not be able to predict the desired constraints of the construction since it will not be empirically adequate. Let us return now to the claim that phrasal ellipsis and gapping have different syntactic properties (cf. Lobeck (1995), Kim (1998), Lasnik (1999), among many others). As mentioned previously, this section explores the properties of nominal elision in Spanish and compares them to the properties of two well-known phenomena, VP ellipsis and V-gapping. The study of the properties of elision in Spanish nominal expressions helps to classify this elision process as nominal ellipsis, where we elide a maximal projection.12 The examples in (20) illustrate different types of elision processes studied in the literature that are relevant here: (20) a. Marı´ a would not eat a hamburger, but Annie would ø b. Anne read Hamlet and Marı´ a ø War and Peace

VP deletion V-Gapping

(20a) shows a case of VP ellipsis, where the ø is understood as eat a hamburger, a VP. (20b) illustrates an example of gapping, where only the head of the VP, read in the example, seems to be elided. A priori, it seems that to classify an elision process as ellipsis or gapping, the factor to take into account is the type of constituent (if any) that is left behind after the elision process has taken place. In line with this, the ellipsis process should affect an entire phrase, while gapping should affect the head of a phrase. If this is the case, examples of Spanish nominal elision such as (21a) should be classified as ellipsis, whereas data such as (21b) should be cases of gapping: (21) a. Compramos varios libros de fı´ sicaobj de Anaposs, pero no (we)bought several books of physics of Ana but Neg compramos muchos [e]. (we)bought many [e] b. Compramos varios libros de fı´ sicaobj de Anaposs, pero no compramos (we)bought several books of physics of Ana, but Neg (we)bought muchos [e] de Quı´ micaobj de Marı´ aposs. many [e] of Chemistry of Mary Nevertheless, the difference between ellipsis and gapping is deeper, since these phenomena differ regarding a number of syntactic properties. In what follows, I compare the properties of Spanish nominal elision with the properties of VP ellipsis and V-gapping. According to many authors (Jackendoff 1971, Lobeck 1995, Kim 1998, among others), gapping is only possible in coordinations involving and and or, as (22) shows. In contrast, VP ellipsis can apply to subordinate or coordinate clauses of all varieties, as (23) shown by:

174

5 Nominal Ellipsis

(22) a. b. c. d.

Jane likes cats, and Suba (likes) dogs Mary put out the trash, or Debora (put out) the recycling bin *Some ate apples today, because others (ate apples) yesterday *Some ate apples, because others (ate) bananas

(23) a. b. c. d.

Jane could like cats, and Suba might [VP] too Mary will put out the trash, or Debora might [VP] Some will eat apples today, because others had [VP] yesterday Some will eat apples today, because others might [VP] yesterday

The examples below show that we do not find these contrasts in grammaticality across nominal elision data in Spanish: (24) a. Juan compro´ varias fotos de Marı´ apossde Parı´ sPPA y Pepe compro´ Juan bought several pictures of Marı´ a of Paris and Pepe bought algunas [e] some [e] b. Juan compro´ varias fotos de Marı´ a de Parı´ s o vendio´ algunas [e] Juan bought several pictures of Marı´ a of Paris or bought some [e] c. Juan compro´ varias fotos de Marı´ a de Paris porque Pepe compro´ Juan bought several pictures of Marı´ a of Paris because Pepe bought algunas [e] some [e] (25) a. Juan compro´ la foto de Marı´ aposs de Parı´ sPPA y Pepe compro´ Juan bought the picture of Marı´ a of Paris and Pepe bought la [e] de Monaposs de Londres the [e] of Mona of London b. Juan compro´ la foto de Marı´ a de Parı´ s o la [e] de Mona de Londres Juan bought the picture of Marı´ a of P. or the [e] of Mona of London c. Juan compro´ la foto de Marı´ a de Paris porque Pepe compro´ la [e] Juan bought the picture of Marı´ a of Paris because Pepe bought the [e] de Mona de Londres of Mona of London As the previous examples show, the two major nominal elision processes in Spanish nominal expressions are possible both in subordinate and coordinate clauses. A second difference between gapping and VP ellipsis pointed out in the literature (cf. Lobeck 1995, Kim 1998, among others) is that gapping is not possible if the gap is not in the same immediate conjunct as its antecedent: (26) a. Jack likes Mary, and Bob said that Bill does [e] too b. *Jack likes Mary, and Bob said that Bill [e] Mona The examples in (27) show that elision processes in Spanish nominal expressions do not follow the same pattern: (27) a. Juan compro´ varias fotos de Marı´ a de Paris, y Ana dijo que Juan bought several pictures of Marı´ a of Paris, and Ana said that

5.2 Nominal Ellipsis

175

Pepe compro´ algunas [e] Pepe bought some [e] a’. Juan compro´ la foto de Marı´ a de Paris, y Ana dijo que Juan bought the picture of Marı´ a of Paris, and Ana said that Pepe compro´ la [e] de Mona de Londres Pepe bought the [e] of Mona of London b. Nosotros compramos muchos libros de fı´ sicaobjde Marı´ aposs y Ana We bought several books of physics of Marı´ a and Ana no sabe por que´ tu´ compraste algunos [e] Neg knows why you bought some [e] b’. Nosotros compramos el libro de fı´ sicaobj de Marı´ aposs y Ana no We bought the book of physics of Marı´ a and Ana Neg sabe por que´ tu´ compraste el [e] de quı´ micaobj de Juanposs knows why you bought the [e] of chemistry of Juan c. Nosotros vimos algunas invasiones de Irakobjde los USag We saw some invasions of Iraq of the US y Ana conoce a un hombre que vio muchas [e] and Ana knows to a man that saw several [e] ´ de Irakobj de los USag y Ana conoce c’. Vimos la invasion (we)saw the invasion of Iraq of the US and Ana knows a un hombre que vio la [e] de Chechenia de Rusia to a man that saw the [e] of Chechnya of Russia Furthermore, as Lobeck (1995) notes, an ellipsis site can precede its antecedent under certain conditions, while a gap cannot: (28) a. Sue didn’t, but John ate meat. b. *Sue the lamb, and John had the salmon. Once more, the nominal elision data pattern with VP ellipsis: (29) a. Porque Juan compro´ [NPaque´llos [e]], Pepe compro´ estos libros because Juan bought those [e], Pepe bought these books b. Porque Juan compro´ aque´llos [e] con pastas azules, Pepe compro´ because Juan bought those [e] with covers blue, Pepe bought estos libros con pastas rojas these books with covers red The grammaticality of the examples in (29) illustrates that both types of nominal elision allow the omitted elements to appear before the antecedent. Finally, Williams (1977) proposes that ellipsis but not gapping, can apply across utterance boundaries, as shown in (30) and (31): (30) A: John likes fish B: *Yes, and Mary [e] meat (31) A: John caught a big fish. B: Yes, but Mary didn’t [e].

Gapping Ellipsis

176

5 Nominal Ellipsis

The grammaticality of the examples in (32) shows that both types of nominal elision can apply across utterance boundaries in Spanish. (32) A: Juan ha escrito un libro largo Juan has written a book long B: Sı´ , pero Marı´ a ha escrito varios [e] yes, but Mary has written several [e] B’: Sı´ , pero el [e] de fı´ sica de Marı´ a es ma´s largo yes, but the [e] of physics of Marı´ a is more long ‘Yes, but the one on physics by Mary is longer’ To conclude this section: a review of the basic properties of the two basic nominal elision phenomena leads to the conclusion that all the cases of nominal elision in Spanish illustrated in Section 5.2.1 are instances of a general phenomenon of ellipsis.13 In other words, although it superficially seems that some of the data illustrated in Section 5.2.1 are the result of gapping a part of a phrasal category, the properties of these data lead us to classify them as cases of ellipsis of maximal projections.

5.3 Putting Together the Analysis Keeping in mind the conclusions of previous discussions, this section proposes an analysis that accounts for the nominal ellipsis data presented in Section 5.2. This section provides a general analysis for all the ellipsis data seen so far and derives the unusual behavior of ellipsis with nominal expressions headed by the definite article in Spanish from the particular morphophonetic properties of the definite article in Spanish. The analysis for the nominal ellipsis data discussed above is based on the notion that ellipsis is a PF phenomenon that deletes redundant phonetic and phonological material and derives a full structure in place of the ellipsis site. The section is organized as follows. Section 5.3.1 deals with general ellipsis data in Spanish nominal expressions, that is, with nominal expressions headed by Ds other than the definite article. Section 5.3.2 analyzes the behavior of Spanish definite nominal expressions in ellipsis contexts. Additional evidence for the analysis of NP ellipsis just presented coming, for instance, from the behavior of partial cliticization is introduced in Section 5.3.3.

5.3.1 General NP Ellipsis This book explores a structure for Spanish nominal expressions that follows the hypothesis of a parallelism between clauses and nominal expressions. This structure (see Chapter 2) assumed a tripartite division of the nominal domain, with a few functional projections to ensure the existence of that division; two possible generation sites for Ds (i.e, Agr or D, depending their strength,

5.3 Putting Together the Analysis

177

see Chapter 3); and restricted locality and anti-locality conditions. The resulting structure is represented in (33). (33)

The combination of the assumptions above enabled me to account for a number of syntactic properties of various elements within nominal expressions so far. More concretely, general word order data, extraction data and modification data have been accounted for under this basic analysis. In this section, I show that this structure is also able to accommodate the nominal ellipsis data presented so far. The analysis I give here claims that the peculiarity of ellipsis contexts is that part of the syntactic structure created during the derivation (i.e., the elements that occupy the ellipsis site) does not have phonological content due to the application of a deletion operation of maximal categories at the PF interface. In other words, I follow here the hypothesis that ellipsis is derived by deleting elements at PF, following Chomsky and Lasnik (1993), Sag (1976), Lasnik (1999), among many others.14 In fact, I assume that this is the only successful account for nominal ellipsis in Spanish since the nominal ellipsis facts in this language point to the need to perform some operations on the syntactic structure occupying the ellipsis site (for instance, some remnant elements must move outside of their maximal projections) prior to the application of the ellipsis operation at PF.

178

5 Nominal Ellipsis

In addition to the previous ones, another assumption of my analysis is that ellipsis processes affect only maximal projections. Section 5.2.2 introduced abundant empirical evidence in support of this hypothesis. More specifically, I will assume that ellipsis processes can affect the lower segment of a segmental maximal projection. Regarding this point, consider the VP ellipsis examples in (34): (34) a. Maria would not eat a hamburger in the classroom, but she would [e] in the cafeteria b. Maria would not completely eat a hamburger, but she would [e] partially The grammatical status of the examples in (34) shows that it is possible to have VP ellipsis leaving stranded a VP-adjunct, such as in the cafeteria and partially. This shows that VP ellipsis can target the lower segment of the VP in the following schematic structure: (35) (. . .) TP T

VP VP

eat a hamburger

Adjunct Ellipsis site

Assuming that ellipsis in nominal phrases operates in a similar fashion, it can target the lower segment of a maximal projection.15 For instance, the ellipsis site affected by NP ellipsis can be represented as shown below: (36) (. . .)

5.3 Putting Together the Analysis

179

That is, only the elements that are base-generated outside of the lower NP node can be left stranded after NP ellipsis has applied. Therefore, we will expect to observe adjuncts, postnominal adjectives, agents, possessors and relational adjectives as possible remnants, while elements such as prenominal adjectives and objects/PPA (cf. Chapters 2 and 4 for detailed descriptions of each type of elements) should not be observed as remnants. The first part of this prediction is borne out by the data in (37): (37) a. Compramos bastantes libros para regalo (we)bought several books to gift y uno [e] [para consulta]adjunct and a [e] to consult b. Compramos varios libros azules y uno [e] rojo (we)bought several books blue and a [e] red c. Compramos varios libros de Luis y uno [e] [de Marı´ a]poss (we)bought several books of Luis and a [e] of Marı´ a d. Compramos varios libros de Cervantes y uno [e] [de Borges]ag (we)bought several books of Cervantes and a [e] of Borges e. Compramos varias novelas policiacas y una [e] roma´ntica (we)bought several books police and a [e] romantic (37) illustrates that adjuncts, postnominal adjectives, agents, possessors and relational adjectives can be remnants of NP ellipsis. Recall that one of the most notorious differences between prenominal and postnominal adjectives is that only postnominal adjectives allow NP ellipsis in Spanish. Thus, our prediction that prenominal adjectives cannot appear as remnant elements is borne out: (38) *Ayer vi a la verdadera terrorista y a la supuesta [e] yesterday (I)saw to the true terrorist and to the alleged [e] ‘Yesterday, I saw the true terrorist and the alleged one’ To this point, the above analysis provides a simple explanation for the contrasts in grammaticality observed in nominal expressions with adjectives. Assuming that NP ellipsis targets NPs, only elements that are base-generated outside of the NP node can be left stranded after ellipsis has applied. Thus, as prenominal adjectives are specifiers of NP, they are contained in the ellipsis site and must undergo NP ellipsis. In contrast, none of the postnominal adjectives are affected by NP ellipsis: postnominal adjectives (Q-As) are generated as adjuncts to NP and relational adjectives are generated in the specifier of nP. None of these positions are contained in the ellipsis site. Hence, postnominal adjectives are expected to be able to survive NP ellipsis. Nevertheless, the analysis does not account for data such as (39), where elements such as de fı´sica (of physics) and sin fotos (without pictures) are left stranded after the deletion process at PF:

180

5 Nominal Ellipsis

(39) a. Compramos varios libros de matema´ticas y alguno [e] [de fı´ sica]obj (we)bought several books of math and some [e] of physics b. Compramos varios libros con fotos (we)bought several books with pictures y alguno [e] [sin fotos]PPA and some [e] without pictures The grammaticality of the examples in (39) illustrates the possibility of having objects and PPA as remnant elements. To account for these data, I propose that the linear position in which we find objects and PPA in (39) is not their basegenerated position. In other words, I follow previous analyses in the literature on ellipsis that postulate that remnant elements in ellipsis processes are the result of the existence of a previous movement of the remnant elements. One of these analyses is described in Jayaseelan (1990). This author analyzes pseudogapping as VP ellipsis with a prior movement of the remnant out of the VP to be elided. A schematic representation of this analysis is shown in (40): (40) Bill ate the hamburger and [Peter [VP[VPdid t1 ] [the hotdog]1] ] Jayaseelan (1990) claims that this prior movement is performed by a heavy NP shift operation, which affects the remnant element, the hotdog and allows it to survive the elision process. Lasnik (1999) proposes a similar alternative analysis of these data. In particular, Lasnik (1999) argues that the data should be analyzed as VP deletion in the PF component, preceded by overt syntactic raising of the object to the spec of AgrO. Another analysis that follows the same spirit is the one given in Kim (1998). Kim (1998) analyzes pseudogapping in Korean and Japanese as the result of VP ellipsis. The appearance of multiple remnants in Korean and Japanese is due to the presence of a focus phrase projection above TP in these languages to which the remnants move before VP ellipsis applies. We return now to the nominal ellipsis data, repeated in (41) for convenience: (41) a. Compramos varios libros de matema´ticas y alguno [e] [de fı´ sica]obj (we)bought several books of math and some [e] of physics b. Compramos varios libros con fotos (we)bought several books with pictures y alguno [e] [sin fotos]PPA and some [e] without pictures The line of analysis discussed above can be implemented for these data by assuming that objects and PPAs move before the NP ellipsis operation targets the lower segment of NP at PF. However, as extensively discussed in Chapters 2 and 3, overt movement of objects and PPAs is severely restricted in Spanish nominal expressions. The descriptive generalization, stated in Chapter 2, is that the presence of possessors and agents in the structure blocks overt movement of objects and PPAs. Therefore, an analysis based on overt movement of objects and PPAs will predict the

5.3 Putting Together the Analysis

181

impossibility of having an object or a PPAappearing with an agent or a possessor as the remnants of a nominal ellipsis operation. That is, this type of analysis will predict the data in (42) to be ungrammatical: (42) a. Compramos el libro de fı´ sicaobj de Marı´ aposs y tu´ compraste el [e] (we)bought the book of physics of Marı´ a and you bought the [e] de quı´ micaobj de Juanposs of chemistry of Juan b. Compramos el libro de fı´ sicaobj de Marı´ aposs y tu´ compraste el [e] (we)bought the book of physics of Marı´ a and you bought the [e] de Juanposs de quı´ micaobj of Juan of chemistry ´ de Irakobj de los USag y tu´ viste c. Vimos la invasion (we)saw the invasion of Iraq of the US and you saw la [e] de Checheniaobj de Rusiaag the [e] of Chechnya of Russia ´ de Irakobj de los USag y tu´ viste d. Vimos la invasion (we)saw the invasion of Iraq of the US and you saw la [e] de Rusiaag de Checheniaobj the [e] of Russia of Chechnya e. Compramos la novela de aventurasobj de Cervantesag de Marı´ aposs (we)bought the novel of adventures of Cervantes of Marı´ a ´ obj de Asimovag de Juanposs y tu´ compraste la [e] de ciencia ficcion and you bought the [e] of science fiction of Asimov of Juan f. Compramos la novela de aventurasobj de Cervantesag de Marı´ aposs (we)bought the novel of adventures of Cervantes of Marı´ a ´ obj de Juanposs y tu´ compraste la [e] de Asimovag de ciencia ficcion and you bought the [e] of Asimov of science fiction of Juan g. Compramos lanovela de aventurasobj de Cervantesag de Marı´ aposs (we)bought the novel of adventures of Cervantes of Marı´ a ´ obj y tu´ compraste la [e] de Asimovag de Juanposs de ciencia ficcion and you bought the [e] of Asimov of Juan of science fiction The grammaticality of the examples in (42) shows that the predictions of that line of analysis are not fulfilled. (42a–b) show that the presence of a possessor does not block the hypothetical movement of the object outside of NP; similarly, (42c–d) illustrate that the presence of an agent does not induce ungrammaticality; and, finally, (42e–g) show that the presence of all argumental PPs is also allowed. Note that any analysis that attempts to account for NP ellipsis taking into account the extraction data from Chapter 2 would face the same problem: the NP ellipsis data seem to force us to remove the object from the ellipsis site while the overt movement of the object cannot be done when a higher element (i.e., a possessor or an agent) is present. To resolve this problem, I propose that the movement that removes the object from the NP ellipsis site is not done in the overt syntactic component of the derivation. In connection with this claim, consider the following data:

182

5 Nominal Ellipsis

(43) a. El libro de fı´ sicaobj de Anaposs the book of physics of Ana b. El libro de Anaposs de fı´ sicaobj the book of Ana of physics Remember that one of the basic properties of Spanish nominal expressions is that argumental PPs can appear in any order to the right of the N. As discussed in Chapter 2, this free order can be due to the effect of a stylistic operation, which applies after all the syntactic operations have been applied to the derivation, i.e., in the PF interface. This last point explains the fact that the different linear positions of argumental PPs do not affect their c-command possibilities, as the data in (44), discussed in Chapter 2, illustrate: ´ [de sı´ mismoi]Theme [de Juani]Agent/Poss (44) a. La descripcion the description of himself of Juan ‘Juan’s description of himself’ ´ [de Juani]Agent/Poss [de sı´ mismoi]Theme b La descripcion the description of Juan of himself ‘Juan’s description of himself’ Accepting the previous remarks on the linear order of arguments within nominal expressions as correct, the proposal is that the stylistic rule that alters the linear order of arguments applies before NP ellipsis. Note that I assume that stylistic rules do not follow the locality constraints stated for overt syntax. This way, the stylistic rule can place objects and PPAs outside of the ellipsis site in the constructions under consideration, probably as adjuncts of some maximal category contained in the nominal expression.16 The derivation for the relevant examples (repeated in (45a)) is given in (45b): (45) a. varios [e] de fı´ sicaobj de Juanposs several [e] of physics of Juan b. (. . .)

5.3 Putting Together the Analysis

183

The derivation in (45) shows that the stylistic rule places the object outside of the ellipsis site, possibly as an adjunct to the maximal category NP. After that, there is a PF deletion of the lower segment of the maximal projection NP. To account for the possibility of the linear order in (46a), where the agent/ possessor linearly precedes the object, I assume that the stylistic rule can attach arguments to any maximal category within the nominal expression. Therefore, the linear order in (46a) is derived as shown in (46b): (46) a. varios [e] de Juanposs de fı´ sicaobj several [e] of Juan of Physics

Therefore, the combination of the stylistic rule and the hypothesis that ellipsis processes take place at PF give us an account for the data where we have more than one argumental PP surviving NP ellipsis. To conclude: the analysis presented in this subsection accounts for all the nominal ellipsis data discussed in Section 5.2 without resorting to any ad hoc operation. The analysis detailed above follows a well-grounded line of research that treats ellipsis as a PF operation, a structure of nominal expressions supported empirically in previous chapters and makes use of the independently needed stylistic movements in Spanish nominal expressions to accommodate all nominal ellipsis data. Then, the proposal put forward in this chapter provides strong empirical evidence for treating ellipsis as a PF operation that targets maximal categories and that can apply after all syntactic and stylistic operations have taken place. Note that the current analysis focuses on the inner part of the proposed structure since it is the one affected by the NP ellipsis operation. Thus, it does not predict the existence of any restriction in nominal ellipsis distribution apart from the one derived from the base-generated position of modifiers within the nominal expression. As a result, some of the restrictions on NP ellipsis in nominal expressions with the definite article need to find an additional explanation. The next section deals with these issues.

184

5 Nominal Ellipsis

5.3.2 On the Restrictions of the Definite Article in NP Ellipsis Contexts Section 5.2 presented some restrictions on the occurrence of the definite article in nominal elision contexts in Spanish. More specifically, we saw that a nominal expression headed by the definite article can undergo nominal elision only if the P de (of) introduces the remnant element. The basic data to explain are repeated below: (47) a. *La [e] sin gafas the [e] without glasses b. La [e] de gafas the [e] of glasses c. Las tres [e] sin gafas the three [e] without glasses

[e] = chica [e] = girl [e] = chica [e] = girl [e] = chicas [e] = girls

As (47) shows, NP ellipsis in nominal expressions with the definite article is ungrammatical17 whenever the only remnant is not a PP headed by de (of). In this section, I suggest that, although ellipsis in nominal expressions with the definite article follows the general pattern of nominal ellipsis in Spanish and its analysis, the clitic status of the definite article causes it to exhibit some differences with respect to other Ds in nominal ellipsis contexts. The main idea to explore is that the presence of full Ps prevents the appearance of the definite article in the relevant structure. There have been a few proposals that aimed to explain the properties of the definite article with respect to NP ellipsis.18 This way, in order to analyze the peculiar behavior that el [NP] constructions have in Spanish, Brucart and Gra`cia (1986) follow the idea of Vanelli (1979), who claims that the definite article must be cliticized to an N or an adjective. Brucart and Gra`cia’s (1986) generalization is that the Spanish definite article must cliticize to a category that is [+N]. The structures proposed are illustrated in (48): (48)

Thus, these authors explain the fact that nominal expressions containing the definite article can appear in elliptical nominals with adjectives and certain PPs by assuming that adjectives are [+V, +N] and that the P de (of) is a false P in

5.3 Putting Together the Analysis

185

Spanish and the category that it introduces is [+N].19 Similarly, Raposo (1999) has noted the clitic-like character that the definite article exhibits in these contexts. In this subsection, I make use of these proposals to complete my analysis of nominal ellipsis in Spanish. The major assumptions of my analysis of definite NP ellipsis follow the spirit of Raposo’s (1999) analysis of Spanish and Portuguese nominal ellipsis.20 Following Raposo (1999), I assume that, due to its unstressed character, the definite article must attach to another element in PF. In other words, I assume that the definite article must cliticize to the element to its right. Along the line of Raposo (1999), I also assume a division of PPs, namely, full PPs headed by Ps such as sin (without), which constitute phases by themselves and false PPs, headed by the P de(of), which do not constitute phases. Note that, in Chapter 2, it was shown that de (of) is a dummy P in Spanish by taking into consideration some syntactic phenomena, such as binding. Here I take those data as additional empirical evidence for the lack of phase in false PPs and propose that de (of) is inserted in PF as a case marker. I depart from Raposo (1999) in the reason why the cliticization of the definite article cannot be satisfied with full PPs: I assume that cliticization processes cannot operate through the boundaries of phases because phases are spell out units, which are coherent with the general properties of cliticization processes (cf. Bosˇ kovic´ (2001) on this issue). The derivation of the data *La [e] sin gafas (The [e] without glasses, where [e] = girl) and La [e] de gafas (The [e] of glasses, where [e] = girl) is as given below (where the symbol ‘||’ stands for phase boundary): (49) a. [DPLa [[NPchica] || [PPsin gafas]]] b. [DPLa [[NPe]] || [PPsin gafas]]] c. *[DPLa [[NPe]] || [PPsin gafas]]]

Spell Out21 NP ellipsis Cliticization

(50) a. b. c. d.

Spell Out NP ellipsis DE(of)-insertion Cliticization

[DP La [[NPchica] [NPgafas]]] [DP La [[NPe] [NPgafas]]] [DP La [[NPe] DE-[NPgafas]]] [DP La[[NPe] + DE-[NPgafas]]]

The ungrammaticality of La [e] sin gafas (the [e] without glasses) is derived from the fact that the cliticization process that applied from (49b–49c) cannot attach the definite article to any element in its phase and cannot operate through the boundaries of the PP phase to cliticize the definite article to the P sin (without). In contrast, the grammaticality of la [e] de gafas (the [e] of glasses) is explained since the cliticization requirement of the definite article can be satisfied within its phase.22 This type of analysis can also explain the contrasts in the following data: (51) a. El libro de fı´ sica con pastas azules the book of physics with covers blue b. El libro con pastas azules de fı´ sica

186

5 Nominal Ellipsis

(52) a. El [e] de fı´ sica con pastas azules b. *El [e] con pastas azules de fı´ sica Under my analysis, the contrast in grammaticality between (51a) and (52b) is due to the fact that the cliticization process applies after the stylistic and NP ellipsis rules apply. Therefore, the article cannot attach into the phase headed by the P con (with) but it can be cliticized within its own phase to the false P de (of), in (52a). To sum up: this section has developed an analysis for the ungrammaticality of NP ellipsis with nominal expressions headed by definite article. I have assumed that these cases follow the general pattern of nominal ellipsis in Spanish and that it is the clitic-like status of the definite article that causes the ungrammaticality. More concretely, the presence of a phase boundary when a full PP (i.e., a PP that is not headed by the P de (of)) follows the definite article in nominal ellipsis contexts prevents the cliticization process of the definite article and causes the ungrammaticality of the relevant structure.

5.3.3 Additional Evidence Section 5.3.1 has presented an analysis of NP ellipsis in Spanish that assumes the existence of a syntactic structure in the ellipsis site that is deleted in the PF interface as the result of the ellipsis process. This section aims to contribute with some additional evidence for the analysis just described that comes from the behavior of partial cliticization in Spanish. As discussed in Chapter 2, cliticization of different parts of a nominal expression is possible in Spanish. The sentences in (53) illustrate different cases of this partial cliticization where we have left stranded different arguments of the N; namely, the agent and the object PP, respectively. (53) a. Compramos varias traducciones [de ese escritor]ag (we)bought several translations of this writer y tu´ las compraste [de aquel polı´ tico]ag and you CL bought of that politician b. He leı´ do varios capı´ tulos [de Don Quijote]obj (I)have read several chapters of Don Quijote y tu´ los has leı´ do [de La Celestina]obj and you CL have read of La Celestina The analysis of partial cliticization pursued in Chapter 2 considers partial cliticization a surface anaphora. This assumption is based on the following data: (54) Hemos comprado varios libros de quı´ mica. (we)have bought several books of chemistry ¿De que´ (piensas que) los ha comprado Ana? Of what (you)think that CL has bought Ana [NP]

5.3 Putting Together the Analysis

187

According to Depiante (2001), the possibility of extraction out of a null element is an indication that the omitted element in question is a surface anaphor; since, as assumed in Hankamer and Sag (1976), among many others, only surface anaphors have at some stage a representation with internal syntactic structure. In other words, the possibility of extraction out of the apparently null NP associated with lo, as in (54), illustrates that there is syntactic structure in the omitted element at some level of representation, which is an indication that the omitted element in partial cliticization is a surface anaphor. Surface anaphors can be treated as the result of either PF deletion or LFcopying. As noted in Bosˇ kovic´ (1994), surface anaphors that are not phonologically null are more amenable to the LF copying analysis. Certain data concerning partial cliticization confirm this. Under the LF copying analysis, movements that affect the LF copying relationship have to take place in the syntax; hence, they should be subject to locality conditions on syntactic movements. This is confirmed by (55). The ungrammaticality of data such as (55a’-a’’) and (55b’-b’’) leads me to explain these data as the result of a locality violation of the elements contained in the antecedent nominal expression: la compraste de Anaposs (55) a. Compramos una casa de Juanposs y tu´ (we)bought a house of Juan and you CL bought of Ana a’.*Compramos una casa de Juanposs con ventanas azulesPPA y tu´ la (we)bought a house of Juan with windows blue and you CL compraste de Anaposs con piscinaPPA bought of Ana with swimming pool a’’.*Compramos una casa con ventanas azulesPPA de Juanposs y tu´ la (we)bought a house with windows blue of Juan and you CL compraste con piscinaPPA de Anaposs bought with swimming pool of Ana b. Compramos varias traducciones [de ese escritor]ag y tu´ las compraste (we)bought several translations of this writer and you CL bought [de aquel polı´ tico]ag of that politician b’.*/?? Compramos varias traducciones [de Don Quijote]obj (we)bought several translations of Don Quijote [de ese escritor]ag y tu´ las compraste [de La Celestina]obj of this writer and you CL bought of La Celestina [de aquel polı´ tico]ag of that politician b’’.*/?? Compramos varias traducciones [de ese escritor]ag (we)bought several translations of this writer [de Don Quijote]obj y tu´ las compraste [de aquel polı´ tico]ag of Don Quijote and you CL bought of that politician [de La Celestina]obj of La Celestina

188

5 Nominal Ellipsis

Under the LF copying analysis, the ungrammaticality of the data in (55a’a’’) and (55b’-b’’) is the result of the overt syntactic movement of the object in the presence of a possessor or an agent in the nominal expression of the first conjunct (see Chapter 2). This overt syntactic movement is motivated by the need of obtaining a nominal expression before spell out that contains the same elements that the empty nominal expression associated with lo contains. Note that these data can also give us additional support for the analysis of nominal ellipsis presented in this chapter. Under my analysis of ellipsis, the impossibility of having overt syntactic movement of the object that removes an object from the NP in the presence of agents or possessors leads me to assume that the cases of remnant objects with nominal ellipsis can only be derived as instances of stylistic (i.e., PF) movement of objects. This stylistic movement of objects can precede the ellipsis process and derives the data, since ellipsis is a PF phenomenon. On the contrary, a process where an overt syntactic movement before PF is required, such as partial cliticization, which I argue involves LF copying, cannot benefit from any stylistic movement of objects. This reasoning derives straightforwardly the ungrammaticality of remnant objects that co-occur with a possessor or an agent in processes involving any type of syntactic movement, such as partial cliticization and the grammaticality of remnant objects with PF operations, such as ellipsis. The analysis presented here provides evidence that surface anaphors can arise through either PF deletion or LF copying. I tentatively suggest that overt surface anaphors require the latter analysis.23

5.3.4 Conclusions The main components of the analysis of NP ellipsis in Spanish detailed in the previous section are the following ones: first, the possibility to leave stranded an element after nominal ellipsis is the result of the different base-generation possibilities of the elements in the nominal expression; second, nominal ellipsis targets exclusively maximal projections; and, third, NP ellipsis is a phenomenon that takes place after the derivation has been shipped to the PF interface, after all syntactic and stylistic operations have taken place. This analysis has great empirical coverage and is consistent with the analysis proposed for the other properties of the Spanish nominal expressions in the book.

5.4 Agreement and Focus in Ellipsis Processes This section deals with the licensing and recoverability conditions for NP ellipsis by paying attention to the role that agreement and focus have in ellipsis

5.4 Agreement and Focus in Ellipsis Processes

189

processes. To do so, this section reviews the main claims made in some analyses proposed in the earlier literature on elision in nominal expressions and contrasts them with the descriptive generalizations reached in Section 5.2 and with my proposal in Section5. 3. I have divided the section into two parts, as follows. In Section 5.4.1 I explore the claim that the nominal ellipsis phenomena in Spanish should be analyzed as the result of its type of agreement and in Section 45..2 I will discuss the relation between focus and nominal ellipsis, paying special attention to some recent proposals that postulate to analyze nominal ellipsis as a complex operation containing a series of movement operations. Apart from introducing the notions of agreement and focus in our discussion of NP ellipsis, the goal of this section is also to provide a background to evaluate my proposal. The discussion in this section will give us the opportunity to see that earlier analyses have provided a solution for some particular cases mentioned in Section 5.2 and that, to my knowledge, there is no analysis that covers all the phenomena described in this chapter as a unitary phenomenon. Then, the lack of empirical coverage can be the first criticism for previous approaches and the first advantage of my proposal, as all previous analyses of elision in Spanish nominal expressions leave some data without explanation. Therefore, through this section, it will become evident that my proposal has greater empirical coverage than previous analysis, is consistent with the analysis proposed for the other properties of the Spanish nominal expressions in the book and is more elegant since no additional device is needed.

5.4.1 The Relevance of Strong Agreement in Nominal Ellipsis Traditionally, ellipsis phenomena have been linked to the presence of rich or strong morphology in a particular language (cf. Lobeck (1995), Torrego (1987), or Bernstein (1993), among many others). Consequently, there is a line of research, the most popular one, which focuses its analysis of ellipsis in languages such as Spanish in the properties of agreement in this language. There are different variants of this type of proposal. I will briefly review here Torrego’s (1987), Bernstein’s (1993) and Lobeck’s (1995) approaches to nominal ellipsis. According to Torrego’s (1987) analysis, the possibility of having nominal ellipsis is related to the null subject parameter. More specifically, Torrego (1987) links the possibility of having nominal ellipsis to the possibility of having pro and proposes the basic structure in (56): (56) a. [DP uno [NPpro]] b. [DPel [NPpro]]

190

5 Nominal Ellipsis

As for the behavior of the Spanish definite article, Torrego (1987) claims that the definite article in Spanish is able to retain the person, number and gender features necessary to license (via government) an NP pro. This type of approach cannot account for restrictions on NP ellipsis seen in previous sections, since any nominal expression in which the N has been substituted for pro and that carries a D expressing the person, number and gender features of the referent of pro, should be grammatical. However, as the contrasts below show, this is not the case: (57) a. Dame el libro verde y que´date el [e] azul give-me the book green and keep-you the [e] blue b. *Dame la verdaderafoto y que´date la supuesta e give-me the true picture and keep-you the alleged e c. Me quedo el e de pastas azules (e = libro) CL (I)keep the e of covers blue (e = book) d. *Me quedo el e con pastas azules (e = libro) CL (I)keep the e with covers blue (e = book) e. *Ana compro´ un libro con pastas azules y Pepe le pidio´ el e Ana bought a book with covers blue and Pepe CLdat asked-for the e The data in (57a–b) illustrate the different behavior of adjectives in nominal elision contexts. The examples in (57c-d) illustrate that some PPs are not possible in nominal elision if the D is the definite article. Finally, the last example (57e) shows that it is impossible to have the definite article in a nominal elision context if the elided element is not followed or preceded by a modifier. Note that all these contexts are predicted as grammatical under Torrego’s (1987) theory, since the presence of pro and the presence of a D that expresses gender, person and number features are enough to license nominal elision. Furthermore, the relation between nominal ellipsis and the null subject parameter is problematic for languages such as French, which has NP ellipsis but does not have null subjects. Finally, Torrego’s (1987) analysis cannot explain data where modifiers remain after the N is deleted. That is, if pro stands for an NP, an addition to this analysis should be postulated to account for the data in which PPs or adjectives are present. As we discussed earlier, this was one of the reasons why my analysis of nominal ellipsis assumed the existence of syntactic structure in the ellipsis site at some stages of the derivation. Bernstein (1993) builds on previous proposals’ insights with respect to the role of agreement for her analysis of nominal ellipsis in Romance. Focusing on Spanish, she divides the data she examines into two groups, depending on the determiner that heads the relevant nominal expression. As for nominal expressions headed by Ds other than the definite article, Bernstein (1993) proposes that these cases of nominal ellipsis receive their explanation in what she calls the UNO strategy.

5.4 Agreement and Focus in Ellipsis Processes

191

Crucially, Bernstein (1993) follows Harris’s (1991) idea that Ns can be inflected in syntax. Thus, according to Bernstein (1993), the underlying form of the indefinite article in Spanish is un (a). The –o affix is generated independently as the head of a functional projection (Word Marker Phrase, WMP) and substitutes into D as a result of head movement. This functional category (WMP) is placed between Number Phrase (NumP) and NP and licenses the null NP via head government in languages such as Spanish. The derivation of sentences such as (58a) is shown in (58b): (58)

Assuming this analysis, the indefinite article raises from the spec of NumP and adjoins to the left of the word marker in D. Bernstein (1993) accounts for the ungrammaticality of (59), where we have a lexical N and the un+o (a+word marker) version of the indefinite article in Spanish, as the result of having the word marker head being doubly filled at deep structure. (59) [*Quiero uno libro grande]. I-want a book big The ungrammaticality of (59) is due to the word marker –o of the lexical N and the word marker –o of the indefinite article occupying the head of word marker at the same time; since we would then obtain a category with two heads, the outcome would violate general X’ theory. This approach and Bernstein’s (1993) assumptions regarding adjective placement as adjuncts or heads (see Chapter 4) allow her to explain some of the restrictions on adjectives and NP ellipsis in Spanish discussed in previous sections. Concretely, prenominal adjectives that are generated as heads in Bernstein’s (1993) proposal would interfere with the WM movement to a higher position, causing NP ellipsis to be ungrammatical. In contrast, postnominal adjectives, generated as adjuncts in her analysis, do not interfere with WM movement and NP ellipsis is grammatical in these cases. The relevant derivations are illustrated in (60) and (61):

192

5 Nominal Ellipsis

(60)

(61)

The analysis should account for nominal ellipsis with any determiner except the definite article. This leads Bernstein (1993) to assume that the feminine version of the indefinite article una(a/an-fem) is ambiguous between un+a (where -a is a WM0 and licenses ellipsis) and un+a (where -a comes from the lexicon as a gender marker and does not license ellipsis). Although Bernstein’s (1993) analysis for NP ellipsis of nominal expressions not headed by the definite article has a greater empirical coverage than previous analyses, it still has to overcome a few problems.24 First, Bernstein’s (1993) analysis of non-definite NP ellipsis in Spanish is based on the existence in Spanish of an indefinite article that appears only in NP ellipsis contexts: uno (a/an). Her proposal is based on the assumption that the so called neuter indefinite article uno (a/an) is composed of un (bare form) and -o (word marker). This analysis raises some problems, as Bernstein (1993) acknowledges, when it is extended to other Spanish Ds. More precisely, this analysis cannot explain why (62a) is ungrammatical: (62) a. *Juan compro´ ese libro y Pedro compro´ esto [e] Juan bought that book and Pedro bought this-neuter [e]

5.4 Agreement and Focus in Ellipsis Processes

193

b. Juan compro´ ese libro y Pedro compro´ e´ste [e] Juan bought that book and Pedro bought this-masc [e] That is, the analysis cannot account for the fact that Spanish does not use the so-called neuter forms of other Ds, for instance, the neuter demonstrative esto (this) in (62a), in cases of non-definite NP ellipsis. More importantly, Bernstein’s (1993) analysis predicts no difference between non-restrictive and restrictive adjectives with respect to NP ellipsis: (63) a. Me gusta {el/aquel} e rojo CL like {the/that} e red (I like the red one ) b. *No me gusta el famoso libro, me gusta el desconocido e Neg CL like the famous book, CL like the unknown e (I don’t like the famous book; I like the unknown one) Under Bernstein’s (1993) analysis, in both cases the adjectives are adjoined to an XP25 and do not interfere with WM movement. However, as she explains in her description of adjectives, there are no cases of NP ellipsis with nonrestrictive adjectives. Finally, let us review Lobeck’s (1995) cross-linguistic analysis of NP ellipsis. This analysis is based on the idea that only strong agreement26 of D or Num can license NP ellipsis. Her structure, following the one proposed by Ritter (1991), is given in (64): (64)

Additionally, Lobeck (1995) postulates the existence of the following parameter: (65) The Ellipsis Identification Parameter (EIP): the number of strong agreement features in Det or Num that is required to identify an empty, pronominal NP is proportional to the number of possible strong agreement features in the agreement system of noun phrases in the language. This analysis allows Lobeck (1995) to account for the fact that German NP ellipsis is broader than English NP ellipsis. This is because German has a richer and more regular agreement (Agr) system than English. These properties of the German Agr allow it to identify empty pronominal NPs in many contexts. In contrast, a poor Agr system, such as the one exhibited by English, cannot generally identify empty pronominal NPs. Following this line of reasoning, Lobeck (1995) argues that in French NP ellipsis is more restricted than in German and English. French expresses more

194

5 Nominal Ellipsis

Agr features than English but its Agr system lacks the regularity of German. According to Lobeck (1995), it is precisely the lack of regularity in the presence of Agr features in French that causes French NP ellipsis to be more restricted than NP ellipsis in German and English. The extension of this analysis to the Spanish data shown above poses an important problem. Namely, the restrictions on NP ellipsis discussed above cannot be explained under this approach, as Spanish shows the same agreement in all the cases mentioned above. To summarize: the traditional link between ellipsis phenomena and the presence of rich or strong morphology in a particular language (cf. Lobeck 1995, Torrego 1987, or Bernstein 1993, among many others) produces undesirable predictions when applied cross-linguistically, since there are languages with weak morphology that allow ellipsis quite freely. Furthermore, this chapter has shown that the presence of agreement is not sufficient to derive the different restrictions on nominal ellipsis. Focusing on the Spanish data, the discussion in this chapter has illustrated that Ds exhibiting the same agreement features do not allow nominal ellipsis in the same structural situations. The restrictions on ellipsis in nominal constructions with the definite article cases illustrated this point previously. Moreover, inner structural factors, such as the type of adjective present in the nominal structure, affect the possibilities of nominal ellipsis, with independence of the agreement pattern established among the relevant elements. It seems that although it is necessary to assume some kind of link between agreement and ellipsis, mostly in the recovery information part of ellipsis, some other structural conditions prevail over this aspect. Recent research (cf. Johnson 2001, Ntelitheos 2003, 2004, Corver and van Koppen 2006, among others) has questioned the relation between agreement and ellipsis and has focused on the relation between the informational properties (contrastive/non contrastive information) of a particular constituent and its possibilities to be elided. This line of research has led to the analysis of ellipsis processes as complex operations, which involve several movements to the left periphery of a maximal projection and deletion operations that prevent us to pronounce multiple copies of an element. This type of analysis will be examined in the next section.

5.4.2 The Relevance of Focus in Nominal Ellipsis Since Rooth (1992), many authors (cf. Swarzschild 1999, Romero 1998, and Merchant 2001, among many others) have proposed that the crucial property of ellipsis processes is related to the fact that an element can be elided only if it is given in a context. Let us consider a typical ellipsis process in English, as illustrated in (66): (66) Gary visited Bill, and then the next day Mary did [e]

5.4 Agreement and Focus in Ellipsis Processes

195

[e] in (66) is understood as the elided VP visit Bill. As illustrated in (66), the elided element must have a particular salient antecedent in the context. Moreover, in cases where there are elements surviving the ellipsis processes (i.e, remnant elements), these elements convey a contrastive focus, as in (67a). (67) a. Bill ate the hamburger and Peter did [e] the hotdog b. *Bill ate the hamburger and Peter did [e] the hamburger

([e]=eat) ([e]=eat)

The grammaticality contrast illustrated in (67) is due to the differences with respect to the information structure of each sentence. In (67a) the nominal expression the hotdog is providing us with new information that contrasts with the relevant information in the first part of the sentence (that is, it is contrastively focused), while in the example (67b) the nominal expression the hamburger repeats the same information given in the first conjunct. Following approaches to contrastive focus (cf. Kiss 1998 for instance) that associate contrastive focus with derived positions, a line of recent research on ellipsis has suggested that ellipsis processes impose contrastive focus on the remnant (cf. Jayaseelan 2001, Lasnik 1999, Kennedy and Merchant 2000, Johnson 2001, among others). The syntactic properties of ellipsis processes thus provide compelling evidence that they involve overt movement of the remnant to a higher discourse oriented position and subsequent PF-deletion of the elided element. Following this type of approach, Ntelitheos (2003) proposes that nominal ellipsis is a complex process that involves three steps: first, the elided NP moves to a topic (i.e. non-focus) position in the periphery of the DP; second, there is a process of NP-deletion of the moved NP; and third, the remnant element moves to a focus projection in the DP periphery. The derivation in (68) illustrates this analysis. (68) a. o Giannis agorase tria vivlia kai o Petros agorase ena.vivlio The Giannis bought-#SG three books and the Petros bought one book ‘John bought three books and Petros bought one book’ b. [XP [TopicP [FocusP [TopicP [DefP ..[FP ena . . .[NP vivlio]]]]]]] c. [XP [TopicP [FocusP [TopicP [NP vivlio] [DefP ...[FP ena . . .tNP]]]]]] d. [XP [TopicP [FocusP [FP ena . . .tNP] [TopicP [NP vivlio] [DefP ...tFP ]]]]] (Ntelitheos (2003:2, (1)) There are several problems that prevent us from adopting this type of approach for the Spanish data in this chapter. On the one hand, an extension of Ntelitheos’s (2003) analysis to Spanish will have to deal with the extraction restrictions discussed in Chapter 2. Remember that any analysis of the basic extraction facts out of Spanish nominal expressions must maintain the thematic hierarchy in proper syntax to be able to account for the extraction possibilities discussed. The proposed movements in Ntelitheos’s (2003) analysis will not be compatible with this requirement regarding the thematic hierarchy. Therefore, an extension of this type of analysis would cause the loss of the uniform approach to ellipsis and extraction, which would imply that some of the most

196

5 Nominal Ellipsis

important properties of Spanish nominal expressions would not receive a uniform analysis. Following this problem of empirical coverage, note that there is nothing in Ntelitheos’s (2003) proposal that can explain the differences seen between nominal ellipsis with the definite article and a PP headed by the preposition de(of) versus nominal ellipsis with the definite article and a PP headed by other preposition. Then, there is abundant evidence against an approach to nominal ellipsis in Spanish based on overt syntactic movement. More concretely, this line of analysis will not be able to derive the ellipsis facts in Spanish and to predict major properties of nominal expressions in this language. Note that the alleged discursive properties of ellipsis (i.e., the claims that an element can be elided only if it is given in a context and that ellipsis processes impose contrastive focus on the remnant) do not automatically imply the existence of overt movement to satisfy these properties. In fact, the main argument (cf. Reinhart (2006) for a discussion on this) to postulate a focus feature triggering focus movement in the syntactic component of our grammar is based on the possibility of having a mismatch between focus stress and nuclear stress in certain sentences. Some authors, such as Cinque (1993), have examined this type of argument and concluded that discursive notions can assign an additional stress to the relevant constituent, or destress the original prominent stress of a sentence. Furthermore, returning to the nominal domain, note that, as Samek-Lodovici (2007) and Bernstein (2001a) have shown, Romance languages like Italian or Spanish can express focus structurally in DP-final position, thus closely paralleling the availability of rightmost focus in the clause attested by many studies since Zubizarreta (1998). Note that, under this approach, the element that receives nuclear stress matches the element that receives focus stress and there is no need to postulate explicit movements to derive these discursive notions in overt syntax. Therefore, assuming that the previous discussion is on the right track, there is no need to propose movement operations in syntax to derive the discursive notions involved in nominal ellipsis.

5.4.3 Conclusions The previous discussion has reviewed the role of agreement and focus on nominal ellipsis through the review of some proposals in the literature. First, we examined the importance of agreement on nominal ellipsis and concluded that an analysis based exclusively on this notion would not be able to account for all the data presented in Section 5.2. Second, we discussed the role of discursive notions such as focus in the analysis of nominal ellipsis. The conclusion was that the relation between ellipsis and information structure does not force us to derive this relation as having overt syntactic movements in the ellipsis data. In fact, the postulation of

Notes

197

overt syntactic movements to elucidate the ellipsis data found problems to explain some other major properties of Spanish nominal constructions. Therefore, my analysis of nominal ellipsis is superior because it is able to provide empirical coverage for all the phenomena described in this chapter as a unitary phenomenon, without resorting to additional assumptions.

5.5 Summary of the Chapter This chapter has introduced an analysis of nominal ellipsis that gives an account for abundant nominal ellipsis data in Spanish. The analysis is based on two main assumptions: first, NP ellipsis is PF-deletion; second, it applies after some stylistic operations have been applied. These two assumptions, combined with the clitic-like character of the definite article in Spanish, provide us with a simple explanation for data that have been problematic for previous analyses, as well as additional data that were not taken into consideration by previous analyses. Furthermore, the analysis defended in this chapter expands the empirical coverage of the main analysis argued in this study, providing us with additional empirical evidence for the hypothesis that ellipsis is a PF phenomenon and for the hypothesis that some surface anaphoras, such as partial cliticization, should be treated in terms of LF copying.

Notes 1

2

3

4

5

Throughout the descriptive part of the chapter, I use the terms nominal elision, elision in Spanish nominal expression, partial nominal elision and partial elision in Spanish nominal expressionas general terms covering ellipsis and gapping phenomena. In Section 5.2.2, I discuss the status of each of the nominal elision phenomena introduced in Section 5.2.1. This descriptive section focuses on Spanish nominal ellipsis. For a description of nominal ellipsis in some other languages, the reader can consult the references in the text. For instance, Brucart & Gra`cia (1986) for Italian nominal ellipsis, Lobeck (1995) for French, English and German nominal ellipsis properties, Ntelitheos (2003) for Greek, etc. Also, Alexiadou et al. (2007: 258-272) present a brief overview of (indefinite) nominal ellipsis in languages such as Spanish, French and Greek, among others. For ease of exposition, I use [e] as representation of the ellipsis site and mark in bold the antecedent of the ellipsis site in the first conjunct (that is, what would correspond to the deleted material in the second conjunct). As Howard Lasnik (p.c.) observes, English shows a similar pattern regarding the possibilities of nominal ellipsis in nominal expressions headed by the definite article. I will assume here that this ungrammaticality is due to a pragmatic or morphological constraint. Note that the fact that other languages show similar restrictions while having different phonological constraints casts doubts on analyses such as Bosque and Demonte’s (1999) proposal. According to the description of nominal elision in Bosque and Demonte (1999), the peculiar behavior of the definite article in Spanish is due to a phonological restriction. That is, the definite article in Spanish is unstressed and it must be supported by a

198

5 Nominal Ellipsis

stressed element. They support their view by the following data with some other unstressed elements in Spanish: (i) a. Mi discurso de ayer fue muy breve my speech of yesterday was very short b. {*Mi e de ayer /El e mı´ o de ayer} fue muy breve {my e of yesterday/ the e mine of yesterday} was very short

6

The ungrammaticality of (ib) with the unstressed possessive pronoun mi (my) and the grammaticality of the same example with the stressed possessive pronoun mı´o (mine) illustrate that unstressed possessives cannot occur in nominal elision contexts. I will show that unstressed possessive pronouns and definite articles do not share the same restrictions in Spanish nominal elision contexts in the descriptive part of this chapter. At this point, I will leave open the issue on the parallelism between the definite article and unstressed possessive pronouns and I will return to it once my description of nominal elision is completed. Note that, as (i) shows, this is not the behavior found with unstressed possessive pronouns: (i) a. *Compramos tus libros sobre Chomsky y tu´ compraste mis [e] (we)bought your books about Chomsky and you bought my [e] sobre Postal about Postal b. *Compramos tus libros de Chomsky y tu´ compraste mis [e] de Postal (we)bought your books of Chomsky and you bought my [e] of Postal

7

The contrast in the grammaticality between the relevant example in the text and (ib) shows that a uniform account for nominal expressions with definite articles and for nominal expressions with unstressed possessive pronouns cannot be adopted (cf. Bosque and Demonte (1999) for a proposal that postulates that the behavior of the definite article and unstressed possessive pronouns in nominal ellipsis contexts can be explained as the result of a phonological restriction.). See note 8 for a non uniform explanation to the data with definite articles and unstressed possessive pronouns in Spanish nominal elision contexts. In relation to this point, some authors (Sa´nchez (1996), among others) have noticed that nominal elision is not possible when it leaves stranded an appositional relative clause: (i) a. La mujer que sabı´ a la verdad lo ayudo´ y la [e] que no lo sabı´ a tambie´n the woman that knew the truth CL helped and the e that not CL knew too ‘The woman that knew the truth helped him and the one that didn’t know it did it too’ b. *La mujer, que sabı´ a la verdad, lo ayudo´ y la [e], que no lo sabı´ a, tambie´n the woman that knew the truth CL helped and the e that not CL knew too

8

9

As seen in Chapter 2, unstressed possessive pronouns can be derived as the result of the internal movement that unstressed possessive pronouns undergo to a specifier position. Assuming that this line of analysis is on the right track, the ungrammaticality of the examples with unstressed possessive pronouns should be linked to the interaction between that movement and the ellipsis process. In fact, it seems that the presence of number is required, as the ungrammaticality of examples (i) and (ii) show. (i) *Entrevistamos a cada amigo de Marı´ a pero no a cada [e] de Ana (ii) *Compramos todo libro de fı´ sica pero no todo [e] de quı´ mica The ungrammaticality of the examples above is due to the lack of number agreement in Ds such as cada (each) or todo (all), according to Ramalle (2005) for instance. to this respect).

Notes 10

199

They support their claim by the following two arguments. First, pro is considered to be a definite pronoun; however, PRO and the elided N can appear in contexts where the relevant interpretation is indefinite, (i): (i) Uno [e] que se habı´ a comido el pastel se sentı´ a mal a [e] that CL has eaten the cake CL felt bad Second, pro can be substituted by a lexical pronoun, whereas PRO and the elided N cannot. The relevant examples are below: (ii) a. pro comieron mucho/ Ellos comieron mucho (they)ate a-lot/ they ate a-lot b. El hijo de Marı´ a y el [e] de Juan no han visto el accidente the son of Marı´ a and the [e] of Juan Neg have seen the accident * El hijo de Marı´ a y el e´l de Juan no han visto el accidente the son of Marı´ a and the he of Juan Neg have seen the accident There are several problems for this analysis: first, Brucart and Gra`cia are not clear regarding the distribution of PRO. It appears that under the most well-known approaches to the distribution of PRO, PRO should be disallowed in the context in question. Furthermore, although it is true that a pronoun cannot substitute for PRO (see relevant data above), a lexical N cannot substitute for it either. Putting aside the above problems, their analysis cannot account for the differences in grammaticality obtained with prenominal and postnominal adjectives, repeated in (iii): (iii) a. *Ayer vi a la verdadera terrorista y a la supuesta [e] yesterday (I)saw to the true terrorist and to the alleged [e] b. Ayer vi la casa azul y la [e] verde yesterday (I) saw the house blue and the [e] green

11

In other words: Brucart & Gracia’s (1986) analysis predicts that data with prenominal As are grammatical, since prenominal adjectives should also be considered [+V, +N] and should therefore allow the cliticization of the definite article in (iiia). Finally, there is another problem related to the empirical coverage of Brucart & Gra`cia’s (1986) proposal. According to their analysis, nominal expressions headed by Ds other than the definite article should be generally grammatical. This is due to the lack of clitic-like properties of those Ds. However, as discussed above in the text, we still find restrictions on nominal elision with DPs headed by Ds other than the definite article; namely, we have the same type of restrictions with adjectives, independently on the D that heads the DP. Note that this claim is not completely correct, since pro can also receive indefinite meaning in cases such as (i): (i) En este paı´ s pro conducen como locos in this country pro drive as crazy ‘People drive like crazy in this country’

12

13

The term ellipsis refers to phrasal elision, elision is being used here as the more general term (see note 1). Note that the tests discussed above do not rule out the possibility of having gapping in Spanish nominal elision data in addition to ellipsis: the results of these tests just show that, in the relevant contexts, the data discussed behave as ellipsis processes. However, I assume from this point on that they are ellipsis processes, since, in Section 5.4, the treatment of As will completely rule out the possibility that Spanish nominal elision is gapping.

200 14

15

16

17

5 Nominal Ellipsis

There are arguments in the literature for this position (in fact, see the discussion below for an argument to this effect); however, as the goal of this chapter is not to provide an analysis of ellipsis phenomena in general, I will not discuss them and simply accept this position as the correct one. As noted regarding the target of partial cliticization, I assume here that an operation can target the lower segment of an NP. As mentioned in Chapter 3, this approach is similar to the one developed by Sportiche (1996) to account for floating quantifiers data. According to Sportiche’s (1996) analysis, floating Qs are adjoined to the NP and are left stranded after the NP (its lower segment) undergoes movement. Cf. also Bosˇ kovic´ (2002) for additional support for Sportiche’s (1996) analysis. I assume here that Grohmann’s (2003) proposal that adjunction cannot be the result of movement does not hold in PF. Of course, definite NP ellipsis is also ungrammatical when a prenominal A is the remnant, as in (i): (i) *Vimos al verdadero asesino y tu viste al presunto [e] (we)saw to-the true killer and you saw to-the alleged [e]

18

These cases are explained by the general analysis introduced in Section 5.3 and they will not be repeated in this subsection. There is a competing line of analysis to the ones described in the text. Bernstein (1993) follows Szabolcsi (1994) and proposes that the definite article is a subordinator of an XP other than NP. In other words, the definite article creates an argument out of (otherwise) predicative categories (Bernstein (1993:176)), such as PPs, CPs and APs, in definite NP ellipsis contexts, which is what explains the grammaticality of the relevant data. Therefore, the structure of a definite DP in these cases is the one in (i), where the definite article takes as its complement an AP, which contains the elided DP in its specifier position: (i)

This approach does not explain the restrictions on definite NP ellipsis with PPs, since, in principle, the definite article should be able to create an argument out of any PP, independently of the P that heads it. Therefore, the contrast in (ii) would be unexpected: (ii) a. El libro de fı´ sica/ El [e] de fı´ sica the book of physics/ The [e] of physics b. El libro sin pastas/ *El [e] sin pastas the book without covers/ the [e] without covers 19

Let me point here to an inconsistency in Brucart & Gra`cia (1986)’s analysis. Recall that their claim of the existence of a null category in the ellipsis site is based on the impossibility of contracting the Italian demonstrative quello with an adjective such as alto (tall). I repeat the relevant examples below: (i) a. Dammi lo sgabello, quello alto give-me the stool, that [e] tall b. *Dammi lo sgabello, quell’alto However, their explanation of the data with the definite article in nominal elision contexts assumes that the cliticization of the definite article can take place across the null N or PRO.

Notes 20

21

22

23

24

25

26

201

Nevertheless, I have modified Raposo’s (1999) original analysis to avoid some problems that this analysis faces with respect to the cycle and constituency. Namely, Raposo’s (1999) analysis appears to violate the cycle, since the structures [tarjeta roja] (red card) and [con tarjeta roja] (with red card) seem to be merged within the DP [las personas](the people) in the derivations he proposes for cases such as Las personas con tarjeta roja (the people with red card). Similarly, the place at which de (of) is inserted in Raposo’s (1999) derivations, at the edge of the higher phase, predicts that the constituency of a DP such as Las personas de tarjeta roja ( the people of red card) is [Las personas de] [tarjeta roja]. This predicts that de tarjeta roja (of red card) would not be a constituent and, as a consequence, it cannot be extracted out of DP. This prediction, as extensively shown in Chapter 2, is not borne out. As noted in Chapter 1, I am assuming that multiple spell out only applies when we obtain a phase. The only category that is a phase in the nominal domain is DP, or the highest projection in the nominal domain (cf. Bosˇ kovic´ (2005)). I also assume that PP is a phase (cf. Bosˇ kovic´ (2002), Bosˇ kovic´ (2004) for discussion on PPs as phases); crucially, I assume that AP is not a phase. Note that Raposo’s (1999) problem with respect to the constituency of false PPs does not arise under my analysis: de (of) is attached to the NP; therefore, they form a constituent. For a different solution to this problem, see Kornfeld & Saab (2004) For other example of overt anaphors, which are apparently rare cross-linguistically, see Lødrup (1994) and Bosˇ kovic´ (1994), who discuss the behavior of the overt surface anaphora det in Norwegian and also Bosˇ kovic´ (1997b), who argues for the presence of an overt surface anaphora in English pseudoclefts. Apart from the problems to Bernstein’s (1993) analysis of nominal ellipsis derived from Spanish data that are mentioned in the text, the reader should also consult Alexiadou et al. (2007: 258–264 and 272–273) where the results of a comparative study of the properties of French, English, Italian, German, Greek and West Flemish nominal ellipsis are used as the base for some other arguments against this approach. On a more technical side, Bernstein’s (1993) analysis of prenominal adjectives and nominal elision appears untenable in the current framework. Recall that her explanation of the impossibility of prenominal adjectives and nominal ellipsis is based on the assumption that prenominal adjectives, which are heads, interfere with the WM0 movement to a higher position (i.e., there is a violation of the head movement constraint). However, the head movement constraint has been abandoned in the current framework, its empirical coverage being derived from more general principles. Thus, in the current framework it is assumed that it is the presence of the same feature in two heads, which determines the impossibility of moving the lower head. Therefore, unless the adjective has the feature that drives WMmovement, the adjective would not interfere with the movement and the ungrammaticality of the examples with prenominal adjectives would remain unexplained. Lobeck (1995) uses the term Strong Agreement to refer to phonologically realized agreement.

Chapter 6

Conclusions and Further Research

6.1 Introduction The past five chapters of this book have described some of the fundamental properties of Spanish nominal expressions and analyzed them within a particular theoretical framework. Although the structure of nominal phrases has been extensively studied in the generative framework and some aspects of the nominal phrases have been analyzed extensively under this approach, this is the first monograph that combines a thorough description of the three main phenomena (namely, modification, extraction and ellipsis), with an analysis that accounts for all the particularities described. This final chapter aims to summarize the main conclusions of the book and to point to directions for further research. The chapter is divided into two main sections. Section 6.2, Major findings of the book, reviews the major descriptive generalizations illustrated in this book, such as the existence of two types of prepositional phrases in Spanish nominal expressions according to their movement possibilities; and the main theories the book has provided evidence for, such as the proposal that ellipsis processes belong to the PF component. Section 6.3, Further Research, discusses some of the directions that need to be explored to expand the analysis of this book. For instance, this section makes explicit the need of more research from a comparative perspective that can help us to determine the plausibility of the analysis described.

6.2 Major Findings of the Book This book shows that a considerable amount of data (some already known, some previously unobserved) can be accounted for in a principled way under a unified analysis that resorts to mechanisms that have been previously claimed in the literature. There are two types of findings in this book: theoretical and empirical. In this section, I will review some of these findings. M.E. Ticio, Locality Domains in the Spanish Determiner Phrase, Studies in Natural Language and Linguistic Theory 79, DOI 10.1007/978-90-481-3398-7_6, Ó Springer ScienceþBusiness Media B.V. 2010

203

204

6 Conclusions and Further Research

Due to the partially descriptive character of this book, there are plenty of empirical contributions. The book started with the discussion that the arguments within Romance nominal expressions (agents, possessors and objects) maintain a hierarchical relation in syntax independently of the linear order in which they occur. Furthermore, we illustrated that some PPs that were traditionally classified as adjuncts (PPA) showed the same behavior as argumental PPs, concretely they behaved as objects. The book also showed that the different types of nominal expressions have different extraction properties depending on the type of D that heads them. This observation led us to examine the differences among the D, divide them into weak or strong (following Milsark 1977) and investigate the different features present in each type of D. This examination was also related to the peculiar properties of the definite article in Spanish. Chapter 3 explored the two features involved in the relevant contrasts: definiteness and specificity. The generalization is then that the definite article in Spanish can have both features, which will produce the mentioned blocking effects, or just the definiteness feature, which will result in the possibility of extracting an argument out of the structure. The investigation of the syntactic properties of attributive adjectives and their relation with other elements within Spanish nominal expressions led me to conclude that there are two different types of attributive adjectives regarding their semantic contribution. Postnominal adjectives combine with the N via predicate modification to produce a new extension of the predicate over which the value of the D applies. On the contrary, prenominal adjectives do not denote properties that combine directly with the N via predicate modification but they denote functions and, in some cases, they must be interpreted outside of the scope of the D. Finally, Chapter 5, Nominal Ellipsis, details the properties of ellipsis in Spanish nominal expressions. It shows that nominal ellipsis is not a generalized phenomenon but that there are constraints to nominal ellipsis in Spanish. It was shown that (total) nominal ellipsis is possible with any D except in certain cases with the definite article and unstressed possessive pronouns and that partial nominal ellipsis can leave any PP as a remnant element except when the definite article heads the construction. Moreover, I illustrated that pre and postnominal adjectives behave differently with respect to nominal ellipsis in Spanish. From a theoretical point of view, this book provides strong support for some recent proposals and makes some important contributions. The most important contribution is to define a basic structure of Spanish nominal expressions that is able to accommodate the diverse data and generalizations of the book. The proposed analysis resorts to an independently needed assumption in order to explain the particular character of the relevant constructions in Spanish. In doing so, the main analysis of this book supports the following theoretical findings.

6.2 Major Findings of the Book

205

First, economy is the notion that guides the entire analysis, as well as the main theoretical framework. The analysis put forward in this book aims to provide the greater empirical coverage without postulating any additional devices or operations in our theory of grammar. The analysis then makes use of commonly accepted functional categories in the proposed structure and assumes restrictions that come from empirically based locality and anti-locality conditions to regulate the distance of dependencies. Second, the limited number of functional categories is also an attempt to be coherent with basic fundamentals of the theoretical framework. Only those categories that have relevant semantic features exist and only those categories that contain features in a given derivation can be present in that derivation. Furthermore, the presence or absence of these features corresponds to the projection (or not) of the functional category DP in the structure. This is crucial under my approach since the presence or absence of DP in the structure will affect the movement possibilities in the structure, due to its interaction with the locality conditions on movement. As shown in Chapter 2, weak determiners allow extraction out of the nominal expression since they do not project the DP category; while the presence of D with strong determiners disallows the extraction out of the DP. Regarding our ontology, the book claims that there are two types of fundamental elements in our theory of grammar: adjuncts and arguments. This distinction proves to be useful to deal with adjectives in Spanish nominal expressions. The proposal being that some adjectives are generated as adjuncts while other adjectives are generated as specifiers and that the different syntactic and semantic properties of adjectives are derived from this basic assumption. The assumption of two different positions to generate adjectives also reduces the need to have additional operations such as noun movement. As shown, this type of operation is currently under debate and it seems to be an added value to my analysis that I do not have to rely on it. Several parts of my analysis point to the existence of PF movements that alter the linear order but do not affect the syntactic or semantic properties of the construction. This type of movement can explain that there is no interaction between the final linear order in which we find the PP constituents and the syntactic dependencies they establish. Furthermore, the analysis defended in this chapter expands the empirical coverage of the main analysis argued in this study, providing us with additional empirical evidence for the hypothesis that ellipsis is a PF phenomenon and for the hypothesis that some surface anaphoras, such as partial cliticization, should be treated in terms of LF copying. Finally, the analysis of Spanish nominal expressions developed in this book has adopted some of the latest developments regarding CP structure, such as the division of clause structure into three domains and it has extended them to account for the properties of Spanish nominal expressions.

206

6 Conclusions and Further Research

6.3 Further Research Many topics are left for further research in this book. Among them, there are three topics that are particularly interesting. First, this book contains occasional references to other cross-linguistic studies such as Alexiadou et al. (2007) and a number of examples dealing with the behavior of some other languages. However, this book does not offer a detailed study of the properties of nominal expressions in other languages, since a comparative analysis was not among its goals. Nevertheless, in order to provide a better understanding of the phenomena described and to test the predictions of the analysis put forward in the book, a careful comparative analysis of different languages is surely required. This way, predictions made by the current analysis such as the one derived from my analysis of adjectives regarding the structural height and interpretation of adjectives must be explored in some other languages. Part of this work has already been started in this book, since many chapters conclude with a series of hypotheses to be examined further. However, a much more detailed crosslinguistic work remains to be done in the future. Second, we have dealt in this book with the parallelism between clausal and nominal domains. The repercussions of this proposal have guided the main analysis of the book and many differences between the clausal and nominal behavior have been noted in different sections of the book but often left unexplored. Further research needs to focus on the properties that make different the clausal and nominal domain. Among the topics to be investigated will then be the repercussions of this type of analysis proposed here for the overall locality constrains in the sentence. Finally, to complete the analysis of Spanish nominal expressions, some further research in the types and properties of these constructions are needed. For instance, it seems to be necessary to test the analysis proposed here against the properties of some additional data in the Spanish nominal expressions that have not yet received enough attention in the literature. For instance, although I investigate in this book the properties of wh-extraction from Spanish nominal expressions, nothing is said regarding the wh-in situ possibilities in the Spanish nominal expressions. This is the topic of Reglero and Ticio (2008) who discuss in detail the properties of wh-in situ in Spanish nominal expressions. Along the same line, further research is needed in topics such as Spanish partitive constructions, which have been treated extensively in some other languages (see Alexiadou et al. (2007, 395–475) for a discussion on the properties and main approaches to partitives, pseudo-partitives and the N-of-N construction crosslinguistically). Similarly, more research is needed in the series of phenomena that relate to the left periphery of Spanish nominal expressions, which are, as many other topics, just barely mentioned in this book.

References

Abels, Klaus. 2001. Clitics as complements of P. What Serbo-Croatian tells us about preposition stranding. Presented at the Acme Balkanica Conference, Concordia University, Montreal. Abney, Steven. 1987. The English Noun Phrase in its Sentential Aspect. PhD dissertation, Cambridge, MA: MIT Press. Aboh, O. Enoch. 1998. From the Syntax of Gungbe to the Grammar of Gbe. PhD dissertation, University of Geneva. Aboh, O. Enoch. 2004. Topic and focus within D. Linguistics in The Netherlands 2004. Amsterdam: Benjamins, 1–12. Alexiadou, Artemis. 2001a. Functional Structure in Nominals: Nominalization and Ergativity. Amsterdam: John Benjamins. Alexiadou, Artemis. 2001b. Adjective syntax and noun raising: word order asymmetries in the DP as the result of adjective distribution. Studia Linguistica 55: 217–248. Alexiadou, Artemis, Liliane Haegeman, and Marita Stavrou. 2007. Noun Phrase in the Generative Perspective. Berlin and New York: Mouton de Gruyter. Alexiadou, Artemis and Chris Wilder. 1998. Adjectival modification and multiple determiners. In A. Alexiadou and C. Wilder (eds.), Possessors, Predicates and Movement in the Determiner Phrase, 303–332. Amsterdam: John Benjamins. Anderson, Mona. 1979. Noun Phrase Structure. PhD dissertation, Storrs, CT: University of Connecticut. Aoun, Joseph. 1985. A Grammar of Anaphora. Cambridge, MA: MIT Press. Baker, Mark. 2003. Lexical Categories: Verbs, Nouns and Adjectives. Cambridge, MA: Cambridge University Press. Barwise, Jon and Robin Cooper. 1980. Generalized quantifiers and natural language. Linguistics and Philosophy 4: 159–219. Belletti, Adriana. 1988. The case of unaccusatives. Linguistic Inquiry 19: 1. Belletti, Adriana. 2001. Agreement projections. In M. Baltin and C. Collins (eds.), The Handbook of Contemporary Syntactic Theory. Blackwell. Bernstein, Judy B. 1993. Topics in the Syntax of Nominal Structure Across Romance. Doctoral dissertation, CUNY, New York. Bernstein, Judy B. 1997. Demonstratives and reinforcers in romance and germanic languages. Lingua 102: 87–113. Bernstein, Judy B. 2001a. Focusing the ‘Right’ Way in Romance Determiner Phrases. Probus 13: 1–29. Bernstein, Judy B. 2001b. The DP hypothesis: Identifying clausal properties in the nominal domain. In Mark Baltin and Chris Collins (eds.), The Handbook of Contemporary Syntactic Theory, 536–561. Malden, MA: Blackwell Publishers.

M.E. Ticio, Locality Domains in the Spanish Determiner Phrase, DOI 10.1007/978-90-481-3398-7, Ó Springer ScienceþBusiness Media B.V. 2010

207

208

References

Bleam, Tonia. 2006. The role of semantic type in differential object marking. Belgian Journal of Linguistics 19: 3–27. Boeckx, Cedric and Sandra Stjepanovic. 2001. Head-ing toward PF. Linguistic Inquiry 32(2): 345–355. Borer, Hagit. 2005. Structuring Sense: Vol. I: In Name Only. Oxford: Oxford University Press. Bosˇ kovic´, Zˇeljko. 1994. D-Structure, D-Criterion, and movement into D-positions. Linguistic Analysis 24: 247–286. Bosˇ kovic´, Zˇeljko. 1997a. The Syntax of Nonfinite Complementation. An Economy Approach. Linguistic Inquiry Monograph 32. Cambridge, MA: MIT Press. Bosˇ kovic´, Zˇeljko. 1997b. Pseudoclefts. Studia Linguistica 51: 235–277. Bosˇ kovic´, Zˇeljko. 2002. Floating quantifiers and D-role assignment. NELS 31: 59–78. Bosˇ kovic´, Zˇeljko. 2004. Be careful where you float your quantifiers. NLLT 22: 681–742. Bosˇ kovic´, Zˇeljko. 2005. On the locality of left branch extraction and the structure of NP. Studia Linguistica 59: 1–45. Bosˇ kovic´, Zˇeljko and Daiko Takahashi. 1998. Scrambling and Last Resort. Linguistic Inquiry 29: 347–366. Bosque, Ignacio. 2001. Adjective position and the Interpretation of Indefinites. In Javier Gutierrez-Rexach and Luis Silva-Villar (eds.), Current Issues in Spanish Syntax and Semantics, 17–37. Berlin/New York: Mouton de Gruyter. Bosque, Ignacio and Jose M. Brucart. 1991. QP Raising in Spanish Superlatives, Ms., Madrid/ Barcelona, UCM/UAB. Bosque, Ignacio and Violeta Demonte. 1999. Grama´tica descriptiva de la lengua espan˜ola. Madrid: Espasa-Calpe. Bouchard, Denis. 2002. Adjectives, Number and Interfaces, Oxford: Elsevier. Bowers, John. 1987. Extended X-Bar Theory, the ECP, and the Left Branch Condition. WCCFL 6: 47–62. Brucart, Jose. and Lluisa Gra`cia. 1986. I Sintagmi Nominali Senza Testa. Rivista di Grammatica Generativa 11: 3–32. Brug_e, Laura. 2000. Categorie Funzionale del nome nelle lingue Romanze. Milan: Cisalpino. Brug_e, Laura. 2002. The positions of Demonstratives in the Extended Nominal Projection. In Guglielmo Cinque (ed.), Functional Structure in DP and IP, 15–54. New York: Oxford University Press. Burzio, Luigi. 1986. Italian Syntax. Dordrecht: D. Reidel Publishing Company. Campbell, Richard. 1996. Specificity operators in SpecDP. Studia Linguistica 50(2): 161–188. Cardinaletti, Anna. 1998. On the deficient/strong opposition in Possessive systems. In Artemis Alexiadou and Chris Wilder (eds.), Possessors, Predicates and Movement in the DP, 17–53. Amsterdam: John Benjamins. Carnie, Andrew. 2002. Syntax. Oxford: Wiley Blackwell Publishers. Carstens, Vicki M. 1991. The Morphology and Syntax of Determiner Phrases in Kiswahili. PhD dissertation, UCLA. Carstens, Vicki M. 2000. Concord in minimalist theory. Linguistic Inquiry 31: 319–355. Cattell, Robert. 1979. On Extractability from Quasi-NPs. Linguistic Inquiry 10: 168–172. Chomsky, Noam. 1957. Syntactic Structures. The Hague: Mouton. Chomsky, Noam. 1965. Aspects of the Theory of Syntax. MIT Press. Chomsky, Noam. 1970. Remarks on nominalization. In R. A. Jacobs and P. S. Rosenbaum Peter S. (eds.), Readings in English Transformational Grammar, 184–221. Waltham, MA, Ginn. Chomsky, Noam. 1973. Conditions on transformations. In S.R. Anderson and P. Kiparsky (eds.), A Festschrift for Morris Halle. New York: Holt, Rinehart & Winston. Chomsky, Noam. 1980. Rules and Representations. New York: Columbia University Press. Chomsky, Noam. 1986. Barriers. Cambridge, MA: MIT Press. Chomsky, Noam. 1995. The Minimalist Program. Cambridge, MA: MIT Press. Chomsky, Noam. 2001. Derivation by phase. In Michael Kenstowicz (ed.), Ken Hale: A Life in Linguistics, 1–52. Cambridge, MA: MIT Press.

References

209

Chomsky, Noam and H. Lasnik. 1993. The theory of principles and parameters. In J. Jacobs, A. von Stechow, W. Sternefeld, and T. Vennemann (eds.), Syntax: An International Handbook of Contemporary Research, 506–569. Berlin: Walter de Gruyter. Cinque, Guglielmo. 1980. On extraction from NP in Italian. Journal of Italian Linguistics 5: 47–99. Cinque, Guglielmo. 1995. Italian Syntax and Universal Grammar. Cambridge Studies in Linguistics. Cambridge, MA: Cambridge University Press. Cinque, Guglielmo. 2005. Deriving Greenberg’s Universal 20 and its exceptions. Linguistic Inquiry 36(3): 315–332. Corver, Norbert and Mario van Koppen. 2006. Let’s Focus on Nominal Ellipsis. Talk given at GLOW. Culicover, Peter and Michael Rochemont. 1992. Adjunct extraction from NP and the ECP. Linguistic Inquiry 23: 496–501. Davies, Willliam D. and Stanley Dubinsky. 2003. On extraction from NPs. Natural Language and Linguistic Theory 21: 1–37. ´ del Adjetivo en el Sintagma Demonte, Violeta. 1999a. El Adjetivo: Clases y usos. La posicion Nominal. In Ignacio Bosque and Violeta Demonte (eds.), Grama´tica descriptiva de la lengua espan˜ola, 129–215, Madrid: Espasa-Calpe. Demonte, Violeta. 1999b. A minimal account of Spanish adjective position and interpretation. In Jon Franco, Alazne Landa, and Juan Martı´ n (eds.), Grammatical Analyses in Basque and Romance Linguistics, 45–75. Amsterdam: John Benjamins. Demonte, Violeta. 2000. Sema´ntica Composicional y Grama´tica: Los Adjetivos en la Interficie Le´xico-Sintaxis. Revista Espan˜ola de Lingu¨ı´stica 29–2: 283–316. Depiante, Marcela. 2001. The Syntax of Deep and Surface Anaphora: A study of Null complement Anaphora and Stripping/Bare Argument Ellipsis. PhD dissertation, Storrs: UConn. Diesing, Molly. 1992. Indefinites. Cambridge: The MIT Press (Linguistic Inquiry Monograph 20). Dimitrova-Vulchanova, Mila and Giuliana Giusti. 1998. Fragments of Balkan nominal structure. In A. Alexiadou and C. Wilder (eds.), Possessors, Predicates and Movement in the Determiner Phrase, 333–361. Amsterdam: John Benjamins. Duffield, Nigel. 1999. Adjectival modifiers and the specifier-adjunct distinction. In David Adger, Susan Pintzuk, Bernadette Plunkett, and George Tsoulas (eds.), Specifiers: Minimalist Approaches, 126–145. New York: Oxford University Press. Erteschik-Shir, Nomi. 1973. On the Nature of Island Constraints. Unpublished Ph.D. dissertation, MIT Press. Enc, Murvet. 1991. The semantics of specificity. Linguistics Inquiry 22: 1–25. Fa´bregas, Antonio. 2007. The internal syntactic structure of relational adjectives. Probus 19: 1–36. Ferna´ndez Soriano, Olga. 1999. El pronombre personal. Formas y distribuciones. Pronom´ In Ignacio Bosque and Violeta Demonte (eds.), Grama´tica descripbres a´tonos y tonicos. tiva de la lengua espan˜ola, 1209–1273. Madrid: Espasa-Calpe. Fiengo, Robert and James Higginbotham. 1980. Opacity in NP. Linguistic Analysis 7: 395–421. Fodor, Janet and Ivan Sag. 1982. Referential and quantificational indefinites. Linguistics and Philosophy 5: 355–398. Fox, Danny and Howard Lasnik. 2003. Successive cyclic movement and Island repair: The difference between sluicing and VP ellipsis. Linguistic Inquiry 34: 143–154. Gallego, Angel. 2007. An L syntax for adjuncts. http://seneca.uab.es/ggt/membres/gallego. htm. Accessed June 2007. Giorgi, Alessandra and Giuseppe Longobardi. 1991. The Syntax of Noun Phrases. Cambridge, MA: Cambridge University Press. Giusti, G. 1993. La sintassi dei determinanti. Padova: Unipress.

210

References

Giusti, G. 2008. Agreement and concord in nominal expressions. In De Cat, Cecile and Katherine Demuth (eds.), The Bantu-Romance Connection, 201–237 Amsterdam: John Benjamins. Godard, Daniele. 1992. Extraction out of NP in French. Natural Language and Linguistic Theory 10(2): 233–277. Grimshaw, Jane. 1990. Argument Structure. Cambridge, MA: MIT Press. Grohmann, Kleanthes K. 2003. Prolific Domains: On the Anti-Locality of Movement Dependencies. Amsterdam: John Benjamins. Grohmann, Kleanthes K. and Liliane Haegeman. 2002. Resuming reflexives. Proceedings of the 19th Scandinavian Conference in Linguistics, Nordlyd 31(1): 46–62. Gue´ron, Jacqueline. 2003. Inalienable possession and the interpretation of determiners. In Martin Coene and Yves D’hulst (eds.), From NP to DP. Volume II: The Expression of Possession in Noun Phrases. Amsterdam/Philadelphia: John Benjamins Publishing Company. Gutie´rrez, Javier and Enrique Malle´n. 2002. NP movement and adjective position in the DP Phases. In Julia Herschenshon, Enrique Malle´n, and Karen Zagona (eds.), Features and Interfaces in Romance: Essays in Honor of Heles Contreras, 107–132. Amsterdam: John Benjamins Publishing Company. Guasti, Maria Teresa. 1997. Romance causative. In Liliane Haegeman (ed.), The New Comparative Syntax, 124–144. London: Longmans Linguistics Library. Haegeman, Liliane. 2004. DP Periphery and clausal periphery: Possessor doubling in West Flemish? Relating nominal periphery to clausal periphery. In David Adger, Cecile de Cat, and G. Tsoulas (eds.), Peripheries, 211–239. Dordrecht: Kluwer. Hagstrom, Paul. 1998. Decomposing Questions. PhD dissertation, Cambridge, MA: MIT. Hale, Ken, and Jay Keyser. 1993. On argument structure and the lexical expression of syntactic relations. In Ken Hale and Jay Keyser (eds.), The View from Building 20: A Festschrift for Sylvain Bromberger, 53–108. MIT Press. Hankamer, Jorge and Ivan Sag. 1976. Deep and surface anaphora. Linguistic Inquiry 7: 391–426. Harris, John. 1991. The exponence of gender in Spanish. Linguistic Inquiry 22: 27–62. Heim, Irene. 1982. The Semantics of Definite and Indefinite Noun Phrases. PhD dissertation, UMass. Heim, Irene and Angelika Kratzer. 1998. Semantics in generative grammar. Oxford: Basil Blackwell. Herburger, Elena. 2000. What Counts: Focus and Quantification. Linguistic Inquiry Monograph 35. Cambridge, MA: MIT Press. Hornstein, Norbert. 1999. Movement and Control. Linguistic Inquiry 30: 69–96. Hornstein, N. 2001. Move! A Minimalist Theory of Construal. Oxford: Blackwell. Hornstein, Norbert, Susan Rosen, and Juan Uriagereka. 1999. ‘Integrals’. Ms. University of Maryland at College Park. Hornstein, Norbert, Kleanthes Grohmann, and Jairo Nunes. 2006. Understanding Minimalism. Cambridge, MA: Cambridge University Press. Huang, James C.-T. 1982. Logical Relations in Chinese and the Theory of Grammar. PhD dissertation, Cambridge, MA: MIT. Ishii, Takako. 1999. Cyclic spell-out and the that-t effects. WCCFL 18: 220–231. Ionin, Tania. 2003. Article Semantics in Second Language Acquisition. PhD dissertation, MIT. Distributed by MITWPL. Ionin, Tania. 2006. This is definitely specific: Specificity and definiteness in article systems. Natural Language Semantics 14: 175–234. Jackendoff, Ray. 1971. Gapping and related rules. Linguistic Inquiry 2: 21–35. Jackendoff, Ray. 1972. Semantic Interpretation in Generative Grammar. Cambridge, MA: MIT Press. Jayaseelan, K. A. 1990. Incomplete VP deletion and gapping. Linguistic Analysis 20: 64–81.

References

211

Johnson, Kyle. 2001. What VP-ellipsis can do, and what it can’t, but not why. In M. Baltin and C. Collins (eds.), The Handbook of Contemporary Syntactic Theory, pp. 439–479. Oxford: Blackwell. Julien, Marit. 2005. Nominal Phrases from a Scandinavian Perspective. Amsterdam: John Benjamins Publishing Company. Kamp, Hans. 1975. Two theories about adjectives. In Edward Keenan (ed.), Formal Semantics of Natural Language, 123–155. London: Cambridge University Press. Karimi, Simin. 1999. The specificity effect: Evidence from Persian. Linguistics Review 16: 125–141. Kayne, Richard. 1984. Connectedness and Binary Branching. Dordrecht: Foris. Kayne, Richard. 1994. The Antisymmetry of Syntax. Linguistic Inquiry Monograph 25. Cambridge, MA: MIT Press. Kennedy, Christopher and Jason Merchant. 2000. Attributive comparative deletion. Natural Language and Linguistic Theory 18: 89–146. Keenan, Edward L. 1987. A semantic definition of ‘‘indefinite NP’’. In Eric J. Reuland and Alice G.B. ter Meulen (eds.), The Representation of (in)definiteness. Cambridge, MA: MIT Press. Kim, J-S. 1998. Syntactic Focus Movement and Ellipsis: A minimalist Approach. PhD dissertation, Storrs: UConn. Kim, Y-K. 1997. Agreement phrases in DP. UCL Working Papers in Linguistics 9. Kiss, K. E´. 1998. Identificational focus versus information focus. Language 74: 245–273. Kolliakou, Dimitra. 1999. De-phrase extractability and individual/property denotation. Natural Language and Linguistic Theory 17: 713–781. Kornfeld, Laura and Andres Saab. 2004. Nominal ellipsis and morphological structure in Spanish. In Bok-Bennema, Bart Hollebrandse, Brigitte Kampers-Manhe, and Petra Sleeman (eds.), Romance Languages and Linguistic Theory. Selected papers from ‘‘Going Romance’’, 183–198, Amsterdam: John Benjamins. Laenzlinger, C. 2000. French adjective ordering: Perspectives on DP-internal movement types. Generative Grammar in Geneva 1: 55–104. Laenzlinger, C. 2005. Some notes on DP-internal movement. GG@G 4: 227–260. Laca, Brenda. 1987. Sobre el uso del acusativo preposicional en eapnol. In Carmen Pensado (ed.), El complemento directo preposicional, 61–91. Madrid: Visor. Lamarche, Jacques. 1991. Problems for N-movement to Num. Probus 3: 215–236. Larson, Richard. 1988. On the double object construction. Linguistic Inquiry 19: 335–391. Larson, R. and S. Cho. 2000. Temporal adjectives and the Structure of Possessive DPs. WCCFL 18. Lasnik, Howard. 1999. Minimalist Analysis. Malden, MA: Blackwell Publishers Inc. Lasnik, Howard and Juan Uriagereka. 1988. A Course in GB Syntax. Cambridge, MA: MIT Press. Lees, R.B. 1960. The Grammar of English Nominalizations. PhD dissertation, Cambridge: MA: MIT. Lees, R. B. and E. Klima. 1963. Rules for English pronominalization. Language 39: 17–28. Leonetti, Manuel. 1999. El artı´ culo. In Ignacio Bosque and Violeta Demonte (eds.), Grama´tica descriptiva de la lengua espan˜ola, 787–883. Madrid: Espasa-Calpe. Leonetti, Manuel. 2004. Specificity and Differential Object Marking in Spanish. http://www2. uah.es/leonetti/papers/Specif&DOM.pdf Lima, Ananda. 2006. A Minimalist View on Differential Object Marking for Specificity. 36th Annual Meeting of the Michigan Linguistics Society, Oakland University. Lobeck, Anne. 1995. Ellipsis. Functional Heads, Licensing and Identification. New York/ Oxford: Oxford University Press. Lødrup, Helge. 1994. Surface proforms in Norwegian and the definiteness effect. In M. Gonzalez (ed.), Proceedings of the North East Linguistic Society 24, 303–315. Department of Linguistics, University of Massachusetts. North East Linguistic Society, GLSA.

212

References

Longobardi, Giuseppe. 1994. Proper names and the theory of N-movement in syntax and logical form. Linguistic Inquiry 25: 609–665. Longobardi, Giuseppe. 2001. The Structure of DPs: some principles, parameters and problems. In M. Baltin and C. Collins (eds.), The Handbook of Contemporary Syntactic Theory, 562–603. Oxford: Blackwell. Luja´n, Marta. 1980. Sintaxis y sema´ntica del adjetivo. Madrid: Ca´tedra. Lyons, Christopher. 1999. Definiteness. Cambridge, MA: Cambridge University Press. Malle´n, Enrique. 2001. Issues in the Syntax of DP in Romance and Germanic. In Javier Gutierrez-Rexach and Luis Silva-Villar (eds.), Current Issues in Spanish Syntax and Semantics, 39–63. Berlin/New York: Mouton de Gruyter. Manzini, Rita M. 1994. Locality, minimalism, and parasitic gaps. Linguistic Inquiry 25: 481–508. Martı´ n, Juan. 1995. On the Syntactic Structure of Spanish DPs. PhD dissertation, University of Southern California. Masullo, Pascual. 1992. Incorporation and Case Theory in Spanish: A Crosslinguistic Perspective. PhD dissertation, Seattle: University of Washington. Matushansky, Ora. 2006. Head-movement in linguistic theory. Linguistic Inquiry 37(1): 69–109. May, Robert. 1977. The Grammar of Quantification. PhD dissertation, Cambridge, MA: MIT. Merchant, Jason. 2001. The Syntax of Silence. New York/Oxford: Oxford University Press. Milsark, Gary. 1974. Existential Sentences in English. Ph.D. dissertation, MIT. Milsark, Gary. 1977. Towards an explanation of certain peculiarities in the existential construction in English. Linguistic Analysis 3: 1–30. Ntelitheos, Dimitrios. 2003. The Syntax of Emphasis: Ellipsis and Discontinuity in the DP. http://www.philology.uoc.gr/conferences/6thICGL/ebook/h%5Cntelitheos.pdf Ntelitheos, Dimitrios. 2004. The Syntax of Elliptical and Discontinuous Nominals. M.A. thesis, UCLA. Ochi, Masao. 2000. Adjunct Wh-in situ and the Nominal Island. NELS 30: 557–569. Ormazabal, Javier. 1991. Asymmetries on Wh-movement and some theoretical consequences. Storrs: University of Connecticut. Ouhalla, Jamal. 2005. Agreement features, agreement and anti-agreement. Natural Language and Linguistic Theory 23: 655–686. Panagiotidis, Paedos. 2000. Demonstrative determiners and operators: The case of Greek. Lingua 110: 717–742. Pereltsvaig, Asya. 2006. Head movement in Hebrew nominals: A reply to Shlonsky. Lingua 116(8): A1–A40. Pesetsky, David. 1987. Wh-in-situ: Movement and unselective binding. In Eric J. Reuland and Alice G. B. ter Meulen (eds.), The Representation of (In)definiteness. Cambridge, MA: MIT Press. Pesetsky, David. 1992. Zero Syntax. Experiencers and Cascades. MIT Press. Picallo, M. Carmen. 1991. Nominals and nominalizations in Catalan. Probus 3: 279–316. Platzack, Christer. 2001. Multiples interfaces. In Urpo Nikanne and Emile van der Zee (eds.), Cognitive Interfaces. Constraints on Linking Cognitive Information, 21–53. Oxford: Oxford University Press. Pollock, Jean-Yves. 1989. Verb movement, universal grammar and the structure of IP. Linguistic Inquiry 20: 365–424. Pustejovsky, James. 1995. The Generative Lexicon. Cambridge, MA: MIT Press. Radford, Andrew. 2000. NP shells. Essex Research Reports in Linguistics 33: 2–20. Rainer, Ferdinand and Soledad Valera. 1991. Compounding in Spanish. Rivista di Linguistica 4(1): 117–142. Raposo, Eduardo. 1999. Towards a Minimalist Account of Nominal Anaphora in Spanish and English. Santa Barbara: University of California.

References

213

Reglero, L. and M. E. Ticio. 2008. Wh-in-situ and the Spanish DP: Movement or no movement? In Proceedings of the 31stPenn linguistics Colloquium 14(1): 311–324. Pennsylvania: University of Pennsylvania. Reinhart, T. 2006. Interface Strategies: Optional and Costly Computations. Cambridge, MA: MIT Press. Ritter, Elizabeth. 1988a. NSO noun phrases in modern Hebrew. NELS 17: 521–537. Ritter, Elizabeth. 1988b. A head-movement approach to construct-state noun phrases. Linguistics 26: 909–929. Ritter, Elizabeth. 1991. Two functional categories in noun phrases: Evidence from modern Hebrew. In S. Rothstein (ed.), Syntax and Semantics 25: Perspectives on Phrase Structure: Heads and Licensing, 37–62. New York: Academic Press. Rizzi, Luigi. 1990. Relativized Minimality, Cambridge, MA: MIT Press. Rizzi, Luigi. 1997. The fine structure of the left periphery. In Liliane Haegeman (ed.), Elements of Grammar: Handbook of Generative Syntax, 281–337. Dordrecht: Kluwer. Ramalle, Rodrı´ guez and Teresa Marı´ a. 2005. Manual de Sintaxis del Espan˜ol. Madrid: Castalia Universidad. Roehrs, Dorien. 2002. The boundary between morphology and syntax: Determiners move into the determiner phrase. IULC Working Papers Online 2: 1–46. Romano, Christine. 1991. Mixed headedness in American Sign Language: Evidence from functional categories. MIT Working Papers in Linguistics 14: 241–254. Romero, Maribel. 1998. Focus and Reconstruction Effects in Wh-Phrases. Ph.D. dissertation, Amherst: UMass. Rooth, Matt. 1992. A theory of focus interpretation. Natural Language Semantics 1: 75–116. Ross, John R. 1967. Constraints on Variables in Syntax. PhD dissertation, Massachusetts Institute of Technology. Published as Ross 1986. Available online at http://hdl.handle.net/ 1721.1/15166 Sag, Ivan A. 1976. Deletion and Logical Form. Ph.D. dissertation MIT. Published in 1980 by Garland Publishers, New York. Saito, Mamoru, and Keiko Murasugi. 1999. Subject predication within IP and DP. In Kyle Johnson and Ian Roberts (eds.), Beyond Principles and Parameters, 167–188. Dordrecht: Kluwer. Samek-Lodovici, V. (2007) Final and Non-final Focus in Italian DPs. Ms. UCL. (To appear a special issue of Lingua on DP-level information structure edited by Enoch Aboh). Sa´nchez, Liliana. 1996. Syntactic Structure in Nominals: A comparative study of Spanish and Southern Quechua. Doctoral dissertation, University of Southern California. Scott, Gary-John. 2002. Stacked adjectival modification and the structure of nominal phrases. In Guglielmo Cinque (ed.), Functional structure in the DP and IP: The cartography of syntactic structures volume I, 91–120. Oxford: Oxford University Press. Shlonsky, Ur. 2004. The form of Semitic nominals. Lingua 114(12): 1465–1526. Siloni, Tal. 1997. Noun Phrases and Nominalizations. The Syntax of DPs. Dordrecht Kluwer Academic Press. Stjepanovic´, Sandra. 1998. Extraction of adjuncts out of NPs. Paper presented at Workshop on Comparative Slavic Morphosyntax, June 5–7. Svenonius, Peter. 1994. On the structural location of the attributive adjective. In E. Duncan, D. Farkas, and P. Spaelti (eds.), Proceedings of the Twelfth West Coast Conference on Formal Linguistics. WCCFL, 439–454. Stanford, CA: Stanford Linguistics Association, CSLI. Swarzschild, R. 1999. GIVENness, AVOIDF, and other constraints on the placement of accent. Natural Language Semantics 7: 141–177. Szabolcsi, Anne. 1983. The possessor that run away from home. The Linguistic Review 3: 89–102. Szabolcsi, Anne. 1994. Functional categories in the noun phrase. In I. Kenesei (ed.), Syntax and Semantics 27: The Syntactic Structure of Hungarian. New York: Academic Press.

214

References

Ticio, M. Emma. 2003. On the Structure of DPs. Ph. D. dissertation, Storrs: University of Connecticut. Ticio, M. Emma. 2009. Putting the Spanish DP in order. In Pascual Masullo, Erin O’Rourke, and Chia-Hui Huang (eds.), Romance Linguistics: Structures, Interfaces, and Microparametric Variation, 261–274. Amsterdam/Philadelphia: John Benjamins. Tellier, Christine and Daniel Valois. 1995. Agreement and Extraction out of DP. Torrego, Esther. 1987. On Empty Categories in Nominals. Boston: UMass. Torrego, Esther. 1998. The Dependencies of Objects. Linguistic Inquiry Monograph 34. Cambridge, MA: MIT Press. Travis, Lisa. 1984. Parameters and Effects of Word Order Variation. Ph. D. dissertation, Cambridge, MA: MIT, 284 pp. Truswell, Robert. 2006. Adjectives and headedness. In Anna McNay (ed.), Oxford Working Papers in Linguistics, Philology and Phonetics, 10: 1–19. Truswell, Robert. 2009. Attributive adjectives and nominal templates. Linguistic Inquiry 40(3): 525–533. Uriagereka, Juan. 1988. On Government. PhD dissertation, University of Connecticut. Uriagereka, Juan. 1999. Multiple Spell Out. In D. Epstein and N. Horstein (eds.), Working Minimalism, 251–282. Cambridge, MA: MIT Press. Uriagereka, Juan. 2001. Adjectival clues’ Presented at the 5th Hispanic Linguistics Symposium. Urbana-Champaign: University of Illinois. Valois, Daniel. 1991. The Internal Syntax of DP. PhD dissertation, UCLA, Los Angeles, California. Vanelli, L. 1979. Una forma supplettiva dell’articolo e la sua fonosintassi. Rivista de Grammatica Generativa 4: 183–206. Vergnaud, Jean-Roger. and M. Luisa Zubizarreta. 1992. The definite determiner and the inalienable constructions in French and English. Linguistic Inquiry 23: 595–652. Von Heusinger, Klaus and Kaiser, Georg A. 2003. Animacy, specificity, and definiteness in Spanish. In K. von Heusinger and G.A. Kaiser (eds.), Proceedings of the Workshop "Semantic and Syntactic Aspects of Specificity in Romance Languages", 67–101 Arbeitspapier 113. Fachbereich Sprachwissenschaft, Universita¨t Konstanz. Williams, Edwin S. 1977. Discourse and logical form. Linguistic Inquiry 8(1): 101–139. Williams, Edwin. 1980. Predication. Linguistic Inquiry 11: 203–238. Williams, Edwin. 1994. Thematic Structure in Syntax. Cambridge, MA: MIT Press. Zamparelli, Roberto. 2000. Layers in the Determiner Phrase. New York: Garland. Zubizarreta, M. Luisa. 1998. Prosody, Focus, and Word Order. Cambridge, MA: MIT Press.

Index

A Abney, 1, 3–5, 15, 54, 56, 61, 102, 118 Adjectives, 1, 6, 16–17, 23, 66, 121–130, 132–162, 179, 184, 190, 201, 204–206 Adjuncts, 1, 15, 17, 20, 23–26, 28, 30, 42–47, 55–56, 77, 124, 135–136, 141–144, 146–149, 178, 183, 204–205 Adverbial adjectives (A-As), 127–128, 132–133, 140, 159 Agent, 3, 13, 16, 24–25, 28, 31, 33–43, 58–59, 63–68, 75, 93, 98, 114, 141–142, 144, 158, 160, 181–183, 186, 188 Agr, 6, 25, 54, 56, 58, 60–61, 64–65, 67, 81, 83, 85, 100–102, 104, 106, 110, 135, 193–194 Agreement, 1, 3, 6, 14–15, 18–19, 24–25, 51, 53, 56, 60–61, 101, 164–165, 188–197, 201 See also Agr; Agreement phrase (AgrP) Agreement phrase (AgrP), 6, 19, 25, 54, 56, 60–65, 67, 80, 101–102, 104, 111, 135, 144 Alexiadou, 7, 19–20, 74–76, 79–81, 117–119, 125, 152, 158, 161–162, 197, 206 A-Marking, 114–115 Anti-locality, 1–2, 8, 12–15, 20, 26, 50–55, 61, 79, 104, 177, 205 Appositive relative clauses, 151, 159 Argument, 13, 15, 16, 23, 30, 39, 43–45, 50–52, 68, 75, 80, 95, 99, 136, 138, 146, 150, 153, 196, 200, 204 Attributive adjectives, 1, 6, 16–17, 19, 72–74, 138, 204

Binding, 15, 23, 29, 31–38, 41–42, 75–76, 139, 185 Bleam, 106–107 Bosˇ kovic´, 12–13, 15, 20–21, 26, 50–52, 66, 77–80, 95, 161, 185, 187, 200–201 Bosque, 130–131, 138–139, 159, 168–169, 197–198 Bouchard, 152 Bowers, 109 Brucart, 19, 165, 172, 184, 199 Bruge`, 116, 118 Burzio, 69

B Belletti, 60, 87 Bernstein, 5, 18–19, 56–57, 60–61, 124, 126, 140, 147–149, 172, 189–192, 194, 196, 200

D Davies, 94–95 Deep Structure (DS), 46, 50, 61–62, 103, 105, 109, 115, 118, 134, 165, 167, 169, 193–194, 198–199

C Catalan, 5, 56, 129 C-command relations, 26, 30, 75 Chain resolution, 59, 75 Checking theory, 13, 51, 80 Chierchia, 117 Chomsky, 1, 3, 6, 11–12, 19–20, 25–26, 28–30, 36, 44–45, 48, 62, 68–69, 94, 98, 136, 138, 156, 198 and Lasnik, 12, 20, 62, 177 Cinque, 3, 7, 39, 72, 122, 124, 149–150, 152, 161, 196 Clitic doubling, 91, 117 Cliticization, 1, 68–71, 81, 134, 137, 143–146, 176, 185–188, 199–200, 205 Condition on Domain Exclusivity (CDE), 53 Context information, 53 Corver, 194

215

216 Definite article, 16, 18, 83–119, 154, 164–167, 171, 176, 183–186, 190, 192, 194, 196–200, 204 Definite effect, 115 Definiteness, 16, 83–119, 204 Definite nominal expressions, 86–88, 92, 97, 113–114, 116 Demonstratives, 57, 92, 96, 118 Demonte, 74, 127–128, 130, 152, 158, 169, 197–198 Depiante, 169–170, 187 Determiner Phrase (DP), 1, 4, 84, 117 Diesing, 103, 115 Differential object marking, 89, 107 Discursive status of the nominal expression, 103 Domains, 1, 6, 8, 12, 14–15, 23–25, 50–55, 57–59, 72, 100–101, 134, 205–206 Donellan, 91 DP hypothesis, 4, 7, 15, 54, 56, 95–96 Dubinsky, 94–95 Duffield, 68, 136, 152 Dummy preposition, 31–33, 75, 97, 185

E Economy, 5–6, 8, 12, 20, 24, 50–51, 57, 60, 205 Ellipsis, 1, 5, 16, 18–19, 56, 68, 134, 137, 143–146, 151, 155, 157, 163–201, 203–205 Ellipsis site, 18, 143–144, 163–164, 166, 171, 175–179, 181–183, 186, 190, 197, 200 Enc, 103, 115 English, 7, 19, 21, 28, 45, 55, 75, 78–81, 85, 92, 94, 96, 123, 137, 152, 154–156, 201 Erteschik-Shir, 94 Expletive definite article, 108, 119 Expression, nominal, 23–81, 83–119, 121–162, 163–201, 203–206 Extensional Q-As, 128–129, 132, 158 Extraction, 1, 12, 15–16, 18–19, 30, 38–43, 58, 61–62, 65–68, 71–72, 94–100, 103–105, 110–119, 142, 146, 149–150, 177, 181, 187, 195, 203–206

F Feature(s), 6, 13–16, 20, 51, 56–57, 84–85, 89–92, 118, 139, 142, 190, 193–194, 204–205 composition, 24, 26–27, 59, 75, 92–93, 115

Index Fernandez Soriano, 91 Focus movement, 196 position, 57, 195 Focus Phrase (FocP), 25, 57–58, 60, 79, 103–106, 111, 135, 144, 180 Free generation analysis, 59, 75 Free linear order, 29 French, 75–76, 80, 109, 122–124, 143, 162, 168, 190, 193–194, 197, 201 Functional categories, 1–2, 4–7, 14, 19, 56–57, 59–60, 79–80, 101–102, 149, 160, 205 Functional configuration and functional categories, 1, 5, 8, 15, 25, 54

G Gapping, 164, 172–176, 180, 197, 199 General NP ellipsis, 176–183 Generic nominal expressions, 108, 111–113, 119 Genitive pronouns, 15, 23, 26–28, 72–74, 81 German, 7, 20, 54, 75, 122, 124–125, 152–153, 160–162, 193–194, 197, 201 Germanic languages, 7, 122, 125, 152–153, 161 Giorgi, 3, 19, 31, 39, 75, 129 and Longobardi, 3, 19, 31, 39, 75, 129 Giusti, 57, 60–61, 79–80, 118 Goal, 1–2, 5, 23, 26, 33, 39, 54, 57, 60, 84, 102, 134, 146, 189, 200, 206 Gra`cia, 19, 165, 172, 184, 199 Greek, 19, 80, 119, 124, 197, 201 Grohmann, 1, 6, 12–15, 20–21, 25–26, 51–54, 61, 68, 72, 78–80, 118, 200 and Haegeman, 15, 25, 54, 60, 72, 79 Gue´ron, 108

H Haegeman, 15, 25, 54, 60, 72, 79 Hankamer and Sag, 187 Hornstein, 13, 20, 30, 51–52, 55, 77, 80

I Intensional Q-As, 128 Ionin, 89, 92 Italian, 31, 72, 75, 80, 109, 124, 170, 172, 196–197, 200

Index J Jackendoff, 18, 164, 172–173 Jayaseelan, 180, 195 Julien, 16, 85, 92

K Kamp, 128 Karimi, 90, 95 Kayne, 59, 68, 136, 150, 158 Kim, 137, 173–174, 180 Korean, 137, 180

L Laenzlinger, 122, 149, 152 Lamarche, 7, 122–123, 152–153 Lasnik, 12, 20, 26, 36, 62, 173, 177, 180, 195, 197 Left periphery, 194, 206 Leonetti, 89–90, 103, 115, 117, 119 LF copying, 19, 69–71, 187–188, 197, 205 LF interface, 8, 18, 53, 163–164 LF movement, 109 Lima, 103 Linear Correspondence Axiom, LCA, 59 Linear word order, 15, 23, 26–27, 30, 75, 121, 154, 161–162 See also Surface word order Lobeck, 173–175, 189, 193–194, 197, 201 Locality, 1–2, 8, 11–16, 20, 23, 26, 48, 50–55, 61–66, 68, 71–72, 79–80, 83–85, 96, 100, 104–106, 121, 160, 177, 182, 187, 203, 205–206 Longobardi, 3, 19–20, 31, 39, 75, 102, 108, 117–118, 129–130 Luja´n, 126 Lyons, 16, 85, 87, 117

M Malle´n, 19, 58, 149 Manzini, 12, 14, 26, 62, 80 Martı´ n, 3, 31, 33, 35, 66, 126, 140, 159 Matushansky, 154, 161 Milsark, 61, 80, 87, 101, 204 Minimalism, 8–11 Minimal link condition, 12, 20 Minimize chain links principle, 12, 20, 62 Mood as specificity marker, 30, 89, 93, 117, 119 Multiple copies, 194

217 N Narrow scope reading, 109, 113 N’-Deletion, 172 N ellipsis, X ellipsis, 18, 164 N’-Gapping, 172 N movement, 1–2, 5, 11–13, 20–21, 58, 68, 79, 84–85, 104–106, 122–124, 142, 146–149, 152–155, 161–162, 205 Nominal expressions adjectives, 121–162 further research, 203–206 nominal ellipsis, 163–201 specificity/definiteness/definite article, 83–119 structure of, 23–81 Nominal expressions with the definite article, 96–97, 108–109, 112, 167, 183–184 Nominal modifiers, 28 Non-definite Ds, 62 Non-definite nominal expressions, 87 Non-locative a (to), 30–31, 75 Non-restrictive reading, 137, 149 Noun Phrase (NP), 3–4, 19, 59, 101, 109, 193 NP ellipsis, XP ellipsis, 18–19, 56, 134, 137, 143–146, 155, 157, 160, 164, 172, 176–183, 188–194, 197, 200 NP movement, 17, 122–124, 147, 162 NP. Small nP, 4, 6, 11, 17–20, 25, 27, 45, 56–60, 64–71, 77, 81, 101, 109, 122–124, 134–135, 137–148, 151, 160–162, 164–165, 172–173, 184–195, 200 Ntelitheos, 194–195, 197 Null NP, 187, 191 Number Phrase (NumP), 19, 56, 60, 107, 148, 161, 191

O Object, 9–12, 20, 24–25, 31–32, 36, 38–42, 45, 48–49, 64–72, 76–79, 81, 84, 88–89, 91, 98–99, 112–114, 116, 140–142, 158, 160, 167, 181, 186, 188 Omega-domain, 25, 58, 65, 67, 83, 101, 104–106, 111, 135, 144 Ormazabal, 19, 26, 29, 38–39, 41–43, 47, 58, 60–61, 76, 108, 112, 160

P Panagiotidis, 118 Partial cliticization, 1, 68–71, 81, 134, 137, 143–146, 176, 186–188, 197, 200

218 Partial nominal elision, 166, 197 Partitive constructions, 116, 206 Pereltsvaig, 154 PF deletion, 19, 183, 187–188, 195, 197 PF interface, 13, 52, 59, 157, 177, 182, 186, 188 Phase, 19–20, 79, 185–186, 201 Phi-domain, 25, 53–54, 58, 64–65, 67, 83, 101, 104, 106, 111, 135, 144, Picallo, 3, 5, 56, 129 Platzack, 1, 6, 25–26, 51 Possessive pronouns, 26, 81, 170–171, 198, 204 See also Genitive pronouns Possessor, 16, 24–25, 28, 34–42, 48–49, 54–56, 59, 61–64, 68, 70–72, 74, 80–81, 84, 97–99, 112, 114–116, 167, 179–181, 183, 188, 204 Postnominal adjectives, 28, 122, 124, 126, 132, 134–135, 137–146, 151, 155–156, 168, 179, 191, 199, 204 PP adjuncts, 38, 42, 46, 138 PP adverbials, 38, 42 PP arguments, 16–17, 23, 28, 39, 43, 59, 75, 105–106, 124, 134, 141, 166 Predicate fronting, 122, 150–151, 160 Predicate modification, 132–134, 137, 140–141, 159, 204 Prenominal attributive adjectives, 17, 73, 138 Preposition(s) a (to), 30, 89, 90, 98, 116–117, 159 de (of), 28–30, 196 por (by), 29, 31, 35, 41–45, 65, 77, 158 Prepositional phrases (PPs), 15, 20, 23, 26–39, 45–49, 72–75, 83–84, 99, 114, 117–118, 184–185, 203, 204 Probe, 59, 68 Proliferation of functional categories, 6, 14, 57 Prolific Domain, 14–15, 20, 52–55, 61–62, 78–79 Pustejovsky, 133 Q Qualitative adjectives (Q-As), 127–131, 137–140, 143, 159, 179 Quantifier-variable readings, 30

R Radford, 56, 59, 79 Raposo, 19, 165, 185, 201 Reduced relative clause, 150

Index Reinhart, 196 Relational adjectives, 17, 124, 127, 134–137, 141–143, 155, 158, 179 Relativized Minimality, 12, 20 Remnant elements, 18, 164–165, 168, 177, 179–180, 195 Restrictive reading, 137, 149 Restrictive relative clauses, 150–151 Resumption, 55 Right periphery, 57 Rightward movements, 59 Ritter, 4, 7, 56, 60, 193 Rizzi, 1, 6, 12, 14, 25–26, 51–52, 56, 60 Romance adjectives, 7, 124, 146–147, 149, 152 Romance languages, 5, 7, 31, 34, 39, 56–57, 60, 75, 97, 108, 119, 121–122, 124, 153–154, 158, 196 Romanian, 154 Rooth, 194 Ross, 11

S Sag, 88, 177, 187 Sa´nchez, 19, 58, 126, 140–141, 149, 159, 198 Scope relations, 153–154, 157 Scott, 122, 149, 156 Shlonsky, 7, 55, 152, 154, 161 Siloni, 7 Specific definite Spanish DPs, 103–105 Specific indefinite nominal expressions, 86, 89, 97–99, 100, 105–107 Specificity, 16, 30, 57, 76, 83–119, 204 effects, 16, 57, 84–85, 96, 103, 106, 109, 115 Specific nominal expressions, 16, 46, 62, 64–65, 84, 93–94, 96–97, 103, 107, 109, 112–113, 118, 159 Specifiers, 17, 20, 40, 58–59, 123, 136, 140, 143, 150, 158, 160–161, 179, 205 Spell out units, 185 Split CP hypothesis, 51 Stressed possessive pronoun, 170–171, 198, 204 Strong determiner, 61, 85, 100–102, 110, 205 Stylistic rule, 29, 182–183 Surface word order, 7–8, 11, 37–38, 58–59, 75 Superlative constructions in Spanish, 112 Surface anaphora, 19, 186, 197, 201, 205

Index Swarzschild, 194 Szabolcsi, 4, 7, 56, 200

T Tellier, 109 Thematic hierarchy, 16, 26, 30, 36, 38–39, 75–76, 195 Theta criterion, 50, 78 domain, 80, 101 positions, 13, 50–52, 78, 80 roles, 4, 13, 50–51, 78, 80 Ticio, 7, 59, 75, 152, 206 Topic Phrase (TopP), 57, 60, 137 Torrego, 3, 26, 41, 58, 61, 89, 95, 108, 112, 160, 189–190, 194 Truswell, 6, 152, 162

U Unstressed possessive pronoun, 170–171, 198, 204 Uriagereka, 19–20, 36, 68, 77–78, 95, 136, 159

219 V Valois, 3, 18–19, 56, 109, 122, 124, 140, 147, 164, 172 van Koppen, 194 Vergnaud, 74, 108 VP. Small vP, 4, 6, 9–10, 13, 20, 52, 56, 60, 66, 70, 173–175, 178, 180, 195

W Weak determiners, 61, 100, 102, 205 Wh-extraction, 16, 26, 38, 43, 55, 84, 96, 206 Wh-island, 43, 46–47, 49 Wh- in situ, 206 Wide scope reading, 109 Williams, 30, 75, 175 Word Marker Phrase (WMP), 5, 56, 191

Z Zamparelli, 19, 61, 80, 102, 130 Zubizarreta, 74, 108, 196