From Words to Discourse: Trends in Spanish Semantics and Pragmatics
Current Research in the Semantics/Pragmatics Interfa Series Editors: K.M. Jaszczolt, University of Cambridge, UK K. Turner, University of Brighton, UK
Related Elsevier books Other titles in this Series: TURNER (ed.) JASZCZOLT GEURTS JASZCZOLT (ed.) PEETERS (ed.) PAPAFRAGOU LEEZENBERG NEMETH & BIBOK (eds.) BRAS & VIEU (eds.)
The Semantics/Pragmatics Interface from Different Points of View Discourse, Beliefs and Intentions Presuppositions and Pronouns The Pragmatics of Prepositional Attitude Reports The Lexicon-Encyclopedia Interface Modality: Issues in the Semantics/Pragmatics Interl Contexts of Metaphor Pragmatics and the Flexibility of Word Meaning Semantic and Pragmatic Issues in Discourse and Dialogue
Related Elsevier journals Journal of Pragmatics Editor. Jacob Mey Language and Communication Editors: Roy Harris and Talbot J. Taylor Language Sciences Editor. Nigel Love Lingua Editors: Johan Rooryck, Neil Smith and Diane Blakemore On-line journal sample copies available at http://www.elsevier.com/locate/linguistics
For further information on the CRiSPI series and for details of how to submit a proposal go to: http://www.elsevier.com/locate/series/crispi
FROM WORDS TO DISCOURSE: TRENDS IN SPANISH SEMANTICS AND PRAGMATICS Edited by Javier Gutierrez-Rexach The Ohio State University, USA
2002 ELSEVIER
Amsterdam - Boston - London - New York - Oxford - Paris San Diego - San Francisco - Singapore - Sydney - Tokyo
ELSEVIER SCIENCE Ltd The Boulevard, Langford Lane Kidlington, Oxford OX5 1GB, UK
© 2002 Elsevier Science Ltd. All rights reserved.
This work is protected under copyright by Elsevier Science, and the following terms and conditions apply to its use: Photocopying Single photocopies of single chapters may be made for personal use as allowed by national copyright laws. Permission of the Publisher and payment of a fee is required for all other photocopying, including multiple or systematic copying, copying for advertising or promotional purposes, resale, and all forms of document delivery. Special rates are available for educational institutions that wish to make photocopies for non-profit educational classroom use.
In the USA, users may clear permissions and make payments through the Copyright Clearance Center, Inc., 222 Rosewood Drive, Danvers, MA 01923, USA; phone: (+1) (978) 7508400, fax: (+1) (978) 7504744, and in the UK through the Copyright Licensing Agency Rapid Clearance Service (CLARCS), 90 Tottenham Court Road, London W1P OLP, UK; phone: (+44) 207 631 5555; fax: (+44) 207 631 5500. Other countries may have a local reprographic rights agency for payments. Derivative Works Tables of contents may be reproduced for internal circulation, but permission of Elsevier Science is required for external resale or distribution of such material. Permission of the Publisher is required for all other derivative works, including compilations and translations. Electronic Storage or Usage Permission of the Publisher is required to store or use electronically any material contained in this work, including any chapter or part of a chapter. Except as outlined above, no part of this work may be reproduced, stored in a retrieval system or transmitted in any form or by any means, electronic, mechanical, photocopying, recording or otherwise, without prior written permission of the Publisher. Address permissions requests to: Elsevier Science Global Rights Department, at the mail, fax and e-mail addresses noted above. Notice No responsibility is assumed by the Publisher for any injury and/or damage to persons or property as a matter of products liability, negligence or otherwise, or from any use or operation of any methods, products, instructions or ideas contained in the material herein. Because of rapid advances in the medical sciences, in particular, independent verification of diagnoses and drug dosages should be made.
First edition 2002 Library of Congress Cataloging in Publication Data A catalog record from the Library of Congress has been applied for. British Library Cataloguing in Publication Data A catalogue record from the British Library has been applied for.
ISBN: 0-08-043991-8 Series ISSN: 1472-7870 ® The paper used in this publication meets the requirements of ANSI/NISO Z39.48-1992 (Permanence of Paper). Printed in The Netherlands.
Current Research in the Semantics/Pragmatics Interface (CRiSPI)
Series Editors: K.M. Jaszczolt, University of Cambridge, UK and K. Turner, University of Brighton, UK Editoral Advisory Board: N. Asher, USA J. van der Auwera, USA C. Casadio, Italy M. Dascal, Israel B. Fraser, USA T. Fretheim, Norway B. Gillon, Canada P. Gochet, Belgium J. Groenendijk, The Netherlands YueguoGu, PRC K. von Heusinger, Germany A. Kasher, Israel M. Krifka, Germany Susumu Kubo, Japan Chungmin Lee, Korea S. Levinson, The Netherlands T. McEnery, UK F. Nemo, France P. Pelyvas, Hungary J. Peregrin, Czech Republic A. Ramsay, UK R. van der Sandt, The Netherlands R. Stalnaker, USA M. Stokhof, The Netherlands G. Ward, USA
The aim of this series is to focus upon the relationship between semantic and pragmatic theories for a variety of natural language constructions. The boundary between semantics and pragmatics can be drawn in many various ways and the relative benefits of each have given rise to a vivid theoretical dispute in the literature in the last two decades. As a sideeffect, this variety has given rise to a certain amount of confusion and lack of purpose in the extant publications on this topic. This series will provide a forum where the confusion within existing literature can be removed and the issues raised by different positions can be discussed with a renewed sense of purpose. The editors intend contributions to this series to take further strides towards clarity and cautious consensus.
This page intentionally left blank
TABLE OF CONTENTS
List of Contributors
ix
1. Introduction Javier Gutierrez-Rexach 2. Spanish Past and Future Tenses: Less (Semantics) is More Silvia Gennari 3. Tensed Complements of Perception Verbs: Issues in their Temporal Interpretation
1 21 37
Alicia Cipria 4. Spanish 'Aspectual' Periphrases: Ordering Constraints and the Distinction Between Situation and Viewpoint Aspect Brenda Laca 5. Non-Declarative Sentences in Spanish: The Case of the Infinitive
61 95
Xose Rosales Sequeiros 6. Aspect and Situations: A Situation Semantics Account of the Semantic Variability of Spanish'/^/-Clauses'
119
Luis Alonso-Ovalle I. Mental Spaces and Epistemic Attitudes: On the Spanish Subjunctive/Indicative Alternation
135
Francisco Aliaga and Eduardo de Bustos 8. Space Accessibility and the Pragmatic Status of Propositions
145
Errapel Mejias-Bikandi 9. Coercion and the Stage/Individual Distinction Victoria Escandell- Vidal and Manuel Leonetti 10. Some Spanish Quantifier Modifiers Richard Zuber I1. Demonstratives in Context Javier Gutierrez-Rexach
181
12. Prepositional NPIs and the Scalar Nature of Polarity
237
Javier Gutierrez-Rexach and Scott Schwenter 13. Degree Quantification and Modal Operators in Spanish
263
Ignacio Basque
159
195
viii
From Words to Discourse
14. The Shifted Reading of the Spanish Simple Past as Dependent on Presupposition Accommodation Rosa J. Garcia Cordoba 15. On the Interaction of Syntax-Semantics-Pragmatics: A Case Study Josep Macia 16. Clitic Doubling, Null Objects and Clitic Climbing in the Spanish of Corrientes Laura Colantoni 17. Spanish no, s/: A Particle of Politeness Robert Vann, Joan Busquets and Dale Koike Name Index Subject Index
289 307 321 337
351 355
CONTRIBUTORS Francisco Aliaga, Universidad Complutense de Madrid Luis Alonso-Ovalle, University of Massachusetts at Amherst Ignacio Bosque, Universidad Complutense de Madrid Joan Busquets, Universite de Bordeaux Eduardo de Bustos, Universidad National de Education a Distancia, Madrid Alicia Cipria, University of Alabama Laura Colantoni, University of Minnesota/CONICET, Argentina Victoria Escandell-Vidal, Universidad National de Education a Distancia, Madrid Silvia Gennari, Brown University Rosa J. Garcia Cordoba, Universidad Autonoma Metropolitana, Mexico Javier Gutierrez-Rexach, The Ohio State University Dale Koike, University of Texas at Austin Brenda Laca, Universite de Paris VIII Manuel Leonetti, Universidad de Alcald, Madrid Josep Macia, Universitat de Barcelona Errapel Mejias-Bikandi, University of Nebraska, Lincoln Xose Resales Sequeiros, University of Greenwich Scott Schwenter, The Ohio State University Robert Vann, Western Michigan University Richard Zuber, Centre National de la Recherche Scientifique, Paris
This page intentionally left blank
1
INTRODUCTION Javier Gutierrez-Rexach, The Ohio State University
1 MEANING AND ITS INTERFACES An examination of the development of linguistic thought in the last fifty years shows that some of the most fruitful linguistic theories and research programs have emerged from crossfertilization processes with other disciplines. Theories such as generative grammar, generalized quantifier theory, dynamic semantics, or the theory of speech acts immediately come to mind as paradigmatic examples of developments in neighboring disciplines (computer science, logic or philosophy) that have had a lasting impact in linguistics. Additionally, linguistic theories may have a broader influence in other realms of inquiry such as the relationship between language and cognition, language an intentionality, or language and communication processes. Nevertheless, the success of a linguistic theory is commonly measured not only by formal criteria such as elegance, simplicity or internal coherence but most importantly by its ability to explain empirical data and to bring to light new correspondences and patterns of data. This back and forth movement between theory and data, or theoretical and empirical concerns, is central to the discipline in more than one respect. The study of language as a faculty or an abstract property is mirrored by the overwhelming reality of linguistic diversity. On the one hand, the consideration of different languages, by themselves or from a comparative viewpoint, has to enrich or probably even make more complex achieving valid generalizations. On the other hand, language intermingles with and is influenced by other aspects of human behavior, including social action and the dynamics of conceptual processes. If we accept this broader and challenging perspective, there is an implicit consensus in the discipline that an area of linguistic research can be considered mature when the validity of theoretical and empirical results is tested
2
From Words to Discourse
cross-linguistically and when predictions from different languages influence and modify the course of theoretical developments. The semantics/pragmatics interface poses a special challenge in this respect because of the heterogeneous nature of its origins and its current interdisciplinary and multi-theoretical status. Semantics and pragmatics are indeed neighboring areas that address several aspects of the meaning of linguistic expressions and their use. From Charles Morris' initial distinction between syntax, semantics and pragmatics to the current state of the art, the field has grown more complex and interrelated and part if not all of that initial "innocence" has been lost. Even the received view from the seventies, after the development of formal semantics and the emergence of Searle's and Austin's proposals, that conceived of pragmatics as the study of meaning minus truth-conditional content (Gazdar, 1979) would probably be considered now by most researchers as an optimistic idealization. Neither semantics can currently be restricted to the study of truth conditions, nor does pragmatics pertain only tc "meaning" in the traditional sense. The development of possible worlds and intensional semantics, as well as the subtheories that took into account higher order operators such as generalized quantifiers, considerably enriched the predictive power and empirical coverage of formal semantics. During the nineties different branches or garden varieties of the trend commonly denominated "dynamic semantics" forcefully showed how the impact of context and context change, an aspect sometimes relegated to pragmatics, reached beyond varieties of use and affected central aspects of semantics such as the interpretation of indefinites and pronouns. Current developments in semantics embrace the treatment of traditional pragmatic issues such as non-declaratives, deixis, the focus-topic articulation, presupposition, metaphor, etc. In turn, pragmatics has also extended its reach considerably. From the Searle-Austin core approach to the intersection points relating meaning, use, action and intentionality, certain unifying proposals such as relevance theory emerged to provide a more comprehensive account of language cognition and information exchange. Other approaches have stressed the social components of the field, exploring more decisively the relationship between saying and doing. In this direction, "micro" sociological approaches, such as those represented by Garfinkel, Schegloff and collaborators, or even Bourdieu, explore the social logic of conversation and discourse in relation to power, gender and various social structures. All of the above seems indeed relevant to the semantics/pragmatics interface and gives an idea of the difficulty or perhaps even the undesirability of the search for a single unification theory of the field. Furthermore, research in the syntax/semantics interface has gained new momentum in recent years, especially after Chomsky's (1995) proposals about the centrality of conceptual and interface issues. The search for a syntax of language is now the search for a Logical Form that captures not only structural facts but also semantic and pragmatic aspects of meaning. What can the study of individual languages contribute to this intriguing and challenging landscape? First, it can help decide which theories have real predictive potential and can be extended or revised to properly accommodate new empirical puzzles. Second, long-standing issues in the grammar of a given language, often described with more or less accuracy in
Introduction
traditional or reference grammars, can receive enlightening accounts when subject to analysis within powerful theoretical frameworks. Finally, as pointed out above, the fine-grained analysis of linguistic data can drive theoretical processes and function as a retro-feeding device vis-a-vis scientific progress. With these issues in mind, Eduardo Bustos (UNED, Spain) and myself decided that the time seemed to be adequate for a conference that explored new problems and perspectives in the semantics and pragmatics of Spanish, from a wide and open variety of theoretical viewpoints and including contributions from scholars at both sides of the Atlantic. The majority of the papers included in this volume are a selection from those presented at the International Conference "Semantics and Pragmatics of Spanish" (The Ohio State University, November 1999) and others are special invited contributions. Among the topics covered are several that have been the subject of intense debate, whereas others represent subtle data patterns not considered so far. The topics include the proper characterization of tense and aspect, verbal periphrases, stage/individual-level predication, the interpretation of infinitives in embedded and adjunct clauses, the subjunctive mood, demonstratives, quantification of excess, exception phrases, binding phenomena, prepositional negative polarity items, particles of politeness, and pronominal doubling. Overall, the analysis of these subjects contributes new findings to prominent theories in the field, such as possible world semantics, relevance theory, mental spaces, type coercion, generalized quantifier theory, dynamic semantics, and the theory of the Logical Form interface. The book will be of interest to students and scholars in the areas of General Linguistics, Romance and Spanish Linguistics, and to all those interested in the crosslinguistic repercussions of current theoretical trends in semantics, pragmatics, and the syntax/semantics interface. Summarizing, this volume attempts to bridge the gap between theory and empirical analysis by focusing on several aspects of the semantics and the pragmatics of Spanish from a variety of theoretical points of view. After a brief tour of the history, trends and directions in Spanish semantics and pragmatics, the remainder of this introduction is devoted to presenting the sixteen contributions to this volume, which have been organized around five major subjects. Not surprisingly, this organization tends to reflect more the topics that have intrigued researchers in Spanish grammar for decades than an organization based on external or theoretical considerations. Thus, issues of tense, aspect and mood/modality concern at least six of the papers, whereas topics such as quantification or the interpretation of pronouns also find significant resonance in others.
2. A QUICK TOUR OF THE SPANISH LANDSCAPE Research in the semantics and pragmatics of Spanish and in Spanish has been particularly lively for decades. The most prominent works of the main analytic philosophers received highly praised translations into Spanish both in Spain and in Latin America (especially in Mexico and
From Words to Discourse
Argentina), and the path breaking works in philosophical semantics by Kaplan, Kripke, Quine, etc. were also translated in a very influential compilation prepared by Tomas Moro Simpson (1973). Sanchez de Zavala (1974) prepared a very useful compilation of the most important studies in generative and interpretive semantics of the early seventies, which allowed Spanish researchers to have access to materials originally published in technical reports and department internal series. Also during the seventies and early eighties, a generation of young linguists at both sides of the Atlantic explored semantic/pragmatic and structural aspects of Spanish grammar. These included, for example, negation (Bosque, 1980), the semantics of the adjective (Lujan, 1980), word order (Contreras, 1976), specificity and indefiniteness (Rivero, 1979), the subjunctive (Manteca, 1981), and some aspects of plurality (Sufler, 1982). In the area of formal semantics, the flourishing of Montague grammar sparked lively interest among several research groups in Spain. The compilation of Montague's papers Formal Philosophy was translated by Daniel Quesada and published in 1977 (Montague, 1977). Juan C. Moreno Cabrera wrote a very insightful and comprehensive monograph on Montague semantics (Moreno Cabrera, 1985), incorporating extensions and applications to Montague's original proposals. Other well-known representatives of this interest in the mathematical analysis of language, especially of the semantic component, are Serrano (1975), and Garrido Medina (1988), a very clear and extensive introduction to the application of logical methods in the semantics of Spanish (cf. also Acero, Bustos, and Quesada, 1982; Quesada, 1985; and Valdes, 1991, for philosophical perspectives). In this same vein, Moreno Cabrera (1994) develops extensive analyses of the semantics of noun phrases and verb phrases incorporating recent trends such as the theory of generalized quantifiers, plurality and event semantics. Ojeda (1991) is one of the best studies of the semantics of plurals, with a pioneering study of the Spanish neuter. Gutierrez-Rexach (1998) is the first introduction to generalized quantifier theory in Spanish. During the nineties, the annual symposium on formal and natural languages (Lenguajes Naturales y Lenguajes Formales) organized by Carlos Martin Vide (Martin Vide, 1986-1997) contributed to a cross-fertilization of formal, computational and logical perspectives in the analysis of natural language and Iberian languages in particular. On the pragmatic front, the first applications of Searle's theory of speech acts and Austin's logic of conversation, and of the debates on mood and presupposition to Spanish came from philosophers interested in linguistic phenomena (cf. Bustos, 1986). During the nineties, a renovated interest in relevance theory (Escandell, 1996) developed on a par with general pragmatic approaches (cf. the excellent introductions by Reyes, 1990, 1995), the study of discourse markers (Martin Zorraquino and Montolio, 1998; Portoles, 1998), colloquial usage (Briz, 1998), metaphor (Bustos, 2000), presupposition (Garcia Murga, 1998), and general pragmatic and praxiological approaches (Sanchez de Zavala, 1994, 1997). The connections between pragmatics and sociolinguistics, i.e. the exploration of the relationships among language use, context and diverse instances of social stratification or socially constructed groups (gender or ethnically based) have also witnessed an incredible expansion, especially in the study of Spanish in the United States (cf. the recent introductions by Moreno Fernandez, 1998 and Silva-
Introduction
5
Corvalan, 2001 for detailed surveys). From this glance at the field of Spanish semantics and pragmatics, it becomes obvious that a selection of papers focused on grammar-centered topics will only offer the reader a partial look at what is going on in the field as a whole. Nevertheless, it can also be taken as evidence of its increasing sophistication, which precludes comprehensive anthologies, an also as a stimulus for future endeavors.
3. TENSE AND ASPECT The tense and aspectual properties of the Spanish verbal system have intrigued grammarians for decades. It is a well-known fact that the Spanish verbal paradigm is a very complex system from morphological viewpoint. Tense classifications that simply correlated morphological differences with vague semantic notions have been deemed unsatisfactory for a long time. Traditional grammarians, from Bello (1847) to Bull (1960, 1965), put forward very insightful proposals that reduced the apparent complexity to more basic semantic correlations. In recent years, following the ideas of Reichenbach (1947) and Hornstein (1977), several researchers (Acero, 1990; Garcia Fernandez, 2000) have deconstructed the ingredients of the Spanish tense system, and a similar impetus can be observed with respect to aspectual properties (cf. the articles in Bosque 1990a, among others). The three chapters in this section take a step further and seek to examine how a combination of insights from semantic/pragmatic theories is needed to explain even very basic temporal and aspectual properties. Silvia German's contribution, "Spanish Past and Future Tenses: Less (Semantics) is More", addresses the issue of how semantic and pragmatic factors are relevant in the determination of simple past/future tenses. Spanish grammarians typically propose a single semantic notion underlying the multiple uses of tense morphemes such as simple future -re/-ra and preterito (simple past). King (1992) and Gili Gaya (1961), for example, propose that -re/-rd invariably conveys a future temporal perspective. However, Spanish simple future can receive both a future reading, as in (la), or a modal non-future one as in (Ib): (1)
a. Juanvendrd. "John will come" b. Juan estard en casa. "John will be at home now"
Similarly, the preterito or simple past is normally assumed to convey semantic perfect aspect, which distinguishes it from the imperfecto (RAE, 1973), which is also a past tense. However, its perfectivity varies depending on properties related to aktionsart. For example, preterito may not imply the completion of the action, as in (2):
From Words to Discourse
(2)
Juan estuvo enfermo. Y todavia lo estd. "John was sick. And he still is"
In these cases, the standard tense definitions fail, leading to the postulation of post-hoc exceptions. Gennari argues that factors other than tense meanings, such as aktionsart and pragmatics, influence temporal interpretation. A new approach to temporal semantics is needed, where these factors combine with semantic definitions to predict tense interpretation and acceptability. Consider the simple future (-re/-rd). First, it refers to a time later than the speech time (ST). The non-future reading of (Ib) follows from the independent fact that (Ib) denotes a state. Note that present readings of -re/-rd only appear with stative sentences. In contrast, non-statives only receive future readings. States have this temporal effect because they trigger the pragmatic inference that they obtain at a larger interval containing the time specified by the tense. If the state is located by -rd/-re at a future time t, its properties generate the inference of its truth at a larger interval i containing t, which can overlap with the ST. The interval associated with (Ib), but not (la), may thus overlap with the ST. Second, -re/-rd also requires that its temporal proposition follow from the context or conversational background (CB) and imposes a pragmatic felicity condition: the CB should be realistic. A realistic CB is a set of propositions true in the actual world (actual facts are taken for granted). Thus, -re/-rd[Q] is true iff FUT[Q] follows from the realistic CB. This definition predicts the reading of (Ib) and (la): the future proposition Q follows from the assumed CB (but Q could overlap with ST in (Ib) due to aktionsart). This also predicts why -rd/-re is not acceptable in non-realistic contexts, as in #57 fuera rica, comprare un auto '#If I were rich, I will buy a car'. As for preterito, Gennari argues that rather than convey perfectivity, it triggers a quantity implicature. This is clear with atelic events (with telic events, telicity rather than preterito entails completion). Because imperfecto, which entails the possible continuation of an event until the ST, is always available, the use of preterito implicates the end of the event before the ST. Preterito, the less informative form, implicates that the more informative imperfecto-reading does not obtain. Note that this implicature is cancelled in (2), thus supporting a pragmatic account. Thus, by including aktionsart and pragmatics in her model of temporal interpretation, Gennari advocates the idea that traditional lexical semantic definitions of tense can be simplified, and it is also possible to consistently account for all possible readings of Spanish past and future tenses. These data support an integrated semantic/pragmatic approach to Spanish tenses, and to temporal semantics in general. In "Tensed Complements of Perception Verbs: Issues in their Temporal Interpretation", Alicia Cipria explores the semantics and pragmatics of past tense complements of perception verbs, such as ver 'to see', oir 'to hear', palpar 'feel by touch', and oler 'to smell'. Complements of perception verbs are particularly interesting due to the different interpretations they receive. Traditional accounts of this class of verbs state that the complement verb must
Introduction
have a simultaneous relationship with the main verb of perception. In cases where this does not obtain, the main verb is interpreted as a cognitive rather than a sensory verb of perception (Gili Gaya, 1961; RAE, 1973). Cipria argues that temporal relations across a complex clause with a main perception verb are not as simple as claimed in some of the existing literature on the topic. Rather, the interaction of aspect/aktionsart and pragmatics with the lexical requirement imposed by the main verb plays a role in the resulting temporal interpretations. Since perception verbs mostly take indicative complements, the issues discussed here bring interesting insights to the general issue of "sequence of tense (s)", which in the Spanish literature has been dealt mostly within the context of subjunctive complements. In general, imperfecto always entails atelicity for its complement clause, regardless of predicate type or other elements, while the preteriio can participate in telic or atelic situations. The import of aktionsart effects in perception contexts lies in the fact that simultaneous readings (be they sensory or cognitive) only arise when the aktionsart of the complement clause is atelic, while there will be backward shifted readings when the complement clause displays telic aktionsart. The fact that the preterito is compatible with both telic and atelic aktionsart gives rise to important consequences in perception contexts. With complements that have no lexical requirements on the temporal interpretation of their complements, the backshifting of the preterite always takes place, regardless of the aktionsart of the complement clause. However, under a verb of sensory perception, the options for the preterito are different: simultaneity is necessary for the sensory interpretation to arise and, in turn, only atelic situations permit simultaneous readings. Under perception verbs, the preterito backshifts when in a telic situation, and it may describe a simultaneous (or seemingly simultaneous situation) when in an atelic situation. Consequently, there is a clash between the lexical requirement of the main verb of sensory perception (for simultaneity) and the preterito embedding effects. For the sensory interpretation to be available, the main verb requires simultaneity of its complement and a condition of "witnessing" or evidentiality also needs to be met. For the cognitive interpretation to arise, however, there are no restrictions as to the type of temporal relationship between main and complement tenses, and there are no restrictions arising from aktionsart effects either. This seems a natural fact, since mental processes are more likely to be associated with any situation, whatever their location in time, whereas a sensory situation requires direct experience. In sum, Cipria claims that the alleged simultaneity requirement of perception verbs does not seem to always work with preterito complements, which may yield a backward shifted reading when embedded under another past. As for imperfecto complements to past main verbs of perception, she concludes that the supposed simultaneity requirement can be overridden by the interaction of aspect, aktionsart and pragmatics, and that the peculiarities of sensory contexts can be better accounted for in terms of evidentiality and actual occurrence, independently of tense specification and aktionsart effects. The study of verbal periphrases is normally one of the topics that one expects to see treated in more detail in traditional and pedagogical grammars, and indeed this is the case. Nevertheless, it is surprising that to date there has been very little theoretical elaboration on the
From Words to Discourse
morphosyntactic and semantic restrictions on the combination of verbs and auxiliaries that form the periphrasis. The proper deconstruction of their linearization, tense/aspectual properties is also mostly an open question. Brenda Laca acknowledges this point in her extensive and detailed study of the semantic and structural properties of Spanish aspectual periphrases, "Spanish 'Aspectual' Periphrases: Ordering Constraints and the Distinction Between Situation and Viewpoint Aspect". In aspectual periphrases, which follow the pattern "finite verb + infinitive/ gerund", the finite verb is either an "aspectualizer" denoting not a situation type but a part of the temporal structure of any situation type, or a verb of motion or location. These "superlexical" verbs express notions such as use (soler), beginning (empezar), or ending (acabar/ terminar). Verbs of motion or location express notions such as (re)turn (tornar/ volver), location (estar), motion (ir/andar), put (poner), leave (dejar), follow (seguir), and stop (parar). Laca builds on current research by Cinque (1998, 1999), who has put forward a syntactic proposal on the ordering constraints that can be observed in periphrastic constructions. More specifically, for Cinque, the higher/finite verbs in such constructions are bona fide auxiliaries that lexicalize a functional (aspectual) head. Cinque argues for a cross-linguistically invariant, highly articulated structure of aspectual functional projections. Together with the rigid order of adverbs and of aspectual particles and affixes, the purported rigid order of "aspectual" periphrases in Romance is taken to provide evidence for the assumed hierarchy of functional projections. Laca distances herself from this type of account and points out several empirical and theoretical inadequacies of such type of syntactic approach. She correlates ordering possibilities with a major distinction which is grounded in semantics, namely the distinction between situation and viewpoint aspect (Smith, 1991). She furthermore argues that a unified account of aspectual periphrases is neither possible nor desirable. Aspectual periphrases divide both syntactically and semantically into two different categories based on the distinction between extrinsic and intrinsic ordering principles, well known in the realm of morphology, where they usually serve to distinguish inflectional from derivational affixes. She establishes two main correlations: (a) between viewpoint aspect and periphrases that are both more peripheral (more external) with respect to the lexical verb they apply to and exhibit extrinsic ordering properties (being rigidly ordered or in complementary distribution among themselves); (b) between situation aspect and periphrases that are both less peripheral, i.e. appear nearer the lexical verb they apply to, and are more freely ordered among themselves. Laca concludes her study examining Spanish periphrases in which semantics and ordering possibilities seem to diverge and periphrases from other Romance languages that apparently exhibit the same semantics but different ordering possibilities than the Spanish ones.
Introduction
4. THE INFINITIVE AND THE SUBJUNCTIVE While the alternations that can be observed between the preterito and the imperfecto, the past and the future or periphrastic constructions are related to temporal, aspectual and aktionsart properties of diverse semantic and pragmatic import, the interpretation of infinitives, imperatives, and the subjunctive requires bringing in subtle considerations about modality and the modal structure of propositions. One of the most salient problems in the semantics and pragmatics of infinitives is how to reduce the diversity of its apparent uses to a single core and how to explain some of its most salient functions, such as the imperative use, in terms of a theory of mood and force. Xose Rosales Sequeiros' chapter "Non-Declarative Sentences in Spanish: The Case of the Infinitive" is concerned with the imperative use of the infinitive in Spanish. In particular, the author examines both its semantics and pragmatics, and provides an account that brings together these two aspects of the infinitive and its imperative use within the theoretical framework of relevance theory. Traditionally, the imperative is seen as a mood in its own right. However, Wilson and Sperber (1988) argue that they see no reason to assume that semantic moods exist. They further argue that the characteristic linguistic features associated with declarative and non-declarative sentences only encode an abstract property of the intended interpretation: the direction in which the relevance of the utterance is to be sought. Wilson and Sperber (1988) claim that the infinitive and the imperative differ semantically in that the latter encodes the property that the state of affairs described is both potential and desirable, while the former encodes the property that the state of affairs is only possible. The semantic meaning of Me encantaria vivir en el campo ('I would love to live in the countryside'), for example, could be paraphrased as Me encantaria tener la posibilidad de vivir en el campo ('I would love to have the possibility of living in the countryside'). Wilson and Sperber (1988) also argue that this does not preclude the use of the infinitive as an imperative when assumptions about achievability and desirability are available in the context. Rosales Sequeiros criticizes this approach because it does not take into account a number of issues:(a) does the imperative meaning become part of the prepositional form of the utterance or is it just a pragmatic contextual assumption?; (b) what pragmatic processes, if any, are involved in this imperative use?; and (c) why is the imperative use of the infinitive more common in Spanish (and other Romance languages such as Galician) than in English? The author argues that whereas the semantics of the infinitive is restricted to possibility, in an imperative use its contribution to the prepositional form of the utterance is enriched via pragmatic processing. This means that its contribution is to the explicit content of the utterance and not to the implicit content, as Wilson and Sperber suggest. This also seems to provide an answer to the third question raised above, namely the repeated use of the infinitive as imperative in German, Spanish and other Romance languages, and its contribution to the explicit content of the utterance has made it possible for this use to be more accessible and become semi-
10
From Words to Discourse
lexicalized. In English, the same (politeness and referential) effects are achieved by other means and, as a result, the infinitive is rarely, if at all, used as an imperative. Luis Alonso-Ovalle's "Aspect and Situations: A Situation Semantics Account of the Semantic Variability of Spanish '^/-Clauses'" analyzes the interpretation of infinitives in adjunct clauses headed by the prepositional complementizer al. The semantic variability of Spanish '^/-clauses' (ACs), a type of infinitival free adjunct (Stump, 1985), challenges a formal analysis. For example, the sentence Al entrar en casa, Maria me sonrie has two interpretations: (a) generic: 'for most situations s such that Maria enters home in s, Maria smiles in s'; and (b) episodic, either temporal ('coming back home, Maria is smiling at me') or causal ('since she comes back home,...'). While ACs headed by non-perfect infinitives can be interpreted as restrictors of adverbs of quantification (QA) or adsentential modifiers (either temporal or causal), neither perfect infinitives nor individual-level predicates can be restrictors. Thus, Pedro (siempre/rara vez) canta al conducir ('Whenever he drives, Peter always sings') contrasts with *Pedro (siempre/rara vez) canta al haber conducido and *Pedro (siempre/rara vez) canta al ser alto. Furthermore, while nonperfect infinitives can be interpreted as either causal or temporal adsentential modifiers (3a), perfect infinitives and individual-level predicates can only yield a causal reading (3b-c). (3)
a. Al volver a casa, Maria me sonrio. "When she came back home, Maria smiled at me" "Since he came back home, Maria smiled at me" b. ...Al ser unapersona de cardcter, creo que terminaria igual que G. "Because he is a bad temper person, I believe that he would end up like G." c. Al haber sido quemado, este elemento genera residues cancerigenos. "Since it has been cremated, this element generates cancerous residues"
Alonso-Ovalle shows that a Kratzerian situation semantics framework (Kratzer, 1989) makes the semantic variability of ACs amenable to a compositional formal analysis. Constituency tests show that ACs can be either infinitival VPs, denoting properties of minimal situations or eventualities, or Aspect Phrases (AspPs), denoting properties of times. Al adjoins either to VPs (all) or AspPs (13/2). Whereascr// is shown semantically vacuous, a/2 requires that the running time of each situation in the denotation of the main clause be included in the set of times denoted by the AspP it adjoins to. The above presented data are shown to be derivable from this hypothesis. First, AspP ACs (including individual-level predicates) cannot be restrictors of AQs, because AQs only take properties of situations as restrictors. AspPs can only be adsentential modifiers. Second, within AspPs, nonperfect infinitives sometimes incorporate an imperfective operator. Given the semantics of ah, imperfective ACs are interpreted as temporally overlapping with the main proposition. However, the aspectual values of both perfect and individual-level predicates make
Introduction
11
this reading tautological. Following Stump (1985), Alonso-Ovalle claims that the causal content they convey is pragmatically inferred. The analysis hinges on the discussion of relevant theoretical issues and offers empirical evidence from Spanish, i.e. the treatment of aspect within situation semantics, the relation between situations and eventualities, and several questions concerning the syntax-semantics interface, especially with respect to aspectual heads. Research on the Spanish subjunctive has also uncovered a wealth of very intriguing data that has been discussed over the years within very different frameworks. Again, the debate has mostly exposed the tension between a variety of uses and an underlying core. Some landmark contributions and approaches are included in Bosque (1990b). One of the most influential recent contributions to the topic is Mejias-Bikandi (1996). He studies what he calls open contexts, within the framework of the theory of mental spaces (Fauconnier, 1985, 1997; Fauconnier and Sweetser 1996). Open contexts in Spanish are constructions where either the indicative or the subjunctive mood can be used, as in (4). (4)
No creo que ha/haya llegado. not I-think that s/he-has-IND/has-SUBJ arrived "I do not think that s/he has arrived"
Mejias-Bikandi (1996) studies several phenomena that illustrate some transfer of information or some connection between different domains in the representation of discourse. In mental spaces terms, the three phenomena illustrate some cross-spatial dependency or sharing of information. He observed that such cross-spatial connections are possible when the complement clause appears in the indicative mood, but they are not possible when the complement appears in the subjunctive mood. In general, mood is viewed as a mechanism that controls the flow of information among different domains in the representation of discourse. An abstract characterization is given in terms of accessibility. Francisco Aliaga and Eduardo Bustos, in "Mental Spaces and Epistemic Attitudes: On the Spanish Subjunctive/Indicative Alternation", examine some of the problematic aspects of Mejias-Bikandi's proposal and present some ingredients of a potential alternative based on the theory of epistemic attitudes. They argue that this theory provides a better explanation of the mandatory nature of indicative or subjunctive mood in certain syntactic environments, and of the discourse or communicative role of mood choice, whenever such election is possible. In turn, Mejias-Bikandi proposes a modification of his 1996's theory in "Space Accessibility and the Pragmatic Status of Propositions" within the information structure framework developed by Lambrecht (1994). Mood in Spanish is not primarily a grammatical mechanism that controls the flow of information in discourse, but rather a grammatical coding of the pragmatic status of the proposition expressed by a clause. Indicative and subjunctive are correlated with the status of a proposition as active or semi-active in a mental space. When a proposition is active, connections across mental spaces are possible; when a proposition is semi-active there are restrictions on the sharing of the information expressed by that proposition among different
12
From Words to Discourse
domains in the representation of discourse. Indicative mood is correlated with the active status of the associated proposition, and subjunctive with the semi-active status.
5. COERCION, QUANTIFICATION AND POLARITY The chapters in this section analyze several aspects in the semantics of predicates, modifiers, polarity expressions and deixis. In "Coercion and the Stage/Individual Distinction", Victoria Escandell and Manuel Leonetti explore a very interesting aspect of the distinction between individual-level predicates (ILPs) and stage-level predicates (SLPs), first introduced by Carlson (1977). Although there are several well-known environments that allow one class of predicates and preclude the occurrence of the other, there are also some syntactic environments in which an ILP appears where a SLP should be expected, and yet the sentence is not ungrammatical. Rather, the ILP is forced to be interpreted as a SLP. Escandell and Leonetti study the circumstances under which this shift happens and propose a systematic explanation of this apparently exceptional behavior. They claim that the mechanism responsible for the shift is coercion, understood as a reinterpretation process set up to eliminate the conflicts between the semantic content of a constituent and the requirements of other elements in the same construction (Pustejovsky, 1995). The effects of coercion processes have been observed in apparent aspectual mismatches between predicates and adverbs, and the reinterpretation of mass nouns as count nouns. Coercion always has a triggering element and the coerced element has to be in the scope of the trigger. The authors study several cases of coercion with different triggers: the copula verbs ser and estar, absolute constructions, secondary predicates and small clauses. Going a step further, they argue that coercion is a transversal operation that affects both the logical form of a sequence (a matter of grammar) and the derivation of explicatures (a matter of pragmatics). Thus, coercion phenomena fit quite well with the overall picture of semantics and pragmatics developed in relevance theory, where enrichment and loosening processes are claimed to apply to conceptual content in deriving explicatures. Some quantifier modifiers have been recently discussed in connection with so-called exceptive or exclusion phrases such as No/Every student except Leo/five. Analyses of such constructions show that they are directly related to such important theoretical problems in semantics as quantifier reducibility, semantic compositionality or categorial polyvalency (Keenan, 1993; Moltmann, 1995). The main goal of Zuber's "Some Spanish Quantifier Modifiers" is to analyze various Spanish constructions with quantifiers modified by this type of expressions within the framework of generalized quantifier theory. The constructions discussed include not only exclusion phrases but also positive inclusion phrases, negative inclusion phrases, and other related constructions. In all these cases, noun phrases are modified by a complex expression formed by a binary connective (salvo, excepto, ademds, aparte de, etc.) and a fixed argument. There is however a categorial difference between modification by salvo,
Introduction
13
incluso, ni siquiera on the one hand and ademds and aparte on the other hand. In first type of modification, a unary determiner is the modified element. In the case of ademds, the modified element is a noun phrase. The connective aparte is categorially polyvalent: it can take as one of its arguments (or form a modifier which takes as its argument) either an NP or a determiner. Various semantic properties are correlated with this categorial difference. There are two modification possibilities: either the first argument of the connective is modified or the second argument is modified. Normally, the functions denoted by modifiers have to be restrictive. Several issues related to compositionality and generality of application arise, including interrogative environments. Zuber studies in detail the interaction of inclusion/exclusion phrases with interrogative quantifiers and determiners and observes that there are interesting consequences that can be derived, especially in connection with the property of exhaustivity. In "Demonstratives in Context", Javier Gutierrez-Rexach explores the semantics of Spanish demonstratives and argues that a detailed analysis of some of the peculiarities of these expressions sheds new light on long standing debates on the nature and properties of deixis. The paper offers a detailed review of current theories of demonstrative content, from the traditional indexical theory, to the Kaplanian theory of character (Kaplan, 1989), and the analysis of demonstratives as defmites. It is argued that when issues of dynamic interpretation are considered, the evidence seems to favor an analysis as determiner functions restricted to context sets, along the lines proposed in generalized quantifier theory. Thus, the theory proposed treats demonstratives as dynamic generalized quantifiers. This allows for an integration of Kaplan's clear considerations about the distinction between demonstrative "character" and truthconditional content and a more conservative view of demonstratives as another type of determiners. Contextual restriction has to be viewed as a dynamic process, sensitive to the information available in discourse and to the shifting background and intentions of the conversation participants. The properties of uniqueness, familiarity, proximity and demonstration are treated as constraints on the determiner's context set, i.e. they are equivalent to presuppositions that have to be satisfied in the context. With regard to Spanish demonstratives, the particular peculiarities of tripartite demonstrative systems (those which encode oppositions among three terms: este-ese-aquel) are argued to be of relevance for the understanding of the mechanics of demonstration. An initial characterization of Spanish demonstratives in terms of the proximal/non-proximal distinction (proximity with respect to the speaker or addressee) is considered and several predictions are tested. Nevertheless, following the insights of the grammarian Salvador Fernandez Ramirez (1951), it is proposed that proximity relations are of a flexible nature and can take place in different domains, as suggested by Karl Biihler. The demonstrative ese works as a neutral term with respect to spatial positioning, i.e. it is unmarked for proximity presuppositions and can be used to express a variety of contrasts. Other topics covered are the interactions between demonstratives and spatial/temporal elements, figurative deixis and discourse deixis, and the interpretation of sequences where more than one demonstrative occur.
14
From Words to Discourse
The last chapter in this section is Javier Gutierrez-Rexach and Scott Schwenter's "Prepositional NPIs and the Scalar Nature of Polarity". The study of polarity sensitivity has been one of the topics that has sparked greater interest in semantics and pragmatics in recent decades. Currently, there are a variety of competing theories in different domains of linguistic research (syntactic, semantic, pragmatic, or some combination thereof). There is also increasing attention being paid to cross-linguistic variation and to polarity phenomena in different languages such as Greek, Italian, Spanish, Catalan, Dutch, etc. The authors argue that there is a class of negative polarity items (NPIs) which have prepositional content and which present new challenges for a general theory of NPI meaning and interpretation. Prepositional NPIs are related to scales of different types, including scales of expectations and those related to speaker's attitudes more generally. Attention to the contribution of semantic pragmatic scales has been the focus of a longstanding tradition in the study of NPIs and ordered entailments, along the lines initially proposed by Horn (1972) and Fauconnier (1975a, 1975b). Gutierrez-Rexach and Schwenter focus on the semantics and pragmatics of a prepositional NPI in particular: que digamos (lit. "that we (might) say [SUBJ]"). Most prepositional NPIs are minimizers along prepositional (epistemic) scales which participate in strong/emphatic refutations. On the other hand, que digamos has a host of very intriguing properties. It can be classified as a strong NPI, licensed by negation and not by merely decreasing environments. The essential property of this item is that it places the associated proposition in a non-endpoint position on a scale that orders degrees of a given property, or lack thereof. Contrastingly, other scalar NPIs such as en absolute 'at all' clearly mark scalar endpoints. Que digamos, in addition to its attenuating value, is a scalar NPI that situates the event or situation described by the predicate on a scale of expectations. The presence of que digamos serves as an indicator that the position of the predicate on the scale is lower than it should have been or lower than would be expected by the participants in the conversation. Additional implicatures of this NPI are also studied, as well as its behavior in interrogative environments and in the scope of other elements.
6. LOGICAL FORM AND INTERPRETATION The impact of semantic and pragmatic aspects in grammar design is an obvious but difficult aspect to handle by theoretical models of grammar/syntax. From Chomsky's initial strong claims on the autonomy of syntax to functionalist approaches that defend grammatical structure as a means to communicate speakers' intentions, there are a wide array of theoretical positions on the relationship between structure and meaning. In general, the negative or controversial views of the fifties, sixties and seventies, including the so-called linguistic wars on the generative or interpretative status of semantics, have been substituted by integrative approaches which precisely try to explore how meaning informs syntax and vice-versa. The three papers in this
Introduction
section address precisely this point: how the consideration of semantic and pragmatic properties of expressions helps us enrich the theory of Logical Form as a level of linguistic representation. Ignacio Bosque discusses some aspects of the expression of modality in grammar in "Degree Quantification and Modal Operators in Spanish". He puts forward a proposal to integrate the modal properties of a number of degree quantifiers and quasi-quantifiers into the restrictive concept of grammar provided by the generative approach to the syntax/semantics/pragmatics interface. First, he analyzes one of the cases in which these quantifiers interact with modal structures, namely degree quantifiers that denote 'excess', such as demasiado(s), ('too much/many'), desmesurado ('disproportionate'), excesivo ('excessive'), and their respective derived adverbs. Bosque argues that some of the essential properties of these quantifiers follow from their intensional nature. Degree quantifiers that denote excess are licensed by an epistemic modal operator. The properties of these quantifiers at Logical Form are the result of the interaction of this operator with other components of the clause and with the syntactic processes which take place therein. Finally, Bosque explores another case in which degree quantification is related to modality. He shows that some of the basic properties of quantifiers of excess extend to a number of adjectives and adverbs of assessment and extreme degree, such as esplendidamente ('splendidly'), de maravilla ('marvelous(ly)'), estupendo ('great'), etc. The relevant configuration retains some basic aspects of the logical form of directive predicates. Additionally, attributive prenominal adjectives share some features of the latter class, so a similar analysis should apply to them. En? (1987) identified two readings for the English past tense in embedded clauses which she called the simultaneous reading and the shifted reading. She noticed that the simultaneous reading was only available when the predicate in the embedded clause was stative. In such simultaneous readings, the time of the complement clause is identical to the time where the matrix sentence is evaluated (and not prior to it, as in the shifted reading). In "The Shifted Reading of the Spanish Simple Past as Dependent on Presupposition Accommodation", Rosa Junia Garcia Cordoba studies the interpretation of Spanish (Mexican dialect) simple past tenses in complement clauses, and observes that whereas the preterito lacks the simultaneous reading, the imperfecto or copreterito, like the English past tense, is ambiguous between a simultaneous (preferred) reading and a shifted reading. As in English, then, this phenomenon is related to aspect, although in Spanish it is grammatical aspect and not verbal aktionsart that matters. The shifted reading of the copreterito requires that a past time other than that in the matrix clause be contextually salient, but this time need not be linguistically expressed. More accurately, it may be expressed by the copreterito alone. Garcia Cordoba follows Stowell (1993) in his analysis of tenses as temporal predicates, but differs from him in that she postulates that these past tenses always contain a Past predicate. The Logical Form context will determine if this Past gets coindexed with the c-commanding embedding Past, giving rise to the simultaneous reading, or whether Past will be calculated with respect to Speech time, and possibly co-refer with other times in the (non) linguistic context, which corresponds to the shifted reading. Crucial for the determination of the position of Tense at Logical Form is a process of presupposition
16
From Words to Discourse
accommodation. Thus, an anti-factive embedding predicate, which forbids accommodation, will determine that the embedded clause remain in situ, where its only possible reading will be the simultaneous one. Under such predicates, the preterite, which never allows simultaneous readings, is excluded, and the copreterito will be unambiguously simultaneous. A clause embedded under other predicates of other classes, on the other hand, may have a shifted reading, for it will be raised so that its tense will be local enough with respect to Speech time. Josep Macia's "On the Interaction of Syntax-Semantics-Pragmatics: A Case Study" proposes a revision of the theory of linguistic binding. It is a well-known fact that the structure of a sentence places constraints on the possible interpretations of the noun phrases (NPs) that occur in it. For instance, (5a) cannot be interpreted so that both NPs refer to Juan, and (5b) cannot be interpreted as meaning that every man loves himself: (5)
a. b.
Juan lo vio. "Juan saw him" Cada hombre lo ama. "Every man loves him"
Principles (B) and (C) of the Standard Binding Theory (SBT) can account for (5a,b) as well as many other phenomena. There are, nevertheless, counter-examples to Standard Binding Theory in which these principles are violated. Macia presents a proposal that accounts for the phenomena that are problematic for SBT and that is based on the following general ideas. First, he distinguishes among the counter-examples to SBT different kinds of phenomena that require different sorts of explanation. Second, contrary to what SBT might lead us to think, the phenomena to be explained is not mainly of a syntactic nature, but of a semantic one: we need to explain why certain interpretations of a sentence are not possible. Finally, Macia develops an alternative to SBT in which principles (B) and (C) are semantic principles that constrain the relation of presupposed co-reference. The application of these binding principles is always with respect to a context (in the sense of Stalnaker, 1978) and subject to a series of pragmatic constraints. Binding principles are just one among several sources of semantic information. Macia describes how these different sources of semantic information interact with the binding principles in order to provide an interpretation for a sentence. Then, we can not only account for the fact that some sentences that SBT would rule out are acceptable to speakers, but also for the fact that these counter-examples to SBT appear to have different degrees of acceptability.
7. PRAGMATICS AND GRAMMATICAL VARIATION The last two chapters of this book address topics in the vast area of pragmatically induced variation. Clitic doubling (CD), clitic climbing (CC) and null objects (NO) are three phenomena
Introduction
17
not commonly correlated in the literature on clitic pronouns. The first two phenomena have been widely studied in Romance languages from many different points of view. NOs (non-overt direct objects) have received less attention and they have only been reported in language contact situations. The main hypothesis of Colantoni's contribution, "Clitic Doubling, Null Objects and Clitic Climbing in the Spanish of Corrientes (Argentina)", is that the three phenomena exhibit positive and negative correlations based on discourse properties and on the degree of animacy of the referent of the pronoun. She shows that CD and NO represent two extreme discourse-pragmatic strategies. Clitics are doubled when the lexical item encodes new referents (Chafe, 1976; Lambrecht, 1994) or when it is a contrastive focus. Clitics are also more often doubled the higher their referent is in the animacy scale. In contrast, NOs are found when the referent is textually active (Lambrecht, 1994), and is low in the animacy scale (all null objects have an inanimate referent). CC is a strategy for enhancing contrastiveness: clitics climb when they receive a contrastive focus. CC is also favored by the doubling of the clitic by a lexical NP in a topicalized position, a fact that relates CC and CD. The two phenomena are also related by another property: both CD and CC are more frequent when the referent of the pronoun is [+human]. Another factor that favors CC is the degree of grammaticalization of the auxiliary verb. CC is not expected when the auxiliary verb retains its original meaning. Extra-linguistic variables, like gender and location, are important in explaining the distribution of the three phenomena. Analyzing twelve hours of speech collected for the Linguistic-Anthropological Atlas of the Argentine Republic, Colantoni found that CD has an even geographical distribution but it is slightly more frequent among men. CC is more frequent among women, and NOs are more frequent in locations where Guarani is more used. These data suggest the need of further studies that systematically consider variables such as gender, location, and language contact. Robert Vann, Joan Busquets and Dale Koike study the pragmatics of a discourse marker or connective in "Spanish no, si: A Particle of Politeness". In everyday conversation, particles of politeness (Brown and Levinson, 1987) serve to indicate how addressees perceive and interpret the pragmatic intents of previous utterances. Such is the case of the Spanish response no, si 'no, yeah', a previously uninvestigated discourse marker that they examine in conversational interaction. Vann, Busquets and Koike argue that no, si represents a pragmatic response to a perceived face-threatening act. They study data drawn from over fifty conversations with speakers of both Peninsular and American Spanish to illustrate their analysis. After identifying the various discourse contexts in which no, si occurs in their data, they propose a pragmatic analysis that considers issues of politeness, pragmatic alignment, and preference structure in accounting for how no, si gives feedback to interlocutors regarding how an utterance or set of utterances has been understood. Following principles of politeness in Brown and Levinson (1987), Goffman (1981), and Schegloff et al. (1977), they demonstrate that the use of no, si represents a dual conversational management strategy of (a) acknowledging a perceived facethreatening act (or dispreferred first) and reacting to it by either correcting the interlocutor in a move of negative politeness (a dispreferred second) or by accepting previous utterances in a move of positive politeness and (b) moving toward alignment, at times hedging the responses
18
From Words to Discourse
that follow. As such, no, si functions as an evaluative expression of footing in which the user can signal that some element of disharmony was perceived and moves to re-establish cooperative harmony, standing ground but avoiding disagreement at the same time. The authors conclude that, more than the mere combination of opposing adverbs, no, si is a compound particle with a set of specialized yet delineable pragmatic functions related through implicature. An important resource for the study of politeness and alignment strategies, this highly specialized particle seems to have formal surface parallels in neighboring Romance languages such as Catalan (no, si) and French (mats, si). These parallels support further investigation of such significant, yet under-investigated, phenomena to better describe and understand the connections between discourse particles and pragmatic theory.
REFERENCES Acero, J. 1990. "Las Ideas de Reichenbach Acerca del Tiempo Verbal." In I. Bosque (1990a), 45-75. Acero, J., E. Bustos, and D. Quesada. 1982. Introduction a la Filosofia del Lenguaje. Madrid: Catedra. Bello, A. 1847. Gramdtica de la Lengua Castellana Destinada al Uso de los Americanos. Tenerife: Cabildo Insular, 1981. Bosque, I. 1980. Sobre la Negation. Madrid: Catedra. Bosque, I. (ed.) 1990a. Tiempo y Aspecto en Espanol. Madrid: Catedra. Bosque, I. (ed.) 1990b. Indicativo y Subjuntivo. Madrid: Taurus. Brown, P. and S. Levinson. 1987. Politeness: Some Universal in Language Usage. Cambridge: Cambridge University Press. Briz, A. 1998. El Espanol Coloquial en la Conversation. Esbozo de Pragmagramdtica. Barcelona: Ariel. Bull, W. 1960. Time, Tense and the Verb. A Study in Theoretical Linguistics with Particular Attentionto Spanish. Berkeley: University of California Press. Bull, W. 1965. Spanish for Teachers. Applied Linguistics. New York: Ronald Press. Bustos, E. 1986. Pragmdtica del Espanol: Negation, Cuantificacion y Modo. Madrid: UNED. Bustos, E. 2000. La Metdfora. Ensayos Transdisciplinares. Mexico: Fondo de Cultura Economica. Carlson, G. 1977. Reference to Kinds in English. New York: Garland. Chafe, W. 1976. Subject and Topic. New York: Academic Press. Chomsky, N. 1995. The Minimalist Program. Cambridge, Mass.: MIT Press. Cinque, G. 1998. '"Restructuring and the Order of Aspectual and Root Modal Heads." Univ. of Venice Working Papers in Linguistics 8. Cinque, G. 1999. Adverbs and Functional Heads. New York: Oxford University Press. Contreras, H. 1976. El Orden de Palabras en Espanol. Madrid: Catedra. En9, M. 1987. "Anchoring Conditions for Tense." Linguistic Inquiry 18. 633-657. Escandell, V. 1996. Introduction a la Pragmdtica. Barcelona: Ariel. Fauconnier, G. 1975a. "Pragmatic Scales and Logical Structure." Linguistic Inquiry 6. 353-375. Fauconnier, G. 1975b. "Polarity and the Scale Principle." CLS 11. 188-199. Fauconnier, G. 1985. Mental Spaces. Cambridge: Cambridge University Press. Fauconnier, G. 1997. Mappings in Thought and Language. Cambridge: Cambridge University Press. Fauconnier, G. and E. Sweetser. 1996. Spaces, Worlds and Grammars. Chicago: University of Chicago Press. Fernandez Ramirez, S. 1951. Gramdtica Espanola. Madrid: Revista de Occidente.
Introduction
19
Garcia Fernandez, L. 2000. La Gramatica de los Complementos Temporales. Madrid: Visor. Garcia Murga, F. 1998. Las Presuposiciones Lingiiisticas, Bilbao: Universidad del Pais Vasco. Garrido Medina, J. 1988. Logicay Linguistica. Madrid: Sintesis. Gazdar, G. 1979. Pragmatics. New York: Academic Press. Gili Gaya, S. 1961. Curso Superior de Sintaxis Espanola. Barcelona: Vox Bibliograph. Goffman, Erving. 1981. Forms of Talk. Philadelphia: University of Pennsylvania Press. Gutierrez-Rexach, J. 1998. Semantica Logica y Cuantificacion Nominal. New Orleans: University Press of the South. Horn, L. 1972. On the Semantic Properties of Logical Operators in English. PhD diss. UCLA. Hornstem, N. 1977. "Toward a Theory of Tenses." Linguistic Inquiry 8. 521-557. Kaplan, D. 1989. "Demonstratives. An Essay on the Semantics, Logic, Metaphysics and Epistemology of Demonstratives and Other Indexicals." In J. Almog et al. (eds.) Themes from Kaplan. New York: Oxford University Press, 481-564. Keenan, E. 1993. "Natural Language, Sortal Reducibility and Generalized Quantifiers." Journal of Symbolic Logic 58.314-325. King, L. 1992. The Semantic Structure of Spanish, Amsterdam: John Benjamins. Kratzer, A. 1989. "An Investigation of the Lumps of Thought." Linguistics and Philosophy 12, 607-653. Lambrecht, K. 1994. Information Structure and Sentence Form. Cambridge: Cambridge University Press. Lujan, M 1980. Sintaxis y Semantica del Adjetivo. Madrid: Catedra. Manteca, A. 1981. Gramatica del Subjuntivo. Madrid: Catedra. Martin Vide, C. 1986-1987. Lenguajes Naturales y Lenguajes Formales I-XII. Barcelona: PPU. Martin Zorraquino, M.A. and E. Montolio. 1998. Los Marcadores del Discurso. Teoria y Analisis. Madrid: Arco. Mejias-Bikandi, E. 1996. "Space Accessibility and Mood in Spanish." In G. Fauconnier and E. Sweetser (eds.) Spaces, Worlds and Grammars. Chicago and London: University of Chicago Press. Moltmann, F. 1995. "Exception Sentences and Polyadic Quantification." Linguistics and Philosophy 18. 223-280 Montague, R. 1977. Elementos de Filosofia Formal. Madrid: Alianza. (Translated by D. Quesada). Moreno Cabrera, J.C. 1985. Logica Formal y Linguistica. Una Introduccion a la Gramatica de R.Montague. Madrid: Ediciones de la Universidad Autonoma. Moreno Cabrera, J. C. 1994. Curso Universitario de Linguisiica General. Vol.11. Madrid: Sintesis. Moreno Fernandez, F. 1998. Principios de Sociolinguistica y Sociologia del Lenguaje. Barcelona: Ariel. Moro Simpson, T. 1973. Semantica Filosofica: Problemas y Discusiones. Buenos Aires: Siglo XXI. Ojeda, A. 1991. Linguistic Individuals. Stanford: CSLI Publications. Portoles, J. 1998. Marcadores del Discurso. Barcelona: Ariel. Pustejovsky, J. 1995. The Generative Lexicon. Cambridge: Cambridge University Press. Quesada, D. 1985. La Logicay su Filosofia. Barcelona: Barcanova. Real Academia Espanola (RAE) 1973. Esbozo de una Nueva Gramatica de la Lengua Espanola. Madrid: EspasaCalpe. Reichenbach, H. 1947. Elements of Symbolic Logic. New York: MacMillan. Reyes, G. 1990.La Pragmdtica Linguistica. Barcelona: Montesinos Reyes, G. 1995. El Abece de la Pragmdtica. Madrid: Arco. Rivero, M. L. 1979. Estudios de Gramatica Generativa del Espanol, Madrid: Catedra, Sanchez de Zavala, V. 1974. Semantica y Sintaxis en la Linguistica Transformatoria. Madrid: Alianza. Sanchez de Zavala, V. 1994. Ensayos de la Palabray el Pensamiento. Madrid: Trotta. Sanchez de Zavala, V. 1997. Hacia la Pragmdtica (Psicologtca). Madrid: Visor. Schegloff, E., G. Jefferson, and H. Sacks, 1977. "The Preference for Self-Correction in the Organization of Repair in Conversation." Language 53. 361-382. Serrano, S. 1975. Elementos de Linguistica Matemdtica. Barcelona: Anagrama.
20
From Words to Discourse
Silva-Corvalan, C. 2001. Sociolingtiistica y Pragmdtica del Espanol. Washington, DC: Georgetown University Press. Smith, C.S. 1991. The Parameter of Aspect. Dordrecht: Kluwer Stalnaker, R. 1978. "Assertion." In P. Cole (ed.) Syntax and Semantics 9. 315-332. Stowell, T. 1993. "Syntax of Tense". Ms. UCLA. Stump, G. 1985. The Semantic Variability of Absolute Constructions. Dordrecht: Reidel. Suner, M. 1982. Syntax and Semantics of Spanish Presentational Sentence-Type Constructions. Washington, DC: Georgetown University Press. Valdes, L. M. (ed.) 1991. La Busqueda del Significado. Madrid: Tecnos. Wilson and Sperber (1988) "Mood and the Analysis of Non-declarative Sentences." In Dancy,Moravcsik and Taylor (eds.) Human Agency: Language, Duty and Value. Stanford, CA: Stanford University Press, 77-101.
21
2
SPANISH PAST AND FUTURE TENSES: LESS (SEMANTICS) IS MORE Silvia Gennari, Brown University
1. INTRODUCTION Spanish simple future and preterite can receive interpretations that are difficult to explain in terms of a single underlying lexical meaning. Consider the case of simple future. It can receive either a temporal or a modal interpretation (Gili Gaya, 1961). The temporal interpretation denotes a time later than the speech time (ST), as exemplified in (1). The modal interpretation conveys a probability or certainty judgment, what Gili Gaya (1961) calls future of probability, as in (2) and (3). Critically, this interpretation is often accompanied by a non-future temporal reading, i.e., an interpretation compatible with a situation in which the sentence is true at the ST: Temporal interpretation: (1) El presidente hablard par television. The president speak-fut. by television. "The president will speak on TV." Modal interpretation: (2) Acabo de ver las luces. Juan ya estara en casa. finish-pres. of see the lights. Juan already be-fut. in home. "I just saw the lights on. Juan will be at home already."
22
(3)
From Words to Discourse
El avion llegaba alas cuatro. Pedro ya estard en Kansas. The plain arrive-imperf. to the four. Pedro already be-fut. in Kansas. "The plain was arriving at 4. Pedro will already be in Kansas."
The modal interpretation is problematic for traditional accounts of the meaning of simple future that propose either a basic future meaning underlying all uses (King, 1992), or the existence of different "senses" composing the meaning of the tense (Gili Gaya, 1961). This is because these proposals fail to systematically explain when and why the non-future reading is available. The case of Spanish preterite presents a similar problem. It can receive a temporal reading together with an aspectual one. Temporally, preterito denotes a time before the ST, while in the aspectual domain, it indicates the completion of the action denoted. This aspectual component of preterito is usually contrasted with that of imperfecto, which does not require completion (Gili Gaya, 1966; Cipria and Roberts, 2000). However, this aspectual reading is not always available: preterito can but need not convey completion. For example, (4) entails that the letter is finished (confirming the completion interpretation), but (6) does not entail or convey the end of the sickness: Completion interpretation: (4) Juan escribio una carta. *Pero no la termino. Juan write-pret. a letter. But not it finish-pret. "Juan wrote a letter. *But he did not finish it." (5) Juan corrio esta manana. #Y todavia corre. Juan run-pret. this morning. And still run-pres. "Juan ran this morning. # And he is still running." Non-completion interpretation : (6) Juan estuvo enfermo ayer. Y todavia lo estd. Juan be-pret. sick yesterday. And still it be-pres. "Juan was sick. And he still is." (6) can receive an interpretation in which the same state referred by the first sentence continues into the present. This is incompatible with the postulation of a meaning for preterito that specifies the completion of the event before the ST. But note that this interpretation only arises under special circumstances. In contrast to the non-future reading of simple future, an isolated preterito sentence does not generate a reading in which the interpretation of the sentence continues into the present. In fact, in isolation, Juan estuvo enfermo ayer "Juan was sick yesterday" receives an interpretation in which Juan is no longer sick. It is the continuation of the second sentence in (6) that raises the possibility of the continuation into the present. These facts
Spanish Past and Future Tenses
23
are puzzling and pose a problem for most accounts, those that attribute to preterito both temporal and aspectual meanings (Bello, 1847; Gili Gaya, 1961; Cipria and Roberts, 2000) and those that attribute to it a simple past meaning (Gennari 1999b). In either case, the absence or presence of the completion reading would need to be systematically explained. Thus, for both simple future and preterito, there are readings that cannot be captured by a single underlying definition of the tense, no matter whether this definition is simple or complex (with one or two basic components). In this paper, I investigate the specific distributions of the temporal readings available for each simple future and preterito in order to identify what is common in all interpretations (i.e., part of the semantic lexical meaning) and what is a result of the context of occurrence. I claim that the correct allocation of the work of semantics and pragmatics can improve our understanding of the meanings and interpretations of these tenses. The basic idea of my proposal is that both simple future and preterito have relatively simple lexical meanings, but independent factors such as aktionsart and conversational implicatures explain the unexpected readings. The proposal is presented in two parts. First, I discuss the temporal interpretation of the tenses. I show that the aktionsart of the sentence involved determines the temporal reading. In particular, it determines whether the future tense receives a future or non-future temporal reading and whether preterito is compatible with a non-completion reading. Thus, the lexical meanings of the tenses need not include an especial specification to treat these readings. Second, I also discuss the distribution of the completion reading of preterito in detail. I argue that its meaning only contains past temporal information but that its completion reading follows from aktionsart facts and a pragmatic conversational implicature. Second, I discuss the modal interpretation of simple future. I show that different occurrences of simple future have a unique meaning with both temporal and modal components, where the modal component is defined as a function from the context of occurrence (as Kratzer 1977, 1979, 1981 has noted for other modals). For any given occurrence, both temporal and modal specifications are present so that there is no temporal use independent from a modal one. The temporal meaning is treated as quantification over future intervals, whereas the modal part of the meaning quantifies over worlds but requires the sentence to follow from the assumptions in the context of utterance (conversational background).
2. ACCOUNTING FOR THE TEMPORAL INTERPRETATIONS 2.1. Temporal Interpretation and Aktionsart The main question addressed in this section is why simple future can receive non-future readings and why preterito can receive non-completion readings. To understand this problem, one needs to look at aktionsart properties, since the temporal interpretation of these tenses is determined in
24
From Words to Discourse
predictable ways by the aktionsart of the sentence involved. Consider first the case of simple future: Stative sentences (7) Juan estard en casa ahora. Juan be-fut. at home now. "Juan will be at home now." (8) El cura sabrd mucho de teologia. The priest know-fut. much of theology. "The priest will know a lot of theology." Eventive sentences (9) Leerd el libra. / Conducird todo el dia. Read-fut. the book./ drive-flit, all the day. "He will read the book." / "He will drive all day long." Stative sentences are correlated with the possibility of non-future readings, although future readings are also possible. In contrast, event sentences (whether activities or accomplishments) are correlated with future readings. A similar situation obtains for the case of the preterito with reference to the non-completion reading: Stative sentences (10) Estuvo en casa esta manana. (Y todavia lo estd.) Be-pret. at home this morning. (And still it be-pres.) "He was at home this morning. (And he still is)." (11) Sospecho que ella lo enganaba. (Y todavia sospecha). Suspect-pret. that she him deceive-imperf. (And still suspect-pres.). "He suspected that she was deceiving him. He is still suspecting." Event sentences (12) Juan escribio una carta. (*Y todavia la estd escribiendo.) Juan write-pret. a letter. (*And still it be-pres. writing). "Juan wrote a letter. (*And he is still writing.)" (13) Camino en silencio. (#Y todavia estd caminando.) Walk-pret. in silence. (# And still be-pres. walking.) "He walked silently. (# And he is still walking)."
Spanish Past and Future Tenses
25
Stative sentences are compatible with an interpretation in which the past state continues into the present, while this interpretation is not possible with event sentences. Thus, for both simple future and preterito, stative sentences correlate with temporal interpretations that can overlap with the ST. This generalization is supported by independent facts. Discourse Representation Theory (e.g., Kamp and Reyle, 1993) and Dowty (1986) have pointed out that stative sentences, including genetics (Carlson and Pelletier, 1995) and progressives, typically generate overlapping readings in discourse contexts. In contrast, event sentences generate sequential readings. Consider some examples: (14) (15)
I went to Bill's office this morning. He was worried. I went to Bill's office this morning. We discussed the budget proposal.
Similarly, the temporal effects of imperfecto in Spanish and progressive in English are attributed to the properties that imperfective aspect shares with stative sentences (cf. Cipria and Roberts, 2000; Dowty, 1979, 1986; Gennari, 1999b). These properties explain why imperfecto and progressive correlate with overlapping readings in subordinate contexts, while simple tenses, which lack these properties, correlate with sequential readings: (16)
(17)
a. Juan dijo que Maria estaba viniendo. Juan say-pret. that Maria be-imperf. coming, b. Juan said that Maria was coming. a. Juan dijo que Maria vino. Juan say-pret. that Maria be-pret. coming, b. Juan said that Maria came.
The temporal effects of aktionsart properties are thus not restricted to simple future and preterito, or to interpretations relative to the ST. Rather, it is a general phenomenon that occurs with several tenses in both English and Spanish and relative to times other than the ST. But why does this phenomenon occur? Following a suggestion made by Dowty (1986), I have proposed that stative sentences generate overlapping readings because they trigger a temporal pragmatic implication: states imply that they are true at a larger interval that includes the interval (past or future) specified by the meaning of the tense. I call this implication the superinterval implication (Gennari, 1999a, 1999b). The intuition behind this implication is the following. If a certain state holds at interval /, one can infer that the state held for some time prior to / and will hold for some time later than /', i.e., that the state holds at a superinterval surrounding i. For example, if I am Argentinean now, you can infer that I was and I will be Argentinean for a while. This inference follows from what we know about states, their very nature. Note that states do not have internal temporal developments like events do, there is no transformation or internal
26
From Words to Discourse
causal structure. Something can cause a state to come into existence (as my being Argentinean was caused by my being born in Patagonia), but once you are in a certain state, it can persist in time until some event stops it. Events, in contrast, are contingent on causal forces acting upon them. This causal dependency makes them temporally bounded. They entail changes that go from an initial state to a final state. Accomplishments entail a single definite change, while activities entail a sequence of changes (Dowty, 1979). Such changes, due to their dependency on external forces, are not taken to go on indefinitely but to start and stop within the interval specified by the tense. Intuitively, if an activity or accomplishment is true at interval i, it does not necessarily lead us to conclude that the event holds at an interval earlier or later than /, since the causal force that applies at i may not obtain ac any other time. Thus, states, but not events, have the superinterval implication. The temporal effects of the superinterval implication can be clearly seen in the temporal interpretation of simple future: (18)
Juan estard en casa ahora. Juan be-fut. at home now. "Juan will be at home now." = 3z [i>st & be-home(j, i)]
According to the standard account of simple future, (18) is true iff there is a future interval / later than the ST at which Juan is home. Now, given the implication associated with states, i.e., that there is a superinterval i' that contains i at which Juan is home, it follows that both the ST and the interval of being home can overlap. This is exemplified in Figure 1 where the slashes represent the superinterval inferred around the future interval specified by the tense:
JJJJJ_\JJJJJJJJJ_IJ_\\\\\\\\\\I_IJJJJJ_ ST be-home Figure 1: schematic representation of the temporal reading of (3) and (18). Slashes represent the superinterval surrounding the future interval of being home.
Note that if the sentence was eventive, the event would be interpreted to happen within the future interval provided by the future tense. So, no overlapping reading would be possible. Note also that the superinterval implication does not necessarily trigger an overlapping reading. The context will help to determine whether such an inference is plausible. For example, John will be at home next week may not trigger the inference, because the state involved is not likely to hold true for such a long period. A similar reasoning applies for the case of the preterite in order to obtain the non-completion reading. The temporal information of the tense is such that it denotes an interval prior to the ST.
Spanish Past and Future Tenses
27
However, under special conditions such as those of (19), the state involved can be inferred to obtain at a larger interval so that this interval is able to overlap with the ST, as represented in Figure 2: (19)
Juan estuvo enfermo. (Y todavia lo estd). Juan be-pret. sick. (And still it is). "Juan was sick. And he still is." = 3i [i<st & be-sick(j, i)]
be-sick Figure 2: schematic representation of the temporal reading of (6) and (19). The slashes represent the superinterval surrounding the past interval of being sick.
Note that this account of the non-completion reading of preterito presupposes that the meaning of this tense is simply past. If the lexical semantic specification of the tense required the completion of the event/state involved, such a specification would contradict the pragmatic inference of a superinterval, and therefore, the superinterval inference as well as the reading of (19) would not be possible. The fact that this inference is possible suggests that the meaning of the preterito does not lexically specify aspectual information. Although there are several remaining questions regarding the interpretation of preterito discussed in the next section, for now it is fair to conclude that the superinterval implication explains why the non-future reading of simple future and the non-completion reading of preterito may arise with states but not with events. Therefore, it is possible to maintain that the semantic lexical specification of the simple future is always future and that of preterito is simply past. Independent aktionsart facts such as the superinterval implication of states will explain the problematic readings for each case.
2.2. Preterito and the Completion Reading In the previous section, I argue that the lexical meaning of preterito only specifies past temporal information because the non-completion reading is possible under certain conditions. This contrasts with Cipria and Roberts's (2000) proposal according to which preterito entails completion, i.e., its meaning specifies what they call the end-point requirement. But if preterito does not lexically specify the end-point requirement, why does preterito typically receive completion readings? And why does the non-completion reading arise in very specific circumstances? In this section, I address these questions by investigating when and why each possible reading is available. On the one hand, I show that the key to understanding the distribution of the completion readings is the type of aktionsart with which preterito combines.
28
From Words to Discourse
On the other hand, I argue that the non-completion reading arises from the cancellation of an otherwise present pragmatic inference, hence the specificity of the contexts in which it occurs. To start with, note that preterito inherits the aktionsart from the phrase it modifies (Cipria and Roberts, 2000). For example, if the verb phrase (VP) denotes an accomplishment or achievement, the result of combining the preterito + VP is an accomplishment or achievement sentence. The same is true for other type of events. This contrasts with imperfecto, which changes the basic VP aktionsart into atelic aktionsart, i.e., into an aktionsart that does not entail a definite change of state. Compare the following: (20)
(21)
Juan escribio una carta. *Pero no la termino. Juan write-pret. a letter. *But not it finish-pret. "Juan wrote a letter. But he did not finish." Juan escribia una carta. Pero no la termino. Juan write-imperf. a letter. But not it finish-pret. "Juan was writing a letter. But he did not finish."
While (20) entails that Juan finished writing the letter, (21) does not necessarily do so. Juan could have been writing a letter without having finished it. This clearly exemplifies the different aspectual effects of preterito and imperfecto. In contrast to imperfecto, preterito does not change the aktionsart of the phrase with which it combines. This means that whichever entailments the basic VP has are inherited by the preterito sentence. Note also that Dowty (1979), Taylor (1977), Verkuyl (1993) and many others have pointed out that accomplishments and achievements entail a change of state. This amounts to entailing the falsehood of the proposition in question at a certain interval i and its truth at an interval / ' later than i (see definition of BECOME in Dowty, 1979). For example, if you write a letter at i, this entails that the letter was not written before i, and that it was written immediately after i. The resulting state must therefore obtain if the accomplishment sentence is true. This explains why preterito sentences with achievements or accomplishments receive a completion interpretation: the event denoted finished within the interval specified by the tense. This also explains why the second sentence of (4) is contradictory: the entailment of this aktionsart is logically incompatible with the truth of the sentence at longer intervals or at the present. Another way to describe the judgment of (4) is in terms of discourse dynamics. This framework will help to characterize the differences in the completion interpretation between states and events. According to Stalnaker (1978), Heim (1988) and dynamic Montague semantics (Chierchia, 1995), every utterance context or conversational background (CB) is a set of propositions (i.e., set of possible worlds) shared by the participants of the conversation. This context updates as conversation proceeds. Every new sentence in discourse causes a change in the CB according to which some new set of propositions is generated. According to Stalnaker
Spanish Past and Future Tenses
29
(1978), this set is the intersection of the CB and the new proposition. As an example, consider a case given by Roberts (1989): (22)
A thief might break into the house. He would take the silverware.
Once the first sentence is added to the utterance context, it generates a set of propositions compatible with the actual situation (the intersection of the initial CB and the proposition expressed by the first sentence). The second sentence in the discourse is evaluated relative to this CB, i.e. relative to worlds in which a thief breaks into the house. The new CB is a set of worlds in which the two propositions of (18) are true. Thus, each new sentence in discourse causes a rearrangement in the CB. In our example (4), the first sentence introduces in the CB a given set of propositions compatible with the proposition expressed. One of these propositions is the entailment that the letter writing was finished at the past interval specified. When the second sentence comes in, the proposition that the letter was not finished logically contradicts that of the current CB. The intersection of the current CB and the new proposition is empty. This, according to Stalnaker (1878), is the characteristic of a self-defeating assertion. The interpreter cannot generate the new CB without concluding that either the first statement or the second one is false. Thus, the discourse is incoherent. Consider now the occurrence of preterite with activities: (23)
Juan bailo. ??Ytodavia estd bailando. Juan dance-pret. ??And still be-pres. dancing. "Juan danced. ?? And he is still dancing."
An interpretation in which there is a continuation of the activity into the present is not available. This is because activities in general have temporal implications that are inconsistent with their truth at intervals beyond the one specified by the tense. To see this, note that activities are normally assumed to be contained within the interval provided by the temporal operator. They typically imply that they take place within some arbitrary initial and end point of their interval of truth (Smith, 1991). Consider the following (where => means implies): (24)
Juan bailo esta manana. => Juan termino de bailar esta manana. Juan dance-pret. this morning. Juan finish-pret. of dance this morning. "Juan danced this morning. => He stopped dancing this morning."
30
From Words to Discourse
When activities occur with simple non-aspectual tenses, they are interpreted to occur within the boundaries specified by temporal adverbs.1 This stems from the fact that a quantity pattern of implicature is triggered by the grammatical elements of the sentence, in particular by the possible alternation betweenpreterito and imperfecto. Consider the following: (25)
(26)
Juan bailo. Juan dance-pret.. "Juan danced." = 3i [i < ST & dance(j, z)] Juan bailaba. Juan dance-imperf. "Juan was dancing." = 3i [i < ST & IMP [dance(j, i)]]
By the definition of imperfecto, (22) logically entails, rather than implicates, that there is a superinterval containing z during which the process of dancing took place (Gennari, 1999b). This is because, like the English progressive (see Dowty, 1979), imperfecto has both an aspectual and temporal component. The temporal component, as preterito, denotes an interval prior to the ST. The aspectual meaning introduces a superinterval surrounding the past interval provided by the past component and possible worlds that continue from this past interval on. IMP in (22), for example, is true iff for some superinterval containing the past interval i and for all worlds that are possible continuations of the actual one from i on, Juan is in the process of dancing. The intuition behind this definition is that it allows the event in question to be in progress at the given past interval, rather than being completed at this interval. This explains why (17) is compatible with a situation in which the letter was not finished and why the activity of (22) can keep obtaining until the ST.2 Sentence (22), thus, clearly contrasts with (21), which simply says that the process of dancing took place at a past interval z. This is represented below:
bailo
ST
Figure 3: schematic representation of the temporal reading o/imperfecto and preterito
1 This is in agreement with the fact that events are temporally bounded, as indicated earlier. However, this does not explain why (19) is not fine. There is nothing in our pragmatic world knowledge that can explain this judgment. Thus, the judgment arises for different reasons. 2 This also explains why the subordinate sentence of (15) overlaps with the time of the matrix: the past component of imperfecto denotes an interval prior to the matrix's interval, but the superinterval of imperfecto is able to overlap with it.
Spanish Past and Future Tenses
31
Imperfecto introduces a superinterval surrounding the relevant past interval. Since imperfecto conveys more temporal information, this is the type of situation that triggers Gricean quantity inferences due to the speaker's choice of preterito over imperfecto. For if the speaker does not choose imperfecto (the most informative form), he/she implicates that the superinterval does not obtain, i.e. that the continuation of the event into the present is not possible. This amounts to a completion reading, in which the event is interpreted as ended in the past.3 The imperfecto sentence is the most informative form because it can be true in more situations than the preterito sentence. In fact, the situation in which the preterito sentence is true is a subset of those in which the imperfecto sentence is true. Thus, when preterito occurs with activities, it yields completion readings due to a Gricean inference triggered by the grammatical elements of the sentence. I will call this Gricean inference the end-point implicature. That preterito activities implicate rather than entail completion is clear from the fact that the implicature in question can be cancelled, a typical characteristic of Gricean inferences (Levinson 1983). Consider the following discourse: (27)
(28)
El presidente hablo por television. En realidad, (creo que) todavia estd hablando. The president speak-pret. by television. In fact, (think, that) still be-pres. speaking. "The president spoke on TV. In fact, (I think) he is still speaking." Hoy llovio. En realidad, todavia no ha parado. Today, rain-pret. In fact, still not have stopped. "Today, it rained. In fact, it has not stopped yet."
Note that the first sentence in both discourses implies that the event in question is finished in the past. However, the second sentence cancels this implication because the speaker corrects the content of his/her statement. The judgments are similar to other cancellation cases. Consider for example, John has three kids. In fact, he has four. This case contrasts with the case of accomplishments such as that of (4) in that the second sentence of the discourse does not make the first one false. There is no logical contradiction involved. The context changes from a set of possible propositions compatible with the first statement to another set intersecting the first and second one (In (23), for example, this is the set of worlds in which the president speaks on TV for a while and continues to do so). These CB sets are different but not mutually exclusive. The cancellation takes place because the end-point inference is compatible with the first set but not the second one. Thus, the proposition corresponding to the end-point inference is dropped from the first CB. This confirms that the completion interpretation associated with preterito activities comes from a pragmatic implicature, rather than from a logical entailment. 3 Note that the pragmatic inference involved in this case is parallel to the quantity inference found between some and all. Some students are good (the less informative form) implicates that Not all student are good (the negation of the most informative form).
32
From Words to Discourse
Consider now the case of preterite occurring with states. Recall from our discussion of the previous section that states can receive an interpretation compatible with their continuation into the present. However, I also noted that state sentences, when occurring in isolation, do not suggest such continuations, as in (29): (29)
Juan estuvo enfermo. Juan be-pret. sick. "Juan was sick."
Sentence (29) seems to imply that the state of being sick ended in the past. This is because if an interpretation in which the possibility of continuation into the present was intended by the speaker, the speaker would have used the imperfecto. As with activities, stative preterite sentences in isolation trigger a quantity inference by contrast with imperfecto. This means that the superinterval implication of states discussed earlier does not arise. In (29) for example, the superinterval implication does not arise because this implication is logically incompatible with the pragmatic inference triggered in the current context. As it stands, it is reasonable for the interpreter to infer the end-point implicature from the speaker's statement, given the speaker's grammatical choice. However, a cancellation is possible and easily available. A continuation such as Y todavia lo estd "And he still is" would cancel the end-point inference and would raise the possibility that states can persist for longer periods than those specified by the speaker. The superinterval implication becomes compatible with the current CB and so becomes the noncompletion reading of the preterito. As with activities, this thus confirms that the end-point requirement is not en entailment but a pragmatic implication. Note that there is a difference between states and activities in the way the cancellation occurs. Activities can be cancelled in situations in which the speaker rectifies his/her assertion. In contrast, states do not require this because the superinterval implication makes the cancellation more accessible. Raising the possibility of a continuation into the present is enough to cancel the implicature. The answers to the questions posed at the beginning of this section are now clear. The completion reading of preterito can occur because of two main reasons: either there is an entailment of completion, in which case this reading always obtains, or there is a pragmatic inference by contrast with the imperfecto. The former occurs with accomplishments and achievements. The latter occurs with activities and states. However, the non-completion reading can arise in specific circumstances if the pragmatic inference is cancelled by incoming discourse.
Spanish Past and Future Tenses
33
3. THE MODAL INTERPRETATION OF SIMPLE FUTURE 3.1. Basic Assumptions In this section, I introduce Kratzer's (1979, 1981) account of modals necessary to understand the modal interpretation of simple future. First, context dependency: modals are context dependent. They are interpreted relative to the CB from which conclusions can be drawn. Formally, CBs are functions that apply to the utterance context and return the set of propositions in it. Thus, CB(st, w), where w is the actual world, is the set of propositions assumed at ST and w. Modals' context dependency is clear in the following English example: (30)
John must not be the student who failed the exam.
Intuitively, the speaker makes a conclusion that follows from the implicit assumptions taken for granted in the context of utterance. The meaning of must thus specifies that (30) is true if it follows from the CB that John is not the student who failed the exam. Types of CBs: different interpretations of modals depend on different kinds of CBs. Consider again (30). The CB can be deontic or epistemic and so can the interpretation: (30) can be uttered in a context where the speaker makes a conclusion from what he/she knows or in a situation in which the speaker (say, John's mother) makes a conclusion from the set of John's responsibilities assumed in the context. Each context thus generates a different interpretation. Restrictions on the CB: the lexical specification of modals may impose certain conditions on the CB in which they occur. For example, must cannot occur in every context, but requires either epistemic or deontic contexts: (30)
??If I was sick, I must be home.
Hypothetical counterfactual situations are not appropriate for the occurrence of must. One way of viewing this sort of restrictions on the context is to say that modals presuppose a certain set of worlds/propositions in their CB for them to be felicitous. (30) is thus infelicitous because the requirement of an epistemic or deontic context associated with must is not satisfied by the current CB, which contains only propositions compatible with the hypothetical situation of the //"-clause.
34
From Words to Discourse
3.2. Simple Future and CBs Consider first some examples: (31)
(32)
Juan estard en casa ahora. Juan be-fut. at home now. "Juan will be at home now." Juan vendrd. Juan come-fut. "Juan will come."
Note that in these cases, as with modals in general, there is a logical relation to the implicit CB. For example, (31) is true iff it follows from the CB that Juan will be at home. It is taken for granted that the speaker knows something from which he/she makes this inference, hence the inferential or probabilistic interpretation. In (32), the implicit inference is less obvious but is still there. In order to make a prediction about the future, the speaker assumes certain knowledge about Juan (his plans, intentions, etc.), otherwise the prediction could not be made. Thus, in all cases there is an inference relation with the implicit current CB. The future sentence is true iff it follows from the CB. Note also that there is something specific about these CBs. A certain kind of CB is required. In the same way that must requires epistemic or deontic contexts (but not, for example, context where a hypothetical world is assumed), simple future requires a realistic or factual CB. A realistic CB is such that propositions that are known true in the actual world are taken for granted. In the examples above, the future sentences follow from sets of propositions known to be facts. I take this to be a presupposition associated with the meaning of the tense. Realistic CBs are required for the future tense to be felicitous and to have a truth value relative to it. This explains why (33) and (34) are not felicitous: (33)
(34)
#Sifuera rica, comprare un auto. If be-subj. rich. Buy-fut-lper. a car. "If I was rich, I will buy a car." ?? Juan quiere venir pero no sabe sipuede. Vendrd el jueves. Juan want-pres. come but not know-pres. if can-pres. Come-fut. the Thursday. "John wants to come but he does not know whether he can. He will come on Thursday."
In (33), the jj-clause introduces a non-realistic hypothetical proposition into the CB so that the CB no longer contains a set of propositions known to be true in the actual world. It rather contains propositions compatible with the non-factual antecedent. In (34), the first sentence of
Spanish Past and Future Tenses
35
the discourse adds to the CB the proposition that John does not know whether he can come. This conflicts with the fact that the future sentence requires this knowledge in the CB to be felicitous. Thus, in both cases, the CB violates the presupposition associated with the tense and therefore the sentences are infelicitous.
4. FINAL PROPOSAL I am now in a position to give the formal semantic definitions and pragmatic requirements of simple future and preterite, according to the discussion above: (35)
(36)
-re/-rd[Q] = MOD[FUT[Q\] is true at ST and w iff it follows from the CB that FUT[Q] = 3i [i > ST& Q(i)], i.e., iff Vw' such that w' belongs to the CB, FUT[Q] is true at w'. Presupposition: CB is factual PRET[Q] = is true at ST iff 3i [i < ST & Q(i)}
(35) says that a sentence Q modified by the simple future is true in the actual world at the ST iff it follows from the (realistic) CB that there is time later than the ST at which Q is true, i.e., iff for all worlds w' in the realistic CB, there is a time later than the ST at which Q is true in w'. (36) says that a preterito sentence is simply true iff there is an interval prior to the ST at which the sentence is true. Note that in (35) the meaning of simple future is composed by a (presuppositionally restricted) modal operator, which quantifies over the worlds of the CB, and a temporal operator, which quantifies over intervals. The non-future reading arises due to the superinterval implication. Note also that there is no aspectual information in the meaning of preterito. The completion reading typically arises for different reasons. On the one hand, it arises with accomplishment-achievement sentences because they entail the resulting state of the event in question. On the other hand, it arises with activities and states because of the end-point pragmatic implicature. This implicature can be cancelled with states because the superinterval implication can generate an interpretation overlapping the ST. Activities lack such an implication so that cancellation depends on the speaker's rectification of his/her statement as in other cases of implicature cancellation. Thus, for all interpretations of these tenses, the pragmatic context of occurrence as well as the aktionsart involved play a role when generating a particular reading.
36
From Words to Discourse
5. CONCLUSIONS In this paper, I have addressed the problem of the multiple interpretations that simple future and preterite can receive. I have shown that because there are systematic restrictions on the temporal interpretations of the tenses, such interpretations cannot be accounted for by complicating the lexical meaning of the tenses. In contrast, I have argued that by including aktionsart and pragmatics in our model of temporal interpretation, we can simplify traditional lexical tense definitions, and consistently account for all possible readings. Thus, the data discussed here support an integrated semantic-and-pragmatic approach to Spanish tenses, and to temporal semantics in general.
REFERENCES Bello, A. 1847. Gramatica de la Lengua Espanola Destinadc al Uso de los Americanos. Madrid: EDAF(1984). Carlson, G. and Pelletier, F. 1995. The Generic Book. Chicago: University of Chicago Press. Chierchia, G. 1995. Dynamics of Meaning. Chicago: University of Chicago Press. Cipria and Roberts. 2000. "Spanish Imperfecto and Preterite: Truth Conditions and Aktionsart Effects in Situation Semantics." Natural Language Semantics 8. 297-347. Dowty, D. 1979. Word Meaning and Montague Grammar. Dordrecht: Reidel. Dowty, D. 1986. "The Effects of Aspectual Class on the Temporal Structure of Discourse: Semantics or Pragmatics?" Linguistics and Philosophy 9. 37-61. Gennari, S. 1999a. "Tense, Aktionsart and Sequence of Tense." Empirical Issues in Formal Syntax and Semantics ed. by Corblin F., Dobrovie-Sorin C., and Marandin J. vol. II, Berne: Peter Lang, 309-332. Gennari, S. 1999b. Tense Meaning and Temporal Interpretation. Ph.D. dissertation. Brown University. Gili Gaya, S. 1961. Curso Superior de Sintaxis Espanola. Barcelona: Vox Bibliograph. Heim, I. 1988. The Semantics of Definite and Indefinite Noun Phrases. New York: Garland Publishing Co. Kamp, H. and Reyle, U. 1993. From Discourse to Logic: Introduction to Model Theoretic Semantics oj Natural Language, Formal Logic and Discourse Representation Theory. Dordrecht: Reidel. King, L. 1992. The Semantic Structure of Spanish. Amsterdam: John Benjamins. Kratzer, A. 1977. "What "Must" and "Can" Can and Must Mean." Linguistics and Philosophy 1. 337-355. Kratzer, A. 1979. "Conditional Necessity and Possibility." Semantics from Different Points of View ed. by Baulerle, Schwarze and Stechow. Berlin: Springer-Verlag Heidelberg N.Y, 117-147. Kratzer, A. 1981. "The Notional Category of Modality." Words, Worlds and contexts ed. by Eikmeyer and Rieser (eds.). Berlin: De Gruyter. 38-74. Levinson, S. 1983. Pragmatics. New York: Cambridge University Press. Roberts, C. 1989. "Modal Subordination and Pronominal Anaphora in Discourse." Linguistics and Philosophy 12. 683-721. Smith, C. 1991. The Parameter of Aspect. Dordrecht: Reidel. Stalnaker, R. 1978. "Assertion." Syntax and Semantics 9. ed. by Cole P. New York: Academic Press, 315-332. Taylor, B. 1977. "Tense and Continuity." Linguistics and Philosophy 1. 199-220. Verkuyl, H. 1993. A theory ofAspectuality. Cambridge, UK: Cambridge University Press.
37
3
TENSED COMPLEMENTS OF PERCEPTION VERBS: ISSUES IN THEIR TEMPORAL INTERPRETATION Alicia B. Cipria, University of Alabama
1. INTRODUCTION Past tense complements are particularly interesting due to the different interpretations they receive when embedded under perception verbs. ' I argue that the temporal relations across a complex clause with a main perception verb are not as simple as claimed in some of the literature. Rather, the interaction of aspect/aktionsart and pragmatics with the lexical requirement of the main verb plays a role in the resultant temporal interpretations. Since perception verbs mostly take indicative complements, the issues discussed here bring interesting points to the general issue of "sequence of tense (s)", which in the Spanish literature has been dealt mostly within the context of subjunctive complements.
2. THE SENSORY AND COGNITIVE INTERPRETATIONS OF PERCEPTION VERBS We consider perception verbs expressions such as ver "to see", oir "to hear", palpar "feel by touch" and and oler "to smell"). Traditional accounts of perception verbs state that the complement verb must have a simultaneous relationship with the main verb of perception. In Most of theoretical assumptions, especially for aktionsart and truth conditions for the tenses are based on previous
38
From Words to Discourse
cases where this does not obtain, the main verb is interpreted as a cognitive rather than a sensory verb of perception.(cf. Gili Gaya, 1961; and RAE, 1973, to name but a few). Along similar lines, Suner and Padilla-Rivera (S&PR) (1987) point out that the complement verb has to match the tense specification of the main verb (i.e. present-present or past-past); otherwise, again, the meaning of the main verb changes to that of cognition. In this and other traditional accounts, there seems to be some confusion in the sense that it is being assumed that a simultaneous reading is equal to a combination of like tenses, i.e. it is assumed that whenever there is a combination of past-past and present-present, a simultaneous reading arises. This is true for some of the examples S&PR present (cf. (la-e)--although cf. discussion of (lb)) but not for the examples under (2): (1)
a. Vi que pasaban. I see-PRET that they go-iMPF by
(matching tense and sensory)
"I saw that they were going by." Or: "I watched as they went by." b. Vi quepasaron. (matching tense and sensory) I see-PRET that they go-PRET by "I saw that they went by." c. Veo quepasan. I see-PRES that they go-PRES by "I see that they go by/are going by." d. Vi que pasardn manana. I see-PRET that they go-FUT by tomorrow "I saw that they will go by tomorrow." e. Veo quepasaron.
(matching tense and sensory)
(different tense and cognitive)
(different tense and cognitive)
I see-PRES that they go-PRET by "I see that they went by." Thus, the argument is that (la-c) only permit combinations with past-past (as in (la,b)) and present-present (cf. (lc)), when ver is used as a verb of sensory perception. So this is taken to entail that main verbs of perception only admit tenses that allow a simultaneous reading, when used in the sensory sense. However, not all embedded past forms (under another past form) yield a simultaneous reading under a verb of perception, cf. (2b): (2)
a. OH que la comida se quemaba. I smell-PRET that the food burn-iMPF "I smelled that the food was burning."
work (Cipria (1996) and Cipria and Roberts (2000)).
(simultaneous)
Tensed Complements of Perception Verbs
b. OH que la comida se quemo. I smell-PRET that the food burn-PRET "I smelled that the food burned."
39
(non-simultaneous)
Both (2a) and (2b) have the same tense specification, viz. past. In (2a), the required simultaneous reading is satisfied but in (2b), with the preterite (which is also a past form), no simultaneous reading seems to arise. Expanding the context of (2a,b) shows the difference more clearly: (2)
a'. Oli que la comida se quemaba pero la pude salvar echdndole mas agua y aceite. "I smelled that the food was burning but I was able to save it by adding more water and oil." b'. #Oli que la comida se quemo pero la pude salvar echdndole mas agua y aceite. "I smelled that the food burned but I was able to save it by adding more water and oil."
(2b) can only be uttered after the food has burned, when the subject sees the situation resulting from the burning (i.e., the "smelling" (or detection of the smell) follows the "burning"). In (2b"), "saving the food" at the "smelling time" is futile. The subject of the main clause infers, by the smell, that the food has burned. The subject only sees the result of the burning and does not witness the actual burning. In this sense, que la comida se quemo ("that the food burned (PRET.)") is interpreted as prior to the "smelling", similar to something like que la comida se habia quemado ("that the food had burned"). Furthermore, not everything that is simultaneous or "matching tense" always yields a sensory interpretation, as shown in the following examples: (3)
a. Oi que entraban en la casa. (simultaneous and sensory) I hear-PRET that they enter-lMPF the house "I heard that they were entering the house/1 heard them entering the house." b. Oi que Maria estaba enferma. (simultaneous and cognitive) I hear-PRET that Maria be-lMPF sick "I heard that Maria was sick."
Both (3a,b) present the complement situation as simultaneous with the reference situation, i.e. the situation described by the imperfecto includes the event time associated with the main clause verb. And we also have the combination past-past in both examples. However, the main verb in (3a) is interpreted as a verb of sensory perception, while the main verb in (3b) is presented as a verb of cognitive perception. We believe that this discrepancy in the interpretation of the main verb is pragmatically motivated. The sensory interpretation does not arise in (3b) because the complement describes a situation that, from what we know about the real world, cannot be actually heard, i.e. we cannot "hear" the sounds of Maria "being sick". Rather, we can hear a
40
From Words to Discourse
report by someone of Maria's being sick. In this case, we can interpret (3b) as expressing something along the lines of "I learned that Maria was sick." Clearly, the difference in readings comes from a pragmatic rather than a semantic source, since both complement situations are simultaneous with the reference situation but only one of them (3a) has a sensory interpretation. We will return to these examples later but for now I am claiming that the interaction of aspect, aktionsart and pragmatics is responsible for the different readings associated with the complement tense. Before closing this section, though, some concepts need to be clarified. In the general literature on Spanish, the terms "sensory" and "cognitive" have been used to describe the interpretations arising in complex clauses with main verbs of perception. By sensory, it should be obvious that we understand those situations that can be directly perceived by the senses, where the subject of the clause with the perception verb actually witnesses a situation, visually, auditorily, or via smell or tact (this has also been called direct perception). Under the cognitive interpretation of perception verbs, evidence is acquired "indirectly" (also indirect perception), i.e. not by actual witnessing of the perceived situation but through hearsay, in the case of oir "hear", or through inference, in the case of ver "see" or sentir "feel, sense". As was seen in the case of oler (and also the case for palpar "to feel by touch") these verbs do not get the cognitive extension, i.e. they can be part of sensory contexts only. This is an important difference with the other verbs, ver/oir/sentir (henceforth, the "ver "see" set"), which always have a possible cognitive interpretation. This is due to historical reasons; some verbs of perception have undergone a semantic change from having a merely sensory meaning to acquiring an additional cognitive sense. There are many studies showing the universality of the cognitive extension for the latter set of verbs (cf. de Haan, 1999; Evans and Wilkins, 2000; Sweetser, 1990; and many others).
3. BACKGROUND ASSUMPTIONS 3.1 Aspect and Aktionsart There is a vast literature on how aspect and aktionsart contribute to the determination and limitation of reference time (a Reichenbachian notion) in temporal interpretation (cf. Hinrichs, 1981,1986; Partee, 1984; and Dowty, 1986; to name but a few). In the general linguistics literature, phenomena included under aspect and aktionsart have been treated in varied and disparate ways. And virtually every author who has written on the subject has her own set of definitions. Even though aspect and aktionsart are closely related, they are two separate notions. Aspect is viewed here as a morphological category, as it is expressed on verbs in Spanish by the use of the preterite and imperfecto morphemes. These morphemes express both aspect
Tensed Complements of Perception Verbs
4\
(perfective and imperfective, respectively)2 and past tense. Aktionsart, on the other hand, is a semantic notion, used here in connection with the propositions expressed by clauses. As we will show, the aspect on a given verb form will affect the overall aktionsart of a proposition in different ways. Thus, the resulting aktionsart of a given proposition (telic or atelic) will be the result of the interaction of the aspect on the verb with various NP complements, adverbial phrases, and so on. One of my main arguments is that the imperfecto will always entail atelicity while thepreterito will not necessarily entail telicity, as is shown further on. Speaking in very broad terms we can say that aktionsarten have been understood, as different classes of events (a la Vendler,1967; Kenny, 1963; and Dowty, 1979), e.g. whether they are stative (e.g. estar enfermo "be sick", gustar "to like") or not (e.g. llegar "to arrive", comprar "to buy"). Vendler's classification is the one which is more widely referred to in the literature, and it is a version of what had been known throughout the centuries as Aristotelian verb classes. Vendler (1967) classifies verbs into states (e.g. know), activities (e.g. drive a car), accomplishments (e.g. paint a picture) and achievements (e.g. recognize)? In this article, the term aktionsart is used with regard to the nature of eventualities correlated to propositions. The reason for considering propositions instead of just verbs or verb classes is the fact that it is not appropriate to evaluate aktionsarten based solely on verb types, given that the aktionsart of a proposition can end up being different from the one borne by the verb in isolated contexts. The aktionsart correlated to a proposition is also influenced by the presence of certain complements, modifiers like adverbial phrases (for example, in an hour vs. for an hour), adverbs of quantification like everyday or, in the case of Spanish, imperfecto/preterito markers. So imperfecto and preterite markers are just one of the elements involved in the overall aktionsart associated with a clause. This is the reason why aspect and aktionsart need to be independently defined. There are two major classes of aktionsarten correlated to propositions: telic and atelic.4 The classification used here is that of Dowty (1987). Atelic aktionsarten may be states (saber "know", querer "want", vivir "live") or processes (Vendler's activities, e.g. correr "run", Hover "rain", escuchar "listen"). In general, telic situations involve the achievement of a goal or some resulting state; they may be simple (e.g. win the contest) (Vendler's achievements} or complex (e.g. write the dissertation) (Vendler's accomplishments). There is one property that centrally distinguishes the atelic from the telic aktionsarten, the "sub-interval property": informally, we can say that if a state/process holds at some interval of time then it also holds at any sub-interval of that interval, so that, for example, if it is true at an interval of an hour that I know something, I also know it at any sub-interval of that hour. Also, its truth at the hour-long interval does not exclude the possibility that there may also be a super" Also called "grammatical aspect". Smith (1983, 1997) calls it "viewpoint aspect". 3 I follow Dowty (1979: 53-54) for the reference to this work and examples therein. Vendler's verb types have also been grouped as types of aktionsarten, inherent lexical aspect; Smith (1983, 1991) groups them under types of "situation aspect". 4 These terms were first introduced by Garey (1957). "Telic" comes from Ancient Greek telos "end".
42
From Words to Discourse
interval, say of two hours, during which the same state or process is true. A formal definition of this property is found in Dowty (1987:18): The Sub-Interval Property for Atelic aktionsarten:5 If 6 is an atelic predicate, then necessarily, 5(xi ,...,xn) is true for interval /if and only if 5(xi ,...,xn) is true for all subintervals /' of/. The predicted entailments of the sub-interval property are illustrated for the process of running in (4), where (4a) entails (4b). Similarly, this permits us to say that if Jaimito runs during the period from four to five and also during the period from five to six, it is also true that he runs from four till six. (4)
a. Jaimito corrio de 4 a 5. Jaimito run-PRET. from 4 to 5. "Jaimito ran from 4 till 5." b. Jaimito corrio de 4 a 4-4:30. Jaimito run-PRET. from 4 to 4:30. "Jaimito was running from 4 till 4:30."
Telic aktionsarten do not have this subinterval property, and in fact if a telic event is true at an interval, none of its proper subintervals will verify an instance of the same type of event. Thus, we can only say that a telic like (5) is true at the maximal interval which it took for Jaimito to write the poem in question; if (5) is true at an interval i, we cannot say that it is true at any subintervals of /, at least for the same poem. From this also follows that if (5) is true of the interval from 4 to 5 and then again of the interval from 5 to 6, it is not true at the interval from 4 to 6, though we could say of the larger interval that it is true there that Jaimito wrote two poems. (5)
Jaimito escribio unpoema. Jaimito write-PRET a poem. "Jaimito wrote a poem"
(telic)
5 Another version, with situations instead of intervals is developed in Cipria and Roberts (2000). Briefly, in order to evaluate a proposition with respect to intervals (for the temporal dimension) we would need an extra index of evaluation, e.g. worlds. Situations have temporal and spatial dimensions directly as part of them. Cipria and Roberts' definitions include more technical assumptions than we are able to provide here. For the purposes of understanding atelicity the version with intervals should suffice. In what follows then, we simply use situations when talking about temporal interpretation.
Tensed Complements of Perception Verbs
43
Compare Dowty's (1987:17) definition of telicity: If 5 is a telic predicate, then the truth of 5 (xl,...,xn) for interval /entails that 5 (xl,...,xn) is false for all proper subintervals /' of/. As expressed before, telicity and atelicity are not properties of verbs alone. Rather, the presence of certain argument NP's, adverbial phrases, or aspectual markers on a verb can yield a different aktionsart from the one suggested by the verb in isolation. In the case of Spanish, the use of the imperfecto will always entail atelicity while the preterite will not necessarily entail telicity. We argue that this follows from the truth conditions associated with each of these verbal aspects, so that in the case of the imperfecto, atelicity is built-in, as it were, in its core meaning. To see how non-verbal elements contribute to aktionsart, observe that in English sentences with a simple past tense verb, an argument NP which is a measure phrase (i.e. whose head is a count noun) will often yield telic aktionsart for the proposition, whereas a mass NP will bring about atelicity; this is illustrated in (6a) and (6b): (6)
a. Oil flowed from the tank. b. 3000 liters of oil flowed from the tank.
(atelic) (telic)
Facts are different for Spanish (given the preterito/imperfecto distinction). Thus, example (7) with an imperfecto verb and the count NP 3000 litros de petroleo "3000 liters of oil" has the same aktionsart (atelic) as the proposition in (8), with the mass noun petroleo "oil": (7)
(8)
Corrian 3000 litros de petroleo por las canerias. flow-3PL.lMPF 3000 liters of oil through the pipes "3000 liters of oil flowed through the pipes." Corria petroleo por las canerias. flow-3sG.lMPF oil through the pipes "Oil flowed/was flowing through the pipes."
(atelic)
(atelic)
Examples (6a), (7) and (8) are atelic because it is true to say of any given subinterval of the flowing process that oil was flowing at that subinterval.6 Also, (7) and (8) have a habitual reading, even without an adverb such as diariamente/por dia "daily/per day". But the preterito is, again, compatible with either aktionsart: In (9), with the preterito and the mass NP argument
6
To imagine when (7) would be true, suppose we have a circular pipe that can hold exactly 3000 liters and oil keeps flowing around in it continuously; then it is true for a given subinterval, that "3000 liters of oil were flowing" during that subinterval.
44
From Words to Discourse
petroleo, the proposition has atelic aktionsart (cf. (6a). The measure phrase in (10) interacts with the end-point requirement of \hQpreterito to entail telicity: (9)
(10)
Corrio petroleo par las canerias. flow-3SG.PRET oil through the pipes "Oil flowed through the pipes." Carrieron 3000 litros de petroleo par las canerias. flow-3PLU.PRET 3000 liters of oil through the pipes "3000 liters of oil flowed through the pipes."
(atelic)
(telic) (can also be iterative)
(6b) and (10) are telic because if it is true at some past interval t that 3000 liters of oil flowed from the tank at i, then it is not true that 3000 liters of oil flowed from the tank at any subinterval of /; instead only some part of the 3000 liters flowed during any subinterval of i. Like (6a), (7) and (8), example (9) has an atelic reading because it is true to say of any given subinterval of the flowing process that oil was flowing at that subinterval. Another well-known fact of aktionsart effects is that an endpoint adverbial such as to the store/a la tienda may similarly suggest telic aktionsart, as illustrated in (11-12), but this effect is overridden by the imperfecto in (13), which has only an atelic (habitual or progressive) interpretation: (11) (12)
Juana ran to the store. a. Juana corrio a la tienda. Juana run.PRET to the store "Juana ran to the store." b. Juana corria a la tienda. Juana run.iMPF to the store "Juana ran/was running to the store."
(telic)
(telic)
(atelic: habitual or progressive)
As can be verified with examples (7, 8,12b), the use of the imperfecto, whatever the predicate or complements, will always result in atelicity for the whole clause. We also mentioned the subinterval (sub-situation) property as the central element of atelicity, which in the case of the imperfecto is part of its core meaning.
3.2. Meanings of Imperfecto and Preterito In order to account for the traditional characterizations of the imperfecto, our formal apparatus includes a modal accessibility relation which, intuitively, chooses between the progressive (or
Tensed Complements of Perception Verbs
45
durative), habitual, and futurate or "intention-in the-past" meanings of the imperfecto, while keeping the atelicity constant (cf. Cipria and Roberts, 2000). Atelicity or the subinterval property is an element shared by all three of the usual imperfecto meanings. A sentence like (13) can have all three readings, depending on contextual variations, compare (14a-c): (13)
Ibamos alaplaya. go-lPLU.iMPF to the beach "We went/ were going /used to go to the beach."
The temporal adverbial clause in (14a) brings out a progressive reading; the adverbial los domingos in (14b) evokes a habitual reading; in (14c) the "intention-in-the-past" reading is clearly available: (14)
a.
b.
c.
Ibamos alaplaya cuando nos encontramos con Miguel. go-lPLU.iMPF to the beach when RECIPR. meet-IPLU.PRET with Miguel "We were going to the beach when we ran into Miguel." Ibamos a la play a los domingos. go-lPLU.iMPF to the beach on Sundays "We went/used to go to the beach on Sundays." Hastaayer, ibamos alaplaya devacaciones. until yesterday go-lPLU.iMPF to the beach on vacation pero hoy Pepa dijo que no hay dinero para eso. but today Pepa say-PRET that not there is money for that "Up until yesterday we were going to the beach on vacation but today Pepa said that there is no money for that."
All the meanings that have been illustrated above for the imperfecto share two characteristics. First, they involve reference to a past time. Second, they display atelic aktionsart. The examples in (14), with the imperfecto, are all atelic. So, in (14a) the subsituations of the event of "going to the beach" are instances of "going to the beach" as well. (14b) says that it was a habit of ours to go the beach on any normal Sunday during some past period of time, and this habit held true also for any sub-period of that period. With the intentional reading illustrated by (14c), the intention to go to the beach holds over some past interval, and also, then, over any sub-interval of that interval. Cipria and Roberts (2000) capture these facts with truth conditions for the imperfecto that involve the sub-situation property as its core meaning, with three sub-cases mediated by a modal interpretation of the imperfecto. Roughly and informally, modal accessibility relations permit for "choosing" between the three contexts.
46
From Words to Discourse
In connection with the intentional or futurate reading, we consider it a sub-case of the progressive interpretation, which, in turn, is related to the notion of "inertia". This notion of inertia has been used in the literature (cf. Dowty's, 1979 "inertia world" for the English progressive; Portner's , 1994 "inertia events"; and Cipria and Roberts, 2000) to capture the fact that out of the different outcomes available for the situation we are considering, there is at least one way in which things continue to be the way they are as in the situation of evaluation, i.e. if no changes occur. If we include the period during which one holds intentions to perform some act as part of the preparatory phase of an extended event,7 then the extended event is in progress during the preparatory phase, during the period when one holds those intentions. And if one's intentions were carried out as planned, then in all the inertia situations corresponding to that period the event itself would come about.8 As for the meaning of the preterito there is no need to make reference to a subinterval property, since we saw that this form is compatible with either telic or atelic aktionsart. The only crucial element in the meaning of the preterito is an end-point requirement (either to indicate culmination or termination, in the case of atelic states or processes). The preterito allows past truth in super or sub-situations (in the case of atelic readings) but all these situations must end at a certain point in the past (relative to Speech Time). Intuitively, the end-point is the final moment of the largest possible situation, ruling out the possibility of on-going situations. Considering the above semantics for the tenses we turn now to explore their behavior in complement clauses.
3.3. General Tense Interpretation of Indicative Complements This section is devoted to main verbs that do not require any particular temporal reference of the situation associated with their complements. We assume that the complement tense describes a situation which is evaluated with respect to the event time corresponding to the main clause verb. This event time serves as part of the reference situation with respect to which the complement situation will be evaluated, i.e. we will examine whether the complement situation is interpreted as prior (past), as simultaneous with or as later than the event time which provides
7 8
See Moens and Steedman (1988) for the notion of preparatory phase of an event. Note that there need not be any intention associated with the preparatory phase, as can be seen in the following:
(i)
Eran las 6. be.IMPF 6:00 El sol se The sun 3-REFL "It was 6 o'clock.
Los campesinos the peasants ponia set.lMPF The peasants started
comenzaron a preparar elfuego. start.PRET to prepare the fire a las 6:50. at 6:50 to prepare the fires. The sun would set at 6:50."
Tensed Complements of Perception Verbs
47
the reference situation. The temporal extension of the reference situation in complex clauses is the event time corresponding to the main clause verb. We will call this event time ET1. (15)
a.
b.
c.
Juana se entero de que Maria estuvo enferma. Juana find out-PRET that Maria be-PRET sick "Juana found out that Maria was sick." Juana se entero de que Maria se caso. Juana find out-PRET that Maria get-PRET married "Juana found out that Maria got married." Juana se entero de que el grupo organize complots. Juana find out-PRET that the group organize-PRET plots "Juana found out that the group organized plots."
In all of the examples above, the situation correlated to the complement is described in the preterito and interpreted as prior to the event time of the situation associated with the main clause. Of course, both main and complement tenses are past so that they are related to situations which lie before speech time. However, the complement tense further "shifts back" and it goes on to describe a situation which is interpreted as past relative to the (past) situation associated with the main clause tense. The readings that arise in these instances will be referred to here as the "backward shifted" readings,9 i.e. when we interpret the situation described by the complement tense as completely prior to time provided by the reference situation. The phrase "completely prior" is key in understanding the interpretation of preterito complements. In embedded preterito clauses, then, there will not be any final sub-situation of the situation associated with the complement (i.e. the situation which is past with respect to the main clause situation) that could possibly be taken to be somehow simultaneous or overlapping with ET1. The end-point element in the meaning of the preterito entails that there is no final sub-situation that could overlap with any other situation or that could be taken as the initial sub-situation of another situation. Imperfecto complements present different facts from the ones arising with preterito complements. Compare (16a-c): (16)
a.
b.
9
Juana se entero de que Maria estaba enferma. Juana find out-PRET that Maria be-lMPF sick "Juana found out that Maria was sick." Juana se entero de que Maria se casaba. Juana find out-PRET that Maria get-iMPF married "Juana found out that Maria was getting married."
The terms "backward shifted" and "forward shifted" were first mentioned in En9 (1987).
48
From Words to Discourse
c.
Juana se entero de que el grupo organizaba complots. Juana find out-PRET that the group organize-lMPF plots "Juana found out that the group organized/was organizing plots."
First, the imperfecto being a past tense, it describes a situation which is past relative to speech time. Second, unlike the examples in (15), none of the complements in (16) describe situations which lie completely in the past with respect to the event time of the main clause tense (ET1). In (16a), the complement describes a situation which is interpreted as "simultaneous" with the event time of the main clause. More specifically, the complement situation includes the event time associated with the main clause tense, this being often described in the literature as "surrounding" or "overlapping" said event time.10 In this case, the situation correlated to the main clause appears as punctual, when compared to the more extended situation related to the complement clause. Thus, in (16a), the state of Maria's being sick started at some time before the main clause event time (ET1) and it may continue to hold during and after ET1. The complement in (16b) describes either: a) a situation in which the actual event of Maria's getting married will occur (if at all) at a time later than ET1 (i.e. the intentional or futurate reading of the imperfecto), or b) a situation where the finding out takes place while the wedding is going on; i.e. a situation simultaneous with the event time of the main clause. In the intentional reading it is not specified whether the actual wedding will take place before or after the speech time, but the actual intention or plan can be interpreted as arising at some time prior to the speech time (cf. discussion of inertia situations).11 In (16c), the complement is correlated to a past habitual situation, i.e. a habitual situation which is past with respect to the speech time. However, this habitual situation cannot be interpreted as completely prior to the event time of the main clause (ET1). At ET1, the situation of "plot-organizing" was still true, even though it may not have been going on at the exact time This description of the effect that a situation described by the imperfecto has with a related situation described by the preterite (i.e. that of overlapping a more punctual situation) appears everywhere in traditional descriptions of Spanish and also in Spanish textbooks. This effect can be seen in any kind of complex clause and also in discourse. Related to this is the description of the imperfecto by the traditional grammarian Andres Bello (1945, first published in 1847), who calls the imperfecto the co-preterito. By this he means, we believe, that the imperfecto can describe a situation which is simultaneous (co-extensive) with a situation described by the preterito and also describe a situation as having started sometime before the event time corresponding to the preterito and even continue forward beyond said event time. 1 ' The intentional reading of the imperfecto arises only with dynamic predicates, i.e. non-states (in the appropriate context). Compare (i): (i) Juana dijo que Mario conocia todos los rincones de la ciudad. Juana say-PRET that Mario know-lMPF all the comers of the city. "Juana said that Mario knew every corner in the city."
In (i) the situation described by the complement, which has a stative verb in the imperfecto, includes ET1 and only has a simultaneous reading with the reference situation. A futurate or intentional reading is not possible in this case.
Tensed Complements of Perception Verbs
49
of ET1. This is so given that, as we have seen, what we understand by habitual situations is the notion of a characteristic situation which does not need to actually happen at the reference time (here, the event time corresponding to the main clause, ET1), as long as the situation can be said to be typical of a period or an era (i.e. a sufficiently large situation) which contains said reference time. Thus, whenever we have a habitual situation described by the imperfecto in the complement to another past tense, the habitual situation receives a sort of "simultaneous-like" reading, since the event described by the imperfecto complement tense, here, the "plotorganizing", is understood as taking place before ET1 and it continues to be a typical situation during and after that event time. There is one common element in all of the examples under (16): all the available readings (habitual, progressive, intentional) can be seen as variations of some sort of a simultaneous reading, where the situation described by the complement imperfecto overlaps at some point with ET1. In the intentional reading, the part corresponding to the preparatory phase can be considered as simultaneous with ET1. Recall that with preterito complements embedded under another past we only had a backward shifted reading. Now the reason why the imperfecto allows for the different variations is related to its inherent atelicity and the lack of a definite end-point. The subsituation property (the essential element in atelicity) makes it possible for the imperfecto to be true for a situation and also for its subsituations. This, in turn, can also make the imperfecto be true at larger situations (supersituations) which include the actual situation at which truth is asserted. This produces an effect such that the complement imperfecto includes ET1, yielding simultaneous readings, while at the same time allowing for the possibility that some of the supersituations of the situation described by the imperfecto can be taken to obtain also at some time prior to ET1. This is compounded by the fact that the imperfecto lacks a specific end-point so that the subsituations or supersituations do not have to end at any specific point. In connection with the preterito we showed that this form is compatible with either telic or atelic aktionsart. We would then have to explain why it is the case that even when the preterito can describe atelic situations we do not get readings in which possible supersituations can overlap ET1. The reason is related to the end-point present in its truth conditions, which requires that all the possible supersituations in which the preterito is true have to come to an end at some time. In the case of embedded contexts, this requirement is seen in the fact that the supersituations in which the preterito could be true have to end before ET1 (or at most end at ET1), without the possibility of extending forward beyond ET1. Because this backward shifting resembles the behavior of relative tenses (in this case, the past perfect), I have called it the "Relative Embedding Effect" (Cipria, 1996). For all of the examples in (15) and (16), similar interpretations for the complements obtain with such main verbs as dijo/decia que "said (PRET./IMPF.) that"; creyo/creia que "believed (IMPF./PRET.) that", and supo/sabia que "knew (PRET./IMPF.) that".
50
From Words to Discourse
4. TENSE INTERPRETATION OF THE COMPLEMENTS OF PERCEPTION VERBS As illustrated earlier, perception main verbs have some kind of lexical requirement on the tense that is used for the complement, in order to yield the desired interpretation, viz. sensory or cognitive. As shown in section 1, the analysis of temporal relations across clauses does not seem to be as easy as equating tense = simultaneous, or sensory = simultaneous, as is claimed in much of the literature. We take the default case for tense interpretation to be that of the verbs in section 2.3, with main verbs that have no specific temporal requirement for the situation associated with its complement. The temporal interpretation of the complement of perception verbs is shown not to follow the same pattern as was indicated for the verbs in section 2.3. This difference, of course, follows from the type of main verb that is involved. In addition, we believe that the different readings arise due to aspect/aktionsart effects and pragmatics, restricted by how we interpret the perception, i.e. sensory or cognitive.
4.1 Aktionsart Effects in Complement Clauses Recall that we argue that imperfecto always entails atelicity for its clause, regardless of other elements or type of predicate, while the preterito can participate in telic or atelic situations. Thus, (17a,b) and (18a) have an atelic complement clause: (17)
a. Oi que Claudia escribia en la computadora. I hear-PRET that Claudia write-lMPF on the computer "I heard that Claudia was writing on the computer." b. Oi que Claudia escribio en la computadora. I hear-PRET that Claudia write-PRET. on the computer "I heard that Claudia wrote on the computer."
In both (17a,b), the situation associated with the complement is interpreted as simultaneous with ET1 and both complex clauses receive a sensory reading.12 In (18a), the complement situation is interpreted as simultaneous with ET1 but it does not have the sensory reading, that is, it is a case of a "hearsay" type report:
{2 (17a) can have a possible "hearsay" reading when interpreted habitually, the atelic characteristic, of course, still applies.
Tensed Complements of Perception Verbs
(18)
5\
a. Oi que Claudia escribia un libra sobre los drusos. I hear-PRET that Claudia write-lMPF a book about the Drusseans "I heard that Claudia was writing a book about the Drusseans." b. Oi que Claudia escribio un libro sobre los drusos. I hear-PRET that Claudia write-PRET a book about the Drusseans "I heard that Claudia wrote a book about the Drusseans."
In (18b), then, we get a complement temporal reading that arises out of the interaction of the end-point of the preterito with the meaning of the predicate (aktionsart effect), which presents the situation described by the complement as a telic event (i.e. an accomplishment). In this case, (18b) can be uttered only after the book has been written, when the subject hears about the results of the writing (i.e. the "hearing" follows the "writing"), therefore the main verb seems to acquire the cognitive interpretation. In (18a), however, because of the use of the imperfecto (which, recall, always entails atelicity), the situation associated with the complement is presented as an atelic process (or activity). Therefore, the feature "past" alone does not account for the differences in readings. The simultaneous reading, then, is only possible with the one past form whose truth conditions say that a simultaneous reading is one of the available interpretations this form can have, that is to say, the imperfecto. The truth conditions for the other past form (the preterito) do not suggest this possibility of a simultaneous interpretation. Compare a similar example to (18b): (19)
Vi que Claudia paso. I see-FRET that C. pass by-PRET. "I saw that Claudia went by."
In relation to this, we argue that example (19), with an embedded preterito, does not present a true simultaneous reading for the complement. The complement crucially includes reference to the state resulting from the "going by". In other words, the subject witnesses the result of "Claudia's going by", after "the going by" has ended, most likely immediately after. Thus, the complement is related to an immediately resulting state, which is what the subject reports, rather than the actual passing or going by. This resultant state, by virtue of its being a state, brings atelicity to the complement clause, which may more easily (when compared with (18b)) give the impression that the complement situation is simultaneous with ET1. This is so given that the resultant state, because of its atelicity may refer to a supersituation of the "going by" which, intuitively, stretches to include ET1; although the implication that it can extend forward beyond ET1 is not available for (19). We argue that the preterito complement in (19) still participates in the Relative Embedding Effect that we attributed to all preterito verbs embedded under another past tense, so we have a
52
From Words to Discourse
backward shifted reading for the preterito complement. Therefore, this reading is in line with all the others illustrated in (15), where the Relative Embedding Effect of the preterito takes place. As mentioned earlier, in connection with the Relative Embedding Effect, a situation described by a preterito complement backshifts and this is due to the fact that the end-point of the preterito requires that, even in atelic cases, all subsituations and supersituations must end at some point. Therefore the situation described by the preterito ends at some point before ET1 or at ET1. We believe that the latter is the case for (19). In a true simultaneous reading, as is the case of, for example, (la) and (17a), no connection with a resultant state need be established, and the situations described by the complements are in progress at ET1 and can also extend forward beyond ET1. At the same time, there is a cognitive reading available for (19), and the relative embedding effect is seeing more clearly under this interpretation, i.e. the "seeing" follows the actual "passing by". Thus, we can say that for (19) the embedded preterito is ambiguous between a sensory and a cognitive interpretation. This state of affairs is different from the one we encounter with (18b), with a completely backward shifted reading. The difference, then, seems to lie in the (a)telic character of the complement clause, in combination with the end-point of the preterito. One might think of a possible counterexample to this generalization backward shifted/telic and simultaneous/atelic: (20)
Vi que los nifios construyeron un castillo en 5 minutos. I see-PRET that the children build-PRET a castle in 5 minutes "I saw that the children built a castle in 5 minutes."
The complement situation is telic (an accomplishment), through the interaction of the end-point of the preterito with a count NP direct object and the adverbial en 5 minutos "in 5 minutes", a usual trigger for telicity. Still, the complement situation does not seem to receive a backward shifted interpretation. The truth of (20) is not possible unless the result of the telic eventuality construir un castillo "build a castle", that is, the actual product, viz. the castle exists (cf. discussion of escribir un poema "write a poem"). But compare the same clause, without the adverbial: (21)
Vi que los ninos construyeron un castillo. I see-PRET that the children build-PRET a castle "I saw that the children built a castle."
(21) is ambiguous between a simultaneous interpretation, so that the sensory reading becomes available (i.e. the actual witnessing of the complement situation), and a backward shifted reading with a cognitive interpretation, in the sense of "I see a newly built (sand) castle therefore I gather the children built the castle."
Tensed Complements of Perception Verbs
53
I claim that the non-availability of a backward shifted reading for (21) is simply a pragmatic effect, triggered by the brief nature of "5 minutes" and the possibility for a sand castle to be built in a short period of time. If the children had built an extremely fancy castle with Lego's, then we could truthfully utter (22): (22)
Vi que los ninos construyeron un Castillo en 10 horas. I see-FRET that the children build-PRET a castle in 10 hours "I saw that the children built a castle in 10 hours."
From what we know about the real world and people in it and their activities, there is not a pragmatically viable situation in which the subject witnesses or "sees" the entire eventuality of "building a castle in 10 hours". Facts like these just reinforce the importance of pragmatic effects in the available temporal interpretations. Still, I believe we can safely claim that the following generalization holds: if the aktionsart of the complement clause is telic, a backward shifted reading arises, whereby the complement situation is interpreted as prior to ET1. If the aktionsart of the complement clause is atelic, then, the simultaneous reading is possible. For a clearer example to this generalization, without a possible cognitive reading, consider: (23)
a. Palpe
que la pintura se salia.
I touch-FRET that the paint come off-lMPF. "I felt (by touch) that the paint was coming off." b. Palpe que la pintura se salio. I touch-PRET that the paint come off-PRET. "I felt (by touch) that the paint came off." These examples have a "true" sensory perception main verb, in the sense that palpar "to feel by touch" is exclusively sensory, i.e. it cannot have a cognitive interpretation. So in (23a) the "coming off of the paint was witnessed. In (23b), however, with a preterito complement, the subject "feels by touch" that the paint has already come off, i.e. the results of the "coming off are witnessed, rather than. This is another example of a "seemingly simultaneous", where what the subject actually witnesses is the results. Notice that the aktionart of the complement clause in (23b) is telic (describing an achievement), while the complement situation in (23a) is atelic (describing a process, the use of the imperfecto makes the whole clause atelic). The import of aktionsart effects in perception contexts lies in the fact that simultaneous readings (be they sensory or cognitive) only arise when the aktionsart of the complement clause is atelic, while there will be backward shifted readings when the complement clause displays telic aktionsart (cf. (19), atelic, and (23b), telic). The fact that the preterito is compatible with both telic and atelic aktionsart, gives rise to interesting consequences in perception contexts.
54
From Words to Discourse
Recall that with complements that have no lexical requirements on the temporal interpretation of their complements (cf. section 2.3), the backshifting of the preterito always takes place, regardless of the aktionsart of the complement clause, but, under a verb of sensory perception, the options for the preterito are different. At this point, it should be clear that simultaneity is necessary for the sensory interpretation to arise and that, in turn, only atelic situations permit simultaneous readings. So we are assuming, for the moment, that atelicity and sensory contexts are related. Now, under perception verbs, the preterito backshifts when in a telic situation, and it may describe a simultaneous (or seemingly simultaneous situation) when in an atelic situation (again, only backshifting occurs with main verbs that impose no lexical requirement on their complements). Consequently, there is a clash between the lexical requirement of the main verb of sensory perception (for simultaneity) and the preterito embedding effects. I suggest this to be the source for the ambiguities in examples such as (19) and (21) with a preterito complement, where both a simultaneous (or even a "seemingly simultaneous") interpretation and a backward shifted interpretation are available. Notice, again, that this ambiguity for the preterito arises with atelic complements. As we see in the next section, for the sensory interpretation to be available the main verb requires simultaneity of its complement, although the condition of "witnessing" or evidentiality also needs to be met. For the cognitive interpretation to arise, however, there are no restrictions as to type of temporal relationship between main and complement tenses, and no restrictions arising from aktionsart effects either. This is, intuitively (and obviously), natural, since mental processes are more likely to be associated with any situation, whatever their location in time, whereas a sensory situation requires direct experience. We turn to the exploration of the nature of these two interpretations.
4.2 Requirements of the Sensory and Cognitive Interpretations First, let us reiterate the kind of cognitive interpretations that are possible for perception verbs, the case of hearsay with oir "hear",13 and the case of inference with ver "see", along with the temporal interpretation of the complement tense, in relation to ET1. It should be clear, then, that whenever we refer to "cognitive" interpretations, it should be understood that both hearsay and inference are included.
13 Although witness: Oi que Mario estaba en el cuarto. "I heard that Mario was in the room". This can be interpreted in the sensory sense but also as an inference, where I assume, by familiar sounds, that is, Mario, who is in the room.
Tensed Complements of Perception Verbs
(24)
55
a. Vi que Juana escribio un libra sobre los drusos.(backward shifted, inference) I see-PRET that Juana write-PRET a book about the Drusseans "I saw that Juana wrote a book about the Drusseans." b. Vi que Juan toco la guitarra. (backward shifted, inference; simultaneous, sensory) I see-PRET that Juana play-PRET the guitar "I saw that Juana played the guitar."
(24b), with an embedded preterito is ambiguous between a sensory and a cognitive interpretation, resulting from the clash in requirements alluded to above. (25)
a. Oi que Maria estaba embarazada. I hear-PRET that Maria be-lMPF pregnant "I heard that Maria was pregnant." b. Oi que Maria se caso. I hear-PRET that Maria get married-PRET "I heard that Maria got married." c. Oi que Maria se casaba. I hear-PRET that Maria get married-lMPF "I heard that Maria was getting married."
(simultaneous, cognitive, hearsay)
(backward shifted, cognitive, hearsay)
(forward shifted, cognitive, hearsay)
The readings for (25a, b) should be clear from earlier discussions. As for (25c), the complement tense is interpreted at later than ET1, i.e. at the time I heard the news, there were indications that Juana was planning to get married. This results from the futurate interpretation of the imperfecto. From the data presented throughout this paper, we can gather that there is an indisputable requirement by the main verb of perception to have a simultaneous interpretation for its complement, if the whole clause is to receive a sensory interpretation. The reverse is not true, however. In other words, in order for a situation to be interpreted as simultaneous with ET1, it does not need to receive a sensory interpretation. Cf. example (17a), which is simultaneous and sensory and (25a), simultaneous but not sensory. As can be seen in examples (25a-c), the cognitive interpretation is not restricted to just one temporal relationship, as the sensory reading is. Thus, the sensory interpretation is the one that seems to be more marked in the sense that it needs a stronger restriction on its interpretation. So we must now examine what that restriction might be. 4.2.1 The Nature of Sensory Contexts In addition to simultaneity, the sensory interpretation is related to both evidentiality and actual occurrence of the situation described by the complement. The concept of evidentiality has been
56
From Words to Discourse
used in the language typology literature for many years (cf. de Haan, 1999) and it deals with the way different languages indicate in the grammar how evidence of information is acquired; i.e. whether the subject has witnessed what is being reported in a direct or indirect way. Thus, there are direct evidentials (the ones discussed above for information strictly acquired through the senses) and indirect evidentials, which include hearsay (also called quotative or reportative) and inference. I use "evidential" in a much looser sense than in typological accounts (which deal mostly with morphological and syntactic ways of encoding evidentials). In order for a perception context to receive a sensory interpretation, then, we need direct evidentiality, in the rigorous sense. However, for our purposes, I will be using the term evidentiality to refer to direct evidentiality. Using a similar concept but within an entirely different context, Dretske (1969) talks of "primary" and "secondary" ways of "seeing". In considering both ways of seeing as epistemic, primary seeing refers to those cases in which "we see that b is P by actually seeing b" (1969: 157 ff, his emphasis). Whereas for "secondary seeing", we are able to see that P by resorting to different background beliefs. In essence, this amounts to the difference between direct experience (witness) and inference from available evidence. Capturing the spirit of this difference but within a more formalized account, Saarinen (1983) develops Hintikka (1969)'s idea of a "perceptual quantifier" (although the data he uses differs slightly from the one discussed here). The basic idea is that in order for someone to see/hear/,etc. (experience with the senses) that an event e took place, we also need the condition that e actually took place and that the subject actually witnessed e. I believe that this is good characterization of what happens in sensory contexts. Any framework or formulation that is used to capture the essence of the sensory reading, as opposed to the cognitive one, needs to incorporate, in one way or other, the idea of evidentiality or witnessing and that there is some situation "out there" (as Saarinen puts it), that actually takes place or obtains, independent of the beliefs (including inferences) of the subject. For our purposes, using only events does not account for examples like: (26)
Vi que Claudia llevaba sombrero/estaba en la esquina. I see-PRET that C. wear-lMPF. a hat/be-lMPF. on the corner. "I saw that Claudia was wearing a hat/was on the comer."
The complement clauses describe states. So the use of situations includes any kind of eventuality (states and events). So a correct characterization of what is witnessed would be that it "obtains", rather than "it takes place", since this would only apply to events. Consider how we would account for the following examples:
Tensed Complements of Perception Verbs
(27)
57
a. Juan vio que Claudia salia del cine. Juan see-PRET that C. get out-lMPF. of the movie theater "Juan saw that Claudia was going out of the movie theater." b. Juan vio que Claudia llevaba sombrero. Juan see-PRET that C. wear-iMPF. a hat "Juan saw that Claudia wore/was wearing a hat."
Assuming the following elements: s: a situation 3 : the Hintikkian perceptual quantifier Vj: Juan ve que 'Juan sees that'; the sentential operator that indicates a complement perception context. The situation in (27a) is true iff: (27a')
(3 s)(s = [ sale del cine] & Vj (s obtains)) "get out of the movie theater"
Similarly, the situation in (25b) is true iff: (27b')
(3 s)(s = [ lleva sombrero] "wear a hat"
& Vj (s obtains))
The first conjuncts in (27a') and (27b'), i.e. sale del cine and lleva sombrero, guarantees that there is some situation "out there" (cf. Saarinen (1983: 124), outside of Juan's "perceptual consciousness" that is independent of Juan or his perceptual viewpoint (as Dretske would put it, free of any "background beliefs"). In other words, the relationship between the perceived situation and the perceiver is not mediated by any kind of inferential process. This is made possible by the perceptual quantifier 3, which binds (from outside) a free variable inside the scope of the perception context. The perceptual quantifier, along with what it binds and the sentential operator V, are all related, under the scope of the perceptual quantifier. The notions of evidentiality and actual occurrence are accounted for. The generalization for any kind of perception situation (i.e. events or states) is thus formulated in (28): (28)
Perceptual quantification for sensory contexts A situation will be sensory iff: (3 s)(s = [ p ] & Px (s obtains))
58
From Words to Discourse
Where: p: a predicate x: the perceiver P: sentential perception (or main verb of perception). This proposal has been sketched for ver "see" but it can, of course, be used with the other sensory perception verbs. Note that with this simple formulation, we do not need to add any specific tense, since the sensory perception will apply whenever there is an evidential relationship between the perceiver and what is perceived. This, in turn, entails that evidentiality will obtain whenever there is some relationship of simultaneity or overlap between the main clause and the complement. In other words, evidentiality entails simultaneity so there is no reason to include a specific index of temporal evaluation to strengthen the context of sensory perception. The perceptual quantification statement captures the lexical semantics of a sensory perception verb with the notions of evidentiality and actual occurrence so that no independent definitions or features need to be postulated. Notice, further, that examples like (3b) and (18a), which are simultaneous but not sensory are, then, ruled out as sensory. As for the cognitive interpretation, no special condition needs to be formulated, since we saw that, just like in contexts where the main verb imposes no lexical requirement on the complement situation, the temporal relationships are not restricted to any particular interpretation. 4.2.2 The Sensory Context and the Two Past Tenses: Recapitulation As was the general case for indicative complements with no lexical requirement from their main verbs, the use of the imperfecto in the complement can potentially (because of its inherent atelicity) be associated with situations which are simultaneous, prior and later with respect to ET1, the different interpretations being reinforced by other contextual elements such as adverbials. We claim that this is still the case for imperfecto complements to main perception verbs. Whether the complex clause receives a sensory interpretation or not will be decided by compliance with the sensory perceptual quantification context, i.e. the simultaneous reading of the imperfecto can arise anyway, but it is not always interpreted as sensory (cf. (25a,c) and footnote 12). The sensory interpretation will only emerge when the requirements for a sensory perceptual quantification context are satisfied (viz. evidentiality and actual occurrence of the perceived situation). As for preterito complements, a true sensory reading will arise only in atelic situations (cf. (19) and (24b)). From all this, then, we can conclude that sensory perception contexts are directly related to atelicity.
Tensed Complements of Perception Verbs
59
Summarizing, the supposed simultaneous requirement of perception verbs does not seem to always work with preterite complements, which may yield a backward shifted reading (and thus, non-simultaneous) when embedded under another past. As for imperfecto complements of past main verbs of perception, we saw that this supposed simultaneous requirement can be overridden by the interaction of aspect, aktionsart and pragmatics, and that we can more accurately account for sensory contexts in terms of evidentiality and actual occurrence, independently of tense specification and aktionsart effects. Note that the way we define the preterito and imperfecto and our account of the aktionsart effects related to these forms have served to provide an accurate explanation of the facts arising in contexts with perception verbs, where the compatibility of the preterito with both telic and atelic aktionsart helped to clarify the differences in readings.
5. IMPLICATIONS FOR SEQUENCE OF TENSE ANALYSES Hopefully, this contribution will add to the sequence of tense discussion by showing the issues involved in the interpretation of indicative complements in general, and perception complements in particular. The elements of aspect, aktionsart, and pragmatic and embedding effects (associated with the complement tenses) bring out a whole new array of factors that need to be considered in the evaluation of the complement tense. Needless to say, this implies that analyses that incorporate notions such as "[-Tense]" for the complement and those that posit copying rules to make the tenses match, will need to revise their accounts in light of this discussion.
REFERENCES Bello, A. y R. J. Cuervo. 1945. Gramatica de la Lengua Castellana.,\s,\.. edition by Niceto Alcala Zamora y Torres. Buenos Aires: Sopena. Cipria, A.1996. The Interpretation of Tense in Spanish Complement Clauses. Ph.D.diss. TheOhio State University. Cipna, A. and C. Roberts. 2000. "Spanish Imperfecto and Preterito: Truth Conditions and Aktionsart Effects in a Situation Semantics." Natural Language Semantics 8. 297-347. De Haan, F. 1999. "Evidentiality and Epistemic Modality: Setting boundaries." Southwest Journal of Linguistics 18. 83-101. Dowty, D. 1979. Word Meaning and Montague Grammar. Dordrecht: Kluwer. Dowry, D. 1986. "The Effects of Aspectual Class on the Temporal Structure of Discourse: Semantics or Pragmatics?" Linguistics and Philosophy 9. 37-62. Dowty, D. 1987. "Aspect and Aktionsart." Ms. The Ohio State University. Dretske, F. 1969. Seeing and Knowing. Chicago: Chicago University Press. (Midway Reprint edition, 1988) Enc, M. 1987. "Anchoring Conditions for Tense." Linguistic Inquiry 18. 633-657.
60
From Words to Discourse
Evans, N. and D. Wilkins. 2000."In the Mind's Ear: The Semantic Extensions of Perception Verbs in Australian Languages." Language 76. 546-92. Garey, H. 1957. "Verbal Aspect in French." Language 33. 91-110. Gili Gaya, S. 1961. Curso Superior de Sintaxis Espanola. Barcelona: Vox Bibliograf. Hinrichs, E. W. 1981. Temporale Anaphora im Englischen. Unpublished Zulassungarbeit, University of Tubingen. Hinrichs, E. W. 1986. "Temporal Anaphora in Discourses of English." Linguistics and Philosophy 9. 63-82. Hintikka, J.I 969. "On the Logic of Perception." In Models for Modalities, J. Hintikka (ed.). Dordrecht: Reidel. Kenny, A. 1963. Action, Emotion, and Will. London: Routledge, Regan and Paul. Kratzer, A. 1989. "An Investigation of the Lumps of Thought." Linguistics and Philosophy 12. 607-653. Moens, M. and M. Steedman.1988. "Temporal Ontology and Temporal Reference." Computational Linguistics 14. 15-28. Partee, B.1984. "Nominal and Temporal Anaphora." Linguistics and Philosophy 1. 243-286. Portner, P. 1994. "A Uniform Semantics for Aspectual -ing." The Proceedings ofNELS 24, GLSA, UMAss. Real Academia Espanola (RAE). 1973. Esbozo de una Nuevu Gramatica de la Lengua Espanola. Madrid: Espasa. Saarinen, E.1983. "On the Logic of Perception Sentences." Synthese 54. 115-128. Smith, C. 1983. "A Theory of Aspectual Choice." Language 59. 479-501. Smith, C. 1991. The Parameter of Aspect. Dordrecht: Kluwer. Suner, M. and J. Padilla-Rivera. 1987. "Sequence of Tenses and the Subjunctive, Again." Hispania 70. 634-42. Sweetser, E. 1990. From Etymology to Pragmatics. Metaphorical and cultural aspects of semantic structure. Cambridge: Cambridge University Press. Vendler, Z. 1967. "Verbs and Times." In Linguistics and philosophy. Ithaca, NY: Cornell Univ. Press, 97-121. Reprinted from Philosophical Review (19'57).
61
4
SPANISH "ASPECTUAL" PERIPHRASES: ORDERING CONSTRAINTS AND THE DISTINCTION BETWEEN SITUATION AND VIEWPOINT ASPECT Brenda Laca, Universite de Paris 8
The main goal of this paper is to propose a correlation between differences in the constraints governing the relative order of "aspectual" periphrases and the semantic and syntactic status of these constructions as expressions of either lexical or syntactic aspectual categories.1 Ordering possibilities will be correlated with a major distinction which is grounded in semantics but can be assumed to project in the syntax, namely the distinction between situation and viewpoint aspect (Smith, 1991). This proposal differs from most other previous treatments of so-called aspectual periphrases in assuming that a unified account of these constructions is neither possible nor desirable. On the contrary, I argue that aspectual periphrases divide both syntactically and semantically into two different categories, and their relative ordering possibilities offer important cues as to this difference. Although this paper will deal mainly with Spanish facts, I will introduce some material from other Romance languages that sheds light on the correlation I am trying to establish. 1 1 would like to thank the audiences at the University of Paris 7 (Tense and Aspect Colloquium, 2000), at the Bosphor University in Istanbul (Word Order Colloquium, 2001), and at the Universidad Menendez y Pelayo in Santander (Escuela de Gramatica Espanola "Emilio Alarcos", 2001) for their valuable remarks and suggestions. I have also greatly benefited from discussions with Mario Barra Jover, Ignacio Bosque, Hamida Demirdache, Anna Gavarro, and Ur Shlonsky, who do not necessarily share the views defended in this paper. I am particularly indebted to Jean Lowenstamm for his insightful comments, and I am certainly more indebted to Olbertz's in depth treatment of Spanish aspectual periphrases than it is possible to acknowledge through references in the text. I gratefully acknowledge the financial support of the Committee Ecos-Sud, action U97H02 France-Uruguay.
62
From Words to Discourse
In section one, I will review some general properties of Romance aspectual periphrases, stressing those which show their monoclausal status most clearly. In section two, a recent proposal by Cinque (1998, 1999, 2000) will be presented in which the relative order of periphrases plays a major role. While this proposal offers a natural explanation for the monoclausal status of these constructions, it will be shown to be descriptively inadequate. In section three, I will offer an alternative to Cinque's explanation for the ordering constraints, which is based on the distinction between extrinsic and intrinsic ordering principles. These two types of ordering principles are well known in the realm of morphology, where they usually serve to distinguish inflectional from derivational affixes. Two correlations will be established: (a) between viewpoint aspect and periphrases that are both more peripheral (more external) with respect to the lexical verb they apply to and exhibit extrinsic ordering properties (being rigidly ordered or in complementary distribution among themselves); (b) between situation aspect and periphrases that are both less peripheral, i.e. appear nearer the lexical verb they apply to, and are more freely ordered among themselves. Section four will address two types of potential counterexamples to my hypothesis: on the one hand Spanish periphrases in which semantics and ordering possibilities seem to diverge, and on the other hand periphrases from other Romance languages exhibiting apparently the same semantics but different ordering possibilities than the Spanish ones.
1. A CHARACTERIZATION OF ROMANCE ASPECTUAL PERIPHRASES The Romance languages exhibit a variety of constructions comprising a higher/finite verb (or verb idiom) and an infinitival or gerundive phrase that are, in very general terms, linked to the expression of time-related notions. Some of the more characteristic of these constructions are given in Table 1: Table 1: Romance aspectual periphrases French
Catalan soler+INF tornar a+INF
Spanish soler+INF volver a+INF
Portuguese Costumar+INF voltar a+INF
etre en train de+INF
estar+GER
estar+GER
anar+GER
ir+GER andar+GER
estar+GER/ a+INF ir+GER andar+GER/ a+INF vir+GER ir+INF estar para/por +INF comefar a+INF, por-se a+INF
aller+INF etre sur le point de+INF commencer a+INF, se mettre a+ INF
anar a+INF estar'a punt de+INF comenfar a/de+INF, posar-se
venir+GER ir a+INF estar por+INF empezar a+INF, ponerse a+INF
Italian solere+INF tornare a+INF stare+GER andare+GER
venire+GER stare per+INF cominciare a+INF, mettersi a+INF
63
Spanish Aspectual Periphrases
cesser de+INF, arreter de+INF## fmir de+INF
venir de+INF continuer a+INF
a+INF deixar de+INF, parar de+INF## acabar de+INF, acabar de+INF 2 continuar+GE R seguir+GER
dejar de+INF, parar de+INF## acabar de+ INF,, terminar deINF acabar de+INF2 continuar+GE R seguir+GER
deixar de+INF, parar de+INF## acabar de+INF, terminar de+INF
smettere di+INF, cessare di+INF## finire di+INF
Acabar de+INF2 continuar+GER/ a+INF
The finite verb in such constructions is either an "aspectualizer", i.e. a superlexical verb denoting not a situation type but a part of the temporal structure of any situation type, or a verb of motion or location. Superlexical verbs correspond to notions such as USE (soler(e)/ costumar), BEGIN (commencer/ empezar/cominciare etc.), FINISH (finir(e)/acabar/terminar). Among the verbs of motion or location, we find (RE)TURN (tornar/volver/voltar), LOCATIVE BE (etre/estar/stare), GO (aller/ anarl ir/ an dare), WALK (andar), COME (venir(e)/ vir), PUT (mettre/posar/poner etc.), LEAVE (deixar/dejar), FOLLOW (seguir), STOP (arreter/ parar). Traditionally, such constructions are classified as "aspectual periphrases", and this classification has carried over to more recent functional studies, in particular to those developing Simon Dik's ideas on the organization of aspectual categories, in which they are conceived of as the expression of different "aspectual operators" (Olbertz, 1998).2 Their behavior, though admittedly not homogeneous, differs from that of typical biclausal constructions containing an embedded or adjunct non-finite clause on several accounts. First of all, the higher/finite verb seems to be transparent as to argument structure and selectional restrictions, so that the whole construction inherits those of the non-finite verb. Thus, gustar 'like' in (la) is a predicate projecting an EXPERIENCER as its dative object and a STIMULUS as its subject, and this argument constellation characterizes the periphrastic combinations containing it: (1)
2
a. Le gustan las peliculas mudas. to-him please the silent movies "He likes silent movies." b. Le dejar on de / empezar on a / volvieron a gustar las peliculas mudas. to-him left of/ began to / (re)turned to please the silent movies "He ceased to like /began to like/ liked again silent movies."
Olbertz (1998) constitutes a major contribution to the description of the semantic and positional properties of Spanish periphrases. Although she does not make a distinction between viewpoint and situation aspect, and she treats all aspectual periphrases as the expression of syntactic operators, the position to be developed in this paper is compatible with her overall results, except for a central point that concerns the status of the progressive periphrasis.
64
From Words to Discourse
In the same vein, the distribution of EN-cliticization in the following Catalan examples shows that the unaccusative / unergative contrast between creixer 'grow' and trucar 'phone' is inherited by the periphrastic combinations that contain those verbs: (2)
a. N'hi creixien moltes. Of-them there grewIMPF many "Many grew there." b. *N'hi trucaven moltes. Of-them there phonedlMPF many "Many phoned there." c. N'hi tornaven a creixer moltes. Of-them there (re)turnedlMPF to grow many "Many grew there again." d. *N'hi tornaven a trucar moltes. Of-them there (re)turnedlMPF to phone many "Many phoned there again."
Finally, suceder 'happen', a predicate requiring an event-denotation as its argument, transmits this property to the periphrastic combinations: (3)
a. Sucedieron accidentes/desgracias/*pensamientos. Happened accidents / tragedies / thoughts "There happened accidents/ tragedies/ *thoughts." b. Suelen suceder/ Estdn sucediendo/ Vuelven a suceder/ Acaban de suceder accidentes/desgracias/*pensamientos. Use happen/ Are happening/ (Re)turn to happen/ Finish of happen accidents/tragedies/*thoughts "Accidents/tragedies/*thoughts usually happen/ are happening/ are happening again/ have just happened."
Second, anaphorization patterns for the non-finite verb and its arguments/adjuncts are variable, and they do not coincide with those of infinitival or gerundive clauses. Some combinations allow for null-complement anaphora, as exemplified in (4c), while others only allow for DO-IT (hacerlo) anaphorization, as in (5).3
3
Languages with a richer inventory of clitics, such as Catalan and French, show that in some cases anaphorization by a clitic is possible, such as in Ja hi torna Already there-returns "Now he's doing it again/ he'll resume doing it".
Spanish Aspectual Periphrases
(4)
(5)
a. Empieza a hablar. Begins to speak "He/she begins to speak." b. *Lo empieza I *Empieza a ello. Begins it / Begins to it c. Ha decidido hablar, pero aim no ha empezado. Has decided speak, but yet not has begun "He/she has decided to speak, but he/she has not begun yet." a. Acaba de hablar. Finishes of speak "He/she has just spoken." b. *Lo acaba I*Acaba de eso Finishes it / Finishes of that c. Ahora le toca hablar a Maria. Pedro acaba *(de hacerlo). Now to-her touches speak to Maria. Pedro finishes (of do it) "Now it's Maria's turn to speak. Pedro has just done it."
Finally, aspectual periphrases are one of the main domains for restructuring effects such as clitic climbing (6a), object agreement with middle-passives (6b) and auxiliary switch in languages such as Italian (6c): (6)
a. Lo terminare de leer hoy. It I-will-finish to read today "I'll finish reading it today." b. Se empiezan a conocer los detalles del accidente. REFL. beginPLUR to know the details of-the accident "The details of the accident are beginning to be known." c. Ha cominciato/E cominciata a cadere la pioggia. Has begun/ Is begun to fall the rain "The rain has started to fall."
That the cliticized argument of the non-finite verb may appear on the higher/finite verb, that in the reflexive middle-passive the finite verb may agree in number with the object of the non-finite verb, and that auxiliary selection on the higher/finite verb may be determined by the nature of the nonfinite verb are facts that point in the same direction as the transparency of the higher/finite verb as to argument structure and selectional restrictions. Both syntactic restructuring and semantic transparency effects can be taken to mean that these combinations involve a monoclausal structure built around a single lexical predicate, the non-finite verb. However, it should be borne in mind that restructuring effects are not coextensive, neither among themselves, nor with the transparency
66
From Words to Discourse
effects or with the bizarre anaphorization patterns that characterize periphrases.
2. CINQUE'S PROPOSAL If periphrases are in fact monoclausal structures built around a single lexical predicate, the question arises as to the syntactic and semantic status of the higher/finite verbs under which this lexical predicate seems to be embedded. In recent work by Cinque (1998, 1999, 2000), an answer to this much debated question is proposed in which constraints on the relative order of periphrases play a major role. The higher/finite verbs in such constructions are, according to this proposal, like bona fide auxiliaries in that they lexicalize a functional (aspectual) head. Cinque argues for a crosslinguistically invariant, highly articulated structure of aspectual functional projections. Together with the rigid order of adverbs and of aspectual particles and affixes, the purported rigid order of "aspectual" periphrases in Romance is taken to provide evidence for the assumed hierarchy of functional projections. In fact, as Cinque accurately points out, periphrases can combine with each other, but some orderings are excluded, as shown in (7): (7)
a. Solia empezar a trabajar a las sets. UsedlMPF begin to work at six "He/she used to start working at six." a'. *Empezaba a soler trabajar a las seis. BeganlMPF to use work at six b. Acababa de terminar de escribir su tercera novela. FinishedlMPF of end of write his/her third novel "He/she had just finished writing his/her third novel." b'. *Terminaba de acabar de escribir su tercera novela. EndedlMPF of finish of write his/her third novel c. Estaba a punto de ponerse a Hover. WasIMPFatpointofput-REFL to rain "It was almost starting to rain." c'. *Seponia aestarapunto de Hover. REFL putlMPF to be at point of rain
The proposed tentative hierarchy is reproduced under (8). The aspectual heads that appear in italics are those for the ordering of which Cinque (1999: 106) adduces further cross-linguistic evidence over and above Romance restructuring constructions, such as positional classes of adverbs, functional affixes and particles. Bold-face indicates the categories that will be dealt with in this paper.
Spanish Aspectual Periphrases (8) ...Asp
ha/>illial
[it. solere] > Asp
delayed (or
[it. fmire per] > Asp
Asprepetitive(l) [it. tornare a] > Asp frequentative (I) > Mod volltlon > Asp di] > Asp
conlinuative
[it. continuare a]> Asp perfec, > Asp
retrospeaive
celeralive (!)
pred,spos,tlonal
[it. tendere a] >
> Asp ,erminative [it. smettere
[ib.-rom. acabar de 2]> Asp prmmmtive
> As
P dUra
A*P progressive [it- stare+GER] > Asp prospective [it. stare per]> Asp [it. inceptive (l) cominciare a]> Mod obhgatlon > Mod ablli(y > Asp frustrative/success [it. (non) riuscire a]> ModperTmssion > Asp conative [it. provare a/tentare di]> Asp complelive(l) [it. fmire di]> (Voice) > Asp celeralive w > As
P inceptive (ID [it- cominciare a]> Asp completive (II) [it. fmire di]> Asp Aspfrei,uenlal,ve(l!>... (Cinque, 1998: 15; 1999:106)
repetitive (II)
[it. tornare a]>
The assumption that periphrases lexicalize functional heads accounts for the monoclausal status of these constructions. This is suggested most strongly by restructuring effects. Together with the additional, and more widely held, assumption that functional structure is rigid, it is also meant to account for the constraints governing their relative order. However, even setting entirely aside the question of the theoretical or conceptual desirability of such a proliferation of functional heads, Cinque's hierarchy does not provide an entirely satisfactory answer to the related questions of the status and ordering constraints of aspectual periphrases. Indeed, the hierarchy can be shown to be descriptively inadequate on at least three accounts. First, obvious cases of acceptable alternative orderings exist, such as those illustrated in (9) and (10): (9)
a. Vuelve a estar Returns to be
a punto de llorar. at point of cry
"He/she's again on the brink of tears." b. Estd Is
a punto de volver a llorar. at point of return to cry
"He/she's about to cry again." (10)
a. Los amigos empezaron
adejar devenir.
The friends began
to leave of come
"Friends were gradually ceasing to come by." b. Dejaron Left
de empezar
a trabajar
al alba.
of begin
to work
at dawn
"They ceased to start working at dawn." Some of these cases of apparently non-rigid order are accommodated in the hierarchy by positing two different aspectual heads with the same expression and roughly the same content, but occupying different positions: a "higher" and a "lower" inceptive head (respectively, Asp
mceptive(I)
and Asp
a
mceptive(ii))' "higher" and a "lower" repetitive head, etc. This strategy seems to be ad hoc and the need to apply it repeatedly deprives the hierarchy of much of its initial attractiveness (note, by the way, that the above examples are not predicted by the hierarchy such as given in (8)). Possible alternative orderings are semantically relevant, and semantic differences among them boil down to what may be
68
From Words to Discourse
expected from composition, as in the case of "normal" embedding constructions. Second, the hierarchy is not sufficiently restrictive, since there are a host of unacceptable combinations that are not excluded by it. Some of them are exemplified in (11): (11)
a. *Termina de empezar a cantar. Finishes of begin to sing "He/she is finishing to begin to sing." b. UVuelve a estar cantando. Returns to be singing "He/she is singing again." c. *Suele acabar de salir. Uses finish of go out "Usually, he has just gone out." d. * Vuelve a acabar de salir. Returns to finish of go out "Again, he has just gone out." e. *Sigue acabando de salir. Follows finishing of go out "He continues having just gone out."
[completive I > inceptive II]
[repetitive I > progressive]
[habitual > retrospective]
[repetitive I > retrospective]
[continuative > retrospective]
Finally, there is at least one case in which the hierarchy makes exactly the wrong prediction. In languages that have a periphrasis expressing what Cinque identifies as retrospective aspect, such as Ibero-Romance acabar de + Inf. and French venir de + Inf., meaning roughly 'have just done something', this periphrasis necessarily precedes, and never follows terminative aspect, contrary to what one would be led to expect from their respective positions in (8): (12)
a. Las campanas acababan de dejar de sonar. The bells finishedlMPF of leave of ring a'. Les cloches venait de cesser de sonner. The bells camelMPF of cease of ring "The bells had just stopped ringing." b. *Las campanas dejaban de acabar de sonar. The bells leftlMPF of finish of ring b'. *Les cloches cessaient de venir de sonner. The bells ceasedlMPF of come of ring
[restrospective > terminative] [restrospective > terminative]
[*terminative > retrospective] [*terminative> retrospective]
These descriptive inadequacies are not of the sort that could be corrected with minor adjustments to the hierarchy. In particular, the need to invoke factors other than functional structure in order to explain the unacceptable combinations in (11) and the need to reduplicate aspectual heads in order to
Spanish Aspectual Periphrases
circumvent the problem posed by the acceptability of alternative orderings undermine the explanatory value of the hierarchy. Furthermore, the initially perceived link between restructuring effects clearly indicating monoclausal status and rigid order is at best a weak one. There is no evidence that the order in restructuring contexts is more constrained than in non-restructuring contexts. As Cinque himself points out (1998, 2000: 19), ordering constraints seem to obtain independently of restructuring phenomena, so that (13a-b) and (14a-b) are equally bad: (13)
(14)
[clitic climbing vs. absence of clitic climbing] a. *Los empezaba a soler leer por la manana. Them beganlMPF to use read by the morning b. *Empezaba a soler leerlos por la manana. BeganlMPF to use read-them by the morning "He began to usually read them in the morning." [object agreement vs. absence of object agreement] a. *Las viviendas prometidas se empiezan a estar por construir. The houses promised REFL begin-PLUR to be by build b. *Seempieza a estar por construir las viviendas prometidas. REFL begin-SING to be by build the houses promised "The promised houses are beginning to be about to be built."
This last point shows that restructuring effects are not a necessary condition for rigid order effects. Neither are they a sufficient condition, since acceptable alternative orderings remain possible even when restructuring has taken place: (15)
(16)
[clitic climbing] a. Los amigos me empezaron a dejar de saludar. The friends me began to leave of greet "Friends were gradually ceasing to greet me." b. Los pacientes me dejar on de empezar a llamar al alba. The patients me left of begin to call up at dawn "Patients ceased starting to call me up at dawn." [obj ect agreement] a. Estos pantalones se empezaron a dejar deusarhace diez afios. These pants REFL began to leave of use it-does ten years "These pants started falling out of use ten years ago." b. Hace anos que las naranjas se dejaron de empezar a cosechar en noviembre. It-does years that the oranges REFL left of begin to pick in November "People have ceased to start picking oranges in November years ago."
70
From Words to Discourse
In fact, in more recent work, Cinque (2000: 15-26) has abandoned the requirement linking monoclausal status to overt restructuring effects, thus suggesting that potentially restructuring constructions are always monoclausal, and that the verbs that are higher than the lowest one in such constructions are always functional. One of the arguments for this revised hypothesis is, precisely, the existence of rigid order in the absence of restructuring effects, such as in (13b) and (14b). In the rest of this paper, I will leave aside the question of overt restructuring effects. The exact nature of their link to monoclausality might be unclear, but it seems clear that they do not relate directly to ordering constraints. Let me briefly summarize the main points of this section. The hierarchy of aspectual functional projections proposed by Cinque and reproduced in (8) does not offer a descriptively adequate account of the constraints governing the relative order of periphrases. It is weakened both by cases of apparently free order, which can only be accommodated by an otherwise unmotivated reduplication of aspectual heads, and by the unacceptability of some of the combinations it predicts. This would seem to require additional constraints over and above the existence of a hierarchy of aspectual heads. However, ordering constraints among periphrases do exist. A more satisfactory account should be able to provide an explanation for the fact that the regularities involved seem to produce partial ordering effects.
3. LEXICAL ASPECT, VIEWPOINT ASPECT AND THE RELATIVE ORDER OF PERIPHRASES 3.1. Extrinsic and Intrinsic Ordering The basic analogy on which the present approach to aspectual periphrases will be built comes from the realm of morphology. In fact, comparably complex patterns involving partial ordering effects are well known in this realm. The approach I will be adopting consists of the following two descriptive hypotheses: (i) the relative order of aspectual periphrases is determined in some cases by extrinsic and in other cases by intrinsic ordering constraints, thus manifesting the same difference in ordering principles that distinguishes inflectional and derivational morphology; (ii) the difference between the extrinsically and the intrinsically ordered domains of aspectual 4
However, "across-the board" restructuring effects are sometimes more difficult to obtain with some combinations of periphrases. This is a point which requires a lot of further empirical work, since data are not clear-cut and we are far from a satisfactory understanding of the array of factors favoring or excluding overt restructuring (Bosque forthcoming). See Cardinaletti and Shlonsky (2000) for an account of clitic climbing in which it is assumed that there are only two clitic positions in a clause, one associated with the lexical verb and the other with the clause's functional architecture, so that "intermediate" clitic climbing, such as in Estaba par volverlo a leer 'WasIMPF by return-it to read' "He/she was about to reread it", would necessarily indicate a bi-clausal structure.
Spanish Aspectual Periphrases
71
periphrases corresponds to the distinction between syntactic (viewpoint) and lexical (situation) aspect. The distinction between extrinsic and intrinsic orderings of affixes is traditionally one of the subsidiary criteria for distinguishing between syntactic morphology (inflection) and lexical morphology (derivation) (see Scalise, 1988; Anderson, 1992). The extrinsic ordering of inflectional affixes illustrated in (17) is fixed and/or semantically irrelevant: it can be assumed to mirror the ordering of functional categories in the architecture of the projection. (17)
a. cant+a+ba+mos 'we sang' sing+TV+IMPF+lPPl b. *cant+a+mos+ba *sing+TV+lPPl+IMPF
On the contrary, the relative order of derivational affixes is potentially free and semantically relevant, as shown in (18). It reflects the order of derivational operations on a basis. (18)
a. [[sombr+ilf]+ero] 'parasol vendor' shadow+DIMIN/MEDIAT+RELAT b. [[sombr+er]+illo] 'small hat' shadow+RELAT+DIMIN/MEDIAT
However, a host of factors conspire to make alternative orderings such as those given in (18a-b) extremely rare. Derivational affixes and/or the semantic operations they are an exponent of impose a number of constraints on the bases they can operate on. In such cases the exclusion of a given ordering can be explained by the intrinsic properties of the elements involved, in particular by the properties of categorial and semantic selection that prevent them from combining with certain bases (together with a number of much more complex factors restricting lexical productivity). Furthermore, it is well known that inflectional affixes are more peripheral (farther from the root) than derivational affixes (see, among many others, Anderson, 1992: 126 and passim). This characteristic ordering, often alluded to as the "inflection outside derivation" universal, is actually a consequence of the fact that functional projections are external with respect to the lexical items they associate with. In what follows, I will try to show that we can find a parallel to the ordering principles of morphology in the realm of aspectual periphrases. That is to say, the same difference between a more peripheral domain of extrinsically ordered elements corresponding to syntactic affixation and a less peripheral domain, corresponding to lexical affixation, where ordering is semantically relevant and constraints are determined by the intrinsic properties of affixes, also holds for aspectual periphrases. I will further argue that this difference corresponds to the difference between viewpoint and situation aspect.
72
From Words to Discourse
3.2. Lexical (situation) and Syntactic (viewpoint) Aspect The distinction between syntactic (viewpoint) aspect and lexical (situation) aspect is due to Smith (1991). In Smith's two-component approach, situation aspect is lexically determined and corresponds to the temporal structure of situations (roughly, to the Vendlerian situation types), whereas viewpoint aspect, expressed by grammatical means, serves to render visible all or parts of a situation, sometimes providing a specific temporal perspective for the whole sentence. Viewpoint aspects determine the interval for which an assertion is made: in Smith's words "only what is visible is asserted". They apply to situations with a given temporal structure (situation aspect) without obscuring this basic structure. Situation aspects, in contrast, are for the most part covert categories, which are evidenced by particularly complex semantic effects in the interaction with viewpoint aspects and with temporal adverbials. These effects mainly involve entailment patterns and regular meaning shifts that can be conceived of as coercion effects produced by marked combinations of the two components. Besides such devices, languages also have lexical means that serve to modify situation types. Essentially, they comprise verbs, affixes or particles that produce "shifted" situation types corresponding to temporal sectors of basic situations or that impose a given situation type on an otherwise underdetermined verb. Thus, start in (19b) produces a "shifted" situation type, in this case an achievement, denoting the initial transition of the accomplishment denoted by (19a). The particle up in (19d) univocally determines the temporal structure of eat as an accomplishment, preventing it from appearing in activity-denoting constellations such as (19c): (19)
a. He ate the apples b. He started eating the apples c. He ate apples d. He ate up the apples / *apples
[accomplishment] [achievement] [activity] [accomplishment / *activity]
As for viewpoint aspect, Smith's "visibility" metaphor and her assumptions as to the span of a situation covered by the different viewpoints can be made more precise if viewpoint aspect is conceived of as a topological relation between the interval corresponding to the situation (situation time: SitT) and the interval for which an assertion is being made (assertion time: AssT). Originally proposed by Klein (1995), this conception of aspect has been integrated into a full-fledged model of temporal relations by Demirdache and Uribe Etxeberria (1997, forthcoming). Under their account, aspectual heads are conceived of as two-place predicates taking SitT and AssT as arguments and expressing the relations BEFORE, AFTER and INSIDE. Temporal heads are two-place predicates expressing the same relations, but taking AssT and UttT (utterance time) as their arguments. For the purposes of this paper, I will set aside some complications related to the aspectual interpretation of tenses and to the degree of recursivity of aspectual heads we are prepared to admit. I will simply exploit the restrictive conception of viewpoint aspect as a relation between SitT and AssT and adopt Demirdache and Uribe Etxeberria's hypothesis as to the three basic predicates involved. This leaves
Spanish Aspectual Periphrases
73
us with the following possibilities for viewpoint aspects: (20)
a. AssT BEFORE SitT: prospective b. AssT AFTER SitT: retrospective c. AssT INSIDE SitT: progressive
3.3. Viewpoint Aspect Periphrases The semantics of some of the aspectual periphrases listed in Table 1 above can be accurately described by assuming that they express the viewpoint aspects enumerated in (20). They are represented in Table 2: Table 2 : Viewpoint aspect periphrases PROGRESSIVE
PROSPECTIVE
RETROSPECTIVE
French
(etre en train de+INF)
aller+INF
venir de+INF
Catalan
estar+GER
Spanish
estar+GER
ir a+INF
acabar de+INF2
Portuguese
estar+GER/a+INF
ir+INF
acabar de+INF2
Italian
stare+GER
acabar de+INF2
Now, these periphrases can precede other aspectual periphrases, as shown in (21), but cannot be preceded by them, as shown in (22): (21)
a. Acaba de {volver a/estar por/empezar a/dejar de/terminar de} corregir las pruebas. RETROSP {return to/be about/begin to/leave of/finish of} correct the proofs "He/she has just corrected again/ been about to correct/ started correcting/ stopped correcting/ finished correcting the proofs."
5 The literature on Romance verbal periphrases is abundant. The reader is referred to Barroso (1994), Bertinetto (1986), Gavarro and Laca (forthcoming), Olbertz (1998), and the literature cited in those works.
74
(22)
From Words to Discourse
b. Va a {volver a/estar por/empezar a/dejar de/terminar de} corregir las pruebas. PROSP {return to/be about/begin to/leave of/finish of} correct the proofs "He/she is going to correct again/ be about to correct/ start correcting/ stop correcting/finish correcting the proofs." c. Estaba {volviendo a/ empezando a/dejando de/terminando de} corregir las pruebas. PROG {return to/ begin to/ leave of/ finish of} correct the proofs "He/she is correcting again/ starting to correct/ stopping to correct/ finishing to correct the proofs." a. *{Vuelve a/Estd por/Empieza a/Deja de/Termina de} acabar de corregir las pruebas. {Returns to/Is about/Begins to/Leaves of/Finishes of} finish of correct the proofs b. *{Vuelve a/Estd por/Empieza a/Deja de/Termina de} ir a corregir las pruebas. {Returns to/Is about/Begins to/Leaves of/Finishes of} go to correct the proofs c. *{Vuelve a/Empieza a/Deja de/Termina de} estar corrigiendo las pruebas. {Returns to/Begins to/Leaves of/Finishes of}be correcting the proofs
The co-occurrence possibilities of these periphrases, which apparently occupy more external positions, are severely restricted. Either they are in complementary distribution with one another, or they exhibit rigid ordering. Thus, prospective and retrospective periphrases are mutually exclusive (23) and they can be followed, but never preceded, by the progressive periphrasis (24). Note, however, that the star before unacceptable combinations signals the unavailability of the pertinent reading for the combination. Since ir a + Inf. also exhibits a reading as a movement verb + directional/final infinitive, and acabar de+ Inf. also has a completive reading (acabar de+ Inf.l 'finish doing something'), the sequences are grammatical, but not with the prospective and retrospective meanings under discussion: (23)
(24)
a. *Acaba de ir a hablar con el. *RETROSP + PROSP talk with him "He/she has just gone to talk with him." b. * Va a acabar de hablar con el. * PROSP + RETROSP talk with him "He/She is going to finish talking with him." a. lAcaba de estar hablando con el. RETROSP + PROG talk with him "He/she's just been talking with him." b. Va a estar hablando con el. PROSP + PROG talk with him "He/she's going to be talking with him."
[only movement reading for ir]
[only completive reading for acabar]
75
Spanish Aspectual Periphrases
c. *Estd acabando de hablar con el. *PROG>RETROSP talk with him "He/she's finishing to talk with him." d. *Esta yendo a hablar con el. *PROG>PROSP talk with him "He/she's going to talk with him."
[only completive reading for acabar]
[only movement reading for ir]
The three Spanish periphrases that are more external and rigidly ordered with respect to one another apparently correspond to the restrictive definition of viewpoint aspect proposed above. However, some precisions are necessary as to their semantics. First of all, the question of the temporal or aspectual nature of ir a + Inf. is a debated one (see, for instance, Bravo, 1999). If these periphrasis were actually an expression of future temporal location, their external position would be neatly accounted for. However, an important argument militates against this view: this construction is possible in some contexts that clearly exclude tenses with future reference, i.e. E after R tenses in the Reichenbachian schema. This is most clearly the case in the antecedents of conditionals, which exclude future and conditional forms, but admit the corresponding forms of ir a + Inf. with the higher verb in the present or in the imperfective past: (25)
a. Si *vendrd / va a venir, nos lo hard saber. If comeFUT / go to come, us it makeFUT know "If he/she will come / is going to come, he/she will let us know." b. [Dijo que ] si *vendria/iba a venir, nos lo haria saber. [Said that] if comeCOND/ wentlMPF to come, us it makeCOND know "[He/she said that] if he/she would come/ was going to come, he/she would let us know." c. Si *vendria / iba a venir, nos lo habria hecho saber. If comeCOND/ wentlMPF to come, us it haveCOND made known "If he/she would come/ were going to come, he/she would have let us know."
We may then assume that ir a + Inf. does not express temporal location, but prospective aspect. Second, we have identified progressive aspect with the expression estar + Ger. This poses the problem of the gerundive periphrases with motion verbs, which we find in the Ibero-romance languages and in Italian and which are often assigned to the same semantic class as the periphrases built with the successors of STARE (Bybee and Dahl, 1989; Coseriu, 1976; Dietrich, 1973). 6
Ir a + Inf. exhibits further uses, not to be discussed here, in which it does not function as an expression of prospective aspect (see Bravo, 1999; Olbertz, 1998). A general problem with the semantics ofperiphrases is that they often exhibit, over and above their aspectual value, what we may call "rhetorical" functions involving discourse sequencing, speakers' attitudes and other pragmatic effects. These effects go well beyond the temporal semantics we are concerned with in this paper. For an insightful account of such effects, see Olbertz (1998).
76
From Words to Discourse
However, the ordering possibilities of gerundive periphrases with motion verbs differ from those of the latter. They can be preceded by periphrases that cannot precede the STARE-periphrases, and they cannot precede some periphrases that can be preceded by the STARE-periphrases, as shown in (26) and (27): (26)
(27)
a. El avion empezaba a ir / *estar perdiendo altura. The plane beganlMPF to go/ be losing altitude "The plane began to gradually lose altitude." b. Vuelve a ir / *estarponiendo los libros uno tras otro en el estante. Returns to go/ be putting the books one after other on the shelf "He's putting the books one after the other back on the shelf." c. Vuelven a andar / *estar arruinando las cosechas. Return to walk / be ruining the crops "They are ruining the crops again." a. El avion ??iba / estaba empezando a perder altura. The plane wentlMPF/ wasIMF beginning to lose altitude "The plane was beginning to lose altitude." b. ??Va / Estd volviendo aponer los libros uno tras otro en el estante. Goes/ Is returning to put the books one after other on the shelf "He's again putting the books one after the other on the shelf." c. *Andan /Estdn volviendo a arruinar las cosechas. Walk/Be returning to ruin the crops "They are ruining the crops again."
There are a number of distributional and semantic facts showing that gerundial periphrases with motion verbs are not instances of a progressive aspect. Progressive aspect expresses proper inclusion of AssT in SitT, so that both endpoints of the situation remain outside the interval for which an assertion is made. This means, on the one hand, that the initial endpoint of the situation is not available for relative temporal location. This is clearly the case for estar + Ger.: the interval denoted by the temporal adverbial or by the w/ze«-clause in (28a) cannot coincide with the initial endpoint of the situation, but must correspond to an interval after this initial endpoint. On the contrary, with ir + Ger. the initial endpoint of the situation remains visible, so that the interval denoted by the temporal adverbial or by the w/ze«-clause in (28b) can coincide with it, specifying the time at which Maria starts setting the table. (28)
a. A las tres/ Cuando llegaba a casa, Maria estaba poniendo la mesa. At the three/ When arrivedlMPF to house, Maria wasIMPF putting the table "At three/When he/she arrived home, Maria was setting the table."
Spanish Aspectual Periphrases
b. A las tres/ Cuando llegaba a casa, Maria iba poniendo la mesa. At the three/ When arrivedlMPF to house, Maria wentlMPF putting the table "At three/When he/she arrived home, Maria started setting he table." As for the final endpoint of the situation, progressive aspect is known to "suspend" telicity, since the natural final endpoint of a telic .situation is not included in the interval for which an assertion is made. In contrast to estar + Ger., ir + Ger. is acceptable in contexts calling for telicity, such as completive contexts (29a) and in X time - adverbials (29b): (29)
a. Cuando termino de ir / * estar poniendo los libros uno encima del otro... When finished of go / be putting the books one upon of-the other "When he finished putting the books one upon the other." b. En veinte anos, fue / *estuvo amasando una inmensa fortuna. In twenty years, went/ *was accumulating a huge fortune "In twenty years, he gradually accumulated a huge fortune."
Thirdly, viewpoint aspects can be assumed to have wide combinatorial possibilities, while also displaying some constraints as to the situation types they can combine with. Combinations that violate such constraints are either unacceptable or lead to regular meaning shifts that can be described as coercion effects. This is precisely what happens with the three periphrases we are discussing. As an expression of prospective aspect, ir a + Inf. can freely combine with any situation type, notably also with states and habituals: (30)
a. Lapeliculale vaagustar. The movie to-him go to please "The movie is going to please him" b. Se iban a ver con frecuencia. REFL wentlMPF to see with frequency "They were going to meet frequently"
The semantic profile of estar + Ger. is very similar to that of the English Progressive. In combination with semelfactives, it leads to derived interpretations as multiple event activities (31 a), and in combination with achievements, it leads to derived interpretations as accomplishments whose "activity" component corresponds to the preliminary stages of the achievement (31b): (31)
a. Estaba tosiendo. WasIMPF coughing "He/she was coughing."
78
From Words to Discourse
b. El tren estaba llegando a la estacion. The train wasIMPF arriving at the station "The train was arriving at the station." In combination with states and with habituals, estar + Ger. leads to dynamic and temporary interpretations: (32)
a. Este chico se estd pareciendo mucho a su padre. This boy REFL is looking like much to his father "This boy is looking more and more like his father." b. Pablo estd yendo muy seguido al teatro. Pablo is going very often to-the theatre "Pablo is going to the theatre very often."
As for acabar de + Inf.2 as an expression of retrospective aspect, the only marked combinations are those with permanent states and habituals. Such combinations give rise to unacceptability, as in (33a-b) or to inchoative shifts, when they are possible, as in (33c). Episodic states are, on the contrary, perfectly acceptable (33d): (33)
a. *Juan acaba de saber ingles. Juan RETROSP know English "Juan has just been able to speak English." b. *Juan acaba de ir muy seguido al teatro. Juan RETROSP go very often to-the theatre "Juan has just gone to the theatre very often." c. Juan acaba de saber la noticia. Juan RETROSP know the news "Juan has just learned the news." d. Juan acaba de estar muy enfermo. Juan RETROSP be very ill "Juan has just been very ill."
Note that this property suffices to differentiate restrospective aspect from perfect aspect, as expressed by the compound tenses and, in particular, by the present perfect, which is acceptable with states and habituals: (34)
a
Juan ha sabido ingles desde su infancia. Juan has known English since his childhood "Juan has been able to speak English since childhood."
Spanish Aspectual Periphrases
b Juan ha ido muy seguido al teatro. Juan has gone very often to-the theatre "Juan has gone to the theatre very often." Whereas perfect aspect requires a resultant state that need not follow the final endpoint of the situation, but can start with the initial endpoint of an atelic situation, retrospective aspect requires that AssT be located after the final endpoint of a situation. Habituals and permanent states are not conceived of as having endpoints (Smith, 1991), and are thus incompatible with it. Episodic states, on the contrary, being conceived of as the result of a change of state and as subject to change, do provide a final endpoint after which AssT can be located. Finally, it is a well known fact that the reading of acabar de + Inf. as an expression of retrospective aspect is not possible with all tenses. The reading is certainly available in the present and in the imperfective past, but it disappears in the perfective past and in all compound tenses, where only acabar de + Inf.l 'finish doing something' surfaces. As expected, acabar de + Inf. does not show the same ordering possibilities when it is in the perfective past, since in this case it is not an expression of viewpoint aspect: (35)
a. ?Acaba / *Acabo de estar hablando con el. Finishes / Finished of be talking with him "He/she has just talked with him / He/she finished talking with him." b. Acaba/*Acabo de {volver a/estar por/empezar a/dejar de/terminar de} corregir las pruebas. Finishes / Finished of {return to/be by/begin to/leave of/finish of} correct the proofs "He/she has just corrected the proofs again / been about to/begun to correct the proofs/ stopped/ finished correcting the proofs."
Less frequently noted is the fact that something analogous happens with ir a + Inf. (see Olbertz, 1998: 365 and passim). When in the perfective past, this periphrasis cannot be combined, for instance, with states and habituals, as in (36): (36)
a. *La pelicula le fue a gustar. The movie to-him went to please b. *Sefueron a ver con frecuencia. REFL went to meet with frequency
7 For the interpretation of the so-called "inclusive anteriority" or "continuative" readings of the present perfect of state predicates illustrated in (34), I'm following the explanation proposed by Kamp and Reyle (1993). See also Demirdache and Uribe-Etxebarria (forthcoming).
80
From Words to Discourse
Simultaneously, its ability to precede some other periphrases vanishes: (37)
a. Va / *Fue a estarpor corregir las pruebas. Goes / Went to be by correct the proofs "He/she is going to be about to correct the proofs." b. Va / *Fue a estar hablando con el. Goes / Went to be talking with him "He/she is going to be talking with him."
This can be interpreted as follows: (a) ir a + Inf. is a bona fide expression of prospective aspect only when in the present or in the imperfective past (and, possibly, with the future when interpreted epistemically); (b) in all other cases, it would correspond to a particular type of lexical aspect focusing the preliminary stages of a situation (see below, 4.1.1.). To summarize, there are three periphrases in Spanish whose semantics can be accurately described by means of a topological relation between AssT and SitT. They exhibit the combinatorial latitude that characterize grammaticalized aspects. As for their relative order, they are more peripheral than other periphrases, and they are rigidly ordered among themselves. 3.4. Situation Aspect Periphrases 3.4.1. Phasal Periphrases The semantics of the periphrases appearing to the right of those we have treated as expressions of viewpoint aspect cannot be described by means of topological relations between two intervals. Some of them obviously serve to create "shifted" situation types that correspond to the beginnings or ends of a given basic situation, such as empezar a + Inf.,ponerse a + Inf., dejar de + Inf., terminar de + Inf. These periphrases, which we will call "phasal", can combine quite freely with one another and are not rigidly ordered among themselves: (38)
a. Los amigos empezaron a dejar de venir. The friends began to leave of come "Friends were gradually ceasing to come by." b. Dejaron de empezar a trabajar al alba. Left of begin to work at dawn "They ceased to start working at dawn."
8 The combinatorial restrictions to some tenses we have just discussed for the expression of retrospective and prospective aspect point out in a similar direction as Rivero's hypothesis (Rivero, 1994), according to which functional (as opposed to lexical) auxiliaries exhibit defective morphology. Space limitations prevent me from developing this point.
Spanish Aspectual Periphrases
However, it is practically impossible to obtain "phases of phases" of a basic situation, i.e. denotations for the initial interval of a final interval, or for the final interval of an initial interval and so on. In most cases, the available readings for the combination of phasal periphrases involve the derivation of a habitual interpretation. The first derived situation, for instance "begin to work at six a.m." in (38b), is reinterpreted as an habitual situation (a pattern of situations of the same nature regularly occurring over a somewhat extended period), and the more external periphrasis, for instance "cease to" in (38b), denotes a phase of this derived habitual situation. Interpretations in terms of "phases of phases" are avoided by way of coercing the habitual interpretation of the lower combination, and completive periphrases such as terminar de + Inf., acabar de + Inf. 1 select for accomplishments. It follows that completive periphrases will not precede other phasal periphrases, as shown in (39): (39)
a. *Los amigos terminaron de dejar de venir. The friends finished of leave of come b. *Termino de empezar a trabajar al alba. Finished of begin to work at dawn
Note that the ordering constraint under discussion is of an intrinsic nature, being determined by the selectional properties of completive phasal periphrases. 3.4.2. Non-phasal Periphrases: Intransformative and Gradual Completion Situations It is sometimes assumed that aspectualizers like continue, keep, go on, etc. are analogous to phasal aspectualizers in that they bring into focus a part of the temporal structure of a basic situation, generating a "shifted" situation that corresponds to the internal temporal sectors of the basic situation (Smith, 1991). If this description were accurate, their semantic contribution would be actually indistinguishable from that of the progressive as a viewpoint aspect. Since the progressive expresses a proper inclusion relation between AssT and SitT, it only makes visible the internal temporal sectors of the situation. But the semantics of the Spanish seguir + Ger. periphrasis, which we classify as an expression of situation aspect, differs significantly from the semantics ofestar + Ger., which we classify as an expression of the progressive. These differences show that seguir + Ger. shifts the basic situation type into a situation type with a peculiar temporal structure that also characterizes some lexical verbs and that we will call "intransformative", following FabritiusHansen(1975). Intransformative situation types involve the comparison between at least two intervals, the denoted interval and a preceding interval, for which it is presupposed that the same (instantiation of the) situation obtained. As illustrated below, seguir + Ger. presupposes that the situation denoted obtained at some interval prior to the interval denoted by the when-clause, whereas estar + Ger. does not carry such a presupposition. Thus, (40c) is entailed by (40a), but not by (40b):
82
(40)
From Words to Discourse
a. Si los musicos hubieran seguido tocando cuando se apagaron las luces... If the musicians hadSUBJ followed playing when REFL extinguished the lights "If the musicians had went on playing when the lights went off..." b. Si los musicos hubieran estado tocando cuando se apagaron las luces... If the musicians hadSUBJ been playing when REFL extinguished the lights "If the musicians had been playing when the lights went off..." c. Los musicos habian estado tocando antes de que se apagaran las luces. The musicians had been playing before of that REFL extinguishedSUBJ the lights "The musicians had been playing before the lights went off."
Now, the contrast between estar + Ger. and seguir + Ger. replicates the contrast between a simple state and an intransformative state. Thus, (41c) is entailed by (4la), which contains the intransformative predicatespermanecer /quedar ('stay, remain'), but not by (41b), which contains the simple state predicate estar. (41)
a. Si Juan hubiera permanecido/se hubiera quedado en Madrid despues del accidente... If Juan hadSUBJ stayed / REFL had SUBJ remained in Madrid after the accident "If Juan had stayed on in Madrid after the accident." b. Si Juan hubiera estado en Madrid despues del accidente... If Juan hadSUBJ been in Madrid after the accident "If Juan had been in Madrid after the accident." c. Juan estaba en Madrid antes del accidente. Juan wasIMPF in Madrid before the accident "Juan was in Madrid before the accident."
Intransformative predicates implicate that a situation 5 obtains at an interval / and presuppose (i) that there is an interval t' prior to t such that a situation s 'obtains at?', and (ii) that s and 5'are subparts of the same situation S. This corresponds exactly to the semantic contribution of seguir + Ger. As for ordering possibilities, seguir + Ger. can precede "phasal" periphrases, but cannot be preceded by them: (42)
a. Sigue poniendose a gritar sin motivo. Follows puttingREFL to shout without reason "He/she still starts to shout without reason." b. Sigue dejando de trabajar a las siete. Follows leaving of work at the seven "He/she stills stops working at seven." c. l*Dejo de seguir cantando. Left of follow singing
Spanish Aspectual Periphrases
o-i
d. ?* Empezo a seguir tocando. Began to follow playing Since phasal periphrases cannot combine with intransformative predicates, as shown in (43), the latter ordering constraint turns out to be an intrinsic one, grounded in the violation of a selectional restriction of phasal periphrases: (43)
a. *Dejo de permanecer en Madrid. Left of remain in Madrid b. *Empezo a permanecer en Madrid. Began to remain in Madrid
The unacceptability of (42c-d) is thus predicted by the hypothesis that seguir + Ger. shifts the basic situation type into an intransformative situation. In Section 3.3, we discarded ir + Ger. as a possible expression of progressive aspect on semantic grounds, and we showed that its ordering possibilities differ from those of viewpoint aspect periphrases. Rather, they correspond to those of situation aspect periphrases. Ir + Ger. can either precede or follow phasal periphrases: (44)
a. Los amigos fueron dejando de venir. The friends went leaving of come "The friends gradually ceased to come by." b. Dejo de ir archivando la correspondencia a medida que llegaba. Left of go filing the mail at measure that arrivedlMPF "He ceased to file the mail as it arrived."
As for its semantics, this periphrasis imposes an incremental interpretation of the basic situation, "shifting" it into the particular temporal structure that characterizes gradual completion predicates. These are known to exhibit mixed properties, being somewhere in between activities and accomplishments (Smith, 1991; Bertinetto and Squartini, 1995). Predicates such as widen, redden, melt, etc are like accomplishments in that they tend to a natural final endpoint, but they are like activities in that each stage of the process leading to this endpoint counts as a proper realization of the situation. Ir + Ger. patterns like these predicates. To summarize: phasal periphrases, as well as seguir + Ger. and ir + Ger., which are more internal than the three periphrases we had identified as the expression of viewpoint aspects, cannot be described by means of topological relations between AssT and SitT. On the contrary, they can be described as means of specifying or shifting the temporal structure of a basic situation, thus qualifying as an expression of lexical (situation) aspect. Their relative order is quite free, being essentially constrained by the intrinsic properties of the elements combined. This substantiates our
From Words to Discourse
claim that "aspectual" periphrases divide into two classes: a more peripheral class of rigidly ordered periphrases expressing viewpoint aspect, and a more internal class of freely ordered periphrases, only subject to intrinsic combinatorial constraints, which express situation aspect.
4. SOME PROBLEMATIC CASES In this section, I will take up the question of the apparent exceptions to the correlation that has just been established. These comprise cases in which semantic and ordering properties seem to diverge, as well as cases from other Romance languages that do not seem to behave as expected.
4.1. Divergences Between Semantics and Ordering For the hypothesis we are trying to substantiate, relying as it does on a very restrictive conception of syntactic (viewpoint) aspect, two types of counterexamples are potentially damaging. One of them involves periphrases whose semantics can be described by means of a topological relation between AssT and SitT, but which occupy less peripheral positions and do not exhibit rigid order. The reverse case would be that of periphrases occupying more peripheral positions and exhibiting rigid order, but whose semantics cannot be accurately described as a relation between AssT and SitT. Both types of counterexamples appear to be attested. 4.1.1. Imminential Periphrases Periphrases like estar por + Inf., estar a punto de + Inf. can be preceded by prospective and restrospective viewpoint aspect periphrases (45), they can only precede phasal periphrases (46), and they can precede or follow seguir + Ger. (47): (45)
(46)
a. Va a estar por salir. Goes to be by go out "He/she's going to be about to leave." b. Acababa de {estar por/estar a punto de} llorar. FinishedlMPF of be by / be at point of cry "He/she had just been about to cry/ He/she had just been on the brink of tears." a. Estuvo por {empezar a/dejar de/terminar de} corregir las pruebas. Was by begin to/ leave of / finish of correct the proofs "He/she had been about to begin / stop / finish correcting the proofs." b. *{Empieza a/deja de /termina de} estar por corregir las pruebas. Begins to/ leaves of / finishes of be by correct the proofs
Spanish Aspectual Periphrases
(47)
8S
a. Estuvo par seguir cantando. Was by follow singing "He/she had been about to continue / resume singing." b. Sigue estando por cantar. Follows being by sing "He/she is still about to sing."
Their relative position is, then, what would be expected of a situation aspect periphrasis, being internal with respect to viewpoint aspect periphrases and freely ordered with respect to some other situation aspect periphrases. Yet, their semantics can seemingly be described as a topological relation between AssT and SitT which is analogous to that expressed by prospective aspect. They even share with the latter the ambiguity that arises in combination with temporal adverbials, which can either locate the situation (when they modify SitT) or function as "perspective point" adverbials (when they modify AssT). Thus, the different readings of (48a) and (48b) mirror those of (49a,b): (48)
(49)
a. Estuvo por marcharse a las ires. Was by leaveREFL at the three "He/she had been about to leave at three." b. A las tres, estuvo por marcharse. At the three, was by leaveREFL "At three, he/she had been about to leave." a. Iba a marcharse a las tres. WentlMPF to leaveREFL at the three "He/she was going to leave at three." b. A las tres, iba a marcharse. At the three, wentlMPF to leaveREFL "At three, he/she was going to leave."
However, prospective aspect as an expression of a precedence relation between AssT and SitT is not to be confused with a modification of situation aspect consisting in focusing on the preliminary stages preceding the initial endpoint of a situation. The latter amounts to a sort of "improper" phasal aspect. Note that imminential periphrases, in contrast with prospective aspect, are incompatible with states and with habituals: (50)
a. La pelicula *estd por / va a gustarle. The movie is by / goes to please-to-him "The movie is about to / is going to please him."
86
From Words to Discourse
b. Se *estaban par / iban a ver con frecuencia. REFL were by / wentlMPF to see with frequency "They were about / were going to meet frequently." We surmise that states and habituals lack preliminary stages previous to their initial endpoint, so that these cannot be focused by imminential periphrases. Furthermore, imminential periphrases require temporal contiguity between the interval they denote and the interval corresponding to the situation. They are thus incompatible with adverbials denoting disjoint intervals, such as de inmediato 'presently', which are acceptable in combination with prospective aspect: (51)
a. *Estaba por marchars e de inmediato. WasIMPF by leaveREFL of immediate "He/she was about to leave presently." b. Iba a marcharse de inmediato. WentlMPF to leaveREFL of immediate "He/she was going to leave presently."
In spite of their superficial semantic similarity with prospective aspect, imminential periphrases remain distinct from it in as far as they do not simply order AssT before SitT. The fact that they are neither possible with states and habituals nor with disjoint time adverbials strongly suggests that they denote a particular temporal sector of a situation, the improper sector corresponding to its preliminary stages. 4.1.2. Soler +Inf. Soler + Inf. necessarily precedes all other aspectual periphrases, even those we have identified as expressions of viewpoint aspect (unless it is in complementary distribution with them): (52)
a. Suele {volver a/estarpor/empezar a/dejar de/terminar de} corregir las pruebas. Uses {return to/be by/ begin to/ leave of/finish of} correct the proofs "He/she usually corrects the proofs again / He/she is usually about to correct the proofs/ He/she usually begins/ stops/ finishes correcting the proofs." b. *{Vuelve a/Estd por/Empieza a/Deja de/Termina de}soler corregir las pruebas. {Returns to/ Is by/ Begins to /Leaves of/ Finishes of} use correct the proofs c. Suele estar esperdndome en la esquina. Uses be waiting-me at the corner "He/she is usually waiting for me at the corner." d. *Estd soliendo esperarme en la esquina. Is using wait-me at the corner
Spanish Aspectual Periphrases
The relative position of soler + Ger. is, then, such as would be expected of a viewpoint aspect periphrasis. But its semantics cannot be described as a topological relation between AssT and SitT. Soler + Ger. denotes a regular pattern of recurrence of the same situation in an extended temporal frame, such as the one that characterizes habitual sentences. Like frequency adverbs, soler + Ger. can quantify over nominal arguments. Thus, in the case of once-only predicates or of permanent states, the quantification expressed by soler + Ger., like that expressed by the frequency adverb in (53b), is roughly equivalent to the adnominal quantification in (53c): (53)
a. Las casas suelen ser de dos pisos. The houses use be of two stories "Houses tend to be two-storied." b. Las casas son a menudo de dos pisos. The houses are often of two stories "Houses are often two-storied." c. La mayoria de las casas son de dos pisos. The majority of the houses are of two stories "Most houses are two-storied,"
It is likely that this periphrasis does not actually belong to the domain of temporal relations in which aspectuality is grounded, but to a much higher domain which would also contain operators of generic quantification. Note that soler + Inf. cannot even be preceded by expressions of epistemic modality, which are among the most peripheral ones: (54)
a. ?* Debe de soler veranear en las Bahamas. Must of use spend-the-summer in the Bahamas "He/she must usually spend the summer in the Bahamas." b. ?* Puede soler veranear en las Bahamas. Can use spend-the-summer in the Bahamas "He/she may usually spend the summer in the Bahamas."
Note also that the incompatibility of soler + Inf. with other aspectual periphrases is not semantically based. As we have seen, periphrases like ir a + Inf., estar + Ger., empezar a + Inf., etc. can combine with habitual situations, provided that habituality is not expressed by soler + Inf. 4.1.3. Volvera + Inf. This periphrasis is internal with respect to viewpoint aspect periphrases (55) and freely ordered with respect to situation aspect periphrases (56):
88
(55)
(56)
From Words to Discourse
a. Va a volver a ponerlos en el estante. Goes to return to put-them on the shelf "He/she is going to put them back on the shelf." b. * Vuelve a ir a ponerlos en el estante. Returns to go to put-them on the shelf "He/she is again going to put them on the shelf." a. Vuelve a estar a punto de llorar. Returns to be at point of cry "He/she's again on the brink of tears." b. Estd a punto de volver a llorar. Is at point of return to cry "He/she's about to cry again." c. Empieza a volver a ponerlos en el estante. Begins to return to put-them on the shelf "He/she starts putting them back on the shelf." d. Vuelve a empezar a ponerlos en el estante. Returns to begin to put-them on the shelf "Once again, he/she begins to put them on the shelf."
[only movement reading for ir]
The position of volver a + Inf. corresponds clearly to that of situation aspect periphrases, and its semantics cannot be described as a topological relation between AssT and SitT. However, it neither "shifts" situation type by focusing on partial temporal sectors of a basic situation, nor does it specify a particular temporal structure for the situation, as is the case with seguir + Ger. and ir + Ger. Nonetheless, its semantics shows important parallels to that of seguir + Ger. As stated above, seguir + Ger. implicates that a situation s obtains at an interval t and presupposes that a situation s' obtains at an interval t' prior to t. We find the same implication and the same presupposition in the case of volver a + Inf. But whereas seguir + Ger. further presupposes that s and s' are subparts of the same situation, volver a + Inf. further presupposes that 5 and 5' are not subparts of the same situation, but different instantiations of the same type of situation. Thus, (57a) entails that the novel has been previously read once, while no such entailment is carried by (57b): (57)
a. Si hubiera vuelto a leer lanovela.... If hadSUBJ return to read the novel "If he/she had read the novel again..." b. Si hubiera seguido leyendo la novela... If hadSUBJ follow reading the novel "If he/she had went on reading the novel..."
Spanish Aspectual Periphrases
°"
Since there is some latitude as to what can count as different subparts of one and the same situation and what can count as different instantiations of the same type of situation, seguir + Ger. and volver a + Inf. can sometimes both be used to describe the same state of affairs, as in (58): (58)
a. Se detuvo un momenta y luego volvio ahablar. REFL stopped a moment and then returned to speak "He/she stopped for a moment, and then resumed speaking." b. Se detuvo un momenta y luego siguio hablando. REFL stopped a moment and then followed speaking "He/she stopped for a moment, and then resumed speaking."
To summarize, the apparent divergences between semantics and ordering constraints do not invalidate the correlation we have established between viewpoint aspect, peripheral position and extrinsic ordering, on the one side, and situation aspect, more internal position and intrinsic ordering, on the other. 4.2. Some Other Romance Periphrases The French periphrasis etre en train de + Inf. seems to correspond roughly to Spanish estar + Ger. It appears in brackets in Table 2 because its status as an expression of progressive aspect is doubtful. This construction is diachronically more recent and much less frequently used than estar + Ger. or, for that matter, the English progressive. Its relative ordering possibilities are more constrained and subject to considerable variation among speakers. Some speakers accept, but others reject, combinations such as those in (59), whereas the corresponding combinations are frequently attested in Spanish: (59)
a. ??// etait en train de commencer a avoir des doutes. He wasIMPF on train of begin to have doubts "He was beginning to have some doubts" b. ??// etait en train definir d'ecrire son roman. He wasIMPF on train of finish of write his novel "He was finishing to write his novel." c. ""// etait en train de cesser de lui faire confiance. He wasIMPF on train of cease of to-him make trust "He was ceasing to trust him."
Thus, etre en train de + Inf. differs in its ordering possibilities both from Spanish estar + Ger. and from the expressions of restrospective and prospective aspect in French, respectively venirde + Inf. and aller + Inf. At the same time however, etre en train de + Inf. lacks the combinatorial latitude
90
From Words to Discourse
that characterizes progressive aspect in the other languages. For instance, its combination with achievements (60a) or with states and habituals (60b-c) are normally rejected by speakers: (60)
a. ??// etait en train d'arriver en retard au bureau. He wasIMPF on train of arrive in delay to-the office "He was arriving late to the office." b. *Cet enfant est en train de ressembler a son pere. This child is on train of look-like to his father "This child is looking like his father." c. *// est en train d'aller tres souvent au theatre. He is on train of go very often to-the theatre "He's going to the theatre very often."
Ifetre en train de + Inf. is not a genuine expression of progressive aspect, then the unacceptability of combinations such as (59) should be explained by selectional restrictions. Its non-peripheral status would go hand in hand with its status as a situation aspect periphrasis. The striking variation observed in speakers'judgments can be interpreted as a symptom of ongoing linguistic change. Catalan, on the other hand, possesses with anar a + Inf. a periphrasis that seems both etymologically and semantically roughly equivalent to Spanish ir a + Inf. However, the Catalan construction cannot be construed as an expression of prospective aspect. Rather, it belongs to the class of imminential situation aspect periphrases focusing on the preliminary stages of a situation. As such, it cannot combine with states or with habituals: (61)
a. *La seva vida va a esser molt tranquilla. The his/her life goes to be very calm "His/her life is going to be very calm." b. *Es van a veure molt sovint. REFL go to see very often "They are going to meet very often."
The use of this periphrasis is subject to two further restrictions: (i) as was the case with imminential periphrases, anar a + Inf. requires temporal contiguity between the interval denoted and the interval corresponding to the situation; (ii) anar a + Inf. requires agentive or intentionally controlled processes. (62a), in which both conditions are violated, is thus considered unacceptable by most speakers: (62)
*el fet que anava aproduir-se al cap depoques hores. the fact that wentlMPF to produce-REFL at-the end of few hours "the fact that was going to materialize in a few hours."
Spanish Aspectual Periphrases
As predicted by our hypothesis, the status ofanar a + Inf. as a situation aspect periphrasis goes hand in hand with a less peripheral position. For instance, anar a + Inf. cannot precede progressive aspect: (63)
*Quan arribis, en Joan va a estar dormint. When you-arriveSUBJ, the Joan goes to be sleeping "When you arrive, Joan is going to be sleeping."
The cases of French etre en train de + Inf. and Catalan anar a + Inf. actually strengthen the correlation we have established between semantics and ordering possibilities, in so far as the comparison with the corresponding Spanish constructions shows both different ordering possibilities and a different semantics. Furthermore, it suggests that the status of etymologically and/or semantically related periphrases is not necessarily the same in different languages. Now, it is an uncontroversial fact that periphrases arise as a result of grammaticalization processes. It is thus tempting to assume that the differences we have just discussed correspond to different stages in such processes. Studies conducted in a functional framework tend to substantiate the hypothesis that grammaticalization processes lead from situation aspect modification to expressions of viewpoint aspect, i.e. from Aktionsart to grammatical aspect (see for instance Squartini, 1998). The formal syntax of grammaticalization is currently assumed to involve the rise of temporal information from lower, lexical layers in a projection to the higher level of functional projections (see for instance Roberts, 1992; D'Hulst, forthcoming). Both types of processes seem involved in the positional and semantic differences we have just described.
5. CONCLUDING REMARKS In this paper, I have tried to show that the relative order of periphrases offers important cues as to their nature, but in a quite different sense than that assumed by Cinque. Instead of across-the-board rigid ordering indicating the existence of a rich hierarchy of functional projections hosting syntactic aspectual heads, we have identified a peripheral domain organized according to extrinsic ordering principles and a more internal domain in which order is semantically relevant and is constrained by the intrinsic properties of the elements combined. This correlates with the semantic difference between viewpoint (syntactic) aspect and situation (lexical) aspect. The semantics of peripheral, extrinsically ordered periphrases can be described by means of a very restrictive conception of viewpoint aspect, in which this category is modeled as a topological relation between two intervals, AssT and SitT. More internal, intrinsically ordered periphrases, produce shifted situation types, and are thus to be classified as expressions of situation aspect. They either bring into focus particular temporal sectors of a basic situation or impose a specific temporal structure onto a basic situation. In most cases, these specific temporal structures are also to be found among simple lexical predicates.
92
From Words to Discourse
Apparent divergences between semantics and ordering constraints were shown not to invalidate the correlation we have argued for. Rather, cross-linguistic divergences with other Romance languages substantiate the correlation, since the observed different ordering possibilities go hand in hand with semantic differences in the predicted sense.
REFERENCES Anderson, S. 1992. A-morphous Morphology. Cambridge: Cambridge University Press. Barroso, H. 1994. O Aspecto Verbal Perifrdstico em Portugues Contempordneo. Porto: Porto Editora. Bertinetto, P. M. 1986. Tempo, Aspetto e Azione nel Verbo Italiano. Florence: Accademia della Crusca. Bertinetto, P. M. and M. Squartini. 1995. "An Attempt at Defining the Class of Gradual Completion' Verbs." Temporal Reference, Aspect and Actionality, Vol. 1. Semantic and Syntactic Perspectives ed. by P.M. Bertinetto et al. Torino: Rosenberg & Sellier. Bosque, I. Forthcoming. "i,Que Sabe El Que Sabe Hacer Algo? Saber entre los Predicados Modales." To appear in Palabras, Victor Sanchez de Zavala in Memoriam ed. by K. Korta and F. Garcia-Murga. San Sebastian: Universidad del Pais Vasco. Bravo, A. 1999. Aspectos de la Gramdtica de!rA+ Infmitivo. Licenciatura Thesis. Universidad Complutense, Madrid. Bybee, J., and O. Dahl. 1989. "The Creation of Tense and Aspect Systems in the Languages of the World." Studies in Language 13. 51-103. Cardinaletti, Anna and U. Shlonsky. 2000. Restructuring in Italian, Clitic Positions and Clausal Strata. Ms,. University of Bologna and University of Geneva. Cinque, G. 1998. "'Restructuring' and the Order of Aspectual and Root Modal Heads." Univ. of Venice Working Papers in Linguistics 8. Cinque, G. 1999. Adverbs and Functional Heads. New York: Oxford Univ. Press. Cinque, G. 2000. "Restructuring" and Functional Structure. Ms., University of Venice. Coseriu, E. 1976. Das Romanische Verbal System. Tubingen: Nan. Demirdache, H. and Uribe-Etxebarria, M. 1997. "The Syntactic Primitives of Temporal Relations." Langues & Grammaires 3, Univ. de Paris 8. Demirdache, H. and Uribe-Etxebarria, M. Forthcoming. "La Grammaire des Predicats Spatio-Temporels: Temps, Aspect et Adverbes de Temps." To appear in Laca ed. (forthcoming). D'Hulst, Y. Forthcoming. "Le Developpement Historique des Propritetes Temporelles du Cconditionnel Franfais et Italien." To appear in Laca ed. (forthcoming). Dietrich, W. 1973. Das Periphrastische Verbalaspekt in den Romanischen Sprachen. Tubingen: Niemeyer Fabritius-Hansen, C. 1975. Transformative, Intransformative und Kursive Verben. Tubingen: Niemeyer. Gavarro, A.and B. Laca. Forthcoming. "Les Perifrasis Temporals, Aspectuals i Modals." To appear in Gramdtica del Catald Contemporani ed. by J. Sola et al. Barcelona: Empuries. Hernanz, M.L. and G. Rigau 1984. "Auxiliaritat i Reestructuracio." Els Marges 31. 29-51. Kamp, E. and U. Reyle. 1993. From Discourse to Logic. Dordrecht. Kluwer. Klein, W. 1995. "A Time-Relational Analysis of Russian Aspect." Language 71. 669-695. Laca, B. 1998. "Aspect - Periphrase - Grammaticalisation. A propos du "Progressif' dans les Langues Ibero-Romanes." Neuere Beschreibungsmethoden der Syntax romanischer Sprachen. Romanistisches Kolloquiums XI (Giessen 1995) ed. by W. Dahmen et al., 207-226. Tubingen: Narr. Laca, B. 2000. "Les Periphrases de Phase en Catalan Contemporain." Linguistica Romanica et Indiana (Festschrift Wolf Dietrich) ed. by B. Staib, 249-270. Tubingen: Narr.
Spanish Aspectual Periphrases
Laca, B. ed. Forthcoming. Temps et Aspect. De la Morphologie a L'Interpretation. Paris: Presses Universitaires de Vincennes. Olbertz, H. 1998. Verbal Periphrases in a Functional Grammar of Spanish. Berlin: Mouton/De Gruyter. Rivero, ML. 1994. "Auxiliares Lexicos y Auxiliares Funcionales." Gramatica del Espanol ed. by V.Demonte.Mexico: El Colegio de Mexico. Roberts, I. 1992. "A Formal Account of Grammaticalisation in the History of Romance Futures." Folia Linguistica Historica 13. 219-258. Smith, C.S. 1991. The Parameter of Aspect. Dordrecht. Kluwer Scalise, S. 1988. "Inflection and Derivation." Linguistics 26. 561-581 Squartini, M. 1998. Verbal Periphrases in Romance. Berlin: Mouton/De Gruyter.
This page intentionally left blank
95
5
NON-DECLARATIVE SENTENCES IN SPANISH: THE CASE OF THE INFINITIVE Xose Resales Sequeiros, University of Greenwich, UK
1. INTRODUCTION This article is concerned with the imperative use of the infinitive in Spanish. In particular, an analysis is presented of its semantics and pragmatics, and an account is provided which brings together these two aspects of the infinitive and its imperative use. Wilson and Sperber (1988) provide a survey of semantic accounts of non-declarative sentences and argue that the infinitive and the imperative differ semantically. According to them, the latter encodes that the state of affairs described is both potential and desirable, whilst the former encodes that the state of affairs is only possible. They also argue that this does not preclude the use of the infinitive as an imperative when assumptions about achievability (i.e. potentiality) and desirability are available in the context. However, their analysis does not take account of a number of issues: (a) what is the theoretical status of the imperative interpretation of the infinitive?, (b) what pragmatic processes, if any, are involved in this type of imperative use?, and (c) why is the imperative use of the infinitive possible in Spanish (and other languages, both Romance (e.g. French) and non-Romance (e.g. German)) and not in languages such as English? It will be argued that whilst the semantics of the infinitive may be restricted to possibility, in an imperative use its contribution is to the explicatures of the utterance, and not to the implicatures. This means that its contribution is to the explicit content of the utterance and not to the implicit content, contrary to what Wilson and Sperber seem to suggest. Hence, any pragmatic
96
From Words to Discourse
processing involved will contribute to the explicit rather than the implicit part of the interpretation, which provides an answer to the second question posed above. Equally, this also seems to suggest an answer to the third question raised earlier, namely, that the contribution of the imperative use to the explicit content of the utterance and the repeated use of the infinitive as imperative in Spanish (and other Romance and non-Romance languages) has made it possible for this use to be more accessible and become semi-lexicalised. In English the same (politeness and referential) effects are achieved by other means and, as a result, the infinitive is not normally used as an imperative. The article is organised as follows. Firstly, a brief survey is provided of the role of illocutionary force and mood in the interpretation of non-declarative sentences. Secondly, there follows a discussion of Sperber and Wilson's model, which is applied to Spanish data (involving both the imperative and the imperatival infinitive), as well as (briefly) to data from other languages. Thirdly, some problems are discussed in relation to Sperber and Wilson's analysis, and then an alternative account is developed. Finally, some conclusions are provided, which suggest a number of areas for further research.
2. MOOD AND ILLOCUTIONARY FORCE IN UTTERANCE INTERPRETATION Speech-act accounts of non-declarative (as well as declarative) sentences have relied on illocutionary force to explain differences between different types of sentence. Illocutionary force has been seen as a semantic category, where the meaning of a sentence was associated with its general potential for illocutionary force. In particular, the semantic meaning of a sentence would be the range of illocutionary forces it could be used to perform. So, for example, non-declarative interrogative sentences have been seen as performing directive speech acts, which involve the speaker making a request for information. Non-declarative imperative sentences were also seen as performing a directive speech act, but this time involving a request for action, not information. From this perspective, the role of pragmatic interpretation was to identify the actual speech act performed from the range of possible ones (i.e. the ones allowed by the specific sentence type). Hence, for instance, an imperative may be used to perform a request for action, which covers a number of possibilities including a request interpreted as a plea, entreatment, order, etc. Pragmatic considerations would help us determine which one of these possible, more specific, requests for action would have been performed in any particular instance. One of the main problems with a speech-act based account of non-declarative sentences are cases in which sentence meaning and illocutionary force potential do not correspond. Thus, for example, many uses of imperatives do not involve a request for action, as predicted by its illocutionary-based semantics. Cases in point are, for example, good wishes, advice, threats, and permission, amongst others. Let us consider good wishes and advice as an illustration of this point:
Non-Declarative Sentences in Spanish
(1)
(2)
97
Good Wishes: (John bidding Mary farewell at the airport) John: Have a safe flight back. Advice: Mary: Is there a petrol station around here? John: Yes. Take the third turn on the left and continue 200 meters.
In both these cases, speech act theory would predict that some action was being requested from the hearer. Thus, for example, in (1) John would be predicted to be requesting Mary to perform the action described, namely, making sure she has a safe flight. However, John is clearly not doing any such thing. In particular, he is not intending to communicate, for example, that Mary must carry out checks to make sure that the aircraft is in flying condition. This action is beyond Mary's control and, probably, ability. John is merely expressing a good wish towards Mary's safe arrival at her destination and, thus, no action needs to be performed to understand successfully the force of the sentence. In (2) Mary is not intending to get John to perform the actions she is describing. She is merely indicating what he could do to get to a petrol station. John may or may not choose to follow her advice. His understanding of the sentence does not depend on his performing the actions described. In these types of case, the imperative cannot be said to change its meaning in any special way, which might prevent it from fulfilling its predicted illocutionary potential. In the light of the above examples, it is difficult to see how illocutionary force, as understood by speech-act theorists, could be a semantic property, rather a pragmatic one, since the precise force with which a sentence is uttered is dependent on pragmatic interpretation (see below for further discussion on this point). In this respect, illocutionary force is a property of utterances (pragmatics), rather than sentences (semantics). However, if illocutionary force is not what distinguishes the meaning of different types of non-declarative (as well as declarative) sentences, then something else must be responsible. Many theorists have resorted to the notion of mood to explain these differences.
3. MOOD-BASED ANALYSIS OF IMPERATIVE FORCE Mood has been defined as referring to the semantic and logical properties that make one sentence type (e.g. imperatives) different from another (e.g. interrogatives). To make a moodbased account work, we need to know (a) how these moods are characterised (i.e. their semantics) and (b) how they relate to force (i.e. their pragmatics). Using mood, as opposed to force, as the main distinguishing feature for (non-)declarative sentences involves establishing the range of speech acts a given mood is conventionally used to perform. As Hare (1970, quoted in Wilson and Sperber 1988:79) states: "When we say that The
98
From Words to Discourse
cat is on the mat' is a typical indicative (when we mention its mood, that is), we identify the type of speech act which it is standardly used to perform. Thus mood signs ... classify sentences according to the speech acts to which they are assigned by the conventions which give meanings to those signs". (My italics) The standard speech acts a given mood is used to perform must be identified by pragmatics. As far as the imperative mood is concerned, it is generally accepted that it is standardly used to perform directive speech acts, which involve an attempt to get the hearer to perform the action described by the proposition expressed. This is, in effect, a parallel characterisation to that of illocutionary force and as such is subject to the same counterexamples. The only difference between the two characterisations is that illocutionary force is seen as a more pragmatically oriented notion (i.e. it emphasises the impact of the sentence on the hearer, its force), whereas mood is seen as a more semantically oriented one (i.e. it emphasises the semantic and logical content of the various sentence types, their, mood). However, in a mood-based approach, pragmatics plays a crucial role in determining which speech acts a given mood is used to perform. As was briefly illustrated above, there are many uses of the imperative which do not involve getting the hearer to perform any kind of action, thereby providing evidence against a moodbased analysis of the imperative. Let us consider some counterexamples to this type of analysis: (3)
(4)
Threats: (John and Mary are fighting) John: Yeah, go on. Punch me. Permission: Mary: Can I use your computer? John: Go ahead. Use it for as long as you need to.
In these cases the imperative is being used to convey threats and permission. However, in neither of them is there an attempt to get the hearer to perform the action described by the proposition expressed. In the case of threats, the use of the imperative is not an attempt to get the hearer to perform the action described by the proposition expressed, rather it is an attempt to stop her doing it. In the case of permissions, the speaker is merely showing his approval towards the action described, but not his intention to get the hearer to carry out the action. In addition to threats and permissions, the earlier cases of good wishes and advice are also counterexamples to the mood-based analysis. In the case of good wishes, the speaker is merely expressing his goodwill towards the hearer. The action described is largely out of the hearer's control, so she could not be taken to understand the sentence as an attempt to get her to perform the action in question. In the case of advice, the speaker is giving the hearer helpful directions, but in no way can he be seen as trying to get her to perform the action described by the proposition expressed. Some scholars have discussed this type of example, putting forward alternative analyses. However, none of them provides an account which covers the full range of cases. Schmerling
Non-Declarative Sentences in Spanish
99
(1982), for example, argues for an account based on the imperative communicating that the speaker is trying to bring about the state of affairs described by the proposition expressed, as opposed to getting the hearer to do it. This analysis could be said to account for good wishes, in that the speaker could be seen as wanting to bring about the good wish; but not for advice, threats, or permission cases, where the speaker does not necessarily want to bring about the state of affairs described. McGinn (1977) departs from speech-act analyses and proposes an alternative using the notion of truth conditions. He argues in favour of the imperatives having not truth conditions as such, but rather fulfillment conditions, which are satisfied when the state of affairs described by the imperative is materialised. The fulfillment of this state of affairs can be achieved by linking mood with desire, which also provides the connection between mood and force. That is to say, the speaker of the imperative (which is the carrier of mood) would be seen as conveying a desire (which is the carrier of force) regarding the realisation of the state of affairs, thereby contributing to the connection between the semantic and pragmatic stages of interpretation. However, this analysis would only account for cases such as good wishes, where a desire may be seen to be involved. In the other cases (threats, advice, and permission), no desire on the part of the speaker is necessarily present. The real breakthrough in accounting for the range of imperatival uses comes when force is taken out of the equation altogether. Huntley (1984) followed this line of analysis when he proposed that the difference between declaratives and imperatives (and other non-declaratives) lies in whether they make indexical reference to the actual world (declaratives) or to some other possible world (imperatives). In this respect, he argues (ibid.: 122), "[imperatives, and other nondeclaratives] can represent a situation as being merely envisaged as a possibility with no commitment as to whether it obtains, in past, present or future, in this world". However, two problems arise in relation to this proposal. Firstly, it is not clear what the link is between envisaging a state of affairs as a mere possibility, and wishing, advising, threatening, permitting, etc., to bring about that state of affairs. In other words, it is not clear what the relationship is between the semantics and pragmatics of non-declaratives from this point of view. However, Huntley did make an important connection, for the purposes of this article, between imperatives and infinitives. He argued that both involve the notion of possibility in their semantics, with the difference that infinitives do not necessarily represent the state of affairs they describe as either achievable or desirable (which imperatives may do). This allows for the possibility of infinitives being interpreted without imperative force, though it does not preclude it altogether. This latter option is the one explored in this article. The second problem with Huntley's analysis is that it predicts that imperatives can be used without imperative force (i.e. by encoding only possibility). However, this seems to be incorrect, as imperatives must somehow be linked to imperative force as part of their meaning (however this force may ultimately be characterised). Otherwise, this would entail that imperatives and infinitives would be semantically equivalent, which seems to be counterintuitive. In order to solve these problems, Sperber and Wilson (1988:84) argue that imperatives are necessarily linked to notions of desirability and achievability, whereas the infinitive is only
100
From Words to Discourse
linked to the notion of possibility. Thus, they distinguish semantically between imperatives and infinitives, thereby retaining their potential for different pragmatic forces, whilst at the same time allowing for the possibility of similarities in their interpretation. Moreover, as will be shown below, they do this by dispensing with the notion of semantic mood altogether.
4. SPERBER AND WILSON'S APPROACH Traditionally, the imperative is seen as a mood in its own right. However, Sperber and Wilson argue that they "see no reason to assume that semantic moods exist" (1988:99). They argue further that the characteristic linguistic features associated with declarative and non-declarative sentences only encode an abstract property of the intended interpretation: "the direction in which the relevance of the utterance is to be sought" (ibid: 101). In the case of infinitives the direction is one of envisaging possibilities. In the case of imperatives the direction is one of entertaining states of affairs as potential (achievable) and desirable. To exemplify this, consider the following examples in Spanish, Galician, and English: (5)
a. Me encantaria vivir en el campo. me would-love to-live in the countryside. "I would love to live in the countryside." b. Querofalar con ela. I-want to-speak with her. "I want to speak to her." c. To be in England in April!
(Spanish)
(Galician)
(English)
In these three sentences there are three infinitives: vivir, falar, and to be, respectively. In each case, the claim is that the speaker is only committing himself to the possibility of the events described. He is envisaging the events as possible and commenting on them at the same time. The semantic meaning of (5a), for example, could be paraphrased as follows: (5')
a. Me encantaria tener la posibilidad de vivir en el campo. me would-like to-have the possibility of to-live in the countryside. "I would love to have the possibility of living in the countryside."
However, in the case of imperatives the speaker is encoding more than this. Consider the following Spanish examples (where SUBJ denotes subjunctive):
Non-Declarative Sentences in Spanish
(6)
(7)
101
Order: - ^Que quieres que haga? what you-want that I-do-SUBJ "What do you want me to do?" - Abre la puerta. you-open-IMP the door "Open the door." Permission: - ^Puedo abrir la puerta? can-I to-open the door "Can I open the door?" - Abrela. you-open-IMP-it "Open it."
In both examples the speaker is encoding that the events are potential (achievable) and desirable. The difference between the two is that whilst in (6) the state of affairs is desirable from the point of view of the speaker, in (7) it is desirable from the point of view of the hearer. Determining from whose point of view the event is desirable is one of the semantic indeterminacies in the interpretation of imperatives that must be resolved pragmatically. Following Sperber and Wilson, the meaning of example (6), for instance, could be paraphrased as follows: (6')
a. It is feasible for you to open the door. b. It is quite desirable, from my point of view, that you open the door.
In fact, Sperber and Wilson's analysis involves three pragmatic indeterminacies regarding the interpretation of imperatives. Firstly, one must determine how potential the state of affairs described is. This may have a bearing on how some utterances are interpreted. For example, it may help us distinguish between requests and challenges: (8)
(9)
Request: (At home) Ana to Pablo: Hazme un cafe. Ana to Pablo: Make-IMP-me a coffee. "Ana to Pablo: Make me a coffee." Challenge: (At home) Ana to Pablo (who's manifestly incapable of making exotic drinks): Hazme un cocktail. Ana to Pablo (who's manifestly incapable of making exotic drinks): Make-IMP-me a cocktail.
102
From Words to Discourse
"Ana to Pablo (who's manifestly incapable of making exotic drinks): Make me a cocktail. " In (8), the imperative can be interpreted as being a request, as the state of affairs described is easily achievable. However, in (9) the state of affairs is manifestly less achievable, because of (the shared knowledge regarding) Pablo's incompetence at making exotic drinks. Thus, the imperative in this case can be interpreted as a challenge more readily than (8). Secondly, one must determine how desirable the state of affairs is. The degree of desirability may also have a bearing on how we interpret some utterances. For example, it may help us distinguish between pleas and requests: (10)
(11)
Request: (Pablo is going to the shop) Ana to Pablo: Trdeme tabaco. Ana to Pablo: Bring-IMP-me tobacco "Ana to Pablo: Bring me tobacco." Plea: (Pablo is going to the shop) Ana, who is suffering from withdrawal symptoms, to Pablo: Trdeme tabaco. Ana, who is suffering from withdrawal symptoms, to Pablo: Bring-IMP-me tobacco. "Ana, who is suffering from withdrawal symptoms, to Pablo: Bring me tobacco."
In (10), in the absence of other contextual assumptions, the imperative is used to make an ordinary request. In particular, there is no reason to believe that a specially high degree of desirability is involved, which means that Pablo is unlikely to give the utterance an interpretation beyond that of a request. However, in (11), where there are relevant manifest assumptions available, the imperative is used with extra force. In particular, due to these available assumptions, Pablo will establish a high degree of desirability as part of the interpretation of the utterance, which will lead him to interpret it as a plea. The third indeterminacy to be resolved pragmatically is the one already mentioned above, namely, the answer to the question from whose point of view is the state of affairs desirable: speaker, hearer, or someone else. As shown above in (6) and (7), the resolution of this indeterminacy may lead us, for example, to interpret utterances as orders or permissions, or in other cases as favors or advice. This resolution is dependent on whether the state of affairs is desirable from the speaker's point of view (orders, favors), or the hearer's (permissions, advice). Hence, to sum up, three elements play a part in Sperber and Wilson's analysis of imperatives: degree of achievability, degree of desirability, and point of view of desirability. Using these parameters, we can account for examples (l)-(4) discussed earlier. Example (1) was a good wish because it involved desirability from the point of view of the speaker (with benefits to the hearer), and lack of achievability (as neither the speaker nor the
Non-Declarative Sentences in Spanish
103
hearer can bring about the state of affairs). Examples (2)-(4) involved desirability from the point of view of the hearer. In particular, example (2) was a piece of advice because the information was desirable from the hearer's point of view in order to fulfil her goal, and was easily achievable. Example (3) was a threat because, although it was potentially achievable, it was not desirable, particularly from the hearer's point of view, since carrying out the state of affairs described would bring negative consequences to her (which would consequently make it not desirable from her point of view). Finally, example (4) was a permission because the state of affairs was desirable from the hearer's point of view, and its potentiality depended on the speaker's acquiescence (which in this case is given). Let us now turn to the imperative use of the infinitive.
5. IMPERATIVE USE OF INFINITIVE It is a well known fact that in Spanish (as well as in many Romance and non-Romance languages, see below) the infinitive can be used as an imperative (see e.g. Butt and Benjamin 1988:277ff). Consider the following examples: (12)
a. In a hospital: Nofumar. not to-smoke "Do not smoke." b. In a recipe book: Trocear las patatas. to-cut the potatoes "Cut the potatoes." c. In an instructions Booklet: Enchufar la impresora. to-plug the printer "Plug the printer." d. In a cashpoint: Introducir la tarjeta. to-introduce the card "Introduce the card."
It seems clear that in none of these examples the infinitive is used simply as a possibility. For instance, in (12a) the hospital personnel do not just want visitors or patients to consider the utterance as communicating the mere possibility of not smoking. In fact, that interpretation would be useless as far as they are concerned, since what they really want is to avoid health risks
104
From Words to Discourse
to their patients. In other words, they want people positively not to smoke. Moreover, if any visitor to the hospital was found smoking, it seems clear that, on the basis of the sign, they would be entitled to require the person to stop smoking or leave the building. This suggests, therefore, that what is being communicated is much more than just a mere possibility. In other words, there are issues of achievability, desirability, and point of view involved in this type of case. In particular, the state of affairs described is easily achievable and highly desirable (from the point of view of the patients and public at large). The same is the case for the other examples, whose force is also that of an imperative. In (12b), the recipe book is not just envisaging the action described as a possibility, but rather if the reader wants to follow the particular recipe, then they must carry out the action described to bring about the overall goal successfully. In (12c) and (12d), the situation is similar. In (12c), the booklet is describing an action which must be followed, if the reader wants to use the printer successfully. In (12d), the cashpoint machine is describing an action which must be followed in order to carry out a transaction. In all these cases, the state of affairs described is achievable and, in the right circumstances, highly desirable from the addressee's point of view. In more general terms, example (12a) can be interpreted as an order, given the social assumptions about the power which hospitals authorities have within their premises. On the other hand, examples (12b-d) can be interpreted as instructions, given their goal-oriented context. In all the above examples, the infinitives could be replaced by imperatives without a change in their interpretation (as far as their directive import is concerned), as shown below: (12')
a. In a hospital: Nofumen. not you-smoke-IMP "Do not smoke." b. In a recipe book: Troceen las patatas. you-cut-IMP the potatoes "Cut the potatoes." c. In an instructions booklet: Enchufen la impresora. you-plug-IMP the printer "Plug the printer." d. In a cashpoint: Introduzcan la tarjeta. you-introduce-IMP the card "Introduce the card."
For instance, imperatival example (12'a) carries the same pragmatic force as the infinitival (12a). Both involve a request not to smoke. Similarly, imperatival example (12'b) carries the
Non-Declarative Sentences in Spanish
105
force of an instruction, as does infinitival example (12b). Both instruct the addressee to follow the action described in order to complete the recipe successfully. The same type of explanation applies to the other examples, where the imperative and infinitival versions share the same pragmatic force. Hence, these paraphrases seem to show that the infinitival versions do, in fact, convey the same direction of relevance sought as the imperatives (in the type of contexts considered). From Sperber and Wilson's point of view, they convey potentiality (i.e. they are achievable state of affairs) and desirability (i.e. they are desired outcomes either from the point of view of the addresser, e.g. at the hospital, on behalf of patients; or from the point of view of the addressee, e.g. in the case of the recipe book). In their framework, Sperber and Wilson allow for the possibility of imperatives and infinitives having equivalent pragmatic interpretations: "The semantic analysis of imperatives must make reference to these notions [i.e. potentiality and desirability]; the semantic analysis of infinitival clauses does not. This is not to say that infinitival clauses can never be used with imperatival force. When can they be so used? When it is clear in the context that the state of affairs "envisaged as a possibility" is both achievable and desirable." (Wilson and Sperber, 1988: 84) If the account provided so far is correct and the infinitival constructions convey the same import as the imperative (as far as the direction of relevance sought is concerned), then the question arises as to whether this equivalence is semantic or pragmatic. In other words, are infinitives semantically ambiguous between possibility, on the one hand, and potentiality and desirability, on the other? Or, do they, semantically, only encode possibility and then become pragmatically interpreted as imperatives? At a first glance, the first option would seem to be counterintuitive, as the infinitive is not normally seen as encoding imperative mood in traditional grammars. From Sperber and Wilson's point of view, the solution would be closer to the second option, i.e. that infinitives encode possibility, and imperatives potentiality and desirability, so that at a linguistic semantic level they are different. However, as stated above, from their point of view, infinitives could be interpreted as carrying imperative force at a pragmatic level, provided assumptions of potentiality and desirability were available in the context (Sperber and Wilson ibid.). This means that infinitives and imperatives could have equivalent pragmatic force. This second analysis would appear to be confirmed when we consider the type of context in which the imperatival infinitive is found. As seen in the above examples in (12), imperatival infinitives are used in contexts in which instructional assumptions are common (e.g. hospitals, recipe books, instructions manuals, etc.). Given their accessibility in these contexts, it would seem natural that assumptions about potentiality and desirability could be used in the interpretation process, with the resulting imperatival upgrade. This would seem to be corroborated by the fact that in other types of context the imperatival interpretation does not seem to be equally available. Thus, consider the following examples where the infinitive is used, but no imperatival interpretation is available (please note that, for ease of presentation, in these and the remaining non-English examples only idiomatic translations are provided, with extra
106
From Words to Discourse
information given when required): (13)
a. On a travel brochure: / Ver la isla desde el avion! Esa es una experiencia iinica. "To see the island from the plane! That is a unique experience." b. At the beginning of a novel: Reflexionar, charlar, observar. Eso era lo que haciamos diariamente. "Reflect, talk, observe. That was what we did daily." c. In questions: ^Marchar ahora? No meparece una buena idea. "Go now? It doesn't seem to me a good idea." d. Intensifier in Repetitions: Comer comimos. Pero la comida era mala. "Eat we did. But the food was bad." e. In descriptions: - iQue hicisteis? "What did you do?" - Primero, ducharnos. Despues, llamar por telefono a Juan. Luego, vernos con el. "First, we showered. Then, we telephoned Juan. Then, we met him."
In all these cases only the narrow infinitival interpretation is available. For example, in (13a) no indication is conveyed about the need for the addresser or addressee to carry out the action described, only the possibility is being envisaged (which may subsequently lead to desiring its actualisation). Similarly, in (13b) no desire is necessarily communicated for the writer or reader to reflect, talk or observe, rather those states of affairs are merely being described (in the past). The same goes for examples (13c-d). Furthermore, general unavailability of imperatival interpretation is not only due to the type of situational context. In certain common grammatical structures, the imperatival interpretation is not available either, e.g.: (14)
a. Embedded Clauses: Quiero ir al cine. "I want to go to the cinema." b. Exclamative Sentences: / Visitor Islandia! Me encantaria. "To visit Iceland! I would love it."
In (14a), the embedded infinitive cannot be interpreted imperatively. It can only be entertained as a possibility. This seems to be the result of the fact that the infinitive is embedded into a higher-level explicature verb (quiero), which determines the type of attitude by which it is governed. In this case, the attitude is of wants, and therefore the infinitive cannot be subjected to
Non-Declarative Sentences in Spanish
107
standard imperatival force. The unavailability of imperatival force is further evidenced by the fact that it would be ungrammatical to construct a sentence with this higher-level explicature type of verb in the imperative, as shown below (where * denotes ungrammaticality, and "formal" polite verbal ending): (14'a) * Quiera ir al cine. "Want-IMP(formal) to go to the cinema" This suggests that not only is imperatival force unavailable in relation to the embedded verb, but also imperative mood is incompatible with the type of higher-level explicature verb used. This seems to suggest further that, in this type of construction with querer-iype verbs, imperatival force is excluded from the whole utterance, not just the embedded verb. Equally, in (14b) the infinitive within the exclamative sentence does not make an imperatival interpretation available. As in the previous case, the state of affairs described can only be envisaged as a possibility. This seems to be for the same type of reason. The exclamative mood under which the infinitive falls determines the attitude which directs the interpretation of the infinitive. This attitude is not one of imperatival force, but rather one of exclamative degree. Therefore, the infinitive cannot give rise to an imperatival interpretation in this type of context. Unlike in the previous example, the attitude in this case was not encoded explicitly by a verb, but rather it was communicated implicitly by the use of exclamative force. In the light of the preceding discussion, it can be concluded that imperatival interpretations of the infinitive can be blocked either pragmatically (by inaccessibility of appropriate contextual assumptions) or grammatically (by insertion in certain grammatical structures). So far we have considered imperatival infinitives in Spanish. However, are they also available in other languages?
5.1. Imperatival Uses of the Infinitive in Other Languages As was suggested above, the imperatival use of the infinitive is indeed not limited to Spanish. Other Romance and non-Romance languages allow it. Here some Galician and German examples will be discussed to illustrate this phenomenon further. In Galician the infinitive is used imperatively as widely as in Spanish (see Alvarez et al., 1986: 387). Let us consider some examples (where INF denotes infinitive): (15)
a. Phone Speaking Instructions: Falar a modo. "Speak (INF) slowly."
108
From Words to Discourse
b. At Doctor's Practice: Non Fumar. "Do not Smoke. (INF)" c. Shopping Message: Traer Ovos do Supermercado. "Bring (INF) Eggs from the Supermarket" d. Notice in Park: Manter Limpo. "Keep (INF) Tidy." As in Spanish, the interpretation of all these infinitival sentences carries imperative force, i.e. not just the force of mere possibility. For instance, (15a) instructs people to speak slowly. In particular, it does not just ask them to consider the mere possibility of doing so. Equally in (15b), the sign is at the very least requesting people not to smoke, if not ordering them to do so. It is certainly not just asking people merely to entertain the possibility of not smoking. The same type of explanation applies to the other examples. Moreover, as in Spanish, these sentences could be reproduced using the imperative without changing their directive import. Thus, consider the following imperatival versions: (15')
a. Phone Speaking Instructions: Falen a modo. "Speak (IMP) slowly." b. At Doctor's Practice: Non Fumen. "Do not Smoke. (IMP)" c. Shopping Messgage: Traian Ovos do Supermercado. "Bring (IMP) Eggs from the Supermarket." d. Notice in Park: Mantenan Limpo (o Parque) "Keep (IMP) Tidy."
The directive import of these imperative sentences is equivalent to the ones above. Both the imperative and infinitive sentences involve not just possibility, but rather achievability and desirability. Both involve interpretations of (non-)actions being instructed, requested, reminded, etc., which are typical of imperatival interpretations. The same is the case for German, where the infinitive is one of the various ways in which imperative force can be expressed (see Drosdowski et al., 1984:293). To illustrate, consider the following infinitival examples:
Non-Declarative Sentences in Spanish
(16)
109
a. On a Grass Lawn: Nicht aufden Rasen Treten. "Do not Step (INF) on Grass." b. Pedestrian Traffic Lights: Rot - Stehen / Grtin - Gehen "Red - Stay (INF) / Green - Go (INF)" c. On a Door Bell: Hier Klingeln "Ring (INF) here" d. On Film Instructions: Kameraanleitung Beachten. "Follow (INF) Camera Instructions."
In these examples the infinitive has been used, and again the intended interpretation goes beyond entertaining the mere possibility of the state of affairs being described. Some kind of imperatival interpretation is intended in each of them. For example, in (16a) the sign does not just ask pedestrians not to step onto the grass, rather it requires them not to do so (or a reprimand might follow). Equally, in (16b) pedestrians are not just invited to envisage the scenarios presented to them by the sign. Rather they are again required to follow the state of affairs described. The same type of imperatival interpretation is available in relation to the other examples. As in Spanish and Galician, the German infinitival sentences can be reproduced using the imperative without affecting the directive import of their pragmatic interpretation. Thus, consider the following imperatival versions of the above examples: (16')
a. On a Grass Lawn: Treten Sie Nicht aufden Rasen. "Do not Step (IMP) on Grass." b. Pedestrian Traffic Lights: Stehen Sie bei Rot / Gehen Sie bei Grun. "Stay (IMP) on Red / Go (IMP) on Green." c. On a Door Bell: Klingeln Sie Hier. "Ring (IMP) here." d. On Film Instructions: Beachten Sie die Kameraanleitung. "Follow (IMP) the Camera Instructions."
As in earlier cases, these imperatival versions convey the same import as the original infinitival examples. For instance, imperatival example (16'a) conveys the same directive import as infinitival example (16a), where in both cases pedestrians are required to stay outside the lawn.
110
From Words to Discourse
From Sperber and Wilson's point of view, the difference between the two is that, as far as the imperative case is concerned, the addressee is directed to an interpretation involving a state of affairs which is both achievable and desirable from the addresser's point of view. Thus, the use of the imperative here counts as a requirement, in that the sign is in a park under the control of the park authorities, who are empowered to set and enforce the rules. On the other hand, in the infinitive case, the addressee is only directed towards envisaging the state of affairs as a possibility. This is then interpreted further, in the light of available contextual assumptions about park regulations, resulting in a pragmatic upgrade of the possibility interpretation into an imperatival one. The same type of explanation obtains in the case of the other examples. Although no further languages will be discussed here, suffice it to say that the imperatival use of the infinitive is fairly widespread within European languages. Just to illustrate this point, consider the following examples in a number of languages taken from the instructions attached to an Afgachrome photographic Film: (17)
a. French: Faire developper immediatement lefilm expose. "Develop (INF) immediately the exposed film." b. Dutch: Belichtingsmeter op ISO 1000/21 instellen. "Set (INF) your exposure meter to ISO 100/21." c. Italian: Osservare le istruzioni allegate all 'apparecchio. "Follow (INF) the instructions provided with your camera."
In each of these examples an infinitive has been used: faire, instellen, and osservare, respectively. As in earlier cases, the interpretation of each of these infinitives goes beyond merely entertaining the possibility of the state of affairs described and involves some kind of imperatival interpretation, where assumptions about achievability and desirability are present. Although the line of explanation developed by Sperber and Wilson seems to account for the differences and similarities encountered here between imperatives and imperatival infinitives. As will be seen below, there seem to be some problems with their account.
5.2. Problems with Sperber and Wilson's Account So far we have observed that Sperber and Wilson's claims are that assumptions of potentiality and desirability constitute the basis for an imperatival interpretation of the infinitive. However, as anticipated above, there are indeed a number of problems with their account. The first is one that Sperber and Wilson leveled at Huntley themselves regarding possibility (Wilson and Sperber 1988:84). Namely, often imperatives are associated with some kind of (non-)action, and
Non-Declarative Sentences in Spanish
111
the gap between potentiality/desirability and action is often too wide to bridge pragmatically and, as a consequence, to explain the resulting directive interpretation. So, for example, simply to say that it is desirable not to smoke in a hospital is not necessarily going to persuade people not to do it, which is the intended (non-)action the hospital authorities wish to bring about. Thus, there must be more than just potentiality and desirability for the desired (non-)action to be brought about. In the case of the hospital, for example, the addressee must recover further assumptions about the desirability of not smoking, so as to encourage the adherence to the rules: (18)
a. If you smoke, the hospital will throw you out. b. If you smoke, patients will suffer the negative effects of smoking. c. If you smoke, you behave antisocially, etc.
In (18) there are some of the assumptions that might be available to the addressee. All of them lead to negative consequences, which might in turn persuade him not to smoke. It is these negative consequences that will make adherence to the rules desirable, and may ultimately carry the force of a warning. This type of example suggests that imperatival directive force requires more pragmatic interpretation for its derivation than seems envisaged by Sperber and Wilson, and may also play an important role in establishing degrees of desirability in the interpretation. Another problem with Sperber and Wilson's account relates to the universality of their claims. Thus, if it is the case that infinitive imperatival interpretations are dependant on the availability of contextual assumptions about potentiality and desirability, then why is it that English infinitives cannot be used imperatively even when those assumptions are available in the context? Consider the following English versions of the examples in (12): (19)
a. b. c. d.
In a hospital. Not to smoke. In a recipe book: To cut up the potatoes. In an instructions booklet: To plug in the printer. In a cashpoint: To introduce the card.
These English infinitival versions are not pragmatically acceptable and, as they stand, also ungrammatical. Therefore, the imperatival interpretations cannot be available. However, contextual assumptions about the potentiality and desirability of the state of affairs described in the English scenarios should be as easily available in the various contexts considered as they are in the Spanish cases. After all, instructions are as easily found in English hospitals, recipe books, and instructions manuals as they are in Spanish ones. This means that, from a contextual point of view, there should not be any differences between the English and Spanish cases (i.e. regarding accessibility of assumptions about potentiality and desirability, and hence interpretation of infinitives with imperatival force). This means further that at some level there must be differences between the two languages which must ultimately stem from linguistic factors, rather than purely contextual ones. If this is correct, Sperber and Wilson's predictions would not be
112
From Words to Discourse
borne out. The question now becomes whether English and Spanish (and other similarly behaved languages) are different at semantic or pragmatic level. Let us consider the first of these two possibilities: that Spanish infinitives are different at semantic level from their English counterparts. This possibility would mean that Spanish infinitives are ambiguous between infinitival meaning and imperatival meaning. As stated above, this possibility does not seem to be an optimal solution, since it creates a number of problems. Firstly, it flouts the Occam-razor principle in that it multiplies meanings unnecessarily. Secondly, as a direct consequence of ambiguity, semantic processing would be doubled, as every time an infinite was used, an addressee would have to disambiguate its meaning. Thirdly, it would make the infinitive equivalent to the imperative at the semantic level, which seems counterintuitive (since their morphology and their meaning are different). If we agree that it is unlikely that in Spanish (and other languages) the infinitive encodes imperative force semantically, then the difference must be pragmatic. Furthermore, if the difference is pragmatic, then crucially it cannot just be availability of contextual assumptions, since they are equally available in both English and Spanish contexts. For us to be able to interpret the Spanish infinitives imperatively, there must be further directions specified pragmatically, in addition to semantic ones, which enable us to find the relevance sought and intended. Moreover, the result must contribute to the explicit meaning of the utterance (as will be seen below). But so far, within the relevance-theoretic paradigm, this has not been allowed for. Encoding is only supposed to impinge on either the linguistic semantic level or pragmatic level, but not both. However, to accommodate the examples in Spanish (and other Romance and non-Romance languages), it has to be conceded that infinitives, in addition to encoding possibility semantically, must also encode pragmatic constraints regarding the type of contextual assumptions that are required in order to arrive at the intended interpretation.
6. A REVISED APPROACH From a theoretical point of view, the possibility of encoding pragmatic constraints (in addition to semantic ones) is not surprising, as the retrieval of appropriate contextual assumptions may require more than just pragmatic principles. The sheer number of potential assumptions available to an addressee makes utterance interpretation even more dependent on any possible linguistic instructions which may help select the appropriate assumptions. This aids pragmatic principles in the process of deriving the intended interpretation. The encoding of instructions to select appropriate assumptions is a common phenomenon in linguistic communication, such as in the case of conjunctions (see e.g. Blakemore 1992, chs. 7-8). From a relevance-theoretic point of view, conjunctions are normally analysed as encoding instructions regarding the direction in which relevance is to be sought, e.g.:
Non-Declarative Sentences in Spanish
(20)
113
Juan nofue al cine ya que llovia. "Juan didn't go to the cinema because it rained."
In the interpretation of this utterance the conjunction ya que ("because") encodes instructions to the addressee to retrieve one crucial contextual assumption which is necessary to understand the utterance, i.e.: (21)
Si llovia, Juan nofue al cine. "If it rained, Juan didn't go to the cinema."
The interpretation of (20) depends on us being able to retrieve the causal connection between the two events described as shown in (21), where if the premise (llovia) holds, then the consequent (nofue al cine) follows. The conj unction ya que allows us to retrieve this causal assumption, by encoding instructions to that effect, and thus enable us successfully to interpret the utterance. In a similar way the infinitive in Spanish encodes instructions about the type of assumptions which the addressee must look for in order to arrive at the intended interpretation. The difference between conjunctions and infinitives is that in the former case linguistic meaning only impinges on the pragmatic processing stage, whereas in the latter it impinges on both the semantic and pragmatic processing stages. This presents us with a new type of linguistic meaning not explicitly argued for so far in the literature. That is, linguistic expressions which encode both semantic and pragmatic meaning. Semantically, the infinitives encode possibility, pragmatically they can constrain further the types of contextual assumption sought. The next question is whether this meaning contributes to the explicatures or the implicatures of the utterance. One might expect that if the constraints are pragmatic they do not make a truthconditional contribution. One accepted test of explicit content contribution (see e.g. Kempson, 1977) is the use of conditionals. Let us take the hospital example as in (12a). Which of the following two situations do we regard as capturing what is being said in the example?: (22)
a. If people are required not to smoke, then the hospital will evict anybody who smokes, b. If people are suggested not to smoke, then the hospital will evict anybody who smokes.
If we regard (22b) as representing the situation envisaged by our example (12a), then the imperative force will not necessarily be part of the explicit content of the utterance. If, on the other hand, we regard (22a) as representing the situation envisaged, then the imperative force will be part of the explicit content of the utterance. It seems clear that the situation that best represents (12a) is (22a) and not (22b). If this is correct, then the imperative content of the interpretation of the utterance must contribute to the explicatures, rather than the implicatures. This would explain why in the interpretation we take the addresser (i.e. the hospital) to be conveying the force of an imperative not just as an implication, but rather as the explication of
114
From Words to Discourse
their message, i.e.: (23)
The hospital requires people not to smoke.
However, this interpretation is available in Spanish but not in English, which means that the recovery of the explicature has not been purely inferential. The Spanish infinitive must direct addressees, in the appropriate contexts, to look for potential and desirable assumptions which then become part of the (higher-level) explicatures of the utterance. This procedural content is not part of the meaning of the English infinitive, which does not allow imperatival interpretations in the same contexts. This means further that words can encode more than one type of meaning and thus, when considered cross-linguistically, the differences between the same grammatical categories may be due to pragmatic, not semantic factors. However, one question remains to be answered: why should an infinitive develop an imperatival use?
7. COMMUNICATIVE ADVANTAGES IN THE USE OF IMPERATIVAL INFINITIVES The imperatival use of the infinitive offers a number of advantages to the communicator. Firstly, it allows him to leave reference to addressees implicit (cf. ponga usted, explicit, vs. poner PRO, implicit). In Spanish, standard imperatives require the speaker to choose either the formal or the informal ending (e.g. coma, formal vs. come, informal). The infinitive allows referential neutrality and detachment, which is particularly suitable in the institutional settings in which it is found. There is referential neutrality because no individual is being referred to overtly. The subject in this case would be a big, un-indexed, PRO, rather than a small pro. There is detachment because the imperative force is derived pragmatically, not semantically, and suggests less personal involvement in the desirability of the state of affairs encoded. This is also suitable in an institutional context where assumptions about shared knowledge between the interlocutors (e.g. hospital-patients, manufacturer-user, bank-client) are generally weaker than in more personal contexts (e.g. between friends, parents-children, teacher-pupil, etc.), and thus it becomes more difficult to determine the origin and degree of desirability experienced in the situation. In other words, we rely on our general knowledge of those situations in order to derive the appropriate level of desirability. The second advantage associated with the imperatival use of the infinitive is that it is less direct than the imperatives proper. That is to say, it leaves assumptions about desirability implicit for the addressee to access and use in the interpretation process. This, paradoxically, increases the mutual understanding between addresser and addressee, as there is an increase in the knowledge that has to be shared between the two interlocutors for the imperatival interpretation to go through. That is, there is a greater implicit appeal to, and increased
Non-Declarative Sentences in Spanish
115
accessibility of, appropriate contextual assumptions, particularly those which relate to rules governing the types of social environment envisaged here (e.g. rules in hospitals, unspoken agreements between friends, etc.). There is a third advantage with the use of the infinitives as imperatives, whose significance varies across languages, namely, processing cost. The imperative infinitive involves less processing cost than the imperative in a number of ways. Firstly, the infinitive is generally not used in conjugated form, unlike the imperative which is in most languages, including Spanish. This adds processing costs to the semantic decoding of the sentence, as the addressee must decode the type of subject being used (e.g. formal vs. informal, singular vs. plural). In the case of the infinitive, a general empty category PRO is readily supplied without having to decode any politeness or number features. Secondly, the infinitive is more amenable to having objects left out in the predicate. Take, for instance, the following example in Spanish: (24)
At a Park: a. Mantener Limpio. "Keep (INF) Tidy." b. Mantengan Limpio el Parque. "Keep (IMP) the Park Tidy."
In Spanish, the infinitival version in (24a) is amenable to having the direct object (el parque) omitted and to recover it from the context. However, the imperative version is less amenable to the same omission process and needs the direct object to be explicit. This means that imperatives again require further processing costs in the semantic decoding stage of the interpretation, as predicate objects must be explicit and, therefore, decoded. In the infinitival version, this is not necessary, as it is supplied as a readily available contextual assumption. A third advantage of the imperative infinitives in terms of processing cost is more language specific. In German, unlike Spanish and Galician, the syntax of some verbs is more complex when they are conjugated than when they are not. This is particularly so in the case of trennbar verbs (i.e. splittable verbs). In this type of verb, verb meaning (and in turn sentence meaning) can only be fixed, when the split part of the verb at the end of the sentence is processed in conjunction with the verb root which appears towards the beginning of the sentence. To illustrate, let us consider an example: (25)
On the Front Door of a House: a. Keine Reklame Einwerfen. "Do not Leave (INF) any Publicity." b. Werf en Sie Keine Reklame Ein. "Do not Leave (IMP) any Publicity."
116
From Words to Discourse
Example (25a) uses the infinitive and the verb appears un-split at the end of the sentence (einwerfen). However, in example (25b) the imperative form of the verb is used and, therefore, the verb must be conjugated. As a result, the splittable verb einwerfen must have its prefix emmoved to the end of the sentence with the root remaining at the beginning. This syntactic complication adds further processing costs to the imperative sentence in relation to the infinitival one. Moreover, this is corroborated by the fact that in German children learn to communicate imperative (and non-imperative) force by using the infinitive, which is easier to acquire and use (native informant). All these arguments in favor of lower processing costs for the infinitive (as opposed to the imperative) predict that preference of the imperative infinitive over the imperative is particularly sensitive to considerations of relative processing cost. That is to say, its use is expected in situations in which low processing costs are crucial to communicative success. Indeed, this seems to be the case. Imperative uses of the infinitive are found in contexts where low processing costs are critical, for example: at pedestrian traffic lights, where simple messages for quick reactions and decisions are highly desirable and at times life saving; product instructions, where space is of the essence and often a large amount of information has to be packed in a very small space (e.g. photographic films, printer cartridges, etc.); at points of service, where a quick, easily readable message is required (e.g. cashpoint machines, telephone booths, etc.); reading instructional material, where simple, clear, direct, instructions allow faster and easier processing (e.g. recipes, manuals, etc.). The possibility of packing equivalent amount of information by providing short, easily readable messages is an obvious advantage, particularly in societies in which speed and efficiency are of paramount importance to information processing. The various advantages and pressures described here may ultimately contribute to the conventionalisation of the imperative use of the infinitive. Moreover, this may be one of the routes available in some languages for associating pragmatic imperative force with the infinitive and making it more accessible during utterance interpretation. The fact that some of these advantages and the resulting pressures are not applicable to other languages may explain, to some extent, why they do not allow the imperatival use of the infinitive. For example, in English there is no distinction between formal and informal verbal endings, nor is the presence of imperative subjects normally required or even allowed. Similarly, the English imperative, unlike the Spanish one, is more amenable to having predicate objects omitted, which further reduces the need for an imperatival infinitive use. In general, all these factors would seem to count against the availability of this infinitival use in languages like English, since the imperative itself exhibits all (or most of) the advantages discussed in this section.
Non-Declarative Sentences in Spanish
117
8. CONCLUSION In this article it has been argued that Spanish infinitives can be used with imperative force in addition to their more common infinitival use. The imperative use sets them apart from English infinitives, in that the latter do not give rise to imperative interpretations, even when the contexts are the same as in the Spanish examples. It has also been shown that imperative interpretation of Spanish infinitives is only possible when they are used in contexts which make instructional assumptions, and their attendant notions of potentiality and desirability, very accessible (e.g. at cashpoints, in manuals, hospitals, etc.). This suggests that the imperative import is not semantic in nature, but rather pragmatic. It suggests further that the meaning of the infinitive in Spanish is complex, including semantic meaning proper and pragmatic constraints. The latter arise from the inclusion of assumptions of potentiality and desirability in context, which the infinitive procedurally encodes. The inclusion of these assumptions in the explicit content of the utterance means that this side of the infinitive's meaning contributes to the explicatures of the utterance. From a cross-linguistic point of view, the difference between the Spanish and the English infinitive is that in the Spanish case there are procedural instructions encoded for the addressee to seek assumptions of potentiality and desirability in the context, whereas in the English case such procedures are not part of the meaning of the infinitive. This means that, in English, without the linguistic help needed to direct addressees towards specific assumptions in the context, those addressees face a gap between possibility, on the one hand, and potentiality, desirability, and imperative force, on the other, which is too wide to bridge, and which will ultimately make the imperative interpretation in the English case not available to him. From a theoretical point of view, Spanish infinitives (as well as infinitives in other Romance and non-Romance languages) unveil a new two-dimensional category of linguistic meaning. From a semantic point of view, they encode conceptual and truth-conditional meaning. From a pragmatic point of view, they encode procedural and truth-conditional meaning. The procedural pragmatic stage may or may not be acted upon depending on whether it does in fact contribute to relevance. Thus, Spanish infinitives encode two types of meaning, one that impinges on the semantic representation of the sentence (semantic dimension), and another that impinges on the pragmatic interpretation of the utterance (pragmatic dimension). Both ultimately play a part in the explicatures of the utterance. One interesting area for further research is to establish how common and wide the Spanish and English behaviors are respectively. As seen above, many European languages seem to allow the imperative use of the infinitive. English, in this respect, seems to stand out in its behavior. However, it would be surprising if English were to be unique in this respect. One possibility may be that this phenomenon is found more regularly in non-European languages. The answer to this question though would require further research, which is outside the scope of this article.
118
From Words to Discourse
REFERENCES Alvarez, R., X.L. Regueira, and H. Monteagudo 1986. Gramatica Galega. Vigo: Galaxia. Bird, G. 1994. "Relevance Theory and Speech Acts." In Tsohatzidis (ed.) 1994, 292-311. Blakemore, D. 1987. Semantic Constraints on Relevance. Oxford: Blackwell. Blakemore, D. 1992. Understanding Utterances. Oxford: Blackwell. Blass, R. 1989. "Are There Logical Relations in a Text?" Lingua 90. 91-110. Butt, J. and C. Benjamin. 1988. A New Reference Grammar of Modern Spanish. London: Arnold. Clark, B. 1993. "Relevance and 'pseudo-imperatives." Linguistics and Philosophy 16. 79-121. Dancy, J., Moravcsik, J. and C. Taylor. 1988. Human Agency: Language, Duty and Value. Stanford, CA: Stanford University Press. Drosdowski, G., ed. 1984. Duden: Die Grammatik. Mannheim: Bibliographisches Institut. Grice, H.P. 1989. Studies in the way of words. Cambridge, MA: Harvard University Press. Hare, R. M. 1970. "Meaning and Speech Acts." Philosophical Review 79. 3-24. Huntley, M. 1984. "The Semantics of English Imperatives." Linguistics and Philosophy 1. 103-33. Jary, M. 2001. "Mood in Relevance Theory: A Re-analysis Focussing on the Spanish Subjunctive". Ms. Kempson, R. 1977. Semantic Theory. Cambridge: Cambridge University Press. Levinson, S. 1983. Pragmatics. Cambridge: Cambridge University Press. McGinn, C. 1977. "Semantics for Non-indicative Sentences." Philosophical Studies 32. 301-11. Papafragou, A. 1988. "Inference and Word Meaning: the Case of Modal Auxiliaries." Lingua 108. 1-47. Fountain, C. 1998. "Person and Voice in the Spanish Infinitive." Bulletin of Hispanic Studies LXXV. 393-410. Searle, J. 1969. Speech Acts. Cambridge: Cambridge University Press. Schmerling, S. 1982. "How Imperatives Are Special, and How They Aren't." Chicago Linguistics Society: Parasession on Nondeclaratives, 202-18. Sperber, D. and D. Wilson. 1995, 2nd ed. Relevance: Communication and Cognition.Oxford: Blackwell. Tsohatzidis, S. 1994. Foundations of Speech Act Theory: Philosophical and linguistic perspectives. London: Routledge. Wilson, D. and D. Sperber. 1988. "Mood and the Analysis of Non-declarative Sentences." In Dancy, Moravcsik and Taylor (eds.) 1988, 77-101.
119
6
ASPECT AND SITUATIONS: A SITUATION SEMANTICS ACCOUNT OF THE SEMANTIC VARIABILITY OF SPANISH '^/.-CLAUSES' Luis Alonso-Ovalle, University of Massachusetts at Amherst
1. THE PUZZLE Frege noticed that free adjuncts and absolutes represent an intricate case of extra meaning that can neither be attributable to overt expressions nor to Gricean effects (Frege 1892). * Thus, (la), for instance, is semantically equivalent to either (Ib) or (Ic), among many other possibilities. (1)
a. Napoleon, who recognized the danger to his right flank, himself led his guards against the enemy position. b. Napoleon, in spite of his recognizing ... c. Napoleon, since he recognized ...
Since their meaning does not seem to be directly composed out of the syntactic combination of their parts, absolutes and free adjuncts directly challenge the Compositionality Principle, which fuels model-theoretic semantics (Partee 1984). Its strongest version, due to Hintikka (1980), is found in (2).
* For useful comments, encouragement and help, I am most grateful to Angelika Kratzer, Barbara Partee and the participants in Kratzer's UMass Seminar on the semantics of embedding. Usual disclaimers apply.
120
(2)
From Words to Discourse
It is only the meaning of the parts and their syntactic mode of combination that matters.
Nevertheless, Stump (1985) has shown that a compositional analysis for these constructions is indeed possible. This paper aims to support his claim by providing such an analysis for a kind of infinitival free adjunct in Spanish: ^/-clauses (henceforth ACs), free infinitival adjuncts of the form of those under (3). The semantic variability of a/-clauses is interestingly restricted to three possible interpretations. First, the line of work associated with David Lewis (1975) and Angelika Kratzer (1991a,b) has shown that certain adverbial clauses (mostly //-clauses) can restrict the domains of various operators. This is indeed the case for ACs with respect to quantificational adverbs like siempre ('always') or ram vez ('rarely'), as (3) shows. (3)
Pedro siempre silba al conducir. Pedro always whistle:3sPres AL to drive "Pedro always whistles when he drives."
Second, they can also be adsentential modifiers. The glosses under (4) show that, as adsentential modifiers, they can have either a 'causal-explicative' or a temporal reading. (4)
Al sonar el telefono, Pedro se callo. AL to ring the phone Pedro shut up:3sPast "Peter shut up when the phone rang." "Peter shut up because the phone rang."
But the interesting property of ACs is that their semantic variability is 'aspectually restricted', as shown by (6-12). First, ACs cannot be restrictors if they are headed by perfect infinitives (6) and/or i-level predicates (7): (6)
(7)
Pedro (siempre/rara vez) canta al haber conducido. Pedro (always/rarely) sing:3sPres AL to have driven *"Pedro always sings after driving."1 Pedro (siempre/rara vez) canta al ser alto. Pedro (always/rarely) sing:3sPres AL to have driven *"Pedro always sings when he is tall."
Moreover, when they are adsentential modifiers, ACs headed by both i-level predicates (8) and perfect infinitives (9) only license causal-explicative readings, excluding temporal ones:
' The mark "*" indicates that the reading is unavailable.
Aspect and Situations
(8)
(9)
121
Al ser tan burro enciende las luces por el dia. AL to be so idiot switch:3s the light during the day *"When he is so idiot, he switches the light during the day." "Because he is so idiot, he switches the light during the day." Al haber sido quemado, este elemento genera residues cancerigenos. AL to have been cremated, this element generate:3s residues cancerous *"When it has been cremated, this element generates cancerous residues." "Since it has been cremated, this element generates cancerous residues."
Finally, the most natural reading for non perfect s-level predicates as adsentential modifiers is the temporal one, even when causal readings, as in (1 Ib), are not excluded. (10)
(11)
Coello se contradijo al ser interrogada por segunda ocasion. Coello rfxcl contradict:3sPast AL to be asked for second time "Coello contradicted herself when/while she was asked a second time." a. Al correr, Juan se cayo (Rigau, 1995). AL to run, Juan rfxvcl fall:3sPast b. "Juan fell down because he ran." c. "Juan fell down while he was running."
That the availability of a causal or temporal reading for an AC is dependent on the aspectual value of the predicate is shown by the fact that, if a predicate has both an i-level and an s-level reading, the i-level reading yields a causal reading and the s-level a temporal one. The predicate serjoven ('to be young'), for instance, is one of those and, therefore, (12) is ambiguous between a temporal and a causal reading. (12)
a.
Piensan que al ser joven tienes que ser drogadicto. (They) think that AL to be young (you) have to be drug addict b. "They think that you have to be a drug addict when you are young." c. "They think that you have to be a drug addict because you are young."
Table 1 summarizes the empirical generalizations presented so far. The explanation of its contents constitutes the minimal goal of any analysis of ACs. The one put forth in the next sections proposes that Spanish infinitives can be either VPs or Aspectual Phrases (AspPs) and that al merges with either. When it merges with VP-infmitives, it is semantically vacuous and introduces properties of situations, which, unlike AspPs, are suitable arguments for quantificational adverbs.
122
From Words to Discourse
Table 1
The semantic variability ofACs. Restrictors (Mod.) temporal (Mod.) causal
Imperfect S ^
i-level * *
^
^
Perfect * x S
The organization of the paper runs as follows: section 2 introduces the basics of a Kratzerian Situation Semantics (Kratzer, 1989), section 3 defends that Spanish infinitives denote sets of minimal situations and shows how to capture the intuition that perfect infinitives denote sets of completed minimal situations, whereas imperfective infinitives denote sets of uncompleted minimal situations. Section 4 defends that al can merge with VPs that, unlike AspPs, are suitable arguments for quantificational adverbs. Section 5 shows how to capture the temporal and causal readings of infinitive AspPs and, finally, Section 6 summarizes the proposal.
2. THE TOOLS: KRATZERIAN SITUATION SEMANTICS For reasons to be seen in the next section, the analysis of ACs will be cast in a Kratzerian Situation Semantics (Kratzer, 1989, 1990).2 This section is devoted to presenting its basics. First, throughout this paper, a situation-based ontology is assumed. Specifically, a model for interpreting natural language is a tuple M := <S,D,W,<, [[ ]]>, where: (13)
S is the set of possible situations. D is the set of possible individuals. D c S W is the set of possible worlds, maximal elements with respect to <. < is a partial ordering on S. [[ ]] is the interpretation function.
The partial ordering on S satisfies at least the following condition: for all s e S there is a unique s'e S such that s < s' and for all s"e S: if s' < s" , then s" = s1. Notice, then, that < imposes a mereological summation structure to S, with each world being the supremum of a complete join semilattice and each situation being part of a world. This on its turn implies that one individual can only be part of one possible world, which requires adopting some version of the counterpart theory, as advocated by Lewis (1968, 1986), to speak about possible alternatives of an actual individual. 2
Portner (1992) includes an interesting comparison between Kratzer's system and the model proposed by Barwise and Perry (1983).
Aspect and Situations
1 23
The type theory of Kratzer's Situation Semantics is standard, except for the fact that the domain of expressions of type e, the set of individuals, is a subset of S and the domain of expressions of type t is p(S). Propositions, then, are sets of situations: the set of situations in which the proposition holds. Finally, the notion of eventualities that exemplify propositions will be relevant for our purposes (Kratzer, 1990, 1998).3 Intuitively, an eventuality that exemplifies or supports a proposition is a situation in which the proposition is true and is small enough so as not to contain anything irrelevant to its truth: (14)
For each s e S and any p e £>(S), 5 is an eventuality that exemplifies p iff for all s' such that s1 < s, there is an s" such that s' < s" < s, and s" is a minimal situation in which p is true.
In order to map propositions into eventualities that exemplify them, I will make use of an operator 4 in the intermediate typed language (see Kratzer, 1999), where: (15)
[[p]] = {s: s exemplifies p}
3. SPANISH INFINITIVES DENOTE SETS OF MINIMAL SITUATIONS In order to posit an analysis of the semantic variability of ACs, after having introduced the basic tools, we still need to introduce a minimal semantics for Spanish infinitives.
3.1. Prepositional and Eventive Infinitives Vendler (1967) noted that English gerunds have either prepositional ( as in (16a), which is intuitively equivalent to John denied that he studied semantics), or event-like denotations (16b). (16)
a. John denied studying semantics. b. Studying semantics was interesting.
The observation can be extended to Spanish infinitives. (17a) is equivalent to its prepositional paraphrase in (17b), whereas the infinitive in (17c) seems to denote an eventuality of studying semantics.
3
In the sense of Bach (1986) that covers both states and proper events.
124
(17)
From Words to Discourse
a. Juan nego haber estudiado semdntica. Juan deny:3sPerPas to have studied semantics "Juan denied to have studied semantics." b. Juan nego que hubiera estudiado semdntica. Juan deny:3sPerPas that (he) have:3sPasSubj studied semantics "Juan denied that he had studied semantics." c. Fue interesante estudiar semdntica Be:3sPerPas interesting to study semantics "Studying semantics was interesting."
Under the position advocated by Vendler (1967), Davidson (1967) and Parsons (1990), among others, there is no direct relation between events and propositions. Therefore, (17) would force us to admit that Spanish infinitives are ambiguous between prepositional and eventive readings. However, by using the tools introduced in Section 2, Portner (1992) closed the gap between propositions and events by treating English gerunds as uniformly denoting a specific type of propositions, those formed by minimal situations that can individually play the role of events. This makes Situation Semantics a specially promising framework for the analysis of ACs. Throughout the paper, I will adopt Former's position and consider that Spanish infinitives denote properties of minimal situations. Specifically, I want to suggest that the infinitival morpheme gets translated into the intermediate language as the Kratzerian minimality operator •I. Consequently, by assuming that VPs denote properties of situations, the denotation of comer una tarta ('to eat a pie') would be the set of eventualities that exemplify the proposition {s: eatsa-pie'(x)(s)}. 4 (18)
Tr (comer una tarta) = ^A,s[eats a pie' (x) (s)] Tr (-ar)
Tr ( VP)= A,s[eat-a-pie' (x) (s)] com- una tarta
3.2. Completed vs. Uncompleted Events: Infinitives as Aspectual Phrases What about the aspectual values of infinitives and gerunds? We know that sometimes -ing forms are interpreted perfectively, as in (19a), and sometimes, as in (19b), imperfectively. (19)
4
a. Mary celebrated climbing the mountain, b. Mary enjoyed climbing the mountain.
The free variable is intended to stand for the trace of the VP-internal subject. This is just a simplification with no significant role in what follows.
Aspect and Situations
125
Portner's treatment of these aspectual differences relies on the notion of point of view: perfect interpretations of gerunds correspond to an external perspective of the minimal situation that it denotes, whereas imperfect gerunds correspond to an internal perspective of the situation. This intuition is formally captured by means of a relation between the situation introduced by the verb of the main clause and that associated with the gerund. Even when I will depart from Portner's formalization of the aspectual values, I propose to extend this treatment to Spanish infinitives in the following sense: whereas /za&er-infinitives denote complete situations, infinitives without an overt aspectual auxiliary are unmarked for completeness, as the following minimal pairs show: (20)
(21)
a. A Pedro le divirtio comer la manzana. to Pedro cl enjoy:3sPast to eat the apple "Peter enjoyed eating the apple."
(completed/uncompleted event)
b. A Pedro le divirtio haber comido la manzana. to Pedro cl. enjoy:3sPast to have eaten the apple "Pedro enjoyed having eaten the apple." a. Al rector le enorgullece dar la conferencia. to the Chancellor cl. is proud of to give the lecture "The Chancellor is proud of giving the lecture." b. Al rector le enorgullece haber dado la conferencia. to the Chancellor cl. is proud of to have given the lecture "The Chancellor is proud of having given the lecture."
(completed event)
(uncompleted event)
(completed event)
This is clearly seen with verbs that denote attitudes towards an event that is over, as in (22): in all these cases an imperfect infinitive is excluded. (22)
a. Conmemoran haber viajado a Roma. commemorate :3pl to have traveled to Rome "They commemorate having traveled to Rome." b. ^Conmemoran viajar a Roma.
Moreover, when the verb expresses an attitude towards an event that one must be involved in or in direct perceptual contact with, the perfect infinitive is definitely ruled out: (23)
(24)
Vi las/lores crecer /*haber crecido. I saw the flowers grow / * to have grown "I saw the flowers grow." Oye la orquesta tocar / *haber tocado. hear:3s the orchestra to play / *to have played
126
From Words to Discourse
"He hears the orchestra playing." Then, we need to formalize the intuitively valid observation that links perfect infinitives with completed events. Instead of reference situations, as in Portner (1992), I will make use of temporal references in order to assume a unified treatment of aspect. Klein (1994) characterized aspect informally as a relation between intervals of time: the running time of an eventuality, what he calls a situation time and the time (interval) about which a sentence makes an assertion, what he calls a topic time. Kratzer (1998) has formalized this notion by using a predicate of eventualities (time) that gives you the running time of an eventuality. Johnston (1995) makes use of a similar device, inspired by the temporal trace function in Krifka (1989) and the running time function in Lasersohn (1990). The analysis proposed in the next sections will use such a device: (25)
[[ time' (s)]] = {t: s is running at t}
Now, if we let aspectual operators existentially close the eventuality variable provided by VPs, a perfect AspP-infinitive would denote an event completed with respect to a reference time t, if the running time of the situations of which the property expressed by the VP can be truly predicated is previous to t. A non-perfect infinitive would be unmarked for completeness if we let the running time of the situation it denotes be non properly included in the reference time. Aspectual operators, which presumably appear in the syntax as heads of the AspPs, map properties of situations (in the case of infinitives of minimal situations) and yield properties of times as a result. In the line of Kratzer (1998), I will also posit existential closure of the situation argument as part of the semantics of the aspectual operators. Haber, the aspectual auxiliary, would be considered the perfect aspectual operator, whose non perfect counterpart is covert in Spanish.5 Assuming all this, we are ready to characterize formally the notion of complete and uncompleted events as follows: (25)
a. Tr ( haber ) =A.pX,t3s [-Ip (s) & time (s) < t] b. Tr ( 0 ) - XpXtBs [Ip (s) & t c time (s) ]6
The semantic tree in (26) is an illustration for the AspP haber fumado ('to have smoked'): (26)
Tr ( AspP ) = A.t3sf [Asmoke1 (x) (s1) & time (s1) < t ] Tr( haber) Tr(FP)= Xs[smoke'(x) (s)] s) & time (s) < t ] XX fum-
5
This operator will be represented in the syntax by means of 0'. For all sets of times P, Q, [[ P c Q ]] = 1 iff every t included in P is also included in Q. Caveat: / would be considered a singleton, so the same semantics are extended to expressions like t cP.
6
Aspect and Situations
127
The denotation of (26) is a set of times such that there is an eventuality of x's smoking whose running time is previous to them. In (27) we have an example of a non-perfect AspP. Its denotation is also a set of times: the set of times such that there is an eventuality of x's smoking whose running time includes them. (27)
Tr (AspP ) = A,t3s* [ismoke1 (x) (s1) & t c time (s1) ]
Tr ( 0-ar ) Tr ( VP) = Xs[smoke' (x) (s)] =XpXt3s[4p (s) & t c time (s) ] /\ fum-
4. Acs AS RESTRICTORS We have seen that infinitives without a perfect auxiliary may be either VPs (18) or non perfect AspPs (27). We are now in the position of explaining why only non perfect s-level predicates can function as restrictors of frequency adverbs. For consider the semantic structure of adverbial quantification (henceforth a-quantification, as in Partee, 1995). In spite of the successful KampHeim approach that claimed that quantificational adverbs are unselective binders (Kamp, 1981; Heim, 1982), Von Fintel (1994, 1995) treats adverbs of quantification as uniformly quantifying only over situation variables. Following Berman (1987) and Heim (1990), he points out that, within a Kratzerian Situation Semantics, adverbs of quantification can be seen as denoting second-order relations between two sets of situations, i.e. between two propositions, the first of which, as usual, can be either contextually supplied or given by a restrictive clause, as in (28). (28)
a. Always, if it rains, Peter takes the bus. b. S always' {s: rain' (s)} {s': Peter-take-the bus (s')}
Intuitively, always is the situation-sensitive correlate of every. Hence, (28) will be true if the set of situations contained in the proposition it rains is a subset of the set of situations in which Peter takes the bus, much as Every cat smiles is true if the set of individuals that are cats is a subset of the set of individuals that smile. Interestingly enough, however, Von Fintel has shown that this is also a naive approach. First, as stated, the proposition denoted by an adverbially quantified statement will either contain all situations or none, because the membership condition on the proposition does not mention the situation whose membership is decided. This has been
128
From Words to Discourse
solved by proposing that the situations in the restrictor have to be evaluated in the actual world (Von Fintel, 1994:19). Second, the mereological structure of the domain of situations in the model makes it very difficult to count situations and, therefore, to compare the cardinalities of the two propositions. This has been solved by restricting the semantics to take care only of minimal situations in the restrictor: more specifically, minimal situations in the restrictor extendable to situations in the nuclear scope of the quantifier. This latter condition makes VPinfinitives suitable arguments for a-quantifiers, since we have proposed that the infinitival morpheme contributes a minimality requirement to the denotation of the VP. Consequently, we are able to claim the following generalization: (29)
a. VP-infinitives are suitable restrictors for Q-adverbs. b. AspP-infinitives cannot be suitable restrictors for Q-adverbs.
As a matter of illustration, consider (30). (30) includes a non-perfect infinitive that may function as a restrictor of a Q-adverb. Applying the proposed translation for VP-infinitives and assuming that al is semantically vacuous, the translation of the infinitive in (30a) is (30b), accepting that the denotation of the e-type argument of the infinitive is somehow controlled (an assumption currently made, see Stump, 1985). (30)
a. Juan siempre silba al conducir. Juan always whistle:3s AL to drive "Juan always whistles when driving." b. Tr (al conducir) = iA.s[ Juan-drive1 (s)]
Sentence (30a) will have the tripartite LF in (31), translated into (32):
(31)
(32)
S siempre al PRO, conducir Juan, silba S Vs Xs [Juan-drive1 (s)] Xs" [Juan-whistles' (s" )]
The LF (32) yields the desired reading: it will be true iff the set of minimal situations of Juan driving is a subset of the set of situations of Juan whistling, i.e. iff the set of situations of Juan whistling includes all minimal situations of Juan driving. Now it is not difficult to derive the prohibition against perfect infinitives and i-level predicates being restrictors. (33) repeats the denotation suggested for perfect (33a) and non perfect (33b) AspPs.
Aspect and Situations
(33)
129
a. Tr ( haberfumado ) = Xt3s' [Asmoke1 (x) (s1) & time (s1) < t ] b. Tr ( 0fumar) = X.t3s' [Ismoke1 (x) (s1) & t c time (s1) ]
From this perspective, it is trivial to explain why these phrases cannot be restrictors: they are just not a suitable semantic type (). It will be clear that aspectual operators existentially close the set of situations that they take as arguments and quantificational adverbs cannot be applied to arguments of type (where i is the type of times ). The same explanation can be given for i-level predicates. Chierchia (1995) considers them inherently generics and does so by letting them incorporate an aspectual habitual operator in the lexicon. If that is the case, then i-level predicates inherently lack the possibility of being restrictors to quantificational adverbs. (34) tries to capture the fact that i-level predicates are temporally stable by letting the running time of an i-level eventuality include any other salient sufficient time. (34)
Tr (ser medico ) =XtVt' [t * t' & 3s' [doctor1 (x) (s1)] & t c time (s1) & t' c time (s') ]
This proposal predicts that every alleged non-perfect s-level predicate that serves as a restrictor is actually a VP-infinitive. And this prediction can be empirically borne out. First, (35) shows what happens if the eventuality variable of the VP-level infinitive is on its turn bound by an aquantifier: the restrictor reading is blocked and the only available reading is the adsentential modifier one. (35)
a. Juan siempre silba al conducir. Juan always whistle:3s AL to drive "Juan always whistles when driving." b. Juan siempre silba al conducir con frecuencia. Juan always whistle:3s AL to drive with frequency "Juan always whistles since he drives frequently." *"Juan always whistles when he drives frequently."
Furthermore, the presence of any element generated in the syntax above the AspP, like temporal adverbs (Ojea, 1994) or negation itself, blocks the restrictor reading, showing that AspPs are not suitable arguments for the a-quantifier. (36)
a. Juan siempre silba al no conducir. *"Juan always whistles when he does not drive." "Since he doesn't drive, Juan always whistles."
130
From Words to Discourse
b. *Juan siempre silba al conducir el ano pasado. "Juan always whistles since he drove last year." Notice that, as expected, restrictors are compatible with VP-adverbs, as (37) shows. (37)
Juan siempre silba al conducir despacio/despreocupadamente... Juan always whistle:3s AL to drive slowly, carelessly... "Juan always whistles when driving slowly, carelessly..."
5. ADSENTENTIAL MODIFIERS The previous discussion leads us to accept the existence of a semantically vacuous al that merges with VP-infinitives, which are suitable arguments for a-quantifiers. I will call it al,. Al, takes sets of minimal situations as syntactic arguments and is semantically vacuous. There are still certain semantic characteristics in this construction that I will not try to deal with here. For instance, it has to be determined why it is the case that in these constructions there are no head restrictor readings as those pointed out in Johnston (1995). Nevertheless, I will now show how to derive the adsentential modifier readings for ACs and their sensitivity to the aspectual values of their infinitives. The strategy Former (1992) (inspired by Stump, 1985) employed for free adsentential modifiers is to assume the existence of a covert operator that denotes a relation between a situation and a proposition. This relation is present in the grammar as a free variable over possible relations of this type. The possible values that the relation takes restrict the interpretation of the modifier. I will sketch here an analysis that makes use of a relation between properties of times and propositions. We have seen that AspPs denote properties of times and that aspectual operators close existentially the eventuality argument. We have also seen that al, when part of an adsentential modifier (henceforth al2), takes AspPs as its argument. I will propose in the spirit of Stump (1985) that al2 is not semantically vacuous, but requires that the sets of times denoted by the AspP it takes as its argument, be included in the running time of the situations in the proposition denoted by the main clause: (38)
Tr(a/2) = A,PA,pA,s'[p(s')«fetime(s') cP]
Notice what happens if al2 combines with a non-perfect infinitive, say sonar el telefono ('to ring the phone'). We obtain: (39)
Tr ( al2sonar el telefono ) = A,pA,s' [p (s1) & time (s1) e [X.t3s[lring' (the phone') (s)] & t c time (s) ]]
Aspect and Situations
131
(39) takes a proposition as its argument and says that the running time of any situation in the proposition is included in the set of times that are included in the running time of an eventuality of ringing the phone. Let this proposition be (40), where I neglect temporal/aspectual content. (40)
Tr ( Pedro se callo )= A,s'" [shut-up1 (Pedro1) (s'")]
Then, by regular functional application we obtain (41). (41) denotes a proposition, namely the set of situations such that they are situations of Peter shutting up and its running time is included in a temporal interval: the set of times t such that there is an eventuality of ringing the phone whose running time includes them. (41)
Tr (al2 sonar el telefono ( Tr (Pedro se callo) ) ) = Xs1 [shut-up1 (Pedro1) (s1) & time (s1) c X.t3s [^ring1 (the phone1) (s) & t c time (s) ]]
(41) summarizes what we have called the 'temporal reading' for ACs as adsentential modifiers. Non perfect AspP infinitivals, then, tend to license this temporal overlapping reading. However, notice that (41) does not exclude a possible causal inference. If a causal chain between the eventuality of ringing the phone and a situation of Peter shutting up is available, then (41) can also have what we have been calling a 'causal-explicative meaning'.7 Finally, notice also that the existential quantification over eventualities included in the AspP denotation forces factivity (Stump, 1985). The denotation of (41) forces us to accept the existence of an eventuality of ringing the phone. Let us see now what happens in the case of perfect AspP-infinitives. (42a) illustrates the type of denotation for a perfect AspP. It denotes a set of times t such that there is an eventuality of x being burnt whose running time is previous to t. (42b) illustrates the denotation of an AC that takes this type of AspP as argument. (42)
a. Tr (haber sido quemado)= X.t3s' [^being bumf (x) (s1) & time (s1) < t ] b. Tr ( al2 haber sido quemado ) = XpA,s" [p(s") & time (s") c A,t3s! [ibeing burnt1 (x)(s!) & time (s1) < t ]]
Now, if we apply a proposition like (9) ( este elemento genera residuos cancerigenos ('this element generates cancerous residues')), we obtain the following denotation:
7
Javier Gutierrez-Rexach (p.c.) has suggested that a normality condition on extendibility of situations can shed some light on the availability of causal readings for some cases. Obviously, this topic deserves far more attention.
132
(43)
From Words to Discourse
Tr (al2 haber sido quemado (Tr (este elemento genera residues cancerigenos ) ))= ^s'[this-element-generates-cancerous-residues' (s1) & time (s1) c [A.t3s[>lbeing burnt' (this element') (s)] & time (s) < t ]]
(43) states that the denotation of (9) is a set of situations of this element generating cancerous residues whose running time is included in the set of times preceded by an eventuality of this element being burnt. It can be seen that a causal link is easily made here on the basis of temporal precedence. And, obviously enough, temporal overlapping is not a possible reading. Finally, let us see what happens when the AspP al2 takes as its argument an i-level predicate. (44a) illustrates the type of denotation suggested for i-level predicates like ser medico ('to be a doctor'). For our purposes, it suffices to assume that it denotes a temporal interval such that the running time of the eventuality of x being a doctor includes any time (of course, the domain of times is to be pragmatically restricted). (43b) shows the denotation of an AC with an i-level predicate. (44)
a. Tr (ser medico ) =ltVt' [ t * t'& 3s' [doctor1 (x) (s')] & tc time (s1) & t'c time (s1) ]] b. Tr ( al2 ( Tr (ser medico ) ))= A.pX,s'[p(s') & time (s1) c XtVt'[ t * t'& 3s[>koctor'(x) (s) & t c time (s) & t'c time(s)]]]
Now, if we apply (44b) to a proposition such as Pedro se salvo ('Peter saved himself), we obtain (45b) as the result: (45)
a. Tr (Pedro se salvo ) = X.s' [save-himself' (Pedro')(s')J b. Tr ( al2ser medico ( Tr (Pedro se salvo) )) = Xs[save-himself' (Pedro1) (s) & time (s) c A,tVt' [t * t'& 3s1 [Idoctor1 (Pedro1) (s')] & t c time (s')& t'c time (s1) ]]]
The denotation of (45b), then, is a proposition: the set of situations of Peter saving himself whose running time is included in the running time of the eventuality of Peter's being a doctor. Since the running time of the eventuality (state) of Peter's being a doctor is temporally stable, temporal inclusion is not informative. I will take this to mean that such an interpretation is pragmatically excluded and that temporal inclusion will have to be reinterpreted as causal dependence. Stump (1985) notes that i-level predicates have a strong tendency to license causal meanings, since they apply to an individual throughout a single, continuous interval whose boundaries are not precise: "The essential properties or dispositions of an individual are, of course, naturally viewed as being among the reasons or causes for that individual behavior; that is, the cause of or reason for some event or state of affairs is routinely sought among the dispositions of its participants. For this reason, individual-level predications are good for specifying reasons and
Aspect and Situations
133
causes; it is, therefore, not surprising that language users tend to infer an explanatory or causative role for strong adjuncts and strong absolutes whose predicates are individual level."
6. SUMMARY Summarizing, the existence of two kinds of infinitives has been argued for: VP-infmitives and AspP-infinitives. Al can take either VP or AspP-infinitives as arguments. When it applies to a VP-infinitive, al has been claimed to be semantically vacuous. Consequently, a VP-level AC will denote a set of minimal situations (eventualities) and can restrict the domain of an aquantifier. However, AspP-infinitives cannot be restrictors of a-quantifiers because their situation variables end up being bound by an aspectual operator that closes them existentially. Rather than denoting properties of eventualities, AspPs denote properties of times. A12 takes AspPs as its arguments andis thought to denote a relation between the set of times the AspPs denote and a proposition p. The semantics of al2 requires that the set of times denoted by the AspP be included in the running time of the situations included in p. In the case of non perfect AspPs, this directly yields temporal overlapping as an accessible reading, without excluding causal dependencies. However, with perfect AspPs temporal overlapping is not possible, and, as a consequence, a causal link is inferred whenever available. I-level AspPs vacuously include the running time of the situations in the denotation of the main proposition and a causal link is inferred. This proposal sketches an explanation for the generalizations in table 1 and opens a set of questions for further research. A natural one to ask ourselves is whether both denotations for al can be reduced to a single one. If one takes temporal overlapping between the restrictor and the nuclear scope in a quantificational structure to be a condition on adverbial quantification, then perfect AspPs cannot satisfy this condition and i-level predicates will vacuously satisfy it (Stump, 1985: 310). As a result, only VP-infmitives will be suitable restrictors.
REFERENCES Bach, E. 1986. "The Algebra of Events." Linguistics and Philosophy 9. 5-16. Barwise, J. and Perry, J. 1983. Situations and Attitudes. Cambridge, MA: The MIT Press. Berman, S. R. 1991. On the Semantics and Logical Forms of WH-Clauses. Ph.D. diss. UMass., Amherst. Chierchia, G. 1995. "Individual-Level Predicates as Inherent Generics." The Generic Book ed. by Gregory Carlson and Francis Jeffry Pelletier. Chicago & London: The University of Chicago Press, 176-224. Davidson, D. 1967. "The Logical Form of Action Sentences." The Logic of Decision and Action, ed. by N. Rescher. Pittsburgh: University of Pittsburgh Press, von Fintel, K. 1994. Restrictions on Quantifier Domains. Ph.D. diss. Umass., Amherst.
134
From Words to Discourse
von Fintel, K. 1995. "A Minimal Theory of Adverbial Quantification." Ms. MIT (draft March 1995). Frege, G. 1892. "Uber Sinn und Bedeutung." Zeitschrift fur Philosophic un Philosophische Kritik 100, 25-50; trans. by Max Black as 'On Sense and Reference.' In P.Geach and M. Black (eds.), Translations from the Philosophical Writings ofGottlob Frege, Oxford: Blackwell, 56-78. Gamut, L.T.F. 1991. Logic, Language and Meaning. Volume 2: Intensional Logic and Logical Grammar. Chicago and London: The University of Chicago Press. Heirn, I. 1982. The Semantics of Definite and Indefinite Noun Phrases. Ph.D. diss. UMass., Amherst. Heim, I. 1990. "E-type Pronouns and Donkey Anaphora." Linguistics and Philosophy 13. 137-177. Hinttika, J. 1980. "Theories of Truth and Learnable Language." In S. Kanger and S. Ohman (eds.) Philosophy and Grammar. Dordrecht: Reidel, 37-57. Johnston, M. 1995. "The Role of Aspect in the Composition of Temporal Adverbial Clauses with Adverbs of Quantification." Proceedings of the North East Linguistic Society 25. 319-333. Kamp, H. 1981. "A Theory of Truth and Semantic Interpretation." In J.Groenendijk, T. Janssen, and M. Stokhof (eds.) Truth, Interpretation, and Information. Dordrecht: Foris. Klein, W. 1994. Time in Language. London: Routdlege. Kratzer, A. 1989. "An Investigation of the Lumps of Thought." Linguistics and Philosophy 12. 607-653. Kratzer, A. 1990. "How Specific is a Fact?" Proceedings of the 1990 Conference on Theories of Partial Information. Center for Cognitive Science and College of Liberal Arts at the University of Texas at Austin. Kratzer, A. 199la. "Modality." Semantik: Bin Internationales Handbuch der Seitgenossischen Forschung, 825-834. Kratzer, A. 1991b. "Conditionals." Ibidem, 651-656. Kratzer, A. 1998. "More Structural Analogies Between Pronouns and Tenses." SALT VIII. MIT. Kratzer, A. 1999. "The Expression of Causality in Natural Languages 2: Because." UMass ms. Krifka, M. 1989. "Nominal Reference and Quantification in Event Semantics." In R. Bartsch, J. van Benthem and P. van Emde Boas (eds.) Semantics and Contextual Expression. Dordrecht: Foris, 75-115. Lasersohn, P. 1990. A Semantics for Groups and Events. Garland Publishers: New York. Lewis, D. 1968. "Counterpart Theory and Quantified Modal Logic." Journal of Philosophy 65. 113-126. Lewis, D. 1975. "Adverbs of Quantification." Formal Semantics of Natural Language, ed. by Edward Keenan. Cambridge: Cambridge University Press. Lewis, D. 1986. On the Plurality of Worlds. Oxford: Blackwell. Ojea Lopez, A. I. 1994. "Adverbios y Categorias Funcionales en Espanol." Revista Espanola de Linguistica 24. 393-416. Parsons, T. 1990. Events in the Semantics of English. A Study in Subatomic Semantics. Cambridge, MA.: The MIT Press. Partee, B. H. 1984 "Compositionality." Varieties of Formal Semantics: Proceedings of the 4th Amsterdam Colloquium, Sept. 1982. Dordrecht: Foris, 281-311. Partee, B. H. 1995. "Quantificational Structures and Compositionality." In E. Bach et al. (eds.) Quantification in Natural Languages. Dordrecht: Kluwer, 541-603. Portner, P. 1992. Situation Theory and the Semantics of Propositional Expressions. Ph. D. diss. Umass., Amherst. Rigau, G. 1995. "The Properties of the Temporal Infinitive Constructions in Catalan and Spanish." Probus. 279301. Stump, G. 1985. The Semantic Variability of Absolute Constructions. Dordrecht: Reidel. Vendler, Z. 1967. Linguistics in Philosophy. Ithaca, NY: Cornell University Press.
135
7
MENTAL SPACES AND EPISTEMIC ATTITUDES: ON THE SPANISH SUBJUNCTIVE/INDICATIVE ALTERNATION Francisco Aliaga, Universidad Complutense de Madrid Eduardo de Bustos, UNED
1. INTRODUCTION In Mejias-Bikandi (1996), an analysis is proposed to explain the alternation between the indicative and the subjunctive mood in Spanish within the framework of the theory of mental spaces of Fauconnier (1985). Mejias-Bikandi claims that such an alternation has as a consequence the construction of different mental spaces that explain or allow a generalization over some linguistic phenomena, such as presupposition inheritance, the introduction of referents through indefinite nominal expressions, and control over infinitival clauses. On the other hand, in Bustos (1986) and Bustos and Aliaga (1996), the alternation is rather related to the epistemic attitude that speakers intend to transmit in communicative interaction and, in those terms, some different phenomena are explained, such as the scope of negation in these contexts, the cancellation of presuppositions, or mandatory mood choice. While there is a basic equivalence between the theories of mental spaces and epistemic attitudes, the latter being a particular case of the former, we will compare both approaches in this paper. We will claim that the relational nature of epistemic theory makes it preferable to explain discourse or pragmatic phenomena.
136
From Words to Discourse
In previous work on the indicative/subjunctive alternation in Spanish (Bustos, 1986; Bustos and Aliaga, 1996), we intended to give an explanation of the following facts: (a) the mandatory nature of indicative or subjunctive mood in certain syntactic constructions; and (b) the discourse or communicative role of mood choice, whenever such election is possible. As for the first issue, we remarked that the functional cause of the mandatory nature of modal choice was the need to preserve contextual coherence, formally defined as the absence of contradictory elements in the contextual frame of the communicative exchange. As for the second, we explained the modal alternation through context modification, intended or not, by the choosing speaker. In other words, the function of choosing the indicative mood was grounded by the speaker's intention to introduce new contextual elements, hence increasing the consistency of discourse. The election of the subjunctive mood, on the other hand, has the effect of allowing the context to remain unaltered, thereby preserving the epistemic neutrality of the speaker. The general pragmatic idea behind this explanation (and similar explanations of other familiar phenomena) was that a given direction of communicative behavior is trying to increase contextual coherence, removing the inconsistencies and introducing new contextual elements. This idea had its most direct precedents in the analysis proposed by G. Gazdar (1979) of the intentional theory of meaning (Grice, 1975) and in the notions developed by the relevance theory of communication (Sperber and Wilson, 1986), particularly the cognitive notion of context.1 In the cognitive dimension, the context would be constituted by a set of beliefs or, in general, by objects of epistemic attitudes. The desires and intentions of expressing such epistemic attitudes are considered the essential fuel of communicative activity, in general, and the grounds for the explanation of mood use.2 On the other hand, the use of the cognitive notion of context was constrained by an elementary principle of theoretical parsimony: do not attribute to the context a more complex structure than is needed for the explanation of general communicative intention and, in particular, for actual linguistic phenomena. The absence/presence of mandatory mood and its alternation in some structures was related to a notion of context essentially construed as an epistemic space. An epistemic space is a set of beliefs dynamically constituted in the interaction between speaker and listeners). The set of beliefs of the speakers, the beliefs ascribed to the listener by the speaker and, among these, the set conforming the shared or common ground beliefs (Smith, 1982) are distinguished as relevant sets. This epistemic view allows us to account for the scalar nature of complementation, oscillating between the low and high certainty of the ' According to Sperber and Wilson, "a context is a psychological construct, a subset of the hearer's assumptions about the world. It is these assumptions, of course, rather than the actual state of the world, that affect the interpretation of an utterance", (1986: 14). 2 As stated by T. Givon, "the pragmatic-communicative interpretation of the four modalities, on the other hand, recasts them in terms of the epistemic states and communicative goals of the two participants in the communicative transaction - speaker and hearer" (1995: 114).
Mental Spaces and Epistemic Attitudes
137
speaker (Givon, 1995), the vortex between indicative and subjunctive complementation, and strong and weak assertion from a semantic point of view (i.e. T. Givon, 1995: 114). However, the mentioned principle of theoretical parsimony clashes sometimes with the generalization principle. The lack of theoretical or conceptual structure might be an obstacle to setting interesting generalizations about linguistic data. Therefore, it is important to analyze whether a more comprehensive hypothesis over the nature of mood alternation, or mood in general, could be stated with a richer representation of meaning and taking discourse progress into account. The theory of mental spaces (Fauconnier, 1985, 1994) or blended spaces (Fauconnier, 1997; Fauconnier and Sweetser, eds., 1996; Fauconnier and Turner 1996) seems to provide such a tool of representation: an enriched frame of an adequate level of generality to establish hypotheses over heterogeneous linguistic data. At the outset, it brings a representation of meaning that is far beyond what is covered by standard theories of logical form (enriched by lexical insertion, determining truth conditions...). Hence, such a theory of grammatical representation can deservedly claim to be cognitive because it accounts for the actual mental processes of the subjects of communication. Finally, the theory of mental spaces attempts to be dynamic and constructive, that is, to account for the contextual elements that shape the flow of communicative interaction. The resulting image of communication, when considered from this theoretical viewpoint, is more complex and finer than what is allowed by any other "orthodox" view. This is not the right moment to present with great detail the representational system of the theory of mental spaces. It is perhaps enough to remember that meaning representations are built through cognitive models that fit in a sequence of cognitive configurations as discourse displays. Such configurations can be thought of as mental spaces, in metaphorical terms, and their relations as relations of accessibility among spaces. Communication is thus conceived as the building and rebuilding of those mental spaces through grammatical and pragmatic means. Grammatical mood is conceived as a grammatical vector indicating the nature of the construed mental space or the connection among mental spaces through accessibility relations (Fauconnier, 1997: 95). The analysis of mood alternation in Spanish by Mejias-Bikandi (1996) analyzes the problems or questions that such an alternation poses for the building or rebuilding of mental spaces. Concretely, it analyzes three kinds of phenomena related to mood alternation: presupposition inheritance, referential determination and control over infinitival or gerundive adverbial clauses, and the effects of such an alternation over the series of cognitive configurations that constitute communication. The general idea grounding this analysis is that the use of indicative mood allows accessibility relations between consecutive mental spaces (what is generally known by the parent and descending mental spaces), while the subjunctive mood blocks those accessibility relations. In other words, the indicative mood allows a good deal
138
From Words to Discourse
of information in the mental space to pass through the successive states of this mental space, while the subjunctive mood questions, at least, this possibility. The general idea and the departing hypothesis of our analysis is not radically different from Mejias-Bijandi's. The parent mental space (or common ground) can be thought of as the counterpart of the initial space of context viewed in epistemic terms. The accessibility relations determining the inherited information among mental spaces have their counterpart in the inference relations (semantic and pragmatic) that specify the element to be introduced or eliminated from the epistemic context. The two approaches or alternative explanations should be analyzed in order to effectively compare their efficiency.
2. PRESUPPOSITION INHERITANCE Mejias-Bikandi (1996) analyzed first the case of mood alternation and the way in which such an alternation affects presupposition inheritance (Fauconnier, 1985). According to this analysis, the difference between (la) and (Ib) is that in the indicative sentence the referential presupposition of the expression his son is inherited in the mental space prompted by the expression maybe in the parent, ground, or actual space of the speaker. However, this presupposition is not inherited in (Ib). The listener cannot conjecture that the speaker believes that the son exists. (1)
a. Talvez "Maybe b. Tal vez "Maybe
su hijo his son su hijo his son
estd is (IND) este is (SUB)
enlacdrcel. in jail." en la cdrcel. in jail."
According to Mejias-Bikandi, the same happens in the cases exemplified by the sentences in (2). (2)
a. No es cierto que los sindicatos han dejado de apoyar al Gobierno. "It is not true that the unions no longer support (IND) the Government." b. No es cierto que los sindicatos hayan dejado de apoyar al Gobierno. "It is not true that the unions no longer support (SUB) the Government."
In these sentences, the presupposition or conventional implicature (3) (Bustos, 1986) is inherited in the mental space corresponding to (2a), which is in the indicative mood. It is not inherited in the subjunctive variant (2b). (3)
Los sindicatos han apoyado al gobierno. "The unions have supported the Government."
Mental Spaces and Epistemic Attitudes
139
However, this analysis of mood alternation might be challenged on pragmatic and discursive grounds. Consider (Ib) in the following discursive context: (4)
S. Hace mucho tiempo que no veo al hijo de C. "It has been a long time since I have seen C's son." L. Tal vez su hijo este en la cdrcel. "Maybe his son is (SUB) in jail."
The referential presupposition 'C has a son' is not affected, in spite of the fact that the sentence is in the subjunctive mood. The listener sums this presupposition up. If this were not the case, he would say something like 'this is not surprising, because C does not have a son'. He restrains himself from challenging or staying epistemically neutral about what is said about the assumed referent. In terms of the theory of mental spaces, the existence of the son of C would still be a part of the cognitive representation built in the communicative exchange. In our opinion, the remarks just stated find additional support in the following fact. For some informants, sentences like (5a) are not acceptable, or quasi-unacceptable, but (5b) is perfect. (5)
a. ISu hijo - si es que tiene un hijo camello - tal vez ha huido. "?His son - if he has a son who is a drug dealer - maybe ran (IND) away." b. Su hijo - si es que tiene un hijo camello - tal vez haya huido. "His son - if he has a son who is a drug dealer - maybe ran (SUB) away."
The reason claimed for this difference in acceptability is not that only the sentence tal vez su hijo estd en la cdrcel ('maybe his son is (IND) in jail') presupposes C tiene un hijo ('C has a son'). The cause(s) of the unacceptability of (5a) would be its incompatibility with the assumed narrow link (Leonetti, 1990) between what is expressed by the predicate in sentence (la) and the assumptions linked to that assertion - for instance, the one that would justify the use of un hijo camello ('a drug dealer son') in the parenthetical construction. Such a link is weaker with the use of the subjunctive mood, which is a fact that may correspond to the weak(er) conjectural nature of those sentences when they are in the subjunctive mood. Similar doubts can be cast on sentence (3a). The common analysis of this sort of structure, from a logical as well as a linguistic point of view, has stressed the fact that external negation, in other words the negation with a scope over the whole sentence, can be fulfilled if the existential presupposition is cancelled. Thus, (2a) would be true if at least one of the following propositions were true. (6)
Los sindicatos siguen apoyando al gobierno. "The unions still support the government."
140
(7)
From Words to Discourse
Los sindicatos nunca apoyaron al gobierno. "The unions never supported the government."
However, the same does not happen if the internal negation is considered: (8)
Es cierto que los sindicatos no han dejado de apoyar al gobierno. "It is true that the unions have not stopped supporting the government."
Sentence (8) would not be true, evidently, if (5) were not. In semantic terms, the difference between both types of negation is scopal. If negation is external, the referential implications can be cancelled; if negation is internal, they cannot be cancelled (in what is out of the scope, namely the subject; cf. Bustos, 1986). In our opinion, what reveals the possibility of challenging Mejias-Bikandi's analysis is the impossibility of stating a general thesis about the function of subjunctive mood in these patterns, without taking into account pragmatic and discursive considerations. In other words, if the sentences are not adequately contextualized, any assertion can be categorically stated over the mental spaces that are their meaning representation. This is due to a property of meanings that mental spaces illuminate. Meanings are ascribed during a communicative exchange, and they are not inherent properties of the lexical, grammatical or logical structure of sentences. For the case at hand, what is at stake is the scope of mood, and that scope can be modified by contextual or discursive considerations, as in (5). It cannot be claimed that, in general, this scope is the whole sentence (or, to put it alternatively, the set of its presuppositions).
3. REFERENTIAL DETERMINATION AND MOOD One of the discursive functions of indefinite NPs is to introduce new referents in communicative exchanges. In the terms of mental space theory, the NP incorporates an element (a point) to the meaning representation. In the frame of the epistemic context, it adds a belief (on the existence of the entity referred by the NP) to the mutual knowledge or pragmatic ground shared by speaker and listener. The referents can be specific or non-specific. When they are specific, they can be mentioned through pronominalization or anaphora in the following discourse, but not when they are non-specific. The hypothesis of Mejias-Bikandi is that the subjunctive mood prompts a nonspecific interpretation of the referent of the NP, while the indicative mood induces a specific one. This hypothesis predicts that, in the last case, sequences of felicitous discourses would be produced if the element introduced were pronominalized, while non-felicitous ones would be produced with the use of the subjunctive mood. This conjecture is illustrated in (9).
Mental Spaces and Epistemic A ttitudes
(9)
141
a. No creo que te he ensenado una foto de mis padres. "I do not think that I have shown (IND) you a picture of my parents." b. No creo que te haya ensenado una foto de mis padres. "I do not think that I have shown (SUB) you a picture of my parents."
According to his analysis, (10) is correct following discourse (9a), but not following (9b). (10)
Espera, que la tengo guardada aqui. "Wait, I have it here."
Mejias-Bikandi is right in claiming that (9b) prompts a non-specific interpretation of una foto ('a picture'), so the only acceptable discourse would be the following: (11)
Espera, que tengo una guardada aqui. "Wait, I have one here."
But his analysis of the indicative case can be challenged. The main problem resides in that, contrary to what it is claimed by Mejias-Bikandi (1996: 167), negation seems to affect the referential interpretation of the NP in (9a). In fact, the very acceptability of (9a) could be challenged. As occurs in similar Spanish examples, (9a) should be judged acceptable only as an echoic or citing construction, that is, as a quasi-literal mention of the words by a speaker or as an immediate answer to a previous assertion. That is, (9a) can only be accepted in discursive contexts of the following type: (12)
S. iQue si conozco a tus padres? ...Me has ensenado una foto de ellos. "Do I know your parents...? You have shown me a picture of them." L. No creo que te he ensenado una fotografia de mis padres. "I do not think that I have shown (IND) you a picture of my parents."
Whatever the case, there is a real correlation between the non-specific reading and subjunctive mood, not only in the case of referential introduction, but in other well-known cases such as quantified clauses. In all those cases, well examined in the existing literature about this topic (cf. Manteca, 1981; Bosque, 1990; and especially Leonetti, 1990: Ch. 6), the use of the subjunctive mood is related to referential indetermination. This being true, a double question arises: (1) the question of whether a better cognitive representation of meaning and the pragmatic and discursive mechanisms that modify it exists; and (2) the question of the introduction of a needed analytical complexity for the explanation of the function of Spanish mood.
142
From Words to Discourse
It seems necessary, then, to stress once more what was said in the previous paragraph: since meaning is a 'fluid' notion, it is not possible to set out contextual generalizations over the meaning of mood and its alternation in Spanish.
4. SUBJECT CONTROL AND MOOD The last example used by Mejias-Bikandi to illustrate the influence of mood in the setting of a mental space is the control exerted by the subject of a main clause over an adverbial embedded sentence (infinitival or gerundive), as in the following example: (13)
a. Habiendo robado el banco, probablemente el ladron se escondio. "Having robbed the bank, the thief probably hid (IND)." b. Habiendo robado el banco, probablemente el ladron se escondiera. "Having robbed the bank, the thief probably hid (SUB)."
The analytical hypothesis is the same as that in the other examples: the indicative mood allows an interpretation in which the subject of the embedded and main clauses is the same, while this does not happen in the subjunctive, as Mejias-Bikandi maintains. The problem with this analysis is that (13b) is judged as unnatural in Spanish, when it is in fact perfectly felicitous and grammatical. The example is not relevant as an illustration of the phenomenon. It seems more relevant to call for an epistemic attitude about the fact stated in the main clause. What is dubious in (13b) is not the thief having robbed the bank, but his hiding. In other words, it is the focus of the assertion, the predicate in this case, that is affected by mood use, and not the peripheral elements or circumstances, namely the referential identity of the subjects of both clauses. Consider the contrast in (14): (14)
a. Quizds el ladron se escondio. "Maybe the thief hid (IND)." b. Quizds el ladron se escondiera. "Maybe the thief hid (SUB)."
As we claimed in previous work (Bustos, 1986; Bustos and Aliaga, 1996), what this contrast reveals is the epistemic attitude of the speaker toward what it is asserted (or the epistemic attitude that the speaker wants to discursively express), and not his attitude toward the referential presupposition of the subject, or towards the co-referentiality with the subject by the embedded clauses. And we should say the same not only in sentences such as (14), but also in sentences like (15).
Mental Spaces and Epistemic A ttitudes
(15)
143
a. Probablemente, el ladron huyo despues de haber sido herido. "Probably the thief flew (IND) after having been wounded." b. Probablemente, el ladron huyera despues de haber sido herido. "Probably, the thief flew after having been wounded."
The identity of reference of the subjects has an essentially discursive ground (Aliaga and Bustos, in press: 5.4). If Mejias-Bikandi's hypothesis were adequate, in principle it would be possible to predict that the indicative mood would be discarded, in favor of the subjunctive, in sentences like the following ones: (16)
a. Probablemente el ladron huyo despues de resultar herido el guardia jurado. "Probably the thief flew (IND) after the security cop had been wounded." b. Probablemente, el ladron huyera despues de resultar herido el guardia jurado. "Probably, the thief flew (SUB) after the security cop had been wounded."
In the above examples, the subjects in the main and embedded clause are different and both sentences are correct, which in our view lends indirect support to the considerations just stated. The co-referentiality between the subject of the main and embedded clause is a default value in Spanish; when there is no such co-referentiality, it is necessary to make it explicit, as in (17). (17)
a. Habiendo sido herido, el ladron se escondio. "Having been wounded, the thief hid." b. Habiendo sido herido el vigilante, el ladron se escondio. "Having the cop been wounded, the thief hid."
We are not claiming that the cause of such a default value is the structural process of derivation, as claimed by Rivero (1971), or the pragmatic constraints on the processing of meaning. What is true is that mood does not affect this fact, contrary to the claims of Mejias-Bikandi.
5. CONCLUSIONS The main goal of our paper has been to show how it is necessary to amend, in some respects, Mejias-Bikandi's (1996) analysis of the indicative/subjunctive alternation in Spanish. In general, and more theoretically, we have tried to show that, when considering an epistemic attitudes model within the theory of mental spaces, calling for a context that is epistemically defined is sufficient to account for a series of phenomena in which the indicative/subjunctive alternation is implied.
144
From Words to Discourse
REFERENCES Aliaga, F. and E. de Bustos. In press. Pragmaticay Lengua Espanola: Fundamentos y Aplicaciones. Madrid: Visor. Bosque, I. Ed. 1990. Indicative y Subjuntivo. Madrid: Taurus. Bustos, E. de. 1986. Pragmdtica del Espanol: Negation. Cuantificacion y Modo. Madrid: UNED. Bustos, E. de and F. Aliaga. 1996. "Indicative, Subjunctive and Context." In J. Gutierrez Rexach and L. Silva-Villar (eds.) Perspectives in Spanish Linguistics, Vol. 1. Dept. of Linguistics, UCLA, 15-33. Fauconnier, G. 1985. Mental Spaces: Aspects of Meaning Construction in Natural Language. Cambridge, Mass.: MIT Press. Fauconnier, G. 1994. Mental Spaces. 2nd ed. of Fauconnier (1985). New York: Cambridge University Press. Fauconnier, G. and M. Turner. 1996. "Blending as a Central Process of Grammar." In A. Goldberg, ed. Conceptual Structure, Discourse and Language, Stanford: CSLI, 113-130. Fauconnier, G. and E. Sweetser, eds. 1996. Spaces, Worlds and Grammar. Chicago: U. of Chicago Press. Fauconnier, G. 1997. Mappings in Thought and Language. Nueva York: Cambridge U. Press. Gazdar, G. 1979. Pragmatics. New York: Academic Press. Grice, H.P. 1975. "Logic and Conversation." In P. Cole and J. Morgan (eds.), Syntax and Semantics, 3: Speech Acts. Nueva York: Academic Press. Givon, T. 1995. Functionalism and Grammar. Amsterdam: John Benjamins. Leonetti, M. 1990. ElArticuloy la Referenda. Madrid: Taurus. Manteca Alonso-Cortes, A. 1981. Gramdtica del Subjuntivo. Madrid: Catedra. Mejias-Bikandi, E. 1996. "Space Accessibility and Mood in Spanish." In G. Fauconnier and E. Sweetser (eds.), 157 178. Rivero, M.L. 1971. "Mood and Presupposition in Spanish." Foundations of Language 1. 305-326. Smith, N. ed. 1982. Mutual Knowledge. Nueva York: Academic Press. Sperber, D. and D. Wilson. 1986. Relevance. Oxford: Basil Blackwell.
145
8
SPACE ACCESSIBILITY AND THE PRAGMATIC STATUS OF PROPOSITIONS Errapel Mejias-Bikandi, University of Nebraska-Lincoln
1. INTRODUCTION In this paper I propose a revision of the analysis for mood in Spanish provided in Mejias-Bikandi (1996). I will argue that the notion of accessibility by itself, as was used in that analysis, is insufficient to give a complete account of the semantic distinction expressed by the indicative/subjunctive contrast. Rather, the observed restrictions on the sharing of information among different domains in the representation of discourse that were associated with the subjunctive mood can be explained as the consequence of the different pragmatic status of the propositions expressed by clauses in the subjunctive. In particular, I will make use of categories used by Lambrecht (1994) to describe the Information Structure component of grammar, and I will argue that the relevant notion to understand accessibility is that of activation: different activation states of the denotata of different propositions result in different accessibility effects. The paper is organized as follows. In section 2,1 will briefly summarize the account given in Mejias-Bikandi (1996). In section 3,1 will point out some aspects of the analysis that I consider unsatisfactory. In section 4, I will briefly discuss the component of Information Structure as discussed in Lambrecht (1994) and introduce some of the notions that I will consider relevant for the revision of the analysis. In section 5 I will present the revised analysis. Conclusions will be presented in section 6.
146
From Words to Discourse
2. SPACE ACCESSIBILITY Mejias-Bikandi (1996) studies, within the framework of Mental Spaces (Fauconnier, 1985, 1997; Fauconnier and Sweetser, 1996), contexts in Spanish where either the indicative or the subjunctive mood can be used. I will refer to these contexts as open contexts. They can be illustrated by sentences (1) and (2): (1)
(2)
No creo que ha llegado. not I-think that s/he-has-IND arrived "I do not think that s/he has arrived." No creo que haya llegado. not I-think that s/he-has-SUBJ arrived "I do not think that s/he has arrived."
The goal of that paper was to characterize explicitly the specific role that mood plays in the semantic distinction of sentences such as (1) and (2). In order to do that, the paper investigates the interaction of mood with three different phenomena: presupposition inheritance (Fauconnier, 1985; Kay, 1992), the establishment of discourse referents by indefinite phrases (Karttunen, 1968) and control into adverbial infinitival clauses. These are three phenomena that illustrate some transfer of information or some connection between different domains in the representation of discourse. In Mental Spaces terms, the three phenomena illustrate some cross-spatial dependency or sharing of information. What is observed is that such cross-spatial connections are possible when the complement clause appears in the indicative mood, but they are not possible when the complement appears in the subjunctive mood. In order to illustrate this analysis, in the following sections I will provide an example of each one of the phenomena discussed. The examples and graphical representations are taken from Mejias-Bikandi (1996).
2.1. Presupposition Inheritance Consider the following two sentences: (3)
(4)
Tal vez incluso Maria estd enferma. maybe even Mary is-IND sick "Maybe even Mary is sick." Tal vez incluso Maria este enferma. maybe even Mary is-SUBJ sick "Maybe even Mary is sick."
Space Accessibility and the Pragmatic Status of Propositions
147
The verb in sentence (3) appears in the indicative mood. There is a presupposition P associated with 'Even Mary is sick'. This presupposition is "Somebody else is also sick". P is a presupposition in the possibility space M, which is built by tal vez 'maybe1. This possibility space is embedded in R, the space that represents the speaker's view of reality. In hearing (3), however, the hearer assumes that there is some other person sick; the hearer assumes that the speaker presupposes that somebody else is sick in the real world. Consequently, it must be that the presupposition P in M is inherited in R. Consider, on the other hand, sentence (4). In this sentence the subjunctive mood is used. As before, we have a possibility space M built by tal vez 'maybe', which is embedded in the space of the speaker's reality R. There is also the presupposition P associated with the word incluso 'even'. In this case, the hearer does not assume that the speaker presupposes that someone else is also sick, rather, the hearer understands that the speaker is just speculating about the possibility that other people might also be sick. In this case, P is a presupposition in M and is not inherited in R. These observations are graphically illustrated in (3') and (4'):
(31)
148
From Words to Discourse
(4')
In both cases, the expression tal vez 'maybe' opens a possibility space M that is embedded within the space R that represents the speaker's perception of reality. The presupposition P is a presupposition in A/ since it is associated with a linguistic expression that introduces material in M. In (3'), which represents the sentence in the indicative mood, the presupposition is inherited in R. This transfer of information is blocked in (41), which represents the sentence in the subjunctive mood.
2.2. Discourse Referents Consider examples (5) and (6). (5)
No creo que te he ensenado unafotOj de mis padres Espera, que la; tengo guardada aqui. not I-think that you I-have-IND shown a picture/ of my parents... Wait, I-have it/ here "I don't think that I have shown you a picture/ of my parents
(6)
Wait, I have it / here.
??No creo que te haya ensenado unafotoj de mis padres Espera, que laj tengo guardada aqui. not I-think that you I-have-SUBJ shown a picture, of my parents ... Wait, I-have it/ here "I don't think that I have shown you a picture/ of my parents
Wait, I have it / here.
Space Accessibility and the Pragmatic Status of Propositions
149
The complement clause of the first sentence in (5) appears in the indicative mood, whereas the complement clause of the first sentence in (6) appears in the subjunctive mood. In both (5) and (6), the negation builds a counterfactual space M that is embedded within R, the space that represents the speaker's perception of reality. The second sentence in both examples establishes a relation between elements in R, since this sentence is outside the scope of the negation in the first sentence. In (5), the use of the indicative in the complement clause of the first sentence allows the anaphoric pronoun in the second sentence to be linked to the indefinite phrase una foto, 'a picture'. In other words, the indefinite phrase in (5) may be interpreted as specific. Since the indefinite is interpreted as specific, it must introduce a new element w in -R, linked to a counterpart w' in M. This new element may be pointed to by the anaphoric pronoun la 'it' in the second sentence. In (6), on the other hand, the anaphoric link between the pronoun in the second sentence and the indefinite in the first sentence results in an unnatural discourse. In other words, the indefinite in this case cannot naturally establish a discourse referent. In Mental Spaces terms, the indefinite phrase una foto 'a picture' in (6) cannot introduce a new element w in R. Consequently, the anaphoric pronoun la 'it' cannot point to w. The situation can be represented in (5') and (6'), where v would represent the element pointed to by the pronoun la:
(51)
150
From Words to Discourse
(6)
The examples in (5) and (6), as those in (3) and (4), illustrate contexts whose representations involve two mental spaces, a counterfactual space M that is opened by the negation in the main clause, and Afs parent space R, the space of the speaker's reality. The verb ensenar 'to show' that establishes a relation in Mean appear either in the indicative or in the subjunctive mood. In each case there is an indefinite expression that introduces a new element which is an argument of the relation expressed by ensenar. When ensenar is in the indicative mood, the indefinite phrase can introduce a new element w in /?, linked to a counterpart w' in M, but this possibility does not exist when ensenar is in the subjunctive mood. When the verb is in the subjunctive mood, the new element w must be introduced directly in M. As a result, when the verb is in the indicative mood, the indefinite phrase can establish a discourse referent. This is not the case when the verb appears in the subjunctive mood.
2.3. Control The third phenomenon discussed involves cases where a preverbal subject in Spanish can be interpreted as the subject of a sentence initial adverbial clause. This is illustrated in sentences (7) and (8): (7)
(8)
Habiendo robado el banco, probablemente el ladron se escondio en este almacen. having robbed the bank, probably the robber hid-IND in this warehouse "Having robbed the bank, probably the robber hid in this warehouse." ??Habiendo robado el banco, probablemente el ladron se escondiera en este almacen. having robbed the bank, probably the robber hid-SUBJ in this warehouse "Having robbed the bank, probably the robber hid in this warehouse,"
Space Accessibility and the Pragmatic Status of Propositions
151
In a Mental Spaces representation of sentences such as (7) and (8), the subject argument of the relation established by escondio 'hid' must be linked to the subject argument of the relation established by robar 'rob'. The hearer of the utterance represented in (7) and (8) must be able to fill the gap in the argument structure of the relation established by robar 'rob' with a counterpart of the element pointed to by el ladron 'the robber'. In (7) and (8), the expression probablemente 'probably' builds a possibility space M. The main clause establishes a relation between elements in M. The adverbial clause habiendo robado el banco 'having robbed the bank' is outside the scope of probablemente 'probably', and thus establishes a relation between elements in the space of the speaker's reality R, the parent space of M. In (7), the verb of the main clause is in the indicative mood and its subject is interpreted as the subject of the adjunct. In this case, the subject argument of the relation established by se escondio 'hid' in M can be linked to the subject argument of the relation established by habiendo robado 'having robbed' in R. This is represented in (7'): (7)
On the other hand, in (8), where the verb esconder 'to hide1 appears in the subjunctive mood, it is difficult to obtain an interpretation under which the subject of the main clause is also the subject of the adjunct clause. As a consequence, (8) is an awkward, unnatural sentence. Informally, the problem with (8) is that it is difficult to relate the content of the adjunct clause to the content of the main clause. As in the case of (7), the main clause in (8) establishes a relation between elements in M and the adverbial clause establishes a relation between elements in R. However, in (8) a link cannot be established between the subject argument of robar 'rob' in R and the subject argument of se escondio 'hid' in M. This is represented in (8'):
152
From Words to Discourse
(81)
2.4. Accessibility On the basis of these observations, an analysis was proposed in which mood is viewed as a mechanism that controls the flow of information among different domains in the representation of discourse. An abstract characterization was given in terms of accessibility. When the clause appears in the indicative mood, information from a Mental Space M is accessible to M's parent space M, and elements within M can be accessed through their counterparts in M. In technical terms, it was said that when the indicative mood is used, although M is the focus space (the space that is elaborated), M' is still the viewpoint space (the space through which M is accessed). On the other hand, when the subjunctive is used, M is both the focus and viewpoint space. The characterization was graphically represented as in (9):
(9)
Space Accessibility and the Pragmatic Status of Propositions
153
This analysis of mood was attractive because it seemed to capture the informal observation that somehow the use of the indicative mood makes the information within the complement clause more "real" than the use of the subjunctive mood. It also situated mood within a wider set of grammatical markers that were analyzed as mechanism to regulate the flow of information among several domains in the representation of discourse (Cutrer, 1994; Doiz-Bienzobas, 1995; Doiz-Bienzobas and Mejias-Bikandi, forthcoming).
3. PROBLEMS However, as an account of the semantic differences associated with mood in Spanish, there are aspects of this analysis that I consider unsatisfactory. First, by restricting the data to the so-called open contexts, the validity of the analysis is necessarily limited. That is, the analysis cannot be taken as a comprehensive analysis of mood in Spanish, since it is claimed to be valid only for cases where either mood is grammatically available. The set of data to be analyzed was arbitrarily restricted and, by doing so, a general perspective on the phenomenon of mood in Spanish was lost. Furthermore, the analysis was restricted to open contexts precisely because it couldn't be expanded to cover other uses of the subjunctive mood in Spanish. Outside open contexts, it is not the case that the use of the subjunctive mood limits accessibility to a space M, as sentence (10) shows: (10)
Quiero que conozcas a un amigOj. Se llama Pedro<,. I-want that you-know-SUBJ a friend is-called-REFL "I want you meet a friend,. His name is Peter,."
In this sentence, the verb quiero 'I want' builds up a wish space M, again embedded within R. The indefinite phrase un amigo 'a friend' in (10) may naturally establish a discourse referent, since it can be anaphorically linked to the definite expression in the second sentence. In other words, in (10) the indefinite may introduce a new element w in R linked to a counterpart w'mM. This example is then similar to (5), but in (10) the complement of the first sentence appears in the subjunctive mood. This sentence did not represent a problem for the analysis in MejiasBikandi (1996) because the sentence does not illustrate an open context and, consequently, was disregarded. Still, sentences such as (10) point to a limitation of the analysis as an account of the semantic differences associated with mood in Spanish. On the other hand, there are also cases outside open contexts where accessibility restrictions do occur with subjunctive complements. This is illustrated in (11), which would be parallel to example (6):
154
(11)
From Words to Discourse
??Lamento que conozcas a un amigOj (mio). Se llama Pedro;. I-regret that you know-SUBJ a friend (mine)is-called Pedro "I regret that you know a friend, (of mine). His name is Peter,."
We can see then that, outside open contexts, the subjunctive mood sometimes allows for crossspatial connections, in the form of the ability to establish a discourse referent, and sometimes it doesn't. So it doesn't seem to be the case that the subjunctive mood per se is a grammatical mechanism that restricts the flow of information in discourse.' The data suggests that the accessibility effects observed are particular manifestations of some other feature that is somehow (and at least partially) related the grammatical category of mood in Spanish.
4. PRAGMATIC PRESUPPOSITION AND ACTIVATION Sentences such as (11), where a discourse referent cannot be established by an indefinite phrase within the complement of verbs like lamentar 'to regret', were discussed in Mejias-Bikandi (1998). There, the inability to establish a discourse referent was explained not in terms of particular mental spaces configurations, but by using the notion of pragmatically presupposed complement (see also Guitart, 1991). It was independently argued that the complement in sentences such as (11), as opposed to complements in sentences such as (10), represents information that is pragmatically presupposed and, as such, it can be considered old information. An intuitively obvious incompatibility was then proposed between establishing a new discourse referent and representing old information. Consequently, under the analysis in Mejias-Bikandi (1998) the inability of the indefinite phrase in the complement to establish a discourse referent is motivated by the pragmatic status of the proposition expressed by the complement, and that pragmatic status is grammatically coded by the subjunctive mood. Finally, an analysis of mood was presented where the indicative mood is associated with pragmatic assertion, and the subjunctive mood with pragmatic non-assertion (a category that includes pragmatic presupposition). However, the notion of pragmatic presupposition also runs into problems to explain the use of mood in some adverbial subordinate clauses. For example, consider sentence (12): (12)
He cocinado antes porque tienes hambre. I-have cooked earlier because you-have-IND hunger "I have cooked earlier because you are hungry."
In addition, notice that in the case of (11) it is not clear, in Mental Spaces terms, what is the space M created by the main clause that is different from the parent space and that is going to be elaborated with the information expressed by the complement clause.
Space Accessibility and the Pragmatic Status of Propositions
155
Sentence (12) can be uttered under circumstances in which the information expressed by the adverbial clause is pragmatically presupposed, where it is part of the shared knowledge of speaker and hearer. Still, the adverbial clause must appear in the indicative mood. Consequently, the notion of pragmatic presupposition seems to be only partially correlated with the use of the subjunctive mood.
4.1. Information Structure Information Structure, as discussed in Lambreclit (1994:5), is "that component of sentence grammar in which propositions as conceptual representations of states of affairs are paired with lexico-grammatical structures in accordance with the mental states of interlocutors who use and interpret these structures as units of information in given discourse contexts." Lambrecht discusses different categories and their relevance for Information Structure. A category that, together with the category of pragmatic presupposition, I consider relevant is the category of activation. The denotatum of a proposition may be active, semi-active or inactive. Quoting Chafe (1987), Lambrecht defines a concept as active (or accessible) when "it is currently lit up, a concept in a persons' focus of consciousness at a particular moment." A concept is semi-active (or accessible/semi-active) when "it is in a persons' peripheral consciousness, a concept of which a person has background awareness, but one that is not being directly focused on." Finally, a concept is inactive if "it is in a person's long term memory, neither focally nor peripherally active." We can apply these notions to account for the different status of the subordinate clauses in (11) and (12). First, notice that the information expressed by a proposition that is pragmatically presupposed must be retrievable or accessible.2 Now consider sentence (11). When a speaker utters (11), the speaker's intention is to communicate his/her emotional reaction to a particular state of affairs. The focus is not on describing that state of affairs, but on describing the main subject's reaction. It is assumed that the state of affairs described by the proposition expressed by the complement clause is background information shared by both speaker and hearer. Then, we can consider that the denotatum of the proposition expressed by the complement clause in (11) is semi-active, speaker and hearer have background awareness of the information expressed by the proposition, but that information is not being directly focused on. On the other hand, consider (12). In (12), the speaker's intention is focused both on the main clause and on the subordinate clause, since the speaker is explicitly establishing a particular semantic relation (causality) between both propositions. Then, the information expressed by the adverbial clause is active,
2
The word 'accessible' is used here to refer to information that is retrievable to both speaker and hearer. This use of the word 'accessible' in Lambrecht (1994) should not be confused with the word accessible or accessibility as they are used in Mejias-Bikandi (1996). In Mejias-Bikandi (1996) the term accessibility is used to refer to the sharing of information among mental spaces.
156
From Words to Discourse
currently lit up; it is information that is being focused on. The information expressed by the subordinate clause must be active since it is relevant to understand the particular causality relation that the speaker is establishing. As correlated with this distinction, consider the following difference between both types of complement. An adverbial clause does not necessarily represent old information; that is, a sentence such as (12) can be perfectly normal even if the adverbial clause represents new information, as in (13): (13)
No he venido antes porque se me ha estropeado el cache. not I-have come because to me has-IND broke down the car "I haven't come earlier because my car broke down."
On the other hand, the complement of a verb like lamentar necessarily must represent (or be assumed by the speaker to represent) old information in order for the sentence to be felicitous.
5. REVISION Bearing the discussion of the previous section in mind, I want to suggest the following revision to the analysis in Mejias-Bikandi (1996). Mood in Spanish is not primarily a grammatical mechanism that controls the flow of information in discourse, but rather, a grammatical coding of the pragmatic status of the proposition expressed by a clause. I will assume, as in MejiasBikandi (1996), the framework of Mental Spaces to represent the structure of a discourse: expressions of natural language build up mental spaces in which information is introduced and relative to which propositions are interpreted. I will also consider that a clause introduces the denotatum of a proposition in some mental space M. If the denotatum of a proposition is semiactive, then there are restrictions on the sharing of the information expressed by that proposition among different domains in the representation of discourse. When the denotatum of a proposition is active, cross-spatial connections are possible. In open contexts, the indicative mood may be used to activate the denotatum of a particular proposition. More specifically, I propose a revision along the following lines: i) When a speaker utters a proposition with the intention of describing his/her or someone else's perception of reality, the proposition is pragmatically asserted (see Mejias-Bikandi 1998) and its denotatum is active, the focus of attention. These propositions are expressed by sentences in the indicative mood in Spanish. ii) When a proposition is pragmatically presupposed, the sentence that expresses that proposition is generally going to appear in the subjunctive mood, and the denotatum of such proposition is generally going to be semi-active. However, when the speaker wants to activate the denotatum
Space Accessibility and the Pragmatic Status of Propositions
157
of a proposition that is pragmatically presupposed, then the sentence that expresses that proposition appears in the indicative mood.3 iii) A proposition that is neither pragmatically asserted nor pragmatically presupposed appears in the subjunctive mood. iv) There are some operators in Spanish, like tal vez 'maybe', probablemente 'probably' or the negation that create open contexts. In open contexts, the use of the indicative mood makes the denotatum of a proposition active. The information expressed by that proposition is then being explicitly focused on and that information can be shared or access across domains. One question that this revision does not answer (and neither does the analysis in MejiasBikandi, 1996) is what makes open contexts possible. Why is it that the operator probablemente 'probably' creates an open context, whereas the matrix es probable que 'it is probable that' doesn't? Leaving that question aside, open contexts offer the speaker an additional grammatical opposition (the opposition between indicative and subjunctive) that may be used by the speaker to explicitly encode information in different ways. It is the view of this paper that the opposition between the indicative and the subjunctive mood in open contexts is used as a mechanism to explicitly activate, or focus on, the denotatum of a proposition. In turn, this activation results in the opening of a particular space M, making this space accessible.
6. CONCLUSIONS In this paper, I have considered that mood in Spanish is a grammatical marker of the pragmatic status of a proposition. The categories relevant to understand the use of mood and the accessibility effects that were related to mood are those of pragmatic presupposition and activation. The accessibility effects that were observed in Mejias-Bikandi (1996) can be explained, not as a direct consequence of a particular grammatical marker (mood), but as a consequence of the pragmatic status of a proposition, which is in turn grammatically coded by mood in Spanish.
REFERENCES Chafe, W. 1987. "Cognitive Constraints on Information Flow." Coherence and Grounding in Discourse, ed. by Russell Tomlin. Amsterdam: John Benjamins.
3
This would explain the differences not only between (11) and (12), but also pairs such as: (i) Me alegra que estes aqui. "I am glad that you are-SUBJ here." (ii) Me alegra que estds aqui. " I am glad that that you are-IND here."
158
From Words to Discourse
Cutrer, M. 1994. Time and Tense in Narratives and Everyday Language.Ph.D. diss., UCSD. Doiz-Bienzobas, A. 1995. The Preterite and the Imperfect in Spanish: Past Situation vs. Past Viewpoint. Ph.D. diss.,UCSD. Doiz-Bienzobas, A and E. Mejias-Bikandi. Forthcoming. "El Aspecto, la Accesibilidad y el Alcance de los Cuantificadores." To appear in Linguistica Cognoscitiva ed. by Ricardo Maldonado. Mexico: UNAM. Fauconnier, G. 1985. Mental Spaces. Cambridge: MIT Press. Fauconnier, G. 1997. Mappings in Thought and Language. Cambridge: Cambridge University Press. Fauconnier, G. and Eve Sweetser. 1996. Spaces, Worlds and Grammars. Chicago and London: University of Chicago Press. Guitart, J. M. 1991. "The Pragmatics of Spanish Mood in Complements of Knowledge and Acquisition-of Knowledge Predicates." Discourse Pragmatics and the Verb. The Evidence from Romance. Ed. by Suzanne Fleischman and Linda R. Waugh. New York: Routledge, 179-193. Karttunen, L. 1968. "What do Referential Indices Refer to." Unpublished ms. Distributed by the Indiana University Linguistics Club. Kay, P. 1992."The Inheritance of Presuppositions." Linguistics and Philosophy 15. 333-379. Lambrecht, K. 1994. Information Structure and Sentence Form. Cambridge: Cambridge University Press. Mejias-Bikandi, E. 1996. "Space Accessibility and Mood in Spanish." Spaces, Worlds and Grammars ed. by Gilles Fauconnier and Eve Sweetser. Chicago and London: University of Chicago Press. Mejias-Bikandi, E. 1998. "Pragmatic Presupposition and Old-Information in the Explanation of the Use of the Subjunctive Mood in Spanish." Hispania 81. 941-947.
159
9
COERCION AND THE STAGE/INDIVIDUAL DISTINCTION Victoria Escandell-Vidal and Manuel Leonetti, UNED, Universidad de Alcala
1. INTRODUCTION As is well known, Carlson's (1977) aspectual distinction between individual-level predicates (ILPs) and stage-level predicates (SLPs) has proven to be relevant to the analysis of a number of grammatical phenomena, such as the interpretation of bare and indefinite noun phrases, the licensing of secondary predicates and absolute clauses, and the distinction between the two copulae ser and estar in Spanish, among others.*'1 There are, however, some syntactic environments in which an ILP appears where an SLP should be expected, and yet the sentence does not result in ungrammaticality; rather, the ILP is forced to be interpreted as an SLP, thus rendering the sequence fully acceptable. The aim of this paper is to provide an account of this phenomenon, usually considered as a case of coercion. In particular, we will address the following questions: * We wish to thank the audiences of the "Semantics and Pragmatics of Spanish" Conference (Ohio State University) and the "XXIX Simposio de la Sociedad Espanola de Lingiiistica" (Universidad de Extremadura) for valuable comments. We are especially grateful to Nuria Alturo, Ignacio Bosque, Violeta Demonte, Maria Jesus Fernandez Leborans, Brenda Laca, Francisco Matte Bon and Begona Vicente for more detailed suggestions on a first draft. Needless to say, none of them is to be held responsible for the ideas finally adopted here. Thanks also to Ted Fernald for providing us with some unpublished material, and to Aoife Ahern for checking the English version. This research is supported by the "Ministerio de Educacion y Cultura" of Spain through a DGICYT project (PB98/0707 "Gramatica e interpretacion en la teoria de la relevancia"). ' It is commonly assumed that ser appears with ILPs and estar appears with SLPs.
160
-
From Words to Discourse
How can an ILP be interpreted as an SLP? Under what conditions is coercion possible? What is exactly the nature of coercion? Is it a syntactic, a semantic or a pragmatic phenomenon? What are the consequences of coercion phenomena for linguistic theory?
2. THE NATURE OF THE ILP/SLP DISTINCTION There is an extensive body of literature on the ILP/SLP distinction.2 In English the effects of the distinction have been much studied in existential //zere-constructions and in sentences with bare plural subjects. In the first case, only SLPs are admitted as codas, whereas in the second case the presence of an SLP or an ILP crucially determines the interpretation of bare plurals, as shown in the examples in (1) and (2): (1) (2)
a. b. a. b.
There are students sick. There are students intelligent. Sharks are visible. Sharks are dangerous.
[Generic or existential reading for the subject] [Only generic reading]
We assume that the ILP/SLP distinction is a matter of semantics, and follow Kratzer's (1988/1995) proposal, according to which SLPs need to be associated with a spatio-temporal variable while ILPs do not. Thus, the anchoring of the predication in space and time is a necessary requirement for SLPs only, and this is why they denote stages. ILPs, on the other hand, do not trigger the search for spatio-temporal locations, and denote properties of individuals. Given this characterisation, it would seem natural to think that ILPs represent essential, temporally stable or permanent properties, while SLPs represent transitory or temporally delimited states. Though this view is adequate for a number of cases, it does not yield the right results for properties, since they are not essentially incompatible with temporal limits. We will start by assuming that ILPs do not denote permanent properties, but classificatory properties. We use the term classificatory to refer to those properties that are used to categorise individuals as belonging to a specific class. From this ascription, a structured picture of the world is obtained, which makes it possible to draw other significant consequences. Considering the following sentence:
2
See, among others, Carlson (1977), Kratzer (1988/1995), Diesing (1992), Ladusaw (1994), Chierchia (1995), McNally (1998), Fernald (1999).
Coercion and the Stage/Individual Distinction
(3)
161
Brutus is a rottweiler.
Uttering (3) immediately gives access to the label of a classificatory concept (rottweiler) to which we associate a number of related properties — some of which can prove crucial. The nature of the concepts we conceive of as classificatory is such that the ascription of an individual to a class tends to produce a lasting result —which explains its cognitive benefits —, but it is not necessarily applied once and for all: we need to allow for some degree of variation. Thus, we probably do not want to say that man or woman are no longer classificatory properties because there are sex changing operations; we would rather say that the individual has changed category. The same holds, of course, for other more usual changes such as those resulting from dying one's hair, changing job or changing one's political ideas. Think, for example, of nouns of role or professional activity, as plumber in (4). (4)
Felipe is a plumber.
In the above sentence we classify Felipe as belonging to the class of plumbers; however, we would be prepared to change this ascription if we came to know that he is now a lawyer. Some adjectives behave in the same way. Consider, for instance, young. We conceive it as denoting a classificatory property. Unfortunately, youth is not a category to which one will belong forever. But the fact that after a certain age one will be removed from this category does not imply that young is not a predicate denoting a classificatory property. Thus, the difference between (5a) and (5b) reflects that the subject has undergone a change of category, not a change of stage. (5)
a. She is young. b. She is no longer young.
A piece of evidence supporting the idea that properties are not necessarily incompatible with category changes is the fact that natural languages often make use of different verbs to express changes of category and changes of stage. In English become requires ILPs, while get requires SLPs, as shown in (6). (6)
a. He became {intolerant / *angry /* frightened}. b. He got {*intolerant / angry / frightened}. c. He {became/ *got} a vampire.
Then, if our view is correct, the distinction ILP/SLP has to be stated in terms of classificatory properties vs episodic stages. The permanent/transitory distinction, therefore, should not be
162
From Words to Discourse
taken as the root of their basic semantic content, but rather as a pragmatic, inferential effect (an implicature), as stated in McNally (1994) and (1998).3
3. THE PROBLEM: ILPS IN SLP CONTEXTS It is commonly assumed that, in Spanish, SLPs are required in contexts such as copulative sentences with estar (Lema, 1996; Fernandez Leborans, 1999); absolute constructions (Bosque, 1990; de Miguel, 1992; Fernandez Leborans, 1995; Marin, 1996; Hernanz, 1991; Hernanz and Suner, 1999); adjunct secondary predicates (Demonte, 1991; Demonte and Masullo, 1999); small clause captions and headlines (Hernanz and Suner, 1999); and small clauses introduced by con ('with') (Fernandez Leborans, 1995; Hemanz and Suner, 1999). Therefore, one should expect that in these environments ILPs, which refer to properties, would give rise to ungrammaticality in a systematic way. However, the occurrence of an ILP in an SLP context does not necessarily result in an anomaly, as the following contrasts show: (7)
(8)
(9)
(10)
(11)
3
a. Estar borracho To-beESTAR drunk b. 'Estar inteligente (Cf. °KSer inteligente) To-beESTAR intelligent a. Seca la ropa, ... Dried the clothes, ... b. * Cara la ropa,... Expensive the clothes a. Caminar descalzo To-walk barefoot b. 'Caminar timido To-walk shy a. Maria en la playa Maria at the beach b. *Maria experta en arte renacentista Maria an expert in Renaissance art a. Con el coche estropeado,... With the car damaged,... b. *Con el coche japones,... With the car Japanese,...
For this view of the distinction, see Fernandez Leborans 1999: 2366-2367
[as a caption] [as a caption]
Coercion and the Stage/Individual Distinction
163
The (a) examples containing SLPs are all grammatical. But notice that not all the (b) sequences with ILPs have the same status: (7b) and (9b) are acceptable, if adequately contextualised, whereas (8b), (lOb) and (lib) are not. The sequence (7b) estar inteligente is interpreted as 'to behave in an intelligent way in a certain situation', and (9b) caminar timido is understood as 'to walk in a shy manner'. Intuitively, what renders these examples acceptable is a reinterpretation process that seems to turn ILPs denoting properties into SLPs denoting stages, in order to "repair" the well-formedness of the construction. This option is available for some environments, but not for others, as the ungrammaticality of (8b), (lOb), and (1 Ib) shows. Given that ILPs can also occur in typical SLP environments, one could be tempted to claim that the ILP/SLP distinction is not relevant at all, since it does not seem to give rise to solid and safe grammatical predictions. Pragmatic explanations based on extralinguistic factors, such as encyclopaedic knowledge or degree of plausibility, could do the job of explaining the remaining contrasts. But if this step is taken, a number of significant generalisations will be missed. A pragmatic approach, if not built on the ILP/SLP distinction in lexical semantics, would fail to account for the distributional data illustrated in (7)-(ll), and would not be able to explain why certain aspectual features, when not occurring overtly, have to be inferentially added in certain environments but not in others. In a few words, pragmatics could not deal with the part of the job that corresponds to syntax and semantics. If the possibility of a purely pragmatic explanation is excluded, we have to resort to an account that makes it possible both to maintain the ILP/SLP distinction and to understand the differences in acceptability shown above. Two questions come to mind: How can an ILP appear in a context that requires an SLP? And under what conditions can this happen?
4. COERCION A suitable answer to these questions can be formulated in terms of coercion4 Coercion is a reinterpretation process set up to eliminate the conflicts between the semantic content of a constituent and the requirements of other elements in the same construction. According to Pustejovsky (1995:111), coercion is one of the generative mechanisms operating at the lexical level: "Type coercion is a semantic operation that converts an argument to the type which is expected by a function where it would otherwise result in a type error." Thus, it produces some sort of conceptual adjustment in order to make sense of the utterance and restore its acceptability. The existence of coercion phenomena is not a surprising fact if coercion is seen as a consequence of the compositional nature of meaning. From a theoretical point of view, we can 1
See Kratzer (1988/1995), De Swart (1998), Fernald (1999).
164
From Words to Discourse
resort to this notion to account for some apparent anomalies in the distribution of linguistic elements without being compelled to abandon independently justified classifications. As Fernald (1999:1-2) puts it, "in general, linguists appeal to coercion when there is reason for believing in a fundamental semantic classification of some sort, but where some wiggle room is possible. Placing a constituent of one class in the syntactic or semantic environment best suited to a member of another class may result in an altered interpretation for the constituent —one that is more like the interpretation typical for a member of the other class— rather than outright ungrammatically." The notion of coercion has proven, in fact, to be a useful tool for solving certain aspectual puzzles. For example, De Swart (1998) invokes it to account for phenomena such as (12): (12)
a. Suddenly, I knew the answer. b. John played the sonata for eight hours.
In (12a) the incompatibility between the stative nature of the predicate know the answer and the adverbial suddenly is solved by coercing the state into an event, which gives the sentence an inchoative reading. Similarly, when an event predicate such as play the sonata is combined with a temporal /or-adverbial, it can only be interpreted as receiving an obligatory iterative reading, being thus coerced into a non-delimited situation. The effects of coercion are also visible in another domain clearly related to aspect as well: the reinterpretation of mass nouns as count nouns, and vice versa, such as in (13) and (14). In (13b) the presence of many imposes a count reading on the mass noun beer, while in (14b) it is the quantifier much that imposes a mass reading on the count noun apple. (13) (14)
a. b. a. b.
There There There There
was much beer on the table, were many beers on the table. are many apples in the salad, is much apple in the salad.
Coercion is obviously a very powerful mechanism, so it has to be adequately constrained to avoid unrestricted application; otherwise, it would hinder any prediction of acceptability and possible and impossible readings. An appropriate use of this notion has to specify the conditions under which coercion processes take place and the roles of semantics and pragmatics in the overall reinterpretation. Our proposal is that coercion is not merely a pragmatic reinterpretation process, but one that is crucially induced by a syntactic trigger. Now, the first question that has to be answered is what counts as an adequate trigger for coercion. Our proposal is that the trigger has to meet two conditions:
Coercion and the Stage/Individual Distinction
•
•
165
First, it has to contain a requirement for a constituent of a specific kind. Generally speaking, the requirement can be of a semantic or of a categorial nature: for instance, a head can select a phrase with the feature [+telic], or a VP (Pustejovsky,1995). Second, the trigger must have scope over the other element. This condition is met either by heads with respect to their complements or by adjuncts with respect to their adjunction domain. The progressive and its VP complement illustrate the first possibility; temporal adverbials and their corresponding VPs illustrate the second one.
This approach has several welcome results. On the one hand, it restricts the coercion mechanism to syntactic constructions with an overt trigger; on the other hand, it ensures that the search for a coerced interpretation will only take place if certain grammatical requirements are met. The hallmark of coercion is, therefore, the occurrence of an adequate inductor.
5. COERCING ILPs Coercion has been defined as a type changing operation: from a mass noun into a count noun and vice versa, from a state into an event, etc. Coercion of an ILP into an SLP is just another instance of type shifting. When the occurrence of an ILP in a context that requires SLPs does not result in ungrammaticality and the ILP receives a stage reading, the ILP has been coerced into an SLP (See Fernald, 1999 for details). And this is what happens, we claim, at least in (7b). As mentioned above, we assume that SLPs are necessarily linked to a spatio-temporal variable, but ILPs are not. The effect of coercion is that of forcing an interpretation in which the ILP is conceived of as denoting a temporally delimited stage, i.e., the predication it represents is relativised with respect to a certain spatio-temporal variable. Thus, though an ILP does not itself require the consideration of this sort of variable to receive an interpretation, the resolution of the aspectual mismatch between the ILP and the requirements of the context obligatorily "adds" such a variable. Once the ILP predicate becomes temporally anchored, the property denoted by it is no longer valid as a criterion for classification. Rather, the coerced predicate is systematically understood as conveying "that the subject is behaving, in some situation, in a manner consistent with having the property denoted by the ILP", as Fernald (1999) puts it. In his description of the readings of ILPs in sentences with adverbs of quantification such as Nancy is rarely clever or Sam is sometimes a bore, Fernald calls "Evidential Coercion" this change in interpretation because "it involves the subject giving behavioural evidence for having the property named by the ILP." When coercion is conceived this way, the existence of coerced readings is no longer a problem for the ILP/SLP distinction. On the contrary, the coercion mechanism itself relies on this distinction and therefore represents a way of maintaining its role in grammar. In other words, speaking about coercion would make no sense unless we assume that there are well-
166
From Words to Discourse
established distinctions and classifications. In what follows, we will concentrate on the possibility of using coercion as an explanatory notion for the data in (7)-(l 1).
5.1. Predications with estar Following Bosque (1990), Mejias-Bikandi (1993), Lema (1996), Fernandez Leborans (1999), among many others, we assume that the distribution of the two copular verbs ser and estar in Spanish reflects the basic ILP/SLP distinction, and that ser and estar are verbal heads that select a small clause, an IL small clause for ser, and an aspectually marked SL small clause for estar. Given this, the occurrence of IL adjectives with estar, as illustrated in (7b), should be a striking phenomenon; and yet it is neither odd nor unusual. The reason is that estar easily forces the interpretation of a property as a stage. In fact, estar meets the necessary conditions to act like a trigger for coercion: it is a head with aspectual selection properties, and it has scope over the ILP. Thus, ILPs such as rojo ('red'), simpdtico ('nice'), inteligente ('intelligent'), dificil ('difficult'), or democrdtico ('democratic') when constructed with estar, are coerced into a reading in which the property is understood as temporally limited with respect to a certain situation or as the result of someone's perception of the situation. (15)
a. estar rojo (cf. Ser rojo) to-beESTAR red "to look red or reddish, to turn red" b. estar simpdtico (cf. Ser simpdtico) to-beESTAR nice "to behave in a nice way, to be perceived as nice" c. estar inteligente (cf. Ser inteligente) to-beESTAR intelligent "to behave in an intelligent way" d. estar dificil (cf. Ser dificil) to-beESTAR difficult "to look difficult" e. estar democrdtico (cf. Ser democrdtico) to beESTAR democratic "to behave in a democratic way, to be perceived by others as a democratic person"
All these examples can mean slightly different things depending on the subject of predication and the context: el sol estd rojo can be translated as 'the sun looks reddish', Juan estd rojo is interpreted like 'John turns red (by shame or anger)' or 'John is acting like a communist'. Along
Coercion and the Stage/Individual Distinction
167
similar lines, Maria es simpdtica/inteligente/dificil characterises the subject as a member of the class of nice/intelligent/difficult persons, while Maria estd simpdtica/ inteligente/ dificil indicates a temporary ascription to that class, typically on the basis of the subject's behaviour in a given situation. The same goes for democrdtico, although it produces a stylistically marked construction, one that involves a register shift and a search for poetic, witty or funny effects. No matter what the type • of the adjectival predicate is, the coercion process applies systematically. Once the spatio-temporal variable is added by the presence of estar, the ILP has to be interpreted as an SLP. As stages are dependent on some spatio-temporal anchoring, it can be easily inferred that a state-of-affairs that is being presented as a stage is one that is being presented as depending on someone's, usually the speaker's, perception of the situation (i.e. it is relativised in that sense). This is the reason why estar is widely used to express judgements on taste and personal evaluation: Esto estd buenisimo ('This tastes really good'), Tu trabajo estd bastante flojo ('Your work is really poor'), John Goodman estd genial en esa pelicula ('John Goodman is great in that movie). An IL predication with estar is therefore situation-dependent. The interpretation of such dependency can be resolved by using two variants of the same strategy: the property is relativised as expressing either personal evaluations and perceptions, or certain behaviour, not an essential feature, of the subject —this second option being, obviously, only available for animate subjects. In any case, the basic strategy for interpreting coerced ILPs is systematically one and the same, and we believe it fits Fernald's notion of Evidential Coercion nicely. Even sequences with relational adjectives that could seem quite odd at first sight receive their interpretation along the lines that we have just established. Consider, for instance, international (international). If a host offers a dinner consisting of food from different countries, a guest can comment on this variety by saying: (16)
jEstas international hoy! You-areESTAR international today!
In (16) the need to create a situational dependency for the ILP induces an interpretation that involves the subject's behaviour and its consequences. As Ignacio Bosque has pointed out to us, some IL adjectives (necesario, 'necessary' ;falso, 'false'; evidente, 'evident') are not coercible by estar. We claim that this fact does not invalidate the systematic nature of the coercion process. In our view, the anomaly of sequences such as *estar necesario and *estar falso is due to the fact that those adjectives always take a prepositional argument as their subject, and prepositional entities do not seem to be conceivable as the object of perception — a necessary step in the resolution of coercion induced by estar. True, an individual can also appear as their subject argument; but when it does, it must be itself coerced into a prepositional reading: thus, Juan es necesario is interpreted as 'it is necessary that
168
From Words to Discourse
John is present/helps'. Intuitively, the limitations on the coercion process induced by estar are to be related to the conceptual cost of the operation involved, but not to any lexical idiosyncrasy or to any syntactic constraint. Coercion occurs only with adjectives. It is excluded for PPs, even when they are equivalent to adjectives in meaning. Thus, only (17a), but not (17b), is possible. Imagine a situation in which Juan approaches wearing a bowler hat and with a walking cane in his hand: (17)
a. jVayaljEstds muy britdnico! Wow! You-areESTAR very British! "Wow! You look/are acting British!" b. *jEstds muy de Gran Bretana! Wow! You-areESTAR very from Great Britain!
Following a suggestion by Brenda Laca, we can assume that the non-coercible nature of PPs with estar is related to the fact that their interpretation is equivalent to that of relational adjectives. When a relational IL adjective gets an SL reading it does it by losing its relational value and becoming purely qualitative — often with the help of a degree quantifier, as in (17)a —; as PPs lack the possibility of being reinterpreted in the same way, they cannot be coerced. As for NPs and DPs, they are syntactically incompatible with estar (cf. *Juan estd (un) gran musico, 'John isESTAR a great musician').5 Hence, it is not surprising that they cannot undergo coercion (but see section 6 below). £star-sentences exhibit, thus, the prototypical features of coercion. They present an explicit syntactic trigger: it is, in fact, the presence of the copular verb estar that, when followed by an IL adjective phrase, activates a reinterpretation mechanism that is systematic.
5.2. Absolute Constructions Absolute constructions with explicit subjects illustrate the opposite case.6 They reject ILPs, being unable to force a stage reading for them. However, this should not be considered as an unexpected result. Actually, the constraints imposed by the construction on the predicate are so strong that even some SLPs give rise to odd sequences. In an absolute construction the predicate has to be a particular type of SLP, one that specifies the resulting end-point of a process. This generalisation accounts for contrasts like the following, where only inmovilizado (immobilised) denotes a final point: 5
The only DPs compatible with estar are those which indicate a position in a scale, such as el primero, 'the first one'; el segundo, 'the second one'; el ultimo, 'the last one'. 6 See Hernanz (1991), de Miguel (1992), Fernandez Leborans (1995), Marin (1996), Hernanz and Sufler (1999).
Coercion and the Stage/Individual Distinction
(18)
169
a. Inmovilizado Juan, ... Immobilised Juan, b. ''InmovilJuan, ... Still Juan,
The immediate prediction is that, when inserted in an absolute construction, a predicate that is ambiguous between a simple stage reading and an end-of-process reading will only have the last one. This prediction is in fact borne out, as shown in (19): (19)
a. Abierta lapuerta,... Opened the door,... b. Sentados los participantes, Seated the participants,.
In Spanish, abierta can mean both 'open' and 'opened', but in (19a) only the second reading is available. The same goes for (19b), where sentados means 'seated', not 'sitting'. 7 All this implies that in absolute constructions the aspectual condition that the predicate has to satisfy is more restrictive than in other typical SLP environments. Therefore, one could think that this fact offers a straightforward explanation for the absence of stage-readings for ILPs. Even if not every SLP is able to fulfil the requirements imposed by absolute constructions, one should not expect that ILPs would, since to reach an end-of-process reading would be more difficult for them. However, according to our proposal, the explanation for the impossibility of coerced readings for ILPs in absolute constructions can be stated in more specific terms: they are excluded because the formal requirements of coercion simply are not met. In fact, an absolute construction is never selected, so there is no explicit trigger that could impose its reading on the predicate. If this condition is not satisfied, then the reinterpretation cannot be allowed. Thus, the impossibility of coercion in unselected environments follows from our proposal in a natural way. However, there is actually a case of coercion in absolute constructions that involves SLPs. This is the case when an SLP that does not meet the requirements of the construction is under the scope of a constituent that requires a culminative predicate. The standard case is one in which an adverbial modifies the small clause. Then, the prediction is that the mismatch between the aspectual features of the adverbial and the SLP in the small clause would be resolved with the adverbial coercing the predicate. And in fact this prediction seems to be born out. Thus, sequences like the ones in (20) easily receive a coerced interpretation.
7
Our judgements on absolute constructions are quite restrictive, and a number of native speakers may disagree with them (see Fernandez Leborans, 1995 for a different view). In our dialect, a perfective predicate is required in the construction; in other variants of Spanish, other SLPs could also do the job.
170
(20)
From Words to Discourse
a. Una vezfurioso elprofesor, Once angry the professor b. Una vez podridas las patatas, Once rotten the potatoes
The coerced interpretation indicates the resulting endpoint of a process where someone tried to exasperate the professor or to deteriorate the potatoes. So, as soon as we introduce an adverbial such as una vez ('once') in the construction, an explicit trigger appears that coerces the interpretation of the predicate — this time from state to end-of-process. Of course, the overall acceptability of such readings will be heavily dependent on pragmatic considerations. But the plausibility of the resulting reading is a matter not of semantics, but of our knowledge of the world.
5.3. Secondary Predicates It is usually assumed that predicative adjuncts are SLPs and yet, at the same time, it is quite easy to find examples of secondary predication in which an ILP is interpreted as a stage. It has to be underlined, however, that the results of introducing an ILP as a secondary predicate are perceived as belonging to a more formal register 01 even to a literary style: (21)
a. Avanzo timido. He-approached shy. b. Saludaba respetuosa. She-greeted respectful. c. Concedio generoso elpermiso. He-granted generous the permission.
The adjectives in (21) seem to receive an adverbial interpretation that can be paraphrased as Avanzo timidamente ('He approached (walking) in a shy manner'), Saludaba respetuosamente ('She greeted in a respectful manner') or Concedio generosamente el permiso ('He granted the permission in a generous manner'). These readings are all of the stage-type. However, things are not as simple as the existence of this sort of examples would suggest. The possibility of reinterpretation is far from being clear and systematic, basically because one can also find examples where ILPs appear as adjunct secondary predicates and yet they are interpreted as indicating properties, and not stages. In (22) and (23) we reproduce some of these examples in Spanish and English (the English ones are taken from McNally (1994:564-565)):
Coercion and the Stage/Individual Distinction
(22)
(23)
171
a. Volvio a su casa millonario. He-returned home (a) millionaire, b. Empezo futbolista y sejubila ministro. He-began (as a) football-player and he-retires (as a) minister. a. They left the Army fervent non-interventionists, b. Poe died a pauper.
Nominal predicates such as millonario, futbolista, ministro, fervent non-interventionists, a pauper, are always ILPs. One could think that a reinterpretation of properties as stages is taking place again, since the predicates tend to be understood as temporally limited. Notice, however, that positing such a modification of predicate type will have to face a number of objections: a) NPs or DPs cannot be converted into SLPs, as demonstrated by their incompatibility with estar in Spanish and with existential constructions in English. (24) (25)
*estar futbolista (cf. OKser futbolista) to beESTAR football-player *There were people republican.
b) The examples admit a paraphrase with ser in Spanish, which means that the predicates still behave as ILPs: (26)
a. b. c. d.
Volvio siendo millonario. Empezo siendo futbolista y sejubila siendo ministro. Dejaron el ejercito siendo antiintervencionistas. Poe murio siendo pobre.
c) As said in section 2, temporal delimitation is not necessarily incompatible with ILP status. In the above examples, all predicative NPs denote classificatory properties, independently of the fact that they can hold before a significant change or after such a change. In fact, most of the cases under discussion involve a special class of nouns, namely those denoting post, professional activity and ideological ascription. Thus, terms such as minister, football-player or noninterventionist are used to classify individuals according to their position in society, and, in this sense, they are ILPs. Nevertheless, as pointed out before, it is perfectly conceivable that an individual could change job or ideals during his/her lifetime, so post or role nouns share some properties with SLPs. In fact, they exhibit a peculiar behaviour in copulative sentences and small
172
From Words to Discourse
clauses —an issue we cannot discuss here.8 Even assuming that these nouns are still ILPs, it should not be surprising if they can be used as adjunct secondary predicates. To sum up, the examples in (22) and (23) show that ILPs can occur as predicative adjuncts without being reinterpreted as SLPs.9 This fact seems to suggest that, as McNally (1994) has argued, coercion is not involved in the interpretation of this particular class of secondary predicates. In addition, the possibility illustrated in (22) and (23) is limited to verbs that denote the beginning or the end of a situation (to go / to return, to begin / to finish, to be born / to die). We can return now to the examples in (21), where ILPs receive a stage interpretation. Our claim is that in those cases there is not a coercion process either, at least in the sense of coercion we have been using, since the basic formal conditions are not met. In fact, there is no trigger that can impose its requirements on the ILP adjunct. According to McNally (1994:570-573), the stage reading is obtained as an effect of a simultaneity condition on the state-of-affairs described by the V and the adjunct. For the condition to be non-trivially met, the adjunct has to be an SLP; otherwise the construction would be infelicitous and uninformative, as happens in McNally's example Fleisher played the piano intelligent. The impossibility of IL readings in adjunct secondary predicates has therefore a pragmatic motivation. The stage reading of the adjuncts in (21) is obtained by a process that resembles coercion in that it produces a conceptual readjustment in the semantics of the secondary predicate. But such a process is not triggered by any grammatical element, so it cannot be defined as coercion in our terms.
5.4. Small Clauses in Headlines and Captions As indicated by the ungrammaticality of (lOb), ILPs cannot be coerced in small clauses in headlines and captions. If we accept the approach to coercion proposed before, this is a totally predictable fact. Small clauses in captions are much like absolute constructions in that they are not selected, so there is no constituent that could be held responsible for triggering the coerced interpretation. Thus, only SLPs are allowed in these constructions. (27)
a. * Hemingway republicano (Cf. Hemingway era republicano.) Hemingway (a) republican (cf. Hemingway wasSER (a) republican.) b. *Hemingway inteligente (Cf. Hemingway era inteligente.) Hemingway intelligent (cf. Hemingway wasSERintelligent.)
(28)
8
a. Hemingway borracho (Cf. Hemingway estaba borracho.) Hemingway drunk (cf. Hemingway wasESTARdrunk.)
See Bosque (1996:59-62), Kupferman (1991), and Stowell (1991). Some additional evidence for this conclusion comes from the analysis of depictive PPs of the kind illustrated in El bocadillo, se lo tomo dejamon, '(As for) the sandwich, (he) ate it (made) of ham'.
9
Coercion and the Stage/Individual Distinction
173
b. Hemingway exultante (Cf. Hemingway estaba exultante.) Hemingway exultant (cf. Hemingway wasESTAR exultant.) The main argument for considering that the need for an SLP in captions is a formal requirement, not merely a matter of pragmatic plausibility, is the following. There are pairs of adjectives (or equivalent PPs) that have the same conceptual content, except for the fact that one is an ILP while the other is an SLP. As expected, only the sequence with the SLP is grammatical in a caption, as the following contrasts show: (29)
(30)
a. °KHemingway enfadado (cf. Hemingway estaba enfadado.) Hemingway angry b. *Hemingway irascible (cf. Hemingway era irascible.) Hemingway irascible a. OKHemingway con barba Hemingway with beard b. *Hemingway barbudo Hemingway bearded
5.5. Small Clauses Introduced by con Small clauses introduced by con ('with') constitute a challenging issue for our proposal. In fact, they contain a preposition that selects an SL small clause, which could be a suitable candidate for a trigger; but at the same time they do not show the typical coercion effects, as noted above (ex. (lib). ILPs give rise to ungrammatical strings when inserted in this environment, except for the usual post or role nouns (cf. Con Mariano embajador en Managua..., ('With Mariano ambassador in Managua...') though the stage version with de ('of) + noun is strongly preferred, cf. Con Mariano de embajador en Managua...). Then, the immediate question that arises is why the preposition cannot act as a coercer. After all, as already mentioned, it selects an SL small clause and has scope over the predicate. A tentative answer can be suggested along the following lines: being a [-V] head, con lacks temporal and aspectual features, and this is possibly the reason why it cannot coerce an adjectival predicate. In some sense, con (as well as sin 'without') is too weak to count as an inductor for a semantic readjustment.
174
From Words to Discourse
6. AN EXTENSION: ILP COERCION INSIDE THE NOMINAL DOMAIN If this proposal is correct, one would expect that coercion of an ILP into an SLP should be possible also in other syntactic environments. And in fact an instance of coercion can be found in the nominal domain which confirms the guidelines of our analysis. It is a coercion phenomenon induced by the combination of the indefinite article un and a proper name. Let's consider the examples in (31): (31)
a. Rivaldo recibio el pase de gol de un generoso Kluivert, "Rivaldo received the goal shot from a generous Kluivert." b. Un irascible Jack Nicholson agredio a losfotografos. "An irascible Jack Nicholson punched the paparazzi."
The phrases un generoso Kluivert and un irascible Jack Nicholson do not refer to the individuals Kluivert and Jack Nicholson but to particular stages they are going through. In (3la), we speak about the generous behaviour of Kluivert in a certain situation, and in (31b) we mention an episode in which Jack Nicholson behaves in an irritable way. The central issue is that generoso and irascible are ILPs, but in (31) they are clearly coerced into a stage reading, thus not introducing properties of individuals. We think that here we have another case of what Fernald (1999) calls Evidential Coercion. In this case, the process is activated by the need to make the proper name compatible with the indefinite article. It is the article that imposes the count reading on the name, as in other phrases like a delicious wine (where the mass noun is given a subclass interpretation), and a Kennedy (where the proper noun is treated as a common, count noun). One of the ways of obtaining a count reading out of a name is by shifting its reference from individuals to stages; once this is accomplished, the adjectival modifier of the name must obligatorily receive a stage interpretation too. What we obtain is a case of indirect coercion, triggered by the indefinite article on the adjective via the name. The final result is comparable to what we find in cases like (32a,b), where the adjectives inspiradisimo ('very inspired') and euforico ('euphoric') are by nature SLPs: (32)
a. Un Pavarotti inspiradisimo hizo las delicias delpublico. "A very inspired Pavarotti delighted the audience." b. Un Benigni euforico se convirtio en elprotagonista de la ceremonia. "An euphoric Benigni became the star of the ceremony."
A significant property of the constructions in (31) and (32) is the obligatory presence of the adjectival predicate. Without the adjectives, these DPs cannot be interpreted as referring to
Coercion and the Stage/Individual Distinction
175
stages, and are assigned alternative readings. For instance, un Pavarotti does not mean 'a stage, or a certain behaviour, of Pavarotti', but something like 'a certain Pavarotti' or 'an imitator of Pavarotti', or even 'a record by Pavarotti'. This is a puzzling fact, because on the one hand the adjective is the coerced element, but on the other hand its presence is necessary if the stage reading of the DP is to be obtained. Such an apparent paradox could be accommodated if we take the adjective both as the coerced element, under the scope of the indefinite, and as the cue that orients the whole reinterpretation process towards a stage-reading, discarding other alternative readings. In fact, it must be assumed that the trigger in this case activates coercion, but it does not specify the precise nature of the resulting interpretation, which depends on the availability of contextual information. A confirmation of this analysis comes from the contrast in (33). Only with the indefinite article is the stage reading made possible, not with a definite determiner. (33)
a. El irascible Jack Nicholson agredio a los fotografos. "The irascible Jack Nicholson punched the paparazzi." b. Un irascible Jack Nicholson agredio a los fotografos. "An irascible Jack Nicholson punched the paparazzi."
In (3 3 a) irascible maintains its basic property value. This is due to the fact that the definite article does not alter the referential properties of names, and therefore it does not induce their stage interpretation. As the name still refers to an individual, the adjectival modifier is not coerced. The indefinite article is thus shown to be responsible for the contextual readjustment of the adjective's reading. Hence, this construction exhibits all the central features of coercion.
7. CONCLUSION In the preceding sections we have provided an account of the possibility of reinterpretation of ILPs as SLPs in terms of coercion. We have considered a number of syntactic constructions in which an ILP occurs when an SLP should be expected, giving rise to different degrees of anomaly.10 Coercion is a very powerful mechanism, so it has to be defined in a way that prevents its unrestricted application. We have argued that coercion is allowed only if certain structural requirements are met. There must be an explicit trigger for the operation and it must have scope over the elements on which it imposes its semantic conditions. In this way, the distribution of coerced readings can be predicted in a motivated way. Thus, in constructions where a head 10 There are a number of constructions that we did not mention, such as perceptual reports and resultative constructions.
176
From Words to Discourse
selects a predicate coercion is systematic: it is the case ofestar and its small clause complement. In unselected environments, coercion is, in principle, not allowed, unless an adjunct is added that supplies the relevant features that trigger coercion." These are the requirements of the coercer. The coercee has to be of a certain kind as well. When the ILP/SLP distinction is involved not just any predicate can be coerced, only adjectives can. Such a restriction is a syntactic, not a pragmatic one, since quasi-synonyms from different categories are not interchangeable. Thus, nouns, DPs and PPs simply cannot be coerced, as shown in 5.1. The above approach has several consequences for semantics, pragmatics and the grammar/pragmatics interface. On the one hand, it is based on the idea that conceptual meaning is of a flexible nature, so as to allow for adjustments issuing from compositionality. The cases of coercion we have discussed arise as a result of the mismatch between the requirements of two elements, one demanding the identification of a spatio-temporal variable, and one lacking this requirement. The element that has scope over the other imposes its conditions, and the conflict is resolved by introducing the variable in the semantic representation. Up to this point, coercion is a semantic operation derived from compositionality. However, the addition of a spatio-temporal variable is not the end of the story. Obtaining the ultimate import of the utterance also involves inferential processes that further specify both the content of the predicate and the value of the added variable. Two facts must be underlined at this point. On the one hand, it is important to realise that such processes are not of a semantic, but of a pragmatic nature: they have to take into account extralinguistic information in order to deduce what is the state of affairs that is being referred to. On the other hand, it should be noticed that such inferential processes have to be completed before a full proposition is obtained. In other words, inference plays the main role in the processes leading from the logical form to the full prepositional form. The explicit content communicated by a sentence — the explicature, in relevance-theoretic terms (Sperber and Wilson 1986/1995) — is obtained as a result of an inferential (hence, pragmatic) development of an abstract, underspecified logical form. Thus, coercion is a transversal operation that affects both the logical form of a sequence (a matter of grammar) and the derivation of explicatures (a matter of pragmatics). For all these reasons coercion phenomena fit quite well with the overall picture of semantics and pragmatics developed in Relevance Theory, where enrichment and loosening processes (Carston 1998, 1999) are claimed to apply to conceptual content in deriving explicatures. Furthermore, the analysis of coercion reveals a significant difference between conceptual and procedural information.12 We claim that coercion affects conceptual information only (though, as " The role of adjuncts is clear in examples such as those considered by Fernald (1999): Nancy is rarely clever Laura is often pedantic Sam is sometimes a bore 12 See Wilson and Sperber (1993) for the distinction between conceptual and procedural content.
Coercion and the Stage/Individual Distinction
177
already noted, not just any kind of conceptual information). Only conceptual content can be enriched, elaborated on and manipulated in different ways under the influence of some linguistic element. Basic procedural content seems to be more rigid and robust, in the sense that it certainly triggers pragmatic inferences, but it cannot itself be elaborated on to satisfy external requirements. As for the elements that are responsible for the coercion effect, they can be both procedural (for instance, tense) and conceptual (for instance, adverbials or lexical verbs) A general question still remains. Why do coercion phenomena exist? After all, producing a sequence with conflicting elements does not seem a safe means of communication, as Fernald (1999) points out. A suitable answer can be formulated in relevance-theoretic terms. The Second Principle of Relevance13 states that when a speaker asks for the attention of another individual s/he is communicating that his/her message will produce a reasonable balance between effects and effort. In other words, s/he communicates that no extra processing effort will be imposed on the hearer that is not rewarded with extra effects. Thus, coercion succeeds because, as predicted by the Second Principle of Relevance, a hearer will not only try to make sense of any utterance, but also will try to figure out the reasons why the speaker chose to produce it the way s/he did. Confronted to a clash between non-matching elements, s/he will do her/his best to recover a relevant interpretation, even though s/he has to resort to conceptual readjustments. The extra effort involved in the resolution of the clash is balanced by additional effects. These are perceived as register or style changes intentionally controlled by the speaker, and appear in all the different kinds of coercion phenomena. Coercion always produces marked utterances; the acceptability degree varies from the most conventionalised cases to the most innovative combinations, obviously the ones that require the greatest processing effort. In this way, we have a pragmatic explanation of the interpretive task. However, as we have seen before, the construccion of coerced readings is not merely a matter of linguistic performance, so we should expect to have an adequate answer also from the grammatical side. And indeed we have one. Mismatches would be simply ruled out as ungrammatical if coercion were not a legitimate mechanism in grammar. In fact, it is a part of our knowledge that meaning can be coerced. We do know, as native speakers of our language, under what structural conditions coercion is possible and what kind of readjustments of meaning have to be produced. For instance, any native speaker of Spanish intuitively knows that any adjectival predicate following estar must be interpreted as an SLP, no matter what its aspectual class is. In this way, we are able not only to infer what an utterance refers to or what is its force in a given set of circumstances, but also to calculate how a string of different elements will be compositionally interpreted. Coercion is thus a mechanism that contributes to the economy of language by allowing a controlled multiplication of senses.
13 "Every act of ostensive communication communicates the presumption of its own optimal relevance." (Sperber and Wilson 1986/1995: 260)
178
From Words to Discourse
REFERENCES Bosque, I. 1990. "Sobre el Aspecto en los Adjetivos y los Participios." In El Aspecto en Espanol ed. by Ignacio Bosque, 177-214. Madrid: Catedra Bosque, I. 1996. "For que Determinados Sustantivos no Son Sustantivos Determinados: Repaso y balance." In El Sustantivo sin Determination ed. by Ignacio Bosque, 13-119. Madrid: Visor. Bosque, I. 1999. "El Nombre Comun." Gramdtica Descriptiva de la Lengua Espanola, ed. by I. Bosque and V. Demonte, vol. 1, pp. 3-75. Madrid: Espasa Carlson, G. 1977. Reference to Kinds in English. New York: Garland. Carston, R. 1998. Pragmatics and the Explicit - Implicit Distinction, PhD dissertation, UCL. Carston, R. 1999: "The Semantics/Pragmatics Distinction: a View from Relevance Theory." In The SemanticsPragmatics Interface from Different Points of View ed. by Ken Turner. Amsterdam: Elsevier. Chierchia, G. 1995."Individual-Level Predicates as Inherent Generics." In The Generic Book ed. by Greg Carlson and Francis Jeffry Pelletier, 176-223. Chicago: The University of Chicago Press. Demonte, V. 1991. "Tiempo y Aspecto en los Predicatives Adjetivos." In Detrds de la Palabra, 116-154. Madrid: Alianza. Demonte, V. and Pascual Jose Masullo. 1999. "La Predicacion: Los Complementos Predicatives." In Gramdtica Descriptiva de la Lengua Espanola ed. by Ignacio Bosque and Violeta Demonte, vol. 2, 2461-2524. Madrid: Espasa. Diesing, M. 1992. Indefinites. Cambridge, Mass.: MIT Press. Fernald, T. 1999. "Evidential Coercion: Using Individual-Level Predicates in Stage Level Environments." Studies in the Linguistic Sciences 29. 43-64. Fernandez Leborans, M. J. 1995. "Sobre Construcciones Absolutas." Revista Espanola de Linguistica 25. 365-395. Fernandez Leborans, M. J. 1999. "La Predicacion: Las Oraciones Copulativas." In Gramdtica Descriptiva de la Lengua Espanola ed. by Ignacio Bosque and Violeta Demonte, vol. 2, 2357-2460. Madrid: Espasa. Hernanz, M. L. 1991. "Spanish Absolute Constructions and Aspect." Catalan Working Papers in Linguistics \. 75128. Hernanz, M. L. and Avellina Suner Gratacos. 1999. "La Predicacion: La Predicacion no Copulativa. Las Construcciones Absolutas." In Gramdtica Descriptiva de la Lengua Espanola ed. by Ignacio Bosque and Violeta Demonte, vol. 2, 2525-2560. Madrid: Espasa. Kratzer, A. 1988/1995. "Stage-Level and Individual-Level Predicates." In The Generic Book ed. by Greg Carlson and Francis Jeffrey Pelletier, 125-175. Chicago: The University of Chicago Press. Kupferman, L. 1991. "Structure Evenementielle de 1'Alternance un/0 Devant les Noms Humains Attributs". Langages 102. 52-75. Ladusaw, W. 1994. "Thetic and Categorical, Stage and Individual, Weak and Strong." In Proceedings of SALT IV ed. by Mandy Harvey and Lynn Santelmann, 220-229. Ithaca: Cornell University Press. Lema, J. 1996. "Aspectual Parametric Effects in Spanish." In Aspects of Romance Linguistics ed. by Claudia Parodi, Carlos Quicoli, Mario Saltarelli and Maria Luisa Zubizarreta, 279-292. Washington D.C.: Georgetown University Press. Marin, R. 1996. "Aspectual Properties of Spanish Absolute Small Clauses." Catalan Working Papers in Linguistics 5/2. 183-212. McNally, L. 1994. "Adjunct Predicates and the Individual/Stage Distinction." In Proceedings of the 12* West Coast Conference on Formal Linguistics ed. by Erin Duncan, Donka Farkas and Philip Spaelti, 561-576. Stanford: CSLI.
Coercion and the Stage/Individual Distinction
179
McNally, L. 1998. "Stativity and Theticity." In Events and Grammar ed. by Susan Rothstein, 293-307. Dordrecht: Kluwer. Mejias-Bikandi, E. 1993. "The Nature of the Stage/Individual-Level Distinction and its Syntactic Reflex: Evidence from Spanish." In Proceedings of the llth West Coast Conference on Formal Linguistics ed. by Jonathan Mead, 326-340. Stanford:CSLI. Miguel Aparicio, E de. 1992. El Aspecto en la Sintaxis del Espanol: Perfectividad e Impersonalidad. Madrid: UAM. Pustejovsky, J. 1995. The Generative Lexicon. Cambridge: Cambridge University Press. Sperber, D. and Deirdre Wilson. 1986/1995. Relevance. Oxford: Blackwell. Stowell, T. 1991. "Determiners in NP and DP." In Views on Phrase Structure ed. by Katherine Leffel and Denis Bouchard, 37-56. Dordrecht: Kluwer. de Swart, H. 1998. "Aspect Shift and Coercion." Natural Language and Linguistic Theory 16. 347-385. Wilson, D. and Dan Sperber. 1993. "Linguistic Form and Relevance." Lingua 90. 1-25.
This page intentionally left blank
181
10
SOME SPANISH QUANTIFIER MODIFIERS R. Zuber, CNRS, Paris
1. INTRODUCTION From a semantic point of view, quantifiers are denotations of possibly complex noun phrases and of nominal determiners. Quantifiers denoted by noun phrases are functions from denotations of verb phrases to truth-values - denotations of sentences. These quantifiers are said to be of type <1>. A nominal (unary) determiner is an expression which when applied to one common noun gives a noun phrase. Semantically, (unary) determiners denote functions from properties, which are denotations of common nouns, to quantifiers of type <1>. Finally, modifiers are (functional) expressions which when applied to an expression of the category C give a resulting expression of the same category C. For any (major) category C we have corresponding modifiers, of category C/C. It follows from this that modifiers denote functions from DC onto DC, where DC are possible denotations of expressions of category C (for a detailed presentation of the theory of generalized quantifiers, with Spanish linguistic examples, see Gutierrez-Rexach, 1998). The modifiers that have been probably most often analyzed are adjectives and adverbs. From the formal point of view, they can be considered as lower order modifiers, since the denotations of their arguments are sets, i.e. objects which are not higher order objects in the sense of type theory or set-theory. The purpose of this paper is to look at some modifiers that modify arguments which are of higher order in the type-theoretical hierarchy. In particular, I want to look at some semantic properties of various constructions in which modification of noun phrases and of determiners is involved. As indicated above, noun phrases and determiners denote quantifiers of type <1> and type respectively, which are objects of higher type.
182
From Words to Discourse
One of the clearest cases where modification of determiners occurs is the case of exclusion and of inclusion phrases. It is convenient to distinguish positive exclusion and inclusion phrases from negative exclusion and inclusion phrases. In (1), we have an example of a positive exclusion phrase, and in (2) an example of negative exclusion phrase. Similarly, in (3) we have an example of a positive inclusion phrase and in (4) an example of a negative inclusion phrase: (1) (2) (3) (4)
Every student except Leo was sleeping. No student, except Leo, was sleeping. All/some students, including Leo, were sleeping. No student, not even Leo, was sleeping.
Complex and modified quantifiers similar to those given in the above constructions played, recently and not so recently (cf. Kretzman, 1982) an important role in the semantic description of natural language. As far as I know, however, recent, theoretically important work on these constructions concerns basically English or Germanic languages (von Fintel, 1993; Hoeksema, 1995; Keenan, 1993; Keenan, 1996; Moltmann, 1995; Moltmann, 1996; Zuber, 1997; Zuber, 1998). In addition, the work in question concerns only a very limited sub-set of constructions involving quantifier modification: in general, the above mentioned authors consider only exception phrases like Every/No... except Leo or Every/no... except five. As I will show, there are many other constructions involving the modification of quantifiers at the semantic level. Roughly speaking, one can distinguish in the literature two approaches in the analysis of exclusion phrases. Keenan (Keenan and Stavi, 1986; Keenan, 1996) considers that syntactically they result from the application of a discontinuous determiner to a common noun. Thus Every student except Leo is a result of the application of the discontinuous determiner Every... except Leo to the common noun student. Of course this approach is compatible with an approach in which the analysis of exclusion determiners is pushed further showing their syntactic and semantic composition. Under the second approach, proposed in particular by Moltmann (1995, 1996), the exclusion phrases result, syntactically, from the application of some functional expressions, in fact modifiers, to quantified NPs. Thus one gets an NP in the form of an exclusion phrase by applying the "exclusion complement" except NP, considered as a modifier, to a NP of the form Every CN or No CN. Thus, although NPs are rarely modified, in this case, according to Moltmann, we have a modification of NPs. Interestingly enough, in order to account for some semantic properties of exclusion phrases Moltmann has to take into account the internal structure of the modified NPs, and in particular the common noun which occurs in this constituent. This means, however, that it is possible to consider that, at least in the case of exclusion phrases, the exclusion complements modify not the full NPs but rather the nominal determiners Every or No. There is another possible way of analyzing exclusion (and inclusion) phrases syntactically and semantically. As we will see, there are some reasons to consider that in such constructions
Some Spanish Quantifier Modifiers
183
there are two determiners (connected by the "inclusion" or "exclusion" connector) and that it is the second determiner, a remnant of ellipsis, which is modified. Although the final semantics obtained in both cases is the same (cf. Zuber, 1998) it might be interesting for empirical and theoretical reasons to keep in mind that there are at least these two possibilities. I will discuss them in some detail below. Given these observations, it is now clearer what kind of data I will take into consideration. I will at least mention, and partially discuss, various Spanish connectors that can be used to form complex modifiers similar to those indicated above. Thus I will consider constructions in which not only a connector like salvo/excepto (except) occurs but also constructions with ademds (in addition), aparte (besides), and possibly others. In addition, I will show that some of these connectors can occur in constructions where modification of interrogative quantifiers occurs (Zuber 2000). This does not mean, however, that any attempt will be made here to present such constructions in a systematic way and even less so in an exhaustive way. I want to stress that my point is more theoretical than descriptive. I want to relate my discussion of these constructions to some general and methodological problems in the semantics of natural languages. Thus various examples of Spanish complex and modified noun phrases and determiners will be used to discuss the phenomenon of categorial polyvalency, of exhaustiveness, and of compositionality.
2. CATEGORIAL POLYVALENCE It is not difficult to find higher order modifiers, in particular modifiers of nominal determiners. For instance numerals, i.e. determiners denoting numbers, have various modifiers. The following expressions are such modified numeral determiners in Spanish: al menos n (at least n), mas de n (more than n), como maxima n (at most n) etc. I will be basically interested in categorially polyvalent (higher order) modifiers. Many functional expressions in natural languages, not only modifiers, are categorially polyvalent in the sense that they can apply to expressions of various categories, deriving resulting expressions which can also be of different categories. Many Boolean connectors are probably the best known examples of such expressions: for instance conjunctions can conjoin two noun phrases, two adjectival phrases, two verb phrases and even two determiners. The so-called focus particles offer an interesting class of categorially polyvalent modifiers. For instance the Spanish particle solo (and its English correspondent only) can modify noun phrases, verb phrases, adverbs, numeral determiners, prepositional quantifiers and possibly even prepositions. As it happens, many of the modifiers of quantifiers I want to analyze are also categorially polyvalent. The following examples illustrate the data I want to explore. Thus I will be interested not only in exclusion phrases as in (5) but also in positive inclusion phrases, as in (6), negative inclusion phrases, as in (7), and other related constructions, as in (8) to (10):
184
(5)
(6)
(7)
(8)
(9)
From Words to Discourse
a. Todos los estudiantes, salvo/excepto Leo/uno, bailan. All the students, except Leo/one, dance "Every student, except Leo/one, dances." b. Todos los estudiantes, salvo los rusos, beben. All the students, except the Russians, drink "Every student, except the Russian ones, drink/are drinking." c. Ningun americano, salvo Bill, duerme. No American, except Bill, sleeps "No American except Bill, is sleeping." Todos los mexicanos, incluso Leo, cantan. All the Mexicans, including Leo, sing "All Mexicans, including Leo, are singing." Ningun mexicano, ni siquiera Leo, canta. No Mexican not even Leo, sings "No Mexican, not even Leo, is singing." Ademds de Maria cinco rusos cantan. In addition to Maria five Russians sing "In addition to Maria five Russians are singing." a. Aparte de Juan (solo) Maria/Maria tambien canta. Besides Juan (only) Maria/Maria also sings "Besides Juan (only) Maria/Maria also sings." b. Aparte de Juan, todos (los estudiantes) dormian. Besides/except Juan all (the students) sleep "Every student, except Juan, is sleeping." Todos los estudiantes, sobre todo Juan. All the students, above all Juan "All the students, Juan in particular/above all."
In all the above examples we have a modification either of a nominal determiner, as in (5), (6), (7), (8) and (9) or of an NP, as in (10). Modification results by application of the complex modifier to its argument. The modifier has in general the form connect NP, where connect is a connector like salvo, excepto, aparte de, ademas, etc. It is interesting that in some cases such a complex modifier can also be conjoined - and of course in this case it is not an modifier anymore. Thus in addition to (10) we also have a "conjoined form", given in (11). Similarly, in addition to (12) we have a corresponding conjoined form given in (13): (11)
Todos los estudiantes, y sobre todo Juan. All the students, and in particular Juan "All the students, and in particular Juan."
Some Spanish Quantifier Modifiers
(12)
(13)
185
Todos los estudiantes, Juan tambien. All the students , Juan also "All the students, also Juan." Todos los estudiantes, y Juan tambien. "All the students, and also Juan."
There is, however, an important semantic difference between both constructions (i.e. conjoined and non-conjoined): (10) and (12), but neither (11) nor (13) presuppose that Juan is a student. In particular, notice that although (14a) is possible, (14b) is excluded for semantic reasons: (14)
a. Todos los estudiantes, y sobre todo sus amigos/profesores. "All the students and in particular their friends/professors." b. *Todos los estudiantes, sobre todo sus amigos/profesores. "*A11 the students, in particular their friends/professors."
This difference is clearly related to the general semantic difference between modification and conjunction. More interestingly, many of the connectors forming such modifiers can be used to form complex modifiers of various categories, depending on the second argument of the connector. The following examples illustrate this fact showing also that not all connectors have such categorially polyvalent properties: (15)
(16)
(17)
(18)
(19)
(20)
Aparte/ademds de pintar, escribe. Besides/in addition to paint writes "Besides/in addition to painting he/she writes." No hay ninguna novedad salvo que/ aparte que Juan ha telefoneado. No is no news except that besides that Juan has called "There are no news besides that Juan called." Vendre "'aparte que/salvo que tu no estes de acuerdo. Come-3d.ps.fut apart from/except if you not be-2ndp. agree "I will come unless you do not agree." Ademds/aparte de alii, tienen un apartamento aqui. In addition/apart from there they have an apartment here "In addition to the apartment there, they have one here." Aparte/ademds de rdpido, trabajamuy barato. Apart from/in addition to rapidly works very cheap "In addition to working rapidly he is not expensive." Aparte/ademds de cansado, llego a casa muy desanimado. Apart from/in addition to tired arrive home very dispirited "In addition to being tired I arrive home very dispirited,"
186
From Words to Discourse
Very often connectors giving rise to complex modifiers are considered as prepositions or adverbials. The following examples, which also show categorial polyvalency, are from Pavon Lucero(1999): (21)
(22)
(23)
(24)
(25)
(26)
Fuera de los estudios, no tiene ninguna otra preocupacion. Out of studies no has no other preoccupation "Besides his studies he/she has no other preoccupation." Ademds/aparte de inteligente, es guapo. In addition/besides (of) intelligent is handsome "In addition to being intelligent he is handsome." Ademds/aparte de con un martillo, puedes hacerlo con una piedra. In addition/besides with a hammer can-2nd do with a stone "In addition to a hammer you can do it with a stone." Tiene casas par todo el mundo, ademds/aparte de en Madrid. Has houses all over the world in addition to/besides in Madrid "He/she has houses all over the world, in addition to Madrid." Ademds/aparte de que no sabe hacer nada, es un pesado. In addition to/apart from not knows do nothing is a boring "In addition to not being able to do anything, he is boring." Ademds/aparte de no hacer nada, es un pesado. "In addition to not doing anything, he is boring."
As these examples show there is a great variety of constructions where complex modifiers formed from the same connector (from the lexical point of view) can be used. It seems that the semantic modification they induce is similar in all cases: speaking very roughly, the semantic effect of such modification is to include or to exclude some specific cases indicated by the second argument of the connector to the general number of cases indicated by the first argument of the connector. It is obvious that the complete and general study of all cases illustrated by the above examples remains to be done. In what follows, I will restrict myself to the modification of expressions denoting quantifiers of type <1> and type as found in various constructions with exclusion and, to a lesser degree, inclusion noun phrases. I will also partially discuss some related modifiers of interrogative determiners as they occur in direct and indirect inclusive questions.
Some Spanish Quantifier Modifiers
187
3. EXCLUSION AND INCLUSION PHRASES As indicated above one can distinguish two types of exclusive (exceptive) noun phrases: positive ones, i.e. those beginning with the determiner a/every and negative ones, i.e. those beginning with the determiner no. As the following examples show, Spanish has at least four connectors salvo (but/except), excepto (except), menos (less) and aparte (besides/apart from) allowing the formation of positive exclusion modifiers: (27)
(28)
Iran todos los estudiantes excepto/salvo/menos yo. Will go-3rdPpl.fut. all students except//but/less me 'All students will go, except me." a. Aparte de Juan vienen todos los estudiantes. Apart from Juan come-3p-pl. all the students "Every student, except Juan, will come." b. Salvo Juan todos los estudiantes vienen. c. Todos los estudiantes vienen, excepto Juan. d. Todos los estudiantes vienen, menos Juan.
Apparently, all the examples in (28) have the same translation in English. Concerning negative exclusion phrases, one notices that the four connectives cannot all be used to construct them. Thus although (29a), (29b) and (29c) are all possible - and have the same translation - (30) is not grammatical: (29)
(30)
a. Ningun Americano, salvo Bill, duerme. "No American, except Bill, is sleeping." b. Ningun Americano, aparte Bill, duerme. c. Ningun Americano, excepto Bill, duerme. *Ningun Americano, menos Bill, duerme.
This fact indicates that menos (less) has a different status. For this reason I will not discuss this item anymore. The first theoretical question one can ask in connection with the above examples where positive or negative exclusion complex noun phrases occur, is what is modified in them and by which modifier. Since all such noun phrases are of the form NPi connect N?2, there are at least two logical possibilities: (1) either the first two elements NPi connect form a complex modifier which applies to the third element, (2) or the first NP is an argument of the complex modifier connect N?2- Furthermore, at least the first NP can be, and in fact should be, of the form Det CN, where Det is either todos (every) or ningun (no). This means that there is a possibility that the complex modifier connect N?2 may modify not an NP but a determiner,
188
From Words to Discourse
since the same modification occurs for all values of the possible common noun CN. So it is possible to analyze exclusion phrases in many different ways. In Zuber (1998) I proposed that it is the second argument of the connector connect, considered as the remnant of an ellipsis, which is modified by the complex modifier NPi connect. This proposal is based on the observation that the sentence in (31a) is equivalent to (31b): (31)
a. Todos los estudiantes, salvo Leo, bailan. All the students except Leo dance "Every student except Leo, is dancing." b. Todos los estudiantes, salvo el estudiante que es Leo, bailan. "Every student, except the student who is Leo, is dancing"
One can consider that the non-elided part the ...who is Leo in (31b) is a determiner denoting a quantifier of type (Zuber, 1998) and that consequently this determiner is modified. As it happens, however, some additional empirical facts make it preferable to consider that it is the determiner forming the quantified NP, i.e. the first argument of connect, the one which is modified. The facts I have in mind have to do with the existence of the so-called inclusion questions (cf. Zuber, 2000b), an example of which is given in (32): (32)
iQue estudiante, ademds de/aparte de Juan, Which student besides of Juan "Which student in addition to Juan, left?"
sefue ? left
It is more justified and simpler to consider that the element that is a modifier is aparte Juan and not a complex expression Que... aparte containing an interrogative determiner Que. Roughly speaking, this means that interrogative words cannot be modifiers but can themselves be modified (by "non-interrogative expressions".) Some other connectors form modifiers which modify noun phrases. This is the case with ademds (in addition) as example (8) above shows. It is interesting, however, to see that in Spanish, as opposed apparently to English, some of the connectors that form modifiers modifying (exclusive or inclusive) determiners can also be used to construct modifiers modifying noun phrases. This is in particular the case with the connector ademds (in addition to), and aparte (apart from). An example of ademds modifying a noun phrase is given in (8) and an example of ademds modifying an (interrogative) determiner is given in (32). In (33a) and (33b) we have two examples with aparte, and (33c) shows that similar constructions are not possible with salvo: (33)
a. Aparte de Juan Maria viene tambien. Besides Juan Maria comes as well "In addition to Juan Maria comes as well."
Some Spanish Quantifier Modifiers
189
b. Aparte de Maria cinco rusos cantan. "In addition to Maria five Russians are singing." c. *Salvo Juan Maria viene (tambien.) Since in (33a) the connector aparte de has proper nouns as arguments, it is obvious that in this case the connector forms a modifier which modifies a noun phrase. A similar situation is in (8) and in (33b). Notice, in addition, that in these examples it is not presupposed that Maria is a Russian. Thus, we have at least two connectors, ademds and aparte, which can form modifiers which modify either a determiner or a noun phrase. Furthermore, depending on what is modified, the connector aparte can induce either the meaning of exclusion, as in exclusion phrases, or the meaning of inclusion, as in (33). These facts give rise to various questions of empirical and theoretical importance, such as a possible difference between these two connectors and the others discussed above, or the validity of the principle of semantic compositionality in this context. I will discuss some of these problems in the next section.
4. QUESTIONS, EXHAUSTIVITY AND COMPOSITIONALITY It is interesting to observe that some of the connectors under consideration do not behave in the same way in declarative sentences and in (inclusive) interrogative sentences. This is in particular the case with the connector ademds. As we have seen, this connector can form various modifiers except modifiers modifying a determiner, unless it is interrogative. We have already seen that (8) does not entail that Maria is a Russian. However, when ademds is used in inclusive questions it can also form a modifier modifying the interrogative determiner. For instance (34a) presupposes that Maria is a Russian: (34)
a. iQue ruso, ademds de Maria, canta? Which Russian besides Maria are singing "In addition to Maria, which (other) Russians are singing?"
Notice also that the following question does not seem to be possible: (34)
b. *lQue rusos, ademds de este mexicano, cantan? Which Russian(s) besides/in addition to this Mexican, is singing *"In addition to this Mexican which Russians are singing?"
We can also test a possible difference between various connectors which can form quantifier modifiers by looking at their behavior in inclusive questions and in related rhetorical questions. As the following examples show, there seems to be a difference in grammaticality between some
190
From Words to Discourse
types of inclusion questions with the connector aparte questions with the connector ademds (in addition to): (35)
(36)
(apart from/besides) and inclusion
iQuien viene aparte/ademds de Pedro? Who comes besides/in addition to Pedro "Who comes besides Pedro?" a. lAparte de Juan quien no ha hecho nada? Besides Juan who has done nothing "In addition to Juan, who did not do anything?" b. *£ Ademds de Juan quien no ha hecho nada?
The semantic difference between various modifier forming connectors can be detected when taking into account the possibility of rhetorical interpretation of inclusion questions on the one hand and at the problem of exhaustivity on the other hand. As is well-known, the rhetorical interpretation of questions becomes easier if the conditional mood is used and this is indeed the case in the following examples: (37)
(38)
(39)
iQuien sino Pedro podria hacer eso? Who but Pedro could do this "Who but Pedro could do this?" iQuien aparte Pedro podria hacer eso? Who apart from Pedro could do this "Who but Pedro could do this?" ? iQuien ademds de Pedro podria hacer eso?
As (39) shows, inclusion questions with ademds do not seem to have a rhetorical interpretation. It is likely that this fact is related to another property of ademds which opposes it to aparte: the property of inducing exhaustivity or non-exhaustivity in the interpretation of questions. Inclusive questions, like "ordinary" wh-questions, tend to be interpreted with an exhaustivity requirement: roughly speaking, the question they convey requires the enumeration of all individuals which have the property denoted by the verb phrase, in addition to the individual (or individuals) referred to by the complement of the inclusion clause (for details concerning exhaustivity in questions see Beck and Rullmann, 1999). Thus, in the following examples (40a) and not (40b) is considered as having an exhaustive interpretation: (40)
a. iQuien viene aparte Pedro? "Who is coming, in addition to Pedro?" b. iQuien viene ademds de Pedro?
Some Spanish Quantifier Modifiers
191
Although exhaustivity has been usually discussed in connection with questions and cleft sentences (Horn, 1981), some of the Spanish connectors can also be considered from this point of view. Thus given a specific conjunction of two proper names in the subject position in a sentence we can ask whether there are exactly two individuals denoted by the proper names which have the property indicated by the VP of the sentence. For instance such a question can be asked also in the context of the following examples: (41)
(42)
a. Aparte de Juan viene Maria. b. Aparte de Juan solo vino Maria. "Besides Juan only Maria came." a. Ademds de Juan viene Maria. b. ?? Ademds de Juan solo vino Maria.
Thus, given the truth of (41a) in a given context, is it reasonable to ask whether there are other persons besides Juan and Maria who came. In other words, does (4la) mean the same as (41b)? Many speakers give an affirmative answer to this question. On the other hand, (42a) is compatible with the situation in which other persons also came. This means that aparte induces exhaustivity and ademds does not. This difference in exhaustivity is related, it seems to me, to the differences in rhetorical readings of questions noted above. The last theoretical question I want to comment on concerns the compositionality of various constructions with modified quantifiers. There are at least two reasons which might lead us to suspect that for such constructions, or for some parts of them, the meaning cannot be compositionally determined. The first reason is that if we consider that in exclusion phrases, as in (5) above, a complex modifier connect N?2 modifies the initial determiner when the meaning of such a complex modifier depends precisely on its argument. If the argument is Every then, roughly speaking, the modifier has an exclusion meaning, as in (5a) and (5b); and if the argument is No then the modifier has an inclusion meaning, as in (5c). In other words, the noncompositionality of the modifier except A can be expressed in the following way. The connector except indicates that its second argument is exceptional, relative to a given set of objects. Now, an object can be exceptional in a given set of objects because either it has a property that other objects do not have or because it lacks the property that all other objects have. In the full exclusion phrase the first type of exception is induced by the initial determiner No and the second type by the initial determiner Every. One could claim that the meaning of the complex sub-part in which these determiners are missing cannot be compositionally determined. In fact, a compositional interpretation of such a modifier is obtainable directly from surface forms, without making use of ellipses (Zuber, 2000a). Indeed, in this case the connector except denotes the function EXCEPT which maps properties (denoted by the complement of except) to a function EXCEPT(P) (where P is the denotation of this complement) which has as its domain the set of two quantifiers, No and Every (by which exclusion phrases can begin). The values of such
192
From Words to Discourse
function are quantifiers of type corresponding to the exclusion determiners. These values are given as follows: (43)
EXCEPT(P)(No)(X)(Y)=l iff PnX=X\nY
(44)
EXCEPT(P)(Every)(X)(Y)=l iff PnX=X-Y
The second reason that may make us doubt the validity of the compositionality principle in the context of quantifier modifiers concerns the categorial polyvalency of aparte. As we have seen, it can give rise to a modifier modifying a noun phrase, as in (9a) or (41), or to a modifier modifying a determiner, as in (9b). Its semantic contribution seems to be different in each case, however. When it forms an NP modifier, it behaves like a conjunction and induces inclusion meaning, with a possibility of exhaustivity: in (4la) both Juan and Maria have the property indicated by the verb phrase. On the other hand, when aparte forms a determiner modifier it induces the meaning of exclusion. Now consider the following sentence: (45)
Aparte de Juan todos los estudiantes vienen. "Except for Juan all the students are coming." "Besides Juan all the students are coming."
For many speakers this sentence is ambiguous and the ambiguity is due to the categorial polyvalency of the connector aparte de: if this connector forms a modifier modifying the determiner todos (every), the phrase has an exclusion meaning according to which Juan did not come and if it is the whole noun phrase todos los estudiantes (every student) which is modified, then the phrase has an inclusive meaning implying that Juan did come. Furthermore, and this is important, the first interpretation entails that Juan is a student and the second interpretation does not have such an entailment but possibly the opposite one. So this ambiguity might suggest that it is not possible to provide a compositional way of computing the meaning of expressions with aparte. But the solution to this difficulty is quite similar to the one mentioned above: a function can be presented by a finite number of cases determined by properties of its argument. So depending on whether the argument of the modifier is a determiner or a noun phrase the function specifying the values of the corresponding complex expressions will be specified accordingly.
5. CONCLUSIONS Spanish has a great variety of "higher order connectors" which can be used to form various complex higher order modifiers. Such modifiers are very often categorially polyvalent. In particular they can denote functions which may take as arguments quantifiers of type <1> and
Some Spanish Quantifier Modifiers
193
quantifiers of type . This preliminary study shows that very often such modifiers can also take other arguments and that their categorial polyvalency may be, at least in some cases, language specific. Their analysis can contribute to an extension of the formal study of questions and various universal constraints which may be related to them (cf. Gutierrez-Rexach, 1997; Zuber, 2000). In addition, the study of various constructions with such modifiers leads to a better understanding of some general semantic and methodological principles such as the principle of semantic compositionality. It has been shown that they give rise to specific semantic phenomena which have been studied in other contexts, such as exhaustivity for instance. Of course, the problem of exhaustivity in questions and in general has many facets and obviously deserves fuller treatment (cf. Horn, 1981; Beck and Rullman, 1999). In particular, it is related to the problem of quantificational strength and variability in indirect "simple" and inclusive questions (Lahiri, 2000). Furthermore, since exhaustivity, exclusion and inclusion phrases are related to the atomicity of Boolean denotational algebras, in the technical sense (Keenan, 1993; Zuber, 1997; Zuber, 2001), the study of complex exclusion NPs and the determiners forming them contributes to our direct knowledge of the algebraic structure of Spanish and natural languages in general. Clearly the observations made in this paper are the beginning of an enterprise which will make this knowledge more complete. * The following persons, on various occasions and with various degrees of perspicacity and enthusiasm, helped me with the final version of this paper: Brenda Laca, Ross Charnock and Josep Quer. The research presented here is related to the programme ECOS-SUD, U97-HO2.
REFERENCES Beck, S. and Rullman, H. 1999. "A Flexible Approach to Exhaustivity in Questions." Natural Language Semantics 1. 249-298 von Fintel, K. 1993. "Exceptive Constructions." Natural Language Semantics 1. 123-148 Gutierrez-Rexach, J. 1997. "Questions and Generalized Quantifiers." In A. Szabolcsi (ed.) Ways of Scope Taking. Dordrecht: Kluwer, 409-452. Gutierrez-Rexach, J. 1998. Semdntica Logica y Cuantification Nominal: Introduccion a la Teoria de Cuantificadores Generalizados. New Orleans: University Press of the South, Inc. Hoeksema, J. 1995. "The Semantics of Exception Phrases." In J van der Does and van Eijck, J. (eds.) Quantifiers, Logic and Language. Stanford, CA: CSLI Publications, 145-178. Horn, L. 1981. "Exhaustiveness and the Semantics of Clefts." In V. Burke and J. Pustejovsky (eds.) Proceedings of the Eleventh Annual Meeting ofNELS. Amherst: U. Mass, 125-142. Keenan, E. L. 1993. "Natural Language, Sortal Reducibility and Generalized Quantifiers." Journal of Symbolic LogicS?,. 314-325. Keenan, E.L. 1996. "The Semantics of Determiners." In S. Lappin (ed.). The Handbook of Contemporary Semantic Theory. Oxford: Blackwell, 41-63. Keenan, E. L. and Stavi, J. 1986. "A Semantic Characterization of Natural Language Determiners." Linguistics and Philosophy 9. 253-326.
194
From Words to Discourse
Kretzman, N. 1982. "Syncategoremata, Exponibilia, Sophismata." In N. Kretzman et al. (eds.). The Cambridge History in Late Medieval Philosophy: From the Rediscovery of Aristotle to the Disintegration of Scholasticism 1100-1600. Cambridge. Cambridge University Press, 211-245. Lahiri, U. 2000. "Lexical Selection and Quantificational Variability in Embedded Interrogatives." Linguistics and Philosophy 23. 325-389. Moltmann, F. 1995. "Exception Sentences and Polyadic Quantification." Linguistics and Philosophy 18. 223280. Moltmann, F. 1996. "Resumptive Quantifiers in Exception Sentences." In M. Kanazawa et al. (eds.). Quantifiers, Deduction, and Context. Stanford, CA; CSLI Publications, 139-170. Pavon Lucero, M. V. 1999. "Clases de Particulas: Preposition, Conjuncion y Adverbio." In I. Bosque and V. Demonte (eds.) Gramatica Descriptiva de la Lengua Espanola. Madrid: Espasa Calpe, 565-655. Zuber, R. 1997. "Some Algebraic Properties of Higher Order Modifiers." In T. Becker and T. Kneger (eds.). Proceedings of the Fifth Meeting on Mathematics of Language. Deutsches Forschungszentrumfur Kunstlische Intelligenz, 161-168. Zuber, R. 1998. "On the Semantics of Exclusion and Inclusion Phrases." Proceedings of SALT 8. Cornell. CLC Publications, 267-283. Zuber, R. 2000a. "Exclusion Phrases and Criticisms of Semantic Compositionality." In A. Ikeya and M. Kawamon (eds.). Proceedings of 14th Pacific Asia Conference on Language, Information and Computation. LogicoLinguistic Society of Japan, 401-412. Zuber, R. 2000b. "On Inclusive Questions." In R. Billerey et al. (eds.). WCCFL 19 Proceedings. Sommerville : Cascadilla Press, 617-630. Zuber, R. 2001. "Atomicity of Some Categorially Polyvalent Modifiers." In de Groote et al. (eds.).Logical Aspects of Computational Linguistics,Volume 2099 of Lecture Notes in Artificial Intelligence. Berlin: Springer Verlag, 296-310.
195
11
DEMONSTRATIVES IN CONTEXT Javier Gutierrez-Rexach, The Ohio State University
1. INTRODUCTION Deictic elements are normally described as those "whose interpretation in simple sentences makes essential reference to properties of the extralinguistic context of the utterance in which they occur" (Anderson and Keenan, 1985).' In the philosophical literature, the term "indexicals" (Peirce, 1955) is more widespread, although both terms tend to be used with the same meaning. One of the most salient features of deictic expressions is their categorial polymorphy. Cross-linguistically, deictic expressions belong to different grammatical categories, and in most languages we can find deictic pronouns, deictic adverbs, deictic determiners, etc. Whereas the content associated to event determination and to time ascription of events is normally instantiated by verbs, auxiliaries and adverbs, we do not find the same degree of uniformity in the case of the expression of deictic content. This property can be taken as accidental or as a clue to the proper understanding of the nature of these elements. Let us assume the latter. If indexicality is a semantic property which can be associated with expressions of several categories, there must be something in that property that ' This article originates from some ideas that I first had in the spring quarter of 1997. At that time, I was completing my dissertation, which was related to context dependence, and I was also working on the syntax of demonstratives with Luis Silva. David Kaplan was a member of my dissertation committee and, in the candidacy examination, I happened to mention, rather uncontroversially it seemed to me at first, how context was incorporated in current dynamic theories of meaning and in the theory of generalized quantifiers. That sparked a very lively discussion between Kaplan and I and, at the end, I realized that the idea of context dependence was more intricate than I supposed at first. Thus, I would like to thank him for getting me to think deeper about context in general. I would also like to thank Luis Silva for all the amazing empirical data on Spanish demonstratives that he brought to our discussions.
196
From Words to Discourse
allows non-trivial categorial polymorphism. Some authors have claimed that indexicals are vehicles of direct reference and any category that can be directly associated with an entity or entities should then be an indexical category. Alternatively, one might view contextual dependence as the relevant property so, in principle, the number of categories which can be dependent on context is not limited. First and second person pronouns are prototypical indexical expressions whose interpretation depends on who the speaker or the addressee is. Consider the following sentence pairs: (1) (2)
a. b. a. b.
Cervantes wrote El Quijote. Javier is talking now. I wrote El Quijote. I am talking now.
The most salient feature of the sentences in (2) is what we can intuitively describe as their meaning "instability". Whereas the truth of sentences (la,b) is not dependent on who utters the sentence, this is not the case in the sentences in (2). (la) is true no matter whether I, Bill Clinton or Cervantes himself utters it. On the other hand, (2a) will be false if I or Bill Clinton uttered it. It is only true if Cervantes, the author of El Quijote, uttered it. With respect to (Ib), we can say that in order to determine its truth or falsity we have to check whether the individual denoted by the proper noun Javier is talking. By contrast, (2b) does not apply only to one individual. It will be true automatically when uttered, since whenever a speaker utters a sentence it must be true that he is talking. Other personal pronouns also have deictic properties. We can easily conclude that the reference of terms such as you, he, etc. is not "stable". It depends on who the addressee is, who is being referred to by the speaker, etc. Temporal and spatial deictics share some properties with pronouns with respect to meaning determination (see Fillmore, 1975, 1997). Temporal and locative adverbs such as now, then, yesterday, tomorrow, here, there, etc. are all dependent on contextual elements related to time reference. Sentence (3) is true if Peter was born on that particular year. The sentence remains true independently of the person or the point in time in which it is uttered. (3)
Peter was bom in 1967.
Consider now (4): (4)
Peter arrived late yesterday.
The above sentence is not true independently of the context of utterance. The difference between deictic adverbs and personal pronouns is that the occurrence of a deictic adverb does not make the determination of the sentence meaning dependent on who the speaker or the addressee is. Rather, it
Demonstratives in Context
197
becomes dependent on the utterance time. For (4) will be true if and only if Peter arrived late the day prior to the moment of the sentence's utterance (now). Spatial deixis is also signaled by adverbial expressions. The determination of the concrete point in space referred to in (5) will again depend on the context, to the extent that somebody who has noi been able to see the indications associated with the utterance will not be able to understand the meaning of the sentence, i.e. where he has to bring the object to and where to put it. (5)
Bring it here and put it over there.
Demonstrative determiners are also intrinsically indexical. They are different from other indexicals in generally requiring an explicit ostension or act of pointing. Thus, the determination of the reference of demonstrative pronouns such as this or that depends on which entities are being demonstrated or pointed at. Sometimes the participants in a conversation might not be able to agree on which point is exactly referred to in an exchange if they did not grasp the demonstration associated with the use of the demonstrative. In the dialogue (6), the demonstrative pronoun this does not refer to the same entity in A's as in B's sentences. (6)
A: This is for you? B: You mean this one. C: No, that one.
Imagine a situation in which A points in one direction where there is more than one object. A's intention is to point at one of those objects and he expects B to recognize this intention. Nevertheless, B's reply is precisely an attempt to clarify the content of A's demonstration, so B would point at the object that he thinks A is demonstrating in a more explicit fashion. But he is wrong, so A's reply will disambiguate the exact demonstrated object. The use of that is justified in A's reply as a way of contrasting the object demonstrated by him with the one demonstrated by B. Although personal pronouns, adverbs and demonstratives are normally the expressions that can serve as vehicles of indexicality, there are certain expressions in other categories which also serve as indexicals and are therefore context dependent. For instance, the verbs come and go (Fillmore, 1997) encode as part of their meaning movement to "here" (come) or movement away from "here" (go). The contextual dependence of these verbs survives even in idiomatic expressions such as go away, which can be only interpreted as related to the physical or mental proximity of the speaker. Adjectives such as local, actual, surrounding are also context dependent, as extensively argued for in the case of local by Mitchell (1986) and Partee (1989). Sentence (7) has a reading in which the interpretation of local is deictic: it refers to the place in which the speaker lives.2 2 The other reading arises when it is possible to establish a dependence between the adjective and a quantifier. The resulting interpretation of local is a "bound" reading, since every journalist under consideration in (i) is associated to an
198
From Words to Discourse
(7) Peter talked to a local journalist. The role of context in the determination of the meaning of expressions cannot be minimized or ignored. As a matter of fact, there are sentences which are completely context dependent such as (8). In this sentence, the utterance context has to provide the relevant individual (the speaker), as well as the location (the location where the speaker is) and time (the utterance time). Thus, the interpretation of (8) is entirely dependent on properties of the utterance situation. There is nothing in this sentence that does not come from that particular context. (8)
I am here now.
Among deictic expressions, the most complex ones and the ones that have received greater attention in the literature are demonstratives. Demonstratives are determiners, so one issue of interest is how much or how little they have in common with other determiners, especially with definites. Secondly, demonstratives not only have deictic uses but also can have anaphoric uses and bound or dependent uses. Thus, their status as fully indexical elements is controversial. Cross-linguistically speaking, demonstratives also exhibit a wide variety of expressive possibilities. There are systems of two demonstratives, like English. Others have three demonstratives, like Spanish. Other languages exhibit systems of more than three terms (see Anderson and Keenan, 1985; Diessel, 1999). The range of deictic uses can be extremely wide, from spatio-temporal deixis to discourse deixis. Other uses of demonstratives that have normally been considered secondary but emerge in a systematic fashion are emphatic uses or as devices to focus attention towards one direction or other. Depending on theoretical approaches, these uses can be taken as ancillary, derived from a main deictic core, or as an essential component of demonstrative interpretation. Spanish demonstratives have certain properties that bring to the fore some of the topics and possibilities described in the previous paragraph. In this paper, I will argue that a detailed consideration of the evidence from Spanish demonstratives shows that a refinement of the semantic and pragmatic content of the notions of indexicality and deixis is needed. This refinement goes in the direction of providing a more comprehensive theory of demonstratives that captures a wide variety of uses and that is dynamic in nature. In section two, I will review the main theories of demonstratives that have been proposed in the literature. In section three, I will address the issue of the context dependency of demonstratives: first, I will explore some ideas about what we should consider as a context and, afterwards, I will propose a theory of demonstratives in which they are athlete of his own town: (i)
Every journalist was talking to a local athlete.
Demonstratives in Context
199
treated as determiner functions inherently restricted to context sets. Finally, in section four I will analyze how data from Spanish forces us to refine and expand the theory of demonstratives proposed in the previous section.
2. THEORIES OF DEMONSTRATIVES 2.1. The Indexical Theory In the Tarskian tradition of model-theoretic semantics, the meaning of a sentence is commonly given by providing its truth conditions in a model. The Fregean distinction between denotation and sense attempted to capture the difference between the object that an expression refers to (its denotation), and the mode or way of presentation, i.e. the circumstances in which it is evaluated. Following Carnap's (1941) parallel distinction between extension and intension, it is obvious that the sentences in (9) have the same extension but not the same intension, since Marcus Tullius is Cicero. Thus, (lOa) is necessarily true but (lOb) might not be. (9) (10)
a. b. a. b.
Marcus Tullius is Marcus Tullius. Marcus Tullius is Cicero. Peter believes that Marcus Tullius is Marcus Tullius. Peter believes that Marcus Tullius is Marcus Tullius.
The extension of a sentential expression is a truth value, the extension of a unary predicate is a set, etc. The intension of a sentence is a proposition: a function from worlds to truth values (the characteristic function of a set of worlds). The intension of a unary predicate is a prepositional concept: a function from worlds to sets, etc. Thus, if the determination of the extension of an expression requires relativizing an interpretation function to models (M), and also to assignments (g) for expressions containing variables, the determination of the intension of a sentence requires adding worlds (w) as another parameter of evaluation. Thus, the interpretation function depends on three parameters of evaluation: models, assignments, and worlds. Montague (1968) called these parameters or circumstances of evaluation an index. An index can be viewed as a complex entitity, namely an n-tuple i = <w, t, p, a, o>, where w is a possible world, / a moment of time, p a place, a an agent, and o an object. Lewis (1972) and Scott (1970) proposed index-based theories of meaning, where the coordinates of an index define the circumstances against which a sentence has to be evaluated as true or false. The meaning "instability" brought about by indexicals is captured as a dependence upon one or more coordinates of the index of evaluation For instance, (1 la) would have different truth conditions depending on the different values of the place coordinate, and (lib) would be sensitive to the different values of the agent (speaker) coordinate.
200
(11)
From Words to Discourse
a. It is cold here, b. I am tired.
Indices are inherently flexible. The number of coordinates that have to be considered is not fixed. There might be more than one agent or object coordinate. Sentence (12a) requires consideration of two agents (speaker and addressee), whereas (12b) requires consideration of two objects, each one of them linked to a different occurrence of this. (12)
a. I am talking to you. b. This is faster than this.
Let us consider (12b) in more detail. This sentence would be felicitous when uttered in a situation in which I am comparing two different cars, two different designs, etc. Lewis (1972) observed that examples like (12b) show that different occurrences of the same demonstrative are associated to different objects. Thus, in the same fashion as we use subscripted variables to indicate that the logical vocabulary has an infinite number of individual variables (jr,... xn), demonstratives should also be indexed (this] ... thisn). The evaluation of sentence meaning is then dependent on sequences of objects. The object coordinate of an index is properly an n-ary sequence of objects (n HI). The representation (13) of (12b) would indicate that thisl is dependent on the first member of the sequence constituting the object coordinate, whereas this2 would be dependent on the second member of the object sequence. (13)
This, is faster than this2
Bennett (1978) proposed a theory of demonstratives which is very similar to the one proposed by Lewis. For Bennett, only places can be proper demonstrata. Regular demonstratives are spelled out as the combination of a definite expression and a locative deictic: this is 'the object here', and that is 'the object there'. There are an infinite number of deictic locative adverbs (here} ... heren, etc.), each one of them indicating a location in the place coordinate of an evaluation index. Bennett notices that this characterization of demonstratives has some inherent problems. If the definitive article picks a unique element and the deictic adverb refers uniquely to a single entity (a place), we have to conclude that demonstratives refer necessarily to the demonstrated place (here or there). Otherwise, they would be referring to two entities, an object and a place, which is incompatible with the content of the definite article. Obviously, to demonstrate a unique place is not the content of a demonstrative. Bennett's solution is to distinguish between semantic reference and speaker's reference. The hearer is able to figure out in most occasions what the reference intended by the speaker is from the content of the improper definite description.
Demonstratives in Context
201
There are other problems with Bennett's approach. In languages such as Spanish, in which the noun phrases in (14) are possible, if we follow Bennett's ideas, it is predicted that all of these expressions would be taken to have an identical meaning ('this man'): (14)
a. este hombre this man b. el hombre este the man this c. el hombre de aqui the man of here d. este hombre de aqui this man of here
This is not the case. Neither (14a) nor (14b) need to make reference to spatial location. This is true of English demonstratives too. It is also evident that anaphoric uses of demonstratives, like in (15), and the discourse deixis phenomenon illustrated in (16) cannot be reduced to spatial location (15) (16)
a. A man came in. This man was carrying a concealed gun. b. Susan died yesterday. This event will haunt us. Juan tuvo una pelea por unos terrenos con Pedro. Este no quedo nada contento y aquel un poco mas. "Juan had a fight with Peter for some land. In the end, the latter was not happy at all and the former a little bit more so."
Another problem of the Lewis/Bennett theory is that it forces us to adopt the hypothesis that demonstratives are infinitely ambiguous. Our lexicon should contain infinitely many entries this^ (1 < i < n) corresponding to the infinitely many potential demonstrata. This kind of ambiguity is different from standard lexical ambiguity. We may expect to find a few different meanings of words such as bank or chair, but not an infinite number of meanings of a functional or grammatical category like a determiner. A more general problem for an index-based theory of demonstratives is that paradoxically it does not seem to explain the very nature of indexicals that it attempts to capture. Consider again (8) repeated here as (17): (17)
I am here now.
Our intuition is that this sentence should be logically valid (in an intended logic of indexicals), in the sense that it becomes immediately true when it is uttered. Nevertheless, for the index theory
202
From Words to Discourse
sentence (17) comes out as contingent. It is true if we consider the values <Javier Gutierrez, Columbus (Ohio), May 23 of 2001> of the agent, place and time coordinates but comes out as false if we consider the values of those same coordinates. Thus, (17) comes out as contingent under the index theory. An amendment that would take care of this problem is to consider only proper indices, namely those such that in the world w, a is located in p at t. An improper index such as the one previously described would be eliminated very much in parallel to the elimination of impossible worlds. But now not only (17) comes out as valid but also (18), by the principle of modal generalization (if/? is true at every proper index then G/» is true at every proper index): (18)
Necessarily I am here now.
Sentence (18) is not logically valid at all. It is not necessarily true that I am here now. The problem arises from the identification of all the parameters in the index as belonging to the same level. As Kaplan observed, one has to distinguish the context of utterance from the circumstances of evaluation. A potential solution within the indexical theory would be to substitute the uniform approach of Lewis and Bennett by a double-index approach (Kamp, 1970; Segerberg, 1973). In this approach, it is recognized that the determination of the meaning of a sentence in a context is a two step procedure. First, we have to establish the prepositional content of a sentence from the sentence's context (the parameters related to the utterance). Second, we have to determine the truth value of the proposition evaluated under different circumstances. Contrastingly, the index theory of Montague/Lewis/Bennett is a one step account that conflates the circumstance (world-time) parameter with the contextual parameter (agent, place, object). According to Kaplan, a standard double index theory would not do the job either because in this approach two indices have to be separated but there is not a clear understanding of the role of those indices. The meaning of a sentence is, in the double index theory, a function from indices (as contexts) to functions from indices (as circumstances) to truth values, i.e. in the function space [ I —» [I —>• 2]]. Kaplan observes that nothing in this theory prevents us from proposing operators on contexts, i.e. functions in the set [[ I —> [I —> 2]] —> [ I —> [I —> 2]]].3 This type of operators on contexts are called by Kaplan "monsters". The logical representation of the following sentences would respectively contain an existential operator on contexts and a universal operator on contexts: (19)
a. In some contexts I am here, b. In all contexts this is broken.
Kaplan claims that it is not possible to have ^-operators on contexts of the form "in Q contexts ... it 3
We write [A —> B] to indicate the set of functions from A to B.
Demonstratives in Context
203
is true that...", which prefixed to a sentence would yield true if there are Q contexts c in which the contained sentence expresses a content which is true in the circumstances of that context c. For (19a) to be true, it would be sufficient that there is a context c in which the speaker of the context c is at the location of c. Obviously, this is not the correct interpretation of (19a), if (19a) is a meaningful English sentence at all. The pronoun 7 and the demonstrative this cannot shift in content as a consequence of being in the scope of an operator. These terms seem to have a directly referential interpretation, by which their reference is fixed and uniquely determined. Kaplan claims that in natural language there are no operators that can meddle with context, so a double indexing theory is flawed. Recently, Schlenker (1999, 2000) has claimed that Kaplan's statement about monsters is empirically incorrect since there are languages in which it is possible to have sentences of the form (20a) interpreted as (20b). (20)
a. John thinks that I am intelligent, b. John thinks that he is intelligent.
In English, the indexical / refers to the speaker of the utterance not to John. The coreferential reading, in which John and the pronominal subject of the embedded sentence refer to the same individual, requires that the embedded pronominal subject be the third person form, as in (20b). In other languages such as Amharic, the interpretation (20b) of a sentence with the form (20a) is possible. In other words, it is possible to obtain a coreferential reading when the embedded pronominal subject is a first person pronoun, so (20a) would come out with the same meaning as the English variant (20b). This requires, according to Schlenker, allowing the possibility of monsters, which quantify over contexts and determine the evaluation of the indexicals in the sentence they have scope over (see Von Stechow, 2001 for a discussion). Israel and Perry (1996) also come to the conclusion that there cannot be pure modal monsters but there can be epistemic monsters.
2.2. Character and Content In a series of articles (Kaplan 1978,1979), Kaplan proposed an alternative view of the semantics of indexicals based on the idea that these terms are directly referential.4 His most prominent contribution to the topic, "Demonstratives", circulated as a manuscript during the seventies and eighties in a version from 1977, and was finally published in 1989 (with an additional piece entitled "Afterthoughts").5 Kaplan's theory addresses the two main shortcomings of the single index theory 4
5
On the theory of direct reference, see Perry (1997), Recanati (1993), and Kripke (1980).
Kaplan's theory has enjoyed wide acceptance in the contemporary philosophical literature. See the contributions in Almog et al. (1989), and also the articles in Perry (1997), and Braun (1994, 1996).
204
From Words to Discourse
and of the double index theory. On the one hand, Kaplan proposes that the parameters that are associated with the context, those that specify the properties of the utterance such as the speaker, the addressee or an act of demonstration that accompanies it, operate at a different level that those that specify the circumstances of evaluation. Parameters such as the speaker or the addressee determine the "character" of an expression. In turn, the circumstances of evaluation determine the "content" of that expression. Using a functional strategy, we can define characters as functions from contextual parameters (possible contexts) to contents, and contents as functions from circumstances of evaluation to extensions. Contents are equivalent to Carnapian intensions. Graphically: (21)
Character: [Contexts —> Contents] Content: [Circumstances —» Extensions]
Indexicals have a context sensitive character, so its content varies with context. In different contexts, the content of an indexical varies. Non-indexicals have a fixed character. In different contexts, the content of a non-indexical will remain constant. The character of a non-indexical is a constant function. Kaplan proposes two principles that determine the interpretation of indexicals and demonstratives: Principle 1. The referent of a pure indexical depends on the context and the referent of a demonstrative depends on the associated demonstration. Principle 2: Indexicals, pure and demonstrative alike, are directly referential. Consider the following sentence: (22)
I am here now.
The character of the indexical 7 is a function that associates a context c with the speaker of c. Once the character is fixed, it cannot be changed. When we determine the content of (22), we cannot say that for each circumstance of evaluation we might have a different individual. Why? Because that would mix character with content. The individual the proposition in (22) is about is determined by the context of the utterance, namely, it is the speaker. In this respect, the resulting proposition is a singular proposition. The rules of linguistic convention on context do not provide a complex which together with the circumstances of evaluation yields an object. They just provide an object. A demonstrative d is accompanied by a demonstration 5. It is this demonstration that determines the character of a demonstrative. In a context c, d[5] is a directly referential term that designates the demonstratum, if any, of 8 in c, and that otherwise designates nothing. Kaplan's theory can account for the meaning of sentence (23) without resorting to lexical ambiguity:
Demonstratives in Context
(23)
205
This is faster than this.
Each occurrence of the demonstrative this will be accompanied by a different demonstration 5, and 82, so the character of £/[8,] and d[?>2] will be different. When we determine the content of (23), the designatum of each of the two occurrences of the demonstrative this will be already fixed (for instance, as the objects a and b). The proposition expressed by (23) is the set of world-time pairs <w,t> such that a is faster than b at <w,t>. The ban on monsters follows in Kaplan's logic of demonstratives from the definition of what the character of an expression is. The character of an indexical is supposed to fix its reference and anchor it to one of the features or participants of the context. Operators on context would interfere with the determination of character, since it would not be possible to associate a unique object with an indexical or a demonstrative. Thus, we can only have operators on content.
2.3. Demonstratives as Definites Kaplan's treatment of demonstratives as directly referential terms introduces an asymmetry among determiners. Most determiners head generalized quantifiers and are quantificational expressions. As such, they can enter into scope relations with each other and with other operators such as negation or modal/intensional operators. The prediction of a theory incorporating Kaplan's proposals is that demonstratives are not quantificational elements and do not participate in scope relations. This property follows from the fact that they are directly referential. Their reference is fixed once an utterance context is provided and they will not permit any variation with respect to other quantifiers B variation is one of the ingredients of narrow scope (Beghelli, Ben Shalom and Szabolcsi, 1993, 1997). In order to show the inadequacy of this prediction, we have to show examples of "varying" demonstratives. Consider the contrasts in (24,25): (24) (25)
a. b. a. b.
I did not read a book. I did not read this book. Somebody must read a book. Somebody must read this book.
The indefinite a book can take narrow scope with respect to negation or a modal verb (must), so that in (25a) any book may satisfy the descriptive content book and the sentence's interpretation is that it must be the case that somebody reads a book, i.e. it constitutes a prescription banning a world of non-readers. On the other hand, (25b) is "about" a particular book. Following Kaplan's theory, the character of the demonstrative determiner this fixes its reference to be the demonstratum (the demonstrated book) in the utterance context, and consequently the demonstrative is forced to be
206
From Words to Discourse
scopally independent.6 The interpretation of (25b) is that for the particular book being demonstrated in a context, somebody must read it (in the circumstances associated with that context). Kaplan's theory predicts that demonstratives and indexicals are like proper names and other rigid designators. The difference with a proper name would be that proper names are rigid not by virtue of their character, since its reference does not vary as a function of context. They have a fixed content and a fixed character. Demonstratives have a fixed content in a given context, but they do not have a fixed character. Their reference depends on the features of context (what the demonstratum is). To this extent, demonstratives and proper names are "rigidifiers" of content. Proper names also pattern with demonstratives in that they are scope independent. The sentences in (26) are not scopally ambiguous and only have one interpretation: (26)
a. I did not talk to John. b. Somebody must talk to John.
A litmus tests for considering demonstratives as non-quantificational elements is thus the absence of scopal interactions. Nevertheless, this and other properties that should follow from a directly referential account can be challenged. As pointed out by King (1999), there are uses in which there is no demonstration or no speaker reference. Suppose a situation in which Greg has done poorly on an exam and he knows that exactly one student received a score of one hundred. Thinking about the difficulty of the exam, Greg could say: (27)
That student who scored one hundred on the exam is a genius.
In this case, Greg employs no demonstration and is not talking about somebody who is present in the context of utterance. This seems to go against a treatment which views demonstratives as contributing only an individual to a singular proposition. Other problematic cases are those in which a bound pronoun occurs in a complex demonstrative:7 (28)
a. Every university professor cherishes that first publication of hers. b. Every father dreads that moment when his oldest child leaves home. c. Most avid snow skiers remember that first black diamond run they attempted to ski.
6
Scope independence is here different from wide scope. Demonstratives should not be taken as quantifiers taking obligatory wide scope. An example of a quantifier which always takes wide scope is each (Beghelli 1995, 1997). Nevertheless, each preserves its quantificational properties (it is a universal quantifier). Contrastingly, a demonstrative is scope independent. 7
Neale (1993) considers this type of construction marginal, as the unnaturalness of Every driver knows that mechanic working for him would indicate. Nevertheless, the existence of examples such as those in (28) seems to indicate that, contrary to Neale's claim, the occurrence of bound pronouns in complex demonstratives is quite natural. It remains as an open problem how to determine why some examples are more natural than others.
Demonstratives in Context
207
In the above examples, the pronouns within the complex demonstrative act as variables bound by the external quantifier. As a result, the embedding demonstratives do not behave as directly referential expressions. There is not a single entity that the demonstratives refer to or actually there might not be a unique entity at all. Finally, King (2001) points out the existence of narrow scope readings of some demonstratives. Consider (29): (29)
a. That professor who brought in the biggest grant in each division will be honored, b. That senator with the most seniority on each committee is to be consulted.
The complex demonstratives in these sentences are clearly not directly referential, and the demonstrative can take narrow scope with respect to each. For example, (30a,b) are possible continuations of (29a,b) respectively, which shows that the demonstrative noun phrase varies with respect to the univeral distributive quantifier each. (30)
a. In total ten professors will be honored. b. In all eight senators have to be consulted.
Demonstratives can also appear in Bach-Peters sentences, in which a sentence has two quantified elements, each one of them containing a pronominal expression bound by the other. These instances of crossing correference are often used as a test for the process of quantifier absorption of two quantificational elements (May, 1985, 1989; Clark andKeenan, 1989). (31)
Every pilot who shot at it hit the Mig which was chasing him.
In the following sentences, the crossing anaphora pattern emerges, not only when the demonstrative heads one of the noun phrases, but also when both noun phrases are demonstratives. (32)
a. Every student of yours who prepared for it passed that test he was afraid of. b. That student of yours who prepared for it passed a test he was afraid of. c. That student of yours who prepared for it passed that test he was afraid of.
More evidence of the behavior of demonstratives as quantifiers comes from weak crossover effects. It is well-known (May, 1985) that quantifiers give raise to weak crossover configurations whereas proper names do not. Sentence (33a) has an interpretation according to which Peter's mother loves him. That is not the case with the sentences (33b,c). (33)
a. His mother loves Peter. b. His mother loves every man.
208
From Words to Discourse
c. His mother loves a man. The standard explanation in terms of Chomsky's (1977) "Leftness Condition" derives the absence of the "bound" reading of the pronoun in the sentences in (33) by a restriction that forbids the coindexing of a pronoun and a trace (variable) left by movement to its left and the resulting semantic co-valuation. Proper nouns do not move at the level of Logical Form so the pronoun is not coindexed with a variable to its left. On the other hand, every man and a man are quantifiers and as such subject to the rule of Quantifier Raising, which will adjoin them to the sentential node, leaving a trace. The resulting structural configuration constitutes a violation of the Leftness Condition. Demonstratives pattern with quantifiers with respect to this property. (34) does not have an interpretation in which the man's mother loves him.8 (34)
HiSj mother loves that man.^
Demonstratives behave like pronouns in allowing E-type readings, as illustrated by the following example from Maclaran (1982): (35)
The Russians had allowed few pieces from their collection to go on show in the West, but these were the highpoint of the exhibition.
They can also have a bound variable interpretation (Roberts, 1999b): (36)
Every dog in my neighborhood, even the meanest, has an owner who thinks that that dog is a sweetie.
If, taking this evidence as a departure point, we were to assume that demonstratives are quantifiers, the next relevant question would be to determine what their quantificational force is. Russell (1905) proposes that defmites have quantificational force. The quantificational force of a definite determiner is broken down in two parts: an existential quantifier and a uniqueness condition on that quantifier, to the effect that only a single entity satisfies the restriction of the determiner: (37)
the king: 3x[king (x) A Vy[king(y) ->• y = x]]
1 King (2001) notes that there are some exceptions in which demonstratives seem to not be sensitive to crossover configurations, such as Someone who liked her asked that woman wearing a red jacket to the dance. He observes that demonstratives seem to pattern here with definites, which would also allow a bound reading in this configuration: Someone who liked her asked the woman wearing a red jacket to the dance. Universal quantifiers are still sensitive to the crossover configuration, since Someone who liked her asked every woman wearing a red jacket to the dance lacks the reading in which her is bound by the quantifier.
Demonstratives in Context
209
Although there is consensus in the literature that existence and uniqueness are the meaning blocks that compose the meaning of a definite, there has been a debate on how to treat them. For Russell, defmites carry existence and uniqueness entailments. Strawson (1950) disagreed with Russell in that he claimed that defmites do not assert/entail existence or uniqueness. Rather, they presuppose these properties, so a sentence which fails these requirements suffers from presupposition failure and cannot be considered false.9 Roberts (1999a) argues furthermore that definite descriptions and pronouns are uniformly interpreted as carrying presuppositions of existence and uniqueness. Defmites presuppose the existence of a familiar entity, which additionally is unique among familiar entities satisfying the descriptive content of the definite. We bring this discussion about defmites in because some proposals on the semantics of demonstratives basically amount to equating them with definite descriptions, with the addition of some conditions that do not operate at the semantic level properly. Maclaran (1982) and Roberts (1999b) propose to treat demonstratives as defmites. According to Maclaran, the main difference between demonstratives and defmites is that the former can denote only individuals already in the context and not those merely inferable from it. Demonstratives delimit the set of possible referents in a context. She furthermore claims that any difference between definites and demonstratives is not truth-conditional. Sometimes the difference is slight, such as in example (38a). In other cases, the use of the demonstrative is inappropriate, as in (38b): (38)
a. I heard Reagan give his acceptance speech last night. The/that man is an imbecile, b. I saw an Italian film last night. The/*this/*that director was Italian.
Roberts (1999b) claims that demonstratives differ from definites and pronouns in carrying a demonstration presupposition. A demonstrative presupposes an associated demonstration. The rigidity or direct reference property of demonstrative noun phrases is a property of the demonstration presupposition. Roberts also argues for a treatment of the dynamics of demonstratives in discourse within Heim's (1982) system of File Change Semantics. In sum, we can conclude that treating demonstratives as directly referential elements is problematic. On the other hand, exploring the similarities between demonstratives and definites requires considering additional facts and an enriched theory of the dynamics of determiners which allows us to incorporate some of Kaplan's insights. In the next section we will present a theory of demonstratives as definite determiners in generalized quantifier theory. But first we have to address an issue which is rather problematic in itself. We can accept as valid that demonstratives are dependent on contexts, but in order to provide a rigorous theory of the meaning of demonstratives, we have to clarify what we understand by context, especially if we are not reducing it to Kaplan's understanding of the term. 9
See Ostertang (1998) for a compilation of the essential readings in definite descriptions, and an excellent introduction to the subject.
210
From Words to Discourse
3. THE CONTEXT DEPENDENCE OF DEMONSTRATIVES 3.1. Context and Dynamics Independently of the approach that one follows, in the theories presented in the previous section there is agreement on the fact that demonstratives are context dependent expressions. Differences among theories are in part traceable to how context is viewed and incorporated in the semantics of expressions. The essential theoretical issue seems to be how to articulate a comprehensive, powerful (although not too powerful), and explanatorily adequate theory of context.10 For the index theory of demonstratives, context is a parameter of evaluation. In this respect, the standard interpretation of variables as terms whose interpretation is fixed by an assignment makes them close to demonstratives in that they are terms dependent on a parameter: the meaning of a variable is the value that an assignment function assigns to it. The interpretation of a formula such as Walk(x) in a model will vary depending on the value which the assignment function assigns to x. In this respect Walk(x) and Walk(y) could express the same proposition if the assignment function g assigns the same individual to x andy. The difference between variables and indexical elements resides only in which coordinate of the index is the one that determines the interpretation of a given term. The interpretation of (1st and 2nd person) pronouns will be fixed by an agent coordinate and the interpretation of demonstratives by the object coordinate. This also explains why free third person pronouns can be adequately represented as free variables. In sum, for the index theory everything in a given index can be taken as part of context. Kaplan's take on this assumption is, as we have explained above, that the resulting notion of context is not fine-grained enough. For Kaplan, the context is the context of utterance and contributes to determining the character of an expression, not its content. This narrow view of context leads us to the direct reference approach. Nevertheless, as we have also seen, there are uses of demonstratives that seem to go against a directly referential view, one in which demonstratives are fixed by the utterance context. Among those uses are quantificational, bound, and discourse dependent uses. This should lead us to a more enriched view of context such as the one defended by Maclaran (1982: chapter 5), or a view of context as discourse context within a dynamic theory of meaning (Heim, 1982; Kamp, 1981; Groenendijk and Stokhof, 1991). In dynamic theories of interpretation the context of a sentence is a discourse state. A discourse state determines the interpretation of a sentence which has been newly added to it. More concretely, in the discourse (39a) with translation (39b), a given previous discourse state modified by the addition of the subformula 3x(Man(x) A Walked_in(x}) is the 10 See Zimmermann (1991) for an excellent survey of context theory. Stalnaker (1998) contains the essential articles tracing back the author's conception of context, assertion and common ground. The special 1998 issue of the Journal of Logic, Language and Information contains several contributions to the theory of context. Aliseda, van Glabbeek and Westerstahl (1998) also contains articles summarizing recent approaches to the theory of context from different points of view (philosophical, linguistic and computational). For recent computational and logical approaches, see also Bonzon, Cavalcanti and Nossum (2000).
Demonstratives in Context
211
context for the second subformula of discourse (39a). (39)
a. A man walked in. He whistles. b. 3x(Man(x) A Walked_in(x)) A Whistle(x)
The term context, in dynamic theories such as elementary Discourse Representation Theory (DRT) or Dynamic Predicate Logic (DPL), is mostly a context of assignment or embedding. For instance, in DPL the dynamic interpretation of the existential quantifier allows for extending its scope beyond the limits of the disambiguated bracketed formula it is prefixed to (its "static" scope) so that the occurrence of the variable x restricted by whistle ends up bound by the existential quantifier, matching our understanding of this discourse, where he is dependent on the existential quantifier. Dynamic binding is not unrestricted. For example, universal quantifiers are externally static and they cannot bind a variable outside their scope. The pronoun he in (40) cannot be bound by the universal quantifier. (40)
Every man walked in. He whistles.
In this respect, a context is a set of assignments, which serves as input to the interpretation of a formula. A formula acts as a test giving as output a modified assignment set satisfying the conditions imposed by the formula. In DRT, a context for a discourse or fragment of a discourse is the Discourse Representation Structure that corresponds to the processing of the preceding discourse. A DRS is a pair < U, Con>, where U is a set of discourse referents and Con is a set of conditions on discourse referents." The context for a sentence will determine anchoring possibilities for the discourse referents introduced by the processing of noun phrases. They will also determine whether discourse referents are accessible to newly introduced discourse referents or not. In update semantics (Veltman, 1996; Groenendijk, Stokhof and Veltman, 1996) acontext is a state (a set of worlds) which represents the meaning or content associated with a fragment of discourse. A formula updates the discourse state and eliminates some of the worlds in the set: those that are not compatible with the content. A state can be made to be more fine grained or informative, by including not only worlds but also "pegs" (Landman, 1986) or discourse referents. Even a more flexible conception of context is the one recently introduced by Guha (1991), MacCarthy (1993), Buvac and MacCarthy (1998) from the point of view of AI applications and knowledge engineering. Contexts are viewed as abstract mathematical objects which may be particularly rich, as situations in a situation calculus or in situation theory (Barwise and Perry, 1983). 1 ' Heim's (1982) conception of context is similar. In File Change Semantics, a context is an ordered pair consisting of a domain (the set of familiar discourse referents) and a Satisfaction Set (a set of world-assignment pairs which represents the information in the common ground about discourse referents).
212
From Words to Discourse
Contexts are first class objects, and the basic form of a formula is ist(c,p), meaning that the proposition p is true in the context c. These formulas are always considered to be asserted in a context, so a formula of the form c': ist(c,p) means that there is an assertion in the context c' that the proposition ;? is true in the context c. This view also distinguishes between the context of assertion and the intended context of a proposition. Several relation among contexts are defined, such as the specialization relation. Contexts can also be entered and exited, etc. The above summary gives us an idea of the diverse notions of context existing in the literature. As Van Benthem (1998) points out, "the term context has a wide variety of uses in logic, linguistics and recently also in Artificial Intelligence. It tastes a little bit like 'system', 'situation', or 'state' in that it combines broad intuitive appeal with a promise of precise structure. Nevertheless, there is such a diversity of uses of 'context' that no single notion may underlie all of them."
3.2. Determiners and Context Sets In this paper, I will adhere to a dynamic view of context. A context for natural language interpretation purposes is a discourse context or conversational state. That state is the common ground (Stalnaker, 1978) shared by the participants in the conversation. In a conversation state, there will be activated discourse referents and information about those discourse referents. Given this view of context, every expression may be potentially context dependent. What might differ is the way on which context dependence is represented. Westerstahl (1985,1989) proposed that a way of capturing the context dependence of certain determiners and generalized quantifiers is by restriction. Concretely, determiner and generalized quantifier functions may be restricted to what he calls context sets. A generalized quantifier Q is a set of sets or equivalently a function from sets or properties of individuals to truth values (Barwise and Cooper, 1981; Keenan and Stavi, 1986). A determiner D is a function from sets to generalized quantifiers. Intensionally, A determiner D is a function from worlds to determiner extensions. Given a determiner D, the restriction of D to a context set C is the determiner Dc such that for any sets A, B for which the determiner is defined B e DC(A) if and only if (iff) B e D(A n C). In the functional view of quantifiers, we say that for any sets A, B, for which a determiner function is defined, DC(A)(B) = 1 iff D(A n C)(B) = 1. Context sets effectively restrict the domain of quantification of a determiner function.'2 Context 12 A similar notion is the notion of relativization, which is commonly used in logics with generalized quantifiers. We say that the relativization of an n-ary determiner D to a context set C, is a n+l-ary determiner. For instance, let D be a binary determiner. The relativization of D, Dr is the ternary function such that Dr(C)(A)(B) = 1 iff D( A n C)(B n C) = 1. The notions of relativization and restriction contrast in that the former, but not the latter, requires an increase in the type or arity of the determiner. The relativization of a binary monadic determiner, i.e. with Lindstrom type <1,1>, is a ternary monadic determiner of type < 1,1 ,'1 >. Nevertheless, if a given determiner D satisfies the properties of conservativity and extension, the properties of relativization and restriction are equivalent. In what follows, we will use these terms interchangeably.
Demonstratives in Context
213
sets represent information available in discourse, and they are presupposed. The composition of the contexts set might also be altered by topic-focus considerations (Partee, 1991, 1995; Von Fintel, 1994). In general, we follow Diesing's (1992) and Partee's (1991) conception of the semantic partition according to which topical elements are mapped in the restriction of a determiner or adverb of quantification, whereas focal elements are mapped in the nuclear scope or restricting argument of the determiner. Since the context set restricts the first argument of a determiner, the information it contains has to refer to topical or presupposed elements. With a given discourse state s, we can associate a set of context sets, those that are available because they have been introduced by a noun phrase, accommodated, or derived by processes such as summation or abstraction (Kamp and Reyle, 1993). Deliberately, we are blurring the distinction between discourse referents, as understood in DRT, and context sets.13 A discourse state is a set of pairs <w, K>, where w is a possible world and K is the set of context sets of w. We call K the generalized context set of w. The dynamic content of a determiner is a state update function: a function from states to functions from determiner denotations to states. For an arbitrary determiner D, state s, and sets A, B, the dynamic content of D is the function Ddta(s)(A)(B) = s' c s such that s' = { <w, K> | for all C e K, Dc(w)(A)(B) = 1 }. Discourse update is an eliminative process, by which possibilities that do not correspond to what is stated (the information added) are eliminated. If a determiner is context neutral, the set of contexts associated to the final state (u l s i < n Kj such that K, is the generalized context set of a world in that state) will be identical to the initial set.14 Determiners that are not restricted to a context set do not eliminate any of the members of K. There are expressions which might be unaffected by contextual restriction. The relativization of a proper name p to a context set does not affect the content of the corresponding generalized quantifier function (the individual generated by p; cf. Keenan, 1996). Whether p has a property or not will not change as a function of different context sets C, C'. In general, a determiner D is context neutral iff for any context set C, it has to be the case that Dc = D. Equivalently and restricting ourselves to determiner extensions, D is context neutral if for any sets A, B, it is has to be the case that DC(A)(B) =1 iff D(A)(B) = 1, so by the definition of context restriction C n A = A. Context dependent determiners can be said to have a determiner lift Dc that restricts them to a context set. That possibility is not available for other context neutral determiners. 13 These differences are not minor. DRT is a representational theory of the dynamics of meaning and discourse referents are entities of the mediating language of Discourse Representation Structures. 14 This point amounts to another difference with DRT semantics. In DRT, certain determiners contribute new discourse referents. Interpretation is an incremental process by which the set of discourse referents is properly expanded with the addition of novel discourse referents and also with the addition of those which are familiar (in this case the discourse referent would be equal to one already available, so the expansion is not proper). Discourse updates are consequently additive, to mirror the process of DRS construction. In an eliminative theory of discourse updates, possibilities are eliminated with discourse growth. Thus, a given noun phrase cannot be viewed as contributing a discourse referent but rather as imposing a test on the context sets of a discourse state. The subset of the discourse state whose context sets satisfy the determiner content "survives", and the remaining subset is eliminated. See Groenendjk, Stokhof and Veltman (1996) for a partially similar view.
214
From Words to Discourse
Westerstahl proposes treating pronouns as pronominal determiners, for which restriction to a context set is not optional. A personal pronoun such as they denotes the pronominal determiner function theyc such that theyc(B) = 1 iff C c B. The relativized version in the higher (determiner) type is they(C)(B) = 1 iff C c; B. The calculation of the composition of the context set has to follow the same criteria for dynamic interpretation posited in current theories of discourse interpretation. Thus, in (41) the set C will be the set of three men who arrived. This view of pronouns is very similar to one based on the E-type strategy (Evans, 1980; Heim, 1990; see also van der Does, 1995). (41)
Three men arrived. They were bald.
Clitic pronouns can also be conceived of as pronominal determiners obligatorily restricted to a context set. As claimed in Gutierrez-Rexach (1999a), in pronominal doubling constructions the context set is retrieved from the doubled quantifier or quantifier associate. In (42), the pronominal clitic los has as its context set the generator set of the principal filter denoted by the noun phrase tus hermanos, i.e. the group of brothers referred to. (42)
Los he visto a tus hermanos. them have-I seen A your brothers "I have seen your brothers."
In Spanish, there are two lexically different existential determiners, which also function as existential pronouns: unos and algunos. If we propose that they both denote a plural existential determiner (a function D such that for all sets A, B, D(A)(B) = 1 iff card(A n B) > 2), we would have to claim that there is a redundancy in the determiner system of Spanish. In Gutierrez-Rexach (200la), it is claimed that there is empirical evidence showing that the two determiners are not equivalent. Algunos can receive a partitive or specific interpretation (En?, 1991) whereas unos cannot. In DRT terms, the discourse referent contributed to a Discourse Representation Structure by algunos can be linked to a previously introduced or accommodated discourse referent. This is not the case with unos, which obeys a "no linking" constraint. As expected, the covert or overt partitive variant with algunos is grammatical in (43) but the variant with unos is not. (43)
a. Llegaron los invitados. Algunos/*unos trajeron regalos. Arrived the guests. Some-pl./-a-pl. brought presents "The guests arrived. Some brought presents." b. Fuse los libros de matemdticas sobre la mesa y algunos/??unos defisica en el cajon. Put-I the books of mathematics on the table and some/apl. of physics in the drawer "I put the mathematics books on the table and some linguistics books on the drawer."
Demonstratives in Context
215
The contrast can be explained if it is assumed that algunos can denote the function algunosc. In other words, it can be restricted to a context set as follows: algunosc(A)(B) = 1 iff Card((C n A) n B) > 2. Unos cannot be restricted by a context set, which predicts that in (43a) the set C = GUESTS n ARRIVED cannot be used to compute the meaning of the second sentence in the discourse. In Heim's (1982) terminology, unos is a novel indefinite. Pronouns and definite determiners receive a similar treatment in terms of context set restriction. A definite determiner can be taken to be a marker of the presence of a context set. A definite determiner is the restriction of a universal (co-intersective) determiner to a context set, which is familiar in discourse. (44)
English: the(A)(B) = allc(A)(B) A Faniiliar(C) Spanish: los (A)(B) = todosc(A)(B) A Familiar(C)
As a matter of fact, in Spanish the simple plural determiner todos cannot take a restriction directly. Rather it has to combine obligatory with a co-intersective determiner (definites, possessives, demonstratives). This seems to corroborate to some extent Westerstahl's methodological proposal about considering definite determiners as context set markers. In the sequence 'universal det + definite determiner' the presence of the co-intersective determiner indicates that the universal determiner is restricted to a context set. (45)
a. *Todos ninos vinieron. all kids came. b. Todos los ninos vinieron. all the kids came. c. Todos mis libros estdn sucios. all my books are dirty. d. Todos esos clientes llegaron tarde. all those customers arrived late.
Salience or uniqueness presuppositions can also be incorporated into the definition of a definite determiner as further conditions that the context set has to satisfy along the following lines: (46)
the(A)(B) = allc(A)(B) A Familiar(C) (A Unique(C))
Familiarity indicates that the context set associated with the definite is already present in discourse. As is well-known (Hawkins, 1978; Heim, 1982), definites can be strongly or weakly familiar. Strong familiarity corresponds to anaphoric uses of definites, where the context set of the determiner is a set already presupposed or available in the discourse state and prominent among available context sets.
216
From Words to Discourse
Weak familiarity only requires that the context set be associated to a presupposed set by a discourse relation, mostly accommodation, bridging, etc. Consequently, we can speak of strong or weak familiarity of the context set of a definite. For instance, the context set of the definite the student is strongly familiar. The context set of the bride is only weakly familiar.15 (47)
a. A student and a professor met at the gym. The student said'hello', b. When I go to a wedding I always kiss the bride.
3.3. Constraints on the Interpretation of Demonstratives Let us now consider how to develop a semantic for demonstratives as definite determiners within generalized quantifier theory. Demonstratives are co-intersective determiners and exhibit the same behavior as defmites and universals in existential contexts. The sentences in (48), where (48b,d) are respectively the Spanish correlates of (48a,c), are not grammatical (understanding this term in its broadest sense to also include semantic anomaly). (48)
a. b. c d.
*There are these students in the garden, *Hay estos estudiantes en el jardin. The table has these legs, *La mesa tiene estas patas.
Indexicals relate the context set to the utterance situation. The semantics of first and second pronouns is identical to the semantics of third person pronouns as described above except for the fact that the context set has to be the speaker of the utterance, in the case of/, and the addressee in the case ofyou: (49)
I C (B)= 1 i f f C c B ACard(C) = l Presupposition: Speakerutt(C) Youc(B) = 1 iff C c B A Card(C) > 1 Presupposition: Addresseeutt(C)
Utterances are not part of the context set. Thus, we will talk about the utterance state (cf. Barwise and Perry's (1983) discourse situation) rather than about the utterance context, as Kaplan does, to 15 Clark (1978) calls this class of defmites "bridging" defmites, and Hawkins calls them "associative" defmites. For more details on the process of bridging in a dynamic framework, see Asher and Lascarides (1996, 1998).
Demonstratives in Context
217
avoid potential confusion. Context sets contain only individuals. If the individual(s) in C is/are the speaker or addressee of the utterance state, this is a property they have which is incorporated as a presupposition to the determiner meaning. The elements of meaning that Kaplan treats as belonging to the character of indexicals, I will treat as presupposed properties of the individuals in the context set. Roberts (1999b) and Zeevat (1999) also treat elements belonging to character as presuppositions. Demonstratives are like definites in having an inclusion condition as the main part of their content. Additionally, they are also obligatorily dependent on a context set. So far we have the same semantics as a definite. What makes the difference is the nature of the additional constraints that the context set of the determiner has to satisfy. Those constraints can be taken again to be presuppositional in nature (Roberts, 1999b; Zeevat, 1999). (50)
thisc(A)(B) Presupposition: Demonstratedutt(C) A Proximalutt (C) A Unique(C) A Weakly familiar(C)
The constraint Demonstratedult(C) indicates that the context set is demonstrated in the utterance state. The remaining properties (proximity, uniqueness and weak familiarity) are also properties of the context set. The property of proximity makes reference to the utterance state, whereas uniqueness and weak familiarity do not. Zeevat's (1999) proposal for indexicals and definites in DRT requires that, for example, in the case of an indexical such as /, the Discourse Representation Structure that is taken as a point of departure is not empty. It contains, by presupposition accommodation, the discourse referents e and x, and the conditions utterance(e) and agent(x,e). In other words, e is the utterance event and x is the speaker of e. The discourse referents are externally anchored (Kamp 1990) to the speaker and speech event.
4. SPANISH DEMONSTRATIVES 4.1. Basic Characterization In this section, I will consider several data on Spanish demonstratives that will allow us to elaborate and make more precise the theory of demonstratives that we have just sketched. Spanish has a threeterm system of demonstrative determiners and pronouns: este - ese- aquel, with their variants in number (singular plural) and gender (masculine, feminine). There is also a neuter series: esto -esoaquello. Usually, the difference among terms is characterized in terms of the proximal/distal distinction in combination with some other parameter. Andres Bello (1847) stated that demonstrative pronouns are used to point at objects indicating their situation with respect to a person: este 'this'
218
From Words to Discourse
denotes proximity of the object with respect to the first person (the speaker); ese 'that' 16 indicates proximity with respect to the second person (addressee); and aquel 'that' demonstrates an object which is distant with respect to the speaker and the addressee. Thus, translating Belle's ideas in terms of feature specification with respect to the features [+/-proximal] and [+speaker]/[+addressee] we obtain the following lexical entries: (51)
este [+proximal] [+speaker] ese [+proximal][+addressee] aquel [-proximal]
Semantically, the corresponding definition would have to associate demonstratives not only with a demonstrated object in the utterance state, but also with a proximity relation associating the context set with the speaker of the utterance. The definition of Spanish demonstratives would require an enrichment of the definitions given in the previous section, to relate the proximity condition to the participants in the utterance state. The resulting characterizations would be the following ones, where the first three conditions are taken to be presuppositions properly. The condition "proximal" relates the context set to an utterance participant or participants. (52)
estec(A)(B) = 1 iff Demonstratedutt(C) A Proximal(C, speakerutt) A Card(C) = 1 A
(C n A) c B esec(A)(B) = 1 iff Demonstratedutt(C) A Proximal(C, addresseeutt) A Card(C) = 1 A
(C n A) c B aquelc(A)(B) = 1 iff Demonstratedutt(C) A -,Proximal(C, participant,,,,) A Card(C) = 1 A (C n A) c B The basic inclusion condition that holds between the set C n A (the intersection of the noun argument A and the context set) and the predicate argument B allows us to assimilate demonstratives to definites in the property of co-intersectivity (Keenan, 1996). We predict that demonstrative noun phrases will not be able to occur in existential constructions, a property characteristic only of generalized existential determiners (Keenan, 1987): (53)
a. *Hay este libra sobre la mesa. there-is this book on the table
16 In what follows, I will translate the Spanish demonstrative ese with the English demonstrative that. Nevertheless, as will be clear from the discussion in the next pages, this translation does not entail that ese is equivalent in meaning to aquel.
Demonstratives in Context
(54)
219
b. *Hay aquellos. There-is those-ones Hay *estos/*los/unos libros. there-is these/the/a-pl. books
When the meaning ofhaber is not existential but locative (Hornstein, Rosen, and Uriagereka, 1995; Gutierrez-Rexach, 2000), demonstratives become licensed, whereas defmites do not. Locative haber constructions simply state the position of one object. Spatial deixis is compatible with the information required. (55)
Hay estos/*los libros sobre la mesa. there-is these/the books over the table
A second natural prediction is that, when demonstratives appear in postnominal or "adjectival" position, the main determiner has to be definite. This fact is explained as a semantic incompatibility between the inclusion condition associated with the demonstrative and the intersection condition associated with an indefinite. (56)
a. el libra este/aquel the book this/that b. *un libra este/aquel a book this/that
Generalized existential determiners do not appear in the determiner position when the demonstrative is in the postnominal or adjectival position either. For instance, (57) shows that muchos and varios are incompatible with postnominal demonstratives. On the other hand, other inclusive or cointersective determiners, such as the possessive mis 'my-pl.' in (58), can appear in the determiner (prenominal) position co-occurring with postnominal demonstratives.17 (57)
*muchos/*varios libros estos many/several books those
17 An exception to this generalization are intrinsically distributive universals, which cannot appear in the determiner position. Compare (i) with (ii):
(i) (ii)
*cada/*todo libra ese each/every book that todos los libros esos all the books those
220
(58)
From Words to Discourse
mis libros esos de los que te hable my books those of the that you talked-I "Those several books of mine that I talked to you about."
A uniqueness presupposition has not been included in the characterization given in (52), because there is evidence that the uniqueness condition might be just a defeasible implicature rather than a presupposition. Leonetti (1996) presents some data suggesting several differences between demonstratives and definites that seem to lead to this hypothesis. First, postnominal evaluative modifiers and elatives are incompatible with definites and not with demonstratives, as shown in (59, 60), a fact also noted by Brucart and Gratia (1986). When these modifiers appear in prenominal position they can co-occur with demonstratives and definites, as in (61). (59)
(60)
(61)
a. *He visto la pelicula malisima de Michael Bay. have-I seen the movie very-bad of Michael Bay "I have seen the very bad movie of Michael Bay" b. *He visto al jugador increible de los Lakers. have-I seen to-the player incredible of the Lakers a. He visto esa pelicula malisma de Michael Bay. have-I seen that movie very-bad of Michael Bay "I have seen the very bad movie of Michael Bay" b. He visto a ese jugador increible de los Lakers. have-I seen to that player incredible of the Lakers a. He visto la/esa malisma pelicula de Michael Bay. have-I seen the/that very-bad movie of Michael Bay b. He visto a ese increible jugador de los Lakers. have-I seen to that incredible player of the Lakers
Evaluative expressions and elatives have typically non-restrictive content. This might be at the root of why the sentences in (59) improve when more restrictive modifiers are added. (62)
a. He visto la pelicula malisima de Michael Bay de la quehablamos el otro dia. have-I seen the movie very-bad of Michael Bay of the that talked-we the other day "I have seen the very bad movie of Michael Bay that we talked about the other day." b. He visto al jugador increible de los Lakers al que todos adoran. have-I seen to-the player incredible of the Lakers to-the all adore "I have seen the incredible Lakers' player that everybody adores."
Demonstratives in Context
221
The contrast betwen (59) and (62) can be attributed precisely to the uniqueness presupposition that definites have to satisfy. This condition is not satisfied by postnominal evaluative expressions, which are not triggers of uniqueness/specificity presuppositions. On the other hand, other elements in the clause such as the restrictive relative clauses in (62) may act as satisfactory triggers of uniqueness presuppositions, since they add the necessary information to uniquely identify an individual. Similarly, Bosque (1996,2001) observes that prenominal adjectives are associated with uniqueness/specificity restrictions. This fact explains the grammaticality of (61). When evaluative or elatives appear prenominally, a uniqueness presupposition is syntactically triggered independently of the lexical nature of the adjective. There are nouns which also are associated with uniqueness requirements, such as fact, height, or weight. In the fact that you arrived late, or in the weight of the parcel, the uniqueness entailment of the noun phrase comes not from the article but is also a requirement triggered by the meaning of these nouns. It is obvious, for example, that a parcel cannot have more than one weight. (63)
a. el/*'este peso del paquete the/*this weight of-the parcel b. la /*esta altura del edificio The/*this height of-the building c. el/*este hecho de que viniste The/*this fact of that came-you "The/*this fact that you came"
The above data pertaining to adjectives, nouns and to their uniqueness properties seems to strongly suggest that demonstratives do not presuppose uniqueness. There are more syntactic facts that point in that direction. There is a type of exclamative/emphatic clauses (see Leonetti, 1996; GutierrezRexach, 1999b) that obligatorily select the definite article: (64)
a. (Lo/*esto alto que es Juan\ the-neut./this-neut. tall that is Juan "How tall John is!" b. (La/*esta cara que tienes\ The/this face that have-you "You have a lot of nerve!"
Gutierrez-Rexach (1999b) claims that in these constructions of "high or extreme degree" (Postma, 1996) there is a hidden Russellian (iota) operator binding a degree variable. (64a) is interpreted as the unique degree of height such that Juan is tall to that degree and the emotive attitude that the speaker expresses towards it (the exclamative content). Matching requirements force that the
222
From Words to Discourse
determiner head be occupied by a definite. If demonstratives and definites were equivalent with respect to their uniqueness presuppositions, demonstratives should be able to appear in (64). Since this is not the case, we can also conclude from this piece of evidence that demonstratives do not presuppose uniqueness.18 Naturally, this does not mean that uniqueness is not implicated in some occasions. For instance, from (65) we infer that the addressee received only one bicycle from the speaker. The implicature only arises when the demonstrative is used non-contrastively. Compare in this respect (65) with (66). (66) has no uniqueness implicature precisely because the demonstrative is used contrastively. The speaker points at one flower pot to separate it from other flower pots in the perceptual environment. Sentence (67a) is a good discourse continuation of (66). The continuation (67b) is not possible for (65). This contrast seems to indicate that a deeper analysis of additional meaning components of demonstratives is needed, especially contrast, as we defend in the next section. (65)
(66)
(67)
lie acuerdas de aquella bicicleta que te regale? You remember of that bicycle that you gave-I "Do you remember that bicycle that I gave you?" Me gusta aquella maceta. me like that flower-pot "I like that flower pot." a. ... pero no esta. but not this-one b. ... pero no de esta. but not of this
Additionally, extraction from noun phrases exhibits the same pattern of contrast (Leonetti, 1996): (i) a. No me nan ensenado la/??esta foto de ningun sospechoso. Not me have-they shown the/??this photo of no suspect "They have not shown me the/this picture of no suspect" b. Carlos es el/??este representante de los alumnos. Carlos is the/this representative of the students "Carlos is the students' representative"
Demonstratives in Context
223
4.2. Extensions 4.2.1. A Basic Bipartite Distinction Bello's (1847) theory would predict that, if (68) is uttered in a conversation between agents Ag, and Ag2, then Ag, (speakerutt) must be making three demonstrations. The first demonstratum has to be close to the speaker (Ag^, the second one close to Ag2 (addresseeult), and the third demonstrated book has to be distant from both the speaker and the addressee. (68)
No quiero que me traigas este o ese libro, traeme aquel. not want-I that me bring-you this or that book, bring-me that "I don't want you to bring me this book or that one, bring me that other one."
Nevertheless, the generalization is not completely accurate. It may be the case that (68) is felicitous and true in the situation described above. However, it may also be true in a situation in which the first two demonstrated books (associated with este and ese) are close to the speaker and the third one is closer to the addressee, as explicitly stated in (69). (69)
No quiero que me traigas este o ese libro que estdn por aqui, traeme aquel not want-I that me bring-you this or that book that are by there, bring-me that-one que estd cerca de ti. that is close of you "I don't want you to bring me this or that book that are close by. Bring me the one that is closer to you."
In this scenario, not only ese is used to indicate proximity to the speaker but aquel is used to indicate proximity to the addressee. Thus, the relevant distinction cannot be characterized only in terms of the [+/-proximal] feature with respect to a conversation participant as stated in the previous subsection. What sentence (69) shows is a contrastive use of determiners, where the three-term system is exploited to indicate a positional contrast between three objects, two closer to the speaker and one more distant with respect to him. Distance with respect to the addressee is secondary in this case. Sometimes, a contrast between ese and aquel can be exploited to indicate two different objects which are distant from the speaker but closer to the addressee: (70)
No me traigas este libro de aqui, traeme ese o aquel que estdn cerca de ti. not me bring-you this book of here, bring-me that or that that are close of you "Don't bring me this book from over here, bring me that one or that one closer to you."
224
From Words to Discourse
In the above example, the book associated with este would be closer to the speaker and those associated with the addressee would be distant from the speaker. Nevertheless, the use of ese as opposed to aquel does not indicate that one object is closer to the addressee than the other. They can be equally distant from him, and two distinct demonstratives are simply used for contrastive purposes. Thus, this evidence indicates that there is a great degree of flexibility in the demonstration associated with Spanish demonstratives vis-a-vis the two parameters used by Bello (proximity/agent). Fernandez Ramirez (1951), in his excellent chapter on demonstratives and pronouns, acknowledges this point. He follows the ideas of the German philosopher Karl Biihler (1937). Biihler distinguishes three types of demonstration: "demonstratio ad oculos" (spatial deixis), "demonstratio reflexiva" (syntactic/ discourse anaphora) and "deixis am phantasma" (figurative or non-spatial deixis). Demonstratives can function as purely deictic terms, as anaphoric terms, and to express demonstrations which are not purely spatio-temporal (cf. also the analysis in Garrido Medina, 1991: chapter 2.1). With respect to spatial deixis, Fernandez Ramirez points out that the term series este-ese-aquel constitutes "a gradation in distance which behaves in this increasing order when relative pointings take place" (§127). In a scenario in which a higher officer is telling a sergeant which soldiers to pick from those standing in formation, the series is used as described by Fernandez Ramirez, with each demonstrative associated with a more distant object. (71)
Quiero que seleccione para la mision a este, ese y aquel soldado. want-I that select-you for the mission to this, that and that soldier "I want you to choose this soldier, that soldier and that other one for the mission."
The strict order theory sketched by Fernandez Ramirez predicts that for a point of reference r and dr the distance of the demonstration 5 with respect to r the following holds: dr($[este]) < dr (8[esej) < dr (5[aquel]). The ordering expressed by (71) probably conforms to this pattern. Nevertheless, if we go back to examples (68-70), we see that strict ordering is not necessary, since we might have este and ese demonstrating equally closer points and aquel a distant point or este as the proximal term, and ese, aquel as the equally distant terms. These facts suggest that ese works as a neutral term with respect to spatial positioning and can be used in situations where the demonstrated object is close to, or distant from, one of the agents of the utterance context. In situations in which there is only one object which is clearly close to the speaker and which is being pointed out by him, este will be used. On the other hand, ese will be used if the object is either not clearly close to the speaker or it is but the speaker does not want to highlight or emphasize this fact, for instance, because the position of the object is in the common ground. Thus, (72a) and (72b) could be both used when the speaker is asking the addressee to give him an object which is close to the speaker. (72a) would be a preferred option unless the speaker does not know the position of the object, that position is in the common ground, or he does not want to make that spatial presupposition salient.
Demonstratives in Context
(72)
225
a. Dame esto. give-me this "Give me this." b. Dame eso. give-me that "Give me that."
4.2.2. Interactions with Other Deictic Terms Demonstratives interact with other space and time deictic elements. The deictic adverbs aqui, alii and alia are also opposed by a proximal/non proximal distinction. Aqui is [+ proximal] and alii and alia are [- proximal]. The adverb ahi plays the same role as ese in the nominal domain.19 Demonstratives can co-occur with adverbial deictics. Interestingly, there has to be matching in the value of the [+/-proximal] features. Semantically, there cannot be presupposition conflict between the specification of the deictic elements. Este can combine with aqui, and aquel with alii but the reverse is not possible. The demonstrative este cannot co-occur with alii and aquel cannot co-occur with aqui. Este and aquel can both combine with the neutral term ahi. (73)
a. este libra de aqui this book of here "this book from here" b. aquel libra de alii that book of there "that book from there" c. *este libra de alii this book of there "this book from there" d. *''aquel libro de aqui that book of here "that book from here" e. Qste/aquel libro de ahi this/that book of (t)here "this/that book from (t)here"
19 There is a weak or clitic proform ahi, which not only acts as a spatially deictic element but also can participate in temporal and nominal pseudo-doubling structures. In Gutierrez and Silva (1998) and Gutierrez-Rexach (2001b), the syntactic and semantics properties of this proform are studied. In the latter paper, Gutierrez-Rexach's (1999a) account of weak or clitic pronouns based on context set retrieval from the doubled associate is extended to this adverbial clitic.
226
From Words to Discourse
The neutral character of ese is again made explicit by the fact that it can combine both with the proximal term aqui and the distal one alii, as well as with neutral ahi. (74)
Dame eso que estd aqui/alli/ahi. give-me that that is here/there/there.
There is also evidence that a matching between demonstratives and temporal deictics is also required. The following examples from Silva and Gutierrez (2001) illustrate how the deictic features may interact with tense and aspectual properties. In (75a), only the demonstrative aquel is compatible with the simple perfect plantee. The reverse situation arises in (75b), where the demonstrative aquel cannot occur with the present form planteo, but the form este is okay in that context. Finally, (75c) is compatible with both temporal forms. (75)
a. iQue resolviste de aquelproblema que te *planteo/plantee? what resolved-you of that problem that to-you pose-I/posed-I "What did you resolve about that problem that I pose/posed to you?" b. iQue resolviste de este problema que te planteo/*plantee! what resolved-you of this problem that to-you pose-I/posed-I "What did you resolve about this problem that I pose/posed to you?" c. iQue resolviste de ese problema que te planteo/plantee? what resolved-you of DEM problem that to-you pose-I/posed-I
The demonstratives este and aquel constitute a binary opposition that is related to the proximal/distal contrast. The demonstrative ese is an unmarked element that may take either value, overriding the basic proximal/distal opposition. The presence of a demonstrative introduces perspectival and aspectual properties. Este associates with tenses that are perceived as close to the speech time (present, progressive, present perfect), whereas aquel associates with those that are perceived as distant (simple past, pluperfect). The tense/aspect correlation with demonstratives is better characterized in these terms than along the past/non-past dimension, as shown in the following examples:20 (76)
20
a. iQue resolviste de este problema que te estoy planteando/he planteado/ what resolved-you of this problem that you am-I posing / have posed / *estaba planteandol*habia planteado? was posing / had posed?
The contrasts shown in (76) are better perceived when demonstratives are used in spatio-temporal or discourse demonstrations. When the use is evocative or figurative, the aspectual distinction gets blurred and they become acceptable in all environments.
Demonstratives in Context
227
b. iQue resolviste de aquelproblema que te *estoy planteando/*he planteado/ what resolved-you of that problem that you am-I posing / have posed / estaba planteando/ habia planteado? was posing / had posed c. iQue resolviste de ese problema que te estoy planteando/he planteado/ what resolved-you of DEM problem that you am-I posing / have posed/ estaba planteando/habia planteado? was posing / had posed? The present progressive estoy planteando and the present perfect he planteado are related to the speech time, thus satisfying a proximity presupposition. The imperfect progressive estaba planteando and the pluperfect habia planteado are aspectually related to a non-proximity presupposition. The demonstrative ese behaves as a neutral form, being able to combine with any aspectual variant. This redistribution of a tripartite demonstrative system leads us to the conclusion that the Spanish demonstrative system is better seen as a bipartite system in which este opposes to aquel, and ese acts like a neutral form. This seems to indicate an underlying convergence of demonstrative systems, where ternary oppositions are recategorized as binary oppositions. 4.2.3. Deixis am Phantasma What Buhler called "deixis am phantasma" (figurative deixis) can be also viewed as part of the presuppositional enrichment of the demonstrative, by which the demonstration and proximity constraints are moved from the spatio-temporal axis in the actual world to involve demonstration of alternative situations or worlds or of context sets in these worlds/situations. Consider (77): (77)
Echo de menos aquellos libros throw of less those books "I miss those books forever lost."
para siempre perdidos. for always lost
The use of the demonstrative aquellos in the above sentence does not require that the speaker have actual spatio/temporal contact with the books in question. The demonstrative is used to evoke them i.e. to make them cognitively "active" in the terminology of Gundel, Hedberg and Zacharski (1989, 1993). The context set is required to be active in a world which is not necessarily the actual world. Additionally, the familiarity constraint that defmites and demonstratives have to satisfy does not have to be satisfied either. Evoked demonstratives can be discourse novel, which brings them closer to the properties of indefinites.21 21
In (i), the use of the demonstrative is adequate when neither the speaker nor the addressee share or able to identify a common set of times, or even if they themselves were not happy in their childhood.
228
From Words to Discourse
The demonstratives este and ese can be used in figurative and evocative contexts, but este cannot as (78) shows. What this suggests is that the capacity of a demonstrative to participate in evocative deixis requires that its use is not associated to a proximity presupposition, as is always the case with este. Furthermore, evocative uses of demonstratives demonstrate context sets in the belief base of the speaker, sometimes rather vaguely. The use of este requires existence in the actual world (the utterance world), which would lead to conflicting presuppositions. (78)
*Echo de menos estos libros para siempre perdidos. throw of less these books for always lost "I miss these books forever lost."
Evoked context sets are non-proximal to the actual (utterance) world, as we have seen. There are other possibilities of exploiting the proximal/distal presupposition in this wider sense. Depreciative and emphatic uses are also instances of "deixis am phantasma." In this case, the non-proximal (distal) presupposition indicates distance with respect to normative criteria in the modal base (Kratzer, 1981) of the speaker. Sentence (79a) can be used to indicate disapproval, and so can (79b) where the demonstrative ese is in postnominal position. (79)
a. Ese idiota de Juan me va a oir. That idiot of Juan me goes to listen "That idiot of Juan is gonna have a piece of my mind." b. El hombre ese es un idiota. the man that is an idiot "That man is an idiot."
4.2.4. Discourse Deixis The "neutral" term ese is excluded in general from discourse deixis. When only one term is present in previous discourse and this term is deictically referred to, este is used. Ese can also be used in one term deixis, but its use is more "charged", in that it brings an emphatic or evocative association as we observed in the previous section. When there are two terms available for discourse deixis, este is used to demonstrate the closest term in discourse from a linear perspective, and aquel is associated with the most distant term.
(i)
Oh, aquellos tiempos de la infancia en los que se era'tanfeliz. "Oh, those childhood times when one was so happy."
Demonstratives in Context
229
For example, in the discourses in (80) when the demonstrative use is purely "anaphoric", i.e. used merely to establish a connection with a term in previous discourse, este is preferred. The use of ese in (80b) may bring in intensional associations: probably the soldier is unknown, etc. (80)
a. Juana estuvo bailando con un soldado. A este soldado, al que llaman Mateo Juana was dancing with a soldier, to this soldier, to-the that call Mateo lefaltaba un brazo. To-him miss an arm "Juana was dancing with a soldier. This soldier was missing an arm." b. En el rio hay un soldado muerto. A ese soldado lo llamaban Mateo. In the river there-is a soldier dead, to that soldier him called-they Mateo "There is a dead soldier in the river. They called that soldier Mateo."
The discourses in (81) contrast in a very revealing fashion. Este and aquel establish a contrast between two terms in the preceding discourse. In all three cases, the condition that este demonstrates the linearly closest element and aquel demonstrates the linearly most distant one is satisfied. Interestingly, the discourse becomes less felicitous when the terms are "too close" and the proximal/distal distinction is blurred. In (81 c), Mateo and Luis are two coordinated constituents, so a discourse distance parameter can only marginally be applied to them. The anaphoric contrasting pronouns uno/otro 'one... the other' are preferred, as in (82). These contrasting pronouns differ from demonstratives in that, as expected, they lack demonstrative character, so the association with the proximal or distal antecedent is flexible. The default association pattern is the one that associates uno and otro respectively with the leftmost and rightmost elements. (81)
a. Mateo le dijo a Juana que habia hablado con Luis. Juana siempre le cree Mateo to-her said to Juana that had-he talked to Luis. Juana always him believes a este pero no a aquel. to this-one but not to that-one "Mateo told Juana that he had talked to Luis. Juana always believes the latter but not the former." b. Mateo siempre se pelea con Luis. A este le molestaba lo que el otro dice Mateo always refl. fight with Luis to this him bothered what that the other says y aquel estd deperpetuo mal humor. and that-one is in perpetual bad mood "Mateo always fights with Luis. The latter is bothered by which the other says and the former is always in a bad mood." c. Juana fue novia de dos soldados, Mateo y Luis. ?A este le faltaba un brazo Juana was girlfriend of two soldiers Mateo and Luis to this him missed an arm
230
(82)
From Words to Discourse
y a aquel una pierna. and to that a leg "Juana was the girlfriend of two soldiers, Mateo and Luis. The former had only one leg and the latter one arm." Juana fue novia de dos soldados, Mateo y Luis. Al uno le faltaba un brazo y Juana was girlfriend of two soldiers Mateo and Luis, to-the one him missed an arm and al otro una pierna. to-the other a leg "Juana was the girlfriend of two soldiers, Mateo and Luis. One was missing an arm and the other a leg."
Discourse anaphoric connections are not possible with all the co-intersective determiners. As is well known, in the Principles and Parameters theory Principle-C of the binding theory (Chomsky, 1981) excludes precisely the possibility that referential expressions be "bound" or co-indexed (co-valued) with a preceding and c-commanding noun phrase. (See Macia, 2002 for a semantic analysis of the binding theory). This is a narrow characterization that captures the restrictions in binding and coreference only partially. More generally, a definite cannot be neither syntactically bound nor coreferential with a preceding referential expression, for example a proper noun:22 (83)
a. *Juani cree que el espanol{ se ird. Juan believes that the Spaniard refl. will-go "Juan believes that the Spaniard will leave." b. *Juan y Pedro han leido Guerray PaZj peroyonohe leido el libra/. Juan y Pedro have read War and Peace but I not have read the book "Juan and Pedro have read War and Peace but I haven't read the book." c. *Juanf vino y el espanolt se fue despues. Juan came and the Spaniard refl. left afterwards "Juan came and the Spaniard left afterwards."
Demonstratives are, as we have seen, an exception to this generalization, although not to principle-C in its narrowest sense. They are clearly "referential expressions"23 but they participate in textual 22
We are excluding those cases in which the definite noun phrase is an "epithet":
(i) Juani queria invitarnos pero el muy idiotat se dejo el dinero en casa. Juan wanted to-invite-us but the very idiot refl. left the money in house "Juan wanted to invite us but the idiot forgot his money at home." 23
Culicover (1997: 60) takes as referential expressions all those not covered by principles A (anaphors) and B (pronouns): "Binding theory characterizes the binding possibilities for NPs that are not pronouns and not anaphors; these are called R-expressions (for referring expressions). An R-expression is not referentially dependent, which means that semantically it cannot have an antecedent, and syntactically it cannot be bound."
Demonstratives in Context
231
deixis. Semantically this goes against a potential generalization that would ban co-intersective elements from anaphoric, bound or coreferential readings. In (84), este libra may be coreferential with the proper noun GuerrayPaz. In the terms of our theory, the context set for the demonstrative is the generator of the generalized quantifier denoted by the proper name. If we compare (84a) and (84b), we see that noun phrases where the demonstrative occurs postnominally tend to behave as defmites with respect to their referential possibilities. (84)
a. Juan y Pedro han leido Guerra y Paztpero yo no he leido este librot. Juan y Pedro have read War and Peace but I not have read this book "Juan and Pedro have read War and Peace but I haven't read this book." b. *Juan y Pedro han leido Guerra y Paztpero yo no he leido el libra estet. "Juan y Pedro have read War and Peace but I not have read the book this."
This fact suggests that we have been calling "discourse deixis" is indeed a deixis phenomenon, more than an instance of mere coreference or textual anaphora. Consequently, not only previous discourse has to provide a suitable context set for the demonstrative but also the contextual presuppositions of the demonstratives have to be satisfied. For instance, an adequate context set C for este has to satisfy the presuppositions Demonstratedut,(Q and Proximal(C, speaker^. These two constraints are obviously restricted to the textual dimension here. 4.2.5. More than One Demonstrative In Spanish, the occurrence of postnominal demonstratives in adjectival position is common. Postnominal demonstratives tend to be associated with emphatic or evocative uses, for which the presupposed demonstration is not spatio-temporal in nature, as we have already shown. (85)
A: Dame la carta. give-me the letter B: iQue cartal what letter? A: La carta esa que te mencione. the letter that that to-you mentioned-I "That letter that I mentioned to you."
Previously, we observed that co-intersective determines can in general occupy the prenominal determiner position when the demonstrative appears in postnominal position. Demonstratives can appear both prenominally and postnominally, so we have structures in which there are actually two demonstratives (Silva and Gutierrez, 2001) and the postnominal demonstrative serves as a focalizer:
232
(86)
From Words to Discourse
aquellos/estos/esos tres coches esos DEM/DEM/DEM three cars DEM
There are several restrictions at work in noun phrases with double demonstratives. First, only ese can occur in the postnominal position as the demonstrative realizing emphasis or focus. Second, an additional determiner can follow the initial demonstrative. In this case, the noun must be preceded by a weak quantifier (87b). Third, the postnominal demonstrative must have a role focusing the noun phrase. Silva and Gutierrez (2001) conclude from these facts that ese is the unmarked focus demonstrative item. This point was already observed by Vincent (1997), who claims that the Latin preform IPSE, form which ese originates, "singles out an item that has been previously mentioned but has lapsed into the background of the discourse and brings it back into focus." (87)
a. *esos tres coches aquellos/estos DEM three cars DEM/DEM b. aquellos *(tres/muchisimos) coches escs DEM *(three/many-superlat.) cars DEM
5. CONCLUSIONS In this paper, I have developed an analysis of demonstratives as contextually restricted determiner and generalized quantifier functions. The context set restricting the demonstrative has to satisfy a series of presuppositions, which are partly different from the presuppositions satisfied by defmites. This analysis addresses Kaplan's concern about properly separating context from content and also incorporates the evidence suggesting that demonstratives are not directly referential expressions and have quantificational properties. Contextual restriction has to be viewed as a dynamic process, sensitive to information in previous discourse and to the shifting background and intentions of the conversation participants. A consideration of a variety of data from Spanish showed that there is a great deal of crosslinguistic variation in the contextual presuppositions of demonstratives. Deixis can take several forms, from spatio-temporal varieties to figurative and discourse deixis, confirming Buhler's basic insights on the subject. Languages with a three-term demonstrative system can exploit this increase in expressive power in a variety of ways. In Spanish, one of the terms is neutral or unmarked with respect to proximity presuppositions and can be used to create a variety of contrasts, including focus-related ones.
Demonstratives in Context
233
REFERENCES Almog, J. (ed.). 1989. Themes from Kaplan. Oxford: Oxford University Press. Aliseda, A., R. van Glabbeek and D. Westerstahl. 1998. Computing Natural Language. Stanford, CA: CSLI Publications. Anderson, S. and E. Keenan. 1985. "Deixis." In T. Shopen (ed.) Language Typology and SyntacticDescription. Cambridge: Cambridge University Press, 259-308. Asher, N. and A. Lascarides. 1996. "Bridging." In Rob van der Sandt, R. Blutner and M. Bierwisch (eds.), From Underspecification to Interpretation. Working Papers of the Institute for Logic and Linguistics. IBM Deutschland. Heidelberg. Asher, N. and A. Lascarides. 1998. "Bridging." Journal of Semantics 15. 83-114. Barwise, J. and R. Cooper. 1981. "Generalized Quantifiers and Natural Language." Linguistics &Philosophy 4. 159-219. Barwise, J. and J. Perry. 1983. Situations and Attitudes. Cambridge, Mass.: MIT Press. Beghelli, F. 1995. The Phrase Structure of Quantifier Scope. PhD diss., UCLA. Beghelli, F. 1997. "The Syntax of Distributivity and Pair-List Readings." In A. Szabolcsi (ed), Ways of Scope Taking. Dordrecht: Kluwer Academic Publishers, 349-408. Beghelli, F, D. Ben Shalom and A. Szabolcsi. 1993. "When do Subjects and Objects Exhibit a Branching Reading?" Proceedings WCCFL XII. Beghelli, F., D. Ben Shalom and A. Szabolcsi. 1997. "Variation, Distributivity and the Illusion of Branching." In A Szabolcsi (ed), Ways of Scope Taking. Dordrecht: Kluwer Academic Publishers, 29-70. Bello, A. 1847. Gramdtica de la lengua castellana destinada al uso de los americanos. Ed. by Ramon Trujillo. Tenerife: Aula de Culrura de Tenerife. 1981. Bennett, M. 1978. "Demonstratives and Indexicals in Montague Grammar." Synthese 39. 1-80 Bonzon, P., M. Cavalcanti, and R. Nossum. 2000. Formal Aspects of Context. Dordrecht: Kluwer Academic Publishers. Bosque, I. 1996. "On Specificity and Adjective Position." In J. Gutierrez-Rexach and L. Silva-Villar (eds.), Perspectives on Spanish Linguistics Vol.1. Department of Linguistics, UCLA, 1-13. Bosque, I. 2001. "Adjective Position and the Interpretation of Indefinites." In J. Gutierrez-Rexach and L. Silva Villar (eds.), Current Issues in Spanish Syntax and Semantics, Berlin: De Gruyter. Braun, D. 1994. "Structured Characters and Complex Demonstratives". Philosophical Studies 74, 193-219. Braun, D. 1996. "Demonstratives and their Linguistic Meaning." Nous 30. 145-173. Brucart, J. M. and L. Gracia. 1986. "I Sintagmi Nominali senza Testa." Rivista di Grammatica Generativa 11. 3-32. Biihler, K. 1937. Theory of Language: The Representational Function of Language. Translated by D. Goodwin. Amsterdam: John Benjamins. 1990. Buvac, S. and J. MacCarthy. 1998. "Formalizing Context." In A. Aliseda, R. van Glabbeek and D. Westerstahl. 1998. Computing Natural Language. Stanford, CA: CSLI Publications, 13-50. Carnap, R. 1941. Meaning and Necessity. Chicago: University of Chicago Press. Chomsky, N. 1977. "On Wh-Movement" In P. Culicover, T. Wasow, and A. Akmajian (eds.), Formal SOyntax. New York: Academic Press, 71-132. Chomsky, N. 1981. Lectures on Government and Binding. Dordrecht: Foris. Clark, H. 1978. "Bridging." In P. N. Johnson-Laird and P. C. Wason (eds.), Thinking: Readings in Cognitive Science. Cambridge: Cambridge University Press. Clark, R. andE. Keenan. 1989. "The Absorption Operator and Universal Grammar." The Linguistic Review 5. 113-136. Culicover, P. 1997. Principles and Parameters. An Introduction to Syntactic Theory. Oxford: Oxford University Press. Diesing, M. 1992. Indefinites. Cambridge, Mass: MIT Press. Diessel, H. 1999. Demonstratives. Form, Function and Grammaticalization. Amsterdam/Philadelphia: John Benjamins.
234
From Words to Discourse
En9, M. 1991. "The Semantics of Specificity."Linguistic Inquiry 22. 1-26. Evans, G. 1980. "Pronouns." Linguistic Inquiry 11. 337-362. Fernandez Ramirez, S. 1951. Gramdtica Espanola. Madrid: Revista de Occidente. Fillmore, C. 1975. Santa Cruz Lectures on Deixis. Bloomington: Indiana University Linguistics Club. Fillmore, C. 1997. Lectures on Deixis. Stanford: CSLI Publications. Garrido Medina, J. 1991. Elementos de Andlisis Lingtiistico. Madrid: Fundamentos. Groenendijk, J. and M. Stokhof. 1991 "Dynamic Predicate Logic." Linguistics and Philosophy 14. 39-100. Groenendijk, J., M. Stokhof and F. Veltman. 1996. "Coreference and Contextually Restricted qquantification." Proceedings from Semantics and Linguistics Theory (SALT) V. Ithaca, NY: Department of Linguistics. Cornell University, 112-129. Guha, R. 1991. Contexts. A Formalization and Some Applications. PhD. Diss., Stanford University. Gundel, J., N. Hedberg and R. Zacharski. 1989. "Giveness, Implicature and Demonstrative Expressions in English Discourse." Chicago Linguistic Society 25: Part 2. Parasession on Language in Context, 89-103. Gundel, J., N. Hedberg and R. Zacharski. 1993. "Cognitive Status and the Form of Referring Expressions in Discourse." Language 69. 274-307. Gutierrez-Rexach, J. 1999a. "The Formal Semantics of Clitic Doubling." Journal of Semantics 16. 315-380. Gutierrez-Rexach, J. 1999b. "The Structure and Interpretation of Spanish Neuter Degree Constructions." Lingua 109. 35-63. Gutierrez-Rexach, J. 2000. "Tener y el Efecto de Definitud." Cuadernos de Lingiiistica 1. 19-37. Gutierrez-Rexach, J. 200 la. "The Semantics of Spanish Plural Existential Determiners and the Dynamics of Judgment Types." Probus 13. 113-154. Gutierrez-Rexach, J. 2001b. "Adverbial Weak Pronouns: Derivation and Interpretation." In J. Gutierrez-Rexach and L. Silva. (eds.) Current Studies in Spanish Syntax and Semantics. Berlin: De Gruyter. Gutierrez-Rexach, J. and L. Silva-Villar. 1998. "Locative and Temporal Weak Preforms." In J. Lema and E. Trevirio (eds.), Theoretical Analyses on Romance Languages, Amsterdam: John Benjamins, 205-226. Hawkins, J. 1978. Defmiteness and Indefiniteness. London: Croom Helm. Heim, I. 1982. The Semantics of Definite and Indefinite NPs. PhD diss., UMass., Amherst. Heim, 1.1990. "E-Type Pronouns and Donkey Anaphora." Linguistics and Philosophy 13. 137- 178. Hornstein, N., S. Rosen, and J. Uriagereka. 1995. "Integral Predication." Proceedings ofWCCFFL 14. 169-184. Israel, D. and J. Perry. 1996. "Where Monsters Dwell." In J. Seligman and D. Westerstahl (eds.), Logic. Language and Information. Stanford, CA: CSLI Publications, 303-316. Kamp, H. 1970. "Formal Properties of 'Now'." Theoria 37. 227-274. Kamp, H. 1981. "Theory of Truth and Semantic Representation." In J. Groenendijk, and M. Stokhof (eds.), Truth, Interpretation and Information. Foris: Dordrecht, 1984, 1-41. Kamp, H. 1990. "Prolegomena to a Structural Account of Belief and Other Attitudes." In A. Anderson and J. Owens (eds.), Prepositional Attitudes: the Role of Content in Logic, Language, and Mind. Stanford, CA: CSLI Publications, 27-90. Kamp, H. and U. Reyle.1993. From Discourse to Logic. Dordrecht: Kluwer. Kaplan, D. 1978. "Dthat." In P.Cole (ed.) Syntax and Semantics , 221-243. Kaplan, D. 1979. "On the Logic of Demonstratives." Journal of Philosophical Logic 8. 81-98. Kaplan, D. 1989. "Demonstratives. An Essay on the Semantics, Logic, Metaphysics and Epistemology of Demonstratives and Other Indexicals." In J. Almog et al. (eds.) Themes from Kaplan. New York: Oxford University Press, 481-564. Keenan, E. and J. Stavi.1986. "A Semantic Characterization of Natural Language Determiners." Linguistics and Philosophy 9. 253-326.
Demonstratives in Context
235
Keenan, E. 1987. "A Semantic Definition of Indefinite NPs." In E. Reuland and A. ter Meulen (eds.) The Representation of (In)defmiteness. Cambridge, Mass.:MIT Press, 286-317. Keenan, E.I 996. "The Semantics of Determiners." In S. Lappin (ed.) Handbook of Contemporary Semantic Theory. Oxford: Basil Blackwell, 41-64. King, J. 1999. "Are Complex That' Phrases devices of Direct Reference." Nous 33. 155-182. King, J. 2001. Complex Demonstratives. A Quantificational Account. Cambridge, Mass: MIT Press. Kratzer, A. 1981. "The Notional Category of Modality." In H.J. Eikmeyer and H. Rieser (eds.) Words, Worlds and Context. Berlin: Walter De Gruyter, 38-74. Kripke, S. 1980. Naming and Necessity. Oxford: Blackwell. Landman, F. 1986. Towards a Theory of Information. Foris: Dordrecht. Leonetti, M. 1996. "Determinantes y Contenido Descriptive." Espanol Actual 66. 5-23. Lewis, D. 1972. "General Semantics." In D. Davidson and G. Harman (eds.), Semantics of Natural Language. Dordrecht: Reidel, 169-218. Maclaran, R. 1982. The Semantics and Pragmatics of the English Demonstratives. PhD dissertation, Cornell University. MacCarthy, J. 1993. "Notes on Formalizing Context." In Proceedings of the 1 Othlnternational Joint Conference on Artificial Intelligence. Macia, J. 2002. "On the Interaction of Syntax-Semantics-Pragmatics: A Case Study." This volume. May, R. 1985. Logical Form: Its Structure and Derivation, Cambridge, Mass., MIT Press. May, R.1989. "Interpreting Logical Form." Linguistics and Philosophy 12. 387-435. Mitchell, J. 1986. The Formal Semantics of Point of View. PhD diss., UMass., Amherst. Montague, R. 1968. "Pragmatics." In R. Thomason (ed.), Formal Philosophy. New Haven: Yale University Press. Neale, S. 1993. Descriptions. Cambridge, Mass.: MIT Press. Ostertang, G. 1998. Definite Descriptions. Cambridge, Mass- MIT Press. Partee, B. 1989. "Binding Implicit Variables in Quantified Contexts." Proceedings ofCLS 25. Partee,B. 1991. "Topic, Focus and Quantification." Proceedings Semantics and Linguistic Theory (SALT) 1. 159-188. Partee, B. 1995. "Quantificational Structures and Compositionality." In E.Bach et al. (eds.), Quantification in Natural Languages. Dordrecht: Kluwer, 541-602. Peirce, C. S. 1955. Philosophical Writings ofPeirce. J. Buchler (ed.). New York: Dover. Perry, J. 1997. The Problem of the Essential Indexical. Oxford: Oxford University Press. Postma, G. 1996. "The Nature of Quantification of High Degree - Very, Many and the Exclamative." In C. Cremers and M. den Dikken (eds.) Linguistics in the Netherlands 1996. Amsterdam: John Benjamins, 207-220. Recanati, F. 1993. Direct Reference. Oxford: Blackwell. Roberts, C. 1999a. "Uniqueness Presuppositions in Definite Noun Phrase." Ms. Ohio State University. Roberts, C. 1999b."Demonstratives as Definites." Ms. Ohio State University. Russell, B. 1905. "On Denoting." Mind 14. 479-493. (Reprinted in Logic and Knowledge. London: Allen and Unwin, 39-56.) Schlenker, P. 1999. "A Plea for Monsters." Ms. UCLA Schlenker, P. 2000. Propositional Attitudes andIndexicality. A Cross-Categorial Approach. Ph.D. dissertation. MIT. Scott, D. 1970. "Advice on Modal Logic." In K. Lambert (ed) Philosophical Problems in Logic. Dordrecht: Reidel, 143173. Segerberg, K. 1973. "Two-Dimensional Modal Logic." Journal of Philosophical Logic 2. 77-96. Silva-Villar, L., and J. Gutierrez-Rexach. 2001. "Demonstratives in a Feature Based Theory of Syntax." In G. M. Alexandrova and O. Arnaudova (eds.) The Minimalist Parameter. John Benjamins, Amsterdam/Philadelphia. 325-344. Stalnaker, R. 1978. "Assertion." In P. Cole (ed.) Syntax and Semantics 9. 315-332. Stalnaker, R. 1998. Context and Content. Oxford: Oxford University Press.
236
From Words to Discourse
Strawson, P. 1950. "On Referring." In Logico-Linguistic Papers. London: Methuen, 1971, 1-27. Van Benthem, J. 1998. "Changing Contexts and Shifting Assertions." In A. Aliseda, R. van Glabbeek and D. Westerstahl (eds.). 1998. Computing Natural Language. Stanford, CA: CSLI Publications, 51-66. Van der Does, J. 1995. "Quantifiers, Contexts and Anaphora." In Paul Dekker and M. Stokhof (eds.),Proceedings of the Iff1' Amsterdam Colloquium. ILLC. University of Amsterdam, 183-202. Veltman, F. 1996. "Defaults in Update Semantics." Journal of Philosophical Logic 25. 221-261. Vincent, N. 1997. "The emergence of the D-system in Romance." In A. van Kemenade and N. Vincent (eds.),Parameters of Morphosyntactic Change. Cambridge: Cambridge University Press, 149-169. Von Fintel, K. 1994. Restrictions on Quantifier Domains, PhD diss. UMass., Amherst. Von Stechow, A. 2001. "Discussion about Monsters." Ms. Universitat Tubingen. Westerstahl, D. 1985. "Determiners and Context Sets." In J. van Benthem, and A. ter Meulen (eds.), Generalized Quantifiers in Natural Language. Foris: Dordrecht, 45-71. Westerstahl, D. 1989. "Quantifiers in Formal and Natural Languages." In D. Gabbay and F. Guenthner (eds.), Handbook of Philosophical Logic, Vol. II (Extensions of Classical Logic). Dordrecht: Reidel, 1-131. Zeevat, H. 1999. "Demonstratives in Discourse." Journal of Semantics 16. 279-314. Zimmermann, T. E. 1991. "Kontextabhangngigkeit." In A. von Stechow and D. Wunderlich (eds.), Semantik/Semantics. Berlin: Springer Verlag, 156-229.
237
12
PROPOSITIONAL NPIs AND THE SCALAR NATURE OF POLARITY Javier Gutierrez-Rexach and Scott A. Schwenter, The Ohio State University
1. INTRODUCTION The study of polarity sensitivity has been one of the topics that has sparked greater interest in semantics/pragmatics in recent decades, its popularity due in large part to Klima's (1964) groundbreaking study of negation in English. The current state of the art on this topic most likely reflects the inherent complexity of the issues involved, with a variety of competing theories in different domains of linguistic research (syntactic, semantic, pragmatic, or some combination thereof). Another interesting feature of recent contributions to the topic is the increasing attention being paid to cross-linguistic variation and to polarity phenomena in different languages such as Greek, Italian, Spanish, Catalan, Dutch, etc. In this paper, we contribute our own perspective to the ongoing debates, by arguing that there is a class of negative polarity items (NPIs) which have prepositional content and which present new challenges for a general theory of NPI meaning and interpretation. Such a theory, we argue, should incorporate attention to scales of different types, including scales of expectations and those related to speaker's attitudes more generally. In this respect, our contribution can be ascribed to a longstanding tradition in the study of NPIs based on scales and ordered entailments, along the lines initially proposed by Horn (1972) and Fauconnier (1975a, 1975b). More interestingly, it also provides arguments in favor of a more fine-grained understanding of the ingredients of polarity licensing, an understanding in which structural factors combine with or are superseded by semantic and pragmatic constraints. We study in detail the semantics and pragmatics of the Spanish NPI que
238
From Words to Discourse
digamos (lit. "that we (might) say [SUBJ]") to provide further evidence for our claims. The rest of this paper is structured as follows: in the second and third sections, we summarize some recent contributions to the analysis of negative polarity licensing and polarity sensitivity; in section four, we analyze prepositional NPIs in Spanish; and in the sections that follow, we study the specific case of the prepositional NPI que digamos in greater detail.
2. CHARACTERIZING THE TRIGGERS 2.1. Negation-Based Theories A polarity sensitive expression is one that can only occur in an environment where a certain polar element also occurs (or, in the case of positive polarity, where a certain polar element is necessarily absent). This expression is called the trigger or licensor. Negative polarity items such as any or yet were initially viewed as expressions that have to appear "in construction with" negation (Klima, 1964). The absence of the negative trigger in (Ib, 2b) results in the ungrammaticality of these sentences. On the other hand, a positive polarity item such as already is incompatible with negation (3b). (1) (2) (3)
a. b. a. b. a. b.
I did not buy any books, *I bought any books. He has not arrived yet. *He has arrived yet. He has already arrived. *He has not already arrived.
Determining which condition(s) have to be satisfied in order for the NPI to be "in construction" with the trigger is one of the central points of the theoretical discussion on polarity items. This line of research, which strives to answer what Ladusaw calls "The Licensing Relation Question" (1996: 326), attempts to characterize the defining property or properties of the trigger. As a result of this perspective, NPIs are characterized indirectly, namely as the elements licensed by the relevant property of the trigger. A first approach in this direction would be to understand the licensing requirement as a syntactic c-command requirement: an NPI has to be syntactically c-commanded by negation. Baker (1970) proposes a slightly different account, more semantic in nature, when he claims that NPIs have to be "within the scope" of negation. Negation need not be actually present in the sentence where the NPI occurs, but a negative statement may be just simply entailed by it. A similar proposal was later made by Linebarger (1980, 1987), who proposed that NPIs are licensed either in the immediate scope of negation in the Logical Form (LF) of a sentence S or when there is a proposition entailed or implicated by an
Prepositional NPIs and the Scalar Nature of Polarity
239
S in whose LF the NPI occurs in the immediate scope of negation. Linebarger's Immediate Scope Constraint (ISC) also requires that no operator intervene between negation and the NPI in the LF of the sentence where the NPI occurs or, if there is no negation in this sentence, the ISC applies to an entailed/implicated sentence where negation actually occurs. Linebarger's account therefore treats NPI licensing as belonging to the syntax/pragmatics interface. On the one hand, the ISC explains the difference in grammaticality between the sentences in (4): (4)
a. He didn't budge an inch, because he was just a wooden statue, b. He didn't budge an inch because Mary pushed him.
The LF corresponding to (4a) is (5a), and the LF corresponding to (4b) is (5b). In (5a), the NPI budge an inch is in the immediate scope of negation, since the CAUSE operator does not intervene between negation and this NPI. Contrastingly, in the LF representation (5b) the CAUSE operator intervenes between negation and the NPI, thus the ISC is violated and the LF is ill-formed. (5)
a. CAUSE( (NOT( he budged an inch )), he was just a wooden statue ) b. NOT( CAUSE( he budged an inch, Mary pushed him ) )
The pragmatic component of Linebarger's theory, which allows indirect licensing when the ISC is satisfied in a sentence which licenses a negative implicature (more accurately, a negative proposition which is derivable via conversational implicature), is claimed to predict why NPIs may occur in non-negative contexts such as in the scope of adversative predicates like surprise, regret and doubt: (6)
I'm surprised that she contributed a red cent to the ACLU.
The above sentence licenses the negative implicature in (7). In this implicated sentence, the NPI occurs in the immediate scope of negation and the ISC is satisfied: (7)
I expected that she wouldn't contribute a red cent to the ACLU.
Although the negative implicature strategy can be extended to other cases, there exists no uniform procedure to uncover the relevant entailment/implicature or to determine its nature. Sometimes it can be a genuine implicature as in (7) above. In other cases, the sentence satisfying the ISC is a semantically equivalent sentence, as in the case of conditionals and comparatives. For example, (8a,b) are equivalent to (9a,b), respectively.1 The problem then becomes In the case of the conditional (8b), it could be argued that it also has a negative implicature namely (i):
240
From Words to Discourse
determining which sentence is the one that satisfies the ISC: the equivalent sentences (9), the implicatures, or both, and whether there is any procedure to decide between the different alternatives. (8) (9)
a. b. a. b.
I loved her more than any other person, If he sees any fierce dog, he screams. I did not love any other person as much as her. Either he does not see any fierce dog or he screams.
The equivalent sentences in (9) obviously satisfy the ISC. This oscillation with respect to the determination of the proper sentence satisfying the ISC is found again in the case of interrogatives. NPIs are licensed both in yes/no questions and in w/z-questions. Although these two types of questions apparently have a uniform illocutionary and syntactic type, the associated sentences satisfying the ISC are generated by different means. The sentences associated with (10a,b) are (1 la,b) respectively. (10) (11)
a. b. a. b.
Do you have any money? Who has any time for volunteer work? It is possible that you don't have any money, Nobody has any time for volunteer work.
There are several properties that distinguish the sentences in (10) and (11) from each other. First, (11 a) can be taken to be an entailment of (lOa) if one adopts a Hamblin-based semantics of questions where the denotation of a question is the set of its possible answers (Hamblin, 1973). Thus, the denotation of (11 a) would be the set { you have (any) money, you don't have (any) money }, and (lOa) would be entailed by that set. With respect to the interrogative (lOb), the associated sentence containing a negative expression is an implicature (lib). Furthermore, the existence of this implicature is what makes this question rhetorical (a question that presupposes or implicates that nobody has time for volunteer work). There is no overt negation in (lib), although there is a "negative" quantificational expression. In sum, a theory of NPI licensing based on the Baker/Linebarger extension of the original claim by Klima (1964) faces several challenges, since it may overgenerate and it does not provide a uniform way of determining indirect licensing.
(i) If he doesn't see any fierce dog, he doesn't scream.
Prepositional NPIs and the Scalar Nature of Polarity
241
2.2. Monotonicity and Downward Entailingness Ladusaw (1979) proposed a theory of NPI licensing which is based on a semantic property of the trigger. This property is downward monotonicity. We say that a function / is downward monotonic or decreasing iff for any arguments A, B, if A < B then^(B)
a. b. a. b.
John ate bananas, John ate. John didn't eat. John didn't eat bananas.
The interest of Ladusaw's theory is that it proposes that an NPI will be licensed in the scope of any downward entailing function. This applies to many other expressions besides sentential negation which can also be taken to denote decreasing functions. For example, the generalized quantifiers denoted by the DPs no employee and/ew employees are decreasing. The set denoted by the VP ate bananas is a subset of the set denoted by ate. Decreasing generalized quantifiers license set to subset inferences, so sentence (14a) entails sentence (14b). (14)
a. No employee/few employees ate. b. No employee/few employees ate bananas.
Ladusaw's theory successfully predicts that generalized quantifiers license NPIs in their scope (i.e. in their VP argument). In the sentences in (15), the NPIs any (raise} and banana can appear in the scope of the generalized quantifiers. (15)
a. No employee got any raise. b. Few employees ate any banana.
Similarly, the first argument of the determiner function denoted by every is a decreasing 2
Technically, the domains in which order is reversed by a decreasing function are not the same. For an arbitrary function/from an ordered domain D to an ordered domain D', we say that/is decreasing iff for every arbitrary elements A, B of D such that A
242
From Words to Discourse
environment and licenses NPIs, as well as comparatives (Hoeksema, 1984): (16)
a. Every student who drank any wine got a headache, b. Two students drank more wine than anybody else.
The theory of licensing based on monotonicity properties has undergone several refinements based mainly on the degree of sensitivity that certain NPIs show (Zwarts, 1993, 1995; Nam, 1994). Some NPIs can be classified as weak, and are licensed by decreasing functions. Others require that the trigger satisfy the stronger property of anti-additivity, and finally others require that the trigger be antimorphic. The strength of the NPI correlates with the "strength" of the monotonicity properties satisfied by the trigger: the stronger the NPI is the more stringent these properties become. Another area that needed refinement was how to explain the variety of environments in which NPIs can appear. This was a problem for the theory of direct or indirect licensing based on negation and it remains a problem for any theory based on the property of monotonicity. For example, given that NPIs are licensed in the protasis of conditional constructions it could be argued that the restriction of a conditional is a decreasing environment. Consider sentence (18) below. It could be argued that the protasis of this conditional is a decreasing environment because (17a) seems to entail sentence (17b). What distinguishes (17a) from (17b) is that the VP go on vacation to the Bahamas entails go on vacation. It could be argued that conditionals reverse this implication order and are hence decreasing, predicting the licensing of NPIs. (17) (18)
a. If you go on vacation, you'll have a good time. b. If you go on vacation to the Bahamas, you'll have a good time. If you ever go on a vacation, you'll have a good time.
Nevertheless, as noticed by Heim (1984), it is not clear that conditionals are monotonedecreasing in a clear-cut sense as in the constructions considered before. Consider sentence (19), which according to the decreasingness hypothesis applied to conditionals should be entailed by (17a). Nevertheless, given our encyclopedic knowledge of the world the conditional would most likely be false: (19)
If you go on vacation to Bosnia, you'll have a good time.
A similar example of non-entailment from (17a) would be (20). The truth of (17a) does not guarantee the truth of (20). The conjunction of the propositions you go on vacation and you have an accident unilaterally entails the proposition you go on vacation, so if the protasis of a conditional were a decreasing environment, (20) should entail (17a), contrary, once again, to our real-world expectations.
Prepositional NPh and the Scalar Nature of Polarity
(20)
243
If you go on vacation and have an accident, you'll have a good time.
What seems to be at play here is the fact that conditionals require the consideration of certain normality criteria (or, in general, modal criteria) that establish the relevant relationship between antecedent and consequent. One possible solution would be to claim that conditionals are not decreasing environments. Another solution would be to require that the worlds or situations under consideration be properly restricted contextually so that undesired entailments are banned but those that seem adequate are allowed. Questions are another potential but controversial environment of NPI licensing. As we stated above, NPIs are licensed in yes/no and w/!-questions. Giannakidou (1999) claims that "in a semantics of interrogatives along the lines of Groenendijk and Stokhof (1984, 1997) interrogatives can be at most non monotonic". Nevertheless, if the trigger of the licensing of NPIs in interrogative sentences is considered to be the w/z-phrase and this expression is treated as a generalized quantifier, a strong parallel with the licensing of NPIs by standard declarative generalized quantifiers such as no student or few students can be drawn. In Gutierrez-Rexach (1997a, 1997b) it is claimed that interrogative generalized quantifiers and determiners are decreasing functions and hence license NPIs.3 (21)
a. Which student has ever been to Moscow? b. Which kid who has ever been to Disneyland forgets it?
A detailed consideration of conditionals and questions seems to reveal that the jury is still out with respect to the issue of whether the monotonicity properties of the trigger, more specifically downward monotonicity, represent the unifying property that triggers must satisfy in order to license NPIs in their scope. Depending on how we view the semantics of these constructions (and potentially others such as intensional and modal verbs, imperatives, generics, etc.), we may or may not be able to claim that they satisfy the decreasingness property.
2.3. Non-Veridicality In more recent investigations, Zwarts (1995) and Giannakidou (1998, 1999) have defended the idea that the unifying property of the triggers of NPI licensing is non-veridicality, a term first 3 The critical property here is the characterization of question entailment (see Groenendijk and Stokhof 1989 and Gutierrez-Rexach 1997a) We say that question (ia) entails question (ib) because the answer set of (ib) is a subset of the answer set of (ia). In other words, an answer to (ia) includes an answer to question (ib). In this respect, interrogation is decreasing, (i) a. Who drank? b. Who drank a beer?
244
From Words to Discourse
characterized by Montague (1969). We say that a prepositional operator is veridical in case Op(p) entails p. If this entailment does not hold, we say that the operator is non-veridical. Perception verbs are clear examples of veridical operators. For example, I see p entails p. On the other hand, verbs such as believe or want are non-veridical, since / believe/want p does not entail p. Giannakidou (1999) claims that all the environments which license polarity sensitive expressions, i.e. those environments which create affective dependencies in general, including questions, the protases of conditionals, the subjunctive mood in Greek and Romance languages, etc., share the property of being non-veridical. This non-veridicality hypothesis seems to cover a wider range of cases than the downward monotonicity hypothesis, but ultimately it faces similar problems. On the one hand, there are some environments that are veridical and seem to license weak NPIs. For example, the verb be lucky is veridical since it presupposes its complement: / was lucky that p entails p. Nevertheless, in (22a) the NPI any talk is licensed in the scope of was lucky. Emotive factive verbs are in general veridical (Lee, 1999). The verb be surprised is likewise veridical (lam surprised thatp entails p) but it also licenses weak NPIs, as in (22b). (22)
a. I was lucky that I didn't miss any talk. b. I am surprised that there are any tickets available.
On the other hand, in order to make the notion of non-veridicality fully comprehensive, it would have to be characterized for each particular construction, with the result that these characterizations might not be subsumable under a blanket notion of non-veridicality. The distinction between direct and indirect licensing permits covering even those cases that apparently violate the non-veridicality condition. For instance, with respect to emotive factives it could be stated that they satisfy the non-veridicality requirement indirectly, since it is the complement that communicates a non-veridical or even negative (a-veridical) implicature.
3. THE CHARACTERIZATION OF THE SENSITIVITY OF NPIS In the previous section, we examined how different theories have tried to tackle the problem of polarity licensing by analyzing various properties of triggers and triggering environments. An equally important problem is the analysis of the contribution of the NPI itself, in other words what makes polarity sensitive expressions sensitive —Ladusaw's "Licensee Marking Question" (1996:326). Besides the obvious theoretical repercussions for the theory of polarity sensitivity, there is an important empirical aspect to solving this problem, namely, figuring out how the differential behavior of NPIs helps us to understand the semantic and pragmatic role of NPIs. More concretely, we are trying to determine what the contribution of a polarity sensitive expression is to the meaning of a sentence. Among others, Fauconnier (1975a, 1975b, 1978), Krifka (1991,1995), Kadmon and Ladman (1993), Lee and Horn (1995), and Israel (1996, 1998)
Prepositional NPIs and the Scalar Nature of Polarity
245
have made important contributions in this direction. The pioneering work of Fauconnier (1975a,b) brought to the fore the role of certain expressions as activators of pragmatic scales. He argues that the function of even is to signal the existence of a pragmatic scale with respect to the context of utterance. The scale is constructed on the basis of a prepositional schema P(x). The Scale Principle states that if x/ is lower than X2 on the scale, then P(xi) implicates P(x2). It follows that if P holds for the lowest element on the scale, it holds by conversational implicature for all elements on the scale. Thus, if a is the lowest element on a scale then P(a) implicates 'for all x, P(x)''. Consider sentence (23): (23)
Even the faintest noise bothers my brother.
A scale is constructed over the domain of noise which is associated with the propositional schema "n bothers my brother". The lowest end of the scale is the faintest noise. Since the propositional schema holds for this lowest element, the sentence implicates that 'for all x such that x is a noise x bothers my brother'. Krifka (1991, 1995) builds on Fauconnier's ideas to craft a very sophisticated semantic and pragmatic account which incorporates elements of scalar reasoning. Polarity items introduce alternatives that lead to an informativity relation with respect to the meanings of the polarity items and the common ground in which they are used. Illocutionary operators make critical use of this information. Krifka's (1991) theory posits a combination of semantic and pragmatic factors for the licensing of NPIs in different constructions. On the semantic side, he presents a lattice-theoretical approach to NPIs. He introduces the concept of a polarity lattice defined as a triple LA = , where A' is the NPI representation, LA is the lattice sort and the following conditions hold: (a) if A' is of type a, LA is of type ; (b) wine'(x) A x is smaller than some quantity e]). Let us consider for example the occurrence of NPIs in questions. The process of question formation consists in attaching the illocutionary operator ERO to the sentence radical. If p' is a proposition, i a world, s the speaker, and h the hearer, then ERO(s,h,i,p') says that 5 asks h whether p' (i) is true. In the case of rhetorical questions, Krifka claims that the speaker wants to show that he is sure to get a negative answer, so he follows the rule: If ERO(s,h,i,A') and A' is an NPI or PPI representation with lattice sort LA, then for any X e LA, s has reasons for -iERO(s,h,i,X). Consider the following interrogative sentence:
246
(24)
From Words to Discourse
Did you ever drink a drop of wine?
According to the rule described above, if a speaker s asks question (24), then for any property X in the polarity lattice La.drop.of.wine, s has reasons for not asking whether a proposition containing X is true at /. The speaker asks a question only about the least element in the lattice, namely a.drop.of.wine'. It is not clear whether the above rule captures the essence of what is a rhetorical question. The speaker is not even asking whether the proposition p containing the least element in the relevant polarity lattice is true. He already knows what the answer is and he is asking it for reasons different from knowing whether/? is true in i. The question is uttered not as a request for information, but with another communicative purpose. In that respect, it seems reasonable to claim that in uttering a rhetorical question, for every X e LA the speaker s has reasons for -^ERO(s,h,i,X). Therefore, rhetorical questions are not properly questions from an illocutionary point of view. In general, a weak NPI such as anything has two semantic properties: first, it introduces alternatives and, second the alternatives induce an ordering relation of semantic specificity, where the NPI itself denotes the most specific element in that order. Strong NPIs, such as anything at all also introduce alternatives but they are non-exhaustive and apply to minor objects to which none of the alternatives can be applied. Consequently, strong NPIs occur in emphatic assertions, which carry an "extreme" or "emphatic" meaning. Consider (25): (25)
a. Mary didn't get anything for her birthday. b. Mary didn't get anything at all for her birthday.
The difference between the above sentences is that (25b) stresses the fact that Mary didn't even get some minor present for her birthday, whereas (25a) just says that Mary got nothing. While the two versions in (25) could be used to describe the same situation in the world, they do clearly link to distinct rhetorical purposes in discourse, e.g. to convey different degrees of sorrow for poor Mary. Kadmon and Landman (1993) and Lee and Horn (1995) focus on the analysis of the weak NPI any. Both proposals amount to the claim that any N is equivalent to the indefinite a N with some extra semantic and pragmatic properties. For Kadmon and Landman those properties are "widening" and "strengthening". In an NP of the form any N, any widens the interpretation of the noun along a contextual dimension. Furthermore, any is licensed only if the widening it induces creates a stronger statement: the statement on the wide interpretation unilaterally entails the statement on the narrow interpretation. Lee and Horn's proposal that any is semantically equivalent to even plus the determiner a captures the properties that relate existentials and any on the one hand, and also the presence of a quantity/kind scale, as required by the presupposition ofeven.
Prepositional NPIs and the Scalar Nature of Polarity
247
Finally, Israel (1996, 1998) posits that polarity sensitive expressions are dependent on two cross-cutting semantic features: quantitative value (q-value) and informative value (i-value). Both values refer to the position of a polarity item within a different scalar ordering. The q-value of a polarity sensitive expression may be high or low, and its i-value may be emphatic or understating. Expressions such as drink a drop, budge an inch, lift a finger, etc. have a low qvalue (they denote scalar endpoints, i.e. minimal quantities) and an emphatic i-value, since they participate in emphatic assertions. The more reduced class of NPIs that have a high q-value and an understating i-value is formed by adverbs such as long in (26a), all that in (26b). (26)
a. He didn't last long, b. He is not all that rich.
To conclude this section, we may summarize the results of recent research on the issue of sensitivity with the claim that different NPIs seem to give raise to different polarity lattices (semantic-pragmatic scales). The occurrence of an NPI is associated with the relevant alternatives in a comparative order and the relative position of the denotation of this expression with respect to the other alternatives is implicated. This type of theory can actually be integrated with a characterization of triggers based on monotonicity, since decreasingness entails reasoning about the properties of an ordering. Still, an important empirical issue that emerges from this provisional conclusions is how far natural language goes in allowing the emergence of semantic/pragmatic scales and the linguistic means used for it. In what follows, we will explore a new class of NPIs that we will call prepositional NPIs and investigate how they are used to evoke speaker-related scales.
4. PROPOSITIONAL NPIS IN SPANISH The majority of Spanish NPIs belong to the class of n-words (Bosque, 1980; Laka, 1990; Vallduvi, 1994). Items such as nada "nothing", nadie "nobody", ninguno "none", nunca "never", etc. are categorially quantifier phrases of nominal or adverbial nature (i.e. they quantify over individuals or times/events). Others, such as ningun N "no N", are determiners heading negative quantifier phrases. All of them are single-world expressions which encode a negative feature and may be subject to syntactic matching requirements.4
4
Following the ideas of Bosque (1994) and Longobardi (1991), NPIs raise at LF to the specifier of the Negative Phrase headed by the negative particle. This can be viewed as a feature matching criterion, the "Neg criterion" of Haegeman and Zanuttini (1991). We do not enter here in the debate on whether n-words are NPIs or negative quantifiers, mainly because our main goal is to characterize a class of expressions that do not belong to the n-word paradigm and because we are exploring the boundaries of polarity sensitivity., Thus, this particular debate is not pertinent to the purposes of this paper.
248
(27)
From Words to Discourse
a. Juan no ha dicho nada. J. NEG has said nothing "Juan hasn't said anything." b. No le he dado un NEG to-him have given a "I haven't given a book to anybody.'" c. No ha venido nunca. NEG has come never "He has never come."
libra a book to
nadie. nobody
There is, nevertheless, a small class of NPIs, which has not received much attention in the literature, characterized by the property of having a full propositional form. We will call this class of NPIs "propositional NPIs". The members of this class are mostly parenthetical expressions headed by the negative particle ni "not even" and containing a verb in the infinitival form (ni sonarlo, ni hablar de ello, etc.) or in its finite form (ni lo pienses, ni lo suenes). They uniformly result in ungrammatically in the absence of negation: (28)
a. Ni pensarlo, Forbes *(no) llegard a presidente. not-even think:it F. NEG will.arrive to president "Don't even think about it, Forbes won't become President." b. Ni lo suenes, *(no) te voy a prestar mi nuevo Lexus. not-even it dream NEG to-you go to lend my new Lexus "Don't even dream about it, I'm not going to lend you my new Lexus."
Because of their sensitivity to negation, these expressions should not be confused with other non-parenthetical forms headed by ni. These latter expressions are not NPIs and structurally they occupy positions in the higher periphery of the clause. In addition, they also have full embedding capacities: (29)
a. Ni sonar con (no) venir hoy. not-even dream with (NEG) come today "Don't even dream of (not) coming today." b. Ni pensar en faltar a close. not-even think in miss to class "Don't even think in missing class."
The primary semantic contribution of propositional NPIs such as these to the content of the sentence in which they occur is to strengthen the associated denial. Compare the reply to (30A) in (30B1) with that in (30B2). (30B1) simply states that A's previous assertion is false: it carries
Prepositional NPIs and the Scalar Nature of Polarity
249
out a straight denial. The denial in (30B2) is stronger. It carries additional entailments and/or implicatures related to the speaker's evaluation of A's assertion. (30)
A: Creo que Forbes llegard muy lejos en su carrera politico.. think that F. will.arrive very far inhis career political "I think that Forbes will go very far in his political career." Bl: No, Forbes no llegard a presidente. NEG F. NEG will-arrive to president "No, Forbes won't become President." B2:(M pensarlo), Forbes no llegard a presidente. not-even think:it F. NEG will.arrive to president "Don't even think about it, Forbes won't become President."
Finally, the members of this class of NPIs have only partial productivity. Not all sequences of ni followed by a verb (of the speech act or propositional attitude class) are permissible, as shown in (31) below. This suggests that propositional NPIs are idiomatic expressions. (31)
*Ni conceb.irlo/expresarlo/ decirlo/entenderlo, etc. "not-even conceive:it/express:it/say:it/understand:it, etc."
An important issue that arises when considering these propositional NPIs is how to separate these idiomatic forms from other well-known polarity idioms such as (ni) un pelo de tonto, (ni) un centimo, etc. As noted by Vallduvi (1994), these latter expressions are minimizers: they denote low points in a scale, i.e. their q-value is low in Israel's (1996, 1998) terms. (32)
a. No tiene (ni) un pelo de tonto. b. For eso no pagaria (ni) un centimo.
Interestingly, the presence of the negative head ni is optional in the case of non-propositional idioms like those in (32). Its absence does not change the minimizing value (q-value) of the NPI. As we have seen, in the case of propositional NPIs, the presence of ni is not optional. The most salient scalar value of the propositional minimizer is not a q-value but rather an i-value of emphasis. Propositional NPIs headed by ni participate in emphatic refutations as in (30) above. In this respect, propositional NPIs are polyfunctional. They can function both as NPIs and as free-standing negative expressions indicating refutation of a contextually-salient proposition. In the previously considered examples, the propositional negative element is a true NPI. Nevertheless, in (33) it is not, since it does not appear in the scope of negation.
250
From Words to Discourse
(33)
A: iMe podrias dejar tu nuevo Lexus para correr el Paris-Dakar! me could let your new L. to run the P.-D. "Could you lend me your new Lexus to run the Paris-Dakar race?" B: jM lo suenes\ not-even it dream "Don't even dream of it!"
(33B) is a strong or emphatic proposition: it refutes the inference by A that B considers lending his Lexus as a possibility. In other cases, the proposition refuted may be entailed by previous discourse: (34)
Ya se que estds pensando en mi Lexus, pero... already know that are thinking in my L. . but \ni lo suenes\ not-even it dream "I already know that you're thinking about my Lexus, but... don't even dream of it!"
Other non-propositional NPIs, such as en absolute "at all" or lo mas minima, "in the least" also have an emphatic i-value that strengthens an assertion, as in (35): (35)
a. No me gusta en absoluto. "I don't like it at all." b. No me gusta lo mas minima. "I don't like it in the least."
These types of items cannot combine with prepositional NPIs, as shown in (36). An explanation of this fact might be that, given the role of these items as conveyors of emphasis (high i-value), it would be redundant to have more than one NPI playing this particular role. Indeed, (36) shows precisely this point: the combination of two non-propositional NPIs with high i-value is also impossible. (36)
(37)
a. *M lo pienses en absoluto. "Don't even think about it at all." b. *Ni lo pienses lo mas minimo. "Don't even think about it in the least." *No me gusta lo mas minimo en absoluto. "I don't like it in the least at all."
The main difference between minimizing idioms and prepositional NPIs is that prepositional
Prepositional NPIs and the Scalar Nature of Polarity
251
NPIs are associated with or evoke epistemic scales related to the speaker's attitude towards the prepositional content of the utterance. As expressions with an emphatic i-value and a low qvalue, the implicature arises that the likelihood that the speaker will accept the proposition under consideration is very low. Rejections/refutations and strong rejections/refutations constitute the lowest point on this scale (the likelihood of acceptance is null). In sum, then, the prepositional NPIs seen so far are minimizers along prepositional (epistemic) scales which participate in strong/emphatic refutations.
5. THE PLACE OF QUEDIGAMOS We claim that there is another prepositional NPI in Spanish, que digamos "that we-say-SUBJ.", which exhibits clearly differential behavior with respect to the other prepositional NPIs considered in the previous section. First, the proposition in question is fully lexicalized as an idiomatic NPI: there is no productivity in the sense that verbs other than decir (not even a limited set of such verbs) can appear in the construction. Compare the idiomatic nature of this NPI to productive constructions which can take relative clauses in the subjunctive, where there is no restriction on the lexical verb: (todo lo) que digas/quieras/comas/supongas "(everything) that you say/want/eat/suppose". In addition, the verb decir must appear in the first person plural present subjunctive form, as the contrast between (38a) and (38b) below shows. And, as (38c) illustrates, the construction is completely lexicalized, insofar as nothing, not even the pronoun nosotros "we", can be interpolated between the complementizer and the verb. Second, que digamos cannot occur as a free-standing negative element (39a) or as a preposed parenthetical (39b):5 (38)
(39)
a. Tu Lexus no es muy barato que digamos. your L. NEG is very cheap that we.say "Your Lexus isn't very cheap que digamos" b. Tu Lexus no es muy barato que *digas/*digdis/*digan/*dijeramos. c. Tu Lexus no es muy barato que (*nosotros) digamos. a. *\Que digamos\ b. *Que digamos, tu Lexus no es muy barato.
From a semantic point of view, que digamos can be classified as a strong NPI (Zwarts, 1995; Giannakidou, 1998). It is licensed by negation (40a), n-words (40b), and by negative particles, 5 Note, as well, that the pronoun cannot appear postposed (*que digamos nosotros), and thus cannot appear at all with the NPI. While there are some fixed expressions in Spanish with "frozen" postposed pronouns (e.g. epistemic/evidential adverbial digoyo lit. 'say I'), these latter forms are not NPIs.
252
From Words to Discourse
such as tampoco (40c) (see also section nine below for more on tampoco). It is not licensed by merely decreasing operators such as the generalized quantifiers headed by the determiners pocos, or menos de n, as the contrast between (41a) and (41b) illustrates. (40)
(41)
a. No es muy listo que digamos. NEG is very smart that we.say "He's not very smart que digamos." b. Nadie ha comido mucho que digamos. Nobody has eaten much that we.say "Nobody has eaten much que digamos." c. Tampoco ha comido tanto que digamos. neither has eaten so-much that we.say "He hasn't eaten so much que digamos." a. *Pocos han comido tanto que digamos. few have eaten so-much that we.say "Few have eaten so much que digamos." b. *Menos de ires estudiantes son muy listos que less of three students are very smart that "Fewer than three students are very smart que digamos."
digamos. we.say
Pragmatically speaking, recall that we argued above that ni pensarlo in (28a) strengthens the assertive content of the main proposition, with the result that sentences with prepositional NPIs express strong refutations or denials. In parallel, que digamos may trigger a conversational implicature of contraposition which amounts to the denial of the proposition expressed by the sentence. For instance, (42a) may implicate something along the lines of (42b). (42)
a. Mart Pili no es muy alta que digamos. "M.P. is not very tall que digamos." b. She is actually short.
This frequent implicature of contraposition notwithstanding, que digamos does not play the same role as the other propositional NPIs. The essential semantic property of this NPI is its scalar nature. However, unlike other well-known NPIs (e.g. en absoluto "at all"), it does not indicate an "extreme" point (i.e. an endpoint) on a contextually-salient scale. Consider (43a, b): (43)
a. Mari Pili no es en M.P. NE(j is in "Mari Pili isn't tall at all."
absoluto absolute
alta. tall
Propositional NPIs and the Scalar Nature of Polarity
b.
No conoces a los cldsicos en NEGknow to the classics in "You don't know the classics at all."
253
absolute. absolute
The function of en absolute is to deny that in (43a) Mari Pili (or the value corresponding to her) should be located on a scale ordering degrees of "tallness" at all, and in (43b) to deny to the fullest extent possible that the addressee has any knowledge of the classics, i.e. to mark his (lack of) knowledge at an extreme point on a pragmatic scale ordering individuals with respect to the degree of knowledge of the classics that they possess. In contrast, que digamos attenuates the scalar value of the proposition it marks. In the terminology of Israel (1998) it belongs to the class of understating NPIs, not to the more well-known (and more well-studied) class of emphatic NPIs, to which an expression like en absoluto belongs. Consequently, que digamos tends to combine with predicates modified by non-endpoint-marking scalar degree expressions (muy "very", tanto "so much", tan "so", poco "little", mucho "much", etc), as in (44), and cannot combine with absolute predicates or one-time predicates, as shown in (45a, b), respectively, or with endpoint-marking scalar degree expressions like nada "nothing" (45c). (44)
(45)
a. Juan no ha comido mucho que digamos. "Juan hasn't eaten much que digamos." b. Mari Pili no ha tornado tanto el M.P. NEG has taken so-much the "Mari Pili hasn't sun-bathed so much que digamos." c. Gana la carrera no es poco win the race NEG is little "To win the race is no small feat que digamos." a. *No han llegado que digamos. NEG have arrived that we. say "They haven't arrived que digamos." b. *Los autores no murieron que the authors NEG died that "The authors didn't die que digamos." c. *Juan no ha comido nada que J. NEG has eaten nothing that "Juan hasn't eaten anything que digamos."
sol sun
que digamos. that we.say
que that
digamos. we.say
digamos. we. say digamos. we.say
An examination of all 93 examples of the NPI que digamos found in the Corpus de Referenda del Espanol Actual (CREA)6 revealed that 82 (88%) of these co-occurred with an accompanying ' The CREA is accessible online (with username and password) at http://www.rae.es.
254
From Words to Discourse
scalar degree expression like the ones listed above. In those instances where it did not co-occur with a degree expression, que digamos was either used to attenuate a contextually-salient scalar value (46a), or to attenuate, as an "afterthought", a seemingly non-scalar negative (46b). (46)
a. El DRAE no es un modelo de sensibilidad sociologica que digamos, aunque su ultima edicion haya recorrido un gran tranco en esa direccion. "The DRAE (Diccionario de la Real Academia Espanola) is not a model of sociological sensibility que digamos, even though its latest edition has gone a long way in that direction." [CREA, Lingiiistica y Lenguaje, Espana, 1994] b. Pues este modelito de requiem de Chanel no es bonito, que digamos, ni siquiera original. "Well, this little requiem model from Chanel isn't pretty, que digamos, or even original." [CREA, testimonies varies, Argentina, 1977]
In (46a) there is a denial of the high-scalar value represented by modelo de sensibilidad sociologica ("model of sociological sensibility"), leading to the interpretation that the DRAE is to be found at a lower point on a scale (lattice) of sensibilities, possibly restricted to sociological sensibilities. However, the aunque clause which follows explicitly indicates that the last edition of the DRAE is getting closer to "model" status, i.e. it occupies a non-endpoint position on the scale in question. In (46b), similarly, the addition of que digamos to the proposition expressed by the preceding clause serves to attenuate its interpretation: whereas the non-attenuated negated predicate no es bonito might potentially lead to a contrary implicature about the model in question, such an interpretation is ruled out by the appended que digamos, which places the negative proposition in a non-endpoint position on a scale which orders degrees of beauty, or lack thereof.
6. ORDERING PROPOSITIONS Once we have determined the status of que digamos as an attenuating scalar element, an important issue arises, namely what kind(s) of scale is(are) involved. Semantically, scales are denotational orderings. The denotation of an expression is associated with an ordering (partial, strict, etc.) of other elements of the same type (with the same denotation). These orderings can be exploited with a variety of purposes such as the creation of informational patterns or the generation of implicatures. Que digamos, in addition to its attenuating value described above, is a scalar NPI which situates the event or situation described by the predicate on a scale of expectations. Let Expect(a, p) be the degree of expectation of the individual a with respect to the
Prepositional NPIs and the Scalar Nature of Polarity
255
proposition/?. The following expectation order is derived: Expect (a, p) < Expect (a, p'). What this means is that the proposition p is less expected by a than p'. For example, when uttering (47), the speaker places the actual event of Mari Pili shopping in a scale of alternative shopping scenarios in which Mari Pili would have bought a bigger or smaller quantity ordered according to speaker-hearer expectations, as depicted in (48). With respect to this scale in particular, the following holds: Expect(a, p n _k) < Expect (a, pn ). The presence of que digamos serves as an indicator that the position of the predicate on the scale is lower than it should have been or lower than would be expected by the participants in the conversation, so (47) and (48) would be equivalent. (47)
(48)
(49)
Mari Pili no ha comprado muchos vestidos que digamos. M.P. NEG has bought many dresses that we.say "Mari Pili hasn't bought many dresses que digamos." pn = "Mari Pili has bought n dresses" pn-i = "Mari Pili has bought n-1 dresses" pn-k = "Mari Pili has bought n-k dresses" Mari Pili bought a quantity of dresses which was lesser than what was expected.
The role of que digamos in invoking expectations can be seen even more clearly in minimal pairtype examples like the following, where only response B2 can be understood as making explicit reference to a "norm" relating to the amount of snowfall or to some contextually-salient scalar expectation shared by the interlocutors about the snow: (50)
A: ^Nieva en tu pueblo? "Does it snow in your town?" B\:Pues no mucho. "Well not much." B2'.Pues no mucho que digamos. "Well not much que digamos."
The reference to a norm or an expected value that is associated with que digamos, furthermore, appears to be one that the speaker assumes to be shared along with his or her interlocutor(s), or, at the least, one which the speaker is presenting as so shared. This restriction on the use of the form is in addition at least somewhat transparently related to its formal properties: the first person plural verb form digamos indexes the expectations of the speaker and some other person(s), most often an addressee. Thus, we find that que digamos cannot be used felicitously in denials like that seen in the dialogue in (51), where speaker B does not agree with speaker A's assertion that Juan has eaten un monton:
256
(51)
From Words to Discourse
A: Juan ha comido un monton. J has eaten a ton "Juan has eaten a ton." Bl:#(Pero si) No ha comido tanto but if NEG has eaten so.much "(But) He hasn't eaten that much que digamos." B2:(Pero si) No ha comido tanto. but if NEG has eaten so.much "(But) He hasn't eaten that much."
que that
digamos. we.say
Given the analysis above of que digamos as an attenuating NPI, this restriction might not be unexpected. However, as response B2 shows clearly through the felicity of tanto "so much", it is not attenuation per se which makes the use of que digamos odd in such a situation, since B2's reply is still attenuated in that it does not fully refute A's statement. Rather, the infelicity of Bl is due to the fact that the NPI is being used where there is clear disagreement between the interlocutors as to how much Juan has eaten with respect to a scale of eating amounts. Contrast this with the case in (50), where response B2 evokes a norm relating to snowfall, and asserts that the snowfall in the town is to be situated at a point that is less than the norm. At the same time, the presence of que digamos in speaker B's reply indexes the agreement between interlocutors as to what constitutes the norm in question, even though the exact content of that norm remains implicit throughout. Non-modified predicates, i.e. those in which there is no accompanying overt scalar expression, can also be interpreted in scalar fashion as the result of the appearance of que digamos. For example, (52) below is only felicitous as a statement that the quantity of food eaten by Juan was less than expected. The strategy is not completely general however, and is restricted by the semantic nature of the predicate itself. Achievement predicates are not gradable according to a degree of expectation, as shown in (53a), but state predicates are (53b). (52)
(53)
Casi no ha comido que almost NEG has eaten that "He almost hasn't eaten que digamos.'" a. lEste zombi no murio que this zombie NEG died that "This zombie didn't die que digamos." b. Este zombi no estd muerto que thiszombie NEG is dead that "This zombie isn't dead que digamos."
digamos. we.say digamos. we. say digamos. we.say
Prepositional NPIs and the Scalar Nature of Polarity
257
The notion of "expectation" at issue here should be taken here in a very general sense, to the effect that on many occasions the use of this attenuating NPI is related to mid-range, or lower than expected, points in epistemic scales of different sorts. For example, the use of que digamos in (54) may communicate uncertainty on the part of the speaker about exactly how much the item will cost, or implicate that the addressee's assumptions about the item being expensive are not well-founded, etc. (54)
Esto no te costard mucho que thisNEG you will.cost much that "This won't cost you much que digamos."
digamos. we.say
1. ASSOCIATED IMPLICATURES As noted above, despite its attenuating semantic content, que digamos may trigger pragmatically other conversational implicatures. Thus, whereas ni pensarlo in (55a) (examples repeated from above) strengthens the content (i.e. the denial) of the main proposition, que digamos may trigger an implicature of contraposition which amounts to the denial of the proposition expressed by the sentence. For instance, (55b) may implicate (indicated by =>) something along the lines of (56). (55)
a. M pensarlo, no te dare mi Lexus. not.even think.it NEG to-you will.give my L. "Don't even think about it, I will not give you my Lexus." b. Man Pili no es muy alta que digamos. M.P. NEG is very tall that we.say "Mari Pili isn't very tall que digamos."
(56)
She is actually short.
The particular scalar nature of this NPI, i.e. the fact that it is not associated with an emphatic or "endpoint" interpretation but rather with an attenuating one, explains the contrary implicature often associated with (55b). Indeed, attenuating expressions may be, and often are, used in understating fashion in order to conversationally implicate full denial (Bolinger, 1972; Horn, 1989). The reasons for doing this are usually associated with politeness: by negating a high scalar value, one leaves open to inference the exact (lower) scalar value of the predication. An inference which is drawn frequently is that the predication holds at an endpoint, i.e. an emphatically low, scalar value.
258
(57)
From Words to Discourse
a. Man Pili no se pinta muchoque digamos. M.P. NEG REFL paint much that we.say "Mari Pili doesn't put on much make-up que digamos" ==>
b. Mari Pili no se pinta (nada). "Mari Pili doesn't put on any make-up." Nonetheless, it is clear that this implicated interpretation is conversational, not conventional, in nature. The relevant piece of evidence for this classification is that it can be canceled in an appropriate discourse context (Grice, 1975). The cancelable, conversational nature of this implicature is illustrated by the use of que digamos in attenuated contexts like (58a), where no contradiction arises between the two clauses opposed by adversative pero, and no implicature corresponding to (58b) is licensed. (58)
a. Juan no es muy listo que digamos pero es lo J. NEG is very smart that we.say but is it suficientemente listo como para aprobar el examen. sufficiently smart as to pass the exam "Juan isn't very smart que digamos but he's smart enough to pass the exam." =/=> b. Juan es (onto. "Juan is dumb."
It should also be noted that an utterance containing que digamos often gives rise to what might be called an "ironic" interpretation. For example, from the assertion that Mari Pili doesn't wear much make-up "que digamos" in (57a) one can extract the purportedly "ironic" interpretation that she doesn't wear any at all. However, as we have shown here, such an interpretation does not always arise, since there is often contextual material available to defeat it. Furthermore, there is no need to recur to a more complex pragmatic concept like "irony" —which may not even be the appropriate one to describe these cases— in order to account for such examples. All that is needed is the analysis of que digamos as an attenuating NPI which, like many other understating expressions (Bolinger, 1972), may often be used to conversationally implicate an emphatic interpretation.
8. NON-DECLARATIVES As we pointed out in the introductory sections of this paper, NPIs are licensed in certain nondeclarative, non-negative environments such as (i) the protasis of conditionals and (ii) questions.
Prepositional NP!s and the Scalar Nature of Polarity
259
For example, the NPIs anybody and nada are licensed in the protasis/antecedent of the conditionals in (59), and in a similar fashion nunca and nada may occur in the questions in (60). (59)
(60)
a. If anybody moves, there will be trouble. b. Si haces nada que ocasione problemas, te expulsardn. If do nothing that causes problems you they.will.expel "If you do anything that causes problems they'll expel you." a. '(jQuien ha estado nunca en Las Bahamas! who has been never in The Bahamas "Who has ever been in The Bahamas?" b. iQuien ha dicho nadal who has said nothing "Who has said anything?"
Sentences (60a, b) have only a rhetorical interpretation, where it is pragmatically presupposed that nobody has ever been in the Bahamas or that nobody has said anything, respectively. This type of rhetorical reading of questions associated with NPIs has been related by GutierrezRexach (1997b, 1998) to the algebraic and scalar properties of NPIs. Concretely, this interpretation is related to the "bottom of the scale" properties of these elements which, in combination with the extensional semantics of questions, derive this particular reading. What is of interest for our present purposes is that que digamos only participates in conditional and interrogative constructions when these are interpreted rhetorically. In the conditional construction (6la) there is a rhetorical question embedded in the consequent. The standard material conditional in (61b) is incompatible with the NPI. Thus, que digamos behaves like other NPIs in triggering rhetorical readings. Additionally, the content of the sz-clause in (6la), i.e. the proposition "Juan no es muy listo que digamos", must be interpreted —NPI included— as being ascribed to someone who cannot be identified as the speaker (such as the addressee or another interlocutor), while the rhetorical question in the main clause must be understood as being the viewpoint of the speaker. (61)
a. Si Juan no es muy listo que digamos, entonces ^corno va a aprobar el examen. if J. NEG is very smart that we.say then how go to pass the exam "If Juan isn't very smart que digamos, then how is he going to pass the exam?" b. *Si Juan no es muy listo que digamos entonces no va a aprobar el examen. "If Juan isn't very smart que digamos then he's not going to pass the exam."
260
From Words to Discourse
9. TAMPOCO AND ATTENUATION Finally, our analysis ofque digamos accounts for still more distributional properties of this NPI. For example, as argued by Schwenter (2000), the n-word tampoco is polysemous between two distinct meanings in Spanish. The first, canonical, meaning (translatable as "(n)either") amounts to a negated coordination and requires that a pragmatically parallel negative proposition be accessible in the discourse context, as in (62). The second, non-canonical, meaning functions as an attenuating scalar particle and does not require a prior negative in the discourse context to be used felicitously as the examples in (63) show. (62) (63)
[El empleado no ha venido...] Eljefe tampoco ha venido. "[The employee hasn't come...] The boss hasn't come either." a. A: \Juan ha comidoun montonl "Juan has eaten a ton!" B: Tampoco ha comido tanto. "He hasn't eaten that much." b. A: Muchas gracias por ser tan puntual. "Thanks for being so punctual." (said ironically) B: Tampoco he llegado tan tarde. "I haven't arrived so late."
Given our analysis of this NPI as a marker of scalar attenuation, the prediction is that que digamos will always be compatible with the second, attenuating, meaning of tampoco, since it is only this one that is inherently sensitive to contextually determined pragmatic scalarity. On the other hand, que digamos will be compatible with the canonical meaning of tampoco only when the former's licensing requirements are met by some other element(s) in the sentence. Thus, the sentences in (63 a, b), which contain the attenuating scalar particle use of tampoco, are fully felicitous as alternative responses in the two discourses in (64). However, the discourse in (65), an alternate version of (62), where the canonical use of tampoco is at issue, is not felicitous. (64) (65)
a. Tampoco ha comido tanto que digamos. b. Tampoco he llegado tan tarde que digamos. [El empleado no ha venido...] #Eljefe tampoco ha venido que digamos.
On the other hand, when the "canonical" use of tampoco is found in a context in which the attenuating scalar properties ofque digamos are satisfied, then the occurrence of the latter is also acceptable, as shown in (66):
Prepositional NPIs and the Scalar Nature of Polarity
(66)
261
{El empleado no ha comido mucho...] Eljefe tampoco ha comido mucho que digamos. "[The employee hasn't eaten much...] The boss hasn't eaten much either que digamos"
10. CONCLUSION Research on NPIs has tended to focus either on the properties of the trigger or on the characterization of the special "sensitivity" of polarity expressions. Less attention has been paid to the meaningful contribution of NPIs to the sentences in which they appear. In this paper, we have contributed especially to this latter research direction by analyzing a class of NPIs which has not received any attention to date, to the best of our knowledge. The importance of prepositional NPIs is that in addition to being sensitive to the semantic properties of the trigger, they also contribute specific semantic and pragmatic properties to the sentence in which they occur. More specifically, we have focused on the prepositional NPI que digamos. We have argued that que digamos is a prepositional NPI with scalar properties, but, unlike other prepositional NPIs in Spanish (as well as in other languages), it conveys as part of its conventional semantic content attenuating meaning instead of emphatic meaning. Based on its possible licensors, it can be classified as a strong NPI. Its main pragmatic function in discourse is to contravene scalar expectations among interlocutors, whether explicit or implicit. It invokes a pragmatic scale of expectations and indicates that the proposition it marks is less expected, or more unexpected, than some alternative proposition(s) located on the same scale.
REFERENCES Baker, C. L. 1970. "Double Negatives." Linguistic Inquiry 1. 169-186. Bolinger, D. L. 1972. Degree Words. The Hague: Mouton. Bosque, I. 1980. Sobre la Negacion. Madrid: Catedra. Bosque, I. 1994. "La Negacion y el Principio de las Categorias Vacias." In V. Demonte (ed.) Gramdtica del Espanol. Mexico: Pubs. NRFH, 167-200. Fauconnier, G. 1975a. "Pragmatic Scales and Logical Structure." Linguistic Inquiry 6. 353-375. Fauconnier, G. 1975b. "Polarity and the Scale Principle." CIS 11. 188-199. Fauconnier, G. 1978. "Implication Reversal in a Natural Language." In F. Guenthner and S. Schmidt (eds.) Formal Semantics and Pragmatics for Natural Languages. Dordrecht: Foris, 289-301. Giannakidou, A. 1998. Polarity Sensitivity as (Non) Veridical Dependency. Amsterdam: John Benjamins. Giannakidou, A. 1999. "Affective Dependencies." Linguistics and Philosophy 22. 367-421. Grice, H. P. 1975. "The Logic of Conversation". Syntax and semantics 3: Speech Acts, ed. by Peter Cole and Jerry Morgan, 41-58. New York: Academic Press. Groenendijk, J. and M. Stokhof. 1984. The Semantics of Questions and the Pragmatics of Answers. Ph.D. diss. University of Amsterdam.
262
From Words to Discourse
Groenendijk, J. and M. Stokhof. 1989. "Type-Shifting Rules and the Semantics of Interrogatives." In G. Chierchia, B. H. Partee and R. Turner (eds.) Properties, Types and Meaning. Vol 2: Semantic Issues. Kluwer. Dordrecht. Groenendijk, J and M. Stokhof. 1997. "Questions." In J. van Benthem and A. ter Meulen (eds.) Handbook of Logic and Language, Cambridge, Mass: MIT Press, 1055-1124. Gutierrez-Rexach, J. 1997a. "Questions and Generalized Quantifiers." In A. Szabolcsi (ed.) Ways of Scope Taking. Dordrecht: Kluwer, 409-452. Gutierrez-Rexach, J. 1997b. "The Semantic Basis of NPI Licensing in Questions." MIT Working Papers in Linguistics 31.359-376. Gutierrez-Rexach, J. 1998. "Rhetorical Questions, Relevance and Scales." International Journal Revista Alicantina de Estudios Ingleses (R.A.E.I.) 11. 139-156. Haegeman, L. and Raffaella Zanuttini. 1991. "Negative Heads and the Neg-Criterion." The Linguistic Review 8. 233-251. Hamblin, C. 1973. "Questions in Montague English." Foundations of Language 10. 41-53. Heim, I. 1984. "A Note on Negative Polarity and Downward Entailingness." NELS 14. 98-107. Hoeksema, J. 1984. "Negative Polarity and the Comparative." Natural Language and Linguistic Theory 1. 403-434. Horn, L. R. 1972. On the Semantic Properties of Logical Operators in English. PhD diss. UCLA. Horn, L. R. 1989. A Natural History of Negation. Chicago: University of Chicago Press. Israel, M. 1996. "Polarity Sensitivity as Lexical Semantics." Linguistics and Philosophy 19. 619-666. Israel, M. 1998. The Rhetoric of Grammar: Scalar Reasoning and Polarity Sensitivity. PhD diss. UCSD. Kadmon, N. andF. Ladman. 1993. "Any." Linguistics and Philosophy 15. 353-422. Klima, E. 1964. "Negation in English." In J. Fodor and J. Katz (eds) The Structure of Language. Englewood Cliffs: Prentice Hall. 246-323 Krifka, M. 1991. "Some Remarks on Polarity Items." In D. Zaefferer (ed.) Semantic Universals and Universal Semantics. Dordrecht: Foris, 150-189. Krifka, M. 1995. "The Semantics and Pragmatics of Polarity Items." Linguistic Analysis 25. Ladusaw, W. 1979. Polarity Senstivity as Inherent Scope Relations. PhD diss. UT-Austin. Ladusaw, W. 1996. "Negation and Polarity Items." In S. Lappin (ed.) The Handbook of Contemporary Semantic Theory. Oxford: Blackwell. 321-341. Laka, I. 1990. Negation in Syntax. PhD diss. MIT. Lee, C. 1999. "Types of NPIs and Nonveridicality in Korean and Other Languages." In G. Storto (ed.) Syntax at Sunset 2, UCLA Working Papers in Linguistics, 96-132. Lee, Y. and Horn, L. 1995. "Any as Indefinite Plus Even". Ms. Yale University Linebarger, M. (1980) The Grammar of Negative Polarity. PhD. diss. MIT. Linebarger, M. (1987) "Polarity and Grammatical Representation." Linguistics and Philosophy 10. 325-387. Longobardi, G. 1991. "In Defense of the Correspondence Hypothesis; Island Effects and Parasitic Gaps Constructions in Logical Form." Logical Structure and Linguistic Structure, ed. by James Huang and Robert May, 149-196. Dordrecht: Kluwer. Montague, R. 1969. "On the Nature of Certain Philosophical Entities." The Monist 53. 159-194. Nam, Seungho. 1994. "Another Type of Negative Polarity Item." In M. Kanazawa and C. Pinon (eds.) Dynamics, Polarity and Quantification. Stanford, CA: CSLI publications, 3-16. Schwenter, S. A. 2000. "Two Kinds of Tampoco." Paper presented at the Fourth Hispanic Linguistics Symposium, Bloomington, Indiana. Vallduvi, E. 1994. "Polarity Items, N-words and Minimizers in Catalan and Spanish." Probus 6.263-294. Zwarts, F. 1993. "Three Types of Polarity." Ms. University of Groningen. Zwarts, F. 1995. "Non Veridical Contexts." Linguistic Analysis 25. 286-312.
263
13
DEGREE QUANTIFICATION AND MODAL OPERATORS IN SPANISH Ignacio Basque, Universidad Complutense de Madrid
1. DEGREE QUANTIFICATION AND MODALITY One of the reasons why the traditional concept of "modality" has not had much success in contemporary formal linguistics is that it seems to be too extensive.* Ever since the linguists from the Geneva school, many authors have coincided in grouping together a wide and wellknown range of grammatical phenomena in the broad area of modality. In spite of their syntactic differences, in one way or another these are said to show aspects related to speakers' attitudes to I would like to thank a number of people for their help in the various stages that this paper has undergone. It was Eustaquio Barjau, who teaches Pragmatics at the Complutense University of Madrid, who first pointed out to me the basic directive effect which I try to analyze here, and also the idea that the meaning of the degree quantifiers that I will deal with is related to speaker's attitudes and should be related somehow to speach act theory. In the present article, I concentrate on the grammatical structures in which these quantifiers appear, but I would like to express my gratitude to Eustaquio for having put me onto the phenomenon that I shall attempt to analyse in this paper (even if I do so with formal tools in which he is not particularly interested). I should also like to thank Luis Saez for his numerous observations and suggestions in almost all the stages of this investigation. He has made an important contribution to this paper, and has also encouraged me to work on an uncommon subject such as this. I also want to thank him for convincing me that I was not wrong in thinking that all this was relevant for current issues in grammatical theory. Many thanks also to Carlos Piera, Carme Picallo, Violeta Demonte, Margarita Suner, Esther Torrego and Maria Luisa Rivero for several remarks (general, specific and/or bibliographical) on this paper. The main parts of the text were presented at the Autonomus University of Barcelona (thanks to Maria-Lluisa Hernanz, Jose Maria Brucart and M. Teresa Espinal for several comments at the presentation) and the 1993 Going Romance Meeting at Utrecht, where I benefited from the observations of Juan Uriagereka, Anne Zribi-Hertz and Dennis Delfitto, among others. A somehow different version was presented at the XXIV Linguistic Symposium on Romance Languages at Los Angeles in March, 1994, and I would also like to thank the audience at this conference for several remarks. A considerably shorter version of this paper was published in the proceedings of LSRL-24. See Bosque (1996).
264
From Words to Discourse
messages by the way they are uttered, interpreted or assessed. Modality has, thus, traditionally been seen as the cognitive, volitive or emotional qualification that the speaker expresses about events or, in more general terms, as a link between his/her thoughts and the facts he or she is verbally describing.1 There is a relatively long list of grammatical phenomena that have, since the 1940s, repeatedly appeared in studies on the concept of modality. It includes modal inflection, a large number of adverbs, many types of speech acts, negation, modal verbs and also some of the socalled discourse connectives. Certain predicates should be added to this list, such as those of propositional attitude and the factives. In other words, a large series of units which, as has often been repeated, have in common the fact that they represent different aspects of the classical "animus loquendi". This linguistic tradition is accompanied by the logical tradition concerning a rather more restrictive variant of the concept of modality which attempts to establish propositional truth conditions. From this point of view, the modality of a proposition introduces a reference framework which allows it to be assessed: it tells us whether it is necessarily or contingently true or false, as regards a certain state of affairs, or within a particular set of possible worlds. I would like to discuss a small fragment of modality in the grammar of Spanish, although I should say that, apart for some aspects of the analysis related to prenominal adjectives and subjunctive clauses, most of the phenomena I shall examine are not exclusive to Spanish, or even to the Romance languages. I shall try to outline a way of integrating the modal properties of a number of degree quantifiers and quasi-quantifiers into the restrictive concept of grammar provided by the generative approach to the syntax/semantics-pragmatics interface. In sections one to five, I will deal with one of the cases in which these quantifiers interact with modal structures. It concerns degree quantifiers that denote EXCESS, such as demasiado(s), ('too much/many'), desmesurado ('disproportionate'), excesivo ('excessive'), and their respective adverbs, should they exist. I will argue that some of the essential properties of these quantifiers follow from their intensional nature. My analysis is sketched in (1):
' Basic references on modality include Palmer (1979, 1986), Coates (1983), Lyons (1975: chap. 17), Hiz (1978), Leech (1971), Jackendoff (1972), Heny and Richards (1983), David and Kleiber (1983), and many more. Several specific issues of certain linguistics journals are devoted to modality, including Folia Linguistica 21: 1. 1987; Langue Franqaise 84. 1989; Langages 43. 1976; and others. See below footnote 4. The existing bibliography covers many more works than I can possibly mention here. Some of the items cited above are overviews or general presentations. Although modal and modal-like structures have been analized in detail in the the GB framework (for Romance, see Picallo (1990), Kempchinsky (1986) and Torrego (1996), among many others), only in recent syntactic work has been some discussion on the role that modal information plays in the modern theory of functional projections. In fact, modal inflection has been claimed to be the head of independent functional projections in works such as Rivero (1991), Terzi (1991) and Pollock (1993). As for non-inflectional modality, Polletto (1993) has presented some syntactic evidence for a specific functional projection between CP and AgrSP, which she names 'Modal Phrase', hosting some modal markers in the syntax of Northern Italian dialects. The main purpose of this paper is to argue for an epistemic non-inflectional modal projection in the level of Logical Form. I will do so by using a very different type of evidence than the type presented in these works.
Degree Quantification and Modal Operators
(1)
265
Degree quantifiers that denote excess are licensed at LF by an epistemic modal operator. Such quantifier phrases move to the Specifier of a Modal Phrase (ModP) headed by an M° epistemic head. The properties of these quantifiers are the result of the interaction of this operator with other components of the clause and with the syntactic processes which take place therein.
My starting point will be a property of these quantifiers that is to some extent unexpected. I am referring to the fact that they are incompatible with imperatives and, in general, directive speech acts. I shall call this property anti-directive polarity (ADP). Let us look at the following contrasts: (2)
(3)
a. Juan trabaja demasiado. "John works too much." b. *Porfavor, trabajen demasiado. "Please, work too much." a. Viajas excesivamente. "You travel excessively." b. *Viajemos excesivamente. "Let's travel excessively."
Other examples can easily be constructed with items cited above. The reader may notice that negation changes things and that certain "ironical" interpretations seem to be possible, but I will return to this later. If we put this aside for the moment, the first obvious and simple questions are "What prevents us from saying these things?" or "What exactly is the type of grammatical incompatibility that these sentences manifest?". I think that a good clue which brings us close to what is happening can be obtained by extending the ADP effect to other non-imperative speech acts: these quantifiers automatically rule out the illocutionary interpretations of directive predicates, even if these verbs fulfill all the necessary requirements for that reading: (4)
(5)
a. Te prometo cuatro cosas. "I promise you four things." b. Hablare con varias personas. "I will speak to several people." a. Te prometo demasiadas cosas. "I promise you too many things." b. Hablare con demasiadas personas. "I will speak to too many people."
[A PROMISE] [A COMMITMENT] [A JUDGEMENT, NOT A PROMISE] [A JUDGEMENT, NOT A COMMITMENT]
266
From Words to Discourse
The sentences in (4) may have illocutionary readings, i.e. they can be interpreted as verbal actions. On the other hand, the sentences in (5) exclude this possibility. Sentence (5a) is only grammatical as regards what I will call "epistemic interpretation", which can be approximately paraphrased as "I think that there are too many things that I promise you", that is, it presents something that is "judged" or "considered". Similarly, the interpretation in which the future tense has an illocutionary force is the one which permits commitative speech acts. That is how (4b) can be read. But this reading is ruled out for (5b) due to the presence of the quantifier demasiadas. This sentence means something close to "I think that the number of people that I will talk to is excessive". For this reason, I shall use the term epistemic quantifiers (henceforth EQs) to refer to items such as demasiado and the rest of excess quantifiers. As C. Picallo and L. Saez have pointed out to me, some notions opposite to "excess" may also be related to epistemic validation. Good candidates are "insufficiency", "scarcity" or "default". In fact, adjectives such as escaso ("scarce") clearly show ADP effects. This issue will not be addressed in this paper. Someone might say that the ADP effect is a "pragmatic phenomenon" of some kind. My claim is that there is a sense in which it is not, and that there is another, somehow looser sense, in which all facts related to asserting domains (including belief contexts, reported speech, evidentials, parentheticals and other structures) have some clear implications for pragmatics. For this reason, I shall attempt to present a grammatical analysis of EQs which has some formal consequences that I think cannot be directly derived from a mere pragmatic approach. Some studies on demasiado or its equivalents in other languages, such as Jayez (1985) and Carel (1992), take into account the evaluation of the pragmatic conditions that have to be fulfilled in order to exceed such an implicit norm, and also argue for an argumentative interpretation of the notion of excess relative to extra-grammatical information. I will not deny that those conditions are relevant, particularly the argumentative ones, but I do not see how one could derive the ADP effect from them. Notice that even the most favorable practical situation is not enough for a sentence such as (2b) to be valid: not even a stress-ridden businessman, unable to keep his business afloat due to low performance, could use (2b) to beg his/her partners or employees to make an excessive effort at work. The context is suitable, but the grammar is against doing so. I think this suggests that sentences as the ones above are strictly "ungrammatical", and not merely "unusual" or "strange". On the other hand, notice that it does not seem to be possible to explain the ADP effect by conversational principles or by using discourse strategies. It might be thought that the irregularity of sentences such as (2b) is due to some principle of conversation, based on an assumed contradiction between this kind of directive expression and the fact that, by using degree quantifiers of excess, we are crossing or overcoming some limit of regular behavior. I simply do not think that we would be on the right track with this idea. There is nothing in the grammar of imperatives to prevent one from asking or ordering someone else to perform actions which are excessive or surpass the implicit patterns of standard behavior. This is easily proved by perfectly 2
At least, it is not mentioned in previous analysis of words like too, including Bolinger (1972), Jayez (1985) and
Degree Quantification and Modal Operators
267
grammatical imperative sentences such as Drink twenty beers in a row! or Do the most awful and unexpected thing that one can imagine! Clearly supporting this idea is the fact that a Spanish verb such as propasarse ('go too far') can be used in the imperative, but its English counterpart can not: Propasemonos ('Let's go too far'). The reason is that the equivalent expression in English contains the degree quantifier too. This amounts to say that any analysis of these expressions must refer to the grammatical properties of these lexical items, and not necessarily to the practical aspects of excess or disproportion.3 The other, somehow looser sense, in which EQs are related to pragmatics is provided by the properties they share with some adverbs in assertive contexts. It is a well-known fact that epistemic adverbs, such as probablemente ('probably') are incompatible with speech acts, a property that has traditionally been related to their weak assertive properties as modal categories. There are also some adverbs that give rise to ambiguities in contexts of transparency, as in cases in which assertions are attributed either to the speaker or to the reporter. In a way close to which en efecto ('in effect') provides this type of ambiguity in (6a), one might say that demasiado gives rise to two similar interpretations in (6b): (6)
a. Juan dijo que, en efecto, no podria venir. "J. said that, in effect, he could not come." b. Juan cree que Pedro trabaja demasiado. "J. believes that P. works too much."
These ambiguities generally show up in contexts with several successive subordinate complements of epistemic predicates. I will not deny that a relation between (6a) and (6b) exists, in the intended interpretations, but I think that from the fact that strong and weak assertive predicates (so-called 'bridge verbs') have some properties of parentheticals in contexts of transparency it follows that a restrictive formal specification of these assertive domains is necessary, rather than a claim that an exclusively pragmatic analysis should do the work here. In any case, my attempt to give some formal concretion to the epistemic domain in which EQs are
Carel (1992). See footnote * above. 3 Another type of approach might relate the ADP effect to the "hidden argument" that demasiado seems to have ('too much/many for a certain individual or for a given situation'). However, I think that the arguments for an epistemic modal projection presented in section two cannot be accounted for from this perspective. 4 Much has been written on this issue, both in the literature on reported speech and in the bibliography on belief contexts. Basic references on the former include Davidson (1988), Richard (1986), Banfield (1982), Coulmas (1986) and Clark and Gerrig (1990), among many others. For Spanish see Maldonado (1991). As for the latter, see Parrel (1983), Hall Partee (1973) --and the comments by Cheng and by Moravsick therein-, Jackendoff (1985), Heim (1992), Farkas (1992), and still many others. Related lines of research are represented by work on subjectoriented adverbials, parentheticals, evidentials, weak assertive predicates and other structures on which an extensive bibliography can be gathered. On parentheticals see Stucky (1987), Espinal (1991a), and many more. The large number of relevant investigations for each of these topics makes almost impossible to mention all the relevant work here.
268
From Words to Discourse
licensed does not imply any rejection of the fact that assertive domains are speaker-oriented frames, in any sense that a pragmatic theory of assertion might consider relevant. Additional evidence for the modal nature of extreme degree comes from the fact that, when used in answers, EQs do not necessarily provide information, but often suggest a type of assessment which frequently gives rise to ironic effects. If someone asks us How many people came to your party? and we answer Too many, we are obviously not identifying a number, but are instead "assessing" or "evaluating" that quantity, which we choose not to identify. This clearly contrasts with other identifying non ironical answers such as the ones with many or several. The essential point to be made here is that we have to account for the fact that EQ phrases have no discursive domain that enables us to refer to them as if they denote a specific group: the expression demasiados libros ('too many books') does not denote a group of items with the property of being books, and therefore we cannot refer to this set extensionally. EQs are necessarily non-specific items, and this makes an essential difference as regards quantifiers such as varios ('several') or pocos ('a few'). The latter are said to be 'weak determiners' in that they allow for the ambiguity between the "cardinal interpretation" and the so-called "quantificational interpretation" (in Milsark's, 1974 terms). EQs, on the contrary, do not really have a quantificational force of their own, but one that is embodied into them through an epistemic predication. I will devote section three to develop the issue of non-specificity. The immediate questions that arise are the following: "How can one explain ADP in grammatical terms?", "How can the differences between EQs and other quantifiers be incorporated into the grammar?" I shall try to provide a plausible answer to these questions in the rest of this paper.
2. A MODAL PROJECTION AT LOGICAL FORM I will take the ADP effect to be a grammatical phenomenon that should be accounted for by sufficiently restrictive linguistic principles. That is, by arguing that excess is a modal notion, we are not liberated from the need to propose an LF for EQ structures. I would like to recall that degree phrases (DegPs) are semantic entities that are not very different from ordinary quantificational phrases. As this been repeatedly pointed out in the literature, degree operators can be interpreted as quantifiers that bind variables whose range are not individuals, but extents, so that a DegP is merely a certain type of QP. But EQs are not simple degree quantifiers. I would like to suggest that, as well as being indefinite, EQs are bound to an epistemic evaluation, i.e. as opposed to varios ('several') or muchos ('many'), a quantifier such as demasiado is not licensed because of the variable that "ordinary QR" originates in LF, but instead needs a modal operator which relates its existence to the need for such an evaluation. It is a well-known fact (May, 1977, 1985) that QPs and w/z-phrases are not referential elements that can receive a theta-role, as opposed to variables, i.e. traces and pronouns locally bound by operators. The reason why QR is compulsory in such cases lies on the fact that if there was no movement, the theta-criterion
Degree Quantification and Modal Operators
269
would in fact be violated, since it requires a strict correspondence between arguments and thetaroles at LF. In other words, LF movement enables ordinary QPs to be recognized as complements or as adjuncts of other predicates. In the analysis I am proposing, EQs move at LF because a part of their meaning is a modal component that has to be syntactically fulfilled. The variable that EQs leave is the one that enables us to understand that their interpretation is "relative to" the intensional frame that the clause is under. In this sense, it is also EQ movement that enables these quantifiers to be recognized as EQs at LF. As regards the formal representation of the relevant structure, there are various possibilities. I shall look at a couple of these: (7)
a. The null modal operator is in CP. b. The null modal operator is in a modal projection under CP.
The two analyses of (3a) according to (7) appear in (8a) and (8b): (8)
a. THE CP OPTION: [CP demasiado; [c [c° [Mod]] [[P Juan trabaja ei ]]]] b. THE ModP OPTION: [CP [ModP demasiado; [M> [M° [IP Juan trabaja et ]]]]]
As can be seen, in (8a) the head of CP is an epistemic abstract modal, and the QP moves to Spec/CP in this structure at LF. In (8b), on the other hand, the EQ moves to the specifier of the modal projection formed by such an operator (Spec/ModP). Let us briefly compare the options: The CP OPTION: This option would, intuitively, appear to be the best, so I shall attempt to justify why I am going to choose the other one. In option (8a), the features of the epistemic modal (shown by [Mod]), are incompatible with those of the imperative form, because they occupy the same position. This analysis apparently shows fairly adequately the relationship between pseudo-assertive verbs of propositional attitude, such as considerar ('consider'), creer ('believe'), etc. and epistemic modals. The connections between them have been argued over consistently in quite a few works on modality: Lyons (1975), Jackendoff (1972), Palmer (1986), Coates (1983), and many others. McDowell (19871 argued that epistemic modals behave as truth conditional operators in a fairly similar way to w/z-phrases. In fact, epistemic modals are incompatible with questions, which leads McDowell to assume that they move to CP at LF. Rivero (1994a, 1994b) shows that imperative verbs move to C° to acquire the illocutionary force of such speech acts. Option (8a) also apparently simplifies the number of functional projections, so that we would just have to assume that performative verbs move to C° at LF. The ADP effect would therefore be easily and naturally explained. THE ModP OPTION: In this option, which has more "structural space" than the first one, it is the relevant features of C° that select a ModP epistemic projection. I shall argue that this option is preferable. We may certainly ask what structural space there is between CP and AgrP. I think that there is some relationship between a ModP under CP and some functional projections that
270
From Words to Discourse
have been proposed between CP and AgrP. I have in mind the the AgrlP proposed by Cardinaletti and Roberts (1991), the Focus Phrase proposed by Uriagereka (1995), the SigmaP proposed by Laka (1990), the ModP proposed by Poletto (1993), the Polarity Phrase proposed by Culicover (1992), the Modal Phrase proposed by Terzi (1991) for Greek subjunctive particles, and also for Rivero (1991) for some futures in Balkan languages, and finally, the non-Case Tense-related position proposed by Zubizarreta (1999). These proposals seem to be coincident, in part, in a number of structures in which focalized elements appear (recall the parenthetical properties of weak assertive verbs mentioned above), although Cardinaletti and Robert's and Uriagereka's proposals are intended to extend to independent cases related to nominative case assignment and person identification. A clear and simple example of the need for such a projection (call it FP) between CP and IP is provided by the fact the so called "bridge verbs" allow for focalized elements in subordinate clauses, with verb raising to F°, but non-assertive verbs, such as factives, clearly reject that movement, since no space is available for the focus phrase to move to:5 (9)
a. Creo que ahi vive Pepe. "I think that P. lives there." b. *Lamento que ahi viva Pepe. "I regret that P. lives there."
There is some relation between this FP being selected by C°, and the ModP being selected by C° that I suggest in (8b). The main difference is related to the epistemic nature of the latter. In a sense, the epistemic ModP is close to a FP plus the relevant epistemic features that are needed to explain the ADP effect. In this sense, ModP is a functional projection with very rich (noninflectional) semantic information. Compared to the CP option, the ModP option looks more like the structure we find in subordinate subjunctive clauses. It seems logical to accept, however, that imperative features are in COMP, so that it may be assumed that a C° with the grammatical information features that belong to imperatives selects no modal projection. In fact, modals are incompatible with imperatives, as is well-known. I would not like to suggest that ModP is exactly an "illocutionary projection", but certainly that it is a "modal projection" in the classic sense of the term modality, i.e. in the sense that modal information introduces a reference framework that assesses or relativizes prepositional contents. In such conditions, the EQ cannot be identified in imperatives either, given the ModP option. Its variable needs to be bound by an epistemic modal operator, but it cannot be bound in the situation described, as no modal projection is selected. Both analyses plausibly reflect the intuitive explanation that one can give to the APD effect: one cannot order an action at the same time one is assessing or judging a prepositional content. In
Degree Quantification and Modal Operators
271
more formal terms, in both cases we obtain a structure in which either the illocutionary interpretation is satisfied --and the EQ variable is not bound and remains uninterpreted- or else the variable is bound, and no illocutionary interpretation is possible. An alternative analysis to this "selectional approach" might be a "head-to-head approach". In this approach one would say that C° always selects M°, so that what the ADP shows is simply the result of M° preventing head movement of V° to C°, in a way closer to Rivero's explanation for the lack of negated imperatives in Spanish. However, since the ADP effect also shows up in non-illocutionary modal environments, including subordinate clauses with subjunctive mood (see section 5), one would have to motivate V-to-C movement in these subordinate clauses. Maybe such a movement takes place at LF, but for the moment I do not think that there is a strong justification for it, so I will not take that option. Comparing the CP and the ModP options, I believe that there are some reasons for choosing the latter, and having the ModP as a lower projection than CP. I suggest the following arguments in favor of this option: a) Epistemic operators that validate EQs introduce a type of modality that shares the speakeroriented properties of weak assertive predicates. However, epistemic operators are not restricted to root clauses. In other words, subordinate imperatives do not exist, but EQs may appear in subordinate clauses. Let us look at the following examples: (10)
a. *For favor, recuerda demasiadas cosas. "Please remember too many things." b. For favor, recuerda que vino demasiada gente. "Please remember that too many people came."
Notice that the imperative recuerda ('remember') does not cause an ADP effect in (lOb). For this, we must assume that the subordinate CP contains a ModP which enables it to be licensed. In (lOa), there is no ModP projection, because the imperative in C° does not select for one. The imperative C° in the main clause of (lOb) prevents its having a ModP present, but the verb recordar ('remember'), has a clause as its complement, therefore nothing prevents the lowest C° selecting that ModP in the subordinate clause. This subordinate effect makes the properties of the operators that license EQs fairly far removed from those we would expect in directive speech acts. If the verb in the subordinate clause is in subjunctive the situation becomes more complex. This will be discussed in section 5 of this paper. b) If we identify the modal projection that licenses EQs with CP, it would be hard to include other operators that independently have to bind their variables, as in the case of relatives. Let us look at the following contrast:
For a list of properties of indicative-taking verbs in Spanish related to strong and weak assertive contexts see
272
(11)
From Words to Discourse
a. *Sepdrame demasiados libros que tengan dibujos. "Set aside for me too many books with drawings." b. Sepdrame los libros que tengan demasiados dibujos. "Set aside for me the books that have too many drawings."
Once again, the existence of a clause makes it possible for EQs to be licensed, but the relative operator in (lib) occupies the position of Spec/CP, without affecting EQ licensing in any way. There appear to be no independent reasons in this case to assume any kind of "absorption" --in the sense used by Higginbotham and May (1981)-- of the relative and modal operators, and more so when these range over very different types of entities. The relevant issue, obviously, is the absence of illocutionary features in C°, so that ModP may appear. It is also interesting to point out that, if ModP and CP are different, then ModP cannot be placed above CP, because, if it were, we would loose adjacency between the relative operator and its antecedent. Once again, this kind of structure cannot be clearly interpreted from option (8a), so I will take (8b) as a better alternative.6 c) The process by which performative predicates acquire illocutionary force is conditioned by a number of well-known lexical properties (belonging to a grammatical class, temporal and personal restrictions, etc.), which suggests that their strictly syntactic nature is not similar to that of imperatives. It may therefore be assumed that it is the projection accommodating the modal operator that enables performative predicates to be identified at LF (see Laka (1990) for a similar idea concerning imperatives). In fact, whereas modals in Romance make up compound predicates with the main verbs that complement them, illocutionary structures cancel out this possibility, i.e. we cannot use Debo prometerlo ('I must promise') as a promise. If we assume that verbs that acquire an illocutionary interpretation move to C° at LF, it is obvious that EQs will not be able to be licensed in such circumstances. From a formal point of view, imperative verbs do not show the same syntactic properties as performatives. The latter have preverbal clitics, as opposed to imperatives, which have postverbal clitics. Thus, we do not have *Lo ved ('You [pi.] see [IMP] it'), but we say Lo prometo ('I promise'). Rooryck (1992) claims that the agreement features of the imperatives in C° (AGR-C0) are different from those that AGR provides in declarative clauses and, on the other hand, they are closer to those which clitics provide, insofar as both are inherent. He suggests that in these sentences AGR-C0 should intervene between the clitic and the part of the complex trace that corresponds to the clitic. I would like to stress that EQs are incompatible with a projection with illocutionary features, but not necessarily with a projection which has the characteristic morphology of imperatives. It is interesting to verify that these two properties can be separated. Let us see how. Bosque( 1990: section 3). 6 Notice that in arguing for the ModP option and against the CP option I am avoiding interrogative sentences, which may look as natural candidates for the comparison. I am doing this because the relevant issue is not whether these
Degree Quantification and Modal Operators
273
There are imperatives that reject preverbal clitics and negation like the rest, but that on the other hand lack illocutionary force. I am referring to the class of rhetorical imperatives that are usually called conditional, which have been studied by various authors, including Bolinger (1967: chap. II.l), Dobrovie-Sorin (1984), Davis (1979) and Clark (1993). I will not go over all their properties here, but I would like to recall that when imperatives function in disjunctive coordination, they have illocutionary force. If, on the other hand, coordination is copulative, they lack illocutionary force and are interpreted as simple conditionals: (12)
a. Di la verdad o te mato. "Tell the truth or I'll kill you." b. Di la verdad y estards per dido. "Tell the truth and you're lost."
In other words, (12a) is equivalent to "if you do not tell the truth, I will kill you", whereas (12b) means "if you tell the truth, you are lost" (no negation). The correspondent sentence with a clitic is Dila Tell it', and not *La di, in both cases. As expected, these sentences can not be negated preserving imperative morphology, but, interestingly, conjunctive imperatives are compatible with EQs, whereas disjunctives show the ADP effect: (13)
a. *Trabaja demasiado. "Work too much!" b. Trabaja demasiado y morirds joven. "Work too much and you'll die young." c. *Trabaja demasiado o te quedards sin vacaciones. "Work too much or you will get no holidays."
It is fairly obvious that disjunctive imperatives have illocutionary force and conjunctives do not, so that in (13b), nobody is being asked to work too much. In this sentence we have to explain the LF position of the imperative and the absence of the ADP effect at the same time. I think that one property of conditional conjunctive imperatives is that they have the main features of structures with "unselected binding" (Heim, 1982), such as the so-called "donkey sentences". It is a well-known fact that in the analysis of indefinites proposed by Heim (1982), partly following on from Lewis (1975), indefinites do not have a real quantifying nature, but are represented by variables that are bound by another operator. As regards Spanish, Hernanz (1994) showed that this analysis can be extended to copulatives with infinitives (see also Lebeaux, 1984) such as querer es poder (literally, "to want is to be able", ie. "where there's a will, there's a way"). As these authors convincingly argue, all these structures contain a generic null operator that binds one or several unspecific variables in conditionals, and in other structures with interrogative sentences are grammatical or not with EQs, but rather what do they exactly mean. See section 4 for
274
From Words to Discourse
quantifying adverbs. As Hernanz shows, this same operator also works as an antecedent of PROarb in infinitive copulatives. We may assume that conditional conjunctive imperatives are not real imperatives, but they coincide with generic conditionals and infinitives with PROarb of copulatives in that they are under the scope of such a generic operator.7 It should be remembered that conditional conjunctive imperatives have the same morphological properties as imperatives, but not the same meaning. I think that in any representation that we might give to (13b), the inflection of the false imperative should exhibit what it has in common with a conditional conjunction, that is, the grammatical manifestation of a possible state of affairs in the scope of the initial generic operator. Let us assume that C° is the right place for both units, and also that the morphological reasons mentioned above force the imperative to be in C°. Notice that if conjunctive imperatives stood in their VP projection due to their lack of illocutionary force, they would be lower than NegP, and we could not explain why they cannot be negated. I will assume that they are in C°. Thus, (13b) would have a partial structure close to the one shown in (14), in which GEN.OP is the generic operator: (14)
[GEN.OP...[cp [c° trabajaj [Modp demasiado; [Agrp--[vp [VP [v°ej ] ...e,...]]]]]]] y moriras joven.
As the illocutionary features of the imperative do not appear in C°, there is nothing to prevent the subordinate ModP to occur, so that the EQ can be licensed. Here we have a C° under the scope of a generic operator. C° selects the ModP with the EQ, something fairly similar to the relationship we have in between si ('if), presumably in C°, and the Negative Polarity Item (henceforth, NPI). See fh. 7. Again, the ModP option turns out to be better than the CP option. some more details on this issue. 7 Unselective generic operators are compatible with other unspecific quantifiers such as weak negative polarity items (NPIs) and Fauconnier's (1975) quantificational superlatives. It is interesting that this licensing process occurs both in conditional conjunctive imperatives (ia) and in the copulatives with infinitives studied by Hernanz: (iia), (iiia): (i)
(ii)
(iii)
a. Haz el menor ruido y te mato. "Make the slightest noise and I'll kill you." b. *Haz el menor ruido o te mato. "Make the slightest noise or I'll kill you." c. Si haces el menor ruido te mato. "If you make the slightest noise, I'll kill you." a. Mover un dedo por el es perder el tiempo "Lifting a finger for him is a waste of time." b. *Movi un dedo por ely perdi el tiempo. "I lifted a finger for him and I wasted my time." c. Si mueves un dedo por elperderds el tiempo. "If you lift a finger for him you are wasting your time." a. Hacer el menor ruido es exponerse a ser descubierto. "To make the slightest noise would be to risk being discovered." b. *Hice el menor ruido y me expuse a ser descubierto. "I made the slightest noise and risked being discovered."
Degree Quantification and Modal Operators
275
d) There is a final argument for the ModP option. Notice that given this option, an EQ is expected to create minimality effects if we have a non-referential QP in CP. A way to prove this is to have a quantificational operator in Spec/CP in a inequality comparative.8 The EQ intervenes and the relative operator cannot bind its variable: (15)
a. Vi las mismas peliculas que habia visto mucha gente. "I saw the same movies that many people had watched." b. *Vi las mismas peliculas que habia visto demasiada gente. "I saw the same movies that too many people had watched."
Again, this is unexpected in the "CP option" because that analysis lacks structural space for the two operators that are needed. Needless to say, all these facts do not follow from a possible strictly pragmatic analysis of EQs either, unless some specific local conditions for EQ licensing are provided in such an approach.9
3. EPISTEMIC QUANTIFIERS AND NON-SPECIFICITY EQs do not lead to the interpretative ambiguities of other quantifiers (wide or narrow scope or de dicto-de re interpretations). This is in fact to be expected if they have to be licensed by an operator. Remember that it is usually said that specificity shows a de-dicto/de re ambiguity with intensional predicates. If in Juan busca varies periodicos ('John is looking for several newspapers'), we have the extensional reading, it may be assumed that wide scope is achieved by movement of the quantifier at LF. As usual, the intensional reading may be interpreted with the QP in situ, so that the function denoted by the intensional verb buscar ('to look for') is applied to the NP intension. The interesting point as regards my argument is that EQs do not show ambiguities of this type. There are not two interpretations for Juan busca demasiados periodicos ('John is looking for too many newspapers'), but only the one I have called epistemic reading. In fact, EQs are indefinite quantifiers that reject the specific interpretation. This fact is particularly relevant for my analysis of EQs, and can be proven in several ways:
8
Thanks to L. Saez for suggesting this argument to me. Notice that the very fact that a quantifier is licensed by a modal is standard in other structures. Sp. cualquiera ('any') can only appear in modal contexts (modal verbs, futures, conditionals, intensional predicates, generic operators, etc.). A simple contrast is found in (i):
9
(i)
a. *Cualquiera entro. ('Anyone came in') b. Cualquiera pudo entrar. ('Anyone might have come in')
Interestingly, cualquiera is historically formed out of a relative plus a modal.
276
From Words to Discourse
a) EQs cannot be followed by the modifiers that usually identify specific readings. I refer to postnominal terms such as en particular ('in particular1), or en concrete ('specifically'): (16)
(17)
a. Varios libros en particular. "Several books in particular." b. Un libra en concrete. "A specific book." a. *Demasiados libros en particular. "Too many books in particular." b. *Excesivas personas en concreto. "An excessive number of people specifically."
b) Specific Qs are able to admit what Enc (1991) calls "covert partitive readings". By establishing the specific reading of quantifiers, what we actually do is define a discourse domain in which we identify a subset. Indefinite NPs have no antecedents in discourse, but they introduce elements that have not been previously mentioned. If they are specific, their reference will be established in relation to the group of items to which they belong. Partitive complements select this group and, thus, the specific reading of indefinites: (18)
a. Busco varios libros "I am looking for several books." b. Busco varios de los libros. "I am looking for several of the books."
[Ambiguous: specific/unspecific] [Unambiguous: only specific]
As opposed to quantifiers such as muchos ('many') or varios ('several'), EQs have no partitive complements, which is expected as they are not specific. In other words, they have no discourse domain from which a subset may be selected: ( 1 9)
{Muchos/ varios / *demasiados} de los ninos. "Many/ several/ too many} of the children."
As is known, indefinite quantifiers with partitive complements lead to definiteness effects in existential sentences, which confirms their specific nature. EQs may appear normally in existential contexts, but never with partitives. This is exactly as expected, as they are unspecific. c) EQs cannot identify a list preceded by "namely", for example after (5a), nor can they serve as antecedents for pronouns. Specific terms require that their discourse referents be linked to previously established discourse entities --Pesetsky's (1987) D-linking; see also Cinque (1991), whereas nonspecifics require the absence of this linking process. As expected, EQs reject appositive clauses, since quantifiers that allow them are specific:
Degree Quantification and Modal Operators
(20)
111
a. Mucha gente, que tieneproblemas, va constantemente al psiquiatra. "Many people, which have problems, constantly go to the psychiatrist." b. *Demasiada gente, que tieneproblemas, va constantemente al psiquiatra. "Too many people, which have problems, constantly go to the psychiatrist."
d) Quantifiers with so-called "personal a" only allow for specific readings, as in Vi a muchos ninos ('I saw many children1). The interpretation that we have in Vi a demasiados ninos has to be either "contrastive" or "ironical", which is consistent with the present approach.10 e) There is an interesting difference between subjects with indefinite quantifiers bound by "unselective binding" and subjects with quantifiers that are not inherently indefinite, such as muchos ('many'), cuatro ('four') or varios ('several'). The former, which act as variables, are compatible with non-temporal reading of operators such as siempre ('always'). See Carlson (1979), Heim (1982) and Lewis (1975). The latter, which are not variables, maintain the temporal interpretation of these quantifiers and have to appear in a peripheral position (either "topic" or perhaps IP adjuncts) with specific interpretation, i.e. outside the scope of the adverbial operator. EQs cannot be specific, so that if we make them appear in such a position this reading should be forced and we will obtain an ungrammatical sentence. This is exactly what happens: (21)
a. Un perro siempre mueve el rabo. "A dog always wags its tail." [NON-SPECIFIC: siempre does not have temporal interpretation] b. Muchos perros siempre mueven el rabo. "Many dogs always wag their tail." [SPECIFIC: Siempre has temporal interpretation] c. *Demasiados perros siempre mueven el rabo. "Too many dogs always wag their tail."
That is, \idemasiado could be specific, (21c) would be well-formed. f) EQs are excluded from other LF positions that also require specific elements. The quantifiers that appear in the complements of inequality comparatives have to be specific. This may be due to the fact that this complement acts as a barrier for extraction, as occurs in overt syntax, or perhaps a higher operator exists that cannot bind a variable that is located in the comparative coda. In any case, the nonspecific nature of demasiado excludes this EQ from comparative codas, which is to be expected: (22)
a. Juan ve mas television que muchas personas. "John watches more television than many people."
Thanks to V. Demonte for pointing to me this effect with personal a.
[Muchas is specific]
278
From Words to Discourse
b. *Juan ve mas television que demasiadas personas. "John watches more television than too many people." Other tests for (non)specificity give similar results.11 In short, we have seen that demasiado is different from other quantifiers such as varios or muchos12 or from numerals in which it does not admit the specific reading. It is reasonable to assume that these other quantifiers do admit it because they do not necessarily have to be under the scope of a modal projection. Such reading is allowed because they may have antecedents in a certain domain of discourse. As we have seen, EQs do not denote sets of objects that can be identified in discourse, but rather sets of objects that only exist when they are modally validated. In other words, there is no reason to believe that EQs should be unspecific if no operator existed under whose scope they must become licensed.
4. NEGATION AND QUESTIONS 4.1. The behavior of EQs in negative sentences is particularly interesting because negation cancels out the ADP effect. That is, we can have EQs with negated directive predicates naturally: (23)
a. No corran demasiado. "Don't run too much." b. No trabajen excesivamente. "Don't work too much."
1 ' For example, the ones related to focus. Notice that if we make purpose clauses to be the focus of negation, quantifiers within the VP will not be licensed by negation and will be interpreted as specific. Demasiado is rejected in these contexts, since the specific reading is forced:
(i)
a. No he comprado varios libros para que los destroces. "I haven't bought several books for you to break them up." b. *No he comprado demasiados libros para que los destroces. "I haven't bought too many books for you to break them up.".
In a similar way, quantifiers under the focus of solo ('only') are specific, and demasiado is rejected in this structure, again as expected: (iij
Solo aparecieron par alii {varios/ *demasiados} amigos tuyos. "Only {several/ too many} friends of yours came over."
12 The case ofmucho(s) ('many') deserves a more detailed study as it behaves like an EQ sometimes and not others. It is clearly different from demasiado in that it admits the specific reading, but I cannot at the moment explain why it appears to be close to EQs in a number of cases.
Degree Quantification and Modal Operators
279
As we have seen, because EQs are not specific, it is not surprising that they fall under the scope of negation if the proper local conditions are met. In other words, (24a) is ambiguous: (24)
a. No he leido muchas novelas de ciencia fiction. "I haven't read many science fiction novels." [Ambiguous: wide or narrow scope] b. No he leido demasiadas novelas de ciencia fiction. "I haven't read too many science fiction novels." [Unambiguous: narrow scope]
Demasiadas in (24b) has lower scope than negation, even if one thinks that the EQ might be outside the scope of negation and inside the scope of the epistemic quantifier. This suggests that there are reasons to suppose that NegP is the real licenser of the EQ in (23), and that LF movement of the EQ does not take place to Spec/ModP but rather to the closer Spec/NegP in these cases. My point will be that a potential focus of negation turns out to be a real one for reasons of economy. Here are some arguments for this interpretation: a) We may first recall that imperative operators license non-specific indefinite quantifiers, but negated imperatives cannot do so. That is, a reasonable explanation for the contrast between Di algo ('Say something') and *No digas algo ('Don't say something') is the fact that the failure of the closer NegP to license algo cannot be "repaired" by the modal projection corresponding to the imperative. That is, the quantifier cannot be specific outside NegP and under the scope of the imperative, nor can it be non-specific inside NegP, since it is not a NPI. Consequently, it cannot be interpreted. (As it is well-known, restrictive relative clauses change the referential interpretation of their NP, but I will not deal with this factor here). In the case being considered, the NegP is preventing algo to be licensed under the scope of the higher imperative operator, which is somehow the reverse of what we have in (23). b) Other neg-triggers license EQs in directive contexts. The following contrast is naturally explained is we assume that the negative projection that sin heads hosts the EQ phrase, but it would remain without explanation otherwise: (25)
a. *Viajen con demasiado dinero. "Travel with too much money." b. Viajen sin demasiado dinero. "Travel without too much money."
There is no higher epistemic modal in these sentences, but even if there were, it could not license the EQ, just as no does not license nada in No viaje sin nada ('I didn't travel without anything'). That is, a closer potential licenser becomes available and turns out to be the right licenser. In general, the relationship between the EQ and negation has to be local, in the same way as the relationship between negative polarity items and their head. As several authors have pointed out (Zanuttini, 1991; Longobardi, 1991; and Acquaviva, 1992 for Italian; Bosque, 1992 for Spanish), NPIs move at LF to a position in which they can enter into an agreement relation with
280
From Words to Discourse
the Neg° head. In the same way that it is not possible to have an NPI in (26a), neither can we have in (26b) the reading in which demasiadas ('too many') falls under the scope of negation: (26)
a. *Juan no dijo que queria nada. "Juan didn't say he wanted anything." b. Juan no dijo que queria demasiadas cosas. "John didn't say he wanted too many things."
That is, the EQ licenser in (26b) is the ModP in the subordinate clause, and not the negative projection. c) Expletive negation does not license NPIs (see Espinal, 1991b). Interestingly, it does not license EQs either. In (Que hambre no tendria yo! ('I sure was hungry'), negation is expletive; no negative subject quantifier is possible (*(Que hambre no tendria nadie!) and EQs are rejected (*(Que hambre no tendria demasiada gente!). This supports the idea that EQs become close to NPIs under negation. d) We saw above that answers with demasiado or with excesivamente are ironic or noninformative. On the other hand, answers with no demasiado ('not too much') or no excesivamente ('not excessively'), are not ironic but quite regular, in the sense that they provide "information" rather than "assessment". Again, this is not surprising, as it is not an epistemic operator, but a negative one, that hosts the degree quantifier in these cases.
4.2. EQs may appear in questions, but they do so in conditions that really make them invalid as real questions, if we take this to mean that they request new information in a neutral manner. EQs in interrogative sentences do not lead to questions that openly request new information, but rather to typically D-linked confirmative questions, in Pesetsky's (1987) sense. EQs in yes/no questions lead either to echo readings, or to a polarized rhetorical reading, such as the one forced by the word acaso ('Perchance, Isn't it true that...?'), es que ('Is it true that...?, Do you really...?'), or an ordinary tag question. The contrast between the (a) sentences —with no EQ— and the (b) sentences —with an EQ— illustrates this: (27)
(28)
a. ^Ha venido mucha gente a tu fiesta? "Have many people come to your party." b. ^Ha venido demasiada gente a tu fiesta? "Have too many people come to your party?" a. ^Has comido bastante? "Have you eaten enough?"
Degree Quantification and Modal Operators
281
b. ^Has comido demasiado? "Have you eaten too much?" Compared to the (a) sentences, the (b) sentences could not be directly used in a neutral way to request information at first. Besides the echo reading, the (b) sentences are polarized questions in the sense already described. An approximate paraphrase of (28b) would be: lAcaso has comido demasiado? ('Have you perhaps (perchance) eaten too much?'), or Has comido demasiado, ^no? ('You've eaten too much, haven't you?'). The syntax of echo questions and, in general, in situ w/z-questions, has been the subject of a great deal of attention over the past few years. Pesetsky (1987) proposed that D-linked in situ w/z-phrases do not move at LF, but instead are unselectively bound by an operator in COMP (an idea that Pesetsky took up from Baker 1970). Sobin (1990) and Domitrescu (1990) extend this idea to w/z-sentences in echo quesions, so that Baker's Q-morpheme performs a very similar role to that of an echo operator. Wh-words in echo questions are typical examples of D-linked items, in the sense that possible replies are restricted to those offered by the speech immediately before. In Domitrescu's analysis (1990, 1992), the highest Spec/CP contains an unselective binding operator in Baker's sense, a sort of quotational w/z-word, and the corresponding C° takes another CP as a complement, the real interrogative clause. As regards EQs in questions, it should be taken into account that a rhetorical question is not actually an interrogation with wh- in situ-, nor it is an echo question, although it does contain some features of both. I would like to suggest that the explanation for EQ behaviour in interrogative sentences is not much different from that which occurs in negatives. Progovac (1993) has remarked that NPIs may be induced by a question, but interestingly enough, this only occurs if the question is rhetorical. That is to say that w/z-words trigger NPIs, but in such cases there is no operator that ranges over different possibilities, but rather one that points out to an empty set of options. In other words, sentences such as Who did Mary ever kiss at the first date? (Progovac's example), can only be interpreted as a rhetorical question. This sentence really means "Mary never kissed anyone at the first date". In Progovac's analysis, the rhetorical interpretation is obtained by treating the w/z-word as a general variable which can be bound by the negation: the polarity operator (which she places in COMP), is merged with the w/z-word, producing a universal negative quantifier. There is some relationship between these facts and those that I am examining here. In a way close to how negation forms complex lexical units with EQs and becomes their licencer (no demasiado 'not too much', no excesivo 'not excessive') we may think that the null operator that is equivalent to acaso ('perchance') performs the same role in questions. So, complex units are formed of a type close to acaso demasiado, acaso excesivo, etc. Acaso can be analysed as the disjunctive operator in rhetorical yes/no questions, and does in fact possess the features of an epistemic marker; i.e. a truth conditional switch operator that converts the disjunctive question into a confirmative one. The result of this conversion is known as a "rhetorical question". It is
282
From Words to Discourse
important to remember, therefore, that the disjunctive variable represented by o no ('or not') in real yes/no questions cannot be bound in such situations, as they are not real yes/no questions: (29)
a. £ Acaso vino Juan? "Didn't Juan come?" b. *£ Acaso vino Juan o no? "Didn't Juan come or not?"
Let us assume, then, that a switch null operator licences EQs in rhetorical yes/no questions. As to the position it should occupy there are at least two possibilities. One is to assume that the illocutionary features of the questions imply that C° does not select ModP. As a result, the presence in Spec/CP of acaso, or the operator representing it, would be what enables EQ to be licensed. The second option is to assume that, as the disjunctive variable cannot be licensed, this is not a real yes/no question. Thus, in such a situation, the epistemic ModP can be selected and the operator represented by acaso would appear in that projection and not in CP. In fact, Culicover (1992), Uriagereka (1995) and Poletto (1993) argue for movement of V° to the respective heads of the "Focus" and "Modal" Phrase that they propose. There is one reason that makes me think that this option is preferable: EQs in indirect questions are interpreted in the same way as in direct ones, but the rhetorical operator overtly appears in the syntax below si ('if), and, as is usually accepted, si occupies the C° position: (30)
Averigua si (acaso) corre demasiado. "See if (perchance) he is running too much."
It would be possible to extend this analysis to wh- questions, as the same switch operator enables them to be turned into yes/no questions (Quien = acaso alguien, or Who = perchance someone), but I will not undertake this task here. In general, I think that one should make a connection between the behavior peculiar to degree quantifiers in negative contexts and in questions. The behavior of EQs in these structures does not cancel out the basic hypothesis of this paper, as both negation and the switch marker (either null or explicit) in rhetorical questions represent appropiate operators for EQ licensing.
5. EQS AND MODALS As I mentioned in section two, EQs can be licenced in complement and relative subordinate clauses, which is enough to consider that the ModP does not necessarily make an "utterance" phrase in the strictest sense of the term. All EQs, however, have to be licensed in a clausal
Degree Quantification and Modal Operators
283
domain.13 We can thus explain why imperatives that appear in the main clause do not lead to ADP effects if the indicative subordinates contain EQs, as we saw in (lOb). Things are more difficult in the case of subjunctive complements. In some sense, this is not entirely surprising, since, as has often been stressed, the subcategorized proposition in the subjunctive mood maintains a temporal and/or modal dependence as regards the main predicate. Let us look at a typical case: that of subjunctives under verbs of will. A sentence like (31) is ungrammatical in one sense and grammatical in another. (31)
Quiero que Juan trabaje demasiado. "I want John to work too much."
This sentence must be excluded in the interpretation in which the EQ is licensed in the subordinate clause, but the sentence is grammatical if it does so in the ModP of the main clause. In the latter case, (31) would mean "I realize that the amount in which I want John to work is excessive". That is, in this reading the speaker is not saying what s/he wants, but rather what s/he believes. In the present analysis, this interpretation is obtained if the EQ raises at LF to the ModP in the main clause across the subordinate in the subjunctive, in a fairly close way to that in which w/z-words do so in similar contexts, or to the way in which a quantifier in a complement clause falls under the scope of a NegP in the main clause: (32)
[Cp [Modp DemasiadOj [M- [M° [IP Quiero [CP que Juan trabaja e; ]]]]]
Let us now take the option that there is no ModP in the main clause. Then, the EQ would remain unlicensed and the sentence would be ungrammatical, as in (33): (33)
*Pidele que trabaje demasiado. "Ask him to work too much."
In the interpretation excluded for (31), the main predicate acquires illocutionary force by V° to M° movement, so that the epistemic ModP projection is not selected (or else is not active). The sentence is, therefore, ungrammatical, as the EQ variable is not bound. We may therefore explain why the EQ is licensed in the main clause, but an interesting question now arises: Why should a subordinate subjunctive prevent the modal validation of an EQ, whereas an indicative subordinate is not doing so? I will suggest that in order to exclude EQ licensing in the subordinate clauses of (31) and (33) we have to interpret subjunctives of will as root modals, as Kempchinsky (1986) does. Since the modal of the subjunctive clause is a deontic one in these sentences, the EQ cannot be licensed therein. Therefore, either the EQ escapes in search of a 13 An interesting question arises as to whether EQ licensing is possible in eventive NPs. The answer is probably yes, specially if modal adjectives, such as probable, are involved. I will not analize these constructions here.
284
From Words to Discourse
higher ModP in the main clause, or else (if it can not find it) it remains unlicensed and the sentence is ungrammatical.14 The syntactic position of epistemic and deontic modals within the framework of the theory of functional projections is a highly complex issue. There appears to be some coincidence on the fact that epistemic modals are generated in a higher position that root ones. Picallo (1990) suggested that epistemic modals are generated as constituents of INFL, whereas root modals are VP adjuncts. She argued that the ungrammaticality of sentences with epistemic modals in Catalan indirect infinitive interrogatives (Ignordvem com deure expresar-li 'We didn't know how we should have expressed it') is due to the PRO theorem because if epistemic modals are in INFL we would have a governed PRO. McDowell (1987) claimed that this kind of incompatibility is due to the fact that epistemic modals coincide with w/z-sentences in that they are truth conditional operators that move to C° at LF to bind a variable. In fact, epistemic modals are incompatible with yes/no questions, as Coates (1983), among other authors, observed. The incompatibility of questions with epistemic modals would thus be, McDowell argues, a result of the Bijection Principle. Absorption, in the sense used by Higginbotham and May (1981), cannot occur in these cases because the w/z-operator ranges over arguments and the epistemic operator ranges over propositions. McDowell's interpretation of epistemic modals as truth conditional operators that move to C° at LF is not much different from the one I am suggesting here by placing them at ModP. As we have seen, there are reasons to assume that CP and ModP are different projections. I would like to recall that there is a certain tradition, well summarized by Lyons (1975: 793, chap. 17), regarding the idea of associating the prospective properties of deontic modals with those of directive speech acts, and the stative properties of statements and other notions related to the prepositional attitude to epistemic modals. Epistemic modals mostly behave like prepositional attitude predicates because they are, as McDowell argues, "hedged assertions". The epistemic modality of possibility is thus interpreted as a subjective qualification that the speaker makes regarding his commitment to the possibility of the proposition being true in terms of its own uncertainty. In this sense, I would like to recall the very well known fact that notions attributed to deontic modals (obligation, ability, permission, etc.), are inherently potentially prospective notions. As Lyons states, they acquire directive interpretation because they implicitly possess a "because-I-say-so" component. Let us look at the three interpretations of (34) from this perspective: (34)
Juan debe trabajar demasiado. "John must/should work too much."
14 An expected consequence is the fact that verbs selecting for both moods (indicative as verbs of assertion, and subjunctive as verbs of influence), reject EQs in the subjunctive complement:
(i)
Insisto en que {trabajas/*trabajes} demasiado. "I insist that you work too much."
Degree Quantification and Modal Operators
285
1) The first reading is the ungrammatical interpretation: deber ('must') is a deontic modal and the sentence has illocutionary force because it has a directive interpretation. The EQs variable cannot be bound either by the modal deber because it is deontic, or by a higher ModP, because it is excluded by illocutionary features. In other words, John's bad-tempered boss cannot use (34) as a way of expressing his/her formal decision to load him unfairly with work. 2) The second reading is grammatical: deber is an epistemic modal and is able to host the EQ in the projection that it heads.15 This is the most natural interpretation. 3) The third reading is also grammatical: deber is a deontic modal, the EQ does not accept it, but looks for the highest ModP because there is no other illocutionary operator to exclude it, as in (32). The sentence would mean, roughly: "I think that the amount in which John has to work is excessive". These three are the desirable interpretations. I would like to add that an essential part of the differences that arise with these interpretations lies in the nature of the tense that modal notions select. More specifically, I think that the non-appearance of EQs in deontic contexts should be backed by the idea that deontic modals coincide with directive speech acts in that they select prospective tenses. In this case, what would really be incompatible in deontic modals is the verb tense of prospective subjunctives and the one required by the epistemic ModP. In a sense, in the same way as English modals contain features of "inherent agreement", it may be said that modals in Romance languages have "inherent tenses". We could then suppose that the main difference between the "deontic subjunctive" and the "epistemic subjunctive" is related to this difference. This idea links up with the classic remark that subjunctives have no specific forms for future tense, even when they refer to prospective situations. An old observation is the fact that the form trabaje ('works [subj]') in (35a) is ambiguous between present and future interpretations. (35)
a. Es posible que Juan trabaje. "It is possible that John is working/works." b. Es posible que Juan trabaje demasiado. "It is possible that John is working too much."
Notice now that trabaje does not have two meanings in (35b), but only the present tense interpretation. That is, the deontic interpretation of the subjunctive mood is incompatible with EQs, but the epistemic interpretation of subjunctives is not.16 EQs are only compatible with the present tense that corresponds to their epistemic meaning. 15 Notice that epistemic adverbs such as quizds ('perhaps'), or tal vez ('maybe'), reject imperatives and made constituents with EQs, which leads us to suppose that maybe they become EQ licensers. This is an interesting issue, but I shall not address it here. 16 See Stowell (1982) for a brief but interesting analysis of infinitives as clauses containing tense operators selected by main predicates. An extension of the "subjunctives as deontic modals" analysis seems to be necessary along these lines in order to exclude EQs in infinitive complements of prospective interpretation, as in (i), and also, I would add, subordinate sentences with future tenses. When these sentences are complements of verbs with illocutionary readings, no access to a higher ModP projection is possible and, again, the result is ungrammatical:
286
From Words to Discourse
6. CONCLUSION The grammar of intuitive quantificational notions such as "excess", "extreme degree" and "appraisal" is a complex phenomenon. I have attempted to show that excess is not to be seen as the identification of a final extent on an extensional scale, but as a modal concept. As such, it introduces an intensional context, i.e.: a reference framework by which the quantified elements are filtered through a prepositional assessment. I have tried to provide a syntactic translation of this idea by proposing an epistemic modal projection between CP and AgrP. This projection bears a clear relation to some focus positions that have been proposed in a similar structural space. I have also attempted to show that this ModP has a branching structure similar to the one found in other functional projections (an idea for which not much evidence has been presented in the literature). EQs' movement at LF to the specifier of ModP is restricted by independent principles of syntax, in such a way that the contexts in which EQs are not appropriate are deduced from this interaction rather than having to be presented independently. See Bosque (2001) for an extension of some of these ideas to the grammar of preposed adjectives and morphological supperlatives in Spanish.
REFERENCES Acquaviva, P. 1992. "The Representation of Negative 'Quantifiers'." Rivista di Linguistica 4. 319-385. Baker M. 1970. "Notes on the Descriptions of English Questions: The Role of an Abstract Question Morpheme". Foundations of Language 6. 197-219. Banfield A. 1982. Unspeakable Sentences. Boston and London: Routledge & Keegan. Bolinger, D. 1972. Degree Words. The Hague: Mouton. Bolinger, D. 1967. "The Imperative in English." In To Honor Roman Jakobson. The Hague: Mouton, 335-362. Bosque, I. 1990. "Las Bases Gramaticales de la Alternancia Modal. Repaso y Balance." In I. Bosque (ed.) Indicativo ySubjuntivo. Madrid: Taurus, 13-65. Bosque, I. 1994. "La Negation y el Principio de las Categorias Vacias." In V. Demonte (ed.) Gramatica del Espanol. Publicaciones de la Nueva Revista de Filologia. Mexico vol. VI, 167-200. Bosque, I. 1996. "On Degree Quantification and Modal Structures." In C. Parodi et al. (eds.) Aspects of Romance Linguistics. Washington, D.C.: Georgetown University Press, 87-106. Bosque, I. 2001. "Adjective Position and the Interpretation of Indefinites." In J. Gutierrez and L. Silva-Villar (eds.), Current Studies in Spanish Syntax and Semantics. Berlin/New York: Mouton-De Gruyter, 17-38. Bosque, I. and J.M. Brucart 1991. "QP Raising in Spanish Superlatives". Ms.Complutense University of Madrid and Autonomous University of Barcelona. Cardinaletti, A. and I. Roberts. 1991. "Clause Structure and X-second." Ms., Univ. of Venice and Univ. of Geneva.
(i) (ii)
*Teprometo formalmente trabajar demasiado. "I hereby promise you to work too much." *Te prometo formalmente que trabajare demasiado. "I hereby promise you that I shall work too much."
Degree Quantification and Modal Operators
287
Carel, M. 1992. "Trap: Enchainements Positifs et Enchainements Negatifs." Chap. 6 of Vers une Formalisation de la Theorie de {'Argumentation dans la Lange. PhD diss. Paris, Ecole des Hautes Etudes en Sciences Sociales. Carlson, G. 1979. "Generics and Atemporal When." Linguistics and Philosophy 3. 49-98. Cinque, G. 1991. Types of A' Dependencies. Cambridge: MIT Press. Clark, B. 1993. "Relevance and 'Pseudo-Imperatives." Linguistics and Philosophy 16. 79-121. Clark, H. C. and R. Guerrig 1990. "Quotations as Demonstrations." Language 66. 764-805. Coates, J. 1983. The Semantics of the Modal Auxiliaries. London: Croom Helm. Coulmas, F. (ed.) 1986. Direct and Indirect Speech. Berlin: Mouton/Walter de Gruyter. Culicover, P. 1992. "Topicalization, Inversion and Complementation in English." Ms., Ohio State University. David, J. and G. Kleiber 1983. La Notion Semantico-Logique de Modalite. Paris: Kliencksieck. Davidson, D. 1988. "On Saying That". Inlnquiries into Truth and interpretation. Oxford: Clarendon Press, 1988. Davis, E. S. 1979. "Some Restrictions on Conditional Imperatives." Linguistics 17. 1034-105. Dobrovie-Sorin, C. 1984. "Imperatifs Conditionnels." Le Francois Moderne 52. 1-21. Domirrescu, D, 1990. The Grammar of Echo Questions in Spanish and Romanian: Syntax, Semantics, Pragmatics. Ph.D. diss. University of Southern California. Domirrescu, D. 1992. "Spanish Echo Questions and their Relevance for Current Syntactic Theory." Southwest Journal of Linguistics 10. 42-65. En9, M. 1991. "The Semantics of Specificity." Linguistic Inquiry 22. 1-25. Espinal, M.T. 1991a. "The Representation of Disjunct Const'tuents." Language 67. 727-762. Espinal, T. 1991b. "On Expletive Negation. Some Remarks with Regard to Catalan." Linguisticae Investigationes 15. 41-65. Farkas, D. 1992. "Two Types of'World-Creating' Predicates." In D. Brentari et al. (eds.) The Joy of Gramar. A Festschrift in Honor of James D. McCawley. Amsterdam: John Benjamims, 35-63. Fauconnier, G. 1975. "Pragmatic Scales and Logical Structure." Linguistic Inquiry 6. 353-375. Hall Partee, B. 1973. "The Semantics of Belief Sentences." In J. Hintikka et. al. (eds.), Approaches to Natural Language. Dordrecht: Reidel, 309-336. Heim, I. 1982. The Semantics of Definite and Indefinite Noun Phrases. Ph.D. Diss. U Mass., Amherst. Heim, I. 1992. "Presupposition Projection and the Semantics of Attitude Verbs." Journal of Semantics 9. 183-221. Heny, L. and B. Richards (eds.) 1983. Linguistic Categories: Auxiliaries and Related Puzzles.. Dordrecht: Reidel. Hernanz, M.L. 1994. "Argumentos Implicitos, Operadores Nulos e Interpretacion Arbitraria: el Caso de de los Infinitives Pseudoecuativos." In V. Demonte (ed.) Gramdtica del Espanol. Publicaciones de la Nueva Revista de Filologia Mexico vol.VI, 315-362. Higginbotham, J. and R. May 1981. "Questions, Quantifiers and Crossing." The Linguistic Review 1. 41-79. Hiz, H. (ed.) 1978. Questions. Dordrecht: Reidel. Jackendoff, R. 1972. Semantic Interpretation in Generative Grammar. Cambridge: MIT Press. Jackendoff, R. 1985. "On Belief Contexts." Linguistic Inquiry 6. 59-93. Jayez, J. 1985. "Trop: L'Exces par Default." Le Franfais Moderne 53. 22-48. Kempchinsky, P. 1986. Romance Subjunctives and Logical Form. PhD diss. UCLA. Koktova, E. 1986. Sentence Adverbials in a Functional Description. Amsterdam: John Benjamins, 1986. Laka, I. 1990. Negation in Syntax: On the Nature of Functional Categories and Projections. PhD diss. MIT. Lebeaux, D. 1984. "Anaphoric Binding and the Definition of PRO." In Proceedings ofNELS 14. 253-274. Leech, G. 1971. Meaning and the English Verb. London: Longman. Lewis, D. 1975. "Adverbs of Quantification." In E. Keenan (ed.), Formal Semantics of Natural Language, Cambridge: Cambridge University Press, 3-15. Longobardi, G. 1991. "In Defense of the Correspondence Hypothesis: Island Effects and Parasitic Constructions in Logical Form." In J. Huang and R. May (eds.), Logical Structure and Linguistic Theory. Reidel: Dordrecht: 149-196.
288
From Words to Discourse
Lyons, J. 1975. Semantics. Cambridge: Cambridge University Press. Vol. 2. Maldonado, C. 1991. Discurso Directo y Discurso Indirecto. Madrid: Taurus. May, R. 1977. The Grammar of Quantification. Ph.D. diss. MIT. May, R. 1985. Logical Form, Its Structure and Derivation. Cambridge: MIT Press. McDowell, J.P. 1987. Assertion and Modality. Ph.D. diss. University of Southern California. Milsark, G. 1974. Existential Sentences in English. Ph.D. diss. MIT . Palmer, F. R. 1979. Modality and the English Modals. London: Longman. Palmer. F. R. 1986. Mood and Modality. Cambridge: Cambridge University Press. Parrel, H. (ed.) 1983. On Believing/De la Croyance. Berlin: Walter de Gruyter. Pesetsky, D. 1987. "Wh-in. Situ: Movement and Unselective Binding." In E. J. Reuland and A. . B. ter Meulen (eds.), The Representation oflndefiniteness. Cambridge: MIT Press, 98-129. Picallo, C. 1990. "Modal Verbs in Catalan." Natural Language & Linguistic Theory 8. 285-312. Picallo, C. 1991. "Nominals and Nominalizations in Catalan." Probus 3. 279-316. Poletto, Cecilia 1993. "Above AGRs." Presented at the WSCRL, University of Barcelona. Pollock, J.-Y. 1993. "Notes on Clause Structure." Unpublished paper, University of Picardie, Amiens. Progovac, L. 1993. "Negative Polarity: Entailment and Binding." Natural Language & Linguistic Theory 16. 149180. Richard, M. 1986. "Quotation, Grammar and Opacity." Linguistics & Philosophy 9. 383-403. Rivero, M.L. 1991. "Long Head Movement and Negation: Serbo-Croatian vs. Slocak and Czech." The Linguistic Review*. 319-351. Rivero, M.L. 1994a. "The Structure of IP and V-Movement in the Languages of the Balkans." Natural Language & Linguistic Theory 12. 63-120. Rivero, M.L. 1994b. "Negation, Imperatives and Wackernagel Effects." Rivista di Linguistica 6. 39-66. Rooryck, J. 1992. "Clitic Ordering in Romance Imperatives: Restricting Relativized Minimality." Presented at the 22nd Linguistic Symposium on Romance Languages, El Paso Texas. Rowlett, P. 1993. "On the Syntactic Derivation of Negative Sentence Adverbials." French Language Studies 3. 3969. Sobin, N. 1990. "On the Syntax of Echo Questions." Lingua 81. 141-167. Stowell, T. 1982. "The Tense of Infinitives". Linguistic Inquiry 13. 561-570. Stucky, S. 1987. "Configurational Variation in English. A Study of Extraposition and Related Matters." In G. Huck and A. Ojeda (eds.), Discontinuos Constituents. Syntax and Semantics. Academic Press, vol. 20. 377-404. Suner, M. 1993. "NPIs, Island Effects, and Resumptive No." Ms. Cornell University. Terzi, Arhonto 1991. "PRO and Obviation in Modern Greek Subjunctives." In The Proceedings of the 10th West Coast Conference in Formal Linguistics. Standford University, 471-482. Torrego, E. 1996. "On Experiencer and Raising Verbs." In R. Freidin (ed.), Current Issues in Comparative Grammar. Dordrecht: Kluwer, 101-120. Uriagereka, J. 1995. "An F Position in Western Romance." In K. Kiss (ed.), Discourse Configurational Languages, Oxford: Oxford University Press, 153-175. Zanuttini, I. 1991. Syntactic Properties of Sentential Negation: A Comparative Study of Romance Languages. University of Pennsylvania, The Institute for Reserch in Cognitive Science, IRCS Report 91-26. Zubizarreta, M.L. 1999. "Word Order in Spanish and the Nature of Nominative Case." In K. Johnson and I. Roberts (eds.) Beyond Principles and Parameters. Dordrecht:.Kluwer, 223-250.
289
14
THE SHIFTED READING OF THE SPANISH PAST TENSES AS DEPENDENT ON PRESUPPOSITION ACCOMMODATION Rosa J. Garcia Cordova, Universidad Autonoma Metropolitana, Mexico
1. INTRODUCTION The subject matter of this paper is the relationship between (A) the shifted reading of the indicative simple past tenses in Mexican Spanish, that is, the preterito or perfective past tense, and the copreterito or imperfective past tense, and (B) the possibility of presupposing a verbal complement clause containing those tenses.' En? (1987) identified two readings of the English Past tense in embedded clauses: the shifted reading in (1), and the simultaneous reading, which is the most salient one for (2): (1)
a. Mary said that she visited John. b. Maria dijo que visito a Juan. Maria say-PRV2 that visit-PRV to Juan "Maria said that she visited Juan."
1 1 want to thank some of the linguists with whom I have had the chance to discuss this topic over the years: Molly Diesing, Jose Lema, Pilar Pinar and Estela Trevino, They have made crucial suggestions or comments, but all errors in what follows are due exclusively to me.
2
Henceforth PRV stands for "perfective past tense" and IPRV for "imperfective past tense".
290
(2)
From Words to Discourse
c. visit say ST a. Mary said that she was pregnant. b. Maria dijo que estaba embarazada. Maria say-PRV that be-IPRV pregnant "Maria said that she was pregnant." c. say ST ... be pregnant ...
As one can infer from (2c), my view of the so called "simultaneous reading" does not really involve simultaneity of the two intervals, but rather Bello's (1841) idea that one of the two intervals is included in the other one. Similar proposals have been developed within Formal Semantics to account for the English progressive. The treatments are similar because both objects of inquiry share a trait: they are imperfective. Aspectual notions are crucial in determining whether a tense will receive a shifted or a simultaneous reading. En? pointed out that the simultaneous reading is excluded when the embedded predicate is eventive. In Spanish it is grammatical aspect, and not aktionsart, that determines the availability of the simultaneous reading: the copreterito allows it, thepreterito does not. Stative sentences in English and Spanish sentences in the copreterito share the characteristic that allows that reading. Now let us consider further data: (3)
(4)
a. * Crei/pense/imagine que Maria te ayudo. I think/imagine-PRV that Maria you help-PRV b. Mencione/supe/dije que Maria te ayudo. I mention/learn/say-PRV that Maria you help-PRV "I mentioned/learnt/said that Maria helped you." (* simultaneous reading ; ^ shifted reading) a. Crei/pense/imagine que Maria te ayudaba. I think/imagine-PRV that Maria you help-IPRV "I thought/imagined that Maria helped/was helping you." (* shifted reading: thought- a minute ago, help- before her accident; •S simultaneous reading) b. Mencione/supe/dije que Maria te ayudaba. I mention/learn/say-PRV that Maria you help-IPRV "I mentioned/learnt/said that Maria helped/was helping you." (/ simultaneous reading; S shifted reading)
In (3) and (4), if the verb in the embedding clause corresponds to an anti-factive predicate, the shifted reading is excluded for both the preterito and the copreterito. Related remarks were first
The Shifted Reading of the Spanish Past Tenses
291
made about English in Costa's (1972) paper. The Spanish data allows us to see that anti-factivity interacts with aspectuality. Thus (3a) is excluded since the perfective preterito never allows a simultaneous reading, which is the only one possible under anti-factives. The embedded copreterito in (4a) is not ungrammatical, but it has lost its shifted reading. Our account of the above facts will involve providing LF structures for the preterito and copreterito in the style of Stowell (1993), as well as proposing two different LF positions for complement copreterito clauses, corresponding to the shifted and the simultaneous readings respectively. The shifted reading arises through extraction of the complement clause, the simultaneous one through in situ interpretation.3
2. THE SIMULTANEOUS READING The simultaneous reading will be analyzed by means of a syntactic coindexation mechanism.4 In order to provide our structures for the copreterito and the preterito, we will start from Stowell's proposal for English past tense clauses, which we see in (5): (5)
TP ZP T Speech /^\ Time T ZP Past /^ "after" Z{ VP
Stowell decomposes the traditional category Tense. His tenses are temporal ordering predicates, which take as arguments interval-denoting phrases called ZP's. The external argument ZP represents the reference time, whereas the internal argument denotes the time of the eventuality and binds a 3 This treatment is similar to Abusch's (1997) extensional analysis of tenses. However, for the Spanish Past tenses, that kind of analysis need not be rejected in favor of her intensional one as she claims it is necessary for English, because in the crucial contexts the Spanish past tenses are replaced by their subjunctive counterparts. 4
A coindexation approach is suggested in De Swart (1997).
292
From Words to Discourse
variable within VP representing the event argument. I claim the copreterito contains a category Asp, which is the counterpart in the temporal field of a Det in the nominal field. Compare (6) with (7): (6)
IP DP Det
(7) VP
XP
AspP VP
NP
Asp
every dog
TP
Imperfective T
ZP
Past The structure for the Spanish tenses, unlike Stowell's, provides us with a constituent, TP, whose semantic interpretation corresponds to the temporal predicate and the reference time argument, but not the eventuality time. This allows us to account for the simultaneous reading of past under past sentences with a simple coindexation mechanism, as shown in (8): (8)
a. Maria dijo que estaba embarazada. Maria say-PRV that be-IPRV pregnant "Maria said that she was pregnant." b. Asp/V-Pl AspPlj
VP1
Speech Time, Aspl Perfective Tl Past
VI
ZP1 VI
Asp/V-P2
e k say AspP2j
VP2
Asp2 TP2j be pregnant Imperfective /^\ T2 ZP2 I I Past ek I assume that TP, but not AspP, may undergo a form of LF coindexation provided that it has a ccommanding antecedent and that both the antecedent and the coindexed TP dominate a Past T head.5 5 1 will not commit as to the binding theory status of TP, given that temporal categories have a totally different distribution from nominal categories.
The Shifted Reading of the Spanish Past Tenses
293
I also assume that the denotation of ZP, which Stowell equates with PRO, varies as follows: In most matrix contexts it will denote the Speech Time, but in narrative or historical texts, it denotes an interval introduced previously in the text. In complement clauses it will undergo control (Stowell (1993), but if a ZP is dominated by a TP which is coindexed with another TP, that ZP will be coindexed with the ZP in the antecedent TP, and thus share its referent, as in (8). Notice that the category dominating AspP and VP is a mixed category with features from both of its daughters. Henceforth by "AspP" I will refer to the highest projection with pure Asp features. This analysis avoids problems of previous treatments, such as the need to introduce a deletion operation, as in Ogihara (1989,1996). Also unlike Stowell (1993), it accounts for the intuition that Past always means the same thing, without having to disregard the intuition that Past always means Past, for our structure for Past tense clauses will be the same as in our analysis of the shifted reading, to which we now turn.
3. THE SHIFTED READING With Abusch (1988, 1997) and Ogihara, I claim that the shifted reading is dependent on an LF movement operation or its equivalent in other frameworks. In our view, this movement is due to presupposition accommodation. We will adopt Diesing's (1990) Mapping Hypothesis, which establishes that the material dominated by VP in the syntax will correspond to the nuclear scope in a Heimian semantic representation, whereas syntactic material located above the VP will correspond to the restrictive clause in the semantic representation. Syntactic constituents corresponding to presupposed material will be found outside the VP at LF. Berman (1991) and Pinar (1995) have implemented these ideas in the analysis of verbal complement clauses. In what follows I will adopt Pinar's analysis for declarative complement clauses. Pinar proposes that pragmatically presupposed declarative verbal complements undergo a process of presuppositional raising. This movement places them at LF within the embedding clause, in the Specifier position of a phrase called Referential Phrase, or RefP, which was proposed by Beghelli and Stowell (1997) as the highest projection dominated by CP. Thus (9), in which the complement clause receives the shifted reading, will have and LF in which the complement clause raises to the higher RefP. (9)
Maria dijo que visito a Juan. Maria say-PRV that pro visit-PRV Juan "Maria said that (she) visited Juan."
One question that remains open so far is why the preteritos we have seen are forced to undergo presuppositional raising, whereas embedded copreteritos may be interpreted in situ. The data in (10)
294
From Words to Discourse
shows that attributing this to an intrinsic property of preterites would be an incorrect move, since perfective past tenses may remain in situ at LF when they are embedded under a future tense clause. (10)
a. Maria pensard que visitaste a Juan. Maria think-FUT that you visit-PRV to Juan "Maria will think that you visited Juan." b. visit ... think ... Speech Time c. Speech time ... visit ... think
Sentence (lOa) allows reading (lOc), in which the visit precedes the thinking but not the Speech time, so the latter is not its reference time. Moreover the fact that the embedding verb belongs to the class that prevents presuppositional raising confirms that in this case the lower tense is not interpreted after its clause undergoes extraction to RefP. Before we proceed with our analysis, let us look at the assumptions on which it is based. First, there is a semantic preference for propositions to be located in the speaker's history (in Dowty's, 1979 sense), and thus take the Speech time as their reference time. This is reflected syntactically in that only TP's or AspP's denoting intervals that are not positively identified as located in the speaker's history are potential controllers for ZP. A perfective clause that is not c-commanded by such categories will be forced to undergo presupposition accommodation to make available the "arbitrary" interpretation for their ZP. Second, anti-factive embedding predicates block presuppositional raising. This "antifactivity" is determined by the pragmatics. Third, only TP's that are the complement of an imperfective Asp may undergo syntactic coindexation. Finally, the temporal relation between the embedding eventuality and the eventuality of an extracted complement clause is determined pragmatically. There can be simultaneity without syntactic coindexation.'
6 An additional problem, pointed out to me by Carl Vogel, is that if copreterito clauses may remain in situ, whereas presupposed material is extracted from the VP, one needs to worry about the LF position of any definite descriptions contained in non-presupposed copreterito clauses. Either one has to allow for those NP's to be extracted on their own from the VP section of the tree, or one has to advocate a proposal parallel to the one put forth by Bustos and Aliaga (2002) for the definite descriptions in subjunctive clauses. That is, one could claim that in neither of the following examples is the existence of a son presupposed: (i) a. ^Porque estd triste ese senor? "Why is that gentleman sad?" b. Tal vez su hijo este en la cdrcel. maybe his son be-PRES-SUBJUNCT in the jail "Maybe his son is in jail." (ii) Juan vio a un hombrepedir un pase de visitante al celador. "Juan saw a man ask the watcher for a visitor's pass." Penso que su hijo estaba en la cdrcel. think-PRV that his son be-IPRV in the jail "He thought his son was in jail." I feel inclined to adopt the second solution.
The Shifted Reading of the Spanish Past Tenses
295
4. WHEN ANTI-FACTIVES PREVENT RAISING AND THE SHIFTED READING I will now expand on the first two assumptions. In (11) we have data that lets us see that there are two classes of tenses: those that allow a clause embedded under their clause to undergo temporal interpretation in situ and those that do not: (11)
7
a. Dije que la cena estuvo deliciosa. I say-PRV that the dinner be-PRV delicious "I said that dinner was delicious." b. * Pense que la cena estuvo deliciosa. I think-PRV that the dinner be-PRV delicious c. (Hace rato) decia que la cena estuvo deliciosa. makes while say-IPRV that the dinner be-PRV delicious "(A while ago) I was saying that dinner was delicious" d. * (Hace rato) pensaba que la cena estuvo deliciosa. makes while think-IPRV that the dinner be-PRV delicious e. Decia que la cena estuvo deliciosa. I say-IPRV(INT7) that the dinner be-PRV delicious "I was saying that dinner was delicious." f. Pensaba que la cena estuvo deliciosa. I think-IPRV(INT) that the dinner be-PRV delicious "I was thinking that dinner was delicious." g. Dire que la cena estuvo deliciosa. I say-FUT that the dinner be-PRV delicious "I'll say that dinner was delicious." h. Pensare que la cena estuvo deliciosa. I think-FUT that the dinner be-PRV delicious "I'll think that dinner was delicious." i. Digo que la cena estuvo deliciosa. I say-PRES that the dinner be-PRV delicious "I say that dinner was delicious." j. Pienso que la cena estuvo deliciosa. I think-PRES that the dinner be-PRV delicious "I think that dinner was delicious."
INT indicates the "interrupted" reading of an imperfective past tense (see below).
296
From Words to Discourse
As we see in (1 la-b), the preterite requires a clause embedded under it to undergo presuppositional raising. What happens is, assuming Dowty's (1979) model of branching time, that when the embedding clause has perfective aspect its eventuality time is located in the same history as the speech time: (12)
Preterito Reference Time D
Speech Time ST
This forces \hepreterito embedded clause to undergo presuppositional raising in (1 la), by our first assumption. However, in (lib) the nature of the embedding predicate prevents raising, thus the sentence is not acceptable. On the other hand, in (1 lg,h) we have a future tense in the embedding clause. This is our potential reference time for the embedded tense. In this case, as we see in (13), we cannot positively locate the interval that this tense denotes as being in the same history as the Speech Time, so a future tense will be an acceptable controller for the ZP of the embedded tense. Thus the embedded clause may remain in situ in (1 Ih) and no conflict arises in spite of the anti-factivity of the embedding predicate. Let us now consider the data in (1 Ic-f) with the imperfective past embedding tense. In general, in this case antifactives are as bad as in thepreterito cases (see 1 Id). However they are allowed under the so-called "interrupted" reading of the imperfective past (lle-f), which implies that the eventuality will continue after the interruption. For instance, suppose I am talking to Juan when Maria suddenly arrives, so I utter (13). Or I am absorbed in thinking and Maria interrupts me, so I say (14). (13)
(14)
Le decia a Juan que la cena estuvo deliciosa. him I tell-IPRV to Juan that the dinner be-PRV delicious "I was telling Juan that dinner was delicious." Pensaba que la cena estuvo deliciosa. I think-IPRV that the dinner be-PRV delicious "I was thinking that dinner was delicious."
These sentences contrast with (15) in which the copreterito eventuality has culminated: (15)
A las cinco pensaba que la cena *estuvc/^habia estado deliciosa. at the five I think-IPRV that the dinner be-*PRVMlPRV- delicious PERFECT
The Shifted Reading of the Spanish Past Tenses
297
"At five I was thinking that dinner *was/>^had been delicious." What triggers the contrast between (14) and (15) is again whether we can infer a culmination for the matrix copreterito eventuality. In that case, the interval it occupies can be positively located in the same history as the Speech Time. If it does not culminate, however, due to the intensionality of imperfectives pointed out by Dowty (1979), it will, like thefuturo, allow the in situ interpretation of a clause embedded under it. Finally, in (1 Ij), we notice that the Spanish present tense also behaves like the future tense in this respect. This is not strange, since the Spanish present tense too is an imperfective, as we can learn from Bello (1841). As expected, and as we see in (18), it is subject to the imperfective paradox: (16)
Gutierre Tibon escribe un libro. Estd tan enfermo que Gutierre Tibon write-PRES a book. Is so sick that "Gutierre Tibon is writing a book. He is so sick that quizd no lo termine. maybe not it finish-PRES-SUBJUNCT maybe he won't finish it."
Thus in (1 Ij) it is also impossible to identify the matrix eventuality as being in the same history as the Speech Time, even though the latter is its reference time. Thus the present tense, in our terms a tenseless imperfective AspP, can be a ZP controller, permitting in situ temporal interpretation of a complement clause. (17) Spanish Present Tense Asp/V-P
Asp ZP Imperfective e
5. DERIVING THE ANAPHORICITY OF IMPERFECTIVES Now let us go back to the problem of why an embedded copreterito, but not an embedded preterito, allows in situ interpretation by LF co-indexation of its TP. To explain this we need to give the lexical meanings of Past and Imperfective. A Past TP will contribute to the sentence's meaning an interval P preceding its reference time, most often the Speech Time:
298
(18)
From Words to Discourse
Denotation of TP with Past head= an interval P such that for every m, m a moment in P, m<ST.
With respect to Imperfective Aspect in (19), it requires that Sentential Time (henceforth SenT), which is the interval identified as the time of the eventuality of the imperfective clause, be a superinterval (proper or non proper) of DH. DH will be the interval containing all intervals denoted by constituents of the complex sentence (whether they correspond to the temporal location of an eventuality or not) minus the interval following Speech Time. (19)
a. Imperfective^ SenT 3 DH unless it can be inferred from the meaning of constituents not dominated by AspP that SenT < ST b. SenT (Sentence Time)= interval identified as the time of the eventuality of the copreterito clause c. DH c (I, u I 2 ... u In) - F where I, ... In are intervals denoted by constituents of the complex sentence, and F is the set of all intervals PF, such that for every moment m in PF, m > ST.
According to (19c), DH is an interval containing all intervals denoted by constituents of the complex sentence (whether they correspond to the temporal location of an eventuality or not) minus what is in the future of the Speech Time. Notice that the definition of the imperfective contains the proviso that it is defeasible if a culmination preceding Speech time can be inferred for the eventuality in question, but the Past TP of a copreterito will not provide it with such culmination. Now, when no culmination is inferred for the copreterito eventuality, SenT will be a superinterval of DH. Also, DH will be a superinterval of P, since the latter is denoted by a constituent of the sentence TP. Therefore, SenT will be a superinterval of P. But since SenT is a superinterval of DH, it will last at least until Speech Time. On the other hand, P ends before Speech time, so P is different from SenT: it is a proper subinterval of it. (20)
copreterito eventuality does not culminate P c SenT Culmination of P ST = Culmination of DH D
D
SenT
...
The Shifted Reading of the Spanish Past Tenses
299
We know SenT is identified as the eventuality time of the imperfective sentence, so P will not be "identified" as the temporal location of any eventuality (only of a part of one). This will give the phrase that corresponds to P, namely TP, an anaphoric character: it will need to get co-indexed with some other temporal expression serving as temporal location of some eventuality in order to be "identified". That is why the TP of an imperfective undergoes syntactic co-indexation. On the other hand, if the material not dominated by AspP allows us to infer that SenT precedes Speech Time, SenT will not be a superinterval of DH, it will be a proper subinterval of it instead. (21)
copreterito eventuality culminates SenT < ST then SenT c DH, and SenT gets identified with P. Culmination of SenT = Culmination of P D ST = Culmination of DH D
D
D
SenT In this case P will be identified with SenT, since this is the most straightforward way to make P correspond to the temporal location of some eventuality. This will be expressed in our notation by co-indexing a TP with the AspP that dominates it. But in this way TP loses its anaphoric character. TP's identified as denoting the eventuality time of their own clause may never undergo syntactic coindexation. They will only receive a temporal interpretation if their ZP gets controlled or, through presuppositional raising, gets to denote the Speech time. Incidentally, apreterito's AspP either has an empty head or is altogether missing. Either way its TP is identified as its eventuality time. (22)
Possible LF Structures for the Preterite a. Asp/V-P b. T/V-P TP
VP
6. VARIATION IN SENSITIVITY TO AKTIONSART What we have said so far about the copreterito applies to those varieties of Mexican Spanish for which the following remark in Moreno de Alba (1978) holds true: "the copreterito may also be used with perfective verbs, giving a duration to the action and presenting it as unfinished". However in spoken Mexican Spanish the copreterito and the simple present tense have been replaced by the
300
From Words to Discourse
imperfective past progressive and the present progressive, except for a limited set of contexts and for sentences with stative verbs or with an iterative meaning.8 Colloquial Mexican Spanish now employs progressive tenses because the simple present tense and the copreterito have, I believe, acquired a behavior parallel to that of the Aspect of the English present and past tenses. In that language those two tenses have an Aspect with a definition akin to that of our imperfectives in (21). Nevertheless there is, I suggest, a parametric difference between English and Colloquial Spanish on one hand, and the variety of Spanish cited by Moreno de Alba on the other. Whereas the English present and past tenses, as well as the copreterito and present tense of the English like Spanish dialects, are essentially imperfectives too, they differ from the copreterito described in this paper in that they are Aktionsart-sensitive imperfectives. That is, whereas in our copreterito SenT refers to the time of the sentential eventuality regardless of the sentence's aktionsart, the SenT of the English Aspect will differ in accordance to the sentence's aktionsart. In case the sentence has an eventive aktionsart, at least two phases will have to be mapped into the English SenT: one phase during which the eventuality takes place, and one during which it no longer does. On the other hand, one will only need to map a single phase into the SenT of stative sentences, the phase during which the state holds. Thus statives in the English present tense will behave aspectually like regular imperfectives, whereas eventives will need to be interpreted as iteratives. This would be due to the fact that the two distinct phases that need to be mapped into SenT will need to be mapped into an interval that is of indeterminate length (recall that we only know that SenT 3 DH): iterativity would be the way out of this problem, since the two phases could be mapped into SenT alternatively an indefinite number of times. (23)
I brush my teeth. phase 1. brushing phase 2. no brushing SenT phase 1. brushing phase 2. no brushing
Perhaps eventive sentences in the English past tense are exempt from requiring iterativity due to the possibility of identifying interval P (c.f. (18)) with phase 1 (that is, the phase mapped into SenT during which the eventuality does hold) given the transparency of SenT in the language. These, as 8 1 thank Javier Gutierrez-Rexach for pointing out to me this divergence between the grammarians' descriptions and the colloquial usage of these tenses.
The Shifted Reading of the Spanish Past Tenses
301
yet, sketchy ideas seem to me to be worth further exploration. After this short diversion, in the next section we return to important issues related to our analysis of the shifted reading.
7. THE PRAGMATICALLY DETERMINED TEMPORAL RELATION BETWEEN EMBEDDING CLAUSES AND EXTRACTED EMBEDDED CLAUSES The shifted reading was so called in the first place because the past tense eventuality of the embedded clause preceded the past tense eventuality of the embedding clause. But I have claimed that there is no syntactic co-indexation involved. Now I will advocate the proposal that the temporal relation between such clauses is a matter of lexical semantics and pragmatics, as Suner (1990) and Zagona (1997) have claimed. Suner, for instance, proposed that verbs are lexically specified with features such as [+/- precedence]. Prometer (to promise) would have the value [-precedence]. Ogihara (1989, 1996) proposes his TEMPORAL DIRECTIONALITY ISOMORPHISM, maintaining that a de RE report about a temporal entity can only be made when the speaker's and the attitude holders' viewpoints match in temporal directionality. Ogihara says this ISOMORPHISM excludes future shifted readings and provides the relative ordering of the two eventualities in double access sentences. Similarly, Abusch (1997) proposed her "Upper Limit Constraint" (henceforth ULC), based on the idea that "forward reference with tenses is impossible because future times are not sufficiently determined from the perspective of the now of an intentional context". Abusch says that the fact that in a branching future model, from the perspective of any point in time, the past is determinate but not the future, prevents tenses from referring to the future, unless they are in the context of a modal. The local evaluation time is an upper limit for the denotation of tenses. Abusch reports that Heim considers the ULC to be a presupposition carried by tense nodes that the interval they denote does not follow their evaluation time. The ULC is generally exemplified by sentences in which the embedding predicate is a predicate of communication or thinking. Curiously Abusch presents counterexample (24) to the ULC as marginally acceptable, and that sentence contains an embedding verb that does not belong to the aforementioned class: (24)
?? As a medical student John hoped to discover a disease he later found the cure for.
Since verbs like "hope" do allow their complements to temporally follow, perhaps the ULC is not due to a characteristic of tense nodes, but of the embedding predicate: the subjects of predicates of communication or thinking are transmitting or mentally manipulating pieces of information. So it is pragmatically plausible that they are acquainted with that information because it involves eventualities in their own past. But then, why are future shifted past tense clauses only marginally acceptable under verbs of the "hope" class?
302
From Words to Discourse
This has to do with the peculiar semantics of this class, in particular, the fact that it is most unusual to conceive of hopes, promises or predictions as presupposed. Recall that, on the other hand, a preterito under preterite sentence requires the embedded clause to undergo presuppositional raising, due to the tense combination. Thus (25a) is ungrammatical whether on a future shifted or on a past shifted reading: (25)
a. *Juan predijo que la bomba explotd. Juan predict-PRV that the bomb explode-PRV b. Juan predijo que la bomba explotaba. Juan predict-PRV that the bomb explode-IPRV "Juan predicted that the bomb was exploding/would explode."
However, if by other cues in the context we can force the embedded sentence to be presupposed, then not only is the sentence grammatical, but it is also understood as future shifted. The text fragment in (26) was judged perfect by all five of my informants, who had previously rejected (25a).
(26) El 4 de marzo de 1970 se organize un concurso de adivinos. Se inscribieron Juan, Luis y Pedro. Las reglas del concurso eran que ganaria el que ese dia hiciera mas predicciones que, con el tiempo.se comprobara que habian sido correctas, lo cualse verificaria en 1998. Hoy estamos en la ceremonia de premiacion, tu llegas tarde y me preguntas quien gano. Yo respondo: "Juan y Pedro hicieron 7 predicciones correctas que Luis tambien hizo, pero Luis ganoporque hizo 8. El sipredijo que mataron a Colosio, mientras que los otros no ". "On March the 4th, 1970, a fortunetellers contest was organized. Juan, Luis and Pedro registered for it. The rules of the contest were that whoever made more predictions on that day that, in due time, could be proved to have been correct, would win (which would be verified in 1998). Today we are in the awards ceremony, you are late and ask me who won. I answer: "Juan and Pedro made 7 correct predictions that Luis made too, but Luis won because he made 8. He did predict (PRV) that they murdered (PRV) Colosio, whereas the others did not".
(note: the 1994 murder of Colosio, candidate to the presidency, is a fact that is well known to Mexicans). Hence, we see the ULC effect in Spanish too, but it seems to work differently according to the properties of the embedding predicate. Verbs of communication and thinking are understood as having complements denoting information about eventualities in their subjects' past, and do not allow future shifting. Future oriented verbs do allow future shifted readings, provided their complement clause may undergo presuppositional raising so that their ZP can be anchored. This is made difficult by their essentially antifactive nature. There are also sentences that do not involve TP co-indexation at LF, but are given a
The Shifted Reading of the Spanish Past Tenses
303
simultaneous reading by pragmatic means: (27)
a. Juan vio que el nino metio la mano alfuego. Juan see-PRV that the kid put-in-PRV the hand to-the fire "Juan saw the kid put his hand into the fire." b. Maria oyo que dieron las tres. Maria hear-PRV that strike-PRV the three "Maria heard [the clock] strike three." c. La victima sintio que la golpearon con un objeto pesado. the victim feel-PRV that her hit-PRV with an object heavy "The victim felt them hit her with a heavy object."
These sentences with verbs of perception allow a simultaneous reading in spite of the extraction of the embedded clause, which we have established is necessary with a preterito under preterito. Simultaneity is here entirely pragmatically determined, for if the embedded eventuality's duration makes it implausible for the subject to perceive it from beginning to end, the perception is still contemporaneous with what is perceived, but what is perceived, however, will no longer be an eventuality: it will be information on an eventuality in the subject's past (see Suner 1990). (28)
a. Juan vio que hubo una fiesta. Juan see-PRV that be-PRV a party "Juan saw that there had been a party." b. Juan oyo que el vecino serrucho un mueble, Juan hear-PRV that the neighbor saw-PRV a furniture "Juan heard that the neighbor sawed a piece of furniture."
In (29), with an embedded copreterito, which allows syntactic co-indexation of its TP, the same pragmatic context does not preclude simultaneity: (29)
a. Juan vio que habia una fiesta. Juan see-PRV that be-IPRV a party "Juan saw that there was a party." b. Juan oyo que el vecino serruchaba un mueble. Juan hear-PRV that the neighbor saw-IPRV a furniture "Juan heard that the neighbor was sawing a piece of furniture."
304
From Words to Discourse
Finally, consider example (30): (30)
a. Juan decia que estudiaba ingles. Juan say-IPRV that he study-IPRV English "Juan said that he studied/was studying English." b. say say say say study study study study c. * study say study say study say
Here, the matrix copreterito behaves as in (12c-d) so a perfective clause under it must undergo presuppositional raising, as the ungrammaticality of (lid) demonstrated. It is known that, as imperfectives, copreteritos may receive an iterative reading. For a sentence with two copreterito clauses, an iterative version of the syntactically determined simultaneous reading becomes available, as we see in (30b). On the other hand, (30c), the iterative version of the past shifted reading is impossible. This is predicted if the effects of the iterativity trigger, probably an operator, can be transmitted through the index shared in the simultaneous reading. But, as we have shown, in the shifted reading the complement clause's reference time is the speech time, there is no co-indexation with the matrix TP. So this reading cannot be iterated. The complement clause is not in the scope of the operator, and the latter cannot affect the pragmatically determined temporal relation between the two clauses.
REFERENCES Abusch, D. 1988. "Sequence of Tense, Intensionality and Scope." Proceedings of the 7th West Coast Conference on Formal Linguistics ed. by Hagit Borer. Stanford: The Stanford Linguistics Association. Abusch, D. 1997. "Sequence of Tense and Temporal De Re." Linguistics and Philosophy 20.1-50. Aliaga, F. and E. Bustos. 2002. "Mental Spaces and Epistemic Attitudes: on the Indicative/Subjunctive Alternation in Spanish." This volume. Beghelli, F. & Timothy Stowell. 1997. "Distributivity and Negation." Ways of Scope Taking ed. by Anna Szabolcsi. Dordrecht: Kluwer. Bello, A. 1841. Andlisis Ideologico de los Tiempos de la Conjugacion Castellana. Valparaiso, Chile: Imprenta de M. Rivadeneyra. Bennett, M. & Barbara Partee. 1972-1978. "Toward the Logic of Tense and Aspect in English." Bloomington: Indiana University Linguistics Club. Berman, S. 1991. On the Semantics and Logical Form of Wh-clauses. Ph.D. diss. UMass., Amherst. Costa, R. 1972. "Sequence of Tenses in That-Clauses". Papers from the 8th regional meeting of the Chicago Linguistic Society ed. by P. Peranteau, J. Levi and G. Phares. Chicago: CLS. de Swart, H. 1997. "Review of T. Ogihara 'Tense, Attitudes and Scope'." Linguistic Analysis 26. 3-4. Diesing, M. 1990. The Syntactic Roots of Semantic Partition. Ph.D. diss. UMass., Amherst.
The Shifted Reading of the Spanish Past Tenses
305
Diesing, M. 1992. Indefinites. Cambridge, Mass: MIT Press. Dowty, D. 1979. Word Meaning and Montague Grammar. Dordrecht: Reidel. £119, M. 1987. "Anchoring Conditions for Tense." Linguistic Inquiry 18. 633-657. Heim, I. 1982. The Semantics of Definite and Indefinite Noun Phrases. Ph.D. diss., UMass., Amherst. Moreno de Alba, J. (1978). Valores de las Formas Verbales en el Espanol de Mexico. Mexico: UNAM. Ogihara, T. 1989. Temporal Reference in English and Japanese. Ph.D. diss., University of Texas, Austin. Ogihara, T. 1996. Tense Attitudes and Scope. Dordrecht: Kluv/er Academic Publishers. Partee, B. 1973. "Some Structural Analogies between Tenses and Pronouns in English." The Journal of Philosophy 0 601-609. Pinar, P. 1995. Negative Polarity Licensing and Negative Concord in the Romance Languages. Ph.D. diss. University of Arizona, Tucson. Stowell, T. 1993. "Syntax of Tense.".Ms., UCLA. Suner, M. 1990. "El Tiempo en las Subordinadas." Tiempoy Aspecto en Espanol ed. by I. Bosque, J. Acero and A. Lopez. Madrid: Catedra. Zagona, K. 1997. "Tenses and Anaphora, Is There a Tense Specific Theory of Coreference?" Ms., University of Washington, Seattle.
This page intentionally left blank
307
15
ON THE INTERACTION OF SYNTAX-SEMANTICSPRAGMATICS: A CASE STUDY Josep Macia, University of Barcelona
1. INTRODUCTION Structural relations among Noun Phrases (or NPs) in a sentence place constraints on the possible semantic relations among those NPs. Consider the following sentences: (1)
a. Juan lo vio. Juan him saw "Juan saw him." b. Cada hombre lo ama. Every man him loves "Every man loves him."
For instance, in (la) Juan and lo 'him' cannot refer to the same individual, and (Ib) cannot mean that every man loves himself, that is, lo 'him' cannot be interpreted as bound by cada hombre 'every man' — bound in the sense analogous to a logical variable being bound by a quantifier. Consider also the examples in (2) —where two NPs in a sentence have the same index if and only if either they both refer to the same individual or one of them is a quantificational NP (e.g, every boy, or somebody) and the other is bound (in the logical sense) to it:
308
(2)
From Words to Discourse
a. b. c. d. e. f. g. h. i. j. k. 1. m. n. o. p. q. r.
Kathririj saw her,. *Kathrin( saw her;. She; saw her; father. Daniel; says that he; is tired. Danieli says that he, is tired. He; says that Jasorij is tired. *He; says that Jason; is tired . HiSj father says that Jason; is tired. *(His father)i says that Richard; is tired. (His father); says that he( is tired. He; saw Robertj. *Hej saw Robert;. His; mother saw Robert;. *Everybody; saw him;. Everybody; saw him,. Every thief, fears that he; gets caught. In Scarsdale, every single boy's mother sends the twerp; off to summer camp. *In Scarsdale, every boy; thought the twerp; would hate summer camp1.
We can explain all these data by appealing to the so called Principles (B) and (C) of the Binding Theory (Chomsky, 1993) (though we will not go into the details of discussing how each sentence above is accounted for by these principles): (B) If a is a pronominal, interpret it as distinct in reference from every c-commanding phrase in its Governing Category; (C) If a is an R-expression, interpret it as distinct in reference from every c-commanding phrase. A succinct explanation of the terminology involved in the formulation of Principles (B) and (C) is in order. The class of pronominals includes the non-reflexive pronouns; R-expressions include proper names, epithets, definite descriptions and quantificational NPs. For the purposes of the present discussion we can set aside some difficulties arising when trying to specify the notion of Governing Category, and we can take the Governing Category (GC) of a node X to be the smallest (i.e. lowest in the tree) NP or IP that contains X. X c-commands Y iff (roughly) the lowest branching node in the syntactic tree (that corresponds to the logical form of a sentence) that dominates (i.e., is 'connected' and 'above') X also dominates Y. In the remaining part of this introductory section we will call the attention to the interplay between syntax and semantics in Binding Theory. As they are formulated in Chomsky (1993), (B) and (C) are semantic (rather than syntactic) principles. They are principles about how certain NPs should be interpreted. In contrast, earlier formulations of Binding Theory (e.g. Chomsky, 1981) might have seemed to make Binding Theory
On the Interaction of Syntax-Semantics-Pragmatics
309
a purely syntactic theory. This is really not so, though. In Chomsky (1981) principles (B) and (C) are formulated as: (B)' Pronominals must be A-free in their Governing Category; (C)' Rexpressions must be A-free. Here is a succinct explanation of the terminology: Node X binds node Y if they are co-indexed and X c-commands Y. To be A-free is not to be bound by anything in an A-position (complement and subject positions are A-positions, adjoined positions and the specifier of CP are non A-positions). We can make three observations regarding (B)'and (C)'. First, according to (B)' and (C)' indexes are not simply a part of some provisional notational convention to help us indicate what the reading of a sentence that we want to consider is. Rather, according to (B)' and (C)' indexes are an essential part of the syntactic representation of a sentence (in the same way that each node in a tree is part of that representation). This is so because Binding Principles (B)' and (C)' are formulated in terms of the relation of binding, which, in turn, appeals to the relation of sameness of indexes. Second, in order to account for the data in (1) and (2) we need to appeal not only to Binding Principles (B)' and (C)', but also to a principle such as: "Referential NPs with the same index must refer to the same individual, referential NPs with different indexes must refer to different individuals; an NP is interpreted as bound by a quantificational NP if and only if they have the same index." This principle about interpretation is crucial. Without it, Binding Theory would not explain, for instance, why (3a) below is unacceptable if Begona is co-referential with la. Even if the syntactic representation (3b)2 is ruled out by Principle (B)', (3c) is not. If it were possible for two NPs with different indexes to have the same referent, then (3c) could give rise to an interpretation that (3a) does not have. (3)
a. Begona la quiere. Begona her likes "Begona likes her." b. Begone laj quiere. c. Begonaj la^ quiere.
Third, as the discussion in the previous paragraph shows, a theory that tries to explain why (3a) cannot have a coreferential reading has to include, in some way or other, a semantic component. The explanation cannot be given in purely syntactic terms (i.e., purely in terms of which LFstructures are or are not acceptable). It might be, as it is the case with the theory of Chomsky (1981), that the bulk of the explanation is carried out by purely syntactic principles (principles that 1
(2 k-r) are taken from Higginbotham (1992). (3c) should be regarded as a partial specification of a tree. We do not explicitly specify the whole tree to simplify the exposition. Analogously for (3c). 2
310
From Words to Discourse
rule out certain LF-structures), and that the semantic component (i.e. the set of rules regarding how to interpret those LF structures) is straightforward. Still the semantic part, even if straightforward, is essential to explain what we want to explain. When we try to account for the data in (1) and (2) what we are trying to do is to explain the kind of constrains that the structural relationships among NPs place on the interpretation of those NPs.
2. COUNTER-EXAMPLES TO BINDING THEORY PRINCIPLES (B) AND (C) Binding Theory accounts for a vast array of data. Nevertheless it has been observed in the literature that there are several different kinds of sentences that are counter-examples to Binding Theory Principles (B) and (C). In (4) we have some examples of sentences that pose a problem for principle (C):3 (4)
3
a. Esa senora es la profesora Rigau. That lady is the professor Rigau "That lady is professor Rigau." b. Todo el mundo se ha dado cuenta finalmente de que Oscar es un All the world self has given notice finally of that Oscar is an incompetente. Incluso il se ha dado cuenta de que Oscar es un incompetente. incompetent. Even he self has given notice of that Oscar is an incompetent. "Everyone has finally realized that Oscar is incompetent. Even he has finally realized that Oscar is incompetent." c. Mary, Betty and John have one thing in common: Mary admires John, Betty admires John, and he himself 'also admires John. d. What do you mean Oscar loves no one? He loves Oscar. e. The logic tutor while trying to explain the law of Universal Instantiation to a student tells him: Look, fathead. If everyone loves Oscar, then certainly he himself must love Oscar. f. Creo que esa mujer que estd hablando por television es Zelda. I-thinkthat that woman that is talking on TV is Zelda. Dice las mismas cosas que Zelda dice en su libro she-says the same things that Zelda says in her book. "I think that woman talking on TV is Zelda. She says the same things that Zelda says in her book."
(4b-e), (5a) and (5c-d) are adapted from Evans (1980), (4e) and (5b) are adapted from Heim (1993).
On the Interaction of Syntax-Semantics-Pragmatics
311
The last clause in each of these examples can be interpreted so that the two italicized NPs have the same referent. In each of these sentences, though, one NP c-commands the other, and, furthermore, the c-commanded NP is an R-expression. The co-referential interpretation is then undesirably ruled out by Principle (C) of the Binding Theory. Analogously, the sentences in (5) pose a problem for Principle (B). (5)
a. Ya se que es lo que Maria, Rosa y Juan tienen en Already I-know what is it that Maria, Rosa and Juan have in comun: Maria admira a Juan, Rosa tambien lo admira y el mismo common: Maria admires to Juan, Rosa also him admires and the self Juan tambien lo admira. Juan also him admires "I know what Maria, Rosa and Juan have in common: Maria admires Juan, Rosa also admires him, and Juan himself also admires him." b. At a Halloween party: Creo que el chico con el disfraz de demonio At a Halloween party: I-believe that the man with the costume of devil es Juan.Es muy sospechoso que (el) lo conozca tan bien. is Juan. It-is very suspicious that (he) him knows so well "At a Halloween party: I believe that the man with the devil costume is Juan. It is very suspicious that he knows him so well." c. What do you mean no one loves Oscar? Oscar loves him. d. What do you mean no one loves Oscar? He loves him.
Some of these sentences, such as (4a) or (5b), are one hundred percent acceptable; others, such as (4c) or (5d), are not perfect but are good enough as to pose a problem for a theory according to which they are simply ruled out. There have been several proposals that try to account for these counter-examples including those of Reinhart (1983), Grodzinsky and Reinhart (1993), Heim (1993), Higginbotham (1992) and Fiengo and May (1994). We do not have enough space here to discuss each of these proposals. What they all have in common is that they try to find a condition or set of conditions that would rule out most of the sentences that Binding Theory rules out but still accept sentences like those in (4) and (5). None of these proposals, I believe, is successful. And they cannot be, since they attempt to deal in a homogeneous way with what are different kinds of phenomena that require different sorts of explanation. Specifically, the sentences in (4) and (5) involve three kinds of counter-examples to (B) and (C): (a) Copular sentences (like (4a)); (b) Non-copular one-hundred percent acceptable sentences (like (4f), (5b));
312
From Words to Discourse
(c) Acceptable though somewhat odd sentences (like (4b), (4c), (4d), (4e), (5a), and (5c), (5d)). Each of these three kinds of examples involves a different kind of phenomena. We will see in the next three sections how we can successfully account for each of these three kinds of data, as well as for all the data that is already successfully accounted for by Binding Theory. In section 3, we will present a new version of principles (B) and (C) and explain how we can account for the counter-examples of type-(b); in section 4, we will consider the type-(a) examples; finally in section 5, we will account for type-(c) counter-examples by explaining both why those sentences are acceptable and why they nevertheless have a somewhat odd character.
3. TAKING SPEAKERS' BELIEFS INTO ACCOUNT I think that (B) and (C) should be modified as in (B)* and (C)*: (B)* If a sentence whose LF representation is of the form ...a--.p-.- (where p is a pronominal, and a is an NP that c-commands p in its GC) is used in a context C, it is not presupposed in C that a and p are co-valued. (C)* If a sentence whose LF representation is of the form ...a-..p... (where p is an R-expression, and a is an NP that c-commands p) is used in a context C, it is not presupposed in C that a and p are co-valued. In stating these conditions we use the notions of 'being presupposed', 'context' and 'being co-valued' in a particular technical sense, which we explain in what follows. The notions of 'presupposition' and 'context' that we use in stating (B)* and (C)* are the same as in Stalnaker (1973, 1974). Linguistic communication always takes place on the basis of a background of common beliefs and assumptions, or context. Using the possible worlds framework we can identify a context in which some instance of linguistic communication takes place with a set of possible worlds: those worlds that as far as the participants in the conversation can tell could be the actual world. The context set consists of those worlds that could be the actual world according to what the participants in the conversation believe, and believe that the others believe, and believe that the others believe that they believe and so on. (If we wanted to make this characterization more precise we should take into consideration that what is relevant is not only what the participants believe but also what they pretend to believe). A proposition p is presupposed in a context C if p is true in each world in C. That is, a presupposition of a conversation is whatever the participants in the conversation take to 4
1 take the idea of appealing to what it is presupposed rather than to what is actually the case in trying to deal with the phenomena related to principles (B) and (C) from Heim (1982,1993). Heim, in turn, credits Postal (1970) for this idea. The main difference between Heim's formulation of the binding principles and mine is that she states her conditions as purely syntactic principles that rule out certain co-indexations among NPs.
On the Interaction of Syntax-Semantics-Pragmatics
313
be part of their common assumptions. For the present purpose, we can take to be co-valued with just to mean to be co-referential with. This is a simplification since then 'to be co-valued with' does not apply to the relation between a quantificational NP and the NP that is interpreted as logically bound to it. We can use this simplified notion here since none of the examples that are important for the discussion in this paper involve quantificational NPs. Using this simplified notion will allow us to keep our explanation shorter and to avoid some technical complications that are not important for the topic that concerns us here. For a treatment of the notion of 'co-valueness', that includes that of 'co-reference', but is not equivalent to it, see Macia (1997). Having explained the notions of being co-valued, presupposition and context we can now explain the central notion in our principles (B)* and (C)*: What does it mean that two NPs are presupposed to be co-valued in a certain context C? As we stated above, we can identify a context with a set of possible worlds: those worlds that are compatible with what the participants in the conversation take to be their shared assumptions. Sometimes a conversation may take place in a context where someone's identity is in question. For instance, it might be an open question whether the person we are looking at is, say, Madeleine Albright. That it is an open question whether she is Albright or not means that there are two kinds of worlds that are both compatible with what the speakers are assuming: worlds where Albright is the person we are looking at, and worlds where the person we are looking at is someone other than Albright. Some terminology is needed at this point. We say that a referential NP picks up an individual I at a world W in the context C, if I is the individual that the NP would refer to if C were the actual world. Let us now consider one example. Suppose I utter the sentence She is a smart person while pointing to a woman who is in front of us. Since all the participants in the conversation will believe (and believe that the others believe, etc.) that the woman is in front of us and that I uttered She while pointing at her, it will be part of the context that the woman is in front of us and that she is the one I am referring to with my use of She. That is, for each world in the context W She will pick up the woman that is in front of us in W. Let's suppose further that we are unsure whether the woman is Madeleine Albright. That means that there will be some worlds in the context where the woman in front of us is Madeleine Albright, but there will also be some worlds in the context where Madeleine Albright is someone other than the woman in front of us.5 She will pick up 5 If I were to utter She is Madeleine Albright, then this sentence would express the necessary true proposition in those worlds where the woman is Madeleine Albright and the necessary false proposition in those worlds where Madeleine Albright is someone else. Not knowing which worlds in the context agree with how the actual world is, we would not know whether the utterance was necessarily true or necessarily false. Still the utterance would be informative since it would indicate to anyone who accepted it, that those worlds where the woman is not Madeleine Albright (and where the utterance expresses the necessarily false proposition) are not compatible with what we take to be the case, and so should no longer be regarded as part of the context. This interesting picture helps us clarify how it is possible that an utterance expresses a necessary proposition but it is nevertheless informative. For a more detailed explanation
314
From Words to Discourse
Madeleine Albright in those worlds where Albright is the woman in front of us, but will not pick up Madeleine Albright in those worlds where someone other than Madeleine Albright is in front of us. This agrees with the intuitive idea that, if we do not know whether the woman is or not Madeleine Albright, then we are unsure as well as to whether my use of She refers to Madeleine Albright or to someone else. We say that NP1 and NP2 are presupposed to be co-valued when used in a context C when NP1 and NP2 pick up the same individual in each world in C. In our example She and Madeleine Albright are not presupposed to be co-valued since in some worlds in the context they pick up different individuals. As we saw, there are worlds in the context where the person in front of us is not Albright, and in those worlds She and Madeleine Albright will pick up different individuals. With our formulation of (B) and (C) it is easy to explain why the last clause of (4f) and of (5b) are completely acceptable sentences. In other words, we can easily account for the type (b) of counter-examples: the non-copular sentences that are one hundred percent acceptable. Let us consider, for instance, (6) (which is the same as (5b): (6)
At a Halloween party: Creo que el chico con el disfraz de demonio es At a Halloween party: I-believe that the man with the costume of devil is Juan. Es muy sospechoso que (il) lo conozca tan bien. Juan. It-is very suspicious that (he) him knows so well "At a Halloween party: I believe that the man with the devil costume is Juan. It is very suspicious that he knows him so well."
The reason why (6) is good even if el 'he' and lo 'him' in the last clause actually refer both to Joe is this: when uttering the last part of (6) it is still an open question whether the man in the devil costume (who is the one el refers to) is or is not Joe (who is the one lo refers to). Whether they are the same or not is precisely what is being discussed. Putting it in terms of the possible worlds framework: in some worlds in the context, Joe is the man that is wearing the devil costume at the Halloween party, but in some other worlds in the context someone other than Joe is the one who is wearing the devil costume. That means that we interpret el and lo so that they pick up the same individual in those worlds in the context where Joe is the one wearing the devil costume, but they pick up different individuals in those worlds where someone else is wearing the costume. Thus, we interpret the two NPs so that it is not presupposed that they pick up the same individual and so it is not presupposed that they are co-valued. There is no problem, then, in interpreting the last part of (6) in accordance to (B)*. An analogous explanation could be given for the acceptability of (5e).
see Stalnaker( 1978).
On the Interaction of Syntax-Semantics-Pragmatics
315
4. COPULAR SENTENCES Consider (4a) repeated here as (7): (7)
Esa senora es la profesora Rigau. That lady is the professor Rigau "That lady is professor Rigau."
Regarding sentences with a so-called equative use of the copula, one might think that the explanation for why they are completely good sentences is the same that we just gave for (6). However this is not really the case. It is true that usually when we utter a sentence such as (7) we are not presupposing that the two NPs refer to the same individual. Nevertheless it still might be perfectly fine to utter an identity sentence in a situation where according to what the participants assume, the two NPs will refer to the same individual. For instance, if Maria, Rosa and Juan are in a room and they all know each other (and know that they know each other, etc), Maria's utterance of (8) is still good. (8)
El es Juan. He is Juan "He is Juan."
It might be a silly and pointless utterance (since what she is saying is something that is already obvious to the three participants in the conversation), but it is still perfectly good from a grammatical point of view. I think that a promising way of trying to account for the complete grammaticality of identity sentences is in terms of the special syntactic character of sentences containing the verb to be. Several works have pointed out and tried to explain the special syntactic properties of the so-called copular sentences6 In particular Heggie (1988) argues that the predicate of copular sentences is not the verb to be, but rather one of the NPs.7 That means that there is only one argument in identity sentences: the NP that is not the predicate.8 If we assume that our binding " See, for instance, Heggie (1988) and Moro (1991). 7 According to Heggie this claim requires some qualifications. 8 Argument and predicate are technical notions in linguistic theory. They are usually characterized in terms of the so-called Theta-theory. One alternative way of characterizing the notions of argument and predicate is in terms of the type of semantic values they can have. We will not commit ourselves here to any particular characterization of the two notions. Among many other alternative ways, one of the simplest characterizations would be the following: We define a predicate ofdegree-n inductively as follows: A predicate of degree-1 is an expression whose semantic value is a function from the set of individuals to the set of truth values; a predicate of degree-n+1 is an expression
316
From Words to Discourse
principles apply only to arguments, then (B)* and (C)* do not provide any information about how copular sentences like (7) should be interpreted. The interpretation of identity sentences does not, then, involve any conflict with the binding principles. Discussing the different arguments that show the special character of copular sentences and that try to prove that the predicate of copular sentences is one of the NPs, as well as examining the possible evidence against this view is by itself a lengthy topic. The reader is referred to the discussion in Heggie (1988) and the other works mentioned in footnote 6. Here we will restrict ourselves to mentioning one kind of data that gives plausibility to the view, without entering in any further details. In Spanish the third person feminine clitic la 'her' corresponds to an argument position; the clitic lo 'it' can correspond to predicates but cannot correspond to any feminine NP in an argument position. This is illustrated in (9): in the second sentence the clitic la stands for the argument Maria in the first sentence, while it is not possible for lo to stand for that argument.9 (9)
A: Ese hombre ama a Maria. B: No, no la/*lo ama. A: That man loves to Maria. B: No, not her/it he-loves "A: That man loves Maria. B: No, he does not."
In the second sentence in (10) the clitic lo stands for the predicate delgada (thin) that appears in the first sentence, whereas la cannot stand for that predicate (10)
A: Maria es delgada. A: Maria is thin. "A: Maria is thin.
B: No, no *la/lo es. B: No, not her / it she-is B: No, she is not. '
In contrast with (10), the argument clitic la in the second sentence of (11) cannot be made to stand for the NP Maria, but the predicate clitic lo can: (11)
A: Esa es Maria. A: That is Maria. "A: That is Maria.
B: No, no *la/loes. B: No, not her/it she-is B: No, she is not."
The data in (9)-(ll) suggest that the NP Maria in (11) is not an argument but a predicate. If the binding principles (B)* and (C)* apply only to argument NPs, then they are irrelevant for the whose semantic value is a function from the set of individuals to the set of predicates of degree-n. X is a predicate if, for some n, X is a predicate of degree-n. X is an argument, if X can be functionally combined with a predicate of degree-1 to yield a truth value. (Notice that under this definition not only expressions whose semantic value is an individual can be arguments; if the semantic value of a quantificational expression is a function from predicates of degree-1 to truth values, then quantificational expressions are also arguments). 9 Heggie(1987) describes similar facts to (9)-(l 1) for French and Italian.
On the Interaction of Syntax-Semantics-Pragmatics
317
interpretation of (11). In particular, there would be no conflict with (B)* or (C)* if we interpreted the two NPs in the first sentence in (11) in such a way that they determine the same individual in each world in the context.10 As we said at the beginning of this section, what we have presented is just one among several pieces of evidence for the conclusion that one of the NPs in a copular sentence is not an argument but a predicate. Taken together these different pieces of evidence make a strong case for this conclusion. We would then see that it is possible to successfully account for another type of the alleged counter-examples to Binding Theory: copular sentences. Notice, finally, that regarding this kind of alleged counter-example our explanation made use of syntactic evidence in order to establish a thesis regarding the semantic type of the NPs involved.
5. THE INTERACTION OF DIFFERENT SOURCES OF INFORMATION An adequate theory for the phenomena we are concerned with in this paper should not only account for the contrast between the sentences which are not acceptable and those which are at least partially acceptable but it should also give an account of why the sentences in (l)-(5) have different degrees of acceptability and, especially, it should account for the difference between fully grammatical sentences like (4f) or (5b) and sentences which are regarded as grammatically awkward and only partially acceptable —like, for instance, (4c) or (5d). I believe we can account for what we called type (c) counter-examples, that is, for the kind of sentences in (4) and (5) which are acceptable even if somewhat awkward, in the following way. Speakers' linguistic judgements should not be seen as directly reflecting the speakers' (implicit) knowledge of binding principles (B)* and (C)*, but rather they should be seen as the final product of the interaction of many factors. When interpreting an utterance of a sentence like (4c) the speakers use several sources of information, the binding principles being just one of them. If the sources of information other than the binding principles determine an interpretation for the NPs in a clear enough way, the sentence will be able to be understood in accordance with that interpretation even if it conflicts with the information provided by the binding theory. The existence of this conflict will explain that the sentences are not regarded as completely good. We are proposing that instead of trying to provide some principles that forbid certain sorts of relations when certain structural relations among NPs obtain, we should rather look at binding theory as describing what information certain structural relations among NPs introduce. That is, 10 1 use the word determine in an ambiguous way. The two NPs will 'determine' the individual in a different way, since, presumably, they will have a different type of semantic value. If the semantic value of the argument NP esa 'that' is an individual, then the semantic value of the predicate NP Maria might, for instance, be a set of individuals -actually, a
318
From Words to Discourse
instead of looking at it from the restrictive side (what cannot be the case and will be ruled out), we could look at it from the side of what is contributed (what information is introduced by certain structural relations). This allows us to see the binding principles as part of a more general picture: that of the interaction of the different semantic factors that give information about the interpretation of the NPs in a sentence. Then, I contend, we have a better and more natural way of explaining the grammaticality judgements of the speakers about some of the problematic sentences. When interpreting the NPs in a sentence we have, on the one hand, the information provided by binding principles (B)* and (C)* (on the basis of the structural relations among the NPs); on the other hand, we have the information provided by other aspects of the sentence or the discourse, and by the context. These two sources of information usually concur, but sometimes they might be in conflict. If they are in conflict and if the second sort of information is strong and unequivocal enough it might overcome the information provided by the binding principles. Then the speakers will judge that the interpretation of the sentence induced .by the second source of information is possible. Nevertheless, and because of the conflict with the binding principles, the sentence will be judged as somewhat awkward or only partially acceptable. Let's examine one specific example. Consider (12), which is the same as (5a). (12)
Ya se que es lo que Maria, Rosa y Juan tienen en comun: Already I-know what is it that Maria, Rosa and Juan have in common: Maria admira a Juan, Rosa tambien lo admira y el mismo Juan tambien lo admira. Maria admires to Juan, Rosa also him admires and the self Juan also him admires "I know what Maria, Rosa and Juan have in common: Maria admires Juan, Rosa also admires him, and Juan himself also admires him."
Because of principle (B)*, the structural arrangement of el mismo Juan 'Juan himself and lo 'him' in the last clause carries the information that we cannot understand el mismo Juan and lo so that it will be presupposed that they refer to the same individual; on the other hand, other factors force us to realize that the individual the two NPs will refer to is the same. The final result is that we can understand the sentence with the co-referential interpretation of el mismo Juan and lo. Because of the conflict between the information that different aspects of the sentence convey, though, the sentence has a somewhat odd character. What are the factors that force that both el mismo Juan and lo be taken to refer to Juan? We can distinguish the following five: (i) el mismo Juan refers naturally to Juan —the same person named Juan that has been referred to in the previous clauses—; (ii) this is reinforced by the fact that we are expecting to be told what Maria, Rosa and Juan have in common, and in the previous two clauses the subject NP refers to, respectively, Maria and Rosa, and so we expect the subject NP in the last clause to refer to Juan; (iii) in addition, the use of el mismo also forces el mismo Juan to refer to the same male named Juan that the previous clauses singleton set, which also determines an individual.
On the Interaction of Syntax-Semantics-Pragmatics
319
refer to, since the use of el mismo carries the presupposition" that the person in question has already been mentioned, (iv) The use of tambien in the last clause carries the presupposition that a property that is being attributed in the last clause has already been considered (the last clause can be seen as involving the attribution of either one of two properties: that of'admiring him', and that of 'being admired by Juan'); the repetition in the previous clauses of the attribution of the property of admiring Juan, makes that property completely salient in order to be the property about which tambien carries the presupposition (as opposed to the property of being admired by Juan); now, if tambien in the last clause carries the presupposition that the last clause is attributing the property of admiring Juan, that means that lo in the VP of the last clause should refer to Juan, (v) Finally, the fact that it has been announced at the beginning that we would be told what Maria, Rosa and Juan have in common, and that we have already been told that what Maria and Rosa both do is admire Juan, makes us expect that in the last clause we will be told two things: first (as we already mentioned above) what Juan does, and, second, that what he does is the same thing that Maria and Rosa do: to admire Juan. That means, again, that both Juan and lo will have to refer to Juan. All these factors force us to interpret the sentence so that we realise that both el mismo Juan and lo refer to the same individual. That is, all these factors together counteract the information conveyed by the binding principles on the basis of the structural relation between the two NPs. Notice, from the account of the (partial) acceptability of (12) that is provided, that an explanation is necessary for why a modification of (12) where by suppressing some of the features that we have said help the sentence to be regarded as good makes the sentence becomes less good. For instance if we omit el mismo or tambien as in (13a) the sentence becomes less good; it becomes even worse if, in addition, we do not mention one of the women as in (13b). Also, if we omit to announce that we are going to tell what it is that the three people have in common, as in (13c), the sentence is certainly worse than (12). (13)
a. Ya se que es lo que Maria, Rosa y Juan tienen en Already I-know what is it that Maria, Rosa and Juan have in comun: Maria admira a Juan, Rosa tambien lo admira y Juan lo admira. common: Maria admires to Juan, Rosa also him admires and Juan him admires "I know what Maria, Rosa and Juan have in common: Maria admires Juan, Rosa also admires him, and Juan admires him." b. Ya se que es lo que Maria y Juan tienen en comun: Already I-know what is it that Maria and Juan have in common:
11 In section 2 we introduced the term 'presupposition' as a feature of contexts. Here we use 'presupposition' in its most common sense. It is a derivative sense, definable in terms of the one in section 2: An expression E is said to carry a presupposition that P if and only if it would be non-felicitous to use E in a context where it is not presupposed that P.
320
From Words to Discourse
Maria admira a Juan, Rosa tambien lo admira y Juan lo admira. Maria admires to Juan, Rosa also him admires and Juan him admires "I know what Maria, and Juan have in common: Maria admires Juan, Rosa also admires him, and Juan admires him." c. Maria admira a Juan, Rosa tambien lo admira y el mismo Juan tambien lo admira. Maria admires to Juan, Rosa also him admires and the self Juan also him admires "Maria admires Juan, Rosa also admires him, and Juan himself also admires him." Similar explanations to the one given here for (12) could be given for sentences such as (4b-e), (5a), (5c) and (5d). In each of these explanations it would be essential to take into account how the semantic information provided by the binding principles (on the basis of the syntactic structure of the sentence) interacts with other sources of information about how to interpret the NPs in a sentence. Some of these other sources of information are semantic (such as the lexical information provided by a noun), some of them are pragmatic (such as the contextual information that determines about which property the word tambien 'also' carries a presupposition).12 "
REFERENCES Chomsky, N. 1981. Lectures on Government and Binding. Dordrecht: Foris. Chomsky, N. 1993. "A Minimalist Program for Linguistic Theory." In The View from Building 20:Essays in Linguistics in Honor ofSylvain Bromberger ed. by K. Hale and S. Keyser, 1-52. Cambridge: MIT Press. Evans, G. 1980. "Pronouns". Linguistic Inquiry 11. 337-362. Fiengo, R., R. May, 1994. Indices and Identity. Cambridge: MIT Press. Grice, P. 1989. "Logic and Conversation." Studies in the Way of Words. Cambridge: Harvard University Press. Grodzinsky Y., and T. Reinhart, 1993. "The Innateness of Binding and Coreference." Linguistic Inquiry 24. 69-101. Heggie, L. 1988. The Syntax ofCopular Structures. Ph.D. diss., University of Southern California, Los Angeles. Heim, I. 1982. The Semantics of Definite and Indefinite Noun Phrases. Ph.D. diss. Umass., Amherst. Heim, I. 1993. "Anaphora and Semantic Interpretation: a Reinterpretation of Reinhart's Approach." Ms. Higginbotham, J. 1992. "Anaphoric Reference and Common Reference." Unpublished manuscript. Macia, J. \997.NaturalLanguageandFormalLanguages.Ph.D.diss. MIT. Moro, A. 1991, "The Anomaly ofCopular Sentences." Venice Working Papers in Linguistics 8. Postal, P. 1970. "On Coreferential Complement Subject Deletion." Linguistic Inquiry 1. 439-500. Reinhart, T. 1983. Anaphora and Semantic Interpretation. London: Croom Helm Stalnaker, R. 1973. "Presuppositions." Journal of Philosophical Logic 2. 77-96. Stalnaker, R. 1974. "Pragmatic Presuppositions." Semantics and Philosophy, ed. by M.Munitz. NY: NY Univ. Press. Stalnaker, R. 1978. "Assertion." Syntax and Semantics 9: Pragmatics, ed. by P.Cole. 315-332.
12
For a more detailed treatment of the three kinds of counter-examples considered here, see Macia (1997). I wish to thank James Higginbotham, Irene Heim, Manuel Garcia-Carpintero, Howard Lasnik, Joe Lau, Alec Marantz, Gemma Rigau and Robert Stalnaker for comments and discussion. Financial support was provided by the research project PB96-1091-C03-3, funded by the DGES, Spanish Department of Education 13
321
16
CLITIC DOUBLING, NULL OBJECTS AND CLITIC CLIMBING IN THE SPANISH OF CORRIENTES Laura Colantoni, University of Minnesota, CONICET (Argentina)
1. INTRODUCTION Spanish clitics have been repeatedly studied, and it might seem difficult to say anything new about them. Nevertheless, I will reconsider the old problems with new data, and with a partly new approach. I will study clitic doubling (CD), null objects (NO), and clitic climbing (CC) in the Spanish of Corrientes, Argentina. These three phenomena have not been reported together in any other Spanish dialect, and, consequently have not been correlated. The hypothesis of the present paper is that the CD, NO and CC can be correlated using discourse-pragmatic properties (new referents, topicality), and the animacy scale (Garcia, 1975; Silverstein, 1976). These two factors are related broadly in this way: identifiable referents (Lambrecht, 1994: 109) will tend to be either not phonetically realized (NO) or the pronoun replacing them will not climb. The presence of new referents (Chafe, 1976; Prince, 1981; Lambrecht, 1994) and/or topicalized elements (Chafe, 1976; Lambrecht, 1994) will favor CD and CC. On the other hand, the higher the referent of the pronoun is in the animacy scale, the more likely that it will be doubled or climb.
322
From Words to Discourse
2. METHODOLOGY I analyze here the speech of six residents of San Luis del Palmar, Itati, and Beron de Astrada, located in the north of the province of Corrientes, Argentina. I collected the data during research for the Linguistic-Anthropological Atlas of the Argentine Republic, directed by O. Kovacci. The speakers (one man and one woman from each location) are lifelong residents of the location under study, have no secondary education, and range in age from 30 to 50 years. The interviews are two hours long and include identification questions (directed to study the vocabulary) and open questions about local customs, in order to elicit free speech. I studied CD, NO and CC for non-reflexive first singular (me), second singular informal (te) and formal (le),1 and third person (le, les, lo, los, la, las) object pronouns, of which there were 575 instances. This total includes every mention of one of these pronouns, and it was used to compare the relative frequency of each phenomenon. I described the linguistic environments in which the three phenomena occur, and I also examined the possible influence of the two extralinguistic variables (sex and location) considered in the original project. However, the methodology used may limit the predictions regarding these two variables.
3. CLITIC DOUBLING The duplication of the clitic by a lexical expression,2 shown in (1), is known as clitic doubling: (1)
Y no mevoy porque le tengomiedo al agua del Parana (SLm).3 And no me I go because CLoAT I-have fear of the water of the Parana. "And I don't go because I'm afraid of the river Parana."
The clitic can have a cataphoric referent, as in (1), or an anaphoric referent as in (2): (2)
A otros les gusta la zanahoria rallada (BAw). To others CLoAT like carrot shredded. "Others like shredded carrot."
In this section, I describe first (cf. 3.1) the frequency of CD by type of pronoun; then, I examine the two factors that favor CD, i.e. the animacy (cf. 3.2) and the status of the referent (cf. 3.3).
1
There were no instances of second person formal accusative pronouns. 1 will use 'lexical expression' in a broad sense, since all indirect objects (IO) and also some direct objects (DO) are prepositional phrases. 2
Clitic Doubling, Null Objects and Clitic Climbing
323
3.1. Type of Pronoun It has been observed that CD is more frequent with third person dative pronouns (Garcia, 1975; Lenz, 1920: 55; Fernandez Ramirez, 1987; Barrenechea and Orecchia, 1979; among others), where it is also considered obligatory. The present data are consistent with this general tendency in Spanish (cf. Figure 1); 78% of the instances of CD occurred with third person dative pronouns. It is important to note, however, that in Corrientes Spanish, dative pronouns are also used instead of accusative pronouns to realize a DO (which makes CD with dative pronouns even more frequent), a phenomenon known among Spanish grammarians as leismo.
10%
n=89
78%
Figure 1: Frequency of pronouns in CD
The third person dative pronoun doubles an IO in 46 instances (66.6%), and a DO in 23 instances (33.3%). This latter case is exemplified by (3): (3)
Hay muchas que no le quieren amamantar a sus hijitos (BAw). There are many-FEM- who not CLoAiwantto nurse their little children. "There are many mothers who don't want to nurse their children."
Here le is used for singular and plural referents, as it is in most of the instances. It is impossible to tell if this is a morphological or a phonological simplification, because in this dialect implosive /s/ is usually deleted or aspirated. The higher frequency of CD with dative pronouns can be considered a disambiguation strategy. Clitics are deictical forms (Garcia 1975:60), which implies that the listener must find the referent of the pronoun. First and second person pronouns' referents are generally situationally accessible; that is why, according to Garcia, they do not show number or gender agreement. Instead, the referent for third person clitics is usually in the discourse context, and therefore is not as accessible. The fact that third person accusative 3
1 indicate first the location, then the sex of the speaker.
324
From Words to Discourse
pronouns show gender and number agreement makes the identification task easier. Dative pronouns, however, especially in Corrientes where the number mark has disappeared, do not encode gender and number information. In consequence, doubling becomes a strategy to facilitate identification, rendering coherence to discourse. Even though duplication is more frequent with dative pronouns, it also occurs with accusative pronouns (11.2%), differentiating Corrientes Spanish from the Spanish in contact with Basque (Landa, 1992; Landa and Franco, 1995). This phenomenon, reported first for Buenos Aires or Rio de la Plata Spanish (Barrenechea and Orecchia, 1979; Jaeggli, 1986), is also found in other dialects (Silva-Corvalan, 1980; Lujan and Parodi, 1996), and it is shown in (4): (4)
Lo banan al muerto (BAw). CLAcc they-wash the dead . "They wash the body."
3.2. Animacy of the Referent Animacy of the referent is relevant in explaining the distribution of accusative and dative pronouns in CD. The former double more frequently inanimate than animate noun phrases (NP); there are 10 instances of CD with accusative pronouns, and in three of them an animate NP is doubled. However, in none of them the referent of the pronoun is [+human].4 On the other hand, dative pronouns show a clear tendency to refer to [+animate], and especially [+human]5 entities (cf. Table 1). Thus, it is possible to conclude that there is an almost complementary distribution between dative and accusative pronouns.
Table 1: Type of referent of dative pronouns [+Animate] [+human] [-human]
[-Animate]
Others
Total
44(63.8%) 9(13%)
14 (20.3%)
2 (2.3%)
69
4
Example (4) is interesting because the referent is an inanimate human. It seems that in the selection of accusative and dative pronouns the first factor to be considered is the animacy of the referent. 5 Fernandez Ramirez (1987:104) points out that dative case tends to be used with human referents in Latin. According to him, this fact became even more frequent in Spanish. 6 'Others' refers to examples as the following, where the animacy of the referent is unclear: Yo, cualquier cosa, lepido a la Santisima Cruz del Milagro (BAw) "If I have some problem, I pray to the Holy Cross of the Miracle."
Clitic Doubling, Null Objects and Clitic Climbing
325
Finally, 11 of the 14 dative pronouns that have an inanimate referent are found with two verbs: llamar (7 tokens) and decir (4 tokens), used with the meaning of 'to name', as in (5): (5)
Alprimario le llaman escuela (BAm). To the primary CLoAT they call school. "The primary school building is called school."
Garcia (1975: 292) proposes an interesting explanation for the use llamar with dative pronouns, which, according to her, is registered even in non-leista dialects. She compares llamar with two other verbs that can be used in the same context with dative pronouns: decir (to say) and denominar (to name). She argues that llamar shares some traits with the two other verbs, and that when it is used with a different meaning (like 'to name' instead of 'to call' as in 5), it tends to take dative pronouns. To summarize, CD is more frequent with dative than with accusative pronouns. The latter doubles more frequently inanimate referents, while the former tends to double [+animate] referents. Since dative pronouns are more frequent than accusative pronouns, it is possible to conclude that if the referent is high in the animacy scale, it is more likely for the clitic to be doubled.
3.3. Discourse-pragmatic Motivations for CD CD can be considered a strategy to render the referent of the pronoun more identifiable. All the cases of CD in this corpus are accounted for by two discourse-pragmatic (or cognitive) motivations. The first motivation occurs when the referent of the pronoun cannot be identified because it is unmentioned. Thus, the lexical item is used to introduce this new referent (Chafe 1976, Prince 1981; Lambrecht 1994),7 and CD arises.8 In 53 of 89 (59.6%) total instances (all of them in third person) the lexical phrase introduces a new referent, as in (6): (6)
Mird, mucha gente habla en guarani directamente. Vienen acd a la municipalidad donde yo trabajo, vienen gente, gente mayor asi que hablan guarani. Me vienen y me hablan en guarani y yo no le entiendo nada. Le digo a la otra chica que me traduzca lo que quiere decir (Iw).
7 For the present paper, I will use Lambrecht's concept of 'unidentifiable referents'. For a discussion about different proposals on new vs. given information, see Lambrecht (1994) and Gundel et al. (1993). 8 This interpretation would be consistent with theories that consider clitics as purely agreement elements, like verbal endings. It predicts that clitics would be doubled for the same reason that subjects are expressed, i.e in order to introduce a new referent or to avoid ambiguity.
326
From Words to Discourse
Look, many people speak Guarani directly. They come here to the City hall where I work, come people, people old, so they speak Guarani. Me they come and me speak Guarani and I do not understand nothing. CLDAT I say to the other girl for me translate what they mean. "Look, a lot of people speak Guarani straightaway. They come here, to the City Hall, where I work, people come, old people so they speak Guarani. They come and talk to me in Guarani and I don't understand anything. I ask the other girl to translate what they
In (6) the speaker is describing the problems that she faces every day at work with speakers of Guarani. Since she does not speak that language, she needs the assistance of a colleague. This colleague was not mentioned previously, so the clitic needs to be clarified with the introduction of the construction, a la otra chica. By contrast, the clitic underlined in the same example is not doubled, because it refers to gente (people), a noun already mentioned. The second motivation occurs when the referent might be identifiable but there is a set of candidates competing for the reference. In such cases, the clitic is doubled for a contrastive purpose (Chafe, 1976; Lambrecht, 1994).9 This motivation accounts for the remaining 36 out of 89 instances of CD, and occurs with first, second and third person pronouns. All the examples of CD with first or second person pronouns belong to this type; in these instances, the referent is always accessible to the listener (it is part of background information). Thus if the clitic is doubled, it is not to solve a referential problem, as in (7): (7)
Hay gente que hace mas o menos cincuenta o sesenta anos que desarrolla esta actividad, y, justamente me toco a mi (en) una etapa de las mas dificiles conducir la comision que se organizo [...] (Im). There are people who ago more or less 50 or 60 years who develops this activity, and, precisely CL-ME happen to me in a period of the hardest to guide the committee that CL-PASSIVE was organized. "There are people who have been working on this for fifty or sixty years, and it happened to me to have to lead the committee in one of the hardest periods. "
Here the clitic is doubled when the speaker centers the attention on himself, emphasizing the importance of being the leader in that situation, when there were many other candidates with more years of experience who could have been chosen instead of him. Cases of doubling for contrastive or identification purposes have something in common; in both contexts the speaker may find that the referent of the pronoun is not clear enough, so the 9
Lambrecht (1994: 289-91) partially rejects Chafe's concept of 'focus of contrast', because it is not realized by any grammatical category. He prefers to consider contrastiveness as a cognitive process, which is the point of view that I will adopt here.
Clitic Doubling, Null Objects and Clitic Climbing
327
lexical phrase is inserted to disambiguate. The difference between the two contexts is that, in the second case, there is some background knowledge of possible referents, and the speaker is choosing one of several candidates. Thus, if doubling is a way of disambiguating (and/or emphasizing), we should expect it to be more frequent when the pronoun dependency on the discursive context is greater. In fact this is what we observe: third person clitics are more doubled than second or first person pronouns (cf. Figure 1), and doubling is also more frequent when the lexical NP encodes new referents. These data show that, in at least one Spanish dialect, clitics are not doubled because of a tendency to pleonasm (Bello, 1980: 919;10 Fernandez Ramirez, 1987: 111; Kany, 1951:107), but because of a pragmatic (or cognitive) need (Silva Corvalan, 1980). This happens because clitics are deictic elements: the higher the deixis, the higher the frequency of doubling. In summary, discursive factors override the animacy of the referent, i.e. if the clitic has an inanimate referent, it will be doubled whenever there is an unidentifiable referent or a contrastive focus.
4. NULL OBJECTS Null objects have not been as intensively studied as the other two phenomena, perhaps because they are not as widespread as CD and CC in Spanish and in other Romance languages. Campos (1986) observes that NOs are accepted in standard Spanish when the DO is marked as [definite], but these are not the environments that I examine here. Kany (1951: 11.4) is the only author who mentions that NOs are very common in Latin American Spanish, especially in conversations, although his examples come from literary works. NOs in Spanish seem to be restricted to language contact situations. They have been reported for Spanish in contact with Guarani, as here (Krivoshein and Corvalan, 1987; Kovacci, 1987), and with other languages like Basque (Landa, 1992; Landa and Franco, 1995), Quechua (Suner and Yepez, 1988), and English (Silva-Corvalan, 1994). The present data reinforce earlier claims that NOs occur only in the specific environments described below.
4.1. Only DOs Are Not Overt All NOs, both in the present data and in the literature,11 are DO (Kovacci, 1987; Suner and Yepez, 1988; Landa, 1992; Landa and Franco, 1995). In addition, NOs appear, in the corpus,
10 It is interesting to mention that Bello (1980:919) offers an exception to the pleonastic use that may be interpreted as a case of contrastive focus. 1 ' Silva-Corvalan (1994: 124-5) finds a different pattern in Spanish in contact with English. According to her, all clitics are eligible for omission, even if the observed omission rate is low.
328
From Words to Discourse
only with highly transitive verbs, such as hacer (to make) (cf. 8), usar (to use), llevar (to carry), ver (to see), truer (to bring), pedir (to ask), sacar (to remove). (8)
Yesperan que se haga el feretro. Hace el carpintero (BAw). And they wait until CL-PASSIVE make-SUBJ the coffin. NULL makes the carpenter. "And they wait for the coffin to be ready. The carpenter makes it."
The second important aspect of this example is that the referent of the NO is a definite NP: el feretro (the coffin). This differentiates Corrientes Spanish from standard Spanish, in which NOs do not occur with definite NPs.
4.2. Type of Referent All 26 NOs in this corpus (Table 3) have inanimate referents (Suner and Yepez 1988; Landa 1992), as in (9).
(9)
Leo "ElLitoral". Me traen todos los dias (BAw). I read "El Literal". CL-lst-SING. NULL they bring every day. "I read "El literal". They deliver it to me every day."
4.3. Thematicity All the referents of NOs in Corrientes speech are highly thematic, as has been previously reported (Kovacci, 1987; Suner and Yepez, 1988; Landa, 1992), (cf. 10 and 8-9 above). Thus, the condition for a DO to be deleted is that it be textually identifiable (Lambrecht, 1994: 109). (10)
Y despues estd el curtidor, que es el que tiene a su cargo elproceso de curtir los cueros, para, a su vez, darle al talabartero, que es el que hace los productos (Im). And then is the tanner, who is who is in charge of the process of tanning the skins, to, later, give NULL CLDAT to the leather worker, who is the one who makes the products. "And then there is the tanner, the one who is in charge of tanning the skins, in order to give them to the leather worker who makes the products."
Clitic Doubling, Null Objects and Clitic Climbing
329
In summary, NOs are present in the opposite discourse contexts as CD, since they are found when the referent is easily identifiable, and also when it is [-animate].
5. CLITIC CLIMBING vs. NON-CLITIC CLIMBING Given a finite verb and a non-finite verbal form (infinitive or gerund), the clitic can be encliticized to the latter or procliticized to the former, as in (11), which is known as CC: (11)
Hay muchas mujeres que no le quieren amamantar a sus hijitos (BAw). There are many women who no CL-DAT want to nurse their little children. "There are a lot of women who don't want to nurse their children."
Variation in the clitic position is an old tendency in Spanish, as Keniston (1937: 89) points out. Descriptive grammars (Real Academia Espanola, 1973: 3.10.6) allow the two positions for clitics, without adding any further comments. Syntactic conditions for CC have been deeply studied, especially within the generative framework, and are briefly reviewed below. The hypothesis in this section is that syntax determines the necessary conditions for CC to take place. However, the variation in the position of the clitic is explained by semantic (cf. II), and, especially, by discourse-pragmatic factors (cf. III).
5.1. Syntactic Conditions It has been established that CC does not take place when the finite verb is impersonal (Lujan, 1980; Kayne, 1989), as in (12), and when there is some intervening material between the finite and non-finite verb, as the interrogative pronoun in (13): (12)
(13)
[...] Hay que saber interpretarla (Im). [...] Have to- IMPERSONAL-PRESENT know to interpret CLACC-fem. "[.. .]You have to know how to interpret it." No sabria como decirle (SLm). No I know -CONDITIONAL- to tell CL-DAT-2nd FORMAL. "I wouldn't know what to tell you."
However, there is no adequate syntactic explanation for the variation of the clitic position with the same verbal complex.
330
From Words to Discourse
5.2. Semantic Explanations Semantic explanations for the distribution of CC and non-CC constructions are based on the different meanings of the finite verb (Napoli, 1981; Myhill, 1988). If the finite verb keeps its original meaning, CC does not take place. However, if there is a change in meaning, CC arises. Semantic explanations have two main drawbacks: i) it is difficult to find in the same corpus enough tokens containing the same verb in CC and non-CC constructions (Myhill, 1988: 234235); ii) sometimes, the different meanings are too subtle to be either perceived or formalized. In the present data only tener que, poder, tratar de, and querer + infinitive were found in both CC and non-CC constructions (one instance of each). Examples (14) and (15) illustrate the difficulties of trying to account for the different meanings of the finite verb: (14)
(15)
Ese ya es media tonto porque no le tiene que dar a todo el mundo (BAw). That-MASC- already is a little silly because no CL-DAT- should give to all the world. "That person is kind of silly, because he shouldn't give money to everybody." Tengo que decirle las calles, el nombre de las calles (SLm). I have to tell CL-DAT-2nd FORMAL- the streets, the name of the streets. "I have to tell you the streets, the names of the streets."
Both cases are verbal phrases expressing obligation. Myhill (1988: 234-35) suggests that in CC constructions tener conveys 'forced separation'. The difference in meaning is not clear in the present data; the verb in (14) seems to convey a stronger obligation than the finite verb in (15). In (14) the speaker is talking about a generous person who may lose all his money if he continues lending money to his friends and relatives. In (15) the speaker is answering an interviewer's question, in what can be considered a hypothetical situation, since his answer is the apodosis of a conditional clause. However, the low frequency of these constructions in both Myhill's data and the present corpus makes any strong conclusion about semantic explanations for CC premature.
5.3. Discourse-pragmatic Factors It is not simple to explain the variation in CC and non-CC constructions, since it seems to be the result of the interaction of many variables, of which discourse-pragmatic ones seem to play the most important role. I examine here four discourse-pragmatic (or cognitive) factors: degree of animacy of the subject and the clitic (5.3.1); presence of a clitic sequence (5.3.2); topicalization (5.3.3), and emphasis (5.3.4).
Clitic Doubling, Null Objects and Clitic Climbing
331
5.3.1. Degree of Animacy: Subject vs. Clitic. This is the only discourse-pragmatic factor that has been considered in the literature. Myhill (1988) suggests that the relative degree of animacy of the subject and the clitic may explain the distribution of CC vs. non-CC. He (1988: 242) proposes an animacy hierarchy for CC according to which second person clitics climb with any type of subjects, first person clitics climb with third person subjects, and third person singular human clitics climb with any other third person subjects. Thus, according to Myhill's animacy hierarchy for Spanish, CC should be favored when the clitic is ranked higher than the subject. However, this tendency is not observed in the present data (cf. Table 2).
Table 2: Frequency ofCCfor third person clitics by type of subject: lst.p.s
CC
-•-•-•
2nd.p.s
3rd.s [+hum]
ls.p.p
1 50%
0
1 50%
0
2
7
9
66.6%
100%
77.8%
100%
Non-
1
0
2
0
CC
33.3%
Total
22.2%
2nd.
3rd.s [-hum]
3rd.p.p
Total
p.p
10
31
83.3%
83.8%
2
6
16.7%
16.2%
12
37
The present data show that the animacy of subject and clitic does not explain the distribution of third person clitics, since they tend to climb with any kind of subject. Indeed, the two examples with second person pronouns, which are the highest in the scale, both show CC. However, Myhill (1988: 244) indicates that his proposal suggests a tendency that may be overruled by discursive factors. Then, it could be said that if the clitic ranks higher than the subject, it can climb, but it does not need to climb. 5.3.2. Clitic Sequences The role of clitic sequences has not been previously analyzed. In the corpus, there are five examples with clitic sequences, and all show CC, as in (16): (16)
[ Q: Que hace el hombrepara que la mujer sepa que gusta de ella? "What does the man do to let the woman know that he likes her?"] A: Se le tiene que declarar (SLw).
332
From Words to Discourse
CL-REFLEX- CL-DAT he-has to declare. "He has to declare his love to her." The presence of a clitic sequence is, of course, a syntactic factor. However, the reason why the sequence climbs more frequently than the clitic alone may be pragmatic or even cognitive. A sequence of pronouns is easier to process, according to Garcia (1975: 473) if they are procliticized, because they are not in an immediate order as when they are enclitic. Garcia uses this explanation to account for the different distribution of clitic pronouns with imperatives and indicative and/or subjunctive forms. Her proposal can be easily extended to account for CC distribution. The clitic sequence that modifies the non-finite form, which in most cases forms a verbal complex with the finite verb, will also be processed more easily if it is encliticized, following the pattern of all other finite verbal forms. While I do not deal specifically with reflexive pronouns here (cf. Methodology), I include this observation, because clitic sequences seem to be an important factor to take into account in further studies of CC. 5.3.3. Topicalization and Focalization When the lexical object is mentioned before the verb, the clitic always climbs, as it is shown in
(17)
(18)
Si, el truco [...], que tambien yo no sejugar, asi que no mepida que le explique. Ahora, la loba, si. Bueno, eso si lopuedo explicar (BAm). Yes, the truco12 [. , .], which also I not know to play, so don't me ask that CL-DAT explain. Now, the loba, yes. Well, that yes CL-ACC I-can explain. "Yes, truco, I don't know how to play it, so don't ask me to explain it. Now, loba, yes. Well, I can explain it." Al intendente, le vamos a preguntar (Iw). To the mayor, CL-DAT we go to ask. "We are going to ask it to the mayor."
Examples (17) and (18) show instances of CC and CD. In (17) the clitic doubles a topicalized pronoun (eso), whose referent is activated (Gundel et al. 1993), because it is present in the immediate linguistic environment (la loba). Thus, there seems to be a reference chain between the referent (la loba), the demonstrative pronoun (eso), and the clitic (lo). In (18), intendente ('mayor') is a new referent introduced in a topicalized position. Since the referent is 'in focus' (Gundel et al 1993), it can be referred by an unstressed pronominal, such as the clitic le. 12
Truco and loba are card games.
Clitic Doubling, Null Objects and Clitic Climbing
333
It is important to remark that there are no cases of non-CC with a topicalized NP. Thus, climbing, as in the examples above, seems to be a strategy to shorten the distance between the clitic and its referent. 5.3.4. Emphasis Finally, CC can be considered a strategy to raise a highly topical form to the main clause for emphatic purposes,13 which is illustrated in (19): (19)
El pay e dicen que es algo que te hace una persona, te voy a decir (Iw). Thepaye they say that is something that CL-2nd-SING does a person, CL2ndSING I am going to tell. "They say that paye is something that a person does to you, I'll tell you."
Here the speaker may want to call attention to the listener because she is talking about a local tradition that might seem strange to someone coming from another province, and CC arises. In (20) and (21) we see a contrasting situation (cf. also 15 above): (20)
(21)
No sabria decirle (BAw). Not I-know-CONDITIONAL to tell CL-DAT-2nd-FORMAL. "I wouldn't be able to tell you." Y, no podria decirle, porqueyo, ahi, en esos temas (SLm). And, no I-can -COND. tell CL-DAT because I, there, in those topics. "I wouldn't be able to tell you, because I, about those topics "
In both examples there is no-CC; the clitic referent is highly thematic, situationally accessible (it is the interviewer), and non-emphatic. The informational focus is on the verbal construction. Therefore, the clitic can be buried in the subordinate clause. There are no cases of CC with nonemphatic clitics, as in (20-21). On the other hand, non-CC is registered in 14 out of 28 instances (50%) in similar contexts. Thus, if we add the instances of non-emphatic pronouns to those of impersonal verbal complexes (9 out of 28), it is possible to account for 82% of the instances of non-CC. In summary, CC is a complex phenomenon that can arise when certain syntactic conditions are met, e.g. the presence of a finite personal form plus an infinitive or a gerund, and/or the absence of interrogative pronouns between the finite and non-finite form. However, the variation in CC and non-CC with the same verbal construction is not accounted for by syntactic conditions. Semantic motivations based on the change in meaning of the auxiliary are
334
From Words to Discourse
interesting, but it is difficult to find examples that clearly demonstrate the different meanings of the auxiliary verb. Finally, discourse-pragmatic motivations account for most of the cases of non-CC (non-emphatic constructions) and explain a significant number of CC constructions.
6. CD, NOs, AND CC: EXTRA-LINGUISTIC VARIABLES Table 3 shows the number of tokens and raw frequency for CD, CC, non-CC and NOs, by speaker location and sex. Table 3: CD, CC, non-CC and null objects: frequency by location and sex. Location CD Sex N SL(w) 2
CC % N 8% 5
-CC % N 20% 1
SL(m)
11
15.9% 3
I(w)
20
12.4% 18 11.1%
I(m)
18
BA(w)
4.3%
Null % N 4% 0
11 15.9% 4
Other clitics Total of clitics % N % 17 68% 25
5.8% 40
58.1% 69
6
3.7% 4
2.5%
20% 3
3.3% 4
4.4% 3
3.3% 62
18
13.4% 9
6.7% 5
3.7% 8
6% 94
BA(m)
20
20.8% 6
6.2%
1% 7
7.3% 62
64.7% 96
Total
89
15.5% 44
7.7% 28
4.9% 26
4.5% 388
67.4% 575
1
113 70.3% 161 69%
90
70.2% 134
Since the extra-linguistic variables were not systematically considered in the original project, I will use the data in Table 3 only to point out two trends, which may merit further study. First, there may be a correlation between CD and CC and sex. Men seem to double clitics with a higher frequency, while CC is more frequent among women. Second, NOs present an interesting distribution by location. They are more common in Beron de Astrada, a town located in the interior of the province, considerably more isolated than the other two locations, where Guarani is widely spoken. Since NOs have been almost exclusively reported in language-contact situation, the degree of bilingualism of the speakers should be considered in a systematic way in future studies.
13 I follow Myhill's intuition (1988: 242) that CC is a way to raise to the main clause the "most prominent and topical NP", although he does not mention about emphasis.
Clitic Doubling, Null Objects and Clitic Climbing
335
7. CONCLUSIONS I have argued in this paper that CD, NOs and CC can be principally explained through the interaction of syntactic, semantic and discourse-pragmatic properties. CD and NOs represent two opposite patterns. CD is a strategy to render more identifiable the referent of the clitic pronoun, either because there is a need to introduce a new referent or because one referent is selected from a set of possible candidates. On the other hand, NOs have highly textually accessible referents (Lambrecht, 1994). Taking into account the position of the referent in the animacy scale, CD and NOs show a complementary distribution. CD is more frequent with animate referents, while NOs are found only with inanimate referents. CC is a strategy to raise the pronoun to the main clause, and does not take place when the referent of the pronoun is situationally accessible. CC is categorical when there is a topicalized NP doubling the clitic (which pairs CC and CD), and when there is a clitic sequence. The three phenomena are more frequent with third person referents, and NOs can only occur in this context. Finally, I have suggested that at least two extra-linguistic variables should be considered in further studies: sex and contact with Guarani.
REFERENCES Barrenechea, A. Ma. and Teresa Orecchia. 1979. "La Duplicacion de Objectos Directos elndirectos en el Espanol Hablado en Buenos Aires." Estudios Linguisticos y Dialectologicos, ed. by A. M. Barrenechea, 73-101. Buenos Aires: Hachette. Bello, A. 1980. Gramatica de la Lengua Castellana. Madrid: Edaf. Campos, H. 1986. "Indirect Object Drop." Linguistic Inquiry 17. 354-359. Chafe, W. 1976. Subject and Topic. New York: Academic Press. Fernandez Ramirez, S. 1987. Gramatica Espanola. 3.2. El Pronombre. Madrid: Arco/Libros. Garcia, E. 1975. Role of Theory in Linguistic Analysis: the Spanish Pronoun System. Amsterdam: North Holland Publishing Company. Gundel, J., Nancy Hedberg and Ron Zacharski. 1993. "Cognitive Status and the Form of Referring Expressions in Discourse." Language 69. 274-307. Jaeggli, 0.1986. "Three Issues in the Theory of Clitics: Case, Doubled NPs, and Extraction." The Syntax of Pronominal Clitics, Syntax and Semantics, v. 19 ed. by Hagit Borer, 15-42. New York: Academic Press. Kany, C.I 951. American-Spanish Syntax. Chicago: The University of Chicago Press. Kayne, R. 1989a. "Null Subjects and Clitic Climbing". The Null Subject Parameter ed by O. Jaeggli and K Safir 239-261. Dordrecht: Reidel. Keniston, H. 1937. The Syntax ofCastillian Prose. The XVI Century. Chicago: The University of Chicago Press. Kovacci, O. 1987. "Algunos Rasgos Foneticos y Gramaticales Observados en la Provincia de Corrientes". Documentos de PREDAL Argentina 2. 14-20. Krivoshein de Canese, N. and Graziella Corvalan. 1987. El Espanol del Paraguay. Asuncion: Centre Paraguayo de Estudios Sociologicos.
336
From Words to Discourse
Lambrecht, K. 1994. Information Structure and Sentence Form. Cambridge: Cambridge University Press. Landa, A. 1992. "Conditions on Zero Object Anaphora in Basque Spanish". Paper read at the second International Conference on Spanish in contact with other languages, Minneapolis, MM. Landa, A. and Jon Franco. 1995. "Two Issues in Null Objects in Basque Spanish." Grammatical Theory and Romance Languages ed by Karen Zagona, 159-168. Amsterdam: John Benjamins. Lenz, R. 1920. La oracion y sus partes. Madrid: RFE. Lujan, M.1980. "Clitic Promotion and Mood in Spanish Verbal Complements." Linguistics 18. 381-484. Lujan, M. and Claudia Parodi. 1996. "Clitic Doubling and the Acquisition of Agreement in Spanish." Perspectives on Spanish Linguistics ed by Javier Gutierrez and Luis Silva. UCLA. Myhill, J.I988."Variation in Spanish Clitic Climbing." Synchronic and Diachronic Approaches to Linguistic Variation and Change ed by T. Walsh, 227-250. Washington D.C.: Georgetown University Press. Napoli, D. 1981. "Semantic Interpretation vs. Lexical Governance: Clitic Climbing in Italian." Language 57. 841887. Prince, E. 1981. "Toward a Taxonomy of Given-New Information." Radical Pragmatics ed by Peter Cole, 223255. New York: Academic Press. Real Academia Espanola.1973. Esbozo de una Nueva Gramdtica de la Lengua Espanola.Madrid: Espasa Calpe. Silva-Corvalan, C. 1980-1981. "La Funcion Pragmatica de la Duplication de Pronombres Cliticos." Boletin de Filologia de la Universidad de Chile 31. 561-570. Silva-Corvalan, C. 1994. Language Contact and Change. Spanish in Los Angeles. Oxford. Clarendon Press. Silverstein, M. 1976. "Hierarchy of Features and Ergativity." Grammatical Categories in Australian Languages ed by R. M. N. Dixon, 112-171. Canberra: Institute of Aboriginal Studies. Suner, M. and Maria Yepez. 1988. "Null Definite Objects in Quitefio." Linguistic Inquiry 19. 511-519.
337
17
SPANISH No, Si: A PARTICLE OF POLITENESS Robert E. Vann, Western Michigan University Joan Busquets, Universite de Bordeaux Dale A. Koike, University of Texas at Austin
1. INTRODUCTION In everyday conversation, particles of politeness (Brown and Levinson, 1987) serve to indicate how addressees perceive and interpret the pragmatic intents of previous utterances. Such is the case of the Spanish response no, si 'no, yeah', a previously uninvestigated discourse marker that this paper examines as a particle of politeness in conversational interaction. The no, si answer to a given utterance in Spanish raises several questions. First, in which contexts of conversation is it appropriate to produce answers with both polarity terms no and si? Second, when used appropriately, what are the functions of no, sil We claim that variants of no, si are used according to the structural properties of the antecedent speech act. More specifically, the particle illustrates different communicative strategies in response to the intentions and expectations that a speaker may infer in the antecedent utterance. The grammaticalized forms of no, si in Spanish provide an indirect response to a previous question or statement, with a set of pragmatically determined meanings. Our semantic-discursive analysis demonstrates that, in terms of structure, a form of no, si occurs in our data as a response to any of five contexts: (a) a question that appears to offer an alternative; (b) a negative question; (c) a non-question (i.e., a statement); (d) an individual's own utterances (internal use); and (e) multiple questions.
338
From Words to Discourse
2. DATA AND STRUCTURAL ANALYSIS The quotations presented in this investigation were drawn from various sources, but mainly from 58 conversations between a linguistic researcher and Spanish speakers in Barcelona, Spain, usually concerning the topic of language use.1 In all quotations, (I) represents the interlocutor and (X) signifies the respondent.
2.1. Discourse Context 1 This discourse context contains a Yes-No-Question (YNQ) and an alternative question (AltQ) in which the second alternative is an implicit or not, in the form of a stylistic variant of o que (or what). X's use of no, si is to confirm or replace what X perceives I's presuppositions to be. Quotation (1), [S1-11, p. 1] (I) £ Te ha costado esto del didlogo mucho o que [= o no]? "Was the dialog hard for you or what?" (X) No-si, porque"No yeah, because-" (I) ^Porque? "Because what?" (X) Mas que nada porque hay muchas eh em, o sea situaciones que nunca, que no lo utilizaria yo nunca. "More than anything else because there are a lot of, um, situations in which I would never use it." (I) ^Porque no hablas castellano mucho o porque tenias que difer- diferir entre como hablas tuy como hablan los demds? "Because you don't speak Spanish very often or because you had to differentiate between how you speak and how others speak?" (X) Si, no como hablan los demds, no, dependiendo de de diferentes, o sea claro que no, diferentes, entes sociales, diferentes closes, todo el mundo supongo que habla mejor o peor. "Yea, not how other people speak, not, depending on different, I mean, of course not, different, social groups, different classes; I suppose everybody speaks a little better or worse." (I) Alfinal, j/0 has hecho como lo dirias tu o como lo dicen los demds? "In the end, did you do it like you would say it or like other people say it?" (X) No, como creo que lo pueden decir, o como seria menos incorrecto. ' We would like to acknowledge the aid of the Program for Cultural Cooperation between Spain's Ministry of Culture and United States' Universities, subvention #1490, whose support of Professor Vann's research allowed him to gather the data on which much of this investigation is based.
Spanish No, si
(I)
339
"No, like I think others could say it, or however it would be less incorrect." I Vale! "OK!"
2.2. Discourse Context 2 This discourse context contains a question-answer pair (QAP) or statement-response pair (SRP) in which negation appears overtly expressed in the question or statement. Here, the no, si response expresses denial of the assumption or expectation that X infers from the negated Q and introduces an explanation that corrects the perceived implicature. Quotation (2), [SI-21, p. 4] (I) Pero ^tienes la impresion que la mayoria de los catalanes piensan como tu, o sea... que son pacificos, y que no no les gusta la guerra, ni la violencia? "But, do you get the impression that most Catalans think like you do, I mean, that they are pacifists and that they don't like war or violence?" (X) Yo creo que hay de todo como todo, j,no?, pero creo que... una inmensa mayoria, es bastante pacifica. "I think that there is a little bit of everything like with anything, you know? But I think that the vast majority of people are pretty pacifist." (I) ^Y los catalanistas tambien? "And the Catalanists too?" (X) Aaaa, creo que si. Siempre hay esa rama radical y tal, pero no encuentro que sean estos una parte representativa de... de los catalanes, esa parte radical y que pueden provocar las destrozas o lo que sea sino, creo que el Catalan es mas bien pacifista, pacifico. "I think so. There is always that radical branch that can provoke trouble, but I don't find those people representative of all Catalans. I think that Catalan people are generally pacifists." (I) O sea, no /no tienes una buena opinion de los catalanistas? "So, you don't have a good opinion of the Catalanists?" (X) No, si si, de los catalanistas si, o sea yo soy catalanista. Lo que no tengo buena opinion es de esta gente catalanista o que se hace llamar catalanista, que...aaaa, bajo ese nombre pues..., que siiiiyo que se, puede llegar a matar a alguien por ser castellano o par ser... espanol o por ser algo asi. "No, I do! Regarding the Catalanists definitely, I mean, I am Catalanist. The people I don't have a good opinion of are the Catalanists, or the people who call themselves Catalanists, who, under that name, well, I don't know, can go so far as to kill someone for being a Castilian or for being Spanish or something like that." (I) lO sea, no los radicales? "So, you don't have a good opinion of the radicals?" (X) Exacto. "Exactly."
340
From Words to Discourse
2.3. Discourse Context 3 In this discourse context no, si appears as an acceptance of and agreement with the propositions implicated in I's non-Q utterance. This quotation is drawn from a meeting between a supervisor (I) and a university teaching assistant (X), both of whom are Mexicans but residents of the U.S., speaking about the feedback the supervisor has for the teacher after a class visit. In this quotation, the use of no, si is related to an acknowledgment of the proposition contained in I's utterance. Quotation (3) (I) Lo hiciste muy muy bien en el ejercicio depreguntas personales. "You did it very well in the exercise on personal questions." (X) Que bien. "That's great." (I) Eh, quizds solamente, bueno esto no tanto en las preguntas personales, pero en la section de cuando ellos hicieron unas traducciones. "Eh, maybe just, well not so much in the personal questions, but in the section where they did the translations." (X) Ah, si, hicieron unas traducciones. "Ah, yes, they did some translations." (I) Quizd, lo que m—...para mi hizo unpoquito defalta es que ellos repitieran a coro, por ejemplo 'eran las cinco', mm 'cuando vivia en Lima ensenaba'. Mas que nada para que ellos acostumbren a oirla. 'Vivia', 'ensenaba'. "Maybe what... in my opinion was lacking is that they repeat in chorus, for example, 'it was five o'clock', 'when I lived in Lima I used to teach'. More than anything else so they get used to hearing it. 'I used to live there', 'I used to teach'." (X) No, si, es cierto, si. "No, yeah, it's true, yes."
2.4. Discourse Context 4 The fourth discourse context in which no, si appears is an internal use in which X re-affirms his own intentions or statements. Quotation (4), [SI-11, p. 11] (I) Pues bueno yo estoy sin preguntas. £ Quieres verlas y si quieres ampliarme mas pues... ? "Well I'm out of questions. Do you want to see them and if you want to add more, well.,.?" (X) No no (risas). "No no (laughter)."
Spanish No, si
(I) (X) (I)
(X)
341
I Ya estds. ya estds! "So you're done then, you're done!" Si, ^no?, ah no no, si. "Yes, aren't I? No, well, yeah." jBueno tu mismo! ^Ha habido algo que has querido comentar algo mas y no has podido? ^o ya estd? "Well, it's up to you! Was there something that you wanted to comment more on but weren't able to, or is that all?" No, en principio. "No, theoretically."
2.5. Discourse Context 5 This discourse context contains separate answers to two independent questions in the context of overlapping turns (<>). That is, X responds to each Q as I poses it, interrupting I's formulation of Q2 with an answer to Ql. Quotation (5), [Sl-18,p.3] (X) No tenemos moneda... "We don't have currency..." (I) ;_Te molesta esto? ^Te gustaria tener una moneda catalana? <> /Que antes la habeis tenido! "Does it bother you? Would you like to have a Catalan currency? <> You had it before!" (X) No. Si. Eeeeh...a mi me es igual la, la moneda, ^sabes?, hay cosas mas importantes que la moneda, ^no?, cosas muchisimo mas importantes que la moneda. "No. Yes. Eeeeh...! don't care about the, the currency, you know? There are more important things than currency, right? things much more important than currency."
2.6. Discourse structure According to our data, grammaticalized variants of no, si can occur only in four of these five discourse contexts (the first four). Semantic analysis reduces these four contexts to three discourse structures. Discourse Context 1 concerns a question or statement without explicit negativity (AltQ) whose structure is given in the tree in Figure (la), where a and (3 are propositions expressed by the utterances produced by the speaker (S) and the addressee (A) in a given dialogue (D). The second discourse structure contains the presence of no in a given question with the proposition y. In this case of polarity reversal or expectation contravention (Michaelis 1994), seen in the tree in Figure (Ib), the structure reflects a correction of all or some previously stated propositions. In Figure (Ic), no, si appears in contexts in which the addressee
342
From Words to Discourse
seeks agreement (acceptance) with the speaker's inferred suggestions and in contexts in which the addressee seeks reaffirmation of his own statements (internal use of no, si).
S:a?
[\
;. R
^.aoque?
*
(t nno-si) °M>-"
S
:a?
S: a
„
n 3(Y)
|
a ALTQs
A: (3
b.
jP
^no' si}
{no, si}... I
NEG-Qs
A
C. Non-Q/Internal
FIGURE 1. - Discourse structure of no, si
3. FUNCTIONAL ANALYSIS Having identified the structures in which no, si occurs in our data, we now turn to function, where we see that in different discourse structures, no, si plays different roles in conversation. Variants of the same linguistic form add different meanings to the context of conversation, depending on the expectations and suggestions implicit in the speaker's utterances. No, si is sensitive to the general structure of the discourse and, in different contexts, it introduces different discourse functions. One function introduced by no-si is Explanation. Explanation is illustrated in Quotation (1), which represents Discourse Structure (a) of Figure 1. Excerpt from Quotation (1), [S 1-1 1, p. 1]
(I)
/; Te ha costado esto del didloso mucho o que [= o nol?
(X)
"Was the dialog hard for you or what?" No-si, porque"No yeah, because-"
In order to satisfy the underlying o no 'or not' of the question, the addressee is, in some way, forced to introduce a causal explanation in the answer. Bolinger (1978:90) points out that the effect of or not in a question is to add a "tinge of greater negativeness". Thus, the speaker is suggesting a no as a response. We believe the addressee in (1) perceives this negativity, which is why he uses a no-si as a response to initiate a clarifying Elaboration. Quotation (2) illustrates, several discourse relations stemming from Contrast.
Spanish No, si
343
Excerpt from Quotation (2), [SI-21, p. 4] (I) O sea, no /no tienes una buena opinion de los catalanistas? "So, you don't have a good opinion of the Catalanists?" (X) No, si si, de los catalanistas si, o sea yo soy catalanista. Lo que no tengo buena opinion es de esta gente catalanista o que se hace llamar catalanista, que...aaaa, bajo ese nombre pues..., que siiii yo que se, puede llegar a malar a alguien par ser castellano o por ser... espanol o parser algo asi. "No, I do! Regarding the Catalanists definitely, I mean, I am Catalanist. The people I don't have a good opinion of are the Catalanists, or the people who call themselves Catalanists, who, under that name, well, I don't know, can go so far as killing someone for being a Castilian or for being Spanish or something like that."
In Correction, no, si denies the presuppositions implicated by the interviewer's question, and corrects the implicated position that 'X does not have a good opinion of Catalanists'. This opens the door to Elaboration and Commentary. X presents additional detail about Catalanists and continues, revealing a personal attitude concerning the subject matter. These discourse relations stem from Contrast. Surprisingly, however, no, si does not always entail Contrast. The same particle that can be used to implicate rejection can also entail Acceptance, as in (3). Excerpt from Quotation (3) (I) Quizd, lo que m—...para mi hizo un poquito defalta es que ellos repitieran a coro, por ejemplo 'eran las cinco', mm 'cuando vivia en Lima ensenaba'. Mas que nada para que ellos acostumbren a oirla. 'Vivia', 'ensenaba'. "Maybe what... in my opinion was lacking was for them to repeat in chorus, for example, 'it was five o'clock', 'when I lived in Lima I used to teach'. More than anything else so they would get used to hearing it. 'I used to live there', 'I used to teach'." (X) No, si, es cierto, si. "No, yes, it's true, yes."
In Acceptance, no, si provides no new information nor evidence that there is a need for clarification in the sequential position in the dialogue. It is a marker of implicit acceptance in a highly specialized set of circumstances that must take into account the roles and expectations of both interlocutors. Both conversants remain coordinated on what constitutes the common ground.2 This is an issue that raises new questions for our analysis. What are the primary cues that speakers might use in identifying no, si as rejection in Figure 1 (b) and as implicit acceptance in Figure 1 (c)? The main difference lies in the implications arising frpm the previous utterances in each excerpt. The content of the antecedent utterances (what is requested/asserted) and the form of antecedent utterances (how information is 2 Schwenter (personal communication, 1999) points out that the acceptance/common ground reading is supported by the epistemic attenuation that the no-si response conveys in Quotation (3), which implies less affirmation.
344
From Words to Discourse
requested/asserted) are essential to indexing the discourse functions of the no, si response, which are given in Formula (1). (1)
Discourse relations = «0acceptance/rejection + '"'affirmation/correction
3.1. Politeness Reinterpreting these same discourse relations in terms of politeness, our pragmatic analysis considers issues of alignment and preference structure in accounting for how no, si gives feedback to interlocutors regarding how an utterance or set of utterances has been understood. Following principles of politeness in Brown and Levinson (1987), Goffman (1981), and Schegloff et al. (1977), we demonstrate that the use of no, si in our data represents a dual conversational management strategy of (a) acknowledging a perceived FTA (or dispreferred first) and reacting to it by either correcting the interlocutor in a move of negative politeness (a dispreferred second) or by accepting previous utterances in a move of positive politeness and (b) moving toward alignment, at times hedging the responses that follow. As such, no, si functions as an evaluative expression of footing in which the user can signal that some element of disharmony was perceived and move to re-establish cooperative harmony, standing ground but avoiding disagreement at the same time. These relations are given in Formula (2). (2)
Politeness = «0ackn0wiedge perceived FTA, initiate counter FTA+ •S'hedge, alignment
Brown and Levinson (1987) specify positive face, the desire to be approved of, and negative face, the desire not to be imposed upon. In this framework, one refers to acts that threaten another's face (FTAs). All the quotations in which no, si appears reveal a response to a previous utterance that has some element of provocation or potential criticism, something that could be heard as an FTA. Such a questioning style, though very frequent in certain modes of discourse and in certain speech communities (cf. Tannen, 1981), can still be interpreted at times as an imposition to the addressee's face (Attardo, 1997). It follows then that the addressees can choose to ignore the FTAs and simply provide the information requested or they can signal that some element of disharmony was perceived and try to work toward regaining full alignment. Such a signal is provided in no, si. It is clear that these concepts obtain for all four of the Discourse Contexts under examination, though for brevity we detail only one representative quotation below. Consider the discourse in Quotation (6), a quotation from Discourse Context 2 that illustrates politeness and alignment strategies common to all the Discourse Contexts. In Quotation (6), X corrects I's mistaken assumption of X's ideas but attenuates his disagreements on the importance of being born Catalan.
Spanish No, si
345
Quotation (6), [SI-27, p.l] (I) Entonces, en orden de importancia, ^cudles serian los requisites que uno tendria que tenerpara ser Catalan? "So, in order of importance, what would the requirements be in your mind for someone to be Catalan?" (X) ^En orden de importancia? "In order of importance?" (I) Si, lo mas importante, segundo mas, tercero... "Yeah, the most important, second most important, third..." (X) Lo mas importante, pues supongo que considerarse Catalan ^no?, hablar bien la lengua, escribirla bien y no se, pensar en Catalan. "The most important thing, I suppose, is thinking of oneself as a Catalan, you know? then speaking the language well, writing it well, and, I don't know, thinking in Catalan." (I) Vale, numero uno es considerarte Catalan. "Ok, number one is thinking of oneself as a Catalan." (X) Si. "Yes." (I) Dos, ^que es? "What is number two?" (X) Hablar, escribir y pensar...aunque si eres castellanoy te consideras Catalan, tambien, no se...y luego, moverte en un entorno. "Speaking, writing, and thinking...although if you're Castilian and you think of yourself as a Catalan, that counts too. I don't know. And also moving in Catalan circles." (I) £ Y nofigura para nada nacer en Cataluna? "And being born in Catalonia doesn't count at all?" (X) Hombre, no creo que quiera decir mucho, <> "Well, I don't think it means much" <> (I) iEh? "Huh?" (X) Pero tambien. Que no creo que quiera decir mucho pero tambien<> "That too, but I don't think it counts all that much, but also"<> (I) ^Ydondeestd? "And where is it?" (X) O sea, tu puedes venir defuera, ^ no ? y vivir muchos anos aqui y considerarte Catalan. "I mean, you can come from outside, right? and live many years here, and think of yourself as a Catalan." (I) O sea, lo de nacer ^o es numero cuatro, o no es ni, ni importante? "So, being bom here, is it number four, or is it not even important?" (X) No, si que entra, lo quepasa es eso. "No, it does count, what happens is that." (I) Que es despues de considerarte Catalan <> "That it's after thinking of oneself as a Catalan" <>
346
From Words to Discourse
(X)
(I) (X)
Si. "Yes." Hablar <> y entorno. "Speaking Catalan and a Catalan background." Si, yo creo que si. "Yes, I think so."
In Quotation (6), I openly asks X if being born in Catalonia doesn't matter at all in defining oneself as Catalan, implying in his question that he himself (I) believes it is quite important. X first answers that it is not important, and has to repeat this answer again for I. But I does not let the issue rest, and presses X to tell him where she would place one's birthplace in the list of defining characteristics. X answers by saying something indirectly related to the question, but I presses to ask if birthplace would just be low on the list (4th), or if it isn't even important. After hearing the same question several times, X counters with no, si and openly corrects I's proposition that birthplace is not even important in a move of negative politeness. Having been pushed finally to admit the importance of birthplace in defining oneself as Catalan, X then attenuates or hedges her own admission by immediately starting an explanation, probably of the relatively minor importance of birthplace.3 The hedging function of the explanation introduced by no, si here is a move of positive politeness. Once he finds common ground that birthplace actually does matter, and once he perceives a counter FTA in the illocutionary force of no, si, I finally backs off and capitulates to X's previously stated position that thinking of oneself as a Catalan is the most important criterion. The situation is resolved when I proposes the compromise regarding the relative importance of criteria, to which X agrees.
3.2. Alignment As stated earlier, the no, si particle can be a linguistic signal that some threat to face was perceived in a previous utterance. Of course, the signal of the perceived threat can be, in and of itself, a counter threat to the person who made the initial threat, and as such it is sometimes redressed to maintain equilibrium in the politeness of the discourse. Thus the use of no, si in Quotation (6) allows its user to stand ground with the other participant(s) of the dialogue and to avoid disagreement with them at the same time. Due to these functions, the particle can be seen 3
The admission is a strategy of positive politeness known as Avoid Disagreement, in which the desire to agree or appear to agree with the addressee leads also to mechanisms for pretending to agree, or instances of 'token' agreement (Brown and Levinson, 1987:113). One example Brown and Levinson cite (p. 114) is: (A) What is she, small? (B) Yes, yes, she's small, smallish, um, not really small but certainly not very big. The addressee (B) obviously does not believe the referent is small, but replies at first in agreement with speaker (A), only then to reveal B's true opinion and hedge it.
Spanish No, si
347
as a local 'aligning action' (Stokes and Hewitt, 1976); that is, any action that is directly relevant to the interlocutor's adjacent utterance and used to clarify or correct the message. From an alignment perspective, the sequence of events shown in Quotation (6) is an excellent case of co-constructed meaning negotiated by the speaker and the hearer in a clearly defined social context, i.e., that of an interview. The importance of the socially variable negotiation of meaning is highlighted by Bourdieu (1991), who proposes a dynamic model that allows simultaneously for agency and interpretation of different meanings in varying social and linguistic markets. In this model, extralinguistic messages (including FTAs) need not be mindfully implied to be inferred, and meaning will be interpreted differently by different interlocutors in different social and linguistic markets. Thompson (Bourdieu, 1991:39, italics in original) reminds us that "each recipient helps to produce the message which he perceives and appreciates by bringing to it everything that makes up his singular and collective experience." X is a free agent who can choose to react to the same stimulus in varying ways. We propose that one such reaction has become conventionalized for some Spanish speakers. Thus, X's own previous experience with the conversational style that I uses in Quotation (6) (i.e., X's pragmatic competence) actually contributes to the status of the repeated question as an FTA for her. Once X goes on record to recognize this discourse style, she co-produces it, defending herself with a counter FTA (no, si), and sets into motion what, in her speech community (as well as those of other Spanish speakers) is a conventionalized, in-your-face negotiation of meaning. That is, the repeated questioning is not actually an FTA until X recognizes it as such on record. X's use of no, si instead of simply saying si consigns overt FTA status to the repeated question and initiates a patterned response mechanism, whose culturally recognized function is to defend oneself baldon-record and then try to find a common ground. In the case of Quotation (6), I recognizes this signal and joins X to negotiate the repair effectively.4 Within adjacency pairs (Schegloff et al., 1977), the first part generally encodes preferred seconds, and addressees can recognize and use the encoded information to provide the preferred response in order to maintain harmony. When the preferred response is not forthcoming, then dispreferred seconds typically are marked by some delay, pause, or discourse marker. In the dialogues examined, the no, si particle appears to mark a dispreferred second of an adjacency pair. For example, in the context of Correction, it can be seen as a presequence to an answer that is generally not one that the interlocutor would usually want to hear. The data we have discussed, however, illustrate a slight variation to the dispreferred second described above. We also actually see a dispreferred first, in the form of a question or statement that provokes or challenges the addressee in some way. 4
It is worth noting that for another individual in this or other speech communities, the repeated question might not be considered even a weak FTA, or it might be taken as an even stronger one. In either case, if the individual were to choose not to respond with some signal such as no, si, then the repeated questioning would not attain the coconstructed status of an FTA, and the common ground repair script would not be played out. This point supports the claim that no, si as a mechanism of politeness is culturally bound.
348
From Words to Discourse
In the cases we have seen, the respondents are put into a position to either affirm something that they are actually reluctant or embarrassed to say or to correct some element of the first part of the pair, expressed by the interlocutor. In all cases, the respondents give dispreferred answers, indicated by no, si. This idea correlates with the politeness reading, since the preferred content of both parts of the QAP or SAP should be of equal politeness polarity. Since I's first part is not perceived to be a polite expression, and X's initial reaction is not preferred, the no, si particle marks the responses as dispreferred expressions. We propose that no, si can displace, or likewise introduce, a dispreferred response. In our data, the no initially signals that the discourse between participants is out of alignment and that what the speaker of no, si heard is not a preferred statement or question. The no is then complemented by the si, which is seen as a move toward alignment with the other participant. This reading is related to Pomerantz's (1984) finding that dispreferred utterances are often comprised of two conflicting propositions, reflective of attention to I's positive and negative face.
4. CONCLUSION Now we can address the issue of direct and indirect answers to a given question. Looking at our linguistic data, we notice that in all cases where it is appropriate to use a no, si answer, a single si 'yes' could be cooperative as a response, serving as a direct affirmative response. It is usually assumed that a direct answer is the most cooperative response in conversational interactions. We believe that the no, si particle shows how in certain circumstances, an indirect answer can be more cooperative than a direct one. We conclude that, more than the mere combination of opposing adverbs, no, si is a compound particle with a set of specialized yet delineable pragmatic functions related through implicature. An important resource for the study of politeness and alignment strategies, this highly specialized particle would seem to have formal surface parallels in neighboring Romance languages such as Catalan (no, si) and French (mais, si). We support further investigation of such significant, yet underinvestigated, phenomena to better describe and understand the functions of language and to further pragmatic theory.
REFERENCES Attardo, S. 1997. "Locutionary and Perlocutionary Cooperation: The Perlocutionary Cooperative Principle." Journal of Pragmatics 27.753-779. Bolinger, D. 1978. "Yes-No Questions Are Not Alternative Questions." In Questions ed. by Henry Dordrecht: Reidel.
Hiz, 87-105.
Spanish No, si
349
Bourdieu, P. 1991. Language and Symbolic Power. Ed. and intro. by J. B. Thompson. Trans. By G. Raymond and M. Adamson. Cambridge: Harvard University Press. Brown, P. and Stephen Levinson. 1987. Politeness: Some Universals in Language Usage.Cambridge: Cambridge University Press. Goffman, E. 1981. Forms of Talk. Philadelphia: University of Pennsylvania Press. Michaelis, L. 1994. "Expectation Contravention in Vietnamese." Journal of Pragmatics 21. 1-36. Pomerantz, A. 1975. Second Assessment: A Study of Some Features of Agreements/Disagreements. Ph.D. diss. University ofCalifornia, Irvine. Schegloff, Emanuel, Jefferson, Gail and Harvey Sacks. 1977. "The Preference for Self-Correction in the Organization of Repair in Conversation." Language 53.361-382. Stokes, R. and John Hewitt. 1976. "Aligning Actions." American Sociological Review 41. 838-849. Tannen, D. 1981. "The Machine-Gun Question: An Example of Conversational Style." Journal of Pragmatics 5. 383-397.
This page intentionally left blank
351
NAME INDEX
Abusch, D., 291,293,301,304 Acero, J., 4, 5, 18 Acquaviva, P., 279, 286 Aliaga, F., 135, 136, 142-144, 294, 304 Ahseda, A, 210,233,236 Almog, J. 203,233 Alvarez, R, 107,118 Anderson, S, 71, 92, 195, 198, 233, Asher.N., 195,233 Attardo, S., 344, 348 Bach, E., 123,133,134 Baker, C, 238,240,261 Baker, M., 281,286 Banfield, A., 267,286 Barrenechea, A., 323, 324, 335 Barroso, H., 73,92 Barwise, J., 122, 133, 211, 212, 216, 233 Beck, S., 190, 193 Beghelli, F., 205, 233, 293, 304 Bello, A., 5, 18, 23, 36, 48, 59, 217, 218, 223, 224, 233, 290, 297, 304, 327, 335 Benjamin, C., 103,118 Bennett, M., 200-202, 233, 304 Berman, S., 127, 133, 293, 304 Bertinetto, P., 73, 83, 92 Bird, G., 118 Blakemore, D, 112,118. Blass, R., 118 Bolinger, D., 210, 257, 258, 261, 266, 267, 273, 286, 342, 348 Bonzon, P., 233 Bosque, I., 4, 5, 11, 15, 58, 61, 70, 92, 141, 144, 162, 166, 167, 178, 221, 233, 247, 261, 263, 272, 279, 286 Bourdieu, P., 347,349 Braun.D., 203,233 Bravo, A., 75, 92 Brown, P., 17,18,346,349 Brucart, J.M. , 220, 233
Briz, A., 4,18 Buhler, K., 224, 227, 232, 233 Bull, W., 5, 18 Bustos, E., 3, 4, 11, 18, 136, 138, 140-144, 294, 304 Butt,!, 103,118 Buvac, S., 211,233 Bybee, J., 75,92 Campos, H., 327,335 Cardinaletti, A., 70, 92, 270, 286 Carel, M., 266,267,287 Carlson, G., 12, 18, 25, 36, 160, 178, 266, 277, 287 Carnap, R., 199,204,233 Carston, R., 176,178 Chafe, W., 17,18,155,157,326,335 Chierchia, G, 28, 36, 129, 133, 160, 178 Chomsky, N., 2, 14, 18, 208, 230, 233, 308, 309, 320 Cinque, G., 8, 18, 62, 66-70, 91, 92, 276, 287 Cipria, A., 22, 23, 25, 27, 28, 36, 37, 38, 42, 45, 46, 49 Clark, B., 118,273,287 Clark, H., 207,216,233,267 Coates, J, 269,284,287 Contreras, H, 4, 18 Cooper, R., 212,233 Coseriu, E., 75,92 Costa, R., 291,304 Coulmas, F., 287 Culicover, P., 230, 233, 270, 282, 287 Cutrer.M., 153,158 David,J., 264,287 Davis, E., 273,287 Davidson, D., 124, 133, 267, 287 DeHaan, F., 40,55,59 Demirdache, H., 61,72,79,92 Demonte, V, 159,162,178 D'Hulst, Y., 91,92 Diesmg, M., 160, 178, 213, 233, 289, 293, 304, 305 Diessel, H., 198,233
352
From Words to Discourse
Dietrich, W., 75,92 Dobrovie-Sorin, C., 273, 287 Doiz-Bienzobas, A., 153,158 Domitrescu, D., 281,287 Dosdrowski, G., 108, 118 Dowty, D., 25, 26, 28, 30, 36, 40-43, 46, 59, 294297, 305 Dretske, F., 56,57,59 En?, M., 15, 18, 47, 59, 234, 287, 289, 305 Escandell, V., 4, 12, 18 Espinal, M.T., 267, 280, 287 Evans, G., 40,59,214,234,310 Fabritius, C, 81,92 Farkas, D., 267,287 Fauconnier, G., 11, 14, 18, 19, 135, 137, 138, 144, 146, 158, 237, 244, 245, 261, 274, 287 Fernald, T., 159, 160, 163, 164, 165, 167, 174, 177, 176 Fernandez Leborans, M.J., 159, 162, 166, 168, 169, 178 Fernandez Ramirez, S., 3, 18, 224, 234, 323, 324, 327, 335 Fiengo, R., 311,320 Fillmore, C., 196, 197,234 Franco, J., 324, 327, 328, 335 Frege, G., 119,134 Gamut, L.T.F., 134 Garcia, E., 321,323,325,332,335 Garcia Fernandez, L., 5, 19 Garcia Murga, F., 4,19 Garey, H., 41,59 Garrido Medina, J., 4, 19, 224, 234 Gavarro, A., 61,73,92 Gazdar, G., 2, 19, 136, 144 Gennari, S., 21,23,25,30,36 Giannakidou, A., 243, 244, 251, 261 Gili Gaya, S., 5, 19, 21, 22, 23, 36, 38, 59 Given, T., 137,144 Goffman, E., 17, 19, 344, 349 Grice, H., 118,258,261 Grodzinski, J., 311,320 Groenendijk, J., 210, 211, 213, 234, 243, 261, 262 Guha, R., 211,234 Guitart, J., 154,158 Gundel, J., 227, 234, 325, 332, 335 Gutierrez-Rexach, J., 1, 4, 13, 14, 19, 181, 193, 195, 214, 219, 221, 226, 231-236, 237, 243, 259, 262
Haegeman, L., 247,262 Hamblin, C., 240,262 Hare, R., 97, 118 Hawkins,!, 215,216,234 Heggie,L., 315,316,320 Heim, I., 28,36, 127, 134, 209,210,211,214,215, 234, 242, 262, 267, 273, 277, 287, 293, 301, 305,310,312,316,320 Heny, F., 264,287 Hernanz, M.L., 85, 92, 162, 168, 178, 263, 273, 274, 287 Hewitt,!, 347,349 Higgmbotham, J., 272, 284, 287, 309, 311, 320 Hinrichs, E., 40,59 Hintikka, J., 56,59, 119, 134 Hiz, H., 264,287 Hoeksema, J., 182, 193, 242, 262 Horn, L., 14, 19, 191, 193, 237, 244, 246, 257, 262 Hornstein, N., 5, 19, 219, 234 Huntley,M., 99, 110, 118 Israel, D., 203, 234, 244, 247, 249, 253, 262 Jackendoff, R., 264,287 Jaeggli, O., 324,335 Jary, M., 118 Jayez, J., 266,287 Johnston, M., 126,130,134 Kadmon,N, 244,246,262 Kamp, H., 25, 36, 79, 92, 127, 134, 202, 210, 213, 217,234 Kany, C., 327,335 Kaplan, D., 4, 13, 19, 195,202-209,211,216,217, 232-234 Karttunen, L., 146,158 Kay, P., 146, 148 Keenan, E., 12, 19, 182, 193, 195, 198, 207, 212, 213,218,233-235 Kempchinsky, P., 264, 287 Kempson, R., 113,118 Keniston, H., 329,335 Kenny, A., 41,60 King,J., 206-208,235 King, L., 5, 19, 22, 36 Kleiber, G., 264,287 Klein, W., 72,92, 126, 134 Klima, E., 237-240,262 Koktova, O., 287 Kovacci, O., 322, 327, 328, 335
353
Name Index
Kratzer, A., 10, 19, 23, 33, 36, 119, 122, 134, 160, 173, 178,228,235 Kretzman, N., 182,194 Krifka, M., 126, 134, 244, 245, 262 Knpke, S., 203, 235 Kupferman, L., 172, 178 Ladusaw, W., 160, 178, 238, 241, 244, 262 Laca, B., 61,63,92,93 Lahiri, U., 193, 194 Laka, I., 247, 262, 270, 272, 287 Lambrecht, K., 11, 17, 19, 145, 155, 158,321,325328,335-336 Landa, A., 324, 327, 328, 335 Landman, F., 211,235,244,246,262 Lascarides, A., 195, 233 Lasersohn, P., 126, 134 Lebeaux, D., 273,287 Lee, C, 244,262 Lee, Y., 244,246,262 Leech, G., 264,287 Lema, J., 162,166,178 Lenz, R. , 323,335 Leonetti, M., 139, 141, 144, 222, 235 Levinson, S., 17, 18, 31, 36, 118, 346, 349 Lewis, D., 120, 122, 134, 199-202, 235, 273, 277, 287 Linebarger, M., 238-240, 262 Longobardi, G., 262, 279, 287 Lujan, M., 4, 19, 324, 335 Lyons,!, 264,284,288 Maclaran, R., 208-210,235 Maldonado, C., 267,287 Manteca, A., 4, 19, 141, 144 Marin, R., 162,168,178 Macia. J, 230,236,313,320 May, R., 207, 211, 235, 268, 272, 284, 287, 288, 311,320 Martin Vide, C., 4, 19 Martin Zorraquino, M.A., 4, 19 McCarthy, J., 211,233 McDowell, J., 269, 284, 288 McGinn, C., 99, 118 McNally, L., 160, 162, 170, 172, 178, 179 Mejias-Bikandi, E., 11, 19, 135-144, 145, 146, 153158, 166, 179 Michaelis, L., 341,349 Myhill, J., 330,331,334,336
Miguel, E. de, 162,168,179 Milsark, G., 268,288 Mitchell,!, 197,235 Moltmann, F., 12, 19, 182, 194 Moens, M., 46, 60 Montague, R., 4, 19, 199, 202, 233, 235, 244, 262 Montolio, E., 4, 19 Moreno de Alba, J., 299, 300, 305 Moreno Cabrera, J.C., 4, 19 Moreno Fernandez, F., 4, 19 Moro, A., 315,320 Moro Simpson, T., 4,19 Nam, S., 242,262 Napoli, D., 330,336 Neale, S, 206,235 Ogihara, T., 293,301,305 Ojea, A., 129,134 Ojeda, A, 4, 19 Olbertz,H., 73,79,93 Orecchia, T., 323, 324, 335 Ostertang, G., 209,235 Padilla-Rivera, J., 38, 60 Palmer, F., 269, 288 Papafragou, A., 118 Parodi, C., 324,335 Parsons,!., 124,134 Parrel, H., 267,288 Partee, B., 40, 60, 119, 127, 134, 197, 213, 235, 267, 287, 304, 305 Pavon, M., 186,194 Peirce, C.S., 195,235 Pelletier, F., 36 Perry, J., 122, 133, 203, 211,216, 233-235 Pesetsky, D., 276,280,281,288 Picallo, C., 264,288 Pinar, P., 305 Polletto, C., 264,288 Pollock,!, 264,288 Pomerantz, L., 348, 349 Portner, P., 46, 60, 122-126, 130 Portoles, J., 4, 19 Postma, G., 221,235 Fountain, C., 118 Prince, E., 321,325,336 Progovac, L., 281,288 Pustejovsky, J., 12, 19, 163, 165, 179 Quesada, D., 4, 18, 19
354
From Words to Discourse
Real Academia Espanola (RAE), 5, 17, 19, 38, 60, 329, 336 Recanati, F., 203,235 Reichenbach, H., 5, 18, 19 Reinhart, T., 311,320 Reyes, G., 4, 19 Reyle, U., 25,36,79,92 Richard, M., 267,288 Rigau, G., 85,92,121,134 Rivero, M. L., 4, 19, 80, 93, 143, 144, 264, 288 Roberts, C., 22, 23, 25, 27, 28, 36, 38, 42, 45, 46, 208,209,217,235 Roberts, I., 91,93,270,286 Rooryck, J., 272,288 Rowlett, P., 288 Rullman, H., 190, 193 Russell, B., 208,209,221,235 Saarinen, E., 56, 57, 60 Sanchez de Zavala, V., 4, 19 Scalise, S., 71,93 Schegloff, E., 2, 17, 19, 344, 347, 349 Schlenker, P., 203,235 Schmerling, S., 98, 118 Schwenter, S., 260, 262 Scott, D., 199,235 Searle, J., 118 Segerberg, K., 202,235 Serrano, S. , 4, 19 Silva-Corvalan, C., 5, 20, 324, 327, 336 Silva-Villar, L. 195, 225, 226, 231 -235 Silverstein, M., 321,336 Smith, C., 8, 20, 29, 36, 61, 72, 79, 81, 83, 93 Smith, N. , 136,144 Sperber, D., 9, 20, 95-99, 102, 105, 110, 111, 118, 176-179 Squartini, M., 83,91,92 Stalnaker, R., 16, 20, 28, 29, 36, 212, 235, 312, 314, 320 Steedman, M., 46, 60 Stokes, R., 347,349 Stokhof, M., 210, 211, 213, 234, 243, 261, 262 Stowell, T., 15, 20, 172, 179, 285, 288, 291-293,305 Strawson, P., 209, 236 Stucky, S., 267-288 Stump, G., 10, 11, 20, 120, 128-134 Sobin, N., 281,288 Suner, M., 4, 20, 38, 60, 263, 288, 301, 303, 305,
327, 328, 336 de Swart, H., 163,164,179,291,304 Sweetser, E., 40, 60, 146, 158 Szabolcsi, A, 205,233 Tannen, D., 344, 349 Taylor, B., 28,36 Terzi,A., 264,288 Torrego, E., 264,288 Uriagereka, J., 270, 282, 288 Uribe,M., 61,72,79,92 Valdes, L. M., 4,20 Van Benthem, J., 212,236 Van der Does, I, 214,236 Veltman, F., 211,213,234,236 Vendler, Z., 41, 60, 123, 124, 134 Verkuyl, H., 28,36 Vincent, N., 232,236 Von Fintel, K., 127, 128, 133, 134, 182, 193, 213, 236 Von Stechow, A., 203,236 Westerstahl, D, 210, 212-215, 233, 236 Wilson, D. , 9, 20, 95, 96, 97, 99-102, 105, 110, 111, 118, 176, 177, 179 Yepez, M., 327,328,336 Zacharski, R, 227, 234, 325, 332, 335 Zagona, K., 301,305 Zanuttini, R., 247, 262, 279, 288 Zeevat, H., 217,236 Zimmermann, T.E., 210,236 Zuber, R., 181-183,188,191-194 Zubizarreta, M.L., 270, 288 Zwarts, F., 242,243,251,262
355
SUBJECT INDEX
absolute constructions, 119-135 accessibility, 142, 144, 152-155 accommodation, 216, 217, 289, 293, 294 accomplishments, 24, 26, 28, 41, 77, 81, 83 achievements, 28,31,41,77,90 activation, 155, 321 adjacency pairs, 347 adjectives, 264, 266, 286 adjunct, 119, 120,132 adverbs of quantification, 127, 129, 165 aktionsart, 28, 35, 37, 40 alignment, 344 anaphora, 319 atelic, 28,41 attenuation, 256, 260 binding theory, 307-320 Catalan, 90-92 Character, 203-206, 210, 217 clitic climbing, 321-335 clitic doubling, 234,321-335 coercion, 165-168, 172 compositionality, 183, 189-194 context set, 199,212-217 conversational background, 348 copreterito, 290, 299, 300, 303, 304 copular sentences, 166, 168,311,315 deixis, 195-223 degree quantifiers, 253-286 demonstratives, 195-233 discourse particle, 248,251 discourse referent, 146-150, 152, 321, 325 discourse structure, 341,342 dynamic semantics, 28, 232 epistemic attitudes, 135, 136, 144 evidentiality, 54-59 exception phrases, 182 excess quantifiers, 264-268, 276 exclusion phrase, 182, 183, 187-191 exhaustivity, 189-193 expectations, 254-257, 261
explicature, 95, 106, 107, 113 focus, 279, 282, 286 French, 62,68,73,316 functional projections, 66, 70, 71, 91, 269, 270, 284, 286 future tense, 21-36 Galician, 100, 115 generalized quantifier, 181, 205, 209, 212 German, 107-109, 115 illocutionary force, 96-98 imperative, 95-116 imperfecto, 43-45 implicature, 95, 162,239 inclusion phrase, 182, 183, 187 indicative, 98, 135-143, 154-157 individual-level predicates, 132, 159 inertia, 46, 28 infinitives, 95-123 information structure, 145,155 Italian, 73,75,316 Logical Form, 268 mental spaces, 135-157 modal operators, 263, 265, 268, negative polarity items, 237-261, 274, 279 null objects, 321-335 open contexts, 146, 147 perception verbs, 37-59 periphrases, 61-90 politeness, 337, 344 presupposition, 135, 137, 146, 215-217, 289 preterito, 21-36,289-295 pronouns, 217,229,308,316 proposition, 28, 41, 123, 167, 242, 244, 245 proximity, 197,217,218 questions, 188,189,245,269,278 relevance theory, 95, 136, 176, 178 scales, 246-149,286 scope, 57, 128, 133, 139, 140, 203, 205, 274 secondary predication, 170 sequence of tense, 59
356
shifted reading, 289-305 situation aspect, 62,71,72 situation semantics, 119,122,123 specificity, 140, 149, 156, 214, 275 speech time, 21,46-48 stage-level predicates, 159 subjunctive, 135-144, 156, 157 telic, 28, 49, 77 uniqueness, 215,217,220 utterance, 163 viewpoint aspect, 77-81
From Words to Discourse