The Architecture of Focus 9783110922011, 9783110185782

Citation preview

The Architecture of Focus

W G DE

Studies in Generative Grammar 82

Editors

Henk van Riemsdijk Jan Köster Harry van der Hülst

Mouton de Gruyter Berlin · New York

The Architecture of Focus

edited by

Valeria Molnär Susanne Winkler

Mouton de Gruyter Berlin · New York

Mouton de Gruyter (formerly Mouton, The Hague) is a Division of Walter de Gruyter G m b H & Co. KG, Berlin.

The series Studies in Generative Grammar was formerly published by Foris Publications Holland.

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

Library of Congress Cataloging-in-Publication

Data

The architecture of focus / edited by Valeria Molnär, Susanne Winkler, p. cm. — (Studies in generative grammar ; 82) Includes bibliographical references and index. ISBN-13: 978-3-11-018578-2 (cloth : alk. paper) 1SBN-10: 3-11-018578-4 (cloth : alk. paper) 1. Focus (Linguistics). 2. Generative grammar. 3. Grammar, Comparative and general. 1. Molnär, Valeria. II. Winkler, Susanne, 1960— . III. Series. P299.F63A73 2005 415—dc22 2005034179

Bibliographic information published by Die Deutsche

Bibliothek

Die Deutsche Bibliothek lists this publication in the Deutsche Nationalbibliografie; detailed bibliographic data is available in the Internet at .

ISBN-13: 978-3-11-018578-2 ISBN-10: 3-11-018578-4 ISSN 0167-4331 © Copyright 2006 by Walter de Gruyter G m b H & Co. KG, D-10785 Berlin. All rights reserved, including those of translation into foreign languages. No part of this book may be reproduced in any form or by any means, electronic or mechanical, including photocopy, recording, or any information storage and retrieval system, without permission in writing from the publisher. Cover design: Christopher Schneider, Berlin. Printed in Germany.

Acknowledgments

This volume grew out of two workshops held at the University of Lund: the workshop on Informationsstruktur - kontrastivt in 2002 and the GLOW workshop on Information Structure in Generative Theory vs. Pragmatics in 2003. We would like to thank the participants of these workshops for the substantial discussions of the ideas and issues presented in this volume. Special thanks are due to the Architecture of Focus authors for their contributions and their participation in the internal reviewing process. We would further like to gratefully acknowledge the fruitful dialogues on the architecture of focus with the following linguists: Katrin Axel, Carsten Breul, Aniko Csirmäz, Bernie Drubig, Hans-Martin Gärtner, Beata Gyuris, Edward Göbbel, Philip J. Jaegger, Mats Johansson, Ewald Lang, Luis Lopez, Läszlo Molnärfi, Doris Penka, Baläzs Suränyi, Kriszta Szendröi, Carla Umbach und Ilse Zimmermann. A special thank you goes to Henk van Riemsdijk for discussing the design of this book with us, for suggesting the title and for his ongoing support in bringing it to its conclusion. We also want to thank Ursula Kleinhenz for her professional assistance and invaluable help in carrying out this project. Particular gratitude goes to Kirsten Brock for copyediting substantial parts of this text and to Daniel Ormelius and Christian Steinrücken for their help in putting this volume together. We also thank Crafoordska Stiftelsen (Sweden), Elisabeth Rausings Minnesfond (Sweden), Vetenskapssocieteten i Lund (Sweden), and the German Science Foundation (DFG) for supporting the workshops and the research of the editors presented in this book. We also express our gratitude to Elisabeth Rausings Minnesfond for providing financial support for the editorial work.

Contents

Exploring the architecture of focus in grammar Valeria Molnar and Susanne Winkler

1

Part 1: Theoretical foundations of focus and information structure On the architecture of topic and focus Nomi Erteschik-Shir

33

Topic, focus and the structural dynamics of language Ruth Kempson, Ronny Cann, Jieun Kiaer

59

On the (in)dependence relation between syntax and pragmatics Joäo Costa and, Maria Cristina Figueiredo Silva

83

Association with focus phrases Manfred Krifka

105

Focus as identification Istvcin Kenesei

137

Focussing as predication Katalin E. Kiss

169

Part 2: The representation of focus, topic and contrast at the syntax-pragmatics interface On different kinds of contrast Valeria Molnär

197

Contrast and movement to the German prefield Werner Frey

235

Noncanonical negation and information structure in Finnish Elsi Kaiser

265

Double Negation and information structure: Somewhere between topic and focus Genoveva Puskas

291

viii

Contents

Part 3: The phonological representation of focus, topic and contrast and its relation to meaning Focus projection and default prominence Daniel Β tiring

321

Focus on contrast and emphasis: Evidence from prosody Nicole Richter and Grit Mehlhorn

347

Topic-focus controversies Nancy Hedberg

373

Part 4: Focus-related constructions in different languages Ellipsis at the interfaces: An information-structural proposal Susanne Winkler

401

Elliptical ifass-clauses in German Kerstin Schwabe

429

On the discourse impact of subordinate clauses Andre Meinunger

459

The influence of operators on the interpretation of DPs and PPs in German information structure Anita Steube

489

Clefts in English and Norwegian: Implications for the grammarpragmatics interface Jeanette

Κ. Gundel

517

The complex functions of /r-clefts Stefan Η über

549

Focus constructions in Hausa Katharina Index

Hartmann

579 609

Exploring the architecture of focus in grammar Valeria Molnär and Susanne Winkler

1. Introduction Over the last four decades linguists have made tremendous progress in the description of information structure and focus. The major breakthrough occurred in the early 70s when Chomsky (1972) and Jackendoff (1972) introduced the concept of focus into generative theory. Chomsky's paper laid the foundation for regarding focus as an essentially phonological phenomenon and Jackendoffs work provided the theoretical tools for formally integrating the focus feature [+F] into the theory of generative grammar. These two proposals constituted the theoretical starting point for many papers which aimed at an explanation of how sentence accent is related to the syntactic anchoring and the semantic interpretation of focus. This first phase of focus research continued to grow steadily and reached its peak within the Principles and Parameters framework. The central research goal was to develop a modular model of grammar which provides a principled account of the theory of focus. The major emphasis was put on the integration of focus into the three main components of the grammatical model: the syntactic component, logical form, and phonological form. The architecture of grammar was based on the T-model with the two-interface hypothesis, which specifies that meaning and sound interact via syntax. In the 90s, with the introduction of the Minimalist Program (Chomsky 1995), the research interest shifted to the investigation of these interfaces, first and foremost the interaction between the syntactic and the phonological components (Ladd 1996, Selkirk 1995, Zubizarreta 1998, Szendröi 2004) and the syntactic and the semantic components (Jacobs 1988, Rooth 1985, Rizzi 1997, Büring 1997, this volume, Szabolcsi 1997, Krifka, this volume). While there is no doubt that prosody and meaning are central to the architecture of focus, the precise role of the syntactic component as mediator remains unclear. With the critical investigation of the two-interface hypothesis, the question emerged as to how syntax-external concepts, such as the essentially pragmatic concepts

2

Valeria Molnär and Susanne

Winkler

connected with focus, can be accounted for under the traditional approach. The implications for focus research have been straightforward: While most researchers working in the generative framework have agreed that information-structural notions are closely tied to the interface issues, the question as to the actual division of labor between the components and their interaction is still controversial. One of the central questions of focus research today is whether focus-related phenomena are best described within the standard generative framework, either as an effect of syntax or as an effect of one of the interpretive components, or whether their explanation should be allocated to the pragmatic component. More concretely, how should derivational processes be modeled with respect to their relation to pragmatics? Do discourse factors actually drive syntactic operations, or does the information-structural component in interaction with pragmatics interpret the output of the syntactic component? Similar challenging questions arise with respect to the interpretive components: Do they impose requirements on syntax or do they simply interpret the structures which are handed over to the interfaces? The great challenge for current linguistic research therefore is to clarify the relation between the internal components (syntax, semantics, phonology) and the external component (pragmatics) of the grammatical model and to integrate information structure into the generative framework, thus providing well-motivated answers to the above questions. In addition to the increased interest in theoretical issues and empirical phenomena related to information structure in recent years, we can also observe that the various approaches to focus and information structure have undergone a radical change in perspective. The theoretical concept of focus and its relation to the notion of topic and contrast has become the core object of research. Along with this, the research questions have shifted: instead of merely locating focus in the architecture of grammar, linguists are investigating the architecture of focus itself. Consequently, the basic underlying aim of this collection is to document this change in perspective and thereby isolate essential keystones and research areas in both the theoretical and empirical domains. The vast number of typological studies and the increased interest in language comparison and focus constitute one of the driving factors for this change. Since the 80s, languages which possess a designated position for the focus constituent, like Hungarian and Basque, have served as evidence for the claim that focus is a syntactic phenomenon which de-

Exploring the architecture offocus in grammar

3

termines the sentence structure of these languages. The detection of language-specific differences in focusing has increased the interest in typological investigations and led to the study of focus-related p h e n o m e n a in a great variety of European, African and Asian languages (cf. Sasse 1987, Ε. Kiss 1995, Drubig 2003, to appear). The typological research has pursued t w o main aims: first, the description of information-structural notions and their structural realization, and second the extraction and identification of language universals. There have basically been two main results. First, linguists have gained deeper insights into the diversity and complexity of the information-structural p h e n o m e n a due to the intense typological interest and the incorporation of a wide variety of data f r o m typologically and genetically diverse languages into the research on i n f o r m a t i o n structure. Second, linguists have realized that only together can the t w o main strands of research on information structure, n a m e l y the g e n e r a t i v e approach and the descriptive t y p o l o g i c a l approach, provide the empirical evidence f o r generalizations within the field of i n f o r m a t i o n structure and m a k e a new architectural design of f o c u s possible. This innovative perspective has been supported by the results of indepth theoretical and empirical investigations of information-structural relations in many languages. Thus information structure has been established as an independent research area, with focus as the core concept. This c h a n g e of perspective allows a new architectural design of inform a t i o n structure, w h e r e the notion of focus is the center of attention. T h e r e are basically four architectural dimensions of focus: (i) The formal syntactic level, (ii) the phonological interpretation depending on accent assignment and phonological phrasing, (iii) the semantic interpretation based on alternative semantics, and (iv) the pragmatic dimension based on the conception that focus is the information center of the sentence containing the new or non-presupposed part of the utterance. In some cases all four dimensions coincide, as illustrated in (1). Capitalization signals pitch accent assignment. (1)

A: B:

Where was Count Dracula born? He was born in TRANSYLVANIA.

In e x a m p l e ( I B ) , the syntactic focus feature is assigned to the accented e l e m e n t Transylvania, which is the pragmatic focus providing the new information. At the same time Transylvania satisfies the semantic focus

4

Valeria Molnär and Susanne Winkler

requirements defined in set-theoretical terms specifying the relevant alternative or subset in the set of locations and thus providing an answer to the question where in (1A). However, there are also cases in which the different dimensions do not coincide. The syntactic and the phonological dimensions diverge in (2), demonstrating that phonological prominence does not necessarily serve as a sign of focus: (2)

A: B:

What happened to Lucy and Mina? BOTH of them were attacked by DRACULA.

In example (2B), there are two accented constituents. However, only the sentence-final constituent, Dracula, is analyzed as the main focus of the sentence, while the sentence-initial constituent is identified as the topic in most approaches. Furthermore, the Hungarian example in (3B) shows that semantic focus defined in the set-theoretical sense does not need to correspond to the pragmatic focus, defined as new information. In focus-prominent languages with designated syntactic focus positions, such as Hungarian, wh-words have the same distribution as focus constituents and occur in the focus position immediately to the left of the finite verb: (3)

A:

KIT ölt meg Dracula? whom kill-past perf Dracula 'Whom did Dracula kill?'

B:

Dracula LUCYT ölte meg. Dracula Lucy-acc kill-past perf 'Dracula killed Lucy.'

While the w/t-phrase in (3A) has the semantics of the focus since it creates a set and operates on alternatives, it cannot be considered the focus in pragmatic terms since it does not provide the new information of the sentence. In the answer (3B), however, all relevant focus requirements are again fulfilled and the four dimensions of focus coincide, leading to the convergence of the four focus dimensions on Lucy. The above discussion shows that the complexity of focus requires the definition of related notions, such as the notions of topic and contrast. Chomsky (1965) defines "the Topic-of the sentence as the leftmost NP immediately dominated by S in the surface structure and the Comment-

Exploring the architecture of focus in grammar

5

of the sentence as the rest of the string" (1965: 221). The most influential d e f i n i t i o n of topic as the category of aboutness goes b a c k to Hockett (1958) and has been further developed by Reinhart (1982). An alternative view holds that the term topic is best described by the notion of frame: " T h e topic sets a spatial, temporal, or individual f r a m e w o r k within w h i c h the main predication h o l d s " ( C h a f e 1976: 51). A n o t h e r definition of topic is given with recourse to old information: The topic is either identified as given information or in weaker versions, by reference to familiarity as in Gundel's (1988) topic-familiarity condition. Within the generative theory, it is generally assumed that an uninterpretable feature in the functional head position triggers l e f t w a r d - m o v e m e n t of the topic constituent (cf. Halliday 1967, M o l n ä r 1991, L a m b r e c h t 1994, Rizzi 1999, Breul 2004). T h e sentence-initial constituents of e x a m p l e s (1) to (3) all q u a l i f y as topics. In e x a m p l e ( I B ) the p r o n o u n he, in e x a m p l e ( 2 B ) both, and in (3B) Dracula, all c o n s t i t u t e g i v e n or presupposed constituents of the sentences describing what the sentences are about. In this sense, the topic is complementary to the focus, which expresses new information. W h i l e the notions topic and focus contribute to the basic requirem e n t s of i n f o r m a t i o n structuring, such as coherence and i n f o r m a t i v e ness, the notion of contrast interacts with topic and focus on a separate level and can affect both of these two notions. Contrast, as argued by M o l n ä r (this volume), is not directly bound to certain discourse pragmatic relations; it can be realized on both topics and foci, as illustrated in (4): (4)

A: B:

What did Count Dracula do to Mina and what to MINA, he ATTACKED but LUCY he KILLED.

Lucy?

M o s t approaches would agree that the predicates attacked and killed in e x a m p l e (4B) are contrastive foci and Lucy and Mina are contrastive topics. H o w e v e r , the exact definitions of these contrast-related notions are still controversial. T w o main dimensions of contrast must be distinguished: First, contrast is used to mark opposition on either the paradigmatic or the syntagmatic level, and second it serves the function of phonological highlighting by accent a s s i g n m e n t (cf. B u ß m a n n 1990: 419). O b v i o u s l y , these two main properties of contrast also characterize the notion of focus. T h e r e f o r e the notions focus and contrast are often regarded as the-

6

Valeria Molnär and Susanne

Winkler

oretical concepts which are closely related. The impact of contrast on information structure is intensely debated in current research. The central question is whether contrast should be regarded as a gradient or as a nongradient notion. If one shares Bolinger's (1961) and Lambrecht's (1994) view according to which focusing is always contrastive, then utterances can only be used contrastively and it is only possible to speak of clear or less clear instances of contrastiveness. Even this gradient approach admits that the "clearest instances of contrastiveness" are those "in which a focus designatum explicitly contradicts a stated or predicted alternative" (Lambrecht 1994: 290). The alternative view, however, holds that contrastiveness is nongradient and that contrastive and noncontrastive cases of focusing must be distinguished (Drubig 1994, E. Kiss 1998). The investigation of the architecture of focus requires the clarification of the linguistic status of contrast in relation to focus. The standard view represented by linguists working within different theoretical frameworks is that contrast cannot be regarded as an independent phenomenon of information structure, but only as a feature of focus and topic (cf. Dretske 1972, Halliday 1967, Rochemont 1986, Rooth 1985, Lambrecht 1994, Biiring, this volume). A different analysis has been suggested by Chafe (1976) and also by recent analyses of Finnish data (Vallduvi and Vilkuna 1998, Molnär 2002, Kaiser, this volume). They propose that contrast is a so-called packaging phenomenon which is relevant in constructing a felicitous discourse. The central research task then targets the precise theoretical characterization of focus in the architecture of grammar. Two connected issues must be pursued: First, we must assess the required universal inventory of information-structural categories and distinguish their precise meaning from cognitive, psychological and functional perspectives. At the same time, we must group them into subclasses and discuss their internal relations. Second, we must aim for generalizations which allow us to arrive at a universally valid theory of information structure. A relevant empirical testing ground is provided by focus-related constructions, often also called stylistic constructions (Rochemont 1986). Focus-related constructions, which are defined as syntactic dislocations with a salient phonological realization and a specific information-structural interpretation, have developed into a cutting-edge research area in generative grammar. They comprise movements to the left periphery of the clause, such as topicalization, w/z-movement, and left dislocation, and

Exploring the architecture offocus in grammar

7

movements to the right periphery, such as extraposition, inversion, and heavy NP shift. In some instances the types of processes involved in the derivation of the construction are controversial, as in the case of ellipsis and specific matrix vs. subordination phenomena (Meinunger, Schwabe, Winkler, see their contributions in this volume). The major theoretical questions relevant to focus-related constructions are: (i) How can the displacements involved in focus-related constructions be described? Are they part of the core syntactic operations of the grammar, in the sense that the syntax provides the optimal configuration for the interpretive components, phonology and semantics? Or are they best described as phonological processes operating in the phonological component? (ii) How are these displacements triggered? What role do morphological markers play? (iii) What role do prosodic aspects (intonation, phrasing, phonological weight) play in displacements of constituents and how do they interact with discourse notions? (iv) How does the focus movement hypothesis interact with the conflicting requirements of the interpretive components, such as ease of processing and interpretation on the one hand and reduction of complexity on the other? Focus-related constructions have long been claimed to encode particular information-structural functions: clefts, for instance, are often analyzed as focusing structures, other leftward movement has been associated with topicalization structures, and elliptical constructions are connected with givenness and anaphoric relations. Central to the analysis of these constructions has been the focus movement hypothesis, that is, the assumption that syntactic displacements are obligatorily associated with focus or topic interpretations. A systematic investigation of these phenomena, their derivation, intonation and interpretation is a prerequisite for a precise characterization of the architecture of focus.

2. Main goal of this collection The main goal of this collection is to investigate the architecture of focus from different perspectives. Research in the field of information structure in the last four decades has shown that the phenomenon of focus is highly complex. The theoretical approaches are manifold, and the data is highly sensitive. The central underlying idea of this collection is to put to test the theories developed within the framework of generative grammar in recent years and evaluate whether they can be conceived of

8

Valeria Molnar and Susanne

Winkler

as essential corner stones of an integrated theory of focus and the related information-structural notions. The main motivation for joining together the contributions in this volume lies in our conviction that an integrated theory of focus can only be achieved within the theoretical framework of generative grammar. W e have therefore brought together researchers working on focus and topic from different perspectives and started a discussion on the state-of-theart of the architecture of focus. Our aim is to present works which investigate relevant and highly current theoretical issues and empirical data from different languages, thereby contributing to a new architectural design of focus. W e have classified the approaches according to the main emphasis of each paper with respect to the interaction of individual components: The general section discusses the theoretical foundations of focus and information structure within grammar. The second section hosts papers which investigate the representation of focus, topic and contrast at the syntax-pragmatics interface. The third section is concerned with the phonological representation of focus, topic and contrast and its relation to meaning. The papers of the final section are concerned with different types of focus-related constructions - among them ellipsis, clefts, and subordinate clauses - in a variety of languages. Thus, the collection of papers reflects the state-of-the-art focus research.

3. The chapters

3.1. Theoretical foundations of focus and information structure The papers in this section investigate the theoretical foundations of focus pursuing two lines of research. First, they address the question of how focus and other discourse functions can be integrated into the grammatical model. Second, they explore the range of theoretical notions which are necessary to determine the discourse status of an expression at the interfaces. The main purpose of the model-theoretical research on information structure is to identify the levels of representation and discuss their relations. For example, approaches within the minimalist framework advocate different levels of representation. According to the standard variant of this theory a formal f(ocus)-feature is introduced at some level of the representation and causes the focus-marked phrase

Exploring the architecture of focus in grammar

9

to move to the specifier position of a corresponding functional focus head, which is projected in the left peripheral domain of the sentence. An alternative to the classic version of this feature-driven approach (cf. Szendroi 2004) allows the focus-marked phrases to move to other Abar-positions in the sentence structure, as well. In both versions of the feature-driven approach, the focus of an utterance is directly and unambiguously represented in the syntactic representation. Since Chomsky (1995) introduced the minimalist program and with it the theoretical requirements (i) that movement must have triggers, (ii) that formal features, such as focus and topic features, must be erased before interpretation, and (iii) that the inclusiveness principle, which requires that "outputs consist of nothing beyond properties of items of the lexicon (lexical features)" (Chomsky 1995: 225), must be observed, various approaches as to how these claims could be implemented have emerged. Erteschik-Shir (this volume) presents one of them. In her article she introduces the notions of topic and focus as part of the lexical selection. An alternative account of the information structural concepts, focus and topic, is set within the framework of Dynamic Syntax. Dynamic syntax, as discussed in the paper by Kempson, Cann, and Kiaer is like minimalism in that the level of logical form, represented as a tree structure, is the only syntactic level, but it differs from it, in the fact that dynamic syntax trees are transparent representations of semantic content which are incrementally built up following the left-right dynamics of natural language processing. The allocation of information structure to this single syntactic-semantic representational level is motivated by the dynamism of parsing. A strict modular view of the interpretation of information-structural notions is proposed by Costa and Figueiredo Silva (this volume), who argue against the syntactic encoding of focus and topic. Interpretation of discourse notions is thereby regarded as a pragmatics-internal process which operates on the structural options provided by the syntactic and phonological systems of different languages. The core question addressed by all articles in this volume concerns the interpretation of the notion of focus. In earlier focus theories two main theoretical lines of focus interpretation were pursued: The syntactically based structured meaning approach involving structural operations at the level of Logical Form (Jacobs 1983, von Stechow 1991) and the alternative semantics approach, which claims that focused constituents can be interpreted in situ (Rooth 1985). While it is still an open

10

Valeria Molnär and Susanne

Winkler

issue whether the movement theory of focus or the in situ theory is valid, Krifka (this volume) argues for the necessity of taking both alternatives into consideration in the analyses of certain phenomena. This volume presents two further novel syntactically based approaches to focus interpretation which constitute alternative versions to a unified account for the interpretation of the different types of focus. The first one is by Kenesei who argues in his article for identification as the common denominator of all focusing, while E. Kiss regards focusing as a function of the predication domain. Nomi Erteschik-Shir: On the architecture of topic and focus The paper by Erteschik-Shir investigates information structure primitives with an emphasis on focus and topic. Her major claim is that focus and topic are introduced as a part of the lexical selection. She argues against the assumption that focus and topic are heads that project their own phrases as assumed by Rizzi (1997), but rather views focus and topic as features which are optionally assigned to elements in the lexicon and enter the derivation very early on (cf. Winkler and Göbbel 2002). An advantage of her assumption that information-structural notions are encoded in the lexicon is that the violation of the inclusiveness principle can be avoided (cf. Zubizarreta 1998). A small set of specific rules determine the focus structure of the sentence, which functions as the input to phonological form and to logical form. Following this line of argumentation, she proposes that the dislocation operations of optional topicalizations are PF-operations. She bases her argument on an investigation of topicalization in Danish and Hebrew, which exhibit morphological marking for agreement and case and argues that topicalization processes are phonological dislocations. Ruth Kempson, Ronny Cann and Jieun Kiaer: Topic, focus, and the structural dynamics of language The paper by Kempson, Cann and Kiaer presents an outline of a dynamic syntax formalism designed to articulate aspects of information structure (topic/focus) without the need to resort to any additional structural representational level. The proposed model incorporates a left-to-right, incremental, monotonic tree growth process producing a single syntactico-semantic representation which encodes both topic and focus structure. The framework in which these generalizations are expressed takes the dynamism of parsing in real time as central to the

Exploring the architecture of focus in grammar

11

syntactic explanation. Phenomena captured include hanging topics, left dislocation, clitic left dislocation, pronoun doubling, topicalization and subject postposing. Joäo Costa and Maria Cristina Figueiredo Silva: On the (independence relation between syntax and pragmatics Costa and Figueiredo Silva investigate the interface relation between syntax and pragmatics. Based on evidence from European and Brazilian Portuguese, the authors argue that notions like focus and topic are not syntactically encoded. Rather syntax feeds the information-structural component. The investigation concentrates on micro-variation facts with respect to word order in answers to w/i-questions in the two varieties of Portuguese, which differ with respect to the types of answers they allow to different w/i-questions. They propose two interface-related hypotheses for the micro-variation observed: (i) There is no parameter specifically related to information structure configurationality; (ii) the following universal holds: Word order is used for codifying focused XPs, whenever syntax allows it. If syntax prohibits it, stress is used as a last resort. The results of their investigation show that the differences between the two languages cannot be grasped by incorporating informationstructural notions into syntax. Rather they assume a strictly syntactic view in a modular framework and propose that the variation is syntax driven and the information-structural interpretation must occur in the pragmatic component. Manfred Krifka: Association with focus phrases Krifka's paper discusses the grammatical architecture of the association with focus phenomenon, contrasting the structured meaning approach and the alternative semantics approach. Thereby, he concentrates on the question of whether association with focus is a structural operation involving L F movement, or an interpretational operation. The author argues for a hybrid association with focus approach which is more complex than either the simple alternative semantics or the simple structured meaning approaches. The central claim is that the association with focus phrases takes place via LF movement, while the association of focus within focus phrases occurs via the projection of alternatives. Three arguments provide more solid evidence for island restrictions in association with focus: explicit contrasts, multiple foci in syntactic islands and elliptical answers to questions. The paper shows that structured meanings

12

Valeria Molnär and Susanne

Winkler

are better designed than alternative semantics to represent association with focus. Istvän Kenesei: Focus as identification Kenesei's paper reexamines the interpretation of focus and puts forth the strong hypothesis that exhaustive/contrastive focus is characterized by identification, and that information focus involves identification as well. The author proposes dispensing with E. Kiss's distinction between information and contrastive focus, and claims that contrastive focus is a subset of information focus. More precisely, the two focus types differ only with respect to the sets they refer to: Information focus involves a subset relation, while contrastive focus makes use of a proper subset relation. In the case of contrastive focus, a complementary set is always created independently of the explicit or implicit character of the contrasting complementary set. Moreover, the paper pursues a related issue: Kenesei develops a uniform characterization of the set of semantic objects that can function as focus by basing these findings on a Fregean ontology. Katalin E. Kiss: Focusing is predication E. Kiss's paper offers a novel analysis of structural focus in Hungarian, arguing that focusing is a function of the predication domain. The basic hypothesis is that the focus occupies the specifier of a predicative projection (PredP) above VP, which also hosts the verbal particle functioning as a resultative, terminative, or locative secondary predicate. However, as focusing displays properties which differ from those of verbal particles or bare nominals, the author proposes that the exhaustive identification associated with structural focus is an instance of a different type of predication in that focus denotes a specificational predicate. The spec,PredP position is also the prosodically most prominent position of the sentence, hence its fillers (verbal particles or specificational predicates) are interpreted as the information focus. In this model, no focus projections or head movement of the verb are needed to derive focusing effects in Hungarian.

Exploring the architecture of focus in grammar

13

3.2. The representation of focus, topic and contrast at the syntax-pragmatics interface The architectural design of focus and information structure requires a clarification of the relation of focus to other relevant concepts which interact with focusing in different ways and at different levels. Besides the notion of topic, often regarded as the complementary notion to focus, the grammatical and interpretational relevance of an additional concept, the notion of contrast, has been recognized in recent linguistic research. The main goal of the second section of this volume is to explore the range of theoretical notions which are necessary to determine the discourse status of an expression at the interfaces. The articles argue for including the notion of contrast in the inventory of the basic information-structural notions in order to capture different syntactic, semantic and pragmatic phenomena. The section starts out with a paper by Molnär, who discusses the relation of contrast to focus and topic and demonstrates the relevance of contrast at the interfaces. Essentially two different notions of contrast as autonomous concepts of information structure are argued for: one intonationally marked, the other syntactically marked, both of which are superimposed on the notions of focus and topic. Frey's paper, which is a representative paper for the focus movement account, investigates the impact of contrast on syntactic movement to the left periphery in German. He claims that contrast is relevant for accounting for the interpretational potential of moved constituents and regards contrast-induced movement as one of three options of displacement to the left edge of the German sentence. The relevance of contrast at the left periphery is also the subject of Kaiser's paper, which explores the interaction of contrastive focusing with negation at the left periphery of the Finnish sentence. She argues for placing fronted negation in the same left peripheral polarity phrase as is also occupied by the contrastive constituent. Contrast is also regarded as essential by Puskas for the analysis of double negation in different languages. She argues for the restriction of double negation to cases where the negative quantifier has identical structural, semantic and pragmatic features to the contrastive topic, and thus functions as an instance of a contrastive topic.

14

Valeria Molndr and Susanne

Winkler

Valeria Molnär: Ott different kinds of contrast Molnär investigates the impact of contrast on linguistic structure and its interplay with other core notions of information structure. The main claim is that contrast is a linguistically relevant phenomenon compatible with both focusing and topicality and cannot merely be regarded as "the result of general cognitive processes referred to as 'conversational implicatures'" (Lambrecht 1994: 291). The linguistic relevance of contrast must, however, be differentiated depending on which languages and which types of constructions are taken into consideration. Contrast can not only serve as a feature for further distinctions within focus and topic, but it can also be superimposed upon these two relevant discourse notions. It is argued that two main types of contrast are superimposed on the functions of topic and focus: (i) the intonationally marked type, called I-CONTRAST, obligatorily realized by a (fall)-rise and connected with a special semantic/pragmatic interpretation (nonexhaustive identification), and (ii) the syntactically marked type, termed S-KONTRAST, in which case besides the prosodic highlighting the left peripheral position is essential for securing the discourse connection with competing (explicitly mentioned or contextually present) alternatives from the relevant set. Werner Frey: Contrast and movement to the German prefield In current syntactic theory, movement is supposed to take place to satisfy certain interface requirements. For left peripheral Α-bar movement the interface involved is the external interface with semantics/pragmatics (Rizzi 2004). Werner Frey's paper concentrates on movements to the prefield in German, here defined as a V2-language with an obligatory prefield position in front of the finite verb in declarative clauses of the clause. Frey investigates the following questions: (i) Which discourse functions do the phrases which are positioned in the German prefield have? (ii) At which stage are these functions assigned to constituents which occupy the prefield? He identifies three different types of left peripheral movements on the basis of the discourse functions expressed by the moved elements in the prefield: (i) A semantically and pragmatically vacuous movement which takes the highest constituent in the middle field, preserving the pragmatic property acquired in the middle field, (ii) base generation of certain adverbials that are not licensed clause internally and are not integrated in the proposition, and (iii) true A-bar movement, which results in the contrastive interpretation of the moved

Exploring the architecture of focus in grammar

15

item. The paper argues that these three options each involve different structural positions in the C-domain of the German clause, which is split up into three different functional projections: CP, KontrP and FinP. Elsi Kaiser: Noncanonical negation and information structure in Finnish Kaiser's paper is concerned with the syntax-pragmatics interface and explores the word order restrictions on Finnish sentences with fronted negation and contrast. The Finnish sentence has on the left periphery a designated position for prosodically focused contrastive constituents. However, preposing negation from its canonical postsubject position to the sentence initial position has a striking effect on the word order and changes the canonical pattern. The main claim is that preposed negation occurs in a left peripheral polarity phrase and is obligatorily associated with the contrastive constituent. Kaiser's account, developed in the optimality theoretic framework, captures the effects of negation by treating it as similar to kontrast, thereby explaining the disappearance of the original left periphery focus position (which should presumably be to the right of negation) in sentences with fronted negation. Genoveva Puskas: Double negation and information structure: Somewhere between topic and focus The paper by Puskas examines the properties of double negation (DN) in Hungarian and argues for the claim that DN is an instance of Contrastive Topic. The DN reading is claimed to only be available if the relevant negative quantifier has the same phonological, semantic and pragmatic properties as the Contrastive Topic (CT): The typical intonational pattern of CT is the L*H contour, and the semantic properties are related to the triggering of the set of possible alternatives with some weakened exclusive feature. The relevant negative quantifier is not only realized with the fall-rise pattern, but its semantic and pragmatic contribution to the interpretation of the sentence seems to match that of DN. Double negation sentences are often claimed to contain some sort of secondary predication. Thus, the semantic and pragmatic dependency of DN on the elements contributing to the main negation suggests that the negative expression responsible for the DN reading is somehow dependent, or parasitic, on the first negative relation.

16

Valeria Molnär and Susanne

Winkler

3.3. The phonological representation of focus, topic and contrast and its relation to meaning The traditional two-interface hypothesis requires that phonology interact with meaning only via syntax. However, it has long been observed that some focus-related phenomena may pose a challenge to this hypothesis (e.g., van Riemsdijk and Williams 1986), in particular those which are based on concepts of anaphora or contrast and do not involve syntactic movement. Therefore, two controversial theories can be distinguished which address the interaction between the phonology of focus, topic and contrast with meaning: First, the traditional syntactic theory, which claims that prosody and interpretation interact via the syntactic component (traditional position defended, e.g., in Bresnan 1971, 1972, Chomsky 1972, Jackendoff 1972, Selkirk 1984, 1995, Winkler 1997, Zubizarreta 1998, among many others); second, the phonology-semantics theory of focus, which diverges from the traditional two-level hypothesis and allows the direct correspondence between sound and meaning (Schwarzschild 1999, Büring, this volume, Reinhart 1995, 1996, Neeleman and Reinhart 1998, Szendröi 2001). The choice between these two theories cannot be made on conceptual grounds alone. In the papers in this section, linguistic evidence and in particular empirical data provide support for both the traditional syntactic theory and the direct correspondence account. This section starts out with Büring's contribution, which is a representative paper for the direct correspondence between sound and meaning. It is followed by the paper by Mehlhorn and Richter, which provides experimental data for the differentiation of contrastive and emphatic accent assignment. This section closes with a paper by Hedberg, which discusses topic and focus controversies on the basis of intonational data from English, Japanese and Korean. Daniel Büring: Focus projection and default prominence Büring's paper addresses the question of whether focus projection rules are needed in the grammar. The paper consists of three parts. In the first part, he provides a summary of Selkirk's (1995) system of rules, which can be conceived of as functions that establish a systematic relation between phonological prominence and the interpretation of pragmatic focus in a specific context. By discussing some common misconceptions of this system, the question arises as to how viable the correlation be-

Exploring the architecture offocus in grammar

17

tween focus marking and pragmatic interpretation actually is. He identifies two processes, namely horizontal and vertical focus projection. The former determines under which circumstances a focus-marked terminal node can remain unaccented; the latter determines which constituents within a larger focus constituent must be focused for structural reasons. The central part of the paper shows that the empirical generalizations about vertical focus projection are inaccurate, and that they should be eliminated. Biiring then investigates the horizontal focus projection rules and shows that the main effect should be thought of as a consequence of default prominence assignment. After sketching a basic system that illustrates how default prominence assignment and focus interact, he proposes a phonologically based theory without any focus projection rules. The theory can be informally described as a focus-to-prominence account, which interacts with general rules of prosodic phrasing in the respective language. Grit Mehlhorn and Nicole Richter: Focus on contrast and emphasis: Evidence from prosody Mehlhorn and Richter's paper investigates the syntactic and prosodic characteristics of emphasis and contrast and their role in the theory of information structure. The empirical basis of the paper draws primarily on speech samples from Russian which have been elicited in production experiments. The main hypothesis is that emphasis and contrast must be differentiated. Emphasis is defined as a prosodic highlighting technique that indicates the emotional involvement of the speaker. Contrast, however, is narrowly defined as occurring in corrections of explicit utterances or in corrections of a presupposition of the situational context. The authors further distinguish between emphatic and focused utterances. They base this distinction on straightforward and instructive tests and measurements of pitch and syllable length in emphatic and focused utterances. Their results show that only contrast, but not emphasis, can be identified as belonging to particular syllables in the prosodic realization of the sentence. Thus, while contrast seems to show features that are determined by the grammatical basis of information structure, emphasis displays a range of features that mark it as a means that is external to the prosodic system of a given language, i.e., it does not seem to be part of information structure. Emphasis thus seems to override focus-background structure.

18

Valeria Molndr and Susanne

Nancy Hedberg: Topic-Focus

Winkler

Controversies

In her paper Hedberg analyzes seven instances of so-called topic-focus controversies. She aims at a clarification of the phenomena in question and a critical discussion of the underlying theories of information structure. The topic-focus controversies relate to aspects of information structure for which conflicting analyses have been suggested that go beyond terminological disagreements. The seven topic-focus controversies are the following: (i) The role of the L+H* accent: Does it mark (all and only) contrastive topics, or does it mark contrast in general? (ii) the role of the topic markers wahiun in Japanese and in Korean, which have both a thematic and a contrastive interpretation: Is contrastive wa/ nun restricted to topics or can it also mark foci? (iii) the role of question words in constituent questions: Do they constitute the topic or the focus of the question? (iv) the question of whether sentences with a single fall-rise accent constitute a theme-only utterance; (v) the role of L+H* accents in negative sentences: Do they mark contrastive topics or foci? (vi and vii) the question of whether the cleft constituent in clefts and inverted pseudoclefts always represents the focus. For each of the seven questions, the author carefully presents the pros and cons of the controversial analyses and supports her arguments with well-chosen natural examples from a corpus of prosodically annotated data.

3.4. Focus-related constructions in different languages This section investigates different focus-related constructions in different languages. Focus-related constructions, defined above as syntactic dislocations with a salient phonological realization and a specific information-structural interpretation, challenge the theory of generative grammar in at least two respects: First, focus-related constructions often seem to constitute an optional phenomenon, which pose a problem to the minimalist program. Second, focus-related constructions are often associated with a marked intonation and specific discourse functions and are therefore claimed to form stylistic constructions (Rochemont 1978) or focus constructions (Taglicht 1984, Rochemont and Culicover 1990), which find an explanation outside of the core syntactic component. The papers collected here address these challenges employing a uniform strategy. First, the underlying working hypothesis of all the papers is that there are no optional phenomena in language. Second, each pa-

Exploring the architecture of focus in grammar

19

per aims at a more precise description of the focus-related effects which are brought about by the construction under discussion. For example, Winkler's paper proposes that elliptical constructions in English and German are only seemingly optional and have clearly defined information-structural effects. She distinguishes two types of ellipses, and proposes that the different interpretations are directly related to the type of derivation of the elliptical clause in the syntactic component. Schwabe's contribution supports this nonoptionality hypothesis with a thorough empirical investigation of elliptical dim-clauses in German. The general result is that elliptical processes are not optional, but serve to highlight contrastively focused remnants by eliding redundant or given material. The new empirical research field which targets the discourse functions of matrix and embedded clauses is also central to Meinunger's article. He concentrates on the discourse status of subordinate clauses in German and shows that seemingly optional alterations are directly connected to their information-structural status. Following the central nonoptionality hypothesis, which links the papers in this section together, Steube investigates the background-focus partition of German sentences and discusses the influence of focus on the interpretation of DPs and PPs. The contributions by Gundel and Huber argue for the existence of different types of cleft sentences in different languages, which exhibit different discourse functions. The final paper of this section raises issue with the nonoptionality-hypothesis on the basis of data from Hausa. Hartmann's paper on different focusing strategies in Hausa shows that our analytic linguistic tools to describe information-structural effects still need to be sharpened. Susanne Winkler: Ellipsis at the interfaces: An information-structural proposal Winkler's paper investigates the phenomenon of ellipsis and its relation to information structure in the theory of grammar, which she developed in more detail in Winkler (2005). Her basic claim is that there are two types of elliptical constructions, one sentence-bound, the other discourse-bound, which differ with respect to their syntactic representation and their correlating information-structural functions. On the basis of intonational evidence, she argues that the former type of ellipsis is derived by syntactic movement, while the latter type is determined by discourse factors. Concentrating on the architecture of focus, she proposes that the contrastive interpretation is directly connected to syntactic movement to the periphery of the phase, while syntactic constructions

20

Valeria Molnär and Susanne Winkler

which do not involve movement of constituents are dependent on discourse factors for their information-structural interpretation. In accordance with the minimalist program, which allows a richer flow of information between the modules, she sketches a grammatical model in which the information-structural component plays a crucial role in distinguishing between syntactically driven and functionally driven ellipses. The central idea is that the information-structural component functions as a relay station between syntax and the interpretive components on the one hand and between syntax and pragmatics on the other. Thus, the proposal shows how the traditional two-interface hypothesis of grammar can be accommodated to the empirical evidence by enriching the interface interaction without giving up the conceptual advantages of modularity. Kerstin Schwabe: Elliptical dass-clauses in German Schwabe discusses the phenomenon of independent dass-clauses in German. Based on an in-depth semantic and information-structural analysis she claims that independent (fass-clauses are elliptical fragments. A comparison to complement dass-clauses shows that independent dassclauses are focus constructions which depend on contextual triggers like w/z-questions. Schwabe proposes an elliptical account for dass-clauses, which adds additional support to the research on ellipsis, hypothesizing that focused material must be phonologically realized, while given material, such as the information of the matrix clause, need not be. However, her analysis differs from syntactic fragment accounts like Merchant's (2004). While Merchant proposes a syntactic deletion account in which the syntactic structure of the fragment is fully projected but not phonologically realized, Schwabe argues for a basegeneration account. She proposes that in the case of dass-clauses only the minimal syntactic structures with empty phonological heads are generated. Thus, her paper contributes to ellipsis research at the semantics-pragmatics interface. Andre Meinunger: The discourse status of subordinate sentences and some implications for syntax and pragmatics Meinunger investigates the information structure of subordinate sentences in German, English and Romance. The central hypothesis with respect to German is straightforward: Embedded V2-clauses constitute assertions which introduce new information into the discourse, as in the sentence Ich glaube, Bens Lehrer hat RECHT. Alternatively, it is possi-

Exploring the architecture of focus in grammar

21

ble to use a subordinate verb-final dass-clause, as in the example Ich glaube, dass Bens Lehrer RECHT hat. If, however, the proposition is given, anaphoric or inferable, V2 is excluded and a (iass-clause is required, as signaled by the contrast between the ungrammatical sentence Ich glaube nicht, Bens Lehrer hat Recht and the grammatical sentence Ich glaube nicht, dass Bens Lehrer Recht hat. With respect to Romance, Meinunger observes that there is a direct correspondence between those grammatical phenomena that block V2 in German(ic) and those that trigger subjunctive mood in Romance. More precisely, he connects the givenness distinction to the presence vs. absence of illocutionary force - a factor which needs to be further investigated in different languages. Anita Steube: The influence of operators on the interpretation of DPs and PPs in German information structure Steube's article is concerned with the background-focus partition of German sentences. The syntactic relations are explained on the basis of Haider's (1997) theory of syntax. The paper aims at an explanation of the subdivision of operators and their positions in fully focused subordinate German clauses. From left to right there are: Nonpropositional operators (sentence mood, attitudinal adverbials, verificational adverbials) and propositional operators (negation and the modal verb können {can)). As long as only wide-scope relations are considered, the scopal relations of the operators agree with their c-command relations. In German, the movement of DPs and PPs has semantic and information-structural effects: DPs and PPs may leave the scope of negation and of verificational operators without being background constituents. When - depending on context - the region between sentence mood and attitudinal operators is reached, the DPs and PPs are to be found in the background. The paper explains what interpretations indefinite DPs, definite DPs, and names have in the scope of which operator(s). The first and the last sections elucidate how grammar can be considered part of a conceptual model that explains how information-structural variation comes about, and the effects it has on each grammatical level. Jeanette Κ. Gundel: Clefts in English and Norwegian: Implications for the grammar-pragmatics interface The paper by Gundel addresses the differences in the use and frequency of (Zi-)clefts in English and Norwegian. The main intention is to cor-

22

Valeria Molnär and Susanne

Winkler

roborate a previous thesis by the author: She argues that the fact that clefts display a higher frequency in Norwegian than in English is due to the strong preference in Norwegian to map information structure onto syntactic structure. Clefts constitute ideal constructions for this purpose. Especially the need in Norwegian to encode presupposed information into a separate syntactic unit - the cleft clause - triggers the more frequent use of cleft constructions in this language, the author claims. In English, in contrast, it is claimed that there is no such need. Stefan Huber: The complex functions of it-clefts Huber's paper investigates the functional potential of clefts in three Germanic languages, German, English and Swedish. Based on a tripartite model of information structure, which differentiates between the layers of topic/comment, theme/rheme and background/focus, he claims that clefts are not exclusively focusing constructions. In addition to the three well-known functions of clefts discussed in the literature (contrastive focusing, continuous topic clefts and all-new-clefts), clefts can be used in thetic sentences, in multiple foci structures and in sentences containing I-topics (also called contrastive topics). Huber argues, however, that there are differences between German, English and Swedish as to the potential of clefts. While clefts with multiple foci and clefts with I-topics are possible in all three languages investigated, thetic clefts are only possible in Swedish. Katharina Hartmann: Focus constructions in Hausa Hartmann analyzes focus constructions in Hausa, a Chadic language spoken mainly in the north of Nigeria. Chadic languages are tone languages which differ from intonational languages with respect to the realization of topic and focus. The paper examines the syntax of two strategies of focus marking in this language, the ex situ and the in situ strategy. That is, the focus appears either in situ, or it is moved (the ex situ strategy). The target of focus movement can be the left or right periphery of the clause, as well as the postverbal position (cf. Tuller 1992). She proposes that the focus constituent moves to the specifier of a focus phrase. The head of FocP is filled by the focus marker nee. Topics, however, are basegenerated in a left peripheral position and can therefore violate the movement constraints. She examines newspaper texts and notices that only a small percentage of all clauses employ the ex situ focus strategy. Under the assumption that every sentence contains a fo-

Exploring

the architecture

of focus in grammar

23

cus, this result points towards the possibility that Hausa allows an in situ focus strategy with all constituents except subjects. The paper points out interesting interpretive relations which will be the subject of further research.

4. Conclusion The above discussion of the contributions shows that the papers in this volume target the architecture of focus in at least five important respects: First, all papers investigate the notion of focus and related informationstructural notions like topic and contrast in their own right. Second, they attempt to analyze focus and information structure as an interface phenomenon. Third, they look at different types of linguistic evidence and at a wide range of data at the phonological, syntactic and semantic levels, as well as evidence from different languages. Fourth, the papers build on current investigations in the theory of grammar and information structure and mirror an important shift in the theoretical characterization of focus. Finally, the papers reflect the interaction of a group of researchers whose recent investigations into the architecture of focus and information structure have brought important results.

References Adger, David, Cecile De Cat, and George Tsoukas (eds.) 2004 Peripheries: Syntactic Edges and Their Effects. (Studies in Natural Language and Linguistic Theory 59.) Dordrecht/Boston: Kluwer. Bolinger, Dwight 1961 Ambiguities in pitch accent. Word 17: 309-317. Bresnan, Joan 1971 Sentence stress and syntactic transformations. Language 45: 257-

281. 1972 Stress and syntax: A reply. Language 48: 326-342. Breul, Carsten 2004 Focus Structure in Generative Grammar: An Integrated Syntactic, Semantic and Intonational Approach. (Linguistik Aktuell 68.) Amsterdam/Philadelphia: Benjamins.

24

Valeria Molnär and Susanne

Winkler

Büring, Daniel 1997 The Meaning of Topic and Focus: The 59th Street Bridge Accent. (Routledge Studies in German Linguistics 3.) London: Routledge. this vol. Focus projection and default prominence. In Valeria Molnär, and Susanne Winkler (eds.), 321-346. Bußmann, Hadumod 1990 Lexikon der Sprachwissenschaft. Stuttgart: Kröner. Chafe, Wallace L. 1976 Givenness, contrastiveness, definiteness, subjects, topic, and points of view. In Subject and Topic, Charles N. Li (ed.), 25-56. London: Academic Press. Chomsky, Noam 1965 Aspects of the Theory of Syntax. (Special Technical Report 11.) Cambridge, Massachusetts: MIT Press. 1971 Deep structure, surface structure and semantic interpretation. In Semantics: An interdisciplinary reader in philosophy, linguistics and psychology, Danny Steinberg and Leon Jakobovits (eds.), 183-216. Cambridge: Cambridge University Press. 1972 Deep structure, surface structure, and semantic interpretation. In Studies on Semantics in Generative Grammar, Noam Chomsky (ed.), 69-119. (Janua Linguarum: Series Minor 107.) The Hague: Mouton. 1995 The Minimalist Program. (Current Studies in Linguistics 28.) Cambridge, Massachusetts: MIT Press. Costa, Joäo, and Maria Cristina Figueiredo Silva this vol. On the (in)dependence relation between syntax and pragmatics. In Valeria Molnär, and Susanne Winkler (eds.), 83-104. Culicover, Peter W., and Michael Rochemont 1983 Stress and focus in English. Language 59: 123-165. Dretske, Fred I. 1972 Contrastive statements. The Philosophical Review 81: 411-437. Drubig, Hans Bernhard 1994 Island constraints and the syntactic nature of focus and association with focus. Arbeitspapiere des Sonderforschungsbereichs 340 'Sprachtheoretische Grundlagen für die Computerlinguistik' 51. Universität Stuttgart/Universität Tübingen. 2003 Towards a typology of focus and focus constructions. Linguistics 41 (1): 1-50. to appear Phases and the typology of focus constructions. In On Information Structure, Meaning and Form, Kerstin Schwabe, and Susanne Winkler (eds.). (Linguistik Aktuell.) Amsterdam/Philadelphia: Benjamins.

Exploring the architecture of focus in grammar

25

Drubig, Hans Bernhard, and Wolfram Schaffar 2001 Focus constructions. In Language Typology and Language Universals: An International Handbook, Martin Haspelmath, Ekkehard König, Wulf Oesterreicher, and Wolfgang Raible (eds.), Vol. 2, 1079-1104. (Handbücher zur Sprach- und Kommunikationswissenschaft 20/2.) Berlin: de Gruyter. Erteschik-Shir, Nomi this vol. On the architecture of topic and focus. In Valeria Molnär, and Susanne Winkler (eds.), 33-57. Frey, Werner this vol. Contrast and movement to the German prefield. In Valeria Molnär, and Susanne Winkler (eds.), 235-264. Gundel, Jeanette Κ. 1988 Universals of topic-comment structure. In Studies in Syntactic Typology, Michael Hammond, Edith Moravcsik, and Jessica Wirth (eds.), 209-239. John Benjamins, this vol. Clefts in English and Norwegian: Implications for the grammar-pragmatics interface. In Valeria Molnär, and Susanne Winkler (eds.), 517-548. Haider, Hubert 1997 Projective economy. On the minimal functional structure of the German clause. In Syntactic Problems - Problematic Syntax, Werner Abraham, and Elly van Gelderen (eds.), 83-103. Tübingen. Halliday, Michael A. K. 1967 Notes on transitivity and theme in English, part 2. Journal of Linguistics 3: 199-244. Hartmann, Katharina this vol. Focus constructions in Hausa. In Valeria Molnär, and Susanne Winkler (eds.), 579-607. Hedberg, Nancy this vol. Topic-focus controversies. In Valeria Molnär, and Susanne Winkler (eds.), 373-397. Hockett, Charles Francis 1958 A Course in Modern Linguistics. New York: Macmillan. Huber, Stefan this vol. The complex functions of ii-clefts. In Valeria Molnär, and Susanne Winkler (eds.), 549-578. Jackendoff, Ray S. 1972 Semantic Interpretation in Generative Grammar. (Studies in Linguistics 2.) Cambridge, Massachusetts: MIT Press.

26

Valeria Molndr and Susanne

Winkler

Jacobs, Joachim 1988 Fokus-Hintergrund-Gliederung und Grammatik. In Intonationsforschungen, Hans Altmann (ed.), 89-134. (Linguistische Arbeiten 200.) Tübingen: Niemeyer. Kaiser, Elsi this vol. Noncanonical negation and information structure in Finnish. In Valeria Molnär, and Susanne Winkler (eds.), 265-290. Kempson, Ruth, Ronny Cann, and Jieun Kiaer this vol. Topic, focus and the structural dynamics of language. In Valeria Molnär, and Susanne Winkler (eds.), 59-82. Kenesei, Istvän this vol. Focus as identification. In Valeria Molnär, and Susanne Winkler (eds.), 137-168. Kiss, Katalin Ε 1995 Introduction. In Discourse Configurational Languages, Katalin E. Kiss (ed.), 3-27. New York: Oxford University Press. 1998 Identificational focus versus information focus. Language 74: 245273. this vol. Focussing as predication. In Valeria Molnär, and Susanne Winkler (eds.), 169-193. Krifka, Manfred this vol. Association with focus phrases. In Valeria Molnär, and Susanne Winkler (eds.), 105-136. Ladd, Robert D. 1996 Intonational Phonology. (Cambridge Studies in Linguistics 79.) Cambridge: Cambridge University Press. Lambrecht, Knud 1994 Information Structure and Sentence Form. (Cambridge Studies in Linguistics 71.) Cambridge: Cambridge University Press. Meinunger, Andre this vol. On the discourse impact of subordinate clauses. In Valeria Molnär, and Susanne Winkler (eds.), 459-587. Merchant, Jason 2004 Fragments and ellipsis. Linguistics and Philosophy 27 (6): 661-738. Molnär, Valeria. 1991 Das TOPIK im Deutschen und im Ungarischen. Stockholm: Almquist & Wiksell International (= Lunder Germanistische Forschungen 58). 2002 Contrast - From a contrastive perspective. In Information Structure in a Crosslinguistic Perspective, Hilde Hasselgärd, Stig Johansson, Bergljot Behrens, and Cathrine Fabricius-Hansen (eds.), 147-161. Amsterdam/New York: Rodopi.

Exploring this vol.

the architecture

of focus in grammar

27

On different kinds of contrast. In Valeria Molnär, and Susanne Winkler (eds.), 197-233. Neeleman, Ad, and Tanya Reinhart 1998 Scrambling and the PF interface. In The Projection of Arguments: Lexical and Compositional Factors, Miriam Butt, and W i l h e l m Geuder (eds.), 309-353. (Center for the Study of Language and Information Lecture Notes 83.) Stanford: Center for the Study of Language and Information Publications. Puskas, Genoveva this vol. Double Negation and information structure: Somewhere between topic and focus. In Valeria Molnär, and Susanne Winkler (eds.), 291-317. Reinhart, Tanya 1982 Pragmatics and Linguistics: An Analysis of Sentence Topics. Bloomington: Indiana University Linguistic Club Publications. 1995 Interface Strategies. (OTS Working Papers.) Utrecht: Research Institute of Language and Speech. 1996 Interface economy: Focus and markedness. In The Role of Economy Principles in Linguistic Theory, Christopher Wilder, Hans-Martin Gärtner, and Manfred Bierwisch (eds.), 146-169. (Studia Grammatica 40.) Berlin: Akademie Verlag. Richter, Nicole, and Grit Mehlhorn this vol. Focus on contrast and emphasis: Evidence from prosody. In Valeria Molnär, and Susanne Winkler (eds.), 347-371. Riemsdijk, Henk van, and Edwin Williams 1986 Introduction to the Theory of Grammar. Cambridge, Massachusetts: M I T Press. Rizzi, Luigi 1997 The fine structure of the left periphery. In Elements of Grammar, Liliane Haegeman (ed.), 281-337. (Kluwer International Handbooks on Linguistics 1.) Dordrecht: Kluwer. 2004 On the form of chains: Criterial positions and E C P effects. Ms., University of Siena. Rochemont, Michael S. 1978 A theory of stylistic rules in English. University of Massachusetts, Amherst: Ph. D. diss. 1986 Focus in Generative Grammar. (Studies in Generative Linguistic Analysis 4.) Amsterdam/Philadelphia: Benjamins. Rochemont, Michael S., and Peter W. Culicover 1990 English Focus Constructions and the Theory of Grammar. (Cambridge Studies in Linguistics 52.) Cambridge, Massachusetts: Cambridge University Press.

28

Valeria Molndr and Susanne

Winkler

Rooth, Mats Ε. 1985 Association with focus. University of Massachusetts, Amherst: Ph. D. diss. Sasse, Hans-Jürgen 1987 The thetic/categorical distinction revisited. Linguistics 25: 511-580. Schwabe, Kerstin this vol. Elliptical i/ass-clauses in German. In Valeria Molnär, and Susanne Winkler (eds.), 429-457. Schwarzschild, Roger 1999 GIVENness, avoidF and other constraints on the placement of accent. Natural Language Semantics 7: 141-177. Selkirk, Elizabeth O. 1984 Phonology and Syntax. (Current Studies in Linguistics 10.) Cambridge, Massachusetts: M I T Press. 1995 Sentence prosody: Intonation, stress, and phrasing. In The Handbook of Phonological Theory, John A. Goldsmith (ed.), 550-569. (Blackwell Handbooks in Linguistics 1.) Cambridge, Massachusetts: Blackwell. Steube, Anita this vol. The influence of operators on the interpretation of DPs and PPs in German information structure. In Valeria Molnär, and Susanne Winkler (eds.), 489-516. Szabolcsi, Anna 1997 Quantifiers in pair-list readings. In Ways of Scope Taking, Anna Szabolcsi (ed.), 311-347. (Studies in Linguistics and Philosophy 65.) Dordrecht: Kluwer. Szendroi, Kriszta 2001 Focus and the syntax-phonology interface. University College London: Ph. D. diss. 2004 Introduction. In Focus and the interaction between syntax and pragmatics, D. Bury, K. Fraud, R. Horsey, and K. Szendroi (eds.), Special issue of Lingua 114 (3): 229-254. Taglicht, Josef 1984 Message and Emphasis: On Focus and Scope in English. (English Language Series 15.) London: Longman. Tuller, Laurice A. 1992 The syntax of postverbal focus constructions in Chadic. Natural Language and Linguistic Theory 10: 303-334. Vallduvi, Enric, and Maria Vilkuna 1998 On theme and contrast. In The Limits of Syntax, Peter Culicover, and Louise McNally (eds.), 79-108. (Syntax and Semantics 29.) San Diego: Academic Press.

Exploring the architecture of focus in grammar

29

Winkler, Susanne 1997 Focus and Secondary Predication. (Studies in Generative Grammar 43.) Berlin/New York: Mouton de Gruyter. 2005 Ellipsis and Focus in Generative Grammar. (Studies in Generative Grammar 81.) Berlin/New York: Mouton de Gruyter. this vol. Ellipsis at the interfaces: An information-structural proposal. In Valeria Molnär, and Susanne Winkler (eds.), 401-428. Winkler, Susanne, and Edward Göbbel 2002 Focus, P-movement and the NSR: A view from Germanic and Romance. Linguistics 40 (6): 1185-1242. Zubizarreta, Maria Luisa 1998 Prosody, Focus and Word Order. (Linguistic Inquiry Monograph 33.) Cambridge, Massachusetts: MIT Press.

Part 1: Theoretical foundations of focus and information structure

On the architecture of topic and focus* Nomi

Erteschik-Shir

Abstract The information structure primitives, Topic (top) and Focus (foe) play a central role in grammar, yet it is difficult to make their properties conform to those of other linguistic categories. In this paper, I argue that top and foe are introduced as a part of lexical selection. Rather than assuming that top and foe are heads that project their own phrases (as in e.g., Rizzi 1997), I view top/foe as features which are optionally assigned to lexical items. On a par with (^-features, they may percolate to the maximal projection of the lexical item they are assigned to. I show how such assignment constitutes the focus structure of a sentence. This focus structure, in turn, predicts phonological properties (among them intonation) as well as interpretative properties (among them quantifier scope). It is further shown that dislocation, in this framework, is best seen as a property of PF.

1. top/foe and inclusiveness The notions of Topic (top) and Focus (foe) play a pervasive role in grammar and interact with phonology (intonation), morphology (top/foe markers), syntax (linear order), and interpretation (quantifier scope). These notions are also critical in information structure in that they constrain the possible sequence of sentences in discourse. 1 In spite of the many roles played by top/foe there is little agreement as to how and where they are introduced in the grammar. It has often been pointed out (e.g., Zubizaretta 1998: 30) that since top/foe are supersegmental, they cannot be lexical and if they are not introduced lexically they violate the condition of inclusiveness (Chomsky 1995), since they will have to be added as annotation in the derivation. 2

34

Nomi

Erteschik-Shir

One way of getting around this is to derive focus from stress which, in turn, is assigned to syntactic configurations (e.g., Selkirk 1984, Cinque 1993, Reinhart 1995, 1996, Szendröi 2001). One of the main problems with this approach is that no account of top (and its interaction with focus) is provided. 3 Other problems include the necessity to add special pragmatic rules to distinguish information focus and contrastive focus (e.g., Cinque 1993). Alternatively top/foe are analyzed as projecting functional features (e.g., Rizzi 1997). These projections are introduced to attract constituents to the left periphery and thus explain word order. There are many problems with this, most importantly that no account is provided for the top/foe status of elements left in situ. Here I present a new approach which introduces top/foe features in the grammar without violating the principle of inclusiveness: by analogy with φ-features, top/foe features are selected from the lexicon. They differ from φ-features, however, in not being associated with particular categories and by not being required. If top/foe are associated with a head, the feature can optionally percolate upwards on a par with phi-features. Topic and Focus projections are thus implemented syntactically according to purely syntactic principles of feature projection and intonation is derived from this structure. 4 No special constraints, lexical or otherwise are needed as is the case in current theories of focus projection in which focus domains are derived from intonation.

1.1. The derivation of focus-structure Focus-structure (f-structure) is thus derived by free selection and association of top/foe features. Each selection of a lexical item licenses an optional assignment of a top or foe feature, (lb) illustrates the set of assignments appropriate to the discourse in (la): the subject is assigned top and the object is assigned foe. 5 These features percolate to the dp with no further percolation possible, (lc) is the merged (rough) tree structure, what I refer to as the f-structure of the sentence. 6 (This structure and the ones below are grossly simplified. Most functional projections, including ip and cp, are ignored for presentational purposes.) (1)

a. Q: What did John eat? A: He ate the cake.

On the architecture of topic and focus b. select select select select

'cake' 'the' 'ate' 'he'

35

assign [foe] -> no assignment no assignment assign [top]

vp he[top] ate the

cake[foc]

In view of the fact that the question identifies 'John' as the topic and forces focus assignment on 'cake', the answer to the w/z-question, the fstructure in ( l c ) is indeed the only appropriate one in the context of this question. How context and f-structure are matched is the topic of the next section.

1.2. Structure and Context The matching of context to f-structure is handled by an extra grammatical interpretive system defined in Erteschik-Shir (1997). In this system top/foe each trigger instructions to manipulate a stack of filing cards, each of which represent a referent available in the discourse. The rules apply to referential constituents within top/foe domains: (2)

F-structure

Rules

a. T O P I C instructs the hearer to locate on the top of his file an existing card with the relevant label. b. F O C U S instructs the hearer to either i) open a new card and put it on the top of the file. Assign it a new label (for an indefinite) or ii) locate an existing card and put it on the top of the file (for a definite)

36

Nomi

Erteschik-Shir

c. U P D A T E instructs the hearer to enter the focus on the topic card and then to copy all entries to all cards activated by the focus rule. 7 These rules apply to the f-structure of (1) in (3) as follows: (3)

Het0p

likes

[cake]foc.

1. Pick the card for 'John' from the top of the file, (top) 2. Enter "he likes cake" on the card for 'John', (update) 3. Open a new card, label it 'cake'. Put it on top of the file (foe i.) 4. Enter "John likes e" on this card, (update) The following cards are now on top of the file and are available as future topics: (4)

'John' he likes cake

'cake' John likes e

Topics are selected from the set of referents previously introduced in the discourse. Topics are therefore necessarily specific: they identify an element in the common ground that the sentence is about. This follows from the topic rule. It also follows that only potentially referential elements, such as DPs, can be assigned topic status. Pronouns, in this system, are necessarily interpreted as topics. This is accomplished by entering the pronoun on a card available on top of the file with a matching heading. If such a card is not available, the pronoun is not interpretable, as expected. As will be shown immediately below (in the discussion of example (7)), contrastive elements are marked both top and foe. Contrastive pronouns are therefore classified as topics even though they are stressed due to their foe feature. Following Strawson (1964) and Reinhart (1981), topics are defined as the 'address' in a file system under which sentences are evaluated. This view of topics as pivotal in truth value assignment has repercussions for semantics as indicated in section 3. 8 It follows from the Focus rule in (2b) that foci are necessarily new in the immediate discourse, i.e., the card for the foe is positioned on top of the file, hence cannot be located there already.

On the architecture of topic and focus

37

From these definitions of top and foe, it also follows that a sentence in which no top nor foe features have been selected will be ruled out. A sentence with no topic is uninterpretable and a sentence with no focus is uninformative. In order to see how lexical assignment of top/foe correlates with these rules, we examine a few more sample derivations: The sentence analyzed in (1) can also merge a different f-structure as shown in (5): (5)

a. What did John do? He ate the cake.

b. select 'cake' select 'ate' select 'he' c.

no assignment assign [foe] -> assign [top]

vp he[tof ateffoc]

dp the

cake

Here foe percolates to the vp rendering vp focus. The fact that this fstructure is appropriate in the context of the question in a and not in the context of the question in (5a), for example, follows from the f-structure rules. Here again the card for 'John' is positioned on top of the file, (by the question or by previous context) allowing its topic status in the following sentence. The question further requires vp focus. The question in (la) similarly defines 'John' as a topic. However, due to the fact that the verb is also introduced in the question it is presupposed in the answer, hence excluded from foe. 9 In the following example, the object is assigned top: (6)

a. What happened John ate it.

to the

cake?

38

Nomi Erteschik-Shir b. select 'it' select 'ate' select 'John' c.

assign [top] assign [foe] no assignment

vp[foc] John ate[foc] it[top]

This example is symmetric to the previous one: in both cases the topic is embedded in the focus domain, yet in the former case if the topic is removed, the remaining focus is a constituent, in the latter it is not. Since the focus rule applies vacuously to the topics contained in it, the effect is that what is interpreted as a focus in the latter case is not a constituent. It follows that even though a syntactic assignment of focus to a nonconstituent is impossible we in fact derive a focus interpretation of a nonconstituent with a purely syntactic assignment procedure. F-structure assignments can also result in a 'subordinate' f-structure. (7) shows how foe embedded in top is interpreted contrastively. In this case, top is contextually construed as a set of two alternatives and foe on one of these alternatives eliminates the other. (7)

a. What did John eat, the cake or the ice-cream? He ate the cake (not the ice-cream). b. select select select select c.

'cake' 'the' 'ate' 'he'

assign [foe] assign [top] no assignment -> assign [top]

vp he[top ate

dp[top]/[foc]

the[top] cake[foc] The contrastive interpretation depends on the contextual availability of a contrast set, i.e., a set introduced overtly or implicitly (here, {the cake; the ice-cream]). The card for this set (positioned on top of the

On the architecture of topic and focus

39

file) consists of a conjunction of two individual cards. The assignment of top to the dp must therefore be interpreted as a reference to the topic set as a whole. The assignment of foe to 'cake' and its percolation to dp, make the same syntactic constituent a focus as well. The focus rule applies to one member of the conjunctive set and positions it, by itself, on top of the file. Update enters the rest of the sentence on the card, and the correct interpretation is derived, selecting 'the cake' and not 'the icecream' as the answer to the question. Spatio-temporal arguments can also play a role in f-structure. Overt references to time and place can be assigned either top or foe as shown in (8): (8)

a. What happened in the morning? [In the morning]t0p, [John ate the cake]foc b. When did John eat the cake ? Het0p ate the cake [in the morningj(oc

Out-of-the-blue statements and answers to questions such as 'What happened?' are anchored to the current contextual spatio-temporal location and are evaluated with respect to it. A card for the current hereand-now, with the location s and the temporal index t contextually construed, must be available on top of the file in view of the fact that reference time is calculated with respect to it. The permanent availability of this card is what allows an interpretation for a merged top feature which is not bound to a lexical item. (9) is an example of a sentence evaluated with respect to the current here-and-now: (9)

a. It is raining. b. select select select select

'raining' 'is' 'it' top

no assignment assign [foc] no assignment

is[foc]

raining

40

Nomi

Erteschik-Shir

There are several possibilities for the structural position of the stage topic in (9). One possibility is spec,cp (enabling a syntactic account of why stage topics generally do not occur in the presence of a c o m p l e mentizer); another possibility is an adjoined ' p e r f o r m a t i v e ' p h r a s e . 1 0 Introducing the two basic information structure primitives lexically in this f a s h i o n derives a variety of different f-structure types d u e to the possibility of multiple top and foe assignments. This approach therefore does not need to stipulate different types of focus (e.g. information focus, contrast, and broad and narrow focus). T o p / f o e assignments are restricted by legibility conditions at the interfaces. The next section discusses the P F interface and shows how prosody and intonation are derived. The following section discusses the interpretation of f-structure.

2. Intonation and Prosody A s noted above many researchers derive stress syntactically and project foci domains f r o m stress. This enables an account of focus which seemingly circumvents the principle of inclusiveness in that stress and focus projection is defined on syntactic structures. Yet, according to C h o m s k y (1995: 381), standard theories are in violation of the condition of inclusiveness in view of the fact that "indications of...intonational structure are not contained in lexical items." A more serious p r o b l e m with this approach is the well-known fact that not all languages mark f o c u s (and topic) intonationally, but rather use morphology or word order for that p u r p o s e . A theory which derives focus f r o m intonation must then e m ploy a d i f f e r e n t way of accounting for focus assignments in those languages. A f u r t h e r p r o b l e m arising f r o m this approach, also m e n t i o n e d above, is that accounts of f o c u s projection do not m a p neatly onto all types of focus structure and require special rules to adjust the results accordingly. T h e derivation of f-structure outlined in the previous section presupposes that (overt) topics and foci are syntactic constituents. I propose in E r t e s c h i k - S h i r (1997, 1999) that intonation is derived f r o m f-structure by a stress rule which assigns stress to foci. This approach directly relates intonation to interpretation by one simple rule. This approach also does not limit focus marking to intonation and allows for alternative markings of both topic and focus by morphology (according to the lexical avail-

On the architecture

of topic and focus

41

ability of such morphological markers) or by displacement, the topic of section 4. The stress rule which assigns stress to foci, applies to all major constituents within the focus domain which are not marked solely as topics. In the stress assignments for the sentences in the previous section are shown (stress is indicated by small caps): (10)

a. hetop

ate [the

CAKE]foc

b. [het0ρ ATE the CAKE]ioc c. [JOHN ATE ittop7foc

d. het0p ate [the e. [it is

CAKE]top/foc

RAINING]{oc

Note first that (10a) and (lOd) are pronounced the same, in spite of their somewhat different f-structures. In (lOd), the contrasted object is stressed by the foe rule which excludes top marked constituents which are not foe marked as well. It follows that only nonconstrastive topics are excluded. In (10b) the vp is stressed. This means that it is not only the object which is stressed but also the verb. The stress on the verb is reduced when the speech-rate is increased, yet a distinction can easily be detected between the pronunciation of (10a) and (10b). This distinction has been ignored since Chomsky (1971) where it was claimed that only the final constituent of the focus is stressed. That this is not the case can be shown even more perceptibly in the following all-focus sentence: (11)

Why are you so upset? [the CHILD ATE the CANDY]foc

Here again, in faster speech, the stresses on the subject and verb are reduced, yet the pronunciation of the three sentences (each in its context) remains distinct even when all three are pronounced with a nonpronominal subject. 1 1 In (10c), stress reduction applies to the stress on the subject. Finally, in (lOe) stress is assigned to the only major category in the focus. ((11) is a better illustration of stress assignment to an all-focus sentence.) Steedman (2000) argues that intonational phrase boundaries and surface syntactic boundaries coincide rendering the 'unconventional' syntactic structure in (12):

42

Nomi

Erteschik-Shir

(12) a.

Marcel

proved

completeness

Marcel proved completeness

According to Steedman the rhythm rule which applies in certain dialects to move the stress onto the first syllable of 'Marcel', applies to the intonational phrase 'Marcel proved' (on a par with 'Marcel Proust'), requiring, within his framework, that the latter be analyzed as a syntactic constituent as in (12b). Instead of positing syntactic structures such as (12b), I propose a mapping of f-structure onto intonational phrases. Since foci are anchored to syntactic constituents, but not to all syntactic constituents, only partial isomorphism is predicted between syntax and prosody. The intonational phrasing in (12a) is consistent with an f-structure assignment in which the subject is the topic and the vp is focused. Since the vp is an fstructure constituent, it must be pronounced as an intonational phrase. (12b), with the intonational phrasing indicated (i.e., 'Marcel proved' forms an intonation phrase) can be an answer to: I know which result Marcel PREDICTED. But which result did Marcel PROVE? This phrasing is also derivable from an f-structure without positing (12b) as a syntactic structure. This is shown in (13). (13)

vp Marcel[top^ proved[top]/[foc]

completeness[foc]

In order for the interpretation to follow, the verb must be contrasted. This is accomplished by assigning both top and foe to it. (In fact, the topic set is {predicted/proved} with the latter focused.) My claim is that intonational boundaries coincide with f-structure boundaries and that the f-structure boundary between the verb and its object enables positing an intonational boundary in that position. In order to account for the full mapping of f-structure to intonational phrases an investigation of which f-structure boundaries are phonologically visible must be pursued. My point here is merely that such an investigation should be made before positing syntactic structures for this purpose.

On the architecture of topic and focus

43

The existence of contrasted syllables provides a challenge to a theory in which stress is derived from foe assignment, and only syntactic constituents can be assigned foe. (14) illustrates such a case: (14) I am advocating PROsecution (not

PERsecution).

As indicated in section 1.2, contrast results from both top and foe assignment: top is assigned to the conjunction of the elements in the contrast set, foe to the element selected from this set as shown in (15): (15)

pro secution foe per secution

f-top

There is a discrepancy between the assignment of foe to the N-constituent prosecution and the assignment of stress to the prefix pro-, which is also interpreted as the contrasted element. I propose that this interpretation and consequent stress assignment is reached by accommodation, since the only way the two elements in (15) can be contrastively interpreted is by limiting the contrast set to the two elements of the constituents which are different. (Such a process is also required to rule out contrast between two identical elements.) 12 In this section, I argued that it is theoretically and empirically advantageous to derive intonation and phrasing directly from f-structure which results from lexical merge.

3. Interpretation In the previous section, it was shown that f-structure must be legible at the PF interface. Here I demonstrate that it must also be legible at the LF interface by showing that quantifier scope can be read off f-structure. Since topics are defined as the 'address' in a file system under which sentences are evaluated (see above section 1.2), it follows that every sentence must have a topic and that truth conditions will vary according to which constituent is assigned top: 13 (16) a. Every man loves SOME woman. b. EVERY man loves some woman.

44

Nomi Erteschik-Shir

In (16a), the object is stressed and the subject is not, indicating that the latter must be interpreted as the topic. The sentence is therefore evaluated for every member of the set of men comprising the topic, i.e., the subject has wide scope. (16b) produces the opposite interpretation, one in which the object has wide scope. This reading is more marked in that it requires heavier contextual framing. That sentences with wide scope objects are marked follows from the fact that subjects are unmarked topics across languages, a fact which has been observed frequently (e.g., Li and Thompson 1976, Reinhart 1981, Andersen 1991, Lambrecht 1994, 2002, Winkler and Göbbel 2002). The unmarked case thus results when subject-predicate structure is aligned with f-structure. We will see in the next section that such isomorphism is a requirement on extraction and provides an explanation for island constraints. If f-structure interpretation derives scopal readings, there is no need for syntactic or semantic scope rules nullifying one of the reasons for LF syntax. In Erteschik-Shir (1997) this approach to quantifier scope is elaborated more fully and further LF syntactic issues are shown to be derivable form f-structure as well. Our architecture will therefore not require LF syntax at all.

4. Syntax It is well known that languages mark topic and focus assignments intonationally, morphologically and/or by word order. In this section, I examine topicalization in Danish and Hebrew as an illustration of the interaction between word order, intonation, morphology and f-structure. The illustrations from Danish and Hebrew topicalization introduced here show that f-structure motivated dislocation (in this case topicalization) is restricted to cases which are unambiguous. Topicalization is a case of optional dislocation as shown in the Danish examples in (17): (17) a. Hun She b. Hun She

hilste pä Ole. Ham havde greeted Ole. Him had

hurt she

ikke not

hilste pä Ole. Hun greeted Ole. She

ikke not

m0dt ham f0r... met him before

havde had

m0dt f0r... met before

On the architecture of topic and focus

45

Since topics may remain in situ (as shown in (17b)), a Rizzi (1997) type view of the left periphery with a TopP projected will fail to explain that not all top features must be checked. Since syntactic dislocation is necessarily triggered by feature checking and is therefore obligatory, topicalization cannot be explained as syntactic dislocation. I therefore pursue an analysis in which optional dislocations of this type are PF operations. PF operations in a minimalist model must also be motivated. Topicalization, i.e., movement to the left periphery is motivated by the need to overtly mark a topic, just as stress is one way to overtly mark a focus. Since topics are also identified by context, moving the topic is optional. Languages may differ in this respect. In English topicalization is rarely employed. In Hungarian (according to one of the reviewers), topicalization is strongly preferred. The reason for this may be the fact that foci in Hungarian are dislocated to the preverbal position and topicalization is needed in order to get the topics out of the way of the foci. In both Hebrew and Danish, topicalization is constrained by the requirement that overt morphology (e.g. agreement, case) enable disambiguation. Since overt morphology is not visible in syntax, this requirement must be formulated as a post-spell-out constraint further strengthening the argument for an analysis of topicalization as phonological dislocation. 1 4 The examples in (18) (from Becker-Christensen 1995: 56) illustrate topicalization in Danish: (18) a. Hun hilste pä Ole. Ham havde hun ikke m0dt f0r... She greeted Ole. Him had she not met before 'She greeted Ole. She had not met him before...' b. A: B:

Pr0v sä at se Try to look Men, der har but, there have

i skujfen. in the drawer jeg ledt. I looked

c. Det regnede jeg ogsä med. that counted I also on Ί also counted on that.' d. Ham gav jeg den til lige him gave I it to just Ί gave it to him just before.'

f0r. before

46

Nomi Erteschik-Shir e. I gär sneede det ogsa. yesterday snowed it also 'Yesterday it also snowed.'

The fronted item in Danish is interpreted as a (noncontrastive) topic and is generally limited to pronouns as in (18a)-(18d) or stage topics as in (18e). Most of these examples are unambiguous due to the predicate: The predicates in (18b)-(18c), for example, select animate subjects. The inanimate fronted topics can therefore only be parsed as objects. Context can also disambiguate in cases in which the predicates are 'symmetrical' in that they allow subjects and objects that are both animate and inanimate. In the following, I use asymmetrical predicates and disregard context. (19) shows that a fronted nonspecific indefinite is unacceptable, even in the presence of a subject pronoun. (19) *En pige mtpdte Peter/jeg a girl met Peter/I 'Peter met a girl yesterday.'

igär. yesterday

This provides further evidence that the fronted constituent must be interpreted as a topic. Since definites qualify as topics, it is surprising that these are unacceptable as shown in (20): (20) a. *Marie/Pigen m0dte Peter Marie/the girl met Peter 'Peter met Marie yesterday.'

igär. yesterday

b. IMarie/Pigen m0dte jeg igär. Marie/the girl met I yesterday Ί met Marie/the girl yesterday.' (20b) is somewhat better than (20a). This, I claim, is due to the nominative subject pronoun which disambiguates the subject/object roles: What is 'wrong' with nonpronominal fronted topics is the (initial) ambiguity of the resulting structure. When an accusative object pronoun is fronted, the result is also unambiguous and therefore acceptable:

On the architecture of topic and focus

(21) Hende m0dte Peter/jeg Her met Peter/I 'Peter met her yesterday.'

47

igdr. yesterday

Surprisingly, even though neither is ambiguous, (20b) is slightly degraded compared with (21). This could be because the correct parse is signaled earlier in the latter than in the former. Dislocation to the left periphery in Danish is restricted to constituents which qualify as topics and must result in an unambiguous parse. Note that the restriction cannot be defined in terms of the fronted element only, in view of the fact that (20b) is only slightly degraded. An analysis restricting fronting to overtly case marked elements is therefore also doomed to failure. It is possible that parsing constraints of this type should be formulated as pure processing constraints, yet other possibly related processes such as pro-drop have been analyzed in purely syntactic terms with serious, albeit unsuccessful, attempts to relate it cross linguistically to the presence of overt agreement morphology. I propose that parsing constraints are most naturally viewed as constraints on the interface with the articulatory-perceptual system, i.e., as PF constraints on the identification of arguments (ID): (22)

ID: In a string, Χ V Y, ID X as subject and Y as object if neither is marked otherwise.

ID would identify the topicalized object in (20a) as a subject and the post verbal argument as an object, indeed the only interpretation of the string. In (20b) the post verbal argument is identified as a subject, allowing the preverbal argument to be interpreted as an object by ID. Similarly, in (21) the initial element is accusative identifying it as the object. The analysis to be proposed should enable an account of cross linguistic variation. An examination of topicalization data in Hebrew shows that this data is also accounted for by the ID constraint. There are two types of Hebrew topicalization, one in which the verb is fronted to V-2 position and one in which it is not. 15 Ambiguity only results with V-2 in which the two arguments are separated by a V rendering a potentially ambiguous string. I therefore discuss only the former type, which leads

48

Nomi Erteschik-Shir

to a contrastive interpretation of the fronted element (capitals indicate stress): (23) a. 1YELED pagash haish. boy met the man 'The man met a boy.' b. YELED pagsha haisha. boy metf e m the woman 'The woman met a boy.' The contrastive stress on the initial element in (23a) is not a disambiguating factor since the sentence can be parsed as having a contrastive subject, rather than an object as intended here. The feminine subjectagreement marking on the verb in (23b) identifies the post verbal element as a subject, the sentences complies with ID and is therefore acceptable. (24) shows that fronting a definite is perfectly acceptable: (24) et hciYELED pagash haish. the-boy met the man 'The man met the boy.' Definite objects in Hebrew are, however, always marked by 'et' (the exact nature of this element is irrelevant here), and definite subjects lack this marking, so if one of the two arguments is definite the sentence is unambiguous. When both the subject and the object are indefinite, the result is never perfect: (25) a. * YELED pagash ish, lo boy met man, not Ά man met a boy, not a girl.'

yalda. girl

b. IIYELED pagsha isha, lo baxur. boy metfem woman, not youth Ά woman met a boy, not a youth.'

On the architecture of topic and focus

49

(25a) violates ID, (25b) does not. It is a w k w a r d since it is hard to c o n t e x t u a l i z e the sentence in the absence of a definite candidate f o r topichood.16 T h e data f r o m Danish and Hebrew thus indicates that overt agreement identifies displaced DPs. It is significant that the overt agreement need not be detected on the displaced element. The constraint must therefore be formulated in such a way that disambiguating overt marking on any element of the sentence qualifies to ID the displaced element. L a n g u a g e s d i f f e r as to how rigidly the ID constraint applies. Danish is extremely rigid in this respect: The S V O reading of an a m b i g u o u s string is strongly preferred and contextual clues can only overrule this i n t e r p r e t a t i o n with d i f f i c u l t y . 1 7 In other languages context can m o r e easily c o m e into play. More research into the role of the ID constraint in a variety of languages is needed to decide what additional factors are res p o n s i b l e f o r these d i f f e r e n c e s a m o n g l a n g u a g e s and w h e t h e r f o c u s displacement is also subject to this constraint.

4.1. Identification in Dependencies In E r t e s c h i k - S h i r (1997), it is argued that a variety of d e p e n d e n c i e s (e.g., w/z-dependencies, binding) are restricted by a single f - s t r u c t u r e constraint: the dependent must be in the focus domain in an u n m a r k e d f-structure, one in which subject-predicate structure is aligned with f structure. (This is the same isomorphism which produces u n m a r k e d scopal interpretations.) (27)

A dependent X is interpretable iff S U B J E C T t 0 p [ . . . X ...]foc sTOP t

T h i s constraint restricts d e p e n d e n c i e s to u n m a r k e d f - s t r u c t u r e s in which the topic is either the subject or a stage topic and the dependent is contained within the f o c u s domain. Crucially a structure in which the object is the topic does not license such a dependency. Let m e briefly illustrate the constraint by examining how it applies to superiority phen o m e n a shown in ( 2 7 ) : 1 8

50

Nomi Erteschik-Shir

(28) a. Who read what? b. *What did who read? (27a) can be uttered in a context in which a set of 'who-s' is under discussion, i.e., in a context of a class of students who were given an assignment to read an article of their choice. It could also be uttered if both the set of students and a set of reading assignments were given in the context. It could, however, not be uttered in a context in which only a set of reading assignments were given. This is due to the constraint on dependents. If the subject w/i-phrase is interpreted as ranging over a discourse specified set, it is licensed as a topic. The object wh-phrase, in turn, is interpreted as dependent on the subject w/i-phrase, in that the question requests a pairing between the two. Under this reading, the multiple w/z-question conforms to the constraint on dependents in that the subject is interpreted as a topic and the dependent is located in the VP which is the focus of the question. The reading in which only the object w/j-phrase is interpreted as the topic, with the subject w/i-phrase dependent on it is blocked because subjects cannot be dependent on their objects as stated in (26). The same f-structure can be assigned to (27b), yet here there are two dependencies to be processed at the same time as shown in (28): (29)

*What did who read t

One dependency is between the fronted w/i-phrase and its trace. The other one is between the two w/i-phrases as in (27a). This leads to a processing constraint of a type similar to the ID constraint since it requires double identification of the trace position because it is a dependent in two separate dependencies. (It is not possible to set up a dependency directly between the two ννΛ-phrases here, because then, even if the subject wh-phrase is interpreted as the topic, the object wh-phrase is no longer within the VP focus domain.) Kayne's (1984) facts in (29) and (30) show that, surprisingly, an extra wh- improves superiority violations: (30)

What did [who]{op [hide t where]foc

On the architecture of topic and focus (31)

Who knows

what whO[0p [saw

51

//foe

This is because the extra w/z-phrase makes it possible to circumvent the analysis in which the trace is doubly identified due to the fact that two separate dependencies can be construed. In both cases, the subject w/i-phrase has to be contextualized as ranging over a topic set and the context has to enable an interpretation with two separate dependencies. Separating out the two dependencies, however, poses a heavy processing load which is why these types of sentences are still quite degraded.

5. Architecture This paper addressed the question as to how to introduce top/foe into the grammar without violating the Inclusiveness principle. The solution I propose is that top/foe are lexically selected. I also argue, based on previous work, that top/foe are interpretable at the PF/LF interfaces. This view results in the following architecture: 19 (32)

Initial

merge

(incl. top/foc)

θ-assignments

/\ Ψ

4

Interpret

PF computation dislocation morphology phonology The output of narrow syntax is delivered to the conceptual-intentional interface, which depends on the configuration of initial merge and f-structure. At the PF interface, structures are linearized and may no longer be transparent to interpretation. 'Interpret' thus applies to the merged structure, a view reminiscent of the Aspects model of deep structure. This is possible since LF functions, under this view, are taken over by f-structure. I have also suggested that Topicalization is to be analyzed as part of the phonological linearization process. 20 Other optional dislocation pro-

52

Nomi

Erteschik-Shir

cesses should probably also be analyzed in this way. This is particularly appropriate in cases such as this one in which dislocation is to the edge of a prosodic domain. Dislocation is therefore a property of P F c o m p u tation. I leave open, here, w h e t h e r cases of syntactic dislocation remain.21 I have further posited several P F Interface requirements, ID, the dependency constraint and the constraint on double identification. T h e idea that ID regulates dislocations is related to a very old idea that m o r p h o l o g i c a l l y 'rich' languages have f r e e r word order, an idea that has been very difficult to implement. M u c h more work on this topic is clearly required. T h e p h o n o l o g i c a l account of displacements also o f f e r s a new perspective on the parameterization of topic and focus positions: L a n g u a g e s d i f f e r with respect to which constituent edges are visible; the left vp edge is the locus for Hungarian foci whereas the right vp edge is the locus f o r Russian foci; the left edge of a sentence is c o m m o n l y the locus of the main topic. It would be interesting to investigate whether different focus positions can be linked to prosodic properties of the languages in question. If this were the case there would be no need for syntactic Topic and F o c u s projections, and the language particular order imposed on t h e m would no longer be required.

Notes *

1. 2.

3.

4.

Thanks are due to the audience of the GLOW workshop Information structure in generative theory vs. pragmatics, Natalia Strahov, Idan Landau and two anonymous reviewers. The list of references relevant to the role of top/foe in the grammar is too long to include here. For some of them see Erteschik-Shir (1997). "Any structure formed by the computation.. .is constituted of elements already present in the lexical items selected for the numeration; no new objects are added in the course of computation apart from rearrangements of lexical properties." (Chomsky 1995: 228) Neeleman and Reinhart (1998) derive 'discourse-linking' (including presupposition), but not topichood from destressing. Although topics are necessarily presupposed, not all presupposed material is a topic as will become clear in section 1.2. I got the idea that top/foe features are assigned to lexical constituents when they are drawn from the lexicon from Winkler and Göbbel (2002). The idea

On the architecture

of topic and focus

53

that these features project on a par with φ-features is f r o m Idan Landau (personal communication). 5. I assume the possibility of percolation from Ν to its extended projection, DP. Similarly, percolation is possible from I to CP allowing foe assignment to a full sentence. Syntactic feature percolation does not allow top/foe assignment to elements smaller than a lexical item since percolation must be upwards. For an account of contrasted syllables (e.g. PROsecution vs. PERsecution), see section 2. 6. Note that in this view the projection of top/foe features is syntactic and that no separate 'level' of f-structure is posited, counter the criticism of Kempson et. al., this volume. 7. In each such entry, the heading is replaced by 'e'. 8. See Cohen and Erteschik-Shir (2002) for an application of this property of topics to the interpretation of bare plurals. 9. Presuppposed elements are defined as entries already present on the cards. For further details of the application of the f-structure rules and how they apply to questions see Erteschik-Shir (1997). 10. The cards permanently available on the top of the file are, in addition to the current stage, cards for the first and second person (the speaker(s) and the hearer(s)). These are the actors in the performative act and should receive an integrated analysis. 11. As shown in my previous work, stress reduction is most pronounced on the intermediate stresses. The initial stress remains slightly stronger yet less strong than the final unreduced stress. This may be a way to clearly mark the initial and final borders of the focus constituent. The intermediate stresses are possibly parallel to what Biiring, this volume, calls ornamental accents. In spite of other seemingly major differences, it is not clear that the two approaches are in fact incompatible, warranting careful comparison. 12. The same approach can also apply to the movement to pre verbal focus position in Hungarian of only part of a contrastive focus constituent, as illustrated in the following example supplied by an anonymous reviewer: (i)

Nora Α FILMET akarta NEZNI (es nem aludni akart) Norah the movie-acc wanted watch-inf and not sleep-inf wanted 'Norah wanted to watch the movie and not sleep' Just as the element the conjuncts have in common is excluded from stress in English, so such elements are excluded from movement in Hungarian. 13. The "tendency" for topics to take wide scope has been noted before (e.g., Ioup 1975, Kuno 1982, Reinhart 1983). All-focus sentences are evaluated with respect to their stage topic, the here-and-now of the discourse. 14. Another option would be to adopt the 'copy' theory of movement in which movement is syntactic, yet PF requirements may force the pronunciation of the copy. (See Boskovic 2001 and the references cited therein for details of this

54

Nomi

Erteschik-Shir

approach and see Erteschik-Shir 2001, 2005a, Erteschik-Shir and Strahov 2004 for arguments against this approach.) 15. Hebrew preverbal subjects are topics, postverbal subjects are foci. 16. The only possibility would be in the context of a stage topic. In ErteschikShir (to appear), I argue that acceptability is linked to the ease of contextualizing a sentence. 17. W%-questions are an exception to this in Danish: (i)

18. 19.

20. 21.

Hvem kender Peter? Who knows Peter 'Who knows Peter?'/'Who does Peter know?' See Erteschik-Shir (2005b) for an explanation of the different behavior of whquestions in this respect. For a detailed f-structure theoretical account of the effect of d-linked w/z-phrases on superiority across languages see Erteschik-Shir (1997), (to appear). See Erteschik-Shir and Rapoport (2000, 2004, in preparation) for a theory in which θ-assignments, structural/lexical case can be read off structure merged and projected by verbal meaning components. See Chung (2003) for an analysis of dislocation in PF as part and parcel of linearization. But see Erteschik-Shir (2001), Erteschik-Shir and Strahov (2004) for arguments that dislocation is a property only of PF.

References Andersen, Torben 1991 Subject and topic in Dinka. Studies in Linguistics 15: 265-294. Becker-Christensen, Christian 1995 Grammatisk oversigt. In Nudansk ordbog & Sprogbrugsleksikon, Christian Becker-Christensen (ed.), 47-67. Copenhagen: Politikens Forlag. Biiring, Daniel this vol. Focus projection and default prominence. In Valeria Molnär, and Susanne Winkler (eds.), 321-346. Boskovic, Zeljko 2001 On the Nature of the Syntax-Phonology Interface: Cliticization and Related Phenomena. North-Holland Linguistic Series: Linguistic Variations. Amsterdam: Elsevier. Chomsky, Noam 1971 Deep structure, surface structure and semantic interpretation. In Semantics: An Interdisciplinary Reader in Philosophy, Linguistics and

On the architecture of topic and focus

55

Psychology, D. Steinberg, and L. Jakobovits (eds.), 183-216. Cambridge: Cambridge University Press. 1995 The Minimalist Program. Cambridge, Massachusetts: MIT Press. Chung, Sandra 2003 The Syntax and Prosody of Weak Pronouns in Chamorro. Linguistic Inquiry 34: 547-599. Cinque, Guglielmo 1993 A null theory of phrase and compound stress. Linguistic Inquiry 24: 239-297. Cohen, Ariel, and Nomi Erteschik-Shir 2002 Topic Focus and the Interpretation of Bare Plurals. Natural Language Semantics 10: 125-165. Erteschik-Shir, Nomi 1997 The Dynamics of Focus Structure. Cambridge: Cambridge University Press. 1999 Focus Structure Theory and Intonation. Language and Speech 42: 209-227. 2001 P-syntactic motivation for movement: Imperfect alignment in Object Shift. Working Papers in Scandinavian Syntax 68: 49-73. 2005a The sound patterns of syntax: The case of object shift. Theoretical Linguistics 31: 47-94. 2005b What is Syntax? Theoretical Linguistics 31: 263-274. to appear What's what? In Gradience in Grammar, Caroline Fery, Gisbert Fanselow, Matthias Schlesewsky, and Ralf Vogel (eds.). Oxford: Oxford University Press. Erteschik-Shir, Nomi, and Tova Rapoport 2000 Aspectual Focus. Paper presented at GLOW, Bilbao. 2004 Bare Aspect: A theory of syntactic projection. In The Syntax of Tense, Jacqueline Gueron, and Jacqueline Lecarme (eds.), 217-234. Cambridge, Massachusetts: The MIT Press in prep. The atoms of meaning: Interpreting verb projections. Ben Gurion University. Erteschik-Shir, Nomi, and Strahov, Natalia 2004 Focus structure architecture and P-Syntax. Lingua 114: 301-323. Ioup, Georgette 1975 Some universals for quantifier scope. In Syntax and Semantics, J. P. Kimball (ed.), 37-58. New York: Academic Press. Kayne, Richard S. 1984 Connectedness and Binary Branching. Dordrecht: Foris. Kempson, Ruth, Ronnie Cann, and Jieun Kiaer this vol. Topic, focus and the structural dynamics of language. In Valeria Molnär, and Susanne Winkler (eds.), 59-82.

56

Νο mi

Erteschik-Shir

Kuno, Susumu 1982 The focus of the question and the focus of the answer. The Parasession on Nondeclarative Sentences, CLS, 276-337. Lambrecht, Knud 1994 Information Structure and Sentence Form. A Theory of Topic, Focus, and the Mental Representations of Discourse Referents. Cambridge Studies in Linguistics: Cambridge University Press. Li, Charles, and Sandra Thompson 1976 Subject and topic: A new typology. In Subject and Topic, Charles Li (ed.), 457-490. New York: Academic Press. Neeleman, Ad, and Tanya Reinhart 1998 Scrambling and the PF Interface. In The Projection of Arguments: Lexical and Compositional Factors, Miriam Butt, and Wilhelm Geuder (eds.), 309-353. Stanford: CSLI Publications. Reinhart, Tanya 1981 Pragmatics and linguistics: An analysis of sentence topics. Philosophica 27: 53-94. 1983 Anaphora and Semantic Interpretation. London: Croom Helm. 1995 Interface strategies. In OTS Working Papers. Utrecht: Utrecht University. 1996 Interface economy: Focus and markedness. In The Role of Economy Principles in Linguistic Theory, Chris Wilder, Hans-Martin Gaertner, and Manfred Bierwisch (eds.), 146-169. Berlin: Academic Verlag. Rizzi, Luigi 1997 The fine structure of the left periphery. In Elements of Grammar, Liliane Haegeman (ed.), 281-337. Dordrecht: Kluwer. Selkirk, Elisabeth 1984 Phonology and Syntax. Cambridge, Massachusetts: MIT Press. Steedman, Mark 2000 Information structure and the syntax-phonology interface. Linguistic Inquiry 31: 649-689. Strawson, P. F. 1964 Identifying reference and truth-values. Theoria 30: 86-99. Szendri, Kriszta 2001 Focus and the Syntax-Phonology Interface, University College London: Ph. D. diss. Winkler, Susanne, and Edward Göbbel 2002 Review article: Zubizaretta, Maria Luisa (1998) Prosody, Focus, and Word Order, Linguistic Inquiry Monograph 33. Massachusetts: MIT Press. Linguistics 40: 1185-1243.

On the architecture of topic and focus

57

Zubizaretta, Maria Luisa 1998 Prosody, Focus and Word Order, vol. 33: Linguistic Inquiry Monograph. Cambridge, Massachusetts: MIT Press.

Topic, focus and the structural dynamics of language1 Ruth Kempson, Ronnie Cann, Jieun Kiaer

Abstract This paper sets out a typology of left- and right-periphery effects in which the differences between the structural and topic/focus effects displayed at the two peripheries of interpretation are characterised without any specific structureparticular concepts as part of the formal explanation. The framework in which these generalisations are expressed takes the dynamics of parsing in real time as central to the syntactic explanation. Phenomena captured include Hanging Topic Left Dislocation, Clitic Left Dislocation, Pronoun Doubling, Topicalisation, Subject Postposing, and the Right Roof Constraint.

1. The Dynamics of Language Processing The concepts of topic and focus involve a context-sensitivity that uneasily straddles the boundaries between syntax, semantics, and pragmatics (see Sgall et al. 1986, Halliday 1967, Chomsky 1971, Vallduvi 1991, Erteschik-Shir 1997). In accounts of the phenomena in the principles and parameters and minimalist frameworks, topic and focus are both treated as associated with one or more projected functional heads (see Rizzi 1997 and others), with no attempt to explain the context-sensitivity of the phenomena. Model-theoretic approaches, to the contrary, address their context-dependence through question-answer congruences (e.g. Biiring 1997, Schwarzschild 1999), but they do not provide a basis for expressing restrictions on syntactic forms of the focusing device, and, despite context-dependency in the explanation, their accounts fail to allow any shift of focus in providing an answer to a question. Vallduvi (1991) and Erteschik-Shir (1997) claim to capture structural and semantic aspects of the construal of topic/focus by articulating an additional level of Information Structure, but this level lacks independent motivation and implies a non-minimal architecture for the grammar.

60

Ruth Kempson, Ronnie Cann, Jieun Kiaer

In this paper, we argue that the Dynamic Syntax framework (DS) - which adopts a Fodorian representationalist stance, 2 with the added twist of reflecting the dynamics of real-time parsing - is able to articulate structural processes of interpretation build-up, from which the reported topic and focus effects can be explained on a cross-linguistic basis. The type of analysis the framework makes available is illustrated by left and right periphery effects in English, Greek, Arabic, Hebrew, Spanish and Romanian. The account to be set out expresses similar insights to Vallduvi and Erteschik-Shir, but in a framework in which the dynamics of the parsing process constitute the grammar formalism and provide all that is needed to explain these effects. The methodology implicit in the Dynamic Syntax formalism is to take the constructs used in semantics, and define them as representations for which a syntax is defined. The process of establishing some such interpretation in context is defined as a monotonic tree growth process. 3 The central claim is that syntactic properties of natural language reside exclusively in the progressive growth of such tree-structure representations strictly following the dynamics of left-right processing (see Hausser 1989 for a related view). N o additional level of syntax needed. The type of tree we assume as interpretation assigned to (1) is (2): (1)

John upset Mary.

(2) Fo(PAST : Ty{e)~~~ Fo(John)

Ty{t) Upset(Mary){John)) t) Fo(Upset(Mary))

Ty(e) Fo(Mciry)

Ty(e —> (e —> i)) Fo(Upset)

Each node in the tree has a concept formula (Fo for Formula) and an indication of what semantic type that concept is. The primitive types are types e and t. All other types are functions on these. 4 All this is no more than a conventional semantic characterisation of how sentences are understood in context as assertions about entities in the world around us, expressed as a predicateargument formula, with the tree showing how the variously typed expressions that are sub-terms of the containing formula combine together to yield that resulting formula - in effect a history of how the parts combine together.

Topic, focus and the structural dynamics of language

61

1.1. Building up Structures The building up of such decorated trees is achieved by a combination of general (computational) actions, and lexical actions. This tree growth process, is not induced on a head-driven basis, but as an incremental parsing process, building up a sequence of partial trees, each one an enrichment of the previous partial tree in the sequence. The overall dynamics is to unfold a skeletal structure in anticipation of concepts which the words progressively provide in the order in which they are presented in the string, and then to combine those concepts together on a strictly bottom-up basis to yield a propositional formula. 5 The starting point, associated with the outset of the parsing process is simply the goal to establish some propositional formula as interpretation, stated as lTy(t) (following Sperber and Wilson 1995), and this overall goal may lead to other subgoals as more information comes in, these represented as additional nodes in the tree, e.g. with requirements ?Ty(e), lTy(e —»· t). In this respect the process is top-down. The progressive building up of interpretation is taken to be a process of unfolding a tree structure in anticipation of what is needed to complete a logical form representing an interpretation using all the words, these words providing actions that enable appropriately decorated structure to be built up (the words providing the bottom-up aspect of the parsing process). 6

1.2. Reflecting context-dependence Pursuing the representationalist stance, all context-dependent aspects of interpretation are represented by underspecified terms to be replaced by substitution of terms as the unfolding structure makes them available. So a pronoun is defined as providing a place-holding device, a metavariable, which has to be replaced during the interpretation process by some OTHER term as made available in the context. For example, in (3), he and her are naturally understood as picking out John and Mary by being replaced by the forms Fo(John) and Fo(Mary)·. (3)

John ignored Mary. He upset her.

The concept of representing the context-dependence of anaphoric expressions in language structurally is familiar enough (see Kamp and Reyle 1993, Ranta 1994) What is less orthodox is the assumption that it is to be defined as a tree-

62

Ruth Kempson, Ronnie Cann, Jieun Kiaer

update process; and with contexts also represented as (partial) trees, anaphora resolution can apply equally to the update of a pronoun from antecedent terms within the structure under construction: (4)

John thinks that he is clever.

For the identification of he in (4), the context relative to which that interpretation process takes place includes the partial structure containing the subject node with its decorations. In canonical uses of pronouns, like all other content words, the pronoun has a restriction that whatever value it is assigned must be taken as decorating a terminal node in the tree, a so-called "bottom" restriction (Cann et al. 2005). As we shall shortly see, there is a range of expletive pronoun uses in which this terminal node restriction is lost.

1.3. The dynamics of long-distance dependency What is novel is the extension of this concept of underspecification and update during the interpretation process to the modelling of long-distance dependency effects: (5)

Mary, John upset.

Sentences such as these have been universally assumed since Chomsky (1965) to display a core syntactic phenomenon not reducible to semantic explanation. The present analysis, however, does not set up such separate syntactic mechanisms, but, instead, follows the parsing dynamics. An expression introduced at an early stage in the interpretation process does not, at this point in the interpretation process, enable its position in the structure to be identified. In (5), the word Mary is construed as providing a term for the resulting logical form, but the node which it decorates does not have its relation within the overall structure yet fixed:7 Tn(a),lTy(t)

{U)T{a),Fo(Mary)S3xTn(x),Q What (ΐ#)77ι(α) expresses is the information that a node in the tree identified as Tn(a) (Γη a predicate that takes tree-node identifiers as its argument) dom-

Topic, focus and the structural dynamics of language

63

inates the node bearing the decoration in question. This only weakly specified tree relation (introduced by a rule called *Adjunction) is updated into a fixed relation later, following an arbitrary number of intermediate steps in the interpretation process. It is only after processing the word upset, that the opportunity to provide this fixed position will arise. At that juncture, there will be a structure with the concept Fo(Upset) labelling a functor node, and a node introduced to provide its internal argument: After processing upset. ?7>(r)

Fo(Mary)..

Fo(John) ''••••.. JTyie),

1Ty{e -»i) 0

Fo(Upset)

Both the type requirement of the object node and the requirement for a fixed position in the tree for Fo(Mary) drive the subsequent step, which is to unify these two nodes (a process which is called Merge). Nothing more will be needed if the role of the formula Mary is identified as being the object of Upset, for such a move will simultaneously satisfy the two outstanding terminal-node requirements. From then on, the process is one of combining the concepts Fo(Upset) and Fo(Mary), and then the predicate Fo(Upset(Mary)) and Fo(John) to yield the logical form Fo(Upset(Mary)(John)), with a resulting tree identical to that derived from the parse of (1). Of course with more than one available strategy, there is the question of feeding-relationships between them. The system is constraint-based, and combinations of actions are excluded only if their application precludes successful completion of the construction process. In the parse of (1), for example, there is no possibility of parsing the first NP as a subject and the second as an unfixed node, since the process of introducing such an unfixed node is defined only to take place if there is no other node dominated by the topnode with requirement lTy{t). Indeed there can't be more than one such unfixed node inhabiting a tree at any one time, as nodes are distinguished only by their relative tree position. So in parsing a sequence of NPs at the left periphery, it turns out that the only option is to analyse the left-peripheral expression as serving to decorate the unfixed node, the second NP as the subject.

64

Ruth Kempson, Ronnie Cann, Jieun Kiaer

This license to introduce an unfixed node into a tree provides us with a means of identifying what might be involved in a focusing device. Defined as it is in terms of left-right processing and progressive update of structure, the first constituent is initially analysed as providing some information, but not the role of that term in the structure. Its structural role is thus initially underspecified and resolved later when some appropriate structure has been introduced. The dynamics of this process, accordingly, involves the isolation of one term at the node in question and the immediately subsequent construction of an accompanying propositional structure, to which that one term provides a subsequent update. The concept of syntactic focus, we shall be suggesting, is no more than this - the provision of some structure identified as providing the update to some constructed structure that yields a completed propositional formula. 8

1.4. Constructing trees in tandem In addition to the projection of such individual trees, paired linked trees are defined. One partial tree, that is, may provide the context for the process of introducing another, so that two trees are introduced in tandem, using a combination of anaphoric substitution processes across a pair of trees and mechanisms for updating an individual tree structure. Relative clause construal provides a core example: (6)

John, who left, cried.

Following the dynamics of how such trees might be incrementally constructed, in the DS account, a LINK transition is defined (using an additional modal operator (L) and its inverse ( L - 1 ) ) that licenses a transition from a node in one tree to the initiation of a second. Expressed as a modal requirement on the way this new emergent tree is to be completed, the tree is constrained so that at some point in its subsequent construction, there must be a copy of the formula from the head from which the LINK relation was constructed. 9 In (6), that relation is constructed from the node decorated by Fo(John),i0 with the copy being provided at an unfixed node by the interpretation assigned to the relative pronoun (Figure 1).

Topic, focus and the structural dynamics of language

65

Tn(a),lTy(t) (t )Tn\u), Fo(John)

?7>(e-f) {.L-l)(\)Tn(a),lTy(t),l(U)Fo(John)

(U)(L-l)(1)Tn{a),Fo(John), Figure 1. Parsing who in (6)

1.4.1. Linked structures from left-peripheral

expressions

This f o r m of relation between obligatorily anaphorically paired trees applies to account for Hanging Topic Left Dislocation structures (Anagnastopoulou et al. 1997), for nothing in the formal statement of the rule licensing the L I N K transition dictates the level of embedding that the head node might be in its containing structure; and nothing prevents it being the top node of some structure. Accordingly, we propose that the left-peripheral expression may be used as the trigger for introducing a tree linked to the rootnode which is to be of type e which that expression duly decorates (Figure 2).

(L)T n(0), Fo(Mary) ,Ty(e)

Tn{0),lTy{t),l(D)Fo{Mary),0

Figure 2. Building a linked structure for left-peripheral expressions The consequence of having constructed and decorated such a linked tree is that the root node with requirement 1Ty{t) must now be constrained to contain a copy of the term projected from the left-peripheral noun phrase so as to satisfy the sharing-of-terms requirement dictated by the L I N K relation between the two structures. This modal form of requirement determines the presence of a suitably construed pronoun in the twinned structure, which the facts of so-called Hanging Topic Left-Dislocation corroborate: (7) (8)

As for Mary, Tom adores her. *As for Mary, Tom adores.

66

Ruth Kempson, Ronnie Cann, Jieun Kiaer

The f o r m of the requirement involves the modal operator (D), which is the union of LINK and daughter relations, i.e. embodying a restriction that there be some subsequently constructed copy of the head term, without any structural restriction on where in the structure such a copy should occur: (9)

I Maria xtes gnorisa ton andra pu tin patreftike. The MariaNOM yesterday I met the man who her married 'As for Maria, yesterday I met the man who married her.' [Greek]

The only exception to this required presence of a lexical pronoun is the case of pro-drop structures, i.e. those cases where the verb projects its argument nodes decorated with a metavariable, exactly as though a lexical f o r m of pronoun had been present.

2. Characterising the left periphery We can now see how, with these two mechanisms for projecting interpretation f r o m left-peripheral expressions, we can define their correspondence with various topic and focus effects. It is standardly recognised (Ross 1967) that long-distance dependency effects can be modelled either as a correlation between two discrete positions in a tree through a process such as movement (or feature-passing), or through a process of anaphoric linkage. There are languages with the left-dislocated expression paired with 'a gap' and displaying island restrictions, diagnostic of movement processes: (10) (11)

Mary, John thinks Tom had upset. *Mary, I dislike the man that

married.

These are reconstructed here as initially unfixed nodes, updated at some point along a sequence of daughter relations, by definition excluding resolution in all structures analysed as projecting linked trees - that is complex NPs, relative clauses, coordination and clausal adjuncts, in effect ensuring that for any left-peripheral expression which is not paired with an anaphoric expression, there is an essential locality restriction associating that expression with the point at which it constributes to the interpretation. 1 1 There are also languages/structures that display pairing of the left-dislocated expression with a pronoun with no subjacency effects; and these have been assumed to be in-

Topic, focus and the structural dynamics of language

67

duced as independent structures - linked to each other solely through anaphoric processes. These correspond straightforwardly with the concept of linked structures: (12)

Il-kita:b da, 'into, tkallimt ma9a l-walacl 'illi katab 9aley-h. The book this, you talked with the boy who wrote on it 'You talked with the boy who wrote on this book.' [Egyptian Arabic]

(13)

As for Mary, I talked to the boy who had scribbled on her book.

The LINK transition imposes the requirement which the anaphoric expression must be construed as meeting, in order to secure a well-formed result. We now expect insensitivity to islands involving anaphorically linked structures, since these are constructed as independent trees. We also expect ordering of topic-construed and focus-construed expressions at the left periphery, in that the building of a LINK transition involves the creation of a shift from one structure to a discrete emergent structure, and the introduction of an unfixed node within an individual structure can only be defined in the face of such a transition, not prior to it. We also expect no iteration of such unfixed nodes, given that the required *Adjunction rule is defined to apply only to those typer-requiring nodes which lack internal structure, and so cannot apply more than once to the same node.

2.1. Linked trees as constructed contexts What might seem to be less expected, given this dichotomy, are the various intermediate forms, involving some kind of interaction between the establishment of a long-distance dependency effect and anaphoric processes, and an apparent blurring of the distinctiveness in the two processes. There are leftperipheral constituents paired with a pronoun which display some but not all properties of movement - their properties include sensitivity to strong island effects, general exclusion of quantified expressions (indefinites only, and with specific interpretation), being associated with a sharp break of intonation following the left-peripheral expression: 12 (14)

Ton Petro, ton nostalgo poli. The Peter^cc, CUcc miss-lsg much Ί miss Peter a lot.'

[Greek]

68 (15)

Ruth Kempson, Ronnie Cann, Jieun Kiaer *Tin

Maria, xtes

gnorisa

tonandrapu

Thence Maria yesterday m e t i s , . ^ , , ^ , the man

tin that h e r ^ c

patreftike. married. 'Mary, yesterday I met the man that her married.' These intermediate effects are problematic for movement accounts, 1 3 as the paradigm leads one to expect a certain diagnostic set of effects associated with movement, and failure to meet such diagnostics is associated with base generated pairing of co-indexed expressions. The two theoretical constructs have been taken to be discrete; and the existence of Clitic Left Dislocation is generally taken to require base generation of the left-peripheral expression and its co-indexed clitic, despite meeting diagnostics of movement. 1 4 The advantage of expressing pronoun construal and long-distance dependency in the same terms of underspecification and resolution through tree growth is the natural characterisation it allows for intermediate effects. According to this analysis, intermediate effects may arise in one of two ways. Either they result f r o m the imposition of locality constraints on the antecedentanaphor relation imposed by the LINK relation, as a structurally restricted requirement on what position in the tree that anaphoric device must decorate. Or they may result f r o m a feeding relation between pronoun construal and long-distance dependency, a node decorated by a pronoun being able to be unified with some unfixed node and so receiving a value through the Merge process as an alternative form of update for pronoun construal. Taking up the form of restriction dictating how two structures are to be taken as linked, in our prototype display, figure 2, we took the form of requirement to be l(D)Fo(a), indicating that there were no locality constraints on where the copy was to be constructed within the individual tree on which this requirement was to be met. However, given the modal form of the requirement, there is every reason to expect variants of this form of constraint through selection of other modal operators from the range (D), ( j * ) , thereby imposing more severe restrictions on where the copy is to be found in this second tree. The imposition of the requirement using the (J,*) operator, where the imposed structural restriction on the antecedent-anaphoric pairing mimics that of an unfixed node, corresponds directly to CLLD structures, and, given the LINK analysis, the attendant pronoun in the second structure and sharp intonational break of CLLD are also expected. 1 5

Topic, focus and the structural dynamics of language

69

The advantage of this approach is not merely in providing a basis for articulating systematic interactions between anaphora and island-inducing structures/processes. It also confirms the correspondence between topic, context and linked tree. From the concept of a pair of linked trees, we can define topic as a direct reflection of a particular form of context, that of a partial tree constructed during a single utterance process. At its most general, then, topic is a structure that provides an antecedent for later identification by some anaphoric device. This may be on a cross-utterance basis: (16)

John came in. He was sick.

But a topic may be realised by presenting some constructed linked tree as the context, whose role is to provide an antecedent for other terms introduced as part of the process of interpretation for the string, as in (17): (17)

As for John, he was sick.

Left-peripheral expressions construed like this may or may not be anaphorically identified from the more general context. Contrastive topics arise when the linked structure so constructed is NOT itself anaphorically linked to the general context, and in so being a departure from it, will be in contrast to it. No structural distinction is needed to reflect this: the relation of a pair of linked structures to the more general discourse context is sufficient.

2.2. * Adjunction and the basis of focus We now turn to the second way in which nodes in a tree may be correlated - through the projection of an unfixed node and its subsequent merge with a distinct node in the tree. Here we find the inverse effect. One term may be isolated from some remainder, because, in virtue of its isolatability, it can be used as an update for a structure once that structure is substantially complete. And once again, there are ways in which structural processes yielding this result interact with anaphora construal to yield mixed effects. The core cases are when an unfixed node can be decorated and left without a fully determined structural relation until later in the construction process. By definition, this is the building of one node within a structure whose position is not fully determined, so that update must be within that structure, precluding the possibility of a weaker restriction, that the update could be

70

Ruth Kempson,

Ronnie Cann, Jieun

Kiaer

across paired structures. Hence a form of update for (18) is also available in which an unfixed node within a S I N G L E structure can be updated by merging it with the node decorated by that pronoun itself: (18)

Ti MariaACC (ti) sinantise xtes. The Maria, (her) I met yesterday 'Maria, I met her yesterday.'

[Greek]

Γη(0),?7>(ί) Fo(i,x,Maria(x)),Ty(e)

,

Γ«(0)

_

.

MJspeaker),

?7>(β

_

Fo(Ufem), FoiXPi

Ρ)

Tv(cn)

13xFo(x),

Fo(Sinant),

0 Fo(x)

Fo(Maria)

Figure 3. Licensing pronoun-merge interaction

Whether this form of update is possible turns on whether the pronoun lexically imposes a terminal-node restriction. If it does, it will not allow such a process (retention of the terminal-node restriction is characteristic of English). 1 6 If it has no terminal node restriction, such an alternative is possible, allowing it to decorate a node which might turn out to be nonterminal in the resulting tree. In Greek, with its possible matching of case-marking on the clitic pronoun and left-peripheral expression, such an analysis appears to be required. 1 7 In standard Spanish, such an analysis appears to be essential to distinguish the dative clitic pronoun, as, out of all the clitics, it alone can be used to duplicate ALL forms of NP, quantified and referential: (19)

(20)

A nadie le devolvio Maria su manuscrito. To no-one CL^AT returned Maria his manuscript. 'To no-one did Maria return his manuscript.'

[Spanish]

A familias de pocos medios les ofrecieron queso y leche. To families of small means to them o f f e ^ / cheese and milk 'To low-income families, they offered cheese and milk.'

Topic, focus and the structural dynamics of language

71

All other clitic doubling, as in (21), involves a specificity restriction indicative of a pair of linked trees established through the anaphoric linkage (Escobar 1997). 18 (21)

A una secretaria que sabe hablar ingles, Pedro la estci AACC secretary that speaks English, Pedro proFem.ACC is buscando. looking for. 'Pedro is looking for a particular secretary that speaks English.'

Further confirmation of the analysis comes from the ambiguity of construal of Spanish subjects. Given the lack of morphological marking of case on the subject, and the pro-drop properties of the verb, projecting a full template of structure and a placeholder in subject position able to be identified contextually, we correctly anticipate that the subject may or may not be construed contrastively. 19 This account of clitic-duplicated left-dislocation structures matches the more familiar base generation vs movement accounts in many respects. We expect the relatively more constrained correlation between a left-peripheral expression and the position of its construal in structures where there is no anaphoric expression making explicit the correlation between what the leftperipheral expression annotates, and the node in the primary structure to which it contributes; as the locality restriction of requiring update within an individual tree is intrinsic to the definition of the unfixed node. We then expect the asymmetry between these structures and those with an anaphoric correlate, where there may or may not be a locality effect depending on the particular item of the particular language, for anaphoric relations themselves simply require accessibility of the requisite formula value from context, and the action which induces the transition from one linked structure to another may or may not itself impose a locality requirement. This account has the edge over the traditional base-generation vs movement dichotomy, in that the intermediate cases emerge as expected variants. 20 In particular, the various anaphoricity effects that are achieved do not need to be formalised with mechanisms distinct from regular processes. The only difference between pronouns which do allow update by node-merging processes and those which do not, is in the loss of the terminal-node restriction. Strong pronouns invariably retain such a restriction, behaving in all respects as regular lexical items. Clitic pronouns may have lost this restriction, though by no means all have done so. In all

72

Ruth Kempson, Ronnie Cann, Jieun Kiaer

other respects, these pronouns function as regular anaphoric expressions, the loss of the restriction merely ensuring that they have a wider distribution. With this mechanism of isolating one node and using that term as the basis for updating some antecedently established structure, we have a structural basis for focus effects. On the one hand, focus can be taken as an update to some prepositional structure as already constructed in context (using prior context like the concept of topic): (22)

A: What did you eat for B: Porridge.

breakfast?

The short answer is an update device that directly modifies the contextually provided structure given by parsing the question. On the other hand, in canonical long-distance dependency structures such as (5), we have the structure where some term is presented as isolated on an initially unfixed node, with some open structure to which it is an update, both presented within a single construction process. So, like the concept of topic, focus is just an aspect of the dynamics of general utterance-interpretation. 2 1

3. At the Right Periphery The less studied right periphery provides a new application area for these tools, where there are additional less well-understood restrictions. The challenge is whether the concepts of linked trees and unfixed nodes can explain the rather different topic and focus effects at the latter stages of the interpretation process, and we argue they indeed do so.

3.1. Pronoun Doubling A natural candidate at the right periphery for a LINK transition f r o m some completed node of type t onto one requiring type e - the inverse of the left periphery effect - is the Pronoun Doubling construction, with its characteristic background topic effect, in which the right-peripheral term serves as a reminder (see Herring (1994), inter al.): (23)

She talks too fast, Ruth

Kempson.

(24)

He's an idiot, that man at the

cashdesk.

Topic, focus and the structural dynamics of language (25)

Lo conosco, Giovanni. H i m I know Giovanni Ί know him, Giovanni.'

(26)

73

[Italian]

Maria lo compro, ese coche. Maria it bought, that car 'Maria bought it, that car.'

[Spanish]

In these structures, in virtue of the LINK relation, an anaphoric expression must be identified as co-referential with the formula annotating the rightperipheral structure which is optional. Tn(0) ,Ty(t), Fo(Talk — fast(RK))

(L-l)Tn(0),7Ty(e),l(U)Fo(RK)

Ty(e),

1t Fo(RK)

Fo(Talk — fast)

Figure 4. Parsing She talks too fast in (23) with LINK transition This, as displayed in Figure 4, accounts directly for optionality of the postposed noun phrase and the co-referentiality of it and the pronoun (23). The background topic effect is directly reflected in the defined sequence of actions. It arises because the pronoun is identified first as some contextually provided value and the right-peripheral expression as a decoration on the topnode of the linked structure has to agree with whatever value the pronoun has been assigned. The lack of availability of this type of reading at the left periphery is what we would expect, for at the left periphery the peripheral expression is interpreted as the context for subsequent identification of the pronoun; but at the right periphery, it is the other way round. The name is interpreted as having to be construed as being assigned the same term as already assigned as the construal of the pronoun. Hence its background topic effect, serving as a reminder. So we get our first taste of the asymmetry between left-right periphery effects derived from paying attention to the incremental dynamics of processing in context.

3.2. "Rightward movement" effects Such postposed "topics" are not the only use of such right-peripheral expressions. 2 2 Expressions postposed after the verb are regularly reported as having

74

Ruth Kempson, Ronnie Cann, Jieun Kiaer

a contrastive use (Herring 1994 and many others): 2 3 (27)

Tutie

wa-ss-ta

Chris.

Eventually c o m Q P A S T , D E C L Chris 'Eventually he came, Chris.'

[Korean]

This is not the end of the variability. In addition, in Italian, for example, Pinto 1997 reports that postposing of an indefinite subject N P buttresses a lack of specificity construal, the quantified expression thereby construed as taking narrow scope with respect to the tense specification: (28)

E' arrivato uno

studente.

Is arrived one student O n e student has arrived.'

[Italian]

In English, too, where the appearance of a subject in the preverbal position is obligatory, there is possible postposing to the post-verbal position of the associated complex subject expression, which appears to be yet another type of structure introduced late on in the interpretation process, that has to merge with the node already decorated by the pronoun: (29)

It is likely that Mary is wrong.

The occurrence of such expletives in English brings out the strictness of the left to right parsing process. Requirements on a node currently being parsed must be met by lexical action or by computational rule. The parse process cannot set aside actions provided by words if they happen not to fit the conditions imposed by the current node under development. Hence in English, like other non-subject-pro-drop languages, the existence of a particular f o r m of subject pronoun: without such a device, the parsing process would break down. Hence its role of projecting a place-holding metavariable and licensing moving on of the pointer. Yet such devices must be UNlike regular anaphoric devices in one particular respect: they must have lost the restriction ensuring that they only decorate a terminal node, for the metavariable they project must be able to be updated by arbitrarily complex structure. This is indeed the observable property of the expletive it in English (see figure 5). The essential right-roof constraint that is the corollary of these processes (Ross 1967) emerges f r o m the dynamics of the left-right parse. W h e n the pointer returns to the dominating type-f-requiring node, having licensed a

Topic, focus and the structural dynamics of language

75

Tn(Q),lTy(t)

.--v-/. ?3x.Fo(x),0

Ty(t-*t), Fo (Likely)

(U)(h)Tn{0),{U)Tn(n),Ty(t),Fo(Wrong(Mary)), ?3x.7n(x)

Ty(e),

Fu\mury)

Fo (Wrong)

Figure 5. Parsing "ft's likely that Mary is wrong"

shift in the parse to the predicate node, no decorations will be able to be compiled at the topnode since an outstanding requirement remains at that daughter node (reflecting strict compositionality on the tree itself). Accordingly the pointer will have to return to that node in order to license actions allowing that requirement to be met. A process of Late *Adjunction thus has to be invoked to enable an appropriate structure to be completed. 2 4 This then provides the decoration, which, with a step of Merge unifying the two nodes, satisfies the requirement which was otherwise preventing successful decoration of the mother node. The right roof constraint is thus an immediate consequence of allowing a node to be but half-decorated, imposing the need of return to that node before decorations further up the tree can be compiled. The result of licensing such a sequence of actions is that we have a mechanism which makes available late addition to skeletal structures lacking one possibly complex piece of structure to yield a new propositional whole. Following the pattern of extraposition, such late introduction of construal for the subject may be associated solely with buttressing some narrow scope effect (as in Italian). Given its presentation late on in the construction process, it may be taken to be a departure f r o m what is in the context, hence contrastive. As we would expect, in languages in which the particular argument is licensed directly by the verb, a pro-drop structure, there will in consequence be processing ambiguity as to whether the postposed term is to be construed contrastively, via application of Late *Adjunction, or as a background topic, via construction of a linked structure. And this is the reported situation for languages such as Spanish. As at the left periphery, the postposed subject may or may not be construed contrastively:

76 (30)

Ruth Kempson, Ronnie Cann, Jieun Kiaer Cantar muy bien Maria. Sings very well Maria 'She sings very well, Maria.'

[Spanish]

Just as there were mixed effects at the left periphery, so we expect the availability of mixed effects. In particular, we can combine left-periphery effects with right-periphery effects. So in Spanish, which displays the freest word order variation of the Romance languages, (31) is wellformed: (31)

Uη coche compro, A car bought

Maria. Maria

' M A R I A b o u g h t A CAR.'

Such examples, notably, involve a contrastive interpretation, and they confirm the analysis proposed. Within the framework, all partial trees along the process of growth must be wellformed, with all nodes uniquely identifiable, and this precludes the presence in any partial tree of more than one node analysed as unfixed at any one time. But, as we would expect in (31), the node decorated by un coche, which is introduced as unfixed, has been assigned a determinate position by the time that the node decorated by Maria is introduced through Late *Adjunction. So (31) is wellformed, a legitimate means of conveying multiple focus interpretations. The presentation of un coche is taken to decorate a node to be treated as a subsequent update for the propositional structure to be set up in parsing compro, but so too is Maria. Hence the availability of a multiple focus interpretation.

4. Conclusion Across a range of left and right-periphery effects, we have shown how the mechanisms of building linked structures and unfixed nodes jointly provide a basis not merely for modelling familiar topic and focus effects at the left periphery, but also at the less-studied right periphery. The explanation has the bonus of explaining the right-roof constraint without special stipulation. M o r e centrally to this paper, there is no superimposed independent level of information structure. All that is necessary to articulate concepts of context and information update which constitute the underpinnings to structural concepts of topic and focus is the exposition of the dynamic architecture intrinsic to natural-language processing.

Topic, focus and the structural dynamics of language

11

Notes 1.

2.

3. 4.

5.

6.

7.

8. 9.

This paper reports collaborative work on a new framework. Thanks to all, especially Wilfried Meyer-Viol, Masayuki Otsuka, Lutz Marten, Eleni Gregoromichelaki, Miriam Bouzouita. This work was supported in part by the Leverhulme Trust, with a research professorship to the first author for which we are extremely grateful. See Fodor (1981), and others following, for the view that human reasoning is essentially syntactic, defined in terms of licensed moves from one representation to another. The context against which utterance interpretation takes place is also represented as a set of trees. The Formula values that decorate the nodes are not words of the natural language. In describing their associated types, type e is for expressions that denote individuals (e for "entity"); type t is for prepositional formulae, that denote truth values (t for "truth-value"). One-place predicates are functions from individuals to truth-values of type (e —> t). Two-place predicates are of type (e —> (e —> ?)). The type cn is used for nominals, analogous to Ν Ρ in theories which distinguish DP and NP (see Kempson et al. 2001). The core of the framework is the Logic of Finite Trees (Blackburn and MeyerViol 1994): (T)X is defined from node η as "X holds at the mother of n", ( | ) X from node η as "X holds at a daughter of «". In any partial tree, there is one node indicated by a pointer, 0 , as the node under development. In this framework all noun phrase construals are taken to be of type e, matching arbitrary names manipulated in natural-deduction proofs. Accordingly, the terms onto which words map are lambda terms within the epsilon calculus, which provides the formal study of arbitrary names (see Meyer-Viol 1997). Formally, this is defined using the Kleene star operation defined on the daughtermother relation: (Τ*)Γη(α) is a node dominated by some node Tn{a), where Tn(a) is along some arbitrary sequence of mother relations from the current node to Tn(a). Adding the requirement 13x.Tn(x) as an additional decoration on that node imposes the requirement that in all successful completions of the tree, this underspecified characterisation is replaced by a fixed tree relation. See Kaplan and Zaenen (1989) for use of the Kleene * in defining 'functional uncertainty' in LFG, equally an account of long-distance dependency. The difference between this and the Kaplan and Zaenen account lies in the concept of growth articulated in Dynamic Syntax as part of the construction process. See Kiaer in preparation for a defence of the view that this explanation extends to phonological stress. Formally, this involves the combination of the LINK transition defined as introducing a LINK relation from the head node decorated with some term Fo(a)

78

Ruth Kempson, Ronnie Cann, Jieun

Kiaer

to a new type-?-requiring node, and the added requirement ? ( | * ) F o ( a ) - viz. the requirement that some daughter node to this top node must be decorated with a copy of the very same term that decorates the head from which the L I N K transition was built. 10. Nodes decorated with type e formulae may have internal structure containing a variable-binding term-operator which is the quantifying device, a variable also of type e, and a restrictor. Restrictive relative construal involves defining a L I N K transition from the node decorated by the variable. Nonrestrictive relative construal involves a transition from the top-node of this structure, also of type e. In a fuller characterisation of natural-language names, these would be analysed as iota terms built up from such sub-structures. 11. One restriction not covered by the assumption that adjunct structures are to be constructed as independent linked trees is the Sentential Subject Restriction. Arguably, this reduces to the restriction of there being only one unfixed node within a tree at a time, but we leave that on one side. 12. This was first identified as Clitic Left Dislocation (CLLD) by Cinque (1991), who argued that despite their sensitivity to strong island effects, these data nevertheless required a base generation account. This range of variation is widespread. For example, Aissen (1992) argued that Mayan languages differ according to whether they have an external topic as an independent structure, separated by an intonational break, or have no such break but are able to occur in subordinate clauses. 13. There may even be variation between forms within a single language, as in Romanian, which has one left-dislocation structure associated with one kind of morphological marker, which displays no island restrictions, a second which does (see Kempson et al. 2001). 14. The difficulty in sustaining this dichotomy is increasingly leading to analyses which depart from it (see Adger and Ramchand 2005). 15. On this analysis, one might expect further that if a language had morphological means of distinguishing more than one type of such structure, these might differentiate between different forms of locality constraint, yielding the type of variation observed in Romanian. 16. The only exception is expletive it, paired with propositional formulae. English relative clauses, in which resumptive use of personal pronouns is an available, albeit marked option, do not provide evidence of any such loss of restriction as in these LINKed structures, it is only the formula value that is copied over into the LINKed structure, and not any ancillary structure (see Cann et al. in prep.). 17. Object clitic pronouns in Greek are standardly analysed as an object agreement marker. The idiosyncratic position of clitics before the verb, which is problematic in other frameworks in occurring in positions in which full noun phrases may not, is here analysed as a lexical calcification of what in the common root to the modern Romance languages, Latin, was a free process of a localised form

Topic, focus and the structural dynamics of language

79

of * Adjunction which, by the licence of introducing an unfixed node whose position in some local predicate-argument structure is immediately determined by case, enables argument nodes to be processed in any order (see Cann et al. 2005 for an account of Bantu argument prefixes and Japanese short scrambling in these terms). 18. The specificity arises because scope is defined relative to individual propositional structures, having been established incrementally during the parse process: indefinite noun phrases are the only apparent exception, their scope being indefinitely extendable (see Kempson et al. 2001, ch. 7). Rioplatense accusative clitics appear to be moving in the direction of the dative, though retaining a specificity restriction on doubling. See Suner (1988). 19. See Belletti (1999) for arguments that Spanish subjects are invariably external to the clause. 20. The Boeckx (2001) account of resumptive pronouns purports to provide a unitary characterisation of resumption. However, it explicitly relies on a separation of "true" resumptive and "intrusive" pronouns, following Sells and Chao (1985), dismissing the optional resumptive use of pronouns in English: (i) The new guy from history who Sue says he's interested in Celtic studies is coming to the party. But the Sells and Chao account which defines so-called intrusive pronouns as having a distinct Ε-type form of analysis, is problematic. Ε-type forms of interpretation, in which the appropriate witness for some set constitutes the denotation, require a semantic computation over the entire clause which provides the antecedent, and are unavailable at the level of providing an interpretation of the pronoun itself. In relying on their analysis, the Boeckx analysis is at best incomplete. 21. As expected, some languages allow mixed effects, with a left-dislocated constituent paired with a pronoun which is itself at a potentially long-distance remove from the site at which it is to be construed. This phenomenon has been observed in Hebrew, but it occurs also in English, combining both topic and focus strategies: (i) As for Shalom, he I think should be given the position. 22. Rightward movement within Minimalism is precluded by Kayne (1994). 23. There are multiple options in verb-final languages as to whether the nominal and doubling clitic are both marked with case, or both marked with a topic-marker, or only the pronominal so marked (see Kiaer in preparation). 24. This process might seem unrelated to *Adjunction. However, this is a natural consequence of such late application, given the update of resolving tree-node underspecification, in which an unfixed node is progressively percolated down the tree, so that at the site of unification there is both the node with open typerequirement, and the unfixed node. So a process introducing an unfixed node of

80

Ruth Kempson, Ronnie Cann, Jieun Kiaer type X from a node of matching type is already implicit, and at this late stage, with all structure otherwise introduced, no other variant is available.

References Adger, David and Gillian Ramchand 2005 Revisiting wh dependencies: Merge and Move. In Linguistic Inquiry 36, 61-93. Aissen, Judith 1992 Topic and Focus in Mayan. Language 1992. (68.1) 43-80. Anagnastopoulou, Elena; H. van Reimsdijk and F. Zwarts (eds.) 1997 Materials on Left Dislocation. Amsterdam: John Benjamins. Belletti, Adriana 1999 "Inversion" as focalization. Catalan Working Papers in Linguistics 7, 9-45. Blackburn, Patrick and Wilfried Meyer-Viol 1994 Linguistics, logic, and finite trees. Bulletin of Interest Group of Pure and Applied Logics 2, 2-39. Boeckx, Cedric 2001 Mechanisms of Chain Formation. Ph.D. dissertation, Department of Linguistics, University of Connecticut. Büring, Daniel 1997 The meaning of topic and focus: The 59th Street Bridge Accent. Routledge. Cann, Ronnie; Ruth Kempson and Lutz Marten 2005 The Dynamics of Language. Oxford: Elsevier. Cann, Ronnie; Tamar Kaplan and Ruth Kempson in prep. Data at the Grammar-Pragmatics Interface: the case of resumptive pronouns in English. In R.D. Borsley (ed.), special volume of Lingua. Chomsky, Noam 1965 Aspects of the Theory of Syntax. Cambridge: MIT Press. Chomsky, Noam 1971 Deep structure, surface structure and semantic interpretation, in D. Steinberg and L. Jacobovits (eds). Semantics- An Interdisciplinary Reader, Cambridge. Cambridge University Press, 193-216. Cinque, Guglielmo 1991 Types of A '-Dependencies, Cambridge, MA: MIT Press. Erteshik-Shir, Nomi 1997 The Dynamics of Focus structure Cambridge: Cambridge University Press.

Topic, focus and the structural dynamics of language

81

Escobar, Linda 1997 Clitic left dislocation and other relatives, in Anagnastopouloue, E., Reimsdijk, H. van Zwarts, F. (eds.) 1997. Materials on Left Dislocation, 233-273. Amsterdam: John Benjamins. Fodor, Jerry A. 1981 Modularity of Mind. Cambridge: MIT Press. Halliday, Michael A. K. 1967 Notes on Transitivity and Theme in English, Part II, Journal of Linguistics 3, 199-244. Hausser, Roland 1989 Computation of Language. Berlin: Springer. Herring, Susan C. 1994 Afterthoughts, antitopics, and emphasis: the syntactization of postverbal position in Tamil in M. Butt, T. King, G. Ramchand (eds.) Theoretical Perspectives on Word Order in South Asian Languages, 119-152. CSLI. Kamp, Hans and Uwe Reyle 1993 From Discourse to Logic. Dordrecht: Reidel. Kaplan, Ronald M. and Annie Zaenen 1989 Long-distance dependencies, constituent structure, and functional uncertainty. In Alternative Conceptions of Phrase Structure, M. Baltin and A. Kroch (eds.), 17-42. University of Chicago Press. Kayne, Richard 1994 The Anti-symmetry of Syntax. Cambridge: MIT Press. Kempson, Ruth and Jieun Kiaer 2004 Japanese scrambling as growth of semantic representation, ms. King's College, London. Kempson, Ruth; Wilfried Meyer-Viol and Dov M. Gabbay 2001 Dynamic Syntax: The Flow of Language Understanding. Oxford: Blackwell. Meyer-Viol, Wilfried P. M. 1997 Instantial Logic. Ph.D. thesis. University of Utrecht. Pinto, Manuela 1997 Licensing and Interpretation of Inverted Subjects in Italian. Ph.D. thesis. University of Utrecht. Ranta, Aarne 1994 Type-Theoretic Grammar. Oxford: Clarendon Press. Rizzi, Luigi 1997 The fine structure of the left periphery. In Haegemann, L. (ed.) Elements of Grammar. Dordrecht: Kluwer. Ross, John Robert 1967 Constraints on Variables in syntax. Ph.D. thesis. MIT.

82

Ruth Kempson, Ronnie Cann, Jieun Kiaer

Schwarzschild, Roger 1999 GIVENness, AvoidF and other constraints on the placement of accent. Natural Language Semantics 7. Sells, Peter and Wynn Chao 1985 On the interpretation of resumptive pronouns. ΝELS 13. Sgall, Petr; E. Hajicöva and J. Panevovä 1986 The Meaning of Sentence in its Semantic and Pragmatic Aspects, Dordrecht: Reidel. Sperber, Dan and Deirdre Wilson 1995 Relevance: Cognition and communication (2nd edition). Oxford: Β lack well. Suner, Margherita 1988 The role of Agreement in clitic-doubled constructions. NLLT 6. Vallduvi, Enric 1991 The informational Component. Ph.D. Pennsylvania.

On the (in)dependence relation between syntax and pragmatics Joäo Costa and Maria Cristina Figueiredo Silva

Abstract This paper presents evidence from European and Brazilian Portuguese for arguing that notions like topic and focus are not syntactically encoded. Instead, it is proposed that syntax feeds information structure. The arguments come from the observation that both varieties of the language display structures in which word order appears to be conditioned by information structure aspects, although, when only focalization of subjects is concerned, it may appear that Brazilian Portuguese is not discourse-configurational. The similarities between the two languages and the fact that the differences can be attributed to syntactic aspects leads us to propose that the source of variation is primarily syntactic and not related to discourse aspects.

1. Introduction In recent literature (e.g. Belletti 2001, Costa 1998, Ordonez 1997, Rizzi 1997, Zubizarreta 1998), the following two generalizations concerning word order and information structure in Romance have been made: (a) Subject-verb inversion is used in focalization contexts; (b) Topic elements may be fronted. According to these generalizations, there is an obvious relation between syntax and information structure. However, there is little consensus regarding the exact nature of this relationship. In particular, this type of statements raises at least the following questions: (a) How can the syntactic effects of information structure be accounted for? Are the several word orders found to be explained in terms of movement to discourse-related functional categories (Rizzi 1997) or do they derive from more general interface conditions between syntax and constraints on pragmatics? (b) Is there a language split, as far as the codification of topic and focus is concerned, as proposed in Kiss (1995)

84

Joäo Costa and Maria Cristina Figueiredo

Silva

and Horvath (1986)? In other words, are foci codified syntactically in some languages and prosodically in other languages, or, alternatively, syntax and prosody must interact in order to account for the patterns of focalization found? (c) What types of information-related notions are relevant in order to properly describe the effects of information-structure alternations found? Is the topic-focus split sufficient? Do definiteness effects reduce to topic-focus alternations? Is the given-new difference enough? The primary goal of this paper is to discuss some micro-variation facts in two varieties of Portuguese (European and Brazilian) that shed light into these questions. It will be shown that it is not possible to identify a uniform behavior regarding an alleged split between the two languages as far as the interaction with focus is concerned. For clarity, we are assuming focus to be the part of the sentence that yields new information, and topic to be the part of the sentence that refers to given information. Of particular interest for this paper will be the difference between "sentence-focus" and "subject-focus". We will be assuming that a sentence-focus context occurs, when the whole sentence conveys new information (as in answers to "what happened?"), while subject focus occurs when only the subject conveys new information. Based on the observation that it is not possible to account for the differences between Brazilian and European Portuguese in terms of a discourse-related parameter, we will propose that there is, in fact, no need for parametrization as far as the relation between syntax and information structure is concerned. On the contrary, we will argue for a modular approach, in which syntax feeds the information structure component, which uses the best syntactic output for compliance with its own constraints.

2. Focalization in Portuguese Let us start by reviewing some data concerning the syntactic distribution of information focus in Portuguese. As shown in Costa (1998), among others, European Portuguese follows the generalization presented above for other null-subject Romance languages: focused constituents are rightmost, and, if a subject is focused, there is subject-verb inversion. In Brazilian Portuguese, in exactly the same context, a subject which is in-

On the relation between syntax and pragmatics

85

formation focus does not invert, remaining in preverbal position. This contrast between the two languages is illustrated in (1): (1)

A: B:

Quern comeu ο bolo? Who ate the cake a. Comeu ο Joäo. Ate Joäo a'. Ο Joäo comeu. Joäo ate

(EP/*BP) (*EP/BP)

It is interesting to note that in this respect Brazilian Portuguese behaves like English: in both languages, an information-focused subject does not invert, as shown in (2): (2)

Who ate the cake? John did.

The lack of correlation between word order and information structure in English has led authors like Horvath (1986), among others, to propose that languages split in the way they encode information structure. More precisely, Horvath (1986) proposes that English marks focus at PF, while languages like Hungarian do so in the syntax. Some authors in Kiss (1995) argue for a similar hypothesis proposing the existence of a discourse-configurationality parameter, according to which languages codify information structure in the syntax only if this parameter is set positively. Extending this reasoning to the difference illustrated in (1) for European and Brazilian Portuguese, which, as shown, is similar to the difference between European Portuguese and English, it is legitimate to ask whether there is a parametric split between the two varieties of Portuguese as far as discourse-configurationality is concerned. These two languages provide a nice empirical ground for testing the parametric approach to discourse-configurationality, since they are very similar, and instantiate a case of micro-variation. So far, we have only considered the behavior of information-focused subjects in interaction with a transitive verb. However, if we take into account what happens in unaccusative contexts, it is possible to observe that the two languages behave partly alike. In sentence-focus contexts, both permit SV or VS, as shown in (3), while in subject-focus contexts,

86

Joäo Costa and Maria Cristina Figueiredo Silva

the behavior is the same as found with transitive verbs: there is subjectverb inversion in European Portuguese, but not in Brazilian Portuguese: (3)

A: B:

(4)

Α: Β:

Ο que aconteceu? What happened a. Chegou ο Joäo. Arrived Joäo a'. Ο Joäo chegou. Joäo arrived 'Joäo arrived.' Quem chegou? Who arrived a. Chegou ο Joäo. Arrived Joäo a'. Ο Joäo chegou. Joäo arrived 'Joäo arrived.'

(EP/BP) (EP/BP)

(EP/*BP) (*EP/BP)

Based on these data, it seems that there is no easy answer to the question raised above, concerning the existence of a discourse-configurationality parameter. The data in (3) appear to indicate that the two languages use word order in the same way in a specific context, while the data in (1) and (4) appear to indicate that the two languages differ. We are, therefore, left with two questions: i) How to understand this difference in behavior, apparently related to verb classes; ii) Can this difference between the two languages reveal something concerning the relation between syntax and discourse?

3. Hypothesis The data presented above appear to indicate that a parameter stating that Brazilian Portuguese is not discourse-configurational does not account for the unaccusative context, in which both European and Brazilian Portuguese use word order to encode discourse-related information. Based on this observation, we make the following hypotheses:

On the relation between syntax and pragmatics A-

87

There is no parameter specifically related to information-structure configurationality.

In other words, we reject the idea that the trigger for surface differences concerning the codification of discourse-related notions is to be explained in terms of a macro-parameter stating that some languages resort to word order to encode topic and focus, while others do not. B-

The following universal holds: Word order is used for codifying focused XPs, whenever syntax allows it. If syntax prohibits it, stress is used, as last resort.

The second hypothesis we are putting forward is that languages with a more flexible syntax use word order for placing XPs in positions that are more suitable for serving information-structure purposes. The condition is that such a position is made available by the syntax of the language. In other words, if syntax generates two or more outputs, information structure requirements (defined independently of syntax) may specify which serves their purposes more adequately (cf. Adger 1994). This may help understanding why prosodic marking always affects foci, while word order is only used in some (constructions of some) languages (as argued for several language by Givon 1990). C-

The topic-focus split is best understood as a scalar relation for which at least three types of information converge: givenness, de finiteness, and quantification. This claim contradicts the categori cal Mapping Hypothesis of Diesing (1992), since it predicts some mismatches between the three scales.

It follows f r o m A and Β that one does not expect categorical mappings between information-structure related categories and syntactic units. At best, we can find tendencies for placing certain constituents with a given discourse function in certain positions, but, as spelled out in B, if the syntax of the language does not allow it, it is predicted that the sentences are still felicitous, if stress is used. In the next section, we will be presenting some further arguments for the hypotheses just presented, and we will show how they account for the differences observed between European and Brazilian Portuguese.

88

Joäo Costa and Maria Cristina Figueiredo Silva

4. Arguments and analysis

4.1. Similarities between European and Brazilian Portuguese. The first aspect to note that points into the direction that drawing a distinction between languages in terms of whether they codify information structure notions in the syntax or not is not a productive tool comes from the observation that Brazilian and European Portuguese share several constructions that must be understood in terms of an interaction between syntax and discourse. In other words, if one were to account for the fact that Brazilian Portuguese patterns like English in the behavior of focused subjects, stating that the former does not use syntax for codifying information-structure notions, the following similarities would require further stipulations. First, both languages have topic-identified null objects (Raposo 1986), as illustrated in (5), that is empty categories in object position, that can be identified by a referent salient in previous discourse: (5)

A: Β:

Ε esse carro? What about that car Ο Pedro afinal decidiu näo comprar. Pedro after-all decided not to buy 'After all, Pedro decided not to buy it.'

As argued in Raposo (1986), Duarte (1987), and Costa and Duarte (2001), among others, null objects and topicalized elements share the same properties. In fact, both languages resort to a syntactic strategy to mark/retrieve contrastive topics, 1 as shown by the availability of topicalization without the presence of a resumptive clitic (Duarte 1987): (6)

Ο The

bolo, cake,

ο Pedro Peter

comeu. ate

Note that, according to Duarte (1987), topicalization displays properties typical of syntactic adjunction. In this sense, both topicalization and null objects instantiate two strategies in which (contrastive) given information is codified in the syntax. If Brazilian Portuguese were a non-

On the relation between syntax and pragmatics

89

discourse configurational language, as the data in (1) might suggest, we would expect topics not to be marked syntactically. Another context in which the two languages behave alike is found in the word order alternations in ditransitive contexts (Salles 1997): (7)

Brazilian Portuguese: A: Ο que ο Joäo deu pra What Joäo gave to Β: Ο Joäo deu pra Maria Joäo gave to Maria

Maria? Maria urn CD. a CD

(8)

Brazilian Portuguese: A: Pra quem ο Joäo deu ο CD? To whom Joäo gave the CD Β: Ο Joäo deu ο CD pra Maria. Joäo gave the CD to Maria

(9)

European Portuguese: A: Ο que έ que ο Joäo deu What Joäo gave Β: Ο Joäo deu ä Maria um Joäo gave to Maria a

(10) European Portuguese: Α: Α quem e que ο To whom Β: Ο Joäo deu ο CD Joäo gave the CD

ä Maria? to Maria CD. CD

Joäo deu ο Joäo gave the ä Maria. to Maria

CD? CD

The data in (7)-(10) make two different points. First, both languages allow for either complement of the verb to surface clause-finally. Second, as shown by the question-answer pairs, the choice between order VDO-IO and V-IO-DO may be made in compliance with discourse requirements, as argued in Costa (2002). This case is of particular interest for the discussion in this paper, since we observe that both languages satisfy the requirement that the focus of the sentence surfaces rightmost. In fact, this is reproduced in (1) for the subject of the transitive verb in European Portuguese. Therefore, positing that the subject in Brazilian Portuguese does not invert in (1) because there is no relation between

90

Joäo Costa and Maria Cristina Figueiredo Silva

syntax and information structure would not account for the behavior of complements of ditransitive verbs. Let us finally consider the case of VP-ellipsis. Both languages display this construction, as shown in Cyrino and Matos (2002): (11)

Ο Pedro

leu

ο

Pedro read the

livro

e

book and

a Maria

tambem

leu.

Maria

also

read.

(BP/EP)

As argued in Matos (1992), elided material must be topical. Therefore, it may be seen that both languages resort to a syntactic strategy of ellipsis in order to delete topics. Again, if only European Portuguese were discourse-configurational, this difference would not be accounted for. Summarizing, these data show that European Portuguese and Brazilian Portuguese do not differ a lot concerning the existence of syntactic constructions for which information-structure requirements are to be taken into account. If one were to assume that Brazilian Portuguese does not use syntax or word order to encode discourse-related information, these similarities would remain unexplained. These data seem to support our hypothesis A, that there is no need for a parameter related to discourse configurationality.

4.2. Inversion vs. non-inversion Since a parametric split does not account for the difference between the two languages illustrated in (1), we must look for a different explanation. In what follows, we claim that the impossibility of inverting focused subjects in transitive and in intransitive contexts in Brazilian Portuguese derives from a syntactic fact. In particular, we will argue that it relates to the fact that Brazilian Portuguese has lost referential null subjects (Figueiredo Silva 1996, among others). Let us first consider two potential alternative hypotheses that would not solve the problem. First, suppose one were to propose that there is no subject-verb inversion in Brazilian Portuguese, because focused constituents are not rightmost in this language. Such a hypothesis would leave the data with ditransitives unaccounted for. On the contrary, if one were to assume that discourse-configurationality is the same in European and Brazilian Portuguese, the ditransitive contexts would be ex-

On the relation between syntax and pragmatics

91

plained, but the lack of inversion illustrated in (1) for Brazilian Portuguese becomes unexpected. Our claim is that subject-verb inversion is, at least in part, a consequence of the null subject parameter. As argued in Rizzi (1982), subject-verb inversion is typical of null subject languages. Assuming this, one expects that subject-verb inversion is only made available in languages in which the null-subject parameter is set positively. Accordingly, it is expected that the syntax of a non null-subject language does not allow subject-verb inversion even for focus purposes. Let us then see how this assumption, together with the proposal made in the previous section, derives the difference between European and Brazilian Portuguese. Let us first consider the case of European Portuguese. Since this is a null-subject language, subjects may be licensed either in Spec,IP or in their base-generated position (see Costa 2003 for a recent proposal on the mechanisms involved for licensing the subject). Accordingly, syntax generates two converging outputs, one with VS order, and another with SV order. Post-syntactically, information-structure considerations apply. As argued by many authors (e.g. Givon 1990), information-structure requirements force information focus to be prominent. A way of assigning prominence is to stress a constituent. 3 According with recent proposals made in the literature on the syntax-prosody interface, the Nuclear Stress Rule assigns most prominent stress to rightmost constituent (Cinque 1993, Nespor and Vogel 1986, among others). Following Reinhart (1995), we will assume that heavy marked stress applies as last resort only if the Nuclear Stress Rule does not yield a legitimate output. Based on these assumptions regarding prosody, it is now possible to explain why a VS order is selected in subject-focusing context in this language: this word order is a legitimate syntactic output. Nuclear stress rule applies stressing the rightmost constituent - the subject. By receiving the sentence nuclear stress, the subject becomes prominent, in compliance with the information structure requirement that focus be prominent. Therefore, it is legitimate to link VS to subjectfocusing context. Let us now turn to the Brazilian Portuguese case. As mentioned above, since referential pro is no longer available in this language, it no longer qualifies as a null-subject language, and VS orders are not legitimate syntactic outputs. Thus, the basic syntactic difference between the two languages is that, unlike European Portuguese, Brazilian Portuguese does not generate two different outputs: only SV is grammati-

92

Joäo Costa and Maria Cristina Figueiredo Silva

cal. As before, information-structure considerations require that information focus be prominent, and prosody assigns prominence in a similar way. Note that, since VS is not a legitimate syntactic output, this word order is not available for linking with the discourse context at stake. Since there is only SV, if the Nuclear Stress Rule applies stressing the rightmost constituent, the verb will bear the sentence nuclear stress. Therefore, as last resort, stress-shift must apply, and a heavy stress is assigned to the preverbal subject. This analysis makes the strong prediction that whenever the syntactic outputs are the same in the two languages, there will be no difference concerning codification of information structure. This is confirmed for ditransitive contexts. Hopefully, this claim will extend to languages other than Brazilian and European Portuguese (see Costa 2004). Note that, under the analysis just presented, syntax proper is not resorting to information-structure categories. The difference in word order is to be explained in purely syntactic terms, as a consequence of the setting of the null subject parameter.

4.3. The case of unaccusative verbs The analysis presented in the previous subsection does not yet account for the fact that Brazilian and European Portuguese partly pattern alike in unaccusative contexts. Recall that inversion is available in sentencefocus contexts in the two languages, while in subject-focus contexts only European Portuguese allows for VS orders. According to the proposal made above, this difference in behavior internal to Brazilian Portuguese should in principle follow from syntactic constraints alone. In fact, it is possible to observe that expletive pro exists in Brazilian Portuguese. The language has lost referential null subjects but not expletive null subjects, as illustrated in the following examples: (12) a. Τά chovendo. Is raining b. Parece que ίά chovendo. Seems that is raining Following Costa (1998), we will assume that VS structures have different syntactic representations, depending on whether the verb is unac-

On the relation between syntax and pragmatics

93

cusative or (in)transitive. In particular, we adopt the proposal that Spec,IP is empty only in the latter contexts, as in (13a). In unaccusative contexts, it may be occupied by an expletive pro, as in (13b): (13) a. [ipV [ypSubj ty]] b. [ippro V [ y p t v D P ] ] Evidence put forward in Costa (1998) for these representations stem from different sets of data. Let us consider two of his arguments. First, there are differences in agreement between unaccusatives and (in)transitives in European Portuguese. In contexts of inversion, an unaccusative verb may fail to agree with the postverbal DP, exhibiting 3 r d person singular agreement, which may be interpreted as agreement with expletive: (14) a. %Chegou Arrived b. *Nadou Swam

muitos many muitos many

meninos. children meninos. children

Second, there are dialects with overt expletives (Raposo and Uriagereka 1990). In inversion contexts, the presence of such expletives is restricted to unaccusative contexts. These two facts permit supposing that the difference between inversion with unaccusatives and (in)transitive verbs lies on the fact that expletive pro is used only in the former. Accordingly, the availability of unaccusative inversion in Brazilian Portuguese follows straightforwardly. Since expletive pro is available in this language, inversion will be possible only when Spec,IP is filled in, in compliance with the null subject parameter. For this proposal to hold, we are assuming with Coelho et al. (2001) that Spec,IP must be filled in in Brazilian Portuguese. In other words, in these terms, the Null subject parameter is to be understood as the interaction between three constraints: a) b) c)

Is Spec,IP projected? Does Spec,IP have to be lexically filled? Does pro exist in the pronominal system?

94

Joäo Costa and Maria Cristina Figueiredo Silva

As shown in Coelho et al. (2001), Brazilian Portuguese instantiates a case in which Spec,IP is projected, and must be filled. Fillers for Spec,IP may be DP subjects or pro. Since there is only expletive pro in the pronominal paradigm, the only pronominal null subjects are expletive. This analysis accounts for the fact that Brazilian Portuguese permits inversion. Note that it still does not say anything concerning the unavailability of inversion in subject-focus contexts. A full answer to this problem will be given in section 5. In short, we can anticipate that the gist of the argument is in the observation that expletive pro is not available in inverted constructions in which the subject is the focus.

4.4. Stress-marking similarities In the previous subsections, we have accounted for the cases in which the two languages differ, showing that such differences follow from our proposal, which predicts that whenever the syntactic outputs are different, we expect to find differences in word order in given contexts. There is, however, another prediction that has not been tested yet. Our hypothesis predicts that European Portuguese should resort to marked stress as last resort, if for some syntactic reason, there is no legitimate syntactic output placing the focused constituent at the position where sentence nuclear stress falls. In what follows, we show that this prediction is borne out. Both in Brazilian Portuguese and in European Portuguese, stress is used to mark information focus whenever the focus is a head, and whenever the focalized constituent cannot surface clause-finally. Let us consider three examples. The first case comes from European Portuguese. As shown in Costa (2002), in ditransitive contexts, when the focused element is a QP binding another XP, it does not surface clausefinally. As shown in the examples below, if no binding is involved, we observe the usual behavior: the focused constituent surfaces at the rightmost position: (15) Α: B:

Α quem e que deste ο to whom did you give the Dei ο livro [p ao Paulo] (I)gave the book to Paulo. #Dei ao Paulo ο livro. '(I) gave to Paulo the book.'

livro? book

On the relation between syntax and pragmatics (16) A: B:

Ο que έ que deste what did you give Dei aο Paulo [f ο (I) gave to Paulo the #Dei ο livro ao Paulo. '(I) gave the book to Paulo.'

95

ao Paulo? to Paulo livro]. book

However, if the focused constituent must bind an anaphor into the non-focused constituent, the syntactic requirement that the binder ccommands the bindee forces the focused constituent not to surface clause-finally:

(18) A: B:

Λ quem to whom Dei [f (I) gave ?*Dei (I) gave

deste did you a cada autor] to each author ο seu livro book his

Ο que έ what Dei [p (I) gave l*Dei (I) gave

que

e que

os livros? give the books? ο seu livro. his book. a cada autor. to each author

deste did you give cada livro] ao each book to ao seu autor to its author

aos to the seu its cada each

autores? authors autor. author livro. book

It is crucial to observe that in (17) and in (18), the non-final elements are stressed, just like the preverbal subject in Brazilian Portuguese is stressed in (4a'). In both cases, the constituents involved do not surface clause-finally for syntactic reasons. The two other cases to be considered are common to European and Brazilian Portuguese. So far, we have just been mentioning cases in which the focused constituent is an XP. With heads, matters are slightly different. Let us first consider what happens when the focus falls on the verb. As shown in (19), there is no word order difference, but just stress on the verb: (19) A:

Ο que fizeste ao What did you do to the

livro? book?

96

Joäo Costa and Maria Cristina Figueiredo Silva B:

Eu QUEIMEI I burnt WO livro, The book, *Eu, ο livro, I the book

ο the eu I

livro. book queimei. burnt queimei. burnt

The same holds for focus on prenominal possessives, as shown in Castro and Costa (2002): (20)

A: Β:

Que livro compraste? Which book did you buy Ο TEU livro. The your book *0 livro TEU the b o o k your

According to Castro and Costa (2002), pronominal possessives are heads, hence we expect them to pattern like the verb in (19). H o w can we interpret these data? Syntactic outputs resulting f r o m syntactic operations that are eligible for being picked for information purposes affect XPs. In other words, scrambling, topicalization, left-dislocation, right-dislocation affect XPs only. Crucially, heads do not enter into any of these constructions. Therefore, we again observe that European Portuguese and Brazilian Portuguese are not different in this respect: in the lack of a legitimate syntactic output, both languages resort to marked stress in order to signal focus. Note that the parametric approach to discourse-configurationality is disfavoured again, since the two languages pattern alike. As in the case of inversion in Brazilian Portuguese discussed above, the context of binding and the data with focus on heads show that in the lack of multiple syntactic outputs or in the lack of a syntactic output placing the focused constituent at the rightmost position, marked stress is used as last resort.

On the relation between syntax and pragmatics

97

5. Relaxing the mapping between syntax and the other levels Let us summarize our results so far. The data presented above permit drawing the following conclusions: a)

b)

The syntactic encoding of focus is primarily conditioned by whether syntax generates multiple outputs (a similar point is argued for Italian subject-verb inversion in Adger 1994). As a consequence, syntax does not need to resort to informationstructure categories.

A second consequence of the proposal made in this paper is that the mapping between syntax and information-structure categories may be relaxed. Since the encoding of information-structure is not directly represented in syntax, we expect there not to be a categorical mapping between syntactic constituents or positions and information-structure categories. W e will provide further arguments in favour of this claim in this section. This discussion will permit accounting for the fact that inversion in unaccusative contexts is not available for subject-focus contexts. Let us consider the inversion constructions discussed so far. As is well known, the subjects in inverted constructions typically convey new information, and are indefinite. However, as shown in (l)-(3), inversion is also possible with other types of DPs, such as proper names. If one follows Diesing (1992), the tendency for indefinites to appear in inversion constructions may be understood as a consequence of the Mapping Hypothesis. According to the Mapping Hypothesis, indefinite material is supposed to stay VP-internally, while definite DPs and strong quantifiers must occur in the IP-domain. However, as it is formulated, the mapping hypothesis is categorical: Definite DPs and strong quantifiers are not expected to surface VP-internally. As just mentioned, this prediction is not borne out, since we find instances of proper names and definite DPs VP-internally. In order to account for the complexity of the data involving information-status, quantificational force and definiteness effects, we propose that at least three scales must be involved. The relevant scales are the following: 2 a) b) c)

Givenness Definiteness Quantification

98

Joäo Costa and Maria Cristina Figueiredo Silva

According to the hypothesis put forward in section 3, the definition of these scales is not a job for syntactic theory. Consider, for instances, the scale of givenness defined in Berlinck (1995), based on Prince (1981), for explaining some corpus data concerning the occurrence of subjects in SY, VSX and VXS order in Brazilian Portuguese: Situationgiven > Text-given > Inferable > Available > partially new > New. Berlinck (1995) shows that this type of distinctions is relevant for a proper classification of the notion of givenness. However, such subtle differences in information status do not correspond to syntactic categories. The same can be said for the other two scales. Under the multi-scalar view advocated here, two predictions are made. First, it is predicted that there are mismatches. For instances, a definite DP conveying new information may surface post-verbally, in compliance with the givenness scale. Second, this type of non-categorical mapping frees syntactic theory from having to worry about precise information-structure related categories, that are best explained within pragmatics. All that is expected is that there are some tendencies for a correlation between word order and discourse-functions (indirectly explainable at the syntax-semantics and syntax-prosody interfaces), but crucially, it is not necessary to resort to covert syntactic operations just to ensure that categorical mappings make the right predictions. Bearing this hypothesis in mind, we can provide an explanation for the status of unaccusative inversion in European Portuguese and Brazilian Portuguese. In Coelho et al. (2001) and Costa et al. (2002), it is argued that inversion with unaccusative verbs in Brazilian Portuguese is an instance of locative inversion. This proposal is based on the following observations. First, there is no true optionality. It is not the case that in sentence-focus contexts, SV and VS alternate freely, as shown in the examples in (21 )-(23): (21) Ο que aconteceu? What happened a. Caiu um aviäo. Fell an airplane b. HUm aviäo caiu. An airplane fell (22) Ο que aconteceu? What happened

On the relation between syntax and pragmatics

99

a. Nasceram 93 bebes. Were born 93 babies b. ΊΊ93 bebes nasceram. 93 babies were born (23) Ο que aconteceul What happened a. Urn frigorifico descongelou. A freezer thawed b. UDescongelou um frigorifico. In fact, both definiteness effects and subtypes of unaccusative predicates seem to favour or disfavour the VS order in this context. The second argument for analyzing VS as locative inversion comes from the comparison with English. It was noted above that, just like in English, VS in Brazilian Portuguese does not provide information-focus on the subject. The same typically holds for English locative inversion and expletive there-constructions (cf. Ward and Birner 1995): (24) Brazilian Portuguese: Quern morreul Who died a. UMorreu ο PM. Died the PM b. Ο ΡΜ morreu. The PM died (25)

English: Who's coming? a. A man comes. b. #There's coming a man.

Based on these two observations, we hypothesize that unaccusative VS sentences are locative inversions in Brazilian Portuguese (cf. Pinto 1997, Ambar 1999, Cornish 2002 for other languages). Assuming with Levin and Rappaport Hovav (1995) that unaccusatives entering locative inversion constructions are associated to (potentially unrealized) locative or temporal argument (cf. Pustejovsky 1995), and that expletive pro may correspond to the temporal / locative argument (Pinto 1997), the Brazil-

100

Joäo Costa and Maria Cristina Figueiredo Silva

ian Portuguese facts follow straightforwardly. Since there is expletive pro in Brazilian Portuguese, we expect this language to instantiate VS only in the context in which locative inversion is adequate: the context of presentational focus in which locative inversions are uttered. This analysis makes some further predictions. First, it is expected that not all unaccusatives invert, since according to Levin and Rappaport Hovav (1995), some unaccusative predicates are not associated to locative or temporal arguments. This would be the case for the predicate descongelar 'thaw', as illustrated in (23). Second, it is predicted that a few verbs other than unaccusatives, selecting the same kind of argument, invert in Brazilian Portuguese. This would be the case for the verb telefonar 'to call', as also argued in Pinto (1997), and Cornish (2002). (26) Ο que e que aconteceu? What happened Telefonou a Maria. Called Maria Finally, since information focus (serving to convey new information) and presentational focus (used for introducing background context) are not identical, it is correctly predicted that they may be linked to different types of syntactic constructions. In fact, English provides independent evidence for distinguishing the two types of VS order in Portuguese. This discussion sheds light on the status of categorical mappings examined in this section. We observe that, in order to account for the status of inversion with unaccusatives in Brazilian Portuguese, it is necessary to acknowledge a distinction between information focus, sentence focus and presentational focus. Such subtle differences in focus-types are not a syntactic matter, although the different focus-types may be linked to different syntactic constructions. The only syntactic aspect that must be taken into account is that Brazilian and European Portuguese do not differ as far as the codification of presentational focus is concerned, because expletive pro is available in both languages. Moreover, it is predicted that categories entering into locative inversion constructions are of different information types and not necessarily indefinite, which strengthens the idea that mappings between syntax and information-structure related notions are to be understood as tendencies.

On the relation between syntax and pragmatics

101

6. Conclusions The discussion of micro-variation facts comparing European Portuguese and Brazilian Portuguese disfavor a parametric approach for defining the relations between syntax and discourse, since such a split would not account for similarities and differences between the two languages as far as the relation between certain syntactic constructions and informationstructure categories is concerned. We hope to have shown that the micro-variation facts studied disfavor reference in syntax to information-structure or pragmatic labels, since the mappings between syntax and such type of information is not categorical. Finally, we have argued that the choice of syntactic outputs for codifying different types of information structures is made post-syntactically, which implies that there are no syntactic operations to serve discourse purposes. Naturally, this claim holds for the cases studied in this paper, and not to all constructions previously studied under hypotheses incorporating discourse notions in syntax. Nevertheless, we hope to contribute to the debate on the relation between syntax and information structure, by casting doubt on some of the constructions that have been previously analyzed as involving direct access to discourse by syntax. It may be the case that future studies on other construction lead to similar conclusions.

Notes 1.

2.

3.

As pointed out by an anonymous reviewer, the notion of contrast is orthogonal to the given/new distinction. We mention that topicalization affects contrastive topics, just to distinguish them from other types of topics. This is partly independent of the type of prosodic strategy for assigning stress. As we will argue, sentence neutral stress and marked pitch accents do not necessarily convey different types of focus. We are assuming with De Hoop (1991) and Ariel (1990), among others, that definiteness and quantification are not strictly binary notions.

102

Joäo Costa and Maria Cristina Figueiredo Silva

References Adger, David 1994 Functional heads and interpretation. CCS, Edinburgh: Ph. D. diss. Ambar, Manuela 1999 Aspects of Focus in Portuguese. In The Grammar of Focus, L. Tuller, and G. Rebuschi (eds.), 23-53. Amsterdam: John Benjamins. Ariel, Mira 1990 Accessing Noun-Phrase Antecedents. London: Routledge. Belletti, Adriana 2001 'Inversion' as focalization. In Inversion in Romance and the Theory of Universal Grammar, A. Hulk, and J. Y. Pollock (eds.). Oxford: Oxford University Press. Berlinck, Rosane 1995 La Position du sujet en Portugals. Leuven, Katholieke Univ./UNICAMP: Ph. D. diss. Cinque, Guglielmo 1993 A Null Theory of Phrase and Compound Stress. Linguistic Inquiry 24: 239-298. Castro, Ana, and Joäo Costa 2002 Weak forms as X°: Prenominal possessives and pre verbal adverbs in European Portuguese. In Romance Linguistics: Theory and Acquisition, A. T. Perez-Leroux, and Y. Roberge (eds.), 95-110. Amsterdam/Philadelphia: John Benjamins. Coelho, Izete, J. Costa, and M. C. Silva 2001 Ordern VS e sujeito nulo em PE e PB. Anais do Encontro Internacional da ABRALIN. Brazil: Fortaleza. Cornish, Francis 2002 Locative inversion in 8 languages: Syntax, semantics, discoursepragmatics, and functional position. Paper presented at 10llt International Conference on Functional Grammar, University of Amsterdam. Costa, Joäo 1998 Word Order Variation: A Constraint-Based Approach. Ph. D. diss. HIL/Leiden University: Holland Academic Graphics. 2002 When Can Objects Bind? Left-to-Right Merge, Scrambling and Binary Structure in European Portuguese. Paper presented at the Incontro di Grammatica Generativa, Lecce [To appear in the Proceedings], 2003 Subjects in Spec,vP: Locality and Agree. Ms., Universidade Nova de Lisboa.

On the relation between syntax and pragmatics

103

Costa, Joäo, Izete Coelho, Maria Christina Figueiredo Silva, and F. de Oliveira 2002 Consideragöcs sobre a Ordern VS e Sujeito Nulo em Portugues Europeu e Portugues Brasileiro. Paper presented at the PEPB Third Colloquium, University of Lisbon, Lisbon. Costa, Joäo, and Ines Duarte 2001 Objectos nulos em debate. In Razöes e emogäo. Homenagem a Maria Helena Mateus, I. Castro, and I. Duarte (eds.) [in press], Cyrino, Sonia, and Gabriela Matos 2002 Syntactic microvariation in VP ellipsis in European and Brazilian Portuguese. Paper presented at 25th GLOW Colloquium, Amsterdam. Diesing, Molly 1992 Indefinites. Cambridge, Massachusetts: MIT Press. Duarte, Ines 1987 Α constru^äo de topicaliza9äo em portugues europeu. Universidade de Lisboa: Ph. D. diss. Figueiredo Silva, Maria Cristina 1996 Α Posigäo do Sujeito em Portugues Brasileiro - Frases Finitas e Infinitivas. Brazil, Campinas: Editora da Unicamp. Givon, Talmy 1990 Syntax. A Functional Typological Introduction. Vol. 2 Amsterdam: John Benjamins. Hoop, Helen de 1992 Case Configuration and Noun Phrase Interpretation, University of Groningen: Ph. D. diss. Horvath, Julia 1986 Focus in the Theory of Grammar and the Syntax of Hungarian. Dordrecht: Foris. Kiss, Katalin (ed.) 1995 Discourse Configurational Languages (Oxford Studies in Comparative Syntax). Oxford: Oxford University Press. Levin, B., and M. Rappaport Hovav 1995 Unaccusatives. Cambridge, Massachusetts: MIT Press. Matos, Gabriela 1992 Constru9Öes de elipse do predicado em portugues. Univ. de Lisboa: Ph. D. diss. Nespor, Marina, and Irene Vogel 1986 Prosodic Phonology. Dordrecht: Foris. Ordonez, Francisco 1997 Word order and clause structure in Spanish and other Romance Languages. New York, CUNY: Ph. D. diss.

104

Joäo Costa and Maria Cristina Figueiredo Silva

Pinto, Manuela 1997 Licensing and Interpretation of Inverted Subjects in Italian. Utrecht University: Ph. D. diss. Prince, Ellen 1981 Toward a taxonomy of given-new information. In Radical Pragmatics, Cole, P. (ed.), 223-256. New York: Academic Press. Pustejovsky, James 1995 The Generative Lexicon. Cambridge, Massachusetts: MIT Press. Raposo, Eduardo 1986 On the null object in European Portuguese. In Studies in Romance Linguistics, Osvaldo Jaeggli, and Carmen Silva-Corvalan (eds.), 373390. Dordrecht: Foris. Raposo, Eduardo, and Juan Uriagereka 1990 Long-distance case assignment. Linguistic Inquiry 21: 505-537. Reinhart, Tanya 1995 Interface Strategies. OTS Working Papers in Linguistics. Rizzi, Luigi 1982 Issues in Italian Syntax. Dordrecht: Foris. 1997 On the fine structure of the left-periphery. In Elements of Grammar, Haegeman L. (ed.), 281-337. Dordrecht: Kluwer Academic Publishers. Salles, Heloisa 1997 Prepositions and the syntax of complementation. University of Wales, Bangor: Ph. D. diss. Ward, Gregory, and Birner, Betty 1995 Definiteness in the English existential. Language 71: 722-742. Zubizarreta, Maria Luisa 1998 Focus, Prosody, Word Order. Cambridge, Massachusetts: MIT Press.

Association with focus phrases* Manfred Krifka

Abstract The paper discusses whether association with focus is a structural operation involving L F movement, or an interpretational operation requiring the simpler mechanism of alternative semantics. It argues for a mixed approach: Association with focus phrases is via L F movement, association of focus within focus phrases is via projection of alternatives. Three arguments are presented that argue for this type of theory, whereas three other arguments are shown to be inconclusive.

1. The nature of association with focus Certain semantic operators, including particles like only, also and even, contribute to the meaning of the sentence in ways that depend on the position of the focal accent in the sentence. This well-known phenomenon, called association with focus, is evident with examples like (la,b). (1)

a. John 'The b. John 'The

[yp only [yp introduced ΒίΙΙψ to Sue]]]. only person John introduced to Sue is Bill.' [yp only [yp introduced Bill to Suep]]. only person John introduced Bill to is Sue.'

According to the central semantic principle of compositional interpretation, the meaning of a complex constituent is a function of its immediate syntactic subconstituents. Hence in (la,b) the meaning of the larger VP starting with only should depend on the meaning of the semantic operator only and the meaning of the smaller YP c - c o m m a n d e d by only, its scope. As the meaning of the complex VPs are different, the meanings of introduced Bill to SueF and introduced Billf to Sue must be different.

106

Manfred Krifka

A number of proposals have been developed that allow for that kind of meaning sensitivity to prosody. Here I will discuss two prominent accounts, Structured Meanings (SM) and Alternative Semantics (AS).

1.1. Structured Meanings / syntactic movement The SM approach, as proposed by Jacobs (1983), von Stechow (1990) and others assumes that focus-sensitive operators like only have access to both the expression in focus and the position within the scope in which the focus is located. In a non-representational semantic theory, which does not allow for syntactic manipulations of expressions of a semantic representation language, the meaning of only must be able to access the meaning of the focus F, the alternatives of the focus A, and a function Β that maps the meaning of the focus to the meaning of the scope, commonly called the background. This is illustrated in the following example: (2)

[vp introduced ΒίΙΙψ to Sue] Focus-background structure: (BILL, Α, λΧ[ΙΝΤΚ00(8υΕ)(Χ)]> Background applied to focus: INTROD(SUE)(BILL)

The set of alternatives must contain the focus, here: BILL e A, and it must contain at least one additional element. Typically, the set of alternatives is restricted by context. Focus-sensitive operators take such background-focus structures and convert them into standard meanings, illustrated here with only: (3)

ONLY«F, Α, Β » = λχΥΥΕ A[B(Y)(x)

F = Y]]

For our example we get the following result; it says that for every alternative y to Bill it holds that if John introduced y to Sue, then y is Bill. Put differently, John didn't introduce any alternative besides Bill to Sue. (4)

ONLY«BILL, A, Xx[INTROD(SUE)(x)]))(JOHN) = V y e A[INTROD(SUE)(y)(JOHN) y = BILL]

The SM theory faces the challenge of how the focus-sensitive particle is informed about the meaning of the focus item and the nature of its

Association with focus phrases

107

contribution to the scope. One prominent answer, proposed by Chomsky (1976), is syntactic movement (provides the acronym SM with another reading). In the case at hand, this is covert movement on logical form: (5)

SS: [yp introduced Bill? to Sue] LF: [Bill;[introduced t; to Sue]] Interpretation: (BILL, A, Xxi[INTROD(SUE)(xi)]}

It is well-known that LF movement as a syntactic operation can be dispensed with in various ways, e.g. operator storage, cf. Cooper (1983), projection of subcategorization features in generalized phrase-structure grammars, cf. Gazdar et al. (1985), or flexible categorial grammar, cf. Hendriks (1993). Such alternatives to LF movement can also be used for the construction of structured meanings; see e.g. Krifka (1992) for a theory couched in a categorial grammar framework. What appears more important than differences in the specific implementations of the SM approach is the implicit claim made by all of them: Namely, that providing the right meanings for focus-sensitive operators is a process comparable to the other task that LF movement has been designed for. The main reason for assuming LF movement is the description of wide scope taking of quantifiers. As an example, consider (7). (6)

A professor [introduced every honors student to the president]. LF for wide-scope reading of every honors student: every h. student \[a professor [introduced t; to the president]]

Here, every honors student is moved and adjoined to the clause, leaving an indexed trace t;. The constituent the quantifier is adjoined to is marked by the index of this trace to indicate that an expression with this index has been extracted. In semantic interpretation, the indexed constituent will be interpreted as a functional expression (cf. Heim and Kratzer (1998) for this view of LF movement). Scope taking of quantifiers and informing an operator of the focus and its position seem to require similar manipulations or enrichments of syntactic structure or semantic interpretation: The quantifier or the focus expression have to be identified, and they have to be related to a position within the scope. So it seems natural to posit the same mechanism for scope taking and for association with focus.

108

Manfred Krifka

Now, it is known since Anderson (1972) and Jackendoff (1972) that scope taking and association with focus appear to have distinct properties: Whereas scope taking of quantifiers is restricted by syntactic island constraints, cf. (9), association with focus appears to be free of it, cf. (11): (7)

A professor introduced [the person that every honors student admires most] to the president. Impossible: 'For every honors student there is a professor that introduced the person that the honors student admires most to the president.'

(8)

John only introduced [the man that JÜIY admires most] to Sue. Possible: 'The only person such that John introduced the man that this person admires most to Sue is Jill.'

I will argue that the problem this raises for the SM account is only apparent. But let us first consider the other prominent approach to the semantics of focus, alternative semantics.

1.2. Alternative Semantics In the AS approach, as developed by Rooth (1985) and Rooth (1992), neither the expression in focus nor its contribution to the meaning of the scope can be accessed directly. Association with focus is achieved in a considerably simpler way than scope taking of quantifiers. What is required is the ordinary meaning of the scope of the focus-sensitive operator and the alternatives to the ordinary meaning introduced by the expression in focus. These alternatives can be compositionally derived by projecting the alternatives of the expression in focus, by the wellknown mechanism first proposed by Hamblin (1973) for w/j-questions. Let [[a]] stand for the ordinary meaning of α and [[a]] A for the set of alternatives to this meaning, then this set can be derived compositionally in the following way: (9)

If [[[α ß]]] = f([[a]], [[β]]), then [[[a ß]]]A = {f(X, Y) | X

e

[ [ a ]]A Ye [Iß]] A}

Association with focus phrases

109

The set of alternative meanings of a simple non-focused expression α is the singleton set of its ordinary meaning {[[a]]}, and the set of alternative meanings of a focused expression α ρ is some non-singleton set ALT([[a]]). We get the following two-level interpretation for our example: (10) a. [[[vp introduced ΒίΙΙγ to Sue]]\ = Xx[INTROD(SUE)(BILL)(x)] b. [[[vp introduced ΒίΙΙγ; to S«e]]] A = {Xx[INTROD(SUE)(y)(x)] I ye ALT(BILL)} The meaning of only can be rendered as follows: (11)

[[ [only [ ν Ρ α ] ] ]] = λχ[[[α]](χ) Λ VYe [[α]] Α [Υ(χ) -> Υ=[[α]]]]

For our example, we get the following interpretation: (12) [[ John [vp only [yp introduced ΒίΙΙψ to Swe]]]] = INTROD(SUE)(BILL)(JOHN) A VYe {INTROD(SUE)(y) I y e ALT(BILL)} [Y(JOHN) —> Y = INTROD(SUE)(BILL)]] That is, it is claimed that John introduced Bill to Sue, and for all properties Y of the type 'introduce χ to Sue', where χ is an alternative to Bill, if John has property Y, then Y is the property 'introduce Bill to Sue'. This approach avoids direct reference to the meaning of the expression in focus; it just makes use of the effects that focus has on the larger expression. For this to work, it is crucial to assume an intensional representation of meanings. To see this, consider a situation in which John introduced Bill and Jill to Sue, and no-one else introduced them to Sue. Certainly, John only introduced ΒίΙΙψ to Sue is false in this situation, but (16), interpreted extensionally, is true, as we have INTROD(SUE)(BILL) = INTROD(SUE)(JILL) = {JOHN}. The intensions of these meanings are different; we have Xi[INTROD(i)(SUE)(BILL)] Φ λ ϊ [ Ι ^ Έ Ο ϋ ( ί ) ( 8 υ Ε ) (JILL)], even in the indicated situation, as Bill and Jill could have been introduced by different persons to Sue. 1 The proper representation we should work with would assign the following reading to sentence (12), applied to the world i* of interpretation:

110

Manfred Krifka

(13)

INTROD(i*)(SUE)(BILL)(JOHN) λ V P e { λ ί [ Ι Ν Τ Κ 0 0 ( ί ) ( 8 υ Ε ) ( γ ) ] I y e ALT(BILL)} [P(i*)(JOHN) - > P= λί[INTROD(i)(SUE)(BILL)]] ]

This says that John introduced Bill to Sue in world i*, and for every property Ρ of the properties of the type 'introduce y to Sue', where y is an alternative to Bill, the only property Ρ that applies at i* to John is the property 'introduce Bill to Sue'.

1.3. Structured Meanings or Alternative Semantics? Alternative semantics offers a more parsimonious approach to focus sensitivity than the structured meaning approach. For one thing, we can derive the focus representation of alternative semantics from structured meanings, but not vice versa. Let (F, A, B) be a structured meaning representation; then the corresponding representation in alternative semantics is B(F) for the ordinary meaning, and {B(X) I X e A} for the alternatives. But there is no way to reconstruct a structured meaning representation out of a meaning m and its alternatives A. Also, notice that the AS does not allow for an operation like binding of variables, in contrast to the SM approach. What AS provides for is something like a semantics for wildcards. In an expression of the type [... * ... * ...], the two occurrences of the wildcard * can be filled independently of each other. In the SM approach, on the other hand, we can distinguish the case in which both positions vary independently, [... χ ... y ...], from the case in which they vary in unison, [... χ ... χ ...]. General considerations of parsimony should favor the AS approach, as the null hypothesis, over the SM approach. In addition, the apparent lack of island sensitivity constitutes a specific argument against the SM approach. However, AS has to face several problems that may tell us that this format is not expressive enough to model focus in natural language. One, which was observed already by Rooth (1985), §2, footnote 13, and is discussed in Blok (1993), is that alternative semantics does not always give us the right interpretation even if couched in an intensional framework. Rooth mentions an example similar to the following one: (14)

Nine only is the square of threes.

Association with focus phrases

111

This sentence should be false if the domain includes negative numbers, as nine is also the square of minus three. But it has a true interpretation under the intensionalized version of (15), as the meanings of the square of three and the square of minus three are the same in all possible worlds - they all have to obey the laws of mathematics. The SM approach has no problem here, as the meanings of three and minus three are different. Another problem of AS, pointed out in Krifka (1992), arises when we consider multiple focus, such as in the following case: (15)

John only introduced Billp to Sue. John also2 only ι [introduced ΒίΙΙψχ to Mdry p2].

The co-indexation expresses the natural reading of the second sentence, saying 'John introduced only Bill to Mary, and there is another person besides Mary such that John introduced only Bill to that person'. Pure AS cannot capture this interpretation, as it cannot identify distinct variables. In AS, the first operator only would associate with both foci, on Bill and Mary, leaving no focus to associate with also. Kratzer (1994) points out another problem that results from the impossibility of co-indexation. (16) A:

B:

What a copycat you are! You went to Block Island because I did, and you went to Tanglewood because I did. No, I only went to Tdnglewoodp because you did

According to standard assumptions about VP ellipsis, B's reply should be spelled out as follows: (17) I only [went to Tänglewoodρ because you went to

Tdnglewoodp].

The only way for pure AS to handle these two foci is that the alternatives, Block Island and Elk Lake Lodge, are introduced in each of the two focus positions independently. But this does not give us the right interpretation. The interpretation AS provides for (17) can be paraphrased as: Of the propositions (a) (b)

Ί went to Block Island because you went to Block Island' Ί went to Block Island because you went to Tanglewood'

112

Manfred Krifka

(c) (d)

Ί went to Tanglewood because you went to Block Island' Ί went to Tanglewood because you went to Tanglewood',

only proposition (d) is true. However, (16) is naturally understood in a way that only propositions (a) and (d) are alternatives. This reading can be achieved in the SM approach if only associated with both foci in (17). (18) I only\ [went to Tdnglewoodpt\ woodp y i]

because

you went to

Tdngle-

A related problem was pointed out by von Stechow (1990), who attributes it to Thomas Ede Zimmermann. It concerns the inability to express any relation between focus values in AS, not just identity. Assume that John gives a formal dinner party. As such parties go, he is obliged to introduce every person to his or her designated partner at table. Assume that the party just started; we have four guests, namely Bill and Sue (who are designated partners) and Charles and Lucy (who are designated partners). John has already introduced Bill to Sue, and Bill to Lucy. At this point, question (19A) can be truthfully answered by (19B): (19) A: B:

Did John introduce every gentleman to his partner at table? No, John only introduced Btllp to Süe-p.

This says: (Bill, Sue) is the only pair of partners-at-table such that John introduced the first to the second. John's introducing Bill to Lucy does not count, as (Bill, Lucy) is not a pair of partners-at-table, and hence not an alternative. Pure AS cannot express this restriction on paired foci, as it introduces the alternatives in each focus position separately. The SM approach, on the other hand, can identify Bill and Sue and combine them to a pair, a process that has been called absorption (cf. Higginbotham and May 1981 for similar cases with multiple questions). These problems with AS suggest that the simple representation framework that it offers for association with focus is too simple. But then we still face the problem of the apparent lack of island sensitivity of association with focus. In the following, we will discuss this in greater detail.

Association with focus phrases

113

2. A hybrid theory of association with focus

2.1. Association with focus phrases I will explore the idea that association with focus is, in fact, sensitive to syntactic islands, following Drubig (1994). An apparent counterexample like (11), repeated here as (25a), resulting in an LF that violates island restrictions under the usual theory, is interpreted as involving the LF (25b), in which the operator associates with the syntactic island that contains the focus. The indexed constituent will be called focus phrase (FP). (20) a. only introduced [the man that Jillρ admires]^ b. only [[the man that Jflip admires] \[introduced

to Sue t\ to Sue] ]

For another example of this type, consider the following: (21) a. only liked [the man that introduced ΒίΙΙψ to Sue]ρ b. only [[the man that introduced Billp to 5iie]pp \ [liked 11]|] Association with the syntactic island containing the focus, a case of piedpiping on LF, makes the focus visible to the focus-sensitive operator without violating syntactic island restrictions. But this can only be part of the story: We still have to explain the contribution of the focus within the focus phrase. For example, the different readings of expressions like the following ones have to be accounted for: (22) a. only liked [the man that introduced ΒίΙΙγ to Sue]pp b. only liked [the man that introduced Bill to Siiep]pp The truth-conditional content of these expressions is different: In the first case, the alternatives are men that introduced someone to Sue; in the second, the alternatives are men that introduced Bill to someone. The assignment of the focus within the focus phrase does matter, even if the focus operator does not associate with it directly.

114

Manfred Krifka

2.2. The role of focus in the focus phrase I would like to propose that the focus within the focus phrase determines the alternatives of the focus phrase, in the manner indicated in Alternative Semantics (see Krifka 1996 for a first presentation of this idea). Then the alternative sets of the two focus phrases in (28a,b) are distinct: In the first case, the alternatives are men that introduced someone to Sue; in the second, the alternatives are men that introduced Bill to someone. (23)

a. [[[the man that introduced ΒίΙΙψ to S«e]pp]] A = (lx[MAN(x) Λ INTROD(SUE)(y)(x)] I y e ALT(BILL)} b. [[[the man that introduced Bill to 5«ep]pp]] A = (lx[MAN(x) Λ INTROD(y)(BILL)(x)] I y e ALT(SUE)}

Combining this with the interpretation of only in the SM approach, we get the following interpretations: (24)

[[John only [[the man that introd. ΒίΙΙψ to Sue] ρ ρ {[liked ti]]]] = V x e [[the man that introd. ΒίΙΙψ to Sue]]A [LIKED(x)(JOHN) —> χ = [[the man that introd. ΒίΙΙψ to Sw