Grammar, Meaning and Pragmatics [1 ed.] 9789027289186, 9789027207821

Citation preview

Grammar, Meaning and Pragmatics

Handbook of Pragmatics Highlights (HoPH) The ten volumes of Handbook of Pragmatics Highlights focus on the most salient topics in the field of pragmatics, thus dividing its wide interdisciplinary spectrum in a transparent and manageable way. Each volume starts with an up-to-date overview of its field of interest and brings together some 12–20 entries on its most pertinent aspects. Since 1995 the Handbook of Pragmatics (HoP) and the HoP Online (in conjunction with the Bibliography of Pragmatics Online) have provided continuously updated state-of-the-art information for students and researchers interested in the science of language in use. Their value as a basic reference tool is now enhanced with the publication of a topically organized series of paperbacks presenting HoP Highlights. Whether your interests are predominantly philosophical, cognitive, grammatical, social, cultural, variational, interactional, or discursive, the HoP Highlights volumes make sure you always have the most relevant encyclopedic articles at your fingertips.

Editors Jef Verschueren

Jan-Ola Östman

University of Antwerp

University of Helsinki

Volume 5 Grammar, Meaning and Pragmatics Edited by Frank Brisard, Jan-Ola Östman and Jef Verschueren

Grammar, Meaning and Pragmatics

Edited by

Frank Brisard University of Antwerp

Jan-Ola Östman University of Helsinki

Jef Verschueren University of Antwerp

John Benjamins Publishing Company Amsterdam / Philadelphia

8

TM

The paper used in this publication meets the minimum requirements of American National Standard for Information Sciences — Permanence of Paper for Printed Library Materials, ANSI Z39.48-1984.

Cover design: Françoise Berserik Library of Congress Cataloging-in-Publication Data Grammar, meaning and pragmatics / edited by Frank Brisard, Jan-Ola Östman, Jef Verschueren.        p. cm. (Handbook of Pragmatics Highlights, issn 1877-654X ; v. 5) Includes bibliographical references and index. 1.  Pragmatics. 2.  Grammar, Comparative and general. 3.  Semantics.  I. Brisard, Frank. II. Östman, Jan-Ola. III. Verschueren, Jef. P99.4.P72G72   2009 306.44--dc22 2009021346 isbn 978 90 272 0782 1 (pb; alk. paper) isbn 978 90 272 8918 6 (EB) © 2009 – John Benjamins B.V. No part of this book may be reproduced in any form, by print, photoprint, microfilm, or any other means, without written permission from the publisher. John Benjamins Publishing Company • P.O. Box 36224 • 1020 me Amsterdam • The Netherlands John Benjamins North America • P.O. Box 27519 • Philadelphia PA 19118-0519 • USA

Table of contents

Preface to the series

xi

Acknowledgements

xiii

Introduction Frank Brisard Theories of grammar  2 1. 2. Topics in pragmatics  9 Naturalizing grammar  12 3.

1

Constructional analysis Kiki Nikiforidou Construction grammar and pragmatic analysis  16 1. 2. The pragmatics of grammar  17 Extending the scope: Conventional 3. pragmatics and conventional discourse  21 4. Constructions in grammaticalization  26 Summary and prospects  28 5.

16

Control phenomena Benjamin Lyngfelt Introduction  33 1. 2. Complement control – object clauses  35 2.1 Control shift  36 2.2 Other kinds of complement control  37 Adjunct control  38 3. 4. Arbitrary control  40 Less discussed control patterns  43 5. 5.1 Control in noun phrases and adjective phrases  43 5.2 Indirect control  45 5.3 Some other control relations  46 6. Outlook  47

33

VI

Handbook of Pragmatics Highlights

Definiteness Ritva Laury 1. Definite descriptions and reference  50 2. Definiteness and identifiability  52 3. Choice between types of definite expressions  55 4. Definiteness and grammar  59 5. Definiteness marking  60 6. Development of definiteness  61 7. Conclusion  62

50

Emergent grammar Marja-Liisa Helasvuo 1. Introduction  66 2. Routinization and the emergence of grammar  67 3. Emergent grammar within linguistics  70

66

Frame analysis Branca Telles Ribeiro & Susan M. Hoyle 1. Introduction  74 2. What are frames?  74 3. Frame and context in interaction  77 4. Frame and footing  79 5. Framing and nonverbal communication  80 6. Framing in everyday talk  80 7. Framing in play  82 8. Framing and institutional discourse  83 8.1 Framing and education  84 8.2 Framing and medicine  84 9. Perspectives for future research  86

74

Functional Discourse Grammar: pragmatic aspects Mike Hannay & Kees Hengeveld 1. Introduction  91 2. Outline of the model  91 2.1 FDG and verbal interaction  91 2.2 The architecture of FDG  92 2.3 Levels and layers  94 3. The interpersonal level  95 4. Discourse Acts and the relations between them  96 4.1 Introduction  96 4.2 Rhetorical functions  97

91



5.

6.

Table of contents

4.3 Illocution  102 Subacts and the relations between them  105 5.1 Introduction  105 5.2 Pragmatic functions  106 5.3 Ascription and Reference  111 Conclusion  114

Generative semantics James D. McCawley† 1. The history of generative semantics  117 2. Tenets of GS  121 2.1 Against deep structure  121 2.2 Derivational constraints  123 2.3 Context and acceptability  123 2.4 Pragmatics integrated in semantics  123 2.5 The status of logic  124 2.6 ‘Transformations’  125 2.7 The ‘base’  126 3. Pragmatics in GS  126

117

Iconicity Elżbieta Tabakowska 1. Introduction  129 2. History  129 3. Iconicity we live by: The state of the art  133 3.1 Iconicity as interpretation  133 3.2 Principles of iconicity  134 3.3 Types of iconicity  136 3.4 Areas of research  139 4. Perspectives  142

129

Information structure Jeanette K. Gundel & Thorstein Fretheim 1. Introduction  146 2. What is information structure?  147 2.1 Referential givenness/newness  147 2.2 Relational givenness/newness — Topic-focus structure  148 3. How do languages express information structure?  150 3.1 Information structure and sentence intonation  150 3.2 Information structure and morphosyntax  153 4. The grammar-pragmatics interface  155

146

VII

VIII Handbook of Pragmatics Highlights

Mental spaces Todd Oakley 1. Meanings are not “in” the words themselves  161 2. What are mental spaces?  162 3. Role and value in reference  165 4. Other features of mental spaces theory  166 4.1 Elements, relations, frames  166 4.2 Space builders  167 5. Spaces and the problems of reference, ambiguity, and presupposition  168 5.1 Referential opacity  168 5.2 Pragmatic ambiguity  169 5.3 Presupposition and optimization  169 6. Mental spaces and perspective in conditionals, counterfactuals, and deixis  171 6.1 Conditionals and counterfactuals  171 6.2 Deictic expressions  172 7. Mental spaces and discourse management  174 8. Conclusion  177

161

Modality Ferenc Kiefer 1. Introduction  179 2. Modality in logic  179 3. Necessity and possibility in linguistics  181 3.1 Epistemic modality  182 3.2 Deontic modality  184 3.3 Some further types of modality  185 3.4 The linguistic tradition  188 4. Evidentials  190 5. A possible synthesis  192 6. Syntactic treatments of modality  194 7. Modality and pragmatics  197 7.1 Two readings of ‘possible’  197 7.2 The illocutionary meaning of modal verbs  197 7.3 Deontic speech acts  199 7.4 Ability and possibility  200 7.5 Modality and grammaticalization  201 8. Prospects  203

179



Table of contents

Negation Matti Miestamo 1. Scope of negation  208 2. Markedness of negation  210 3. The expression of negation in the world’s languages  214 4. Negative polarity items  219 5. Negation and scalarity  220 6. Metalinguistic negation  221 7. Negative transport  223 8. Negation in diachrony  224 9. The acquisition of negation  226

208

Prague school Petr Sgall 1. Historical overview  230 2. Main concepts and fields of research  232 3. Prague functionalism and pragmatics  235

230

Role and Reference Grammar Robert D. Van Valin, Jr. 1. Introduction  239 2. Historical background  239 3. Central concepts  240 3.1 Clause structure  240 3.2 Semantic structure  242 3.3 Focus structure  245 3.4 Grammatical relations and linking  245 4. Some implications of RRG  247

239

Semantics vs. pragmatics Ken Turner 1. Fregean beginnings  250 2. From then until now  250 3. Current manoeuvres: (Neo-(Post-))Gricean pragmatics  253 3.1 Relevance Theory  254 3.2 The Least Effort Hypothesis  254 3.3 The Q-, I- and M-Principles Hypothesis  255 3.4 Pragmatic intrusion  257

250

IX

X

Handbook of Pragmatics Highlights

4. 5. 6.

Current manoeuvres: (Neo-(Post-))Kaplanean semantics  259 Postscript: The logical basis of the semantics-pragmatics interface  263 Conclusion  263

Tense and aspect Robert I. Binnick The semantics of markers of tense and/or aspect  268 1. 1.1 Tense  268 1.2 Aspect  268 1.3 Aktionsart  269 1.4 Underspecification and the pragmatics of tense and aspect  270 2. Discourse functions in MTA choice  272 2.1 Genre  272 2.2 Focalization  272 2.3 Function  274 Discourse coherence in the interpretation of MTAs  274 3. 3.1 Discourse coherence  274 3.2 The linguistic level  275 3.3 The intentional level  278 3.4 The attentional level  283

268

Word order Mirjam Fried Syntactic typology  289 1. 2. Pragmatic functions of word order  290 Cognitive correlates of theme/rheme notions  293 3. 4. Word order in grammatical descriptions and linguistic theory  294 Diachronic perspective  296 5. 6. Concluding remarks  297

289

Index

301

Preface to the series

In 1995, the first installments of the Handbook of Pragmatics (HoP) were published. The HoP was to be one of the major tools of the International Pragmatics Association (IPrA) to achieve its goals (i) of disseminating knowledge about pragmatic aspects of language, (ii) of stimulating various fields of application by making this knowledge accessible to an interdisciplinary community of scholars approaching the same general subject area from different points of view and with different methodologies, and (iii) of finding, in the process, a significant degree of theoretical coherence. The HoP approaches pragmatics as the cognitive, social, and cultural science of language and communication. Its ambition is to provide a practical and theoretical tool for achieving coherence in the discipline, for achieving cross-disciplinary intelligibility in a necessarily diversified field of scholarship. It was therefore designed to provide easy access for scholars with widely divergent backgrounds but with converging interests in the use and functioning of language, in the topics, traditions, and methods which, together, make up the broadly conceived field of pragmatics. As it was also meant to provide a state-of-the-art report, a flexible publishing format was needed. This is why the print version took the form of a background manual followed by annual loose-leaf installments, enabling the creation of a continuously updatable and expandable reference work. The flexibility of this format vastly increased with the introduction of an online version, the Handbook of Pragmatics Online (see www.benjamins.com/online). While the HoP and the HoP-online continue to provide state-of-the-art information for students and researchers interested in the science of language use, this new series of Handbook of Pragmatics Highlights focuses on the most salient topics in the field of pragmatics, thus dividing its wide interdisciplinary spectrum in a transparent and manageable way. The series contains a total of ten volumes around the following themes: – – – – – – – –

Key notions for pragmatics Philosophical perspectives Grammar, meaning and pragmatics Cognition and pragmatics Society and language use Culture and language use The pragmatics of variation and change The pragmatics of interaction

XII

Handbook of Pragmatics Highlights

– Discursive pragmatics – Pragmatics in practice This topically organized series of paperbacks, each starting with an up-to-date overview of its field of interest, each brings together some 12–20 of the most pertinent HoP entries in its respective field. They are intended to make sure that students and researchers alike, whether their interests are predominantly philosophical, cognitive, grammatical, social, cultural, variational, interactional, or discursive, can always have the most relevant encyclopedic articles at their fingertips. Affordability, topical organization and selectivity also turn these books into practical teaching tools which can be used as reading materials for a wide range of pragmatics-related linguistics courses. With this endeavor, we hope to make a further contribution to the goals underlying the HoP project when it was first conceived in the early 1990s.  

Jan-Ola Östman (University of Helsinki) & Jef Verschueren (University of Antwerp)

Acknowledgements

A project of the HoP type cannot be successfully started, let alone completed, without the help of dozens, even hundreds of scholars. First of all, there are the authors themselves, who sometimes had to work under extreme conditions of time pressure. Further, most members of the IPrA Consultation Board have occasionally, and some repeatedly, been called upon to review contribu­tions. Innumerable additional scholars were thanked in the initial versions of handbook entries. All this makes the Handbook of Pragmatics a truly joint endeavor by the pragmatics community world-wide. We are greatly indebted to you all. We do want to specifically mention the important contributions over the years of three scholars: the co-editors of the Manual and the first eight annual installments, Jan Blommaert and Chris Bulcaen, were central to the realization of the project, and so was our editorial collaborator over the last four years, Eline Versluys. Our sincerest thanks to all of them. The Handbook of Pragmatics project is being carried out in the framework of the research program of the IPrA Research Center / Antwerp Center for Pragmatics at the University of Antwerp. We are indebted to the university for providing an environment that facilitates and nurtures our work.  

Jan-Ola Östman (University of Helsinki) & Jef Verschueren (University of Antwerp)

Introduction Meaning and use in grammar Frank Brisard University of Antwerp

The topic of this volume is the relationship between the field of grammar, as the study of the structural features of linguistic objects (sentences, phrases, words), and that of pragmatics, which broadly speaking studies aspects of language in use (utterances/ speech acts). Since the nature of grammar is held essentially to resolve into issues of the knowledge of so-called rules of composition (or competence) and, on the other hand, pragmatics is concerned with characterizing the behavior of language users (as performance), one of the main challenges in bringing the two disciplines together will be to investigate the possible links between typically human, rational knowledge and purposeful, for the larger part culturally acquired behavior. That is, instead of more or less presuming that such behavior, in a particular context, is always simply determined by already available knowledge (of what to do/say), we might also consider the potential influence that repeated patterns of behavior, as initially observed in others, have on “emerging” knowledge systems, even if they are highly schematic as in the case of natural language. This would not make language less rational a phenomenon, at least not in the sense of being less meaningful or motivated. And indeed, if meaning is what makes people jump (i.e., makes them pay closer attention in the form of an interpretation and, in certain situations, imitate), then it should come as no surprise that the key to relating grammar and pragmatics lies in discovering the very subtle and abstract meanings behind grammatical structures, which have more often than not been thought to be devoid of any kind of functionality other than formal. So, while in the not so distant past the encroachment of pragmatics upon grammar was limited to establishing domains where “rules” did not appear to apply (lexically prompted “exceptions” in syntax, context-dependent expressions in semantics), we have now reached a point where certain grammatical theories adopt a fully pragmatic perspective, usually referred to as “usage-based”. This means that they address the formative impact of actual instances of language use on the system as a whole, and that meaning intentions, as a result of them being intertwined with form in any one such instance, play a crucial role at every level of organization, from the morpheme, over idioms and formulae, to constructional templates. This is how meaning (purpose), use (behavior), and linguistic knowledge can be seen as interrelated, and in the remainder of this

2

Frank Brisard

Introduction I will try to sketch some of the historical background against which the developing relationship between grammar and pragmatics is situated. Primarily, the volume offers an overview of a number of older and more recent, generally functionally oriented, models of grammar that, in one way or another, acknowledge the relevance of pragmatic themes. Each of the contributing authors asks how this affects the general outlook on language structure of these models, whether issues of language use inform their very makeup or are merely included as possible subject matter, and how far the actual integration of pragmatics ultimately goes (is it a module/layer or is the model truly usage-based?). In these chapters, which will be presented in Section 1, systematic care has been taken to highlight the relevant problems and focus on the implications of considering pragmatic phenomena from the point of view of grammar. Furthermore, a limited number of chapters deal with traditional topics in the grammatical literature, and specifically those which are called pragmatic because they either are not strictly concerned with truth (semantics) but rather operate on it – to some people, this is the definition of modality –, or receive their (truth) value only from an interaction with context. These will be briefly discussed in Section  2. Finally, a concluding Section 3 reflects on the apparent naturalization of grammar that goes together with the increased attention paid to extra-linguistic motivation, and points out some consequences for the conception of language as a symbolic system. 1.  Theories of grammar Logical Positivism is a major instance of the linguistic turn that has marked 20th-century philosophy, because it construes the notion of a scientific theory in terms of a set of sentences, or a language. The structure of this language can be analyzed formally by concentrating on its logical properties (syntax) or on its relationship with the world of objects (semantics). Pragmatics, in this picture, tends to be a theoretical wastebasket, as per Richard Montague’s views on the relation between linguistic meaning and indices (Montague 1974). Indeed, pragmatics qua contextualized semantics offers little more than the complement to a logic-based view of semantic content. In this perspective, the full meaning of a sentence is analyzed as an indexical function from a proposition to a time, place, and/or “possible world”.1 As such, Logical Positivism has also had a .  If modality is what the speaker does with a proposition, then any propositional modification, including volitional, emotive, and evaluative ones, comes under this heading and illocution becomes the only real object of study. Thus, the traditional dividing line between semantics and pragmatics will be blurred, and most of what one can say about linguistic meaning considered pragmatic. If, on the other hand, modality is only related to categories of necessity and possibility, linking the validity of propositions to a set of possible worlds, then there is a clear division of



Introduction

significant impact on linguistic pragmatics (Brisard & Bultinck 2006), if only in managing to restrict the range of possible pragmatic topics to issues of contextual “enrichment” for a long time. Alternative conceptions of pragmatics have of course been around for a while, including a more action-oriented approach that is chiefly inspired by one of the early semioticians of the past century, Charles Morris. As an associate editor of the International Encyclopedia of Unified Science, Morris placed an emphasis on the uses that scientific theories are put to. In this specific (rather Kantian) take on pragmatics, users of signs, or interpreters, can best be described as actors with an interest in what they say and do, within science or elsewhere. It constitutes a perspective that can be said to have given rise to interactionist, ethnomethodological, and otherwise ethnographically inspired accounts, typically in exchange with such neighboring disciplines as sociology, anthropology, and psychology (all of this treated elsewhere in the present series). Nevertheless, in his overview of the current state of affairs in the debate over the semantics-pragmatics interface, Turner (this volume) goes back to a (post-Fregean) philosophical formulation of pragmatics as explicating meanings that have not been directly expressed by the speaker, and thus as enriching a semantics that is itself still very much based on (first-order) logic. This debate is obviously more relevant to the study of meaning than it is to that of grammar, but some of the more important problems investigated in this tradition do touch upon central grammatical concerns, including the behavior of anaphoric expressions and all kinds of scalar phenomena, for instance with connectives and quantifiers. When it comes to syntax, George Lakoff and colleagues like Postal, the late Jim McCawley, and Ross, but mainly Lakoff (1965) argued, in his dissertation On the nature of syntactic irregularity, that there were, or had to be, lexically governed rules (like the raising rules and other transformations), whereby only certain of the seemingly eligible predicates occur in a particular syntactic pattern. Special new theoretical devices, such as rule features, had to be invented to deal with them in a generative grammar, and these were required precisely because the assumption that language factors neatly into regular and irregular components has turned out just that: an aprioristic assumption. This, in generativism, led to an inadmissible critique or extension — whichever way one cares to look at Generative Semantics and what came out of it —, and ultimately to the well-known schism between the formalist and functionalist paradigms in linguistics. McCawley (this volume) explains how in the study of grammar, as a result of the Generative-Semantic experiment, meaning starts to trickle in at more levels than a sentence’s deep structure, eventually leading

labor between semantic, or lexically and syntactically predictable, meaning and pragmatic issues of relating linguistic structure to context, i.e., time, location, social setting, and participants’ roles, as well as the interlocutors’ strategies, plans, goals, and intentions. See Kiefer (this volume).

3

4

Frank Brisard

to the “interpretivist/lexicalist” reinterpretation we all know has marked Chomskyan thinking ever since. And while the success of Generative Semantics itself may have been relatively short-lived, in part due to the unattractiveness of ever more excessively complicated deep-structure representations, other new approaches to syntax developing in the early 1970s, among which Montague/categorial grammar and relational grammar, would soon fill in the void. At the same time, and as McCawley notes too, it was not until the community was already in a period of decline that generative semanticists came to see meaning as at least partly having to do with not strictly semantic (i.e., with pragmatic) issues; the small body of literature on the subject from a GS point of view covers rather traditional and well-delimited topics like tense meaning and pronoun choice. In a way, we have had to wait until the 1980s and ‘90s before any of the truly radical insights from Generative Semantics, and notably its attribution of meaning to grammatical constructs, were actually developed. This happened especially in the newly emerging movement of cognitive linguistics, uniting theories that all, in one way or another, adhere to a symbolic conception of grammar (cf. below). In the meantime, generativism had also been challenged by other paradigms, mainly originating in European structuralism. Most of the models involved, exploring both the systemic and goal-oriented nature of language (synchronically and diachronically) and the interactive features of communication, are called functionalist, a term that is primarily related to a tradition known as the Prague School, prominent since the 1930s. It is, of course, quite arbitrary to select only a couple of these models here, but many of the following paradigms discussed have at least played an important historical part in the gradual introduction of pragmatics into the study of grammar. The Prague School, presented in Sgall (this volume), though initially mostly successful in the domain of phonology, offering a functional description of a language’s sound patterns as specific subsystems, has had a lasting impact on the analysis of sentence structure (cf. also Gundel & Fretheim, this volume) as well, which it sees as conditioned by the patterning of sentences in a discourse, i.e., by their use in context. In systematically distinguishing between form and function, or between grammar and logic, the Prague Linguistic Circle was clearly influenced by earlier sources, including Danish functionalism (Hjelmslev’s work on case, but mainly Jespersen) and Polish and Russian formalism, with Roman Jakobson as a central figure in theory formation. Mathesius (1929) was the first to come up with the idea of Functional Sentence Perspective, offering a structural account of the topic-focus articulation. In turn, this Prague School has inspired later functionalist models, such as Simon Dik’s Functional Grammar in Amsterdam and the development of cognitive semantics. Its understanding of language as a set of levels/layers, ordered from meaning intention to expression (speaker-related) or vice versa (hearer-related), reappears in subsequent work, among others of the Australian linguist Michael Halliday, who combines a



Introduction

Prague-School approach with the ethnographically inclined research by J.R. Firth. Systemic Functional Grammar sees the paradigmatic aspects of language (language as a meaningful choice between oppositions) as primary (see Halliday’s scale and category grammar from 1961), defining meaning as the function of a unit in a context, and always treats syntagmatic structure in relation to specific (sub)systems of a language, which form elaborate networks. These networks are seen, in a very Firthian vein, as serving purposes of “social semiotics”, since they are tightly linked with particular registers that have some semiotic importance within a given cultural context (e.g., the “register” of mother-child interaction). Like most functional models, then, it tries to introduce pragmatics into the study of grammar by linking the meaning of a sentence to an illocutionary intention (and thus, the sentence itself to its utterance), as well as by indicating the special kind of relationship between the level of the sentence and that of the discourse/text and the communicative setting in which it is embedded. Systemic grammar can still be called generative in the sense that it provides descriptions of the system networks that represent the meaning potential from which the speaker may select in generating the structure for a given linguistic unit, like the clause. As such, it illustrates the tension that has existed since the 1960s between Halliday’s and others’ action-oriented work and a more classically philosophical conception of pragmatics (as voiced, for instance, in Leech 1983), or, in other words, between principle- vs. rule-based accounts of phenomena of language use. Two distinctly functionalist contemporary theories of grammar are Functional Discourse Grammar (Hannay & Hengeveld, this volume) and Role and Reference Grammar (Van Valin, this volume). They, too, might, in a more neutral use of the word, be called generative, since they aim for some kind of formalized algorithm to generate linguistic structures, from intention to articulation. But at the same time, both models are also good examples of a very explicit stance against Chomskyan generativism, denying the autonomy of grammar and stressing the context of verbal interaction, in the form of interpersonal and discourse-related effects on linguistic structure. They are both highly centered on “descriptive adequacy” and thus found to be well-suited for typological research as well. Functional Discourse Grammar is the heir to Dik’s Functional Grammar, with an additional specification of various levels (pragmatic, semantic, morpho-syntactic, and phonological) which a given structure needs to run through in order to go from pre-linguistic conception to linguistic output. It takes the speech or discourse act as its basic unit of analysis and can accommodate structures both smaller and larger than the sentence. The starting point is always an underlying meaning representation, whereby it is pragmatics that governs semantics and all the rest, rather than the other way around. The link between grammar and context is ensured by specifying the various ways in which the surrounding discourse, the physical speech setting, and/or the social relationships

5

6

Frank Brisard

between participants interact with the grammatical operations of “encoding” and “formulating” a thought. And similarly, Role and Reference Grammar is a primarily semantico-pragmatically oriented theory of grammar in that it has lexical representations (“logical structure”) map into syntactic surface structures without any intervening level of syntactic representation and with potential interference from discoursepragmatic factors. It descends from Fillmore’s (1968) Case Grammar, especially in how it views clause structure (which is layered), and as with Fillmore, it does not assume the universality of grammatical relations like subject and direct object. Finally, Role and Reference Grammar treats semantic roles rather differently from most other theories: it posits two types or tiers, the first of which are the traditional specific thematic relations (agent, theme, etc.), while the second are generalized semantic “macroroles”, the so-called actor and undergoer which correspond to the two primary arguments in a prototypical transitive relation (cp. with Langacker’s 1987 trajector and landmark). While the linking between argument position and macroroles is said to be universal, the theory locates cross-linguistic variation in the mapping between macroroles and syntactic functions (like “pivot”). Thus, Role and Reference Grammar posits a distinction between more semantic “role-dominated” languages and “reference-dominated” ones, permitting discourse-pragmatic influences on the choice of pivot. One of the principal ideas behind Functional Discourse Grammar is its rigorous top-down architecture, which is directly motivated by a concern to reflect psycholinguistic, experimental evidence on online processes of language production in (the mind of) the speaker (mainly following Levelt 1989). That sounds very cognitive indeed, yet it is still a relatively far cry from the outspoken psychological orientation of contemporary cognitive linguistics, which claims that all of language structure is ultimately shaped by extra-linguistic, cognitive-psychological principles, usually those playing at very low levels of processing. Once again we must go back to Fillmore and his original formulation of Frame Semantics as relating to the conceptual systems involved in the interpretation and production of discourse (see Petruck 1995 for an overview). Frames, or scripts or schemas,2 refer to those cognitive constructs that are responsible for the surplus of meaning that any message conveys, over and above its literal meaning, i.e., what can be computed on the basis of lexical meaning and sentence structure. Frames, like Gestalts, organize experience by affecting the way .  Frames are appearing pretty much everywhere now, from artificial intelligence over cognitive science to text linguistics. In Telles Ribeiro & Hoyle (this volume), the emphasis lies on socioanthropological uses of this notion and its application in studies of institutional discourse, following Bateson’s and especially Goffman’s conception of frames as systematically transforming all forms of human activity. This is a particularly dynamic conception, in which verbal interaction consists in constant activities of reframing, or footing.



Introduction

we categorize, remember, and revise what we know and say/hear. As holistic structures of expectation, they allow us to predict what new experiences and relationships will be like and interpret them against our prior knowledge. But more crucially for grammar, finding out about the importance of frames and unpredictable (in the sense of noncompositional) meaning elements in the organization of the lexicon has led more or less directly to a logical next step, viz., the assumption that, if grammatical constructions carry meaning (have a “semantic pole”; cf. Generative Semantics), then frames must also have a part to play in the organization of grammar. And in fact, in various types of Construction Grammar3 (see Nikiforidou, this volume), purely semantic information characterizing “conventionalized aspects of a construction’s function” (Croft & Cruse 2004: 258) is complemented with equally conventional situational/discursive/textual/register characteristics, so that a strict dichotomy between semantics and pragmatics is no longer maintained. Furthermore, constructional approaches, like others in cognitive linguistics, present themselves as nonmodular, non- or transderivational, and typically employing some unification-based formalism which treats structural aspects of signs as integrated parts, rather than distributed over different modules. Their revolutionary potential resides in the combination with a truly usage-based philosophy, meaning that the entire grammatical system of a language is taken to be made up, on the basis of individual experiences, of taxonomies of concrete pairings of meaning and form, the latter including phonological aspects like prosody and intonation, at progressively more schematic levels until nothing is left but a constructional template, the ultimate explanatory goal of generativism. In contrast with generativism, however, usage-based models pay attention to, and derive their predictions from, considering fully schematic and intermediate instances for any (family of) construction(s), with a special interest in idiomatic expressions, which are by definition half-lexical, half-grammatical.4 In the end, such an approach forces one to abandon artificial, theory-laden contrasts between rules of regular expressions (grammar) and lists of irregular expressions (lexicon). Or, in other words, it suggests the relevance of positing a continuum of symbolic expressions, with extremes going from the productive to the idiosyncratic. In principle, construction grammarians look at all sorts of form-meaning pairing, as long as its properties cannot be predicted from general syntactic/semantic rules or from an all-purpose .  This is another term that has a wider application than will be highlighted here. Specifically, it also applies to explicitly generative models, including Generalized Phrase Structure Grammar, Head-Driven Phrase Structure Grammar, and Lexical-Functional Grammar. .  Construction Grammar shares this interest with the phraseology and collocation tradition in lexicography (see, e.g., Sinclair 1991). In combination with the development of sophisticated statistical techniques for the investigation of (large) corpora, it has also given rise to a new paradigm researching (families of) “collostructions” (e.g., Gries & Stefanowitsch 2006).

7

8

Frank Brisard

pragmatic reasoning module, and the pairing occurs with sufficient frequency (which it may only acquire over the course of time, as witnessed in historical studies of so-called grammaticization). Obviously, cognitive linguists tend to argue that issues of meaning should be related to the mind conceiving of that meaning. Several theories within the overall cognitive-linguistic paradigm attend to different aspects of this mental reality: e.g., the “contemporary theory of metaphor”, developed by Lakoff and his colleagues in the 1980s and ‘90s, proposes that metaphorical structures do not only pervade language, but also human thought. The theory of Mental Spaces and Conceptual Integration (Oakley, this volume) follows up on the idea that frames provide stereotypical roles, values, and relations that may be mobilized in the online interpretation of discourse. According to Gilles Fauconnier, they are so mobilized in the form of elements included in mental models or spaces, which are “partial structures that proliferate when we think and talk, allowing a fine-grained partitioning of our discourse and knowledge structures” (Fauconnier 1997: 11), and in bindings across these spaces, called “blends”. As we talk or think, our focus thus flows from one space to the next, transporting and mapping entities according to points of view, beliefs, changes of mood or tense, analogical counterfactuals, and so on, and with each giving birth to a different mental space (see also Fauconnier & Sweetser 1996). What is most notable is that the meaning which is thought to be everywhere now (i.e., at all levels of linguistic structure from the morpheme up) does actually not reside in any significant way in those symbolic structures, but is triggered by them. That is, mental-space theorists and other cognitive linguists in general see grammatical devices (and ultimately even lexical ones) as vastly underspecified in terms of meaning themselves, acting rather as triggers or instructions, for instance to set up a particular configuration of mental spaces. The actual content of these configurations (e.g., which types of mental space are involved, or referential properties of the elements contained in them) is to be inferred from co- and contextual clues. This is part of what Fauconnier (1994) has called the “backstage cognition” at work in the interpretation of linguistic meaning, and it demonstrates that language, and especially grammar, is a huge “compression machine”, working mainly with stereotypical conceptions and attributing most of the interpretive detail to contextual enrichment, including the recruitment of general world knowledge.5 In this respect, it is clear that the study of grammar has moved beyond the level of individual sentences and has come to consider discourse and its management as an integral

.  Research into the gradual conventionalization of the link between a given expression and the context(s) in which it appears (or, in other words, of the gradual inclusion of pragmatic information into an expression’s semantics) is part of the agenda in the grammaticization literature. For an introduction, see Heine et al. (1991).



Introduction

and indeed essential part of its domain. What is more, in this pragmatic perspective it has turned out that the sentence is but one, possibly epiphenomenal, level among many “lower” and “higher” ones, and that it is the routinization of a simple or complex “chunk” of symbolic information that constitutes the real object of study for grammarians, regardless of such issues as compositionality or how concretely or schematically that chunk might be specified in terms of meaning and/or form. This is not just a pragmatic, but also an essentially functionalist, turn in grammar, because it makes “reference to external factors” (Levinson 1983: 40), especially communicative causes and functions, to explain grammatical behavior and in fact promotes the study of performance, governed by psychological principles, as a necessary preliminary for being able to say anything scientifically relevant about such a thing as linguistic competence. 2.  Topics in pragmatics Mental-space theory is particularly suited for dealing with phenomena, usually described in terms of “projection” problems, that involve role- vs. value-readings (cf. Donnellan’s 1966 referential vs. attributive readings), referential opacity and ambiguity, pragmatic presupposition, deixis, and the behavior of tenses in conditionals, counterfactuals, and narrative. It is, precisely because it sees all of these as following general properties of (real or fictive) cognitive models, including the reference mental-space theory makes to principles of mental access (which seem to have linguistic consequences): pragmatic connectors linking entities in one space to various types of metaphorical, metonymic, analogical, etc. representation in another. As such, the theory also provides an understanding of linguistic meaning and reference as experientially grounded (perspectival), viz., in the epistemic position (knowledge, beliefs) of the speaker/conceptualizer, that is a true alternative to the various traditionally truth-conditionally oriented sentence logics, where reference is a direct relation with the world. Thus, invariably the chapters in the present volume discussing such classic semantic topics as definiteness, modality, negation, and tense and aspect point out the pragmatic, usage-related nature of many, if not all, known constraints governing the use of specific markers or constructions, instead of them being, as long maintained, primarily syntactically motivated. For instance, the definiteness effect observed with English there constructions, as in There’s a/*the rat in the kitchen, is suggested to relate to the epistemic (hearer-new) status of the post-verbal referent, sometimes actually allowing definite nominals (Laury, this volume). Moreover, it seems promising and fruitful to investigate the cognitive correlates of definiteness (in particular the notions of identifiability and mental accessibility; cf. Ariel 1990), which, again, is basically a pragmatic concern. And with modality, too, syntactic differences do not correlate systematically

9

10

Frank Brisard

with semantic differences (see Kiefer, this volume, on raising vs. control verbs and how they cannot be reduced to the distinction between epistemic and root modality). Traditional research in this domain has centered around problems, among more logically inspired ones, of the “illocutionary act potential” of modal expressions, a discourseanalytic subject that touches upon the interactional functions of modality. In this regard, Searle’s (1975) observation of formulaic, so-called indirect speech acts involving modal verbs, as in May I? or Could you check these figures (please)?, has important implications for assessing the degree of conventionality with which an illocutionary force can be said to be carried by these expressions. This, the typical grammaticization paths taken by modal markers (roughly following schemes like ability Æ root possibility Æ epistemic possibility), and the role of conversational implicatures therein are also increasingly being studied from a comparative point of view. Similarly, negation as a modal operator changing a proposition’s truth value, a subject closely associated with scalarity, invites a pragmatic approach insofar as the seemingly inevitable markedness of negative utterances needs to be addressed, which a strictly logical account cannot in principle (Miestamo, this volume). Instead, as a rule recourse is taken to Grice’s (1975) “logic of conversation” to explain why negative propositions are usually denials of positive ones present in the context, and especially to the maxims of quantity and relevance (Horn 1989). Even tense and aspect markers, which would appear to be the referential categories par excellence in language, do not escape cognitive/functional and discourse-based treatments, and within formal circles it is at least conceded that much of the semantics of tense and aspect is typically underspecified (Binnick, this volume). Relevance Theory, as a neo-Gricean model, therefore proposes that missing or implicit temporal information needs to be filled in by the hearer and that, consequently, tense and aspect markers cannot be seen so much to encode propositions-in-time as to direct the hearer in the process of recognizing the speaker’s communicative intention and filter out impossible readings (Wilson & Sperber 1998): for instance, in I didn’t turn off the stove, the definite reference point in the past, which is said to be required by the use of the past tense, is not part of the meaning of (i.e., given by) the past tense but must be pragmatically inferred from context.6 These, and others referred to in the relevant chapters, are just a few examples of the interesting submission that the pragmatics somehow relevant to the grammatical facts at issue may in fact be seen to underlie their semantics.

.  The principle of iconicity, introduced in Section 3, plays an important role in establishing the default interpretation of the relative sequence of events in a narrative: the order of the eventualities described is that of the clauses describing them. This principle is thus a prime example of a pragmatic/psychological force driving grammatical processes.



Introduction

Recently, it has been Construction Grammar that has paid closest attention to the rich pragmatics conventionally associated with specific grammatical constructions, as in the domain of information structure (but also, for instance, re scalarity in language).7 In expressing some basic informational content, the speaker is always confronted with a variety of (structural and intonational) options offered by the language to arrange that content in function of what the speaker wants an utterance to be about and what she may assume the addressee already knows or believes and/or is attending to herself (Gundel & Fretheim, this volume; see also Mathesius’ groundbreaking work in this field, referred to in the discussion of the Prague School in Section 1). Naturally, information-structural concepts such as topic, focus, and various degrees of referential givenness are pragmatically relevant categories with clear pragmatic effects, yet ones that are at the same time heavily constrained by linguistic form across languages. Thus, the debate on whether or not it is possible to reduce the expression and interpretation of information structure completely to general principles governing human interaction, or to other cognitive abilities that are independent of language, is probably put in too extremist terms. Still, it is evident that not all the phenomena associated with information structure can be directly attributed to, or built into various modules of, grammar. Here, too, rather than relegating problems of distinguishing between properties that are grammar-driven and those that are purely pragmatic to some conceptual-intentional “interface”, it might in the end be more promising to assume a continuum of less and more grammaticized formal strategies for arranging semantic content. The same holds for the linear dimension of word order (Fried, this volume), which, as a fundamental property of human language, is all together at odds with the hierarchically organized, multidimensional nature of a grammatical system. The sometimes strikingly different linearization patterns that languages are by now known to exhibit have led to the primarily syntactic typological research initiated by Greenberg (1963), which runs into trouble when trying to ascertain the universal nature of the notion of basic word order. Functional typologists like Sandra Thompson (1978) have questioned this status and are more concerned with explaining why certain orders are disproportionately less frequent across the world’s languages (OSV) than others (SVO, SOV, VSO), or with investigating what discursive functions might be served by variable word order and what cognitive mechanisms might be at work in accounting for the distribution of word-order patterns. It is, once again, mainly due to the Prague School’s heritage that we now have the insight that any linguistic structure, from sentence to

.  An illustration of the former is Lakoff ’s analysis of there constructions, which appeared as Case Study 3 in Lakoff (1987) and is generally considered to be the earliest study in (proto-)Construction Grammar. The latter, scalarity, is the topic of another early classic in Construction Grammar, Fillmore et al.’s (1988) paper on let alone.

11

12

Frank Brisard

text, ought to be analyzed as a reflection of the speaker’s communicative strategies in structuring information flow. Finally, even the study of such a supposedly hardcore syntactic concept as control in nonfinite clauses (Lyngfelt, this volume) needs to allow for patterns of “pragmatic control”, where it is some contextually salient element, rather than the usual matrix subject/object, that controls PRO, as in After PRO*i finishing her studies, an overseas jobi was a welcome change. As in most contemporary domains of grammar, this one is governed by the opposition between configurationalist (syntactic) accounts and lexicalist ones, which, like Jackendoff (1972), demonstrate that control patterns essentially depend on the meaning of the matrix verb. This focus on semantics, and the idea that verbs in the lexicon not only select for the number and types of arguments they are construed with but also for the control relations associated with these arguments, is in line with the general interpretivist move in generative grammar noted earlier, while the realization that contextual bias cannot be neglected as a motivating factor for certain “syntactic” relations (like control) follows an equally notable trend to involve discourse ever more closely in grammatical analysis. 3.  Naturalizing grammar If grammatical constructions are in a way just formally complex, semantically very abstract words (or, vice versa, words relatively simple and concrete constructions), then a truly symbolic conception of grammar seems within reach, or may actually be developing at the moment. This would connect modern (structuralist) linguistics with extra-linguistic disciplines in the so-called life sciences, including (social) psychology, anthropology, and, ultimately, biology, much more tightly than, in my view, ever before, because such an approach generalizes the concept of the sign in dealing with linguistic structures of varying complexity, thereby reaffirming the semiotic character of this enterprise, and, furthermore, it focuses on the natural principles responsible for organizing complex signs. One of those principles is, of course, that of iconicity (Tabakowska, this volume), which immediately raises the Platonic question of the limits of the symbolic, or arbitrary, status of the linguistic sign, something of a Saussurean dogma if there is one.8 The existence of systematic correspondences between form and meaning (over and above onomatopoeia and basic sound symbolism), as well as between expressions and their contexts of use, might lead to the not unreasonable .  A seminal paper challenging this dogma is Jakobson (1965), which includes the classic reference to Caesar’s dictum veni, vidi, vici. More recently, Haiman (e.g., 1985) has revived the interest in the at the time “somewhat unfashionable” topic of iconicity. For the important implications for work on proto-grammars and the origins of language, once considered an infamous question, see Givón (1995).



Introduction

hypothesis, against Saussure’s autonomous and radically synchronic conception of language, that diachronic linguistic development shows a dynamic shift of class membership which may in fact explain the state of a language at a certain time. This would make the conventional symbols constituting language secondary in the sense of motivated by primary iconic and indexical relations, and it would perhaps even make sense for describing processes of language acquisition (ontogenetic diachrony). And given that icons and indices by definition stand closer to the natural world than symbols do, principles of that natural world might leak through to the highest levels of grammar via the iconic and indexical properties of every utterance we produce. Bolinger’s (1977) remarks on the function in discourse of repetition, deletion or pronominalization of nominal phrases can serve as a nice illustration of how iconicity, formerly the privilege of literary theories of mimesis, has at long last crossed over, through various forms of text linguistics, to the very core of grammar, syntax. One particularly strong demonstration of the naturalizing study of grammar as recurrent discourse patterns is provided by Emergent Grammar (Helasvuo, this volume). Together with colleagues like Thompson and John Du Bois, Paul Hopper proffers an approach to grammar as continually provisional and rooted in repeated instances of concrete language use. [T]he Emergence of Grammar (EOG) attitude […] has come to view grammar as the name for a vaguely defined set of sedimented (i.e., grammaticized) recurrent partials whose status is constantly being renegotiated in speech and which cannot be distinguished in principle from strategies for building discourses. (Hopper 1988: 118; emphasis in the original)9

Hopper also gives the term grammaticization a wider meaning: instead of referring to the historical processes whereby lexical items become more and more grammatical, it designates the fundamental ontology of grammar, i.e., the ways in which grammar, as movement toward structure, can be said to exist. Gradually, situationally defined (one could say, indexical) meaning and features of use, characterizing collaborative acts of communication, fade away and an expression crystallizes into a grammatical routine. Moreover, since grammar is fundamentally social (“distributed”) and interactive in nature, it is not individual intuition but real-time negotiated activity in context that provides the relevant data. Ideal(ized) though this position may be from a strictly

.  As Helasvuo (this volume) rightly observes, it should be noted that the term Emergent Grammar does not so much identify Hopper’s (or any other specific linguist’s) own position as represent, for rhetorical purposes, an extreme attitude opposed to what is called the “a priori grammar attitude” in his article. Hopper and his colleagues, who all look at discourse — spoken or written — for emerging regularities and have developed “Discourse Transcription” conventions along the way, are sometimes associated with the Santa Barbara School of ethnomethodology.

13

14

Frank Brisard

linguistic/grammatical point of view, it could, and perhaps even should, stimulate further collaboration with scholars working in interactional sociolinguistics, conversation analysis, and discourse analysis.

References Ariel, M. (1990). Accessing noun-phrase antecedents. Routledge. Bolinger, D. (1977). Meaning and form. Longman. Brisard, F. & B. Bultinck (2006). The interface of linguistics and pragmatics: Its development during the second half of the 20th century. In S. Auroux, E.F.K. Koerner, H.-J. Niederehe & K. Versteegh (eds.), History of the language sciences: An international handbook on the evolution of the study of language from the beginnings to the present, vol. 3: 2510–20. Walter de Gruyter. Croft, W. & D.A. Cruse (2004). Cognitive linguistics. Cambridge University Press. Donnellan, K. (1971 [1966]). Reference and definite descriptions. In D. Steinberg & L. Jakobovits (eds.), Semantics: 100–14. Cambridge University Press. Fauconnier, G. (1994). Mental spaces: Aspects of meaning construction in natural language. Cambridge University Press. [augmented paperback edition] ——— (1997). Mappings in thought and language. Cambridge University Press. Fauconnier, G. & E. Sweetser (eds.) (1996). Spaces, worlds, and grammar. University of Chicago Press. Fillmore, Ch. J. (1968). The case for case. In E. Bach & R.T. Harms (eds.), Universals in linguistic theory: 1–88. Holt, Rinehart & Winston. Fillmore, Ch. J., P. Kay & C. O’Connor (1988). Regularity and idiomaticity in grammatical constructions: The case of let alone. Language 64: 501–38. Givón, T. (1995). Functionalism and grammar. John Benjamins. Greenberg, J. (1963). Some universals of grammar with particular reference to the order of meaningful elements. In J. Greenberg (ed.), Universals of language: 58–90. MIT Press. Grice, H.P. (1975). Logic and conversation. In P. Cole & J.L. Morgan (eds.), Syntax and semantics, vol. 3: Speech acts: 41–58. Academic Press. Gries, S.Th. & A. Stefanowitsch (eds.) (2006). Corpora in cognitive linguistics. Mouton de Gruyter. Haiman, J. (ed.) (1985). Iconicity in syntax. John Benjamins. Halliday, M.A.K. (1961). Categories of the theory of grammar. Word 17: 241–92. Heine, B., U. Claudi & F. Hünnemeyer (1991). Grammaticalization: A conceptual framework. University of Chicago Press. Hopper, P.J. (1988). Emergent grammar and the a priori grammar postulate. In D. Tannen (ed.), Linguistics in context: 117–34. Ablex. Horn, L. (1989). A natural history of negation. University of Chicago Press. [reprint with supplemental bibliography, CSLI Publications, 2001] Jackendoff, R. (1972). Semantic interpretation in generative grammar. MIT Press. Jakobson, R. (1971 [1965]). Quest for the essence of language. In Selected writings, vol. 2: 345–59. Mouton. Lakoff, G. (1965). On the nature of syntactic irregularity. Ph.D. dissertation, Indiana University. ——— (1987). Women, fire and dangerous things: What categories reveal about the mind. University of Chicago Press. Langacker, R.W. (1987). Foundations of cognitive grammar, vol. 1: Theoretical prerequisites. Stanford University Press.



Introduction

Leech, G.N. (1983). Principles of pragmatics. Longman. Levelt, W.J.M. (1989). Speaking: From intention to articulation. MIT Press. Levinson, S.C. (1983). Pragmatics. Cambridge University Press. Mathesius, V. (1929). Zur Satzperspektive im modernen Englisch. Archiv für das Studium der neueren Sprachen und Literaturen 155: 202–10. Montague, R. (1974). Formal philosophy. Yale University Press. Petruck, M.R.L. (1995). Frame semantics and the lexicon. In M. Shibatani & S. Thompson (eds.), Essays in semantics and pragmatics: 279–96. John Benjamins. Searle, J.R. (1975). Indirect speech acts. In P. Cole & J.L. Morgan (eds.), Syntax and semantics, vol. 3: Speech acts: 59–82. Academic Press. Sinclair, J. (1991). Corpus, concordance, collocation. Oxford University Press. Thompson, S.A. (1978). Modern English from a typological point of view: Some implications of the function of word order. Linguistische Berichte 54: 19–35. Wilson, D. & D. Sperber (1998). Pragmatics and time. In R. Carston & S. Uchida (eds.), Relevance theory: Applications and implications: 1–22. John Benjamins.

15

Constructional analysis Kiki Nikiforidou University of Athens

1.  Construction grammar and pragmatic analysis Construction Grammar developed as a non-modular, non-derivational, unificationbased grammatical theory aimed at covering the facts of language as a whole (Fillmore, Kay & O’Connor 1988). Alongside the original approach, developed further by Kay and Fillmore (1999), Michaelis and Lambrecht (1996), and Fried and Östman (2004a), other constructional approaches have since emerged, Cognitive Construction Grammar (Goldberg 1995, 2006) and Radical Construction Grammar (Croft 2001) being prominent versions of the constructionist paradigm. Although the different constructional approaches are not intended as pragmatic theories or methods as such, they all assign a special place to pragmatic research and have often yielded fine-grained, detailed analyses of the pragmatics associated with particular linguistic forms. In this respect, constructional analysis allows for the natural integration of pragmatics into grammatical theory. The present chapter explores some of the ways in which this integration has been realized in different studies, illuminating the potential of construction grammar for describing (and even formalizing) all kinds of conventional linguistic knowledge, including pragmatic and discoursal.1 All constructional approaches emphasize the role of grammatical constructions, i.e., conventionalized pairings of meaning and form, in stating language-specific and cross-linguistic generalizations and in accounting equally for regular and semi-regular patterns. Both the adults’ knowledge of their language and the child’s learning of language (Tomasello 2003; Goldberg 2006) are assumed to be structured in terms of such conventional pairs. The meaning, i.e., the semantic pole of any construction (lexical or syntactic), is in turn defined in terms of frames; in this sense, construction grammar subsumes frame semantics and incorporates valuable insights from frame-semantic theory (Fillmore 1985; Fillmore & Atkins 1992), not in the least the rejection of a strict dichotomy between semantics and pragmatics. Information, therefore, about the pragmatic/discoursal/textual/register characteristics associated with a particular form can be represented in the meaning pole of the corresponding construction alongside purely semantic information (Goldberg 1995: 7; Fried & Östman 2004a). Meaning

1.  Construction Grammar is used here as a cover term for all constructional approaches.



Constructional analysis

in construction grammar thus stands for “all the conventionalized aspects of a construction’s function, which may include not only properties of the situation described by the utterance, but also properties of the discourse in which the utterance is found […] and of the pragmatic situation of the interlocutors” (Croft & Cruse 2004: 258). The special interest in the pragmatics of constructions can be traced to the concern of construction grammarians to find a place for idiomatic expressions in the grammar of a language. In Fillmore, Kay and O’Connor’s (1988) classification of idioms, the category of formal (or schematic) idioms (part of which can be filled by the usual words or phrases syntactically and semantically appropriate for a given slot), poses a challenge to any componential model with “productive rules” vs. “lexicon” components. Such expressions, while semi-regular or productive, are crucially associated with syntactic, semantic, and often pragmatic properties which cannot be predicted from the general syntactic or semantic rules of the language, or from an all-purpose pragmatic component; as such, they must be directly assigned to the construction as a whole. For example, the formal idiom exemplified by Him be a doctor?! associates a morphosyntactic pattern with a special pragmatic effect of ‘incredulity’. To the extent that the latter is not derivable by general conversational principles, it should be accounted for by the grammar.

2.  The pragmatics of grammar The requirement of (at least partial) unpredictability of the syntactic, semantic or pragmatic properties associated with a particular pattern has dominated early constructional analyses, both in the choice of the phenomena analyzed and in the resulting kind of analysis. In short, construction grammarians sought to identify semi-schematic patterns2 with “interesting” semantics and/or pragmatics, i.e., a meaning and function which are not fully compositional given their component parts. The seminal let alone paper (Fillmore, Kay & O’Connor 1988) highlighted the rich pragmatics associated with the pattern exemplified in (1), as well as the special discourse setting required for such utterances:

(1) Fred won’t eat shrimp, let alone squid.3

Interpreting (1) requires the construction of a “scalar model”, i.e., a set of propositions ordered by unilateral entailment against the background of a presupposed scale (in

2.  Semi-schematic patterns are constructions that are productive and regular enough to merit inclusion in the grammar, while also having fixed parts and idiosyncratic properties. 3.  The capital letters represent prosody peaks.

17

18

Kiki Nikiforidou

this case, Fred’s unwillingness to eat shrimp unilaterally entails his unwillingness to eat squid). In other words, the let alone construction forces an interpretation in which the two propositions are seen as points on a scale; the exact nature of the scale is not derived from the form of the sentence but is determined pragmatically in the conversationally common ground. In addition, the felicitous utterance of (1) requires that the proposition that ‘Fred eat squid’ has been uttered in the conversation and be on the floor. Although the let alone construction seems to serve simultaneously the conflicting demands of Relevance (by asserting the informationally weaker second conjunct) and of Quantity (by asserting the more informative first conjunct), exclusive reference to these maxims can predict neither the pragmatically based scalar interpretation nor the discoursal requirement of a context that somehow includes the less informative proposition. These, we must conclude, are built into the constructional pattern as such. In the same line, Kay and Fillmore’s (1999) analysis of the What’s X doing Y? construction, e.g., (2), reveals that a special property associated with this pattern is the presupposition of ‘incongruity of the event’:

(2) What are your children doing playing in my garden?

The constructional (non-derivable) morphosyntactic and semantic characteristics of this pattern include the following: (a) the presence of the verb do in the present participle is obligatory, (b) the form doing must be the complement of the copula be, (c) the be + doing combination does not express progressive aspect, (d) the construction does not permit negation of either be or do. The incongruity reading of this pattern cannot be derived via conversational reasoning, since it is very often the case that such sentences do not have a literal reading at all (on which to base the flouting of a maxim), nor is this reading, when available, contextually suspended, e.g., (3): 

(3) What are your children doing playing in my garden? How adorable! (Kay & Fillmore 1999: 5)

However, it was recognized early in the constructional literature that it is not always easy, or even possible, to distinguish between conversational and constructional pragmatics. Fillmore (1996), for instance, does not give a definite answer to whether negative questions in English, e.g., (4)–(5), represent a grammatical construction associated with a special pragmatic interpretation, or a case of regularly derivable pragmatics on the basis of a non-logical semantics for natural-language negation (as suggested by Givón 1979):

(4) Don’t you like my dress? (5) Won’t you give me some help?

Such sentences convey that the speaker is expecting a negative answer, and this negative orientation is further accompanied by an implication of surprise or disbelief suggesting



Constructional analysis

that the speaker had originally expected a positive response. A constructional approach would argue that there exists a complex grammatical construction in English, consisting of the yes-no inverted pattern and the negation construction; while this construction “inherits some of the semantics of each of the other two, it has a package of interpretation principles that are, by separate convention, associated with the whole” (Fillmore 1996: 61). Even if there are instances of negative questions that do not have this special interpretation, this does not undermine the constructional analysis, since every sentence bearing this syntax can be the product of either the “special” negative-question construction, or the combination of the two source constructions independently (in the latter case, we do not expect to have the special pragmatics); however, the evidence for assigning constructional status to negative questions, suggests Fillmore, is rather scant. A stronger case can be made for Japanese, where negative questions expressing requests (compare (5) above) are the standard way of expressing polite requests, e.g., (6) (Fillmore 1996: Example (16)): (6) Otya o moo sukosi kure.mas.en ka? tea.obj more a-little give.pol.not QUES ‘Won’t you give me some more tea?’

Although there is a possibility that even here general compositional principles are at work (presumably interacting with or motivated by different rules of politeness), the pragmatic idiosyncracy supports a constructional status for negative questions when used as requests. In subsequent studies, and as the constructional paradigm grows, the focus seems to shift from attempting to tease out the degree of predictability of constructional pragmatics to identifying additional form-meaning/function patterns and their locus within the constructional hierarchy of a given language. In this context, Michaelis and Lambrecht (1996) analyze a type of exclamative sentence they call the nominal extraposition construction, e.g., (7)–(8), associated with special syntactic but also pragmatic properties: 

(7) It’s amazing the difference. (8) It’s astonishing the age at which they become skilled liars. (Michaelis & Lambrecht 1996: (1a,f))

While the nominal extraposition construction shares a number of features with the more familiar right-dislocation one, e.g., (9),

(9) It’s amazing, the difference.

it is, in fact, distinguished from it on the basis of particular discourse-pragmatic constraints: the post-predicate NP in examples like (7) and (8) must denote a referent which (a) is not yet active but is accessible from the discourse topic (see also Lambrecht

19

20

Kiki Nikiforidou

1994), and (b) must be a focal element, as opposed to a topical one (as in (9)). In other words, while in right dislocation the predicate alone is in focus, in nominal extraposition both the predicate and the argument are focal. Given the prosodic features of nominal extraposition, its focal properties at least seem fairly well motivated. A detailed analysis of other exclamative constructions and of the abstract template they all share shows that in fact several of the semanticpragmatic properties of the nominal extraposition construction are motivated, or, in construction-grammar terms, inherited from the abstract exclamative construction. These include (a) the assertion of an expectation contravention, (b) the identifiability (availability in context) of the referent denoted by the post-predicate NP, and (c) the scalar interpretation associated with the pattern. In relation to the scalar interpretation, it often happens (similarly to the case of let alone) that both the relevant scalar parameter and the corresponding property scale have to be inferred in context, as for instance in (10): (10) It’s amazing the odd people my sister knows.

A ‘variety of people’ and a ‘number of people’ interpretation, at least, are possible for this example arising from a metonymic reference of the post-predicate NP (i.e., reference to an entity — the odd people my sister knows — is used to evoke a relevant scalar property of that entity). Importantly, however, the emphasis on the motivated aspects of the construction does not in the least undermine or appear at odds with its constructional status; as argued by the authors, nominal extraposition represents “a gestaltlike interaction of formal, semantic and pragmatic constraints” (Michaelis & Lambrecht 1996: 215). Summing up, constructional analysis has yielded valuable insights as to the nature of the relationship between grammatical, most notably semi-schematic, patterns of the type discussed and pragmatic meaning. The studies cited are but a few among many by now and were chosen precisely for their illustration of the interesting pragmatic and discourse features (pertaining, for instance, to scalarity, incongruity, information structure, and different kinds of affective stance) conventionally and, for the most part, unpredictably associated with particular formal patterns. It is worth noting that although construction grammarians are primarily interested in the relationship between form and pragmatic meaning, it is undeniably the case that their works have significantly enriched the understanding of the corresponding pragmatic domains per se. Scalarity, for instance, has for a long time been a central issue in the pragmatic literature (Fauconnier 1975; Ducrot 1980; Horn 1989, etc.). But the development of the scalar model for addressing the semantics-pragmatics of the let alone construction has helped formalize the notion of two-dimensional scalarity and illuminate the concept of informativeness in fine-grained detail and in ways not previously considered. The same holds for the in-depth analysis of exclamative constructions and other sentence



Constructional analysis

types, which yielded whole new insights into the study of information structure (see in particular Lambrecht 1994, 2004). In this respect, constructional analysis contributes directly to pragmatic analysis, and provides a principled way for integrating the two.

3. Extending the scope: Conventional pragmatics and conventional discourse Construction grammar has sought to provide a clear alternative to Chomskyan mainstream linguistics with its rigid division between grammar and lexicon, productive rules and total idiosyncracy. Understandably, therefore, a central place was assigned to phenomena that exhibited some degree of productivity and, also, syntactic, semantic and pragmatic properties that could not be derived from the productive syntacticsemantic rules or from general pragmatic principles (cf. the emphasis on idiosyncratic properties and subpatterns in Jackendoff 1997, 2008 and Goldberg and Jackendoff 2004). As the constructional approaches multiply, however, the scope of constructional analysis has extended to include all conventional patterns, regardless of their degree of internal predictability. In the words of Goldberg (2006: 5), “[a]ny linguistic pattern is recognized as a construction as long as some aspect of its form or function is not strictly predictable from its component parts or from other constructions recognized to exist. In addition, patterns are stored as constructions even if they are fully predictable as long as they occur with sufficient frequency”. With the parallel development of corpus linguistics in recent years and of statistical methods specifically geared to linguistic analysis (e.g., Gries & Stefanowitsch 2006), constructional approaches have indeed yielded results that fully incorporate the commitments of a usage-based analysis. Constructional status has thus come to be derived from the sufficiently entrenched, i.e., conventionalized, relationship of a particular form with a particular semantics or pragmatics. Kay (2005) recasts as constructions examples like (11)–(14), where the illocutionary force of ‘challenging the wisdom of a suggested course of action’ is conventionally associated with questioning the mental health of the suggestor (Kay 2005: Example (61)):

(11) (12) (13) (14)

Are you crazy? Have you lost your mind? Are you out of your gourd? Is he out of his gourd?

Such examples are not characterized by any particular morphosyntax (other than the interrogative pattern) and the semantics of questioning the suggestor’s mental health is of course fully derivable; so is most likely the pragmatic force of challenging the wisdom of a suggested course of action (cf. Morgan’s (1978) original treatment of these

21

22

Kiki Nikiforidou

cases as “short-circuited implicatures”). Nevertheless, as Kay suggests, they may well be treated as interpretational constructions (or in Morgan’s terms, “conventions of usage” — see also Stefanowitsch 2003). In the same line, the much discussed Can you VP? pattern exemplified in (15), (15) Can you pass the salt?

is conventionally associated with the expression of a request and is even characterized by some degree of non-detachability (cf. *Are you able to pass the salt?), while exhibiting lexico-semantic features normally associated with direct requests (cf. Can you please pass the salt?). The conventionality of these cases is not undermined by the fact that such aspects of meaning may be derived, as noted, by general pragmatic principles. In (16), for instance, (discussed in Langacker 2001: 165–66), (16) Harold has finished his thesis. And I was just elected pope.

we have a familiar pattern: the speaker pretends to make an assertion but follows it with another assertion, so obviously false that the hearer is bound to realize that it is not the intended meaning. While this may be regularly derivable by Gricean Quality or Relevance, it is still the case that speakers learn and use this pattern (of refuting a previous apparent assertion) as a matter of established convention. In Langacker’s view, being a conventionally sanctioned facet of the ability to speak and understand English, this “higher-order” unit cannot be excluded from the linguistic system (unless by an arbitrary decision to restrict conventionality to words and traditionally-defined idioms). While not all construction grammarians necessarily agree with this assessment of what constitutes the grammar of a language, it is clear that several trends within the constructional paradigm have extended their interest to “simple” encoding idiomaticity. According to Fillmore’s original classification (in Fillmore, Kay & O’Connor 1988; see also Croft & Cruse 2004: 231), an “encoding idiom” is an expression that can be interpreted following the standard rules for sentence or phrase interpretation, but is conventional for the particular expression with the particular meaning (e.g., answer the door, bright red, wide open, etc.); even if the meaning of encoding idioms could be figured out by anyone who knows English, they are still idiomatic in that a (nonnative) speaker could not have guessed that this is the conventional, natural-sounding way of expressing these concepts. Consider, for example, Lambrecht’s (2004) analysis of the French right-detached comme-N pattern, e.g., (17): (17) C’ est pas marrant, comme histoire. it is neg funny as story ‘That’s not a funny story.’



Constructional analysis

Both the predicative function of the noun following comme and the fact that this constituent is known by speakers to be non-focal and active (or semi-active) in the discourse are inherited from other constructions in the language (the comme + bare noun construction and the right-topic construction, respectively). Its meaning and discourse function can be seen as resulting from these two features. Nevertheless, it qualifies as a construction since “there is no grammatical rule according to which a sentence expressing this meaning under these discourse conditions must have this particular form” (Lambrecht 2004: 192). As apparent from the discussion in this section, encoding idiomaticity may characterize formal-pragmatic pairings of different types, expanding considerably the domain of conventional, and hence constructional, pragmatics. At the same time, the preceding examples have made clear that often the relevant pragmatic features are really discoursal in nature (e.g., (17), also (7) & (8)) and, as such, they may be linked to formal patterns that transcend sentence boundaries (e.g., (16)). In accordance with its background in Frame Semantics and the early realization that a conceptual frame may well contain discoursal information (including genre and text type — cf. Fillmore 1981), constructional analysis has extended in this direction as well, attempting to integrate conventionalized discourse phenomena into grammar. Östman (2004) thus suggests that construction grammar can and should extend to structures larger than the sentence, even if these were explicitly excluded from its original formulation and in practice left out from most subsequent work. To the extent that much of discourse is also conventionalized in the ways discussed, the systematicity and regularity of discourse patterns needs to be addressed, incorporated in a theory of grammar, and if possible formalized in a way consistent with the rest of formalized grammatical structures. So, for example, recipes, as argued by Östman (2004), may well be taken to represent such a discourse pattern/construction, conventionally associating a particular text type with a particular genre. The text type in this case is the instructive type, coupled, however, with a visual display of ingredients in a list-like fashion, both of these figuring centrally as a gestalt-like template at the formal level of the construction. Crucially, these conventional associations are claimed to have a cognitive basis, giving rise to expectations that govern the interpretation of all relevant texts. Cross-linguistic variation is manifested, as expected, at the level of conventionalized discourse patterns, just as it is manifested with other grammatical constructions; while Finnish contact ads, for example, instantiate the commercial-advertisement pattern, their English counterparts are either instances of the story-with-a-plot pattern or the human-interest one (at a more general level, they all instantiate the descriptive text type). The need to recognize such conventional associations is even stronger in cases where the form of a discoursally licensed pattern is not derived from any other construction in the language and may, in fact, be predicted to be impossible. Such is the case of

23

24

Kiki Nikiforidou

headlines, whose morphosyntax, as shown by Östman (2004), is only possible in the particular pattern, e.g., (18):4 (18) Dog saves baby.

In this case, the absence of articles (both definite and indefinite) is fully conventional and expected and any article-bearing version, even if possible, would certainly be less conventional. This type of licensing is, according to Östman, in principle formalizable in ways that directly reflect the relevant discourse dependencies and even preserve the unification-based formalism of some versions of the theory (Kay & Fillmore 1999; Fried & Östman 2004a). It is interesting to note that just as the studies mentioned in the previous section bring together grammatical inquiry with established topics in the pragmatic literature, the approach outlined above incorporates categories and concepts originating in discourse analysis, and more generally in discourse studies (e.g., genres, text types). The purpose is not to replace the work and insights produced in a discourse-analytic framework, but rather to integrate those which are conventional enough into grammatical theory. As in the case of sentence-level constructions, the constructional method depends on identifying conventional associations between meaning and form. In the cases discussed here, the conventional part of their meaning simply includes knowledge of a discoursal, context-dependent nature. Allowing for suprasentential units to fall within the scope of grammatical research as well as for a focus on spoken language, the constructional approach extends in yet another direction of discourse analysis, namely conversation analysis (Linell 1998). The potential of a constructional approach for analyzing conversational patterns has been noted by researchers within the conversational paradigm itself, e.g., Fox and Thompson (1990) for relative clauses and Couper-Kuhlen and Thompson (2000) for concessive patterns. Importantly for present purposes, the constructional method has been applied to the analysis of dialogic data by construction grammarians as well. Fried and Östman (2005) analyze a set of pragmatic particles in spoken Czech and Solv, arguing that construction grammar is well equipped to deal with spoken language, provided it extends its unit (the construction) to include dialogue chunks. In some of their uses, the particles under investigation indeed display properties that involve dialogic coherence, while all of their uses refer to some broadly understood pragmatic parameter such as politeness or involvement. In a constructional framework, such parameters can be expressed as attributes, incorporating all discourse-relevant information, each having a range of potential values. The coherence attribute, for instance, can be analyzed as having four subtypes, each corresponding to specific values, e.g., Table 1 (Fried & Östman 2005: 1774): 4.  Strictly speaking, it is also possible in two other settings as well, family conversation and interlanguage, although none of these seems to raise such conventional expectations.



Constructional analysis

Table 1.  Coherence attribute: Subtypes and values Attributes

Values

Coherence Type of speech act

Question Request Assertion, etc.

Speaker information (sex, age, …)

M/F, …

Shift of discourse topic

Yes (=> new topic) No (=> keep old topic, or topic specification not relevant)

Expectations vis-à-vis hearer

Yes (expecting response, specific response) No

While the details of the analysis (e.g., the choice and number of parameters, types and subtypes) may be open to revision, it is important that such studies attempt to systematize pragmatic/discoursal knowledge in a way analogous to other kinds of speaker knowledge which is reflected in grammar. In a constructional framework specifically, all kinds of syntactic and semantic information relevant to the description of a lexical or grammatical (usually sentence-level) construction may be represented in terms of attribute-value pairs, which collectively specify the internal and external syntactic and semantic behavior of the construction at hand. The discoursal/pragmatic properties outlined above complement such descriptions in a compatible formalism. The goal is not to project construction grammar as a framework designed for modeling human interaction (Fried & Östman (2005) readily acknowledge that it cannot substitute a primarily dialogical approach such as conversation analysis), but rather to integrate into grammar those aspects of conversation that are conventionally tied to recognizable formal clues. Attempts to integrate discoursal parameters into grammar can also be found in certain cognitive analyses, which do not fall explicitly in the constructional paradigm, yet are clearly construction-based. Verhagen (2005) focuses, for instance, on intersubjective constructions, e.g., negation and discourse connectives, whose purpose, he argues, lies precisely in cognitive inter-speaker coordination; as such, they require reference to discoursal parameters. This analysis is also usage-based, a feature of constructional analysis that will be taken up in the next section. In analyzing conditional constructions in English, Dancygier & Sweetser (2005) adopt a construction-grammar framework, though without any attempt at formalization, and for certain conditional patterns they refer to discoursal/pragmatic parameters of the sort proposed in Fried and Östman (2005). Therefore, a common theme in all the research discussed in this section is that linguistic units can be identified as constructions not only on the basis of their morphosyntactic and semantic conventional makeup, but also on the basis of their discoursal and/or sociocultural currency.

25

26

Kiki Nikiforidou

4.  Constructions in grammaticalization Ever since its initial formulation in Meillet (1912), grammaticalization (the process whereby lexemes or lexical items become grammatical) has been conceived as a context-dependent, holistic process; its typical instances simultaneously affect the phonological, morphosyntactic and semantic properties of the segment(s) undergoing the change. Crucially, even changes affecting single lexical items really take place in very specific, multi-parameter constructional environments; whole constructional patterns may also become more grammaticalized. Soon enough, the early grammaticalization literature (Lehmann 1992; Hopper & Traugott 1993) recognized the role of the constructional environment in this kind of change, even if the concept of the construction was not given a technical definition. In contrast, therefore, to diachronic studies that sought to describe syntactic change in isolation (e.g., Lightfoot 1979), most work on grammaticalization essentially adopts a constructional approach to language, consistent with its origins in traditional grammar and functional linguistics.5 The purpose of this section is not to cover all constructionally oriented work in grammaticalization, but to discuss certain approaches which explicitly adopt one of the constructional frameworks in analyzing diachronic data. An overlapping area of interest between construction grammar(s) and much work in grammaticalization lies precisely in conventional/constructional pragmatics. A dominant view in the literature is that grammaticalization involves pragmatic strengthening, i.e., is characterized by a unidirectional tendency for meanings to increasingly code speaker attitude or evaluative stance toward a situation, whether this is belief, personal commitment to the assertion, or assessment of truth and all kinds of epistemic relation among propositions. This strengthening of the speaker’s pragmatic viewpoint has been termed “subjectification” (Traugott 1989, 2003). Pragmatic strengthening occurs through the rise, gradual entrenchment, and subsequent conventionalization of implicatures created when a word or particular linear string appear in specific contexts, characterized both formally (in terms of intonation and morphosyntax) and semantically-pragmatically. An example of subjectification is the well studied reanalysis of be going to from the meaning ‘be in motion with a particular purpose’ to the meaning of ‘planned futurity’ (Langacker 1990; Hopper & Traugott 1993; Bybee 2003). As Bybee et al. (1994: 297) observe, “context is all important” in the development of motion verbs into future markers and must include both an allative syntactic and semantic component (‘movement toward’) and a progressive, present or imperfective verb form (‘movement in progress’). In such environments,

5.  Traugott (2003) contains an overview of different grammaticalization approaches with a view to their more or less constructional orientation.



Constructional analysis

the pattern is associated with the inference that the subject intends a future occurrence of the planned action. It is obvious that a constructional approach to grammar provides appropriate methodological tools for dealing with this type of change. It allows for an integrated account of both the input to the diachronic process, including all relevant formal, semantic and on occasion pragmatic features that license the change, and the output, which in grammaticalization is bound to include a meaning which is essentially pragmatic (Nikiforidou 1996). This is acknowledged in recent work (Traugott 2003, 2007), where approaches like Goldberg’s (1995), Kay and Fillmore’s (1999) and Croft’s (2001) are cited as having “obvious relevance” and where non-constructional approaches are considered to espouse a view of grammar that “has not envisaged the importance of studying interfaces with pragmatics” (Traugott 2003: 627). What is perhaps less obvious is that a constructional perspective may illuminate the grammaticalization process as such, offering insights into the intermediate phases of the process and fleshing out the concepts of entrenchment and conventionalization of an implicature. Being usage-based, most constructional approaches take into account frequency effects (see Section 3), arguing that frequency may be enough to ensure the conventionalization and mental storage of a pattern. Recent refinements in corpus methods have further made it possible to measure the strength of the association of a particular item with a particular context, directly reflecting the effect of the licensing conditions and the potential priming of a word in a particular construction, i.e., the “collostructional measurement” (Stefanowitsch & Gries 2003; see also Lindquist & Mair 2006). This is the perspective adopted, for instance, in Bybee (2006) for the analysis of be going to, where it is argued that the high frequency of go(ing) in the more general purpose construction (e.g., I was sending to use Lord Timon myself…, Shakespeare, Timon of Athens, II.2) is the basic motivating factor for the development of the autonomous future construction. The approach is developed further in Hilpert (2008), who analyzes future constructions in the Germanic languages (including English be going to) with an emphasis on the main verbs that occur in such environments; statistical co-occurrence preferences between verbs and constructions are shown to figure in the grammaticalization process, allowing further refinement of the parameters that underlie the creation and entrenchment of implicatures. Lexically filled, instance-based constructional patterns occurring with sufficient frequency are in fact shown to play an important role both in the (simple) conventionalization of meaning as in the case of “prefabs” (Bybee’s term for encoding idioms), which result from relatively low levels of repetition, and in grammaticalization, which results from higher frequencies. Therefore, they constitute arguments for constructional approaches, like those in Kay and Fillmore (1999), Goldberg (1995, 2006), Lakoff (1987), Langacker (1987), and Croft (2001) (Bybee 2006: 715).

27

28

Kiki Nikiforidou

It is worth noting that constructionally oriented historical work may conversely contribute to the refinement of construction-grammar conceptual and methodological tools. As an example, consider the case of coercion, a mechanism assumed to operate in some versions of construction grammar (Michaelis 2004; Fried & Östman 2004a) in order to resolve semantic differences between a word and its constructional (morphosyntactic) context. In Michaelis’ approach, coercion is claimed to be realized in two distinct ways, namely as explicit or implicit “type-shifting”. Explicit type-shifting is the shift in the designation of a lexical item by a grammatical construction with which the particular word is conventionally combined. An example of explicit type-shifting is the change undergone by a word such as bread in the partitive construction (a piece/bit/ slice of bread). In Michaelis’ (2003) analysis, the partitive requires an unbounded/mass complement and bread has indeed mass semantics (at least prototypically); the shifting involved (from the unbounded/mass construal of the complement to the part/portion interpretation) is, therefore, the conventional effect of the construction. Implicit typeshifting, on the other hand, refers to the shift in the designation of a lexical item in order to reconcile semantic conflicts between word and construction. For example, a typically count/bounded noun, such as sheet, is said by Michaelis (2003: 174) to be coerced into a mass/unbounded reading by the partitive construction (e.g., a shred of sheet). As argued, however, by Traugott (2007), historical corpus data might lead one to question the validity of some of the assumptions above, and may consequently force one to reconsider the validity of the associated distinctions. Data from Early and Middle English, as well as from modern use, show, for instance, that the partitive does not really require unbounded complements (e.g., a piece of a chromosome may break off … (Traugott 2007: 533)), while several lexical items which commonly occur in this construction (e.g., words denoting vegetables or cooked foods) may be systematically construed as either bounded or unbounded; this calls into question the generality of type-shifting, at least for such constructions. In the same line, the absence from the corpus of examples like a shred of sheet may call into question whether “coercion of this type actually occurs, or if it does, whether it occurs frequently enough to make the distinction between explicit and implicit type-shifting theoretically important” (Traugott 2007: 547). In short, while this particular distinction may be motivated by a variety of other data, the grounding and perspective offered by historical analysis are valuable in constraining the possibilities and refining the explanations. 5.  Summary and prospects Constructional approaches have uncovered the rich pragmatics associated with semi-schematic, grammatical patterns and illuminated the subtle relationships between the formal components of such patterns and the pragmatic meaning they



Constructional analysis

evoke. The productive character of the constructions at hand, in conjunction with the non-predictability of their semantic and/or pragmatic properties, has been the foundational argument for constructional theories. In arguing for the constructional organization of linguistic knowledge, constructional analyses have thus contributed genuinely to the understanding of topics central to the pragmatics literature, e.g., scalarity and information structure. The realization that conventionality (in the sense of entrenchment through repeated use) is enough to ensure construction status has extended considerably the domain of the constructional method. Any conventional form-meaning pairing, regardless of the predictability of its meaning or function, can be recognized as a construction, provided it occurs with sufficient frequency and the form automatically, or even prototypically, evokes a particular meaning. This, in turn, increases considerably the range of data which in earlier literature would have been thought of as “conventional implicatures” and provides a direct lifeline to work in grammaticalization. Indeed, the constructional approach offers the necessary theoretical couching for the analysis of changes which are characteristic of grammaticalization. As shown repeatedly in the relevant studies, both the input and the output of such changes must be stated in terms of a holistic constructional environment, simultaneously including morphosyntactic, semantic, and pragmatic properties. At the same time, the usage-based character of most constructional approaches may realistically accommodate the process of “conventionalization of an implicature”, which has been argued to characterize most of grammaticalization, by taking into account precisely the co-occurrence frequencies of words and constructions. The preoccupation of construction grammarians with meaning and the extension into aspects of encoding idiomaticity in language inevitably brings construction grammar close to the phraseology and collocations tradition (e.g., Sinclair 1991, 2004). It has recently been recognized by scholars in this tradition that construction grammar has much to contribute (to phraseology and lexicography) “by way of theoretical insights into the nature of the lexicon, and a synthesis would benefit both theoretical linguistics and practical lexicography” (Hanks 2008: 228; see also Moon 1998; Fillmore 2008). However, construction grammar has also been criticized by phraseology scholars for not being usage-based (enough) nor adopting a suprasentential perspective on discourse and pragmatics (Hanks 2008: 228). The studies cited in Sections 3 and 4 clearly show that such criticism is no longer valid and, if anything, restricted to the seminal analyses of sentence-level constructions, which were not necessarily relying on any corpus for evidence of conventionality. As discussed above, constructional approaches have already crossed sentence boundaries, and the constructional method has been applied to the conventional links of discoursal and conversational properties with (possibly) larger formal patterns. In addition, most constructional approaches, as said, are truly usage-based and fully exploit the explanatory power of frequency and relative

29

30

Kiki Nikiforidou

entrenchment. In harboring a real concern for conventionality, constructional approaches can therefore converge with phraseology research, providing a much needed, theoretical, and all-encompassing framework for the integration of the description of conventional knowledge.

References Bernal, E. & J. Decesaris (eds.) (2008). Proceedings of the XIII Euralex International Congress. Universitat Pompeu Fabra. Bybee, J. (2003). Mechanisms of change in grammaticization: The role of frequency. In B.D. Joseph & R.D. Janda (eds), The handbook of historical linguistics, 602–23. ——— (2006). From usage to grammar: The mind’s response to repetition. Language 82: 711–33. Bybee, J., R. Perkins & W. Pagliuca (1994). The evolution of grammar: Tense, aspect, and modality in the languages of the world. University of Chicago Press. Cole, P. (ed.) (1978). Syntax and semantics 9: Pragmatics. Academic Press. ——— (ed.) (1981). Radical pragmatics. Academic Press. Couper-Kuhlen, E. & S.A. Thompson (2000). Concessive patterns in conversation. In E. Couper-Kuhlen & B. Kortmann (eds.), Cause-condition-concession-contrast: Cognitive and discourse perspectives, 381–410. Couper-Kuhlen, E. & B. Kortmann (eds.) (2000). Cause-condition-concession-contrast: Cognitive and discourse perspectives. Mouton de Gruyter. Croft, W. (2001). Radical construction grammar: Syntactic theory in typological perspective. Oxford University Press. Croft, W. & A. Cruse (2004). Cognitive linguistics. Cambridge University Press. Cuyckens, H., R. Dirven & J.R. Taylor (eds.) (2003). Cognitive perspectives on lexical semantics. Mouton de Gruyter. Dancygier, B. & E. Sweetser (2005). Mental spaces in grammar: Conditional constructions. Cambridge University Press. Ducrot, O. (1980). Les échelles argumentatives. Minuit. Fauconnier, G. (1975). Pragmatic scales and logical structure. Linguistic Inquiry 6: 353–75. Fillmore, C. (1981). Pragmatics and the description of discourse. In P. Cole (ed.), Radical pragmatics, 143–166. ——— (1985). Frames and the semantics of understanding. Quaderni di Semantica 6: 222–54. ——— (1996). The pragmatics of constructions. In D. Slobin et al. (eds.), Social interaction, social context, and language, 53–69. ——— (2008). FrameNet meets construction grammar. In E. Bernal & J. DeCesaris (eds.), Proceedings of the XIII Euralex International Congress, 49–68. Fillmore, C. & S. Atkins (1992). Toward a frame-based lexicon: The semantics of RISK and its neighbors. In A. Lehrer & E. Feder Kittay (eds.), Frames, fields, and contrasts: New essays in semantic and lexical organization, 75–102. Fillmore, C., P. Kay & M.C. O’Connor (1988). Regularity and idiomaticity in grammatical constructions: The case of let alone. Language 64: 501–38. Fox, B. & S.A. Thompson (1990). A discourse explanation of the grammar of relative clauses in English conversation. Language 66: 297–316. Fried, M. & J.-O. Östman (2004a). A thumbnail sketch of construction grammar. In M. Fried & J.-O. Őstman (eds.), Construction grammar in a cross-language perspective, 11–86.



Constructional analysis

Fried, M. & J.-O. Östman (eds.) (2004b). Construction grammar in a cross-language perspective. John Benjamins. Fried, M. & J.-O. Östman (2005). Construction grammar and spoken language: The case of pragmatic particles. Journal of Pragmatics 37: 1752–78. Givon, T. (1979). On understanding grammar. Academic Press. Goldberg, A. (1995). Constructions: A construction grammar approach to argument structure. University of Chicago Press. ——— (2006). Constructions at work: The nature of generalization in language. Oxford University Press. Goldberg, A. & R. Jackendoff (2004). The English resultative as a family of constructions. Language 80: 532–68. Gries, S. Th. & A. Stefanowitsch (eds.) (2006). Corpora in cognitive linguistics. Mouton de Gruyter. Hanks, P. (2008). The lexicographical legacy of John Sinclair. International Journal of Lexicography 21: 219–29. Hilpert, M. (2008). Germanic future constructions: A usage-based approach to language change. John Benjamins. Hopper, P. & E. Traugott (1993). Grammaticalization. Cambridge University Press. Horn, L. (1989). A natural history of negation. University of Chicago Press. Horn, L. & G. Ward (eds.) (2005). The handbook of pragmatics. Blackwell. Jackendoff, R. (1997). Twistin’ the night away. Language 73: 534–59. ——— (2008). Construction after construction and its theoretical challenges. Language 84: 8–28. Joseph, B.D.& R.D. Janda (eds.) (2003). The handbook of historical linguistics. Blackwell. Kay, P. (2005). Pragmatic aspects of grammatical constructions. In L. Horn & G. Ward (eds.), The handbook of pragmatics, 675–700. Kay, P. & C. Fillmore (1999). Grammatical constructions and linguistic generalizations: The what’s X doing Y? construction. Language 75: 1–33. Lakoff, G. (1987). Women, fire, and dangerous things: What categories reveal about the mind. University of Chicago Press. Lambrecht, K. (1994). Information structure and sentence form. Cambridge University Press. ——— (2004). On the interaction of information structure and formal structure in constructions: The case of French right-detached comme-N. In M. Fried & J.-O. Östman (eds.), Construction grammar in a cross-language perspective, 157–99. Langacker, R. (1987). Foundations of cognitive grammar, Vol. I: Theoretical prerequisites. Stanford University Press. ——— (1990). Subjectification. Cognitive Linguistics 1: 5–38. ——— (2001). Discourse in cognitive grammar. Cognitive Linguistics 12: 143–88. Lehmann, W. (1992). Historical linguistics. Routledge. Lehrer, A. & E. Feder Kittay (eds.) (1992). Frames, fields, and contrasts: New essays in semantic and lexical organization. Lawrence Erlbaum Associates. Lightfoot, D. (1979). Principles of diachronic syntax. Cambridge University Press. Lindquist, H. & C. Mair (eds.) (2006). Corpus approaches to grammaticalization. John Benjamins. Linell, P. (1998). Approaching dialogue, talk, interaction and contexts in dialogical perspectives. John Benjamins. Meillet, A. (1912). L’évolution des formes grammaticales. Scientia 12. (Reprinted in A. Meillet 1948, Linguistique historique et linguistique générale 1, 130–48.) Michaelis, L. (2003). Word meaning, sentence meaning, and syntactic meaning. In H. Cuyckens, R. Dirven & J.R. Taylor (eds.), Cognitive perspectives on lexical semantics, 163–210. ——— (2004). Entity and event coercion in a symbolic theory of syntax. In J.-O. Östman & M. Fried (eds.), Construction grammars: Cognitive grounding and theoretical extensions, 45–88.

31

32

Kiki Nikiforidou Michaelis, L. & K. Lambrecht (1996). Toward a construction-based theory of language function: The case of nominal extraposition. Language 72: 215–47. Moon, R. (1998). Fixed expressions and idioms in English: A corpus based approach. Clarendon Press. Morgan, J. (1978). Two types of convention in indirect speech-acts. In P. Cole (ed.), Syntax and Semantics 9: Pragmatics, 261–80. Nikiforidou, K. (1996). Modern Greek as: A case study in grammaticalization and grammatical polysemy. Studies in Language 20: 599–632. Östman, J.-O. (2004). Construction discourse: A prolegomenon. In J.-O. Östman & M. Fried (eds.), Construction grammars: Cognitive grounding and theoretical extensions, 121–44. Östman, J.-O. & M. Fried (eds.) (2004). Construction grammars. Cognitive grounding and theoretical extensions. John Benjamins. Panther, K.U. & L. Thornburg (eds.) (2003). Metonymy and pragmatic inferencing. John Benjamins. Sinclair, J. (1991). Corpus, concordance, collocation. Oxford University Press. ——— (2004). Trust the text: Language, corpus and discourse. Routledge. Slobin, D., J. Gerhardt, A. Kyratzis & J. Guo (eds.) (1996). Social interaction, social context, and language. Lawrence Erlbaum Associates. Stefanowitsch, A. (2003). A construction-based approach to indirect speech-acts. In K.-U. Panther & L. Thornburg (eds.), Metonymy and pragmatic inferencing, 105–126. Stefanowitsch, A. & S. Th. Gries (2003). Collostructions: Investigating the interaction between words and constructions. International Journal of Corpus Linguistics 8: 209–43. Tomasello, M. (2003). Constructing a language: A usage-based theory of language acquisition. Harvard University Press. Traugott, E. (1989). On the rise of epistemic meanings in English: An example of subjectification in semantic change. Language 65: 31–55. ——— (2003). Constructions in grammaticalization. In B.D. Joseph & R.D. Janda (eds.), The handbook of historical linguistics, 624–47. ——— (2007). The concepts of constructional mismatch and type-shifting from the perspective of grammaticalization. Cognitive Linguistics 18: 523–57. Verhagen, A. (2005). Constructions of intersubjectivity. Oxford University Press.

Control phenomena Benjamin Lyngfelt* University of Gothenburg

1.  Introduction Control, as the term is generally understood in syntax, concerns the interpretation of non-finite clauses such as infinitivals or gerunds.1 It may be defined as a coreference relation between the understood subject of a non-finite clause and some other element that provides its interpretation. This element is called its controller. Some different types of controllers are illustrated in (1): (1) a. They asked him to leave the party. b. By working extra shifts she could finally afford that trip. c. To be honest, it is worth a lot more.

In the ditransitive structure in (1a), the understood subject of leave the party is controlled by the matrix object him; and in the initial adjunct of (1b), the understood subject of working extra shifts is controlled by the matrix subject she. In (1c), there is no syntactically realized controller, since the understood subject of be honest is controlled by the speaker (or writer). The controlled element, the understood subject, is often referred to as PRO and represented as such in linguistic examples (e.g., Theyi asked himj to PROj leave the party). In some generative frameworks, this PRO is assumed to correspond to an invisible pronoun. For expository reasons, I will henceforth follow the common practice of representing the controlled element as PRO. I will not, however, subscribe to the assumption that it marks an actual invisible pronoun, but rather aim at as theory-neutral a presentation as possible. An often recognized distinction among control structures is the one between obligatory and non-obligatory control. This distinction has been pictured in many

*Benjamin Lyngfelt is a Swedish Academy research fellow, sponsored by the Alice and Knut Wallenberg Foundation. Thanks to Maia Andréasson and an anonymous reviewer for valuable comments on earlier versions of this article. 1.  The term control has also been applied to event semantics. This notion of control is employed in Dik’s (1997) Functional Grammar as a State of Affair feature.

34

Benjamin Lyngfelt

different ways, but a fairly representative definition is that “[t]he controller and the infinitive must be clausemates” (Landau 2000: 3; for a more elaborate definition, see, e.g., Williams 1980: 209). Thus, (1a) is a typical example of obligatory control, and (1c) a clear case of non-obligatory control (although the interpretation of speaker control is of course obligatory in some sense). As for adjuncts such as the one in (1b), their status with respect to this distinction is less clear; we will return to this issue in Section 3. Another central distinction is the one of “ordinary” control patterns versus so-called arbitrary control, essentially a distinction between control and non-control. Arbitrary control refers to cases where there is no controller and PRO instead receives a generic or arbitrary interpretation, as illustrated in (2): (2) a. PROarb Playing golf is good for you. b. John is easy to PROarb talk to.

Both sentences in (2) are presented as general statements, supposedly meant to be true for anyone playing golf or anyone talking to John, respectively. In other words, there is no controlling referent, and PRO is assigned generic or arbitrary reference. As is the case with other generic expressions (e.g., the pronoun one), the context may provide a more specific reading, often associated with the speaker, but the expressions as such do not call for a specific interpretation of PRO. Notice the contrast with (1c), where there clearly is a controller, although not linguistically realized. Arbitrary control will be treated further in Section 4. Finally, control should be distinguished from so-called raising structures. Compare the sentence pairs (3a–b) and (3c–d): (3) a. They asked Maryi to PROi sing at the reception. b. They heard Mary sing at the reception. c. Maryi promised to PROi handle the situation just fine. d. Mary seemed to handle the situation just fine. e. It seemed Mary handled the situation just fine.

The difference, indicated by there being no PRO in (3b) or (3d), has to do with semantic roles. In (3a), Mary fills two semantic roles, both as askee (Patient of asked) in the matrix and as singer (Agent of sing) in the subordinate infinitive (cf. the finite variant They asked Mary if she could sing at the reception, where each role is expressed separately). In (3b), by contrast, Mary plays no semantic role in the matrix clause. The Theme of heard consists of the whole string [Mary sing at the reception], and thus Mary fills a semantic role only in relation to the subordinate predicate (cf. They heard that/how Mary sung at the reception). The same kind of difference concerns (3c–d); again, Mary fills two semantic roles in (3c) but only one in (3d). In (3c) she is both promiser (Agent of promised) and handler (Agent of handle) (cf. Mary promised that she would handle the situation just



Control phenomena

fine). In (3d), although Mary is the grammatical subject of seemed, she does not fill any semantic role with respect to it (there is no such thing as a “seemer”). Instead, the Theme of seemed is [Mary handle the situation just fine], as illustrated in (3e), where there is an expletive subject and the whole Theme is expressed as a complement. Thus, control is a relation between distinct elements, whereas raising is not. Regardless of whether PRO is assumed to be an invisible subject or merely a feature of the verb, it corresponds to a semantic role separate from that of its controller.2 There is a large theoretical literature on control, primarily focusing on obligatory control and how to account for it within a certain framework. Most syntactic frameworks have their own sub-theories of control, such as Manzini (1983) for Government & Binding, Hornstein (1999) and Landau (2000) for the Minimalist Program, Bresnan (1982) for Lexical-Functional Grammar, and Sag & Pollard (1991) for Head-Driven Phrase Structure Grammar, to name but a few. An overview of the development of control theories within Chomskyan grammars is given in Davies & Dubinsky (2004). I will not review these various proposals here, since this is an overview article on control phenomena rather than control theories. Examples will be mainly taken from English and to some extent Swedish. The latter language is included due to the availability of a comprehensive corpus study of control phenomena in Swedish (Lyngfelt 2002), including some patterns that are often disregarded in the control literature.3 2.  Complement control – object clauses The control issue that has been most thoroughly treated in the theoretical literature is complement control, in particular obligatory control of non-finite object clauses (other complement types will be addressed in Sections 2.2 and 5). The basic descriptive generalizations are the following. In monotransitive structures, PRO is normally controlled by the matrix subject, as in (4a). In ditransitive structures, PRO is usually controlled by the matrix object, as in (4b) – except for some verbs, notably promise, where it is rather controlled by the matrix subject, as in (4c): (4) a. Ii hate to PROi admit it. b. Shei persuaded himj to PROj stay. c. Shei promised himj to PROi stay.

2.  There have been proposals to collapse the distinction between control and raising (e.g., Hornstein 1999). I will not go into this debate here (cf. some of the papers in Davies & Dubinsky 2007), but merely observe that most accounts of control maintain this distinction. 3.  Lyngfelt (2002) is written in Swedish. For a preliminary presentation in English, see Lyngfelt (2000).

35

36

Benjamin Lyngfelt

The various theories proposed to account for obligatory control in verb complements are essentially of two types: configurational or semantic/lexicalist. Configurational accounts, following Rosenbaum (1967), are typically based on the so-called Minimal Distance Principle (MDP), according to which PRO is controlled by the closest NP higher in the phrase structure (in generative terms, the closest “c-commanding” NP). In sentences such as (4a), the subject is closest and hence, the controller. In ditransitives, there is an intervening object, usually resulting in object control as in (4b) – but not in (4c). Thus, MDP accounts for the dominating patterns of subject control in simple transitive structures and of object control in ditransitives, but leaves promise as a lexical exception. On the semantic/lexicalist approach, first pursued by Jackendoff (1972), these control patterns are essentially semantic and depend on the meaning of the matrix verb. In other words, the verbs in (4) do not only select for the number and types of arguments they are construed with but also for the control relations associated with these arguments. Assuming it follows from the verb hate that it takes an infinitival complement in (4a), it also follows from the meaning of hate that this complement receives subject control. Of the ditransitive structures, the meaning of persuade specifies object control in (4b) whereas the meaning of promise yields subject control in (4c). Lyngfelt (2002) introduces the term orientation for this aspect of verb meaning; verbs like promise are oriented towards the subject, and verbs like persuade are oriented towards the object (cf. also Sag & Pollard 1991; Culicover & Jackendoff 2005). Verbs with similar meanings display similar control patterns in other languages as well. Both the configurational and the lexicalist approach are well attested in the literature, although it may be that the configurational approach has been losing ground lately. Even from a Minimalist perspective, a tradition usually favoring configurational solutions, Landau (2000) concludes Overall, the MDP is untenable, and the question of controller choice (inside the O[bligatory] C[ontrol] domain) is best handed over to semantic/pragmatic considerations. (Landau 2000: 24)

2.1  Control shift One of the main arguments for a semantic approach to complement control concerns control shift (or coercion, cf. Sag & Pollard 1991). This notion pertains to cases where the control patterns usually associated with certain verbs are reversed, as illustrated in (5): (5)

a. b. c. d.

Hei begged herj to PROj give a talk. Hei begged herj to PROi be allowed to give a talk. Shei promised himj to PROi play a few songs at the party. Shei promised himj to PROj get to play a few songs at the party.



Control phenomena

Verbs like beg are object-oriented and thus usually occur with object control, as in (5a), but can under certain circumstances get subject control, as in (5b). Conversely, a subject-oriented verb like promise (cf. (5c)) may occasionally be construed with object control, as in (5d). In both (5b) and (5d), the control relation is shifted, at least from a syntactic perspective, since PRO associates with a different element than expected. However, the orientation of the matrix verbs does not shift, and neither do the semantic control relations. It is still the expected controllers who are in control of the subordinate events. In (5b), the object of beg (i.e., her) is not coreferential with the syntactic subject (PRO) of be allowed but with its Agent, and it is the Agent who is in control of the event. Likewise, in (5d), the subject of promise (i.e., she) is not the controller of PRO but is nonetheless in control of who gets to play. In other words, one may conclude that the control relation specified by the meaning of beg and promise, respectively, does not directly concern the syntactic subject of the subordinate clause, but rather a semantic element in control of the subordinate event. Farkas (1988) introduces the feature RESP(onsibility) to account for this element, and Sag & Pollard (1991) assume an X-ARG feature for the same purpose. Given that the control relation primarily concerns a semantic element, determining the understood subject of the subordinate clause then becomes a matter of linking syntax and semantics. Usually, RESP/X-ARG associates with the subject, but in (5b) and (5d) it does not – precisely because the understood subject is not the element in control of the event. Control shift is favored by passivization of the subordinate clause, as in (5b) – but it does not require passive per se, since a shift also occurs in (5d), which is at least formally active. Given the right circumstances, it may even occur with an active Agent as the understood subject, as illustrated in (6):

(6) The pupili asked the teacherj to PROi/j leave early. (example from Farkas 1988)

Although ask is an object-oriented verb, usually assigning object control, the pragmatic circumstances enable a subject-control reading in (6), due to the well-known authority relations between pupil and teacher. Note, however, that even under a subject-control reading, the object (the teacher) is still in control of the event, being in a position to allow (or forbid) the pupil to leave early. It should be noted that control shift is a fairly uncommon phenomenon in actual language use, but it is fairly widespread across languages. It is also theoretically relevant as a strong argument in favor of a semantic approach to controller assignment – at least regarding complement control. 2.2  Other kinds of complement control Most accounts of complement control focus on objects. As for other complement types, the above reasoning applies to some but not all of them. First, it should be noted that English wh-infinitivals differ from ordinary infinitival objects in that they often

37

38

Benjamin Lyngfelt

receive arbitrary rather than obligatory control. One example of each type is given in (7) – examples from Barrie (2007: 263). There are also some few cases of arbitrary control in ordinary non-finite object clauses (cf. Section 4). (7) a. Johni knows when to PROi wash the dishes. b. Maryi learned how to PROarb fly a 747.

Prepositional verb complements follow the general pattern for objects, i.e., subject control in single complement structures and object control in two-complement structures, as illustrated by the Swedish examples in (8): (8) a.

Har dui försökt med att PROi starta om datorn? (Sw.) have youi tried with to PROi start again computer-def ‘Have you tried rebooting the computer?’

b. Allmänt brukar jagi avråda folkj från att PROj köpa alltför generally use Ii advice-against peoplej from to PROj buy too

små bilar. (Sw.) small cars



‘I usually recommend people not to buy too small cars.’

Predicative complements (also called subject complements) behave entirely differently from other verb complements, and they are interpreted according to a pattern of indirect control (cf. Section 5.2). As for complements of nouns and adjectives, these are treated in Section 5.1. 3.  Adjunct control If obligatory control in object clauses has been dominating the theoretical control literature, adjunct control is a larger issue in traditional grammar – and especially in normative grammar. Usually, PRO in adverbial adjuncts and free modifiers is controlled by the matrix subject, as in (9): (9) a. After PROi finishing her studies, shei decided to try out her luck overseas. b. Youi must apply to the embassy to PROi get a visa. c. PROi Just happy to get a job, hei didn’t mind the low salary.

Exceptions to this pattern display what may be called pragmatic control, where – instead of the subject – PRO is controlled by some other pragmatically salient element in the context.4 This is illustrated in (10) (note that an ungrammaticality star on an index means that the indicated coreference relation is unavailable): 4.  Other terms for what is called pragmatic control here include logophoric control (e.g., Williams 1992) and control by sentence topic (Kawasaki 1993).



Control phenomena

(10) a. After PRO*i finishing her studies, an overseas jobi was a welcome change. b. An applicationi must be sent to the embassy to PRO*i get a visa. c. PRO*i Just happy to get a job, the salaryi was a lesser concern.

There is some disagreement regarding the acceptability of such sentences. Therefore, they make for a controversial topic in normative grammar and are sometimes discouraged by usage guides (under the label of dangling or unattached modifiers). It is clear, however, that pragmatic control is way too common and too widely accepted to be simply disregarded as ungrammatical. At most, it may be considered a fault of style. The acceptability issues regarding pragmatic control are not typical of English but also concern the Scandinavian languages and, to varying degrees, presumably all languages with similar constructions. I will not dwell further on the normative aspects of adjunct control here, but rather focus on the descriptive facts. In a corpus study of Swedish, Lyngfelt (2002) distinguishes three properties that significantly favor pragmatic (non-subject) control in adverbial adjuncts. In fact (except for a couple of special types to be illustrated in (13) below), it is hard to come across cases of pragmatic control that do not exhibit at least one of these three characteristics:

A. sentence-initial adjunct B. passive matrix clause C. expletive matrix subject

The relevance of sentence-initial position (A) has been stressed in a number of works and with respect to several languages. It is quite clear that pragmatic control is both more common and less restricted in sentence-initial adjuncts. In non-initial adjuncts, pragmatic control is relatively rare and typically restricted to cases where the matrix subject is ruled out as a controller for semantic reasons. For instance, subject control is unavailable in (10b) since an application cannot acquire a visa. In sentence-initial adjuncts, by contrast, pragmatic control seems to be pretty much freely available as long as it suits the context. An example where subject control would be semantically possible, but where pragmatic control is still the preferred interpretation, is given in (11): (11) After PROj finishing the dissertation, theyi couldn’t blame herj for taking a vacation.

Passivization (B) also favors pragmatic control, as again illustrated in (10b), where PRO is controlled by the Agent of be sent rather than the subject. In passive sentences, both the subject and the Agent are viable controllers; and control by the matrix Agent, as in (10b), is fairly common. Finally, expletive subjects (C) make poor controllers, which leaves the field open for pragmatic control, as in (12): (12) After PROj/*i reading this book, iti was hard for mej to eat any type of meat for a while.

39

40

Benjamin Lyngfelt

Common to passives and expletives is that they represent non-agentive subjects. This property is also common in other cases of pragmatic control, like (10a) and (10c). It is not, however, non-agentivity as such that is the decisive property but rather the low (pragmatic) saliency of these subjects. If the subject is pragmatically salient, we do get subject control. As an additional observation, it should be noted that there are some cases of pragmatic control that are entirely uncontroversial from a normative perspective, for instance the two types exemplified in (13): (13) a. Theyi sent mej back to PROj get another one. b. PRO Judging from experience, some students are bound to get this wrong.

As shown in (13a), it is perfectly acceptable for PRO in a sentence-final purpose clause to be controlled by the matrix object instead of the subject. As for speaker-oriented adverbials, as in (13b), pragmatic control (more specifically, control by the speaker) is the only interpretation. Even from a strict normative viewpoint, these cases must be regarded as exceptions to the subject control pattern rather than violations of it. More to the point, they are usually disregarded altogether – presumably precisely because they are uncontroversial. Finally, with respect to the distinction between obligatory and non-obligatory control, the dominance of subject control has prompted a normative rule that PRO should be controlled by the matrix subject. Quirk et al. (1985: 1120) refer to this as the attachment rule. According to this rule, adjuncts are subject to obligatory subject control. However, the mere need for a normative rule shows that this pattern is not really obligatory. Taking the full range of data into account, it is quite clear that adjuncts fall outside the domain for obligatory control. 4.  Arbitrary control As mentioned above, arbitrary control may be characterized as non-control; there is no controller, and PRO receives an arbitrary or generic interpretation. This is the standard pattern in non-finite subject clauses (14a), including extraposed subjects (14b), as well as in so-called tough-constructions (14c). It is also very common in independent non-finite clauses, typically used as headings (14d): (14)

a. b. c. d.

PROarb Getting a new passport takes about a week. It’s not easy to PROarb tell two laurel species apart. This puzzle isn’t that hard to PROarb solve. PROarb Winning at poker

All these sentences are referring to activities without indicating who is performing them, which yields a generic interpretation of PRO. The statements in (14a–c) are presented



Control phenomena

as valid for any arbitrary individual, whereas headings like (14d) simply indicate what kind of activity is being treated in the text in question. In none of these cases is there a controller of PRO, not even implicitly. Accordingly, structures of arbitrary control are best treated as generic expressions, where PRO is assigned direct arbitrary reference. However, just like with other generic structures, there is a bias towards salient interpretations. There may often be a specific reading provided by the context, typically associated with the speaker. Both (14b) and (14c), for instance, report subjective judgements concerning the speaker and are not necessarily true for anyone else. There are also cases that refer more explicitly to specific situations. Consider, for example, the following sentences: (15) a. In this weather, it’s really hard PROarb to get a fire going. b. It was no use PROarb complaining. c. Do you know where to PROarb get a decent lunch around here?

A sentence like (15a) typically refers to a specific event, where a referent corresponding to PRO is provided by the context – and (15b) definitely does. Still, both are instances of arbitrary control. This is because they are presented as general statements. Although the intended subjects of get a fire going and complaining, respectively, may be quite obvious for each utterance, the use of these constructions (instead of, e.g., I couldn’t get a fire going) implies – correctly or not – that the same facts would hold for anyone in the same situation. Likewise, although a question like (15c) is typically asked with the speaker in mind, it is expressed as pertaining to anyone. This is also why arbitrary control should be distinguished from pragmatic control (as described in the previous section), where there is no genericity involved. In both cases, the interpretation of PRO is determined by the context, but in the case of arbitrary control that is just a secondary effect. PRO is assigned arbitrary reference, and the contextual bias involved has more in common with other generic expressions than with other control structures. I emphasize this because from a viewpoint focusing on the distinction between obligatory and non-obligatory control, arbitrary control is easily seen as just a special case of non-obligatory control. When put more into focus, however, it is shown to be more distinct than that. Whereas pragmatic control is a form of control, arbitrary control is essentially non-control. Having established this, it is important to note that typical arbitrary control structures can still be assigned specific reference. In English, this is done by including a for-phrase, as in (16a). The interpretation may also be specified by a reflexive, as in (16b): (16) a. That’s a difficult thing for me/a kid/anyonei to PROi do. b. PRO Shaving myself/himself/oneself was not an option.

PRO in (16a) is obligatorily coreferential with the complement of for. This element may be a specific entity (e.g., me), a class (e.g., kid), or an open variable (e.g., anyone). In

41

42

Benjamin Lyngfelt

the latter case, the for phrase may seem redundant, since the resulting reading is one of arbitrary reference; but it adds emphasis by making this explicit. When the non-finite clause contains a reflexive pronoun, as in (16b), the interpretation of this pronoun also determines the interpretation of PRO; and the arbitrary reading is maintained only in combination with a generic pronoun like oneself. As mentioned above, arbitrary control typically appears in non-finite subject clauses, tough-constructions, and independent non-finite structures. Of these, toughconstructions may require some explanation. These involve gapped infinitives embedded in an adjective phrase, as in (17a), or a noun phrase, as in (17b). In the latter case, the infinitive is still associated with the adjective: (17) a. This nuti is hard to PROarb crack [ ]i. b. This is a hard nuti to PROarb crack [ ]i. c. It is hard to PROarb crack this nut. d. To PROarb crack this nut is hard.

A distinguishing feature of these constructions is the gap in the infinitive (the object of crack), which corresponds to the noun nut. It is also striking how they resemble infinitival subjects, as in (17c–d); and they are indeed semantically equivalent to these. In transformational frameworks, tough-constructions are assumed to be derived from subject clauses, and it seems quite clear that there is some kind of structural relation. In any case, the similarity explains why PRO in tough-constructions is interpreted the same way as in non-finite subject clauses, i.e., by arbitrary control. Other structures similar to tough-constructions are interpreted in the same way. For example, Swedish employs what is probably best described as a tough-verb, namely the ‘be possible’ sense of the verb gå, as in (18), where the interpretation of PRO is, as expected, arbitrary control: (18) Bilen går inte att PROarb laga. car-def goes not to PROarb fix ‘The car isn’t possible to fix/It isn’t possible to fix the car.’

Finally, there are also some cases of arbitrary control in verb complements, which otherwise usually receive obligatory control (cf. Section 2). Such cases include prepositional complements of certain verbs, as in (19a), and English wh-infinitivals, as in (19b): (19) a. He has lectured on PROarb translating poetry at the university for several years. b. Who discovered how to PROarb milk a cow?

If tough-constructions are similar to infinitival subjects, then these constructions rather resemble headings (cf. example (14d) above), especially the type in (19a), where Translating poetry might well be the title of a book or a lecture (wh-infinitivals sometimes



Control phenomena

appear as independent utterances as well). Arbitrary control in verb complements is quite restricted and seems to be licensed by certain specific verbs. The distribution of arbitrary control structures, which typically appear in argument positions, may be contrasted with that of pragmatic control, which only occurs in adjuncts.

5.  Less discussed control patterns Although there is a huge literature on control, it has mainly focused on complement control (Section 2) and to some extent adjunct control (Section 3). There are, however, several other types of control structures. These types have been extensively discussed from a structural perspective, especially various kinds of infinitives, but the control literature rarely ventures beyond objects and adverbial adjuncts. This section will account for control patterns in noun phrases and adjective phrases (5.1), indirect control (5.2), and some other control relations (5.3). Some of the generalizations are originally based on Swedish control patterns (cf. Lyngfelt 2002), although most examples will be in English. On the few occasions where the two languages are known to differ, this will be noted. 5.1  Control in noun phrases and adjective phrases In adnominal non-finite clauses, i.e., non-finite clauses modifying nouns, most cases adhere to the same interpretive pattern: possessive control. This means that PRO is controlled by whoever is in possession, in a broad sense, of the matrix noun, as illustrated in (20): (20) a. They assigned mei another essay to PROi write. b. Shei was just getting into the habit of PROi reading. c. There may be reasons to reconsider youri promise to PROi retire.

Although the controllers are realized differently in all three sentences, as an object in (20a), a subject in (20b), and a genitive modifier in (20c), they all share the same possessive relation between PRO and the noun phrase containing it. In (20a), the controller is the person being assigned the essay. In (20b), the one who has got into a habit can be said to have acquired it. And in (20c), although the controller is the person who gave rather than received the promise, it is still his or her promise, as indicated by the genitive. As shown by these examples, the notion of possession involved includes more than just ownership; in many cases the relevant property is rather access to the element in question. As of yet, no precise semantic characterization of this possessive relation

43

44

Benjamin Lyngfelt

has been proposed. Roughly, the possessor corresponds to a genitive modifier – and may well be realized as such, as in (20c). The examples in (20) represent three different types of adnominals: an infinitival relative clause (20a), an identifying apposition (20b), and a complement (20c). Starting with the infinitival relatives (or relative infinitives), these normally contain a gap, which is coreferential with the matrix noun (as indicated by the j index in (21a) and (21c)). In addition to this general pattern, English also employs infinitival relatives without a gap, as in (21b). The controller in these structures is not the possessor of the matrix noun, but the noun itself. In Swedish, such structures are ungrammatical, and a sentence like (21b) would have to be construed with a finite relative clause.5 Even in English, the pattern of possessive control is obligatory if the infinitive includes a gap. The matrix noun correponds to the gap and cannot at the same time be the controller of PRO, since PRO and the gap are referentially distinct arguments of the same predicate: (21) a. They assigned mei another essayj to PROi write [ ]j. b. You need someonei to PROi help you. c. Youi need someonej to PROi help [ ]j.

Identifying appositions (20b) are distinguished by the head and the modifier being coreferential. In (20b), the habit is reading. For this type, there is no alternative to the standard pattern of possessive control. As for the complement type (20c), it usually occurs with deverbal nouns. The control pattern is parallel to that of the corresponding verbs, where the possessor of the noun corresponds to the subject of the verb, as shown in (22). Nevertheless, these complements typically display possessive control: (22) a. Hei promised to PROi resign/hisi promise to PROi resign b. Shei intends to PROi win/heri intention of PROi winning

Turning to control in adjective phrases, there are two control patterns, corresponding to two different structure types. These two types are often discussed with reference to the contrasting examples easy to please and eager to please. As mentioned in Section 4, tough-constructions receive arbitrary control, as in (23a). In other adjective phrases, PRO is controlled by the base of predication of the adjective; whatever the adjective is predicated of is also the controller of PRO. This is illustrated in (23b–c): (23) a. Johni is easy to PROarb please [ ]i. b. Johni is eager to PROi please. c. That almost made mei ready to PROi give up.

5.  There are some exceptional cases of infinitival relative clauses without gaps in Swedish as well; and this usage, although still marginal, seems to be spreading. Most likely, this is due to influence from English.



Control phenomena

As in infinitival relatives, the gap seems to be the crucial difference here. In toughconstructions, the base of predication of the adjective corresponds to the gap; if there is no gap, the base of predication is the controller. The patterns of possessive control in noun phrases and control by the base of predication in adjective phrases are structurally similar, in that the controller equals the external argument of the matrix head (a property they also share with subject control of verb complements and adjuncts). As a local control relation, it is quite consistent throughout both structure types. However, this local relation is often implicit, since arguments of nouns and adjectives are usually unexpressed. The actual interpretation of the possessor/base of predication, and consequently of PRO, depends on external factors that vary according to the context. The most common outcome is coreference with the matrix subject. 5.2  Indirect control As mentioned in Section 2.2 above, predicative complements (or subject complements) behave differently from ordinary verb complements with respect to control. Instead of being directly related to an element in the matrix clause, PRO in predicative complements is controlled by a corresponding element within the subject of the matrix clause. This two-step relation (the controller is an argument of an argument), which is illustrated in (24), may be called indirect control: (24) a. PROarb,i Seeing is PROi believing. b. Myi original intention was to PROi write a mystery novel.

If the subject itself is a non-finite clause, as in (24a), PRO in the predicative is coreferent with PRO in the subject. If the subject is a noun phrase, PRO is controlled by the possessor of the noun (cf. Section 5.1), which in (24b) is realized as a genitive modifier. What these two controllers have in common is that they are the external arguments of their respective heads; subjects (including PRO) are the external arguments of verbs, and possessors/genitives are the external arguments of nouns. Thus, the external argument of the predicative (i.e., PRO) is controlled by the external argument of the subject. This pattern of indirect control follows from the nature of predicative complements. Since copular verbs basically equate the predicative with the subject, it follows that predicative non-finite clauses co-occur with subjects with some sort of sentential content. Accordingly, both “events” have corresponding “subject” arguments. Even when the matrix subject is a noun phrase, it carries some kind of eventive meaning and is often headed by a deverbal noun. Notice, for instance, how the noun intention in (24b) corresponds to the verb intend, and how the genitive of the noun equals the subject of the verb (my intention/I intended).

45

46

Benjamin Lyngfelt

As shown in Section 4, PRO in subjects is interpreted by arbitrary control. Hence, in sentences where both the subject and the predicative complement are non-finite clauses, as in (24a), the arbitrary interpretation extends to the predicative as well. However, the predicative does not receive an arbitrary reading by itself; PRO in the predicative is still obligatorily controlled by whatever is the external argument of the subject. Another kind of indirect control seems to occur when a non-finite clause is embedded in another non-finite clause: (25) Shei asked mej to PROj take care of the applications before PROj attending to anything else.

In (25), PRO in the embedded adverbial (before attending) is controlled by its matrix subject, which itself happens to be PRO (the understood subject of take care). This, in turn, is controlled by its matrix object (since ask is an object-oriented verb), and thus the deepest embedded PRO is indirectly controlled by the main-clause object. I would, however, rather treat this as two separate control relations. Unlike predicative complements, where indirect control is a core property, cases like (25) are simply a consequence of the recursiveness of language. 5.3  Some other control relations Most syntactic control relations hold within the same finite clause. However, when the subject of a subordinate clause is itself a non-finite clause, PRO may in some circumstances be controlled by an element outside the immediately superordinate clause, as illustrated in (26). Since Grinder (1970), this phenomenon is called super-equi:6 (26) a. Johni knew that PROi making a fool of himself in public disturbed Sue. b. *Johni knew that it disturbed Sue to PROi make a fool of himself in public. c. Johni knew that it would help Sue to PROi behave himself in public.

As illustrated in (26), the availability of super-equi depends both on position and on the predicate of the intermediate clause. For some predicates (like disturb in (26b)), super-equi is not available if the non-finite subject is extraposed, while for other predicates, like help in (26c), it works just fine. It may be worth noting that the structure in (26a) is ungrammatical in some other Germanic languages (e.g., Swedish, Icelandic, and German), where non-finite clauses are disallowed in the ordinary subject position

6.  In early transformational grammar, control was treated as equi-NP deletion, on the assumption that the subject of a non-finite clause was deleted due to coreference with another element. Control across a finite clause boundary was consequently called super-equi. Although the analysis as such is no longer maintained, the name super-equi has stuck.



Control phenomena

of subordinate clauses and hence are obligatorily extraposed (in Scandinavian, this restriction concerns finite clauses as well). For some reason, the discussion about super-equi is usually focused on sentences where a reflexive pronoun specifies the intended meaning, and reflexives are known to specify or sometimes even overrule otherwise expected control patterns (cf. example (16b) above). Without a reflexive, as in (27a), the interpretation is less fixed: (27) a. John knew that it would help Sue to PRO make a good impression. b. He learned that PROarb throwing a party didn’t have to be such a big deal.

In (27a), PRO could be coreferential with either John or Sue, or maybe even someone else. Since the normal pattern for non-finite subjects in subordinate clauses is arbitrary control, as in (27b), just like in other subjects (cf. Section 4), super-equi might simply be a special case of arbitrary reference being specified by the context. On the other hand, the wealth of data and literature on super-equi suggests that it is more than that. Especially the fact that there are syntactic constraints involved indicates that this is not merely a pragmatic issue. In any case, although super-equi has been extensively discussed for decades, we still have not reached a proper understanding of it. Finally, let us consider a couple of sentences where PRO and the controller do not match entirely. Usually, there is a one-to-one relation between PRO and its controller, but in some cases the controller only covers part of the reference of PRO, as in (28a); and in some cases PRO is jointly controlled by the subject and the object, as in (28b). In Landau (2000), these relations are called partial control and split control, respectively: (28) a. The chairi preferred to PROi+ gather at 6. (Landau 2000: 5) b. Johni asked Maryj whether to PROi+j get themselves a new car. (Landau 2000: 53)

Some scholars also recognize implicit control as a category of its own, pertaining to cases when the controller is not expressed in the syntax, for instance control by an implicit agent in a passive clause (cf. Section 3). I have not done so here, since some control relations hold regardless of whether the controller is explicit or not, e.g., possessive control in adnominals (cf. Section 5.1). It may also be the case that the local control relation is implicit, although a coreferential element is expressed elsewhere in the syntax; again, a typical example is (local) possessive control in adnominals.

6.  Outlook Although there is a huge literature on control, most of it has centered around a fairly small subset of control structures. The main issues in theoretical syntax seem to be (a) the distinction between obligatory and non-obligatory control and (b) how to

47

48

Benjamin Lyngfelt

account for the obligatory control patterns. In practice, this usually leads to a focus on control in verb complements, notably infinitival objects. There are admittedly good reasons for this. For syntactic theory, it is a highly relevant question what is determined by the syntax and what is not. And in attempting to explain that which falls within the domain of syntax, it is a high priority to account for core phenomena like the relation between verbs, subjects, and objects. It is also a matter of fact that this task is far from finished. For instance, control shift remains a particularly elusive phenomenon. Nevertheless, it seems that the ripest field for future research on control would concern other control structures than the standard cases. The control patterns in noun phrases, adjective phrases, and predicative complements should fall within the realm of even the most narrow syntactic theory, but have been largely disregarded in the literature so far. As for more specific topics, wh-infinivals are among the phenomena that warrant more attention. From a more semantic (or pragmatic) viewpoint, notions like arbitrary control, pragmatic control, and possessive control are still poorly understood. Arbitrary control is worth exploring both in relation to genericity in general and in relation to pragmatic control. For instance, just what is the difference between pragmatic control and the contextual bias involved in arbitrary interpretation? Also, to what extent is genericity involved when typical arbitrary control structures are assigned specific reference elsewhere in the syntax? As for pragmatic control, we know fairly well under which circumstances it may occur, but far too little of the pragmatic interpretation as such. Turning to possessive control, the semantics of this notion has yet to be made precise, a venture that would coincide with seeking a better understanding of the genitive. Cross-linguistically, the control literature is dominated by a focus on English, but includes quite a lot of data on other languages as well. Some interesting observations include control into finite complements in Portuguese, control of objects in Chinese, and backwards control – where the explicit element resides in the subordinate clause and the implicit element in the matrix clause – in Romanian and Malagasy. Not all of these cases fit standard definitions of control. As a general remark on the cross-linguistic data, it is striking that, although the distribution of control structures varies greatly across languages, the interpretation of the control relations is remarkably similar. Whether this hints at universal interpretation strategies or merely depends on what data are available remains to be seen. Acquisitional studies of control are few but not nonexistent, cf. for example Goodluck & Behne (1992). As for the diachronic aspects, it is hard to find studies of the development of control per se. Rather, control data may be extracted from diachronic studies of various kinds of control structures, such as infinitival clauses, gerunds, etc. All in all, there is a great deal left to be learned about control.



Control phenomena

References Barrie, M. (2007). Control and wh-infinitivals. In: Davies & Dubinsky (eds.): 263–279. Bresnan, J. (1982). Control and complementation. Linguistic Inquiry 13: 343–434. Also published in: J. Bresnan (ed.), The mental representation of grammatical relations. MIT Press. Culicover, P. & R. Jackendoff (2005). Simpler Syntax. Oxford University Press. Davies, W.D. & S. Dubinsky (2004). The grammar of raising and control. A course in syntactic argumentation. (Studies in Natural Language and Linguistic Theory 71.) Blackwell. Davies, W.D. & S. Dubinsky (eds.) (2007). New horizons in the analysis of control and raising. Springer. Dik, S.C. (1997). The theory of Functional Grammar. Part 1: The structure of the clause. Mouton de Gruyter. Farkas, D. (1988). On obligatory control. Linguistics and Philosophy 8: 27–58. Goodluck, H. & D. Behne (1992). Development in control and extraction. In: J. Weissenborn, H. Goodluck & T. Roeper (eds.), Theoretical issues in language acquisition: 151–171. Lawrence Erlbaum. Grinder, J.T. (1970). Super equi-NP deletion. Papers from the Regional Meeting of the Chicago Linguistic Society 6: 297–317. Hornstein, N. (1999). Movement and control. Linguistic Inquiry 30: 69–96. Jackendoff, R. (1972). Semantic interpretation in generative grammar. MIT Press. Kawasaki, N. (1993). Control and arbitrary interpretation in English. Diss. University of Massachusetts, Amherst. Dissertation Abstracts International 54/2. Landau, I. (2000). Elements of control: Structure and meaning in infinitival constructions. (Studies in Natural Language & Linguistic Theory 51.) Kluwer. Lyngfelt, B. (2000). OT semantics and control. Rutgers Optimality Archive 411. Lyngfelt, B. (2002). Kontroll i svenskan. Den optimala tolkningen av infinitivers tankesubjekt. [‘Control in Swedish: The optimal interpretation of implicit subjects in infinitival clauses.’] Acta Universitatis Gothoburgensis. Manzini, M.R. (1983). On control and control theory. Linguistic Inquiry 14: 421–446. Quirk, R., S. Greenbaum, G. Leech & J. Svartvik (1985). A comprehensive grammar of the English language. Longman. Rosenbaum, P. (1967). The grammar of English predicate complement constructions. MIT Press. Sag, I.A. & C. Pollard (1991). An integrated theory of complement control. Language 67: 63–113. Williams, E. (1980). Predication. Linguistic Inquiry 11: 203–238. Williams, E. (1992). Adjunct control. In: R. Larson et al. (eds.), Control and grammar: 297–322. (Studies in Linguistics and Philosophy 48.) Kluwer.

49

Definiteness Ritva Laury University of Helsinki

There is a vast literature on definiteness, and it is possible to discuss only some of the research in an article of this scope. For book length discussions on the topic, see, for example, Hawkins (1978), Chesterman (1991), and Lyons (1999). The discussion here will proceed in roughly chronological order from early treatments of definiteness by philosophers to descriptions of the use of the definite articles and other markers of definiteness by grammarians. I will then discuss the choice between different types of definite expressions and the research on definiteness in grammar and discourse, the morphosyntactic expression of definiteness in the languages of the world, and finally, the development of definiteness in language acquisition and grammaticalization. 1.  Definite descriptions and reference Definiteness is a topic where linguistics and philosophy meet. Philosophers have mainly been interested in investigating what the use of definite descriptions entails or implies; to answer this question, they have focused their attention, for the most part, on singular, definite noun phrases in declarative sentences. An early observation made by Frege (1892: 214) was that the use of referring expressions in general assumed the existence of a referent. A slightly different point was made by Russell (1905) in his highly influential treatment of definiteness. Russell claimed that what he called ‘definite descriptions’, for example, the king of France, implied both the existence of such an entity (a quality shared by indefinite descriptions) and, furthermore, its uniqueness (this being what distinguished definite descriptions from indefinite descriptions, such as a man). Russell did acknowledge that the use of the does not always involve uniqueness, so that we do “speak of the son of So-and-so even when So-and-so has several sons” (1905:  481), but he suggests that “it would be more correct to say ‘a son of So-and-so’ ”. Accordingly, Russell suggested that we cannot in fact speak of the present king of France, since no such entity exists, and we also cannot speak of the inhabitant of London, since there is more than one inhabitant of London (1919: 176). More controversially, Russell claimed that since definite descriptions such as the man can be true of any individual matching the description, they do not refer directly; that is, unlike



Definiteness

proper names, they do not directly pick out particular referents (1919: 173). This last claim was contested by Strawson (1950), who argued that definite (as well as indefinite) descriptions are in fact used to refer, and also that, in contrast to what Russell had claimed, their use does not imply existence, but only presupposes it. Strawson makes a number of interesting points. Russell had argued that denoting phrases by themselves lack meaning, although “every proposition in whose verbal expression they occur” does have meaning (1905: 480), but Strawson takes this argument further and notes that mentioning or referring is not something that expressions do, but rather something that expressions can be used to do — “it is people who mean, not expressions” (1950: 328). By thus separating the potential of definite expressions such as the king of France to refer to an actual person and the use by speakers of such an expression to do so (a distinction similar to Guillaume’s (1919) le nom en effet and le nom en puissance; see also Hjelmslev 1928). Strawson is able to point out that whether the expression refers or not, and what it refers to, is a function of the context in which it is used. Contextual factors which affect reference, according to Strawson, are, “at least, the time, the place, the situation, the identity of the speaker, the subjects which form the immediate focus of interest, and the personal histories of both the speaker and those he is addressing” (ibid.: 336). Thus, for example, a definite expression such as the king of France could be used to refer to a number of different individuals depending on the context in which the reference is made. Strawson also stresses the importance of convention in definite reference; he suggests that “one of the conventional functions of the definite article is to act as a signal that unique reference is being made” (ibid.: 331), i.e., the definite article signals, but does not assert uniqueness, as had been claimed by Russell. Strawson also discussed the distinction between the referring use and the predicative or attributive use of words and expressions, which, he argued, is “primarily one between different rôles or parts that expressions may play in language, and not primarily one between different groups of expressions” (ibid.: 337). However, Strawson does not make it entirely clear how one could tell the difference between the two uses, although he implies that the distinction has to do with the kind of sentence the expression appears in, as well as the type of definite expression used; he suggests that when definite expressions appear as subjects of ‘singular subject-predicate sentences’ (ibid.:  320), they exemplify the referring use, and also that pronouns are mainly used referentially (ibid.: 337). However, Donnellan (1966 [1971]) has argued that definite noun phrases functioning as subjects can be used either referentially or attributively. According to Donnellan, whether a definite noun phrase is used referentially or attributively is “a function of the speaker’s intentions in a particular case” (Donnellan 1966: 110). One of Donnellan’s examples is the sentence The murderer of Smith is insane. If the speaker intends the noun phrase to refer to a particular person, for example to the person being accused of Smith’s murder, Donnellan considers the use of the noun phrase a

51

52

Ritva Laury

referential use. This type of use would be referential regardless of whether the speaker actually believed that the person accused was guilty or whether he was in fact guilty; what counts is the speaker’s intention to refer to a particular person, whether the description used fits or not. On the other hand, in the attributive use, the description is what counts; in this case, if Smith actually committed suicide, the speaker would actually not succeed in referring to anyone. Like Donnellan and Strawson, Searle (1969: 82) also argues that definite descriptions can be used to refer, and states explicit conditions for ‘successful’ definite reference, uniqueness and identifiability. Interestingly, Searle considers identifiability a function of the type of expression used in terms of the ability of the hearer to be able to identify the referent on the basis of the expression uttered by the speaker; this issue has recently been taken up by Chafe, who discusses the issue in terms of what he calls “sufficiently identifying language” (1994: 97–100). However, Searle’s account of definiteness, like the accounts of other philosophers before him, fails to provide a complete account of definiteness because, like other philosophers before him, he restricts his analysis to singular definite descriptions. It is, of course, not the aim of philosophers to provide a complete account of the use of the definite article in English or other languages, let alone a complete account of definiteness in language. While it is not always clear whether and to what degree the issues discussed by the philosophers (such as Donnellan’s referential/attributive distinction) are linguistically relevant, they do, however, form a baseline from which many linguistic studies of definiteness begin. However, it turns out, as Strawson noted, that “problems of use are wider than problems of analysis and meaning” (1950: 337). 2.  Definiteness and identifiability The most basic function of definite articles and other markers of definiteness in languages of the world is to signal a speaker’s expectation that the addressee can identify what is being referred to. The insight that definiteness has to do with assumed ‘identifiability’, ‘familiarity’ or ‘locatability’ of the referent informs most modern work on definiteness (for example, Chafe 1976; Hawkins 1978; Du Bois 1980; Clark & Marshall 1981; Heim 1983) and appears to be equally true of the use of all definite expressions such as demonstratives, personal pronouns, and noun phrases explicitly marked for definiteness with devices such as the definite article. Identifiability, in fact, has become so strongly associated with definiteness that some scholars (Du Bois 1980; Prince 1992; Chafe 1994) have made the decision to consider definiteness a formal category (so that noun phrases with the definite article, pronouns, and proper names are considered definite by virtue of their form) and consider identifiability the pragmatic and semantic feature which normally (though not without exception) characterizes the referents of definite forms.



Definiteness

There is also a fair degree of agreement on the main sources of definiteness. A speaker may expect his or her addressee to identify the referent(s) of a definite expression either on the basis of linguistically encoded information which either precedes or follows in the discourse at hand (endophoric reference in terms of Halliday & Hasan 1976), or on the basis of extra-linguistic, situational information (exophoric reference). Endophorically, the identifiability of a referent may be established through previous mention in the discourse. Subsequent mentions of the same referent in the same text or conversation, even if that referent was initially introduced with an indefinite mention, will ordinarily be definite. This is a very common, though not the most common (Fraurud 1990) source of definiteness, and, accordingly, examples of this phenomenon are easy to find in both spoken and written texts; a classic example is the following, attributed to Jespersen by Christophersen (1939: 29): Once upon a time there lived an old tailor in a village. The tailor was known all over the village as ‘Old Harry’. In this example, both the tailor and the village are introduced with a formally indefinite mention, and subsequently referred to with definite mentions. Obviously, anaphoric (backward referring) use of definite expressions is not restricted to the use of the same form; for example, subsequent mention of the tailor could be made by using the pronoun he, a noun preceded by a demonstrative such as this tailor, or a different noun, such as the man. It has also been shown that although the nature of a particular referent may change considerably, the pressure toward referential continuity is so strong that speakers in the same conversation or text may continue referring to it as if it were the ‘same’ thing (Charolles 1998). For example, Charolles shows that in a cooking recipe, the referent is originally referred to in the plural (les têtes de 24 écrevisses ‘the heads of 24 crayfish’), and although the recipe calls for grinding them in a mortar until they are reduced to a paste, and thus the heads of the crayfish are no longer separable and hence not plural in any concrete sense, they (!) continue being referred to with les ‘they’, a plural pronoun, in the rest of the recipe. Another type of much studied endophoric reference involves a definite mention which relies for its identifiability on an earlier mention of another, related referent. For example, an earlier mention of a car makes it possible to refer subsequently to the steering wheel, the horn, the battery, and other car-related items; presumably, the mention of a car would trigger certain expectations regarding the components of a car. In other words, it is possible to talk about the steering wheel after mentioning a car because speakers can expect their addressees to share information which would lead him or her to expect a car to have a steering wheel, even if there has not been any previous mention of a steering wheel. This type of endophora has been called ‘bridging inference’ by Clark & Haviland (1977), ‘associative anaphora’ by Hawkins (1978) ‘inferrability’ by Prince (1981), ‘accommodation’ by Heim (1983), ‘indirect anaphora’ by Erkü and Gundel (1987) and ‘indirect sharing’ by Chafe (1994; 1996). Exactly what type of inferencing processes are involved, and how previously mentioned, related

53

54

Ritva Laury

information licenses a subsequent definite mention has been much debated. Clark & Haviland (1977) provide experimental evidence indicating that definite mentions were processed at a slower speed when they were not preceded by an explicit antecedent, and had to be inferred from earlier mentions of related referents. Their subjects read pairs of sentences such as the following:

(1) (2)

Esther got some beer out of the car. The beer was warm. Esther got some picnic supplies out of the car. The beer was warm.

Clark and Haviland found that the subjects needed more time to compute the reference of the beer in the second sentence pair, where the definiteness of the beer is inferrable from the mention of the picnic supplies, while the first sentence pair, which contained an explicit antecedent, was processed more quickly. Chafe (1996) argues, based on prosodic evidence, that even among definites whose identifiability is based on inference, the directness of the inference varies. Gundel (1996) discusses the ways in which speakers choose among and interpret different types of definite expressions which lack explicit antecedents, and Hellman (1996) compares cognitive theories designed to account for this type of inferencing. Epstein (1997) notes that while it is clear that stereotypical notions shared within a speech community are responsible for much use of inferrables (on this, see also Prince 1981: 242; Kleiber 1994), definite mentions are also used by speakers without a previous mention of another, related referent which would have triggered a frame within which the definite mention would be interpretable. In other words, sometimes we can talk about the steering wheel without mentioning a car first. An example Epstein gives has to do with the departure from the city of Los Angeles of the two professional football teams that used to play there, the Rams and the Raiders. (3)

So we lost the Rams and the Raiders. Lost our innocence. But hold the flowers. Put away the handkerchiefs. Stop the sobbing. We still have the Rose Bowl, don’t we?

Epstein points out that the definite mentions of the flowers, the handkerchiefs, and the sobbing are all stereotypically related to the ‘funeral’ frame, which nevertheless is not explicitly mentioned prior to the use of the definites, showing that “the frame itself need not be directly evoked at all (…) but can be activated simply by mentioning some of its salient elements” (Epstein 1997: 5). In fact, the definiteness of the nouns is crucial in creating the implication the writer seeks to convey. It indicates that they all belong to the frame which they function to evoke. Epstein also shows that speakers employ definite noun phrases on an ad hoc basis for first mentions in such a way that they create an entirely novel association by forcing an association with a frame with which they are not stereotypically linked with. In these uses, speakers can be observed using definite expressions for referents which are not actually identifiable at all to the addressee.



Definiteness

The association between definiteness and identifiability, familiarity, and uniqueness has also been critically examined by Birner & Ward (1994), Du Bois (1980) and Epstein (1994). Birner & Ward (1994) point out that unique identifiability within the discourse context is a sufficient but not necessary condition for the use of noun phrases marked with the definite article, so that if the referent of a noun phrase is undifferentiated and not relevantly differentiable, it can be referred to with a definite noun phrase even though it is not uniquely identifiable. A very similar point was made by Du Bois (1980), who noted that in

(4)

The boy scribbled on the living-room wall

the noun phrase the living-room wall can be used even if the addressee is not expected to be able to identify precisely which wall is meant. Du Bois suggests that such uses are explained by what he calls ‘the curiosity principle’: “A reference is only counted as identifiable if it identifies an object closely enough to satisfy the curiosity of the addressee” (Du Bois 1980: 233); in other words, a definite noun phrase can be used because the exact identity of the wall is not relevant (see also Kadmon 1990). Du Bois also notes that “speakers have facultative control over definiteness” (ibid.: 219) so that they are, for example, free to use formally indefinite forms for referents which are perfectly identifiable to their addressee. And Epstein (1994) shows through numerous examples that in addition to serving as a marker of identifiability, the definite article also has what he calls an expressive function; definite noun phrases are used by speakers for non-identifiable referents to express prominence, emphasis, and viewpoint. Prominence appears to be very important in accounting for the historical development of definite articles, as shown by Epstein (1994, 1995) for French, Faingold (1996) for Vulgar Latin, Spanish, Portuguese and Rumanian, and Laury (1995, 1997) for Finnish.

3.  Choice between types of definite expressions While much work concerned with definiteness has been implicitly or explicitly confined to explicating the use of the definite article, many scholars have also focused on the choice speakers make between different types of definite referential expressions, for example, the choice between a full noun phrase (the man) and a pronoun (he). Factors that are thought to affect such a choice are the cognitive accessibility of the referent, the range of referential forms available to speakers of a particular language, discourse structure, topicality of the referent, and interactional and social factors. Thus, for example, Chafe (1976, 1979, 1987, 1994) suggests that the form chosen by a speaker to refer to something he or she expects the addressee to be able to identify depends on the speaker’s beliefs regarding the current status of that referent in the consciousness of the addressee. Referents which the speaker believes to be in the addressee’s

55

56

Ritva Laury

current, focal consciousness are called given information by Chafe, while referents which the addressee is not currently conscious of at all (but may have a representation for in his or her long-term memory) are called new information. Referents which the addressee is peripherally conscious of, because they have been mentioned earlier, or are present in the environment, are called accessible information. According to Chafe, given information is ordinarily expressed with attenuated forms such as pronouns, while new and accessible information is expressed with full noun phrases. For Chafe, these three statuses represent the ‘activation cost’ (1994: 71–81) of a referent; new referents are the most costly to activate in terms of mental energy required to bring them to active consciousness, accessible referents are somewhat less costly, and given referents are the least costly of all. Chafe strictly separates the current cognitive status of a referent from its identifiability and points out that even new referents can be referred to with a definite noun phrase if the speaker expects that the addressee can identify the referent (although the referent is not activated in his/her consciousness prior to being mentioned). Prince’s (1981, 1992) approach is similar to Chafe’s except that, unlike Chafe, she does not separate identifiability from accessibility, but rather incorporates ability to identify the referent with its current cognitive accessibility. In Prince’s taxonomy, referents (which she calls ‘entities’) are separated into ‘new’ and ‘evoked’ depending on whether they have already been introduced into the discourse or not. Newly introduced referents are further divided into ‘unused’, referents for which the hearer already has a representation for, and ‘brand-new’ referents for which the hearer must “create a new entity” (Prince 1981: 235). Brand-new entities can be anchored or unanchored. Anchored entities are connected to another discourse entity which, Prince notes, is not itself brand-new. One example of an anchoring which Prince gives is a guy I work with, where a guy is linked, with a relative clause, to I, representing the anchor. Evoked entities, on the other hand, may be either textually or situationally evoked. Prince also acknowledges another category, ‘inferrable’ (see above). Within this category are items which have not actually been previously mentioned, but are related to another, already evoked ‘trigger’ entity. A special type of inferrables are ‘containing inferrables’ which contain the trigger entity in their description, for example, the pages of that book I bought, where the hearer needs no prior knowledge of the pages referred to; all that is required is that the hearer know about the book. Prince suggests (1981:  245) that there may be a preferred hierarchy or scale of types of expressions used by speakers depending on their hypotheses regarding what a particular hearer knows. Prince (1981) proposes that different types of expressions represent points on her Familiarity Scale, and gives examples. However, she does not make it entirely clear exactly what types of expressions are linked with what statuses. More explicit claims along these lines have been made by Ariel (1988, 1990, 1991) and Gundel, Hedberg & Zacharski (1993). Ariel suggests that in



Definiteness

choosing referring expressions, speakers must take into consideration the degree of mental accessibility of the referent for the addressee. She assumes that referring expressions are ranked on what she calls an Accessibility Marking Scale so that expressions at one end of the scale (zeros, reflexives and agreement markers) are used to refer to highly accessible referents, while expressions at the other end of the scale (long definite expressions, full names) are used to refer to less accessible referents. Accessibility of referents, on the other hand, is affected by four types of factors: saliency of the referent, recency of previous mention, the number of referents ‘competing’ for the role of antecedent, and the degree of cohesion between the units that the expression and its antecedent occur in. The higher the cohesion between the two, the more accessible the referent. Gundel, Hedberg & Zacharski’s Givenness Hierarchy (1993) is similar to Ariel’s theory in that types of referring expressions are also ranked with respect to the cognitive status of the referent. However, in the GHZ framework, each type of referring expression is paired with a particular status, so that, for example, a lexical noun with a definite article is associated with the status ‘uniquely identifiable’, while pronouns require that their referent be ‘activated’. Additionally, statuses higher on the scale entail all the lower statuses, so that a form or forms associated with a particular status can be used for a referent that has a status higher on the hierarchy. For example, the status ‘activated’, which is the highest on the scale, entails the status ‘uniquely identifiable’, and for this reason, a lexical noun with a definite article can also be used for a referent which has the status ‘activated’. Gundel et al. support their proposal through an empirical study of the distribution of referring expressions in five languages. Their finding is that while the six statuses correspond directly to particular types of English referring expressions, this is not true of any of the other languages they studied. Thus, for example, neither Japanese nor Russian have forms for which the status ‘uniquely identifiable’ is both necessary and sufficient, that is, they lack formal indicators of definiteness such as the English article (1993: 284–285). Clancy (1980) also makes the important point out that the choice among types of referential forms is strongly affected by the resources available in a particular language, as well as discourse structure. In her study of Japanese and English narratives, she found that Japanese speakers, whose referential system involves a choice between total ellipsis and explicit nominal reference (since the use of pronouns in Japanese is quite limited), took longer to establish characters as old (‘given’) information and were more likely than English speakers to allow minor discourse boundaries to trigger a return to full nominal reference. Fox’s (1987) important work discusses the influence of discourse structure and interactional factors on the choice between full noun phrases and pronouns in English. She bases her discussion on a corpus of conversations and written texts, and suggests that while the rules of anaphoric patterning are genre-specific, so that there is no single

57

58

Ritva Laury

rule of anaphora for all types of English, the general pattern that emerges from her research is that mentions in the same discourse sequence are coded with pronouns, while mentions that are not in the same sequence are coded with full noun phrases. Fox stresses that in different genres (such as conversation versus written expository prose), what counts as a sequence may be quite different. Additionally, she points out that although structural factors may be considered to determine anaphoric choice, on the other hand, use of a particular type of form by a speaker also accomplishes structure; for example, by choosing a full noun phrase instead of a pronoun, a speaker can indicate that a previous sequence has been closed and a new one is now starting. Thus speakers are able to manipulate the existing patterns of anaphoric choice to accomplish certain interactional tasks (such as disagreeing). Much recent work concerned with choice of referential form has indeed stressed interactional factors. This is especially so for articles contained in Fox (1996). Ford & Fox (1996) point out that a speaker’s choice between forms such as he and this guy does not depend solely on a speaker’s evaluation of the hearer’s current cognitive awareness of the referent, but also has to do with “continuing or shifting recipiency, such as attracting the attention of a non-engaged recipient, or acknowledging the attention of a participant” (Ford & Fox 1996:  165). Downing (1996) suggests that the choice between proper names and other referential options depends on a speaker’s stance relative to the topic of talk, in addition to discourse structure and topicality of the referent. The fact that proper names implicate that both the speaker and the addressee have some degree of knowledge of the referent (in Downing’s terminology, proper names are ‘co-recognitionals’) “makes them especially useful in contexts where the speaker wishes to display his or her authority to speak about a particular referent, as is common in disagreement contexts” (Downing 1996: 136). Another important strand of research on referential choice has considered the effect of the topicality of the referent. Especially influential has been the crosslinguistic research in Givón (1983). This work showed that the topicality of the referent, as measured in part by the recency of the previous mention in terms of clauses (RD, or referential distance) correlated with the choice of referential form in such a way that pronouns and ellipsis were used for highly topical referents, while full noun phrases were used for less topical referents. Givón (1983) was also important in terms of the use of rigorous, quantitative methodology in accounting for choice of referential form. For an innovative quantitative approach to the study of referential choice, see Kibrik (1996). Intersecting in an interesting way with the issues of topicality and discourse structure in referential choice is the concept of grounding. In general, backgrounded, or less topical, old information will be assumed to be already shared or presupposed, and hence definite. Its discourse function is to anchor the foregrounded, more topical, new information, which may be previously unknown and hence indefinite (Givón 1984:  256). These



Definiteness

facts have important consequences on sentence level; old information is more likely to be found in the subject position, while new information is expressed in the predicate. 4.  Definiteness and grammar The most general statement that can be made about definiteness and grammar is that by and large, but not without exception (see Mithun 1992), definite elements in sentences tend to occur before indefinite elements. Thus, for example, in Mandarin Chinese, where definiteness is not marked morphologically, preverbal elements tend strongly to be interpreted as definite, while postverbal elements are interpreted as indefinite or ambiguous with respect to definiteness (Li & Thompson 1981). It has also been observed that sentence constituents are differently treated in terms of marking for definiteness. In particular, objects appear to be more likely to be explicitly marked for definiteness than subjects. In many languages, case marking and definiteness marking interact in such a way that only definite objects are distinctively marked. It then becomes an interesting question whether the marker is in fact a case marker or a definiteness marker or whether it represents both categories. Many genetically and typologically unrelated languages exhibit this pattern, for example, Turkish and Chickasaw (Anderson 1985) and Persian (Ghomeshi 1997). There is also some evidence that when a definite article develops in a language, object noun phrases are more heavily marked with the emerging article than subjects (Laury 1997). Comrie (1978) suggests that definiteness marking in objects is due to referential prominence, rather than definiteness per se. In any case, it appears that definiteness in objects is a marked feature. This is likely to be due to the discourse functions typically performed by different constituents. It has been shown that objects, along with obliques, tend to be the sentence constituent where new information is introduced (Du Bois 1987). Although definiteness and identifiability, as noted above, are a separate category from the information statuses new and given, only new information can be unidentifiable. Thus, if objects are more likely than other constituents to express new information, they are also more likely to be indefinite, and definite (identifiable) objects are unusual and therefore likely to receive distinctive marking. Conversely, as already noted by Jespersen (1924:  150), subjects are more likely to be formally definite than other constituents. Since subjects in discourse tend to be thematic, or topical, they usually express anaphorically or situationally recoverable information, and are thus likely to be referred to with pronominal, definite NPs (subjects tend to be ‘light’ in terms of Chafe 1994). There is also a large literature on ‘existential’ constructions, such as the English ‘there is/are’ construction, suggesting that noun phrases which follow the presentational element are usually indefinite. This phenomenon is commonly called ‘the definiteness

59

60

Ritva Laury

effect’ after Milsark (1977). In Mandarin, according to Huang (1987), definite NPs such as proper nouns and nouns modified by demonstratives are ungrammatical in presentational constructions, and bare nouns in such constructions are interpreted as indefinite. However, in other languages this restriction appears not to be absolute, and there is much debate on the nature of the definiteness effect. It has been suggested that the constraint is pragmatic, not syntactic in nature (Hannay 1985; Abbott 1993). Prince (1992) and Ward & Birner (1995) propose that the postverbal NP in theresentences has to have a hearer-new referent (thus allowing either definite or indefinite NPs). However, Abbott (1997) presents contrary evidence, suggesting that postverbal NPs in there-sentences can also be hearer-old. 5.  Definiteness marking Conceptual definiteness, taken as ‘identifiability’, ‘uniqueness’ or ‘familiarity’ of a referent is clearly separate from overt marking of definiteness in the languages of the world. Even within the same language, noun phrases with equally definite (identifiable) referents are not consistently marked for definiteness in the same way. Consider, for example, the fact that while names are considered definite, in English, some are equipped with articles (the Caucasus), while others are not (Asia), with no difference in degree or type of definiteness. Likewise, as Croft (1990: 5) has pointed out, even closely related languages may differ in what is overtly marked for definiteness and what is not: in French, mass nouns and generics are marked with a definite article (les français aiment la gloire; J’aime les artichauts) while in English they are not (The French love glory, I love artichokes). Many languages have no markers of definiteness, and could thus be argued to lack this category in their grammar (Lyons 1999). On the other hand, as has been discussed above, word order can affect the interpretation of noun phrases as definite or indefinite in languages which lack morphological markers of definiteness. In such languages, it is also possible for morphemes whose main function is not the marking of definiteness to nevertheless have as a secondary function the marking of definiteness or indefiniteness in certain contexts. For example, in standard written Finnish, where there are no articles, partitive noun phrases in certain sentence types are interpreted as indefinite, while accusative and nominative noun phrases are interpreted as definite (Vilkuna 1992: 45–51; Chesterman 1991: 110–122). Those languages which do mark definiteness overtly exhibit a variety of morphosyntactic patterns. Definiteness markers can occur on noun phrases as bound morphemes, as in Jamul Diegueño (Epstein 1994), where the definite marker is a postposed clitic, as seen in uumall ‘book’ and uumall-pu ‘the book’; as free standing words (articles), as in English; or both, as in Scandinavian languages. In other languages, definiteness is expressed through an agreement morpheme on the verb. This is the



Definiteness

case, for example, with the Australian language Ngiyambaa, and certain Uralic languages (Lyons 1999: 86–87). As Anderson (1985) points out, even in closely related languages there are differences in how definiteness is marked. Swedish and Danish have both a preposed, freestanding definite article and a postposed bound definite marker. Both Danish and Swedish require the postposed marker in definite nouns which are not modified, as in hus ‘house’ and huset ‘the house’ (thus contrasting with English), and both (like English) also require the preposed marker in definite nouns which are modified. However, in Danish, definite nouns which are modified have no postposed definite (bound) marker, as shown by det lille hus ‘the little house’ while the Swedish det lilla huset has both the preposed article and the postposed definite marker. In some languages, articles are invariable and not inflected; this is true of both English, where there is very little inflectional morphology, as well as Hungarian, although the latter language is otherwise inflectionally complex. However, in certain other languages, articles have complex paradigms; in Classical Greek, articles are inflected for four cases and three numbers, while in Catalan and German they are inflected for case and gender. It is also possible for adjectives to have separate definite and indefinite forms. In some languages, for example Hebrew, adjectives agree in definiteness with the noun which is separately marked as definite, but in other languages the adjective may be the only marker of definiteness in the noun phrase.

6.  Development of definiteness In first language acquisition, most studies have indicated that adult-like expression of definiteness is slow to develop (for a useful review of this research, see Dasinger 1995). The early use of the articles by children under 6 or 7 years of age appears to be predominantly deictic in nature, indicating that young children depend heavily on extralinguistic context for referential strategy (Karmiloff-Smith 1979). In particular, young children have trouble with definiteness in contexts where the interlocutor’s state of knowledge differs from the child’s, so that they are likely, for example, to use definite forms for referents which are not identifiable to their interlocutor (Brown 1973). The use of definiteness for the expression of intertextual relationships appears to be challenging for children as well. In experimental studies where young children have been asked to retell narratives to both knowing and naive listeners, they have been shown to use indefinite forms for referents already introduced into the narrative, and are also less likely than older children and adults to make initial mentions with indefinite forms (Hickmann 1982). The marking of indefiniteness is in general slower to develop than the marking of definiteness, both in languages which have explicit markers of definiteness and ones that do not (Dasinger 1995). Definiteness marking is

61

62

Ritva Laury

late-acquired by second-language learners as well. Even advanced learners whose first language does not contain articles make many errors in their use, the most common type being the omission of both the definite and indefinite article in obligatory contexts (Master 1995).

7.  Conclusion The study of definiteness has continued to expand in scope and depth since the early treatments by philosophers. While the early work concerned singular, definite noun phrases mainly in English, we now know much more about the formal marking of definiteness, and the use of various types of definite noun phrases, in many different languages, as well as the pragmatic and cognitive correlates of definiteness such as identifiability, and the interaction of grammar and definiteness. However, no unified theory of definiteness which would cover all uses of definite noun phrases exists. A fruitful direction for further research appears to be work on manifestations of definiteness in a wide range of languages. It also seems quite clear from presentday studies that interesting, profound and novel findings on this aspect of language will emerge from an examination of the use of noun phrases by real speakers in real contexts of use.

References Abbott, B. (1993). A pragmatic account of the definiteness effect in existential sentences. Journal of Pragmatics 19: 39–55. ——— (1997). Definiteness and existentials. Language 73: 103–108. Anderson, S. (1985). Inflectional morphology. In T. Shopen (ed.), Language typology and syntactic description, v. III: 150–201. Cambridge University Press. Ariel, M. (1988). Referring and accessibility. Journal of Linguistics 24: 65–87. ———(1990). Accessing Noun-Phrase Antecedents. Routledge. ——— (1991). The Function of Accessibility in a Theory of Grammar. Journal of Pragmatics 16: 443–63. Birner, B. & G. Ward (1994). Uniqueness, familiarity and the definite article in English. BLS 20: 93–102. Brown, R. (1973). A first language: The early stages. Harvard University Press. Chafe, W. (1976). Givenness, Contrastiveness, Definiteness, Subjects, Topics, and Point of View. In C.N. Li (ed.), Subject and Topic: 25–55. Academic Press. ——— (1979). The Flow of Thought and the Flow of Language. In T. Givón (ed.), Discourse and Syntax: 159–181. Academic Press. ——— (1987). Cognitive Constraints on Information Flow. In R. Tomlin (ed.), Coherence and Grounding in Discourse: 21–51. John Benjamins.



Definiteness

——— (1994). Discourse, Consciousness, and Time: The Flow and Displacement of Conscious Experience in Speaking and Writing. Chicago. ——— (1996). Inferring Identifiability and Accessibility. In T. Fretheim & J.K. Gundel (eds.): 37–46. Charolles, M. (1998). Fonctionnement du pronon dans les contextes évolutifs. Paper presented at the 6th International Pragmatics Conference, Reims. Chesterman, A. (1991). On definiteness: A study with special reference to English and Finnish. Cambridge University Press. Christophersen, P. (1939). The Articles. A study of their theory and use in English. Munksgaard. Clancy, P.M. (1980). Referential choice in English and Japanese narrative discourse. In W.L. Chafe (ed.), The Pear Stories: Ablex. Clark, H.H. & S.E. Haviland (1977). Comprehension and the given-new contract. In R.O. Freedle (ed.), Discourse production and comprehension: 1–40. Ablex. Clark, H.H. & C.R. Marshall (1981). Definite reference and mutual knowledge. In A.K. Joshi, B.L. Webber & I.A. Sag (eds.), Elements of discourse understanding: 10–63. Cambridge University Press. Comrie, B. (1978). Definite direct objects and referent identification. Pragmatics Microfiche 3.1.D3. Croft, W. (1990). Typology and universals. Cambridge University Press. Dasinger, L.K. (1995). The Development of Discourse Competence in Finnish Children: The Expression of Definiteness. Ph.D. diss., Berkeley. Donnellan, K. (1971 [1966]). Reference and definite descriptions. In D. Steinberg & L. Jakobovits (eds.), Semantics: 100–114. Cambridge University Press. Du Bois, J.W. (1980). Beyond definiteness: The trace of identity in discourse. In W.L. Chafe (ed.), The pear stories: Cognitive, cultural and linguistic aspects of narrative production: 203–74. Ablex. ——— (1987). The Discourse Basis of Ergativity. Language 63: 805–55. Downing, P. (1996). Proper names as a referential option in English conversation. In B. Fox (ed.): 95–143. Epstein, R. (1994). Discourse and definiteness: synchronic and diachronic perspectives. Ph.D. diss., San Diego. ——— (1995). The later stages in the development of the definite article: evidence from French. In H. Andersen (ed.), Historical Linguistics 1993: 159–176. John Benjamins. ——— (1997). Roles, Frames and Definiteness. Paper presented at ICLA5. Erkü, F. & J.K. Gundel (1987). The pragmatics of indirect anaphors. In J. Verschueren & M. Bertucelli-Papi (eds.), The Pragmatic Perspective: 533–545. John Benjamins. Faingold, E. (1996). Demonstrative Pronouns and the Definite Article in Latin and the Romance Languages. Papiere zur Linguistik 54: 67–82. Ford, C.E. & B. Fox (1996). Interactional motivations for reference formulation: He had. This guy had, a beautiful, thirty-two Olds. In B. Fox (ed.): 145–168. Fox, B. (1987). Discourse structure and anaphora: Written and conversational English. Cambridge University Press. Fox, B. (1996). (ed.), Studies in anaphora. John Benjamins. Fraurud, K. (1990). Definiteness and the processing of noun phrases in natural discourse. Journal of Semantics 7: 395–433. Frege, G. (1892). Über Sinn und Bedeutung. Zeitschrift fùr Philosophie und Philosophische Kritik 100: 22–50. Fretheim, T. & J.K. Gundel (eds.) (1996). Reference and Referent Accessibility. John Benjamins. Ghomeshi, J. (1997). Topics in Persian VPs. Lingua 102: 133–167. Givon, T. (1983). (ed.), Topic Continuity in Discourse: Quantified Cross-Language Studies. John Benjamins.

63

64

Ritva Laury Guillaume, G. (1919). Le problème de l’article et sa solution dans la langue française. Hachette. Gundel, J.K. (1996). Relevance Theory Meets the Givenness Hierarchy. An Account of Inferrables. In T. Fretheim & J.K. Gundel (eds.): 141–154. Gundel, J.K., N. Hedberg & R. Zacharsky (1993). Cognitive Status and the Form of Referring Expressions in Discourse. Language 69: 274–307. Halliday, M.A.K. & R. Hasan (1976). Cohesion in English. Longman. Hannay, M. (1985). English existentials in Functional Grammar. Foris. Hawkins, J. (1978). Definiteness and indefiniteness: A study in reference and grammaticality prediction. Croom Helm. Heim, I.R. (1983). File Change Semantics and the Familiarity Theory of Definiteness. Definite and indefinite noun phrases. In R. Bäuerle, C. Schwartze & A. von Stechow (eds.), Meaning, Use, and Interpretation of Language: 164–189. Walter de Gruyter. Hellman, C. (1996). The ‘price tag’ on Knowledge Activation in Discourse Processing. In T. Fretheim & J.K. Gundel (eds.): 193–212. Hickmann, M. (1982). The development of narrative skills: Pragmatic and metapragmatic aspects of discourse cohesion. Ph.D., University of Chicago. Hjelmslev, L. (1928). Principes de grammaire générale. Host. Huang, C-T.J. (1987). Existential sentences in Chinese and (in)definiteness. In E.J. Reuland & A.G.B. ter Meulen (eds.), The representation of (in)definiteness: 226–253. MIT Press. Jespersen, O. (1924). The Philosophy of Grammar. Allen & Unwin. Kadmon, N. (1990). Uniqueness. Linguistics and philosophy 13: 273–324. Karmiloff-Smith, A. (1979). A functional approach to child language: A study of definiteness and reference. Cambridge University Press. Kibrik, A.A. (1996). Anaphora in Russian narrative prose: A cognitive calculative account. In Fox (ed.): 255–303. Kleiber, G. (1994). L’anaphore associative, d’une conception à l’autre. In C. Schnedecker, M. Charolles, G. Kleiber & J. David (eds.), L’anaphore associative. Aspects linguistiques, psycholinguistiques et automatiques: 5–64. Klincksieck. Laury, R. (1995). On the grammaticalization of the definite article SE in spoken Finnish. In H. Andersen (ed.), Historical Linguistics 1993: 239–250. ——— (1997). Demonstratives in interaction: The emergence of a definite article in Finnish. John Benjamins. Li, C.N. & S.A. Thompson (1981). Mandarin Chinese: a functional reference grammar. University of California Press. Lyons, C. (1999). Definiteness. Cambridge University Press. Master, P. (1995). Consciousness raising and article pedagogy. In D. Belcher & George Braine (eds.), Academic Writing in a Second Language: 183–204. Ablex. Milsark, G.L. (1977). Toward an explanation of certain peculiarities of existential constructions in English. Linguistic Analysis 3: 1–29. Mithun, M. (1992). Is Basic Word Order Universal? In D.L. Payne (ed.), Pragmatics of Word Order Flexibility: 15–61. John Benjamins. Prince, E. (1981). Toward a taxonomy of given-new information. In P. Cole (ed.), Radical Pragmatics: 223–55. Academic Press. ——— (1992). The ZPG Letter: Subjects, Definiteness, and Information Status. In W.C. Mann & S.A. Thompson (eds.), Discourse Description: Diverse Linguistic Analyses of a Fund-Raising Letter: 295–325. John Benjamins. Russell, B. (1905). On denoting. Mind 14: 479–93. ——— (1919). Introduction to Mathematical Philosophy. Allen & Unwin.



Definiteness

Searle, J.R. (1969). Speech Acts. Cambridge University Press. Strawson, P.F. (1950). On referring. Mind 59: 320–44. Ward, G. & B. Birner (1995). Definiteness and the English existential. Language 71: 722–42. Vilkuna, M. (1992). Referenssi ja määräisyys suomenkielisten tekstien tulkinnassa. Suomalaisen Kirjallisuuden Seura.

65

Emergent grammar Marja-Liisa Helasvuo University of Turku

1.  Introduction Emergent grammar is based on the assumption that grammar is not a discrete set of rules that exist a priori, detached from the contexts of use, but rather, that grammar is provisional and that it emerges in discourse (Hopper 1987, 1988, 1998). This approach has been described by Hopper (1988: 118) in the following way: [T]he Emergence of Grammar (EOG) attitude […] has come to view grammar as the name for a vaguely defined set of sedimented (i.e. grammaticized) recurrent partials whose status is constantly being renegotiated in speech and which cannot be distinguished in principle from strategies for building discourses.

According to Hopper, grammar is not a prerequisite for generating discourses, but rather, it is rooted in and evolves from discourse. Neither is it complete and predetermined, but instead, its structure is open and in flux (see also Hopper 1998: 157). Furthermore, grammar is dialogic: its forms are being adapted to the needs of the co-participant in conversation or the audience of a written text. In the case of writing, this cannot be witnessed directly, but when analyzing spoken interaction, the constant monitoring and adaptation is evident on many levels, e.g. in the organization of repair, pausing, prosody, etc. (see e.g. Fox & Jasperson 1995; Schegloff 1979, 1996). Hopper gives the term grammaticization a wider meaning than the standard textbook definition in the literature: the term does not just refer to the historical process whereby a lexical item becomes grammatical or a less grammatical item becomes more grammatical (see e.g. Heine et al. 1991:  3), but also, grammaticization refers to the more fundamental ontology of grammar, i.e. to the ways in which grammar can be said to exist (see Helasvuo 2001a: 1–2): grammar is — to use Hopper’s terms — “a vaguely defined set” of grammaticized recurrent patterns of discourse (Hopper 1988: 118). This way of understanding grammaticization as the crystallization of recurrent discourse patterns is common in functionalist thinking (cf. Bybee et al. 1994; Chafe 1994; Du Bois 1985, 1987; Givón 1979; Hopper & Thompson 1984; Mithun 1991; Ono & Thompson 1995; Helasvuo 2001a; see also Hopper & Traugott 1993; for a critical discussion of this usage of the term grammaticization, see Campbell 2001: 154–157). Hopper has laid out the basic tenets of emergent grammar in two articles (1987, 1988; cf. also Hopper 1998). In several later articles he has analyzed data in concrete



Emergent grammar

detail and shown how the approach can be implemented (see e.g. Hopper 1991, 1997; Thompson & Hopper 2001; cf. also Hopper 1979; Hopper & Thompson 1980, 1984; and papers in Bybee & Hopper 2001). The 1987 and 1988 articles are written in a dialogic fashion, outlining two basic approaches to grammar that are in polar opposition. Hopper describes them as “attitudes”, namely, the “A Priori Grammar attitude” and the “Emergence of Grammar attitude”. He remarks that “what is being described are extreme positions; many linguists in practice occupy a place somewhere between these two poles, and either explicitly or tacitly draw from a middle ground” (1988: 117–118). This remark has gone unnoticed by some critiques of Hopper’s work: inter alia Givón (1999) reduces Hopper’s eloquent rhetorics into a simple opposition between Chomsky (a priori grammar) and Hopper (emergent grammar) and then accuses Hopper of taking an extreme position that cannot possibly represent the facts of grammar in natural language. Givón (1999: 84) concludes — using almost exactly Hopper’s wording (see above) —, “the facts of grammar in natural language use tend to uphold a middleground position, one akin to that of Sapir and Jespersen”.

2.  Routinization and the emergence of grammar In the emergent grammar approach, grammar is not a discrete, predefined set of rules which are presupposed by discourse, but rather, it is an open-ended collection of forms that emerges in actual language use (see Hopper 1998:  158–159). There is a heavy emphasis on repetition or routinization: Becker (1979) speaks of “prior texts”: previous actual utterances form the basis of new utterances. When an action is constantly being repeated in discourse, its situationally defined meaning fades away gradually and it becomes routinized or grammaticized (see Haiman 1994). Furthermore, grammar is fundamentally social and interactive in nature. As an example of the routinization of a frequent discourse pattern, Hopper (1998:  159–160) discusses the English preposition of. Sinclair (1991) shows that the preposition of does not function like other prepositions in English, but instead, it appears almost always in a qualifying expression modifying a noun (e.g. in a possessive construction such as a head of a prepositional phrase). It behaves textually unlike other prepositions in English and therefore, it is questionable whether it should be classified as a preposition at all. (Sinclair 1991: 81–98.) Hopper cites (1999: 159) frequency counts from the Lancaster–Oslo–Bergen corpus of English which show that of is by far the most frequent preposition (38% of occurrences of prepositions, cf. the next most frequent preposition in with 22%). Thus, of stands out in terms of discourse frequencies also. While prepositions in English normally function as the head of a prepositional phrase that itself functions as a sentence adjunct, the preposition of does not occur in this function. The preposition of most often appears in constructions which modify a

67

68

Marja-Liisa Helasvuo

noun (e.g. the city of Atlanta where the of-phrase is an appositional modifier), whereas other prepositions normally occur in prepositional phrases which function as adjuncts. Sinclair (1991) concludes that of is not really a preposition, but rather, it forms a category in which it is the sole member. Hopper (1998: 159–160) takes this movement of the preposition of away from the other prepositions as illustrative of a discourse pattern that has already become sedimented in the grammar of English. Thompson & Mulac (1991) look at the grammaticization of epistemic phrases in English from a quantitative perspective. They study what is called ‘that-deletion’ (see Ex. 1) in English syntax and its relationship to epistemic phrases (Ex. 2 and 3; Ex. 1–3 are taken from Thompson & Mulac 1991: 313).

(1) I think that we’re definitely moving towards being more technical. (2) I think 0 exercise is really beneficial, to anybody. (3) It’s just your point of view you know what you like to do in your spare time I think.

‘That-deletion’ is usually thought of in terms of an alternation between constructions like (1) and (2). Thompson & Mulac argue for an analysis in which there is an alternation of constructions like (1), where I is subject and think is the main verb and that introduces a complement clause, and constructions like (2) and (3) in which I think is an epistemic phrase that expresses the speaker’s stance. They show that in constructions like (2) and (3) the epistemic phrase functions like an epistemic adverb such as maybe with respect to the rest of the clause. In cases like (3), the epistemic phrase does not appear in a position typical of main clauses introducing a complement, i.e. before the complement, and Thompson & Mulac conclude (1991: 326) that epistemic phrases are “free to float to various positions in the clause” just like other epistemic particles do in English. Thompson & Mulac show (1991: 319–321) that among verbs that could take a that-complement, think and guess are by far the most frequent verbs (53% and 12%, respectively), and the first person pronoun I the most frequent subject (83%). Thompson & Mulac argue (1991: 326) that the frequent occurrence of I (in declaratives) and you (in questions) without the complementizer that has led to their re-interpretation as epistemic phrases with verbs expressing belief (such as think and guess). Scheibman (2001) shows that in her data from conversational American English, 1st and 2nd person subjects occur primarily with verbs of cognition and, for 1st person, with verbs of saying. A closer look at these subject + verb combinations reveals that an overwhelming majority of the combinations are conventionalized expressions such as I guess, I don’t know, you know, and I mean, which are grammaticized expressions of epistemic stance (Scheibman 2001: 85; see also Östman 1981; Kärkkäinen 2004). Routinization comes up on all levels of language, not just in clausal syntax, as in the case of the English preposition of and epistemic phrases like I think, but also for example in the sequential structuring of conversation. In an in-depth study of response



Emergent grammar

particles in Finnish conversation, Sorjonen (2001) discusses two response particles niin and joo. These particles have very similar functions: they are both used as positive answers to polar (yes/no) questions, as responses to directives, and as so-called backchannel elements. Studying carefully the action environments of these two particles, Sorjonen is able to show that each particle has its favorite sequential and activity contexts, and in this ‘home environment’ the particle expresses a certain kind of interpretation of prior talk (e.g. the particle may show that prior talk is analyzed as expressing shared information; Sorjonen 2001: 279–280). For the particle niin, Sorjonen shows a grammaticization pathway from an element integrated in clausal morphosyntax, namely, the plural instructive form of the demonstrative se ‘that/it’, into a response particle forming a responding utterance of its own, with its own grammaticized meanings (Sorjonen 2001: Ch. 2). Goodwin (1979) has analyzed the emergence of the sentence as a collaborative enterprise. He notes (1979: 97–98) that “[…] sentences in natural conversation emerge as the products of a process of interaction between speaker and hearer and that they mutually construct the turn at talk”. In a study on the interactive organization of assessment sequences Goodwin & Goodwin (1987) demonstrate how speakers constantly monitor their co-participants when producing assessments. The monitoring concerns not only the verbal and non-verbal responses by the co-participants but also the possible lack of response. Furthermore, Goodwin & Goodwin show how the trajectory of the assessing turn may change as a result of this monitoring. It is in this way that we can say that sentences emerge as an interactive process. Helasvuo (2001b, 2004) looks closely at the syntactic organization of co-constructions, i.e. syntactic constructions jointly produced by two or more conversation participants. In a co-construction, one speaker starts a syntactic construction and another speaker completes it. Example (4) illustrates this. (4) 1. Olli:

[1nythän1] sä voi-t ostaa [2nyt,2]  now you can-sg2 buy  now [now] you can buy [now,]

2. Pekka: [1 nii  so [so

ny,1] now now]

[ 2 nii, 2]  so [so]

→ 3. Tor: toise-n-ki pullo-n. another-acc-cli bottle-acc yet another bottle. (Taken from Helasvuo 2004: 1321–1322)

Olli and Tor produce a transitive construction together, with Olli (line 1) providing the subject pronoun sä ‘you’, the modal auxiliary voit ‘can’ which is inflected in the 2nd person singular as required by the subject, and the transitive verb ostaa ‘buy’ and Tor

69

70

Marja-Liisa Helasvuo

producing the object NP toisenki pullon ‘yet another bottle’ (line 3). The object NP is coded with the accusative case as required by the transitive construction initiated by Olli (line 1). At the point where line 1 has been produced, there is a projection for an object NP, and Tor comes in and provides the NP. He formulates his contribution so that it fulfills the formal criteria set up by the construction so far, i.e. exhibits the required case ending. Thus, co-constructions can reveal how interactants attend to each other and monitor their speech on the basis of recipient response. (For a more thorough discussion, see Helasvuo 2004.) Co-constructions provide evidence for the claim that clauses emerge in an interactive process on the part of the conversation co-participants. In sum, emergent grammar emphasizes the role of repetition and routinization in grammar. The focus is on grammar as it is being used for communication. Therefore, in order to come up with meaningful facts about language, one has to look at actual language use. Hopper (1998: 165) states: “Structure can only be seen as emergent if forms are examined in their naturally occurring contexts”. Instead of being situated in the individual mind, grammar is seen as distributed among co-participants in interaction.

3.  Emergent grammar within linguistics Emergent grammar could be seen as belonging to the field of functional linguistics. In no way, however, can it be described as representing mainstream functionalism. With its emphasis on naturally occurring data studied in context, it has little in common for example with the kind of typologically-oriented functionalism that relies on isolated sentence examples taken from reference grammars and textbooks. In this section, I will discuss where emergent grammar stands in the field of linguistics, looking for possible connections. In his more theoretical articles (1987, 1988, 1998), Hopper builds a contrast between emergent grammar and what he calls a priori grammar according to which grammar is a discrete set of rules which are logically and mentally presupposed by discourse (Hopper 1988:  118). The a priori grammar position holds that grammar can be — and perhaps also should be — studied as detached from its contexts of use. Relevant data is supplied by intuition. Moreover, grammar is a static entity which is fully present at all times in the mind of the speaker. In contrast, from the perspective of emergent grammar, language is a real-time activity and its regularities are subject to negotiation or even abandonment (Hopper 1988: 120). Hopper’s description of a priori grammar covers various strands of linguistics. He emphasizes (1988: 117–118) that it is not meant to describe any particular linguist or school of thought. The same is true of his description of the Emergence of Grammar attitude: many linguists — be it of functional or some other persuasion — certainly



Emergent grammar

subscribe to the idea that grammar, like speech, is a real-time, social phenomenon (Hopper 1998: 156), but not very many share the view that there is “continual movement toward structure, a postponement or ‘deferral’ of structure”, or even less the idea that “there is […] no ‘grammar’ but only ‘grammaticization’ — movements toward structure” (Hopper 1987: 142, 144). It is, indeed, an “extreme position”, as Hopper himself puts it (1988: 118; for a criticism of this position, see Givón 1999). One of Hopper’s most crucial contributions to the field has been his insistence on the importance of questioning our received notions and categories (see e.g. Hopper & Thompson 1984; Hopper 1997 on lexical categories; Hopper & Thompson 1980; Thompson & Hopper 2001 on transitivity; Hopper 1988, 1998 on the notion of sentence). Instead, one should look at discourse — spoken or written — for regularities that emerge from actual language data. Becker (1979) talks about ‘figures’: a figure is a phrase or clause which is highly standardized in its format and which permits substitution, but the substitutions are restricted to some few places. For the most part, discourse consists of assemblages of such figures (see also Du Bois 2001a on dialogic syntax). As an example of such figure, we could take Lambrecht’s work on spoken French. Lambrecht (1987) shows that there is an emergent clause type in spoken French which he calls the preferred clause structure. It is of the form V (X) where the subject is an incorporated pronoun which is also the unmarked topic of clause. Full lexical NPs, if there are any, follow the verb (marked with X in the formula) and are semantically nonagentive and pragmatically non-topical. He notes that there are two major construction types that function to preserve the preferred clause structure, namely the so-called dislocations and clefts, which serve to manipulate the argument structure so that full lexical NPs fall outside the clause. They also manage information flow by presenting new referents and handling topics. (Lambrecht 1987, 1988; cf. also Du Bois 1985, 1987, 2001b on preferred argument structure.) Later on, Lambrecht has developed his ideas further in the framework of construction grammar (see e.g. Lambrecht 1994); with its interest in the interplay between lexical and idiomatic routinization and syntactic regularities, construction grammar has much in common with emergent grammar. Interactional linguistics (as presented in Selting & Couper-Kuhlen 2001) searches for syntactic and sequential routinization in conversation. In a fashion similar to emergent grammar, there is no meaningful distinction between competence and performance, which is based on the idea that all linguistic knowledge is stored in the individual speaker’s mind, and this knowledge is independent of and detached from the actions and the contexts of use. Couper-Kuhlen & Selting (2001: 4–5) note that interactional evidence goes against the view of language as an abstract system of pre-established discrete elements that comprise grammar. They argue that “language forms and structures must be thought of in a more situated, context-sensitive fashion as actively (re)produced and locally adapted to the exigencies of the interaction at hand. In this sense they can be conceived of as arising or emerging in use”.

71

72

Marja-Liisa Helasvuo

Hopper claims (1998: 162) that grammar is “distributed among the various participants in a collaborative act of communication”. In Hopper’s writings, however, it is not easy to find any concrete methods for analyzing this collaboration in closer detail; this, then, is a task for scholars working in interactional linguistics, conversation analysis and discourse analysis. In sum, emergent grammar can be viewed as an understanding of the nature of human language (Hopper 1988: 117); it does not offer a concrete framework of analysis for all facets of language it discusses (inter alia, a method for analyzing language as a tool for social interaction). It provides a detailed view on the ontology of grammar and language, and an insightful discussion of the implications this ontology for the study of different languages.

References Becker, A.L. (1979). The figure a sentence makes: An interpretation of a classical Maly sentence. In T. Givón (ed.): 243–260. Bybee, J. & P. Hopper (eds.) (2001). Frequency and the Emergence of Linguistic Structure. John Benjamins. Bybee, J., R. Perkins & W. Pagliuca (1994). The Evolution of Grammar: Tense, Aspect and Modality in the Languages of the World. The University of Chicago Press. Campbell, L. (2001). What’s wrong with grammaticalization? Language Sciences 23: 113–161. Chafe, W.L. (1994). Discourse, Consciousness, and Time. The Flow and Displacement of Conscious Experience in Speaking and Writing. The University of Chicago Press. Couper-Kuhlen, E. & M. Selting (2001). Introducing Interactional Linguistic. In Selting & Couper-Kuhlen (eds.): 1–22. Du Bois, J.W. (1985). Competing Motivations. In J. Haiman (ed.), Iconicity in Syntax: 343–365. John Benjamins. ———(1987). The Discourse Basis of Ergativity. Language 63: 805–855. ——— (2001a). Towards a Dialogic Syntax. Ms. University of California, Santa Barbara. ——— (2001b). Discourse and Grammar. In M. Tomasello (ed.), The New Psychology of Language: Cognitive and Functional Approaches to Language Structure. Vol. 2: 47–87. L. Erlbaum. Fox, B. & R. Jasperson (1995). A Syntactic Exploration of Repair in English Conversation. In P.W. Davis (ed.), Alternative Linguistics: Descriptive and Theoretical Modes: 77–134. John Benjamins. Givón, T. (1979). From Discourse to Syntax: Grammar as a Processing Strategy. In T. Givón (ed.): 81–112. ———(1999). Generativity and variation: The notion of ‘Rule of Grammar’ revisited. In B. MacWhinney (ed.), The Emergence of Language: 81–114. Erlbaum. Givón, T. (ed.) (1979). Discourse and Syntax. Academic Press. Goodwin, C. (1979). The interactive construction of a sentence in everyday conversation. In G. Psathas (ed.), Everyday language: Studies in ethnomethodology: 97–122. Irvington. Goodwin, C. & M.H. Goodwin (1987). Concurrent operations on talk: Notes on the Interactive Organization of Assessments. IPRA Papers in Pragmatics 1: 1–54. Haiman, J. (1994). Ritualization and the development of language. In W. Pagliuca (ed.), Perspectives on Grammaticalization: 3–28. John Benjamins. Heine, B., U. Claudi & F. Hünnemeyer (1991). Grammaticalization: A Conceptual Framework. University of Chicago Press.



Emergent grammar

Heine, B. & E. Traugott (eds.), Approaches to Grammaticalization. John Benjamins. Helasvuo, M.-L. (2001a). Syntax in the Making. John Benjamins. ——— (2001b). Emerging syntax for interaction: Noun phrases and clauses as a syntactic resource for interaction. In M. Selting & E. Couper-Kuhlen (eds.): 25–50. ——— (2004). Shared Syntax: The Grammar of Co-Constructions. Journal of Pragmatics 36: 1315–1336. Hopper, P.J. (1979). Aspect and foregrounding in discourse. In T. Givón (ed.): 213–242. ——— (1987). Emergent grammar. Papers of the 13th Annual Meeting, Berkeley Linguistic Society: 139–157. Berkeley Linguistic Society. ——— (1988). Emergent Grammar and the A Priori Grammar Postulate. In D. Tannen (ed.), Linguistics in Context: 117–134. Ablex. ——— (1997). Discourse and the category ‘verb’ in English. Language & Communication 17: 93–102. ——— (1998). Emergent grammar. In M. Tomasello (ed.), The New Psychology of Language: 155–175. Erlbaum. Hopper, P.J. & S.A. Thompson (1980). Transitivity in Grammar and in Discourse. Language 56: 251–299. ——— (1984). The Discourse Basis for Lexical Categories in Universal Grammar. Language 60: 703–752. Hopper, P.J. & E. Traugott (1993). Grammaticalization. Cambridge University Press. Kärkkäinen, E. (2004). Epistemic stance in English conversation. A description of its interactional functions, with a focus on I think. John Benjamins. Lambrecht, K. (1987). On the status of SVO sentences in spoken French discourse. In R. Tomlin (ed.), Coherence and Grounding in Discourse: 217–261. John Benjamins. ——— (1988). Presentational cleft construction in spoken French. In J. Haiman & S.A. Thompson (eds.), Clause combining in grammar and discourse: 135–179. John Benjamins. ——— (1994). Information Structure and Sentence Form: A Theory of Topic, Focus, and the Mental Representation of Discourse Referents. Cambridge University Press. Mithun, M. (1991). The role of motivation in the emergence of grammatical categories: the grammaticization of subjects. In B. Heine & E. Traugott (eds.), Vol. II: 159–184. Ono, T. & S.A. Thompson (1995). What can conversation tell us about syntax? In P. Davis (ed.), Alternative linguistics: Descriptive and theoretical modes: 213–271. John Benjamins. Östman, J.-O. (1981). You know: A Discourse Functional Approach. John Benjamins. Schegloff, E.A. (1979). The Relevance of Repair for Syntax-for-Conversation. In T. Givón (ed.): 261–288. ——— (1996). Turn Organization: One Intersection of Grammar and Interaction. In E. Ochs & E.A. Schegloff & S.A. Thompson (eds.), Interaction and Grammar: 52–133. Cambridge University Press. Scheibman, J. (2001). Local Patterns of Subjectivity in Person and Verb Type in American English Conversation. In J. Bybee & P. Hopper (eds.): 61–89. Selting, M. & E. Couper-Kuhlen (eds.) 2001. Studies in Interactional Linguistics. John Benjamins. Sinclair, J. (1991). Corpus, concordance, collocation. Oxford University Press. Sorjonen, M.-L. (2001). Responding in Conversation. John Benjamins. Thompson, S.A. & P.J. Hopper (2001). Transitivity and Clause Structure in Conversation. In J. Bybee & P. Hopper (eds.): 27–60. Thompson, S.A. & A. Mulac (1991). A quantitative perspective on the grammaticalization of epistemic parentheticals in English. In B. Heine & E. Traugott (eds.), Vol. II: 313–329.

73

Frame analysis Branca Telles Ribeiro & Susan M. Hoyle Lesley University/National Library of Medicine

1.  Introduction Frame analysis, a way of studying the organization of experience, is an approach to cognition and interaction that focuses on the construction, conveying, and interpretation of meaning. Frames affect the way in which we categorize, remember, and revise what we know, as well as what we say, how we mean it, how others hear it, and how we do things together linguistically and otherwise. The term frame and the related terms script and schema have been used similarly in the social sciences as well as artificial intelligence. Some usages are more static and some more dynamic, but all have to do with structures of expectation: we predict what new experiences and relationships will be like and interpret them against our prior knowledge. A frame is the interpretation that is essential to any activity or text. We identify, both to ourselves and to our interactants, how to define a situation, how its parts fit into the whole, how the whole relates to larger structures of experience, and how what is unfolding at a given moment affects what will come next. Language users, whether conversationalists, readers, or viewers of a film, must decide whether a message — a statement, command, question, or laugh — is hostile, friendly, aggressive, or flirtatious; intended to amuse, inform, annoy, or persuade; intended as a main point or a side remark. Framing, then, is a matter of conveyed meaning more than literal meaning.

2.  What are frames? Tannen ([1979] 1993) reviews key concepts from different knowledge domains, pointing out similarities in the way the terms frame, script, and schema have been used. In one important sense, these terms all refer to cognitive constructs, networks of related ideas or images. One strand of research derives from artificial intelligence. Minsky (1975) discusses the representation of stereotyped situations and multi-level networks of knowledge in terms of frame, while Bobrow & Norman (1975) characterize the organization of computer memory in terms of schema, and Rumelhart (1975) proposes a schema for stories to aid in their automatic parsing. Schank & Abelson (1977) introduce



Frame analysis

the term script to refer to a schema for events that proceed in a particular temporal order and that include slots for actors, actions, and props. Their classic restaurant script, originally designed to indicate what a computer would need to presuppose in order to infer meaning from text, exemplifies organized knowledge about an everyday event. In this example, actors include customers and waiters; actions include sitting down, ordering, eating, and paying, in that order; props include menus and food. Because the elements form a connected web, not all elements need be explicitly introduced in a text: in the restaurant script, the waiter and the check may be referred to without preamble, as may the teacher in a classroom script (Nelson 1986). Such cognitive constructs let us engage in, remember, and classify experiences. In psychology, Bartlett (1932), contending that schemas are flexible, active, and developing, argued that memory is actively constructed and organized, with details subsidiary to an overall general impression. More recently, schemas have been nominated as the ‘building blocks of cognition’ (Rumelhart 1980), a view implying that pattern recognition and pattern building are fundamental to the workings of the mind. Cognitive anthropologists have emphasized that different social groups have different ‘cultural models’ (Holland & Quinn 1987); these need not reside fully within individuals’ minds, for pieces of a single cultural schema may be distributed among different members. Linguists have discussed how cognitive frames guide the interpretation and production of discourse. In frame semantics (Fillmore 1976; Petruck 1995), it is proposed that both the prototypical meaning of a word and knowing how to use it depend on how it and related words and referents form a conceptual system. In text analysis, van Dijk (1980) claims that schemas organize knowledge and produce textual organization, operating on microstructure (words, clauses, sentences), macrostructure (topics, themes), and superstructure (text types). A narrative superstructure, for instance, organizes and constrains the content of stories, epics, parables, advertisements, myths, and rumors. Expectations about narrative structure and content are culturally relative: Tannen (1979) shows how, in storytelling, cross-cultural differences in frames affect surface linguistic form. Recent research in anthropology, sociology, and sociolinguistics has called for a more dynamic use of the notion of frame — essential to any speech activity and any interaction. As Bateson ([1954] 1972) originally pointed out, all communicative moves are framed by implicit messages about how the content is to be taken. He explains frames using the analogies of picture frames (where the frame tells the viewer what to pay attention to) and, more abstractly, of mathematical sets (how logical implications of membership may apply). Participants define an activity by relying on the following communicative and psychological premises: (1) any sign is only a signal and any signal may have more than one interpretation (e.g., in one situation a bite may signal ‘play’, while in another situation it may signal ‘combat’); (2) one set of messages (‘this is play’) excludes other messages (‘this is combat’); (3) messages within a set share common

75

76

Branca Telles Ribeiro & Susan M. Hoyle .

assumptions (‘this is play’ may imply ‘we are friends’, ‘this is fun’, and shouting and punching are to be understood as a ‘game’; (4) there is no communication without metacommunication (messages cannot be understood unless they are framed); (5) frames are embedded within one another (the ‘outer frame’ or ‘ground’ delimits how the message — which becomes figure — is to be understood); (6) framing is dynamic (what set of messages become figure, while other sets become ground, is a jointly constructed process in an ongoing activity). These processes capture how indirectness and ambiguity work in everyday conversations. Of particular interest are paradoxical framings such as ‘play’, ‘irony’, ‘threat’, or ‘fantasy’. Goffman’s influential Frame Analysis (1974) expands on Bateson’s definitions and provides numerous examples of framing in everyday life, revealing how people transform activities in systematic ways and cue such transformations. A frame consists of “principles of organization which govern events — at least social ones — and our subjective involvement in them” (1974: 10). There is no out-of frame activity. Furthermore, participants in a social encounter are continuously reframing talk and thereby changing the ongoing interaction. Simultaneous meanings and multiple realities are not exceptional: there is no limit to the number of ‘rekeyings’ of an activity. Each framing, however, has its own realm of existence; following James (1950), Goffman points out that selective attention can make real several different worlds. ‘Frames’, ‘laminations’, ‘keyings’, and ‘fabrications’, among other concepts, constitute a ‘framework of frameworks’. Goffman’s notion of frame lamination operates in much the same way as Bateson’s psychological frames, where some situational contexts are brought to the foreground while others become background. Distinguishing keyings (systematic transformations) from fabrications (intentional efforts to manage activity so as to elicit a false belief) provides meaningful information about the nature of reframings in settings such as psychiatric wards, dramatic presentations, or storytelling. Recent analyses of interaction stress even more the dynamic, processual potential of frames. Tannen & Wallat ([1987] 1993) propose a useful distinction between interactive frames and knowledge schemas. Knowledge schemas are cognitive entities consisting of “participants’ expectations about people, objects, events and settings in the world” (1993: 60). Interactive frames, on the other hand, have to do with the ways in which people signal and interpret what activity they are jointly constructing. Frames and schemas work together to structure an interaction, as Tannen & Wallat show through analysis of a videotaped pediatric examination, in which the participants juggle several frames: that of a social encounter (in which the doctor entertains the child and chats with the mother), a medical examination (in which the doctor not only examines the child but also explains what she is doing to a video audience, ignoring the mother), and a consultation (during which the doctor shares information with the mother). Analysis shows that when knowledge schemas are not shared, the flow of an exchange can be disrupted and the interactive frame of the moment modified: a mismatch of



Frame analysis

schemas about health on the part of the doctor and the mother repeatedly triggers frame switches. Crucial to the findings, though, is that the frames are interactively achieved; whether it is mismatched schemas or another trigger that results in a shift in frame, it is not done unilaterally. Also crucial is how much work is involved in accomplishing each frame shift as participants negotiate the appropriate register, discourse topic, and verbal and nonverbal moves. 3.  Frame and context in interaction A frame perspective on interaction takes the view that context is not separate from interaction. Context does not reside in the physical setting or in the enduring characteristics or relationships of participants. Rather, interactants help create the very context in which they act; they “dynamically reshape the context that provides organization for their actions within the interaction itself ” (Goodwin & Duranti 1992: 5). People may orient to their surroundings, past experiences, assumptions about their interlocutors, expectations about the future (and so on in an infinite list), but just which such things will matter in any given encounter cannot be predicted. Context, in this view, may alter from moment to moment: a frame perspective shares with an ethnomethodological perspective (e.g., Atkinson & Heritage 1984) the assumption that context is not given a priori but is continually renewed and modified. As Erickson & Shultz (1981: 147) say, “the capacity to assess when a context is as well as what it is” is an essential part of social and communicative competence. Interactants must signal and recognize transition points in activities in order to produce behavior appropriate to each stage as well as taking into account sequential relationships among behaviors (e.g., ‘fast’ speech is fast only in comparison to what preceded it). In the last analysis, though, context boundaries are notoriously ambiguous, and participants are usually unable to identify on the fly the exact moment when the definition of a situation changes. Such analysis, then, takes a participants’ perspective on context while emphasizing that not all potentially relevant factors are in fact so treated. Many elements of a frame are encoded and understood linguistically, or at least paralinguistically. In this regard, Gumperz’s (1982, 1992) ‘contextualization cues’ are crucial. Contextualization cues include code and variety choice, topic, tempo, rhythm, voice quality, formulaic expressions, and nonverbal behavior such as laughing, smiling, and postural shifts. A shift in register — lexical, syntactic, or prosodic features conventionally associated with particular settings or participants — generally functions as a cue to a change in frame. Contextualization cues indicate, subtly but powerfully, how speakers intend for their hearers to interpret what they say. Interpretation is always situationally based rather than predictable by grammatical rules, and inferencing by hearers is as important as encoding by speakers. Where cues are misinterpreted, often

77

78

Branca Telles Ribeiro & Susan M. Hoyle

because of cultural or other background differences, an interaction may well break down. Even when conventions are shared, cueing may be problematic, as Dorr-Bremme (1990) demonstrates in a study of an elementary school classroom: when the teacher omitted her customary contextualization cues at activity junctures, she inadvertently relinquished control of the classroom talk. Topic is a key component in defining contextual boundaries and is part of the definition of an interactional situation. To apply Tannen & Wallat’s (1993) distinction, knowledge about topics, as well as knowledge about when and where to discuss certain topics, are schemas (relatively stable cognitive constructs). But topic affects and is affected by interactive framing. In the analysis of a conversation, figuring out what the topic is, why it is relevant, and which participants develop it helps understand how specific framings are created and sustained. If speakers avoid controversial topics such as religion or politics, it is in order to avoid reframing their talk as heated argument; if strangers talk about the weather, it is in order to frame their talk as sociable and unthreatening. In institutional interactions, such as those in educational, clinical, or judicial settings, topic is often dictated by an institutional agenda. For example, in doctor/patient interviews, the doctor brings a topic agenda that helps establish the major interactional contexts (Ribeiro 1994). Even such topical talk, however, can trigger more or less subtle frame shifting: a doctor’s request for information about the patient’s social history may trigger a personal story and thus a personal framing and a change in participation structure. The interrelation of frame and topic has implications for coherence (Ribeiro 1993, 1994). Speakers and hearers have multiple options for coherence as they select candidates from different levels of talk (Schiffrin 1987). The following illustrates these choices:

Doctor: Patient: Doctor: Patient:

you were born on what date? on January 11. of what year? of 1921. I am sixty-one.

Coherence is achieved within the exchange system as speakers follow turn-taking rules and alternate in the speaking role. It is achieved simultaneously in the action structure, when the next speaker provides the expected next action (a request for information is followed by a statement). Finally, the propositional content coheres as participants jointly refer to the ongoing topic (date of birth and age). To make sense of the above interaction, however, one needs to attend to the situation: what is going on here and now? This is the opening of a standard medical (psychiatric) interview. For a psychiatric patient, making the proper choice among social and linguistic options has definite implications for the doctor’s evaluation and diagnosis. Thus the patient must jointly construct with the doctor the institutional frames of the psychiatric interview. Coherence



Frame analysis

is achieved when both participants share an understanding of the ongoing talk/activity, its purposes, and ways of interaction (Ribeiro 1994).

4.  Frame and footing People continuously reframe their ongoing activities by manipulating footing — Goffman’s (1981) term for the way in which framing is accomplished in verbal interaction. Footing is the stance that speakers and hearers take toward each other and toward the content of their talk. In Goffman’s (1981: 128) words, it is “the alignment we take up to ourselves and the others present as expressed in the way we manage the production or reception of an utterance.” A frame, then, is constructed through participants’ signaling their own and recognizing and ratifying one another’s footing — which may and usually does change from moment to moment. Furthermore, Goffman suggests that most often interactants do not simply change footing, but rather embed one footing within another. Such embedding of interaction arrangements, or ‘lamination’ of experience, permeates talk, and “within one alignment, another can be fully enclosed” (1981: 155). Footings and footing shifts are constituted and evidenced in large part through changes in the ‘participation framework’ of talk (Goffman 1981), the complex relations among speakers, hearers, and utterances. A speaker, for instance, may at any one time be acting as ‘animator’, ‘author’, and/or ‘principal’ of his or her utterance. The animator speaks the words, the author chooses and arranges the words, and the principal is responsible for the position behind the words. These roles may be lodged in one and the same person (the normal and unmarked situation), or they may be distributed across persons. On the hearer’s side, one may be either a ratified or an unratified participant; those ratified include both addressed and unaddressed recipients, while unratified persons include overhearers and eavesdroppers. Although overhearers are, by definition, unaddressed, they may well be taken into consideration by speakers and may in fact be the target of a message. For instance, Morgan (1996) describes ‘pointed indirectness’, part of the African American practice of ‘signifying’, in which a speaker directly addresses a ‘mock receiver’ but intends the utterance for someone else who is present. Levinson (1988) argues that Goffman’s decomposition of the traditional notions of speaker and hearer may be carried further, and he identifies numerous participant roles (such as ‘Relayer’, ‘Sponsor’, ‘Ghoster’, ‘Target’). Irvine (1996), on the other hand, points out that such a list of participant roles would be infinite — and would not be universally applicable — because roles are creatively constructed to fit each novel situation. Scholars who apply the notion of framing to data share the assumption that looking at naturally occurring, connected discourse yields a fuller understanding than looking at isolated sentences or constructed texts. In what follows we mention some recent work that uses the concepts central to frame analysis.

79

80

Branca Telles Ribeiro & Susan M. Hoyle

5.  Framing and nonverbal communication Kendon’s (1990) work on focused encounters extends frame theory to nonverbal communication. Particularly relevant is the F-formation (or face formation), a system of spatial distribution, postural configuration, and orientation. It is an interactional unit of behavior that “clearly demarcat[es] the ‘world’ of the encounter from the rest of the ‘world’ ” (1990: 250). Spacing and orientation work in the process of ‘frame-attunement’, constituting a ‘scaffold’ for whatever people will jointly construct (1990: 253), signaling major and minor adjustments in frame. Positioning arrangements may ‘preframe’ (1990: 253) a new activity (for example, to initiate a conversational closing, one may step back slightly, after which others may also step back, thus acknowledging the proposal for closure). Central, too, to Kendon’s analysis is the distinction between what participants interpret as signal (intended communication) and noise (background information). Participants continuously select ‘attentional tracks’ (Goffman’s story-line tracks) as well as ‘directional tracks’ (Goffman’s signs which regulate components and phases of activities). ‘Disattended tracks’ play an important part as well, for what was previously jointly understood as noise may be reframed as signal if given the proper attention. Others have also pointed to the importance of nonverbal activity in framing encounters. Playground games (M. Goodwin 1985, 1995, 1998), for instance, may be an arena in which players take authoritative stances toward each other’s missteps. Such stances are simultaneously and sequentially enacted both verbally — through challenges, teasing, insults, and rejoinders, often issued with dramatic intonation — and nonverbally — through challenging postures, gestures such as accusatory pointing, or replaying of another’s jump to demonstrate its invalidity. C. Goodwin (1995) similarly demonstrates the centrality of postural shifts, combined with intonational shifts, in an aphasic man who is nonetheless able to communicate. In everyday conversation as well, coordination among speakers and hearers is an accomplishment dependent upon both vocal and nonvocal elements such as gaze (C. Goodwin 1981; Erickson 1996; Schegloff 1999). Participants, then, define their activities — as play, as conflict, as conversation — through the ways in which they align to each other not only verbally but also physically. 6.  Framing in everyday talk Not only do interactants frame their talk as constituting a particular overall sort of activity, but they frame each moment of talk. Storytellers, for instance, can use a particular pattern of tense switching to indicate to their audiences that characters in a story are on unequal footings (Johnstone 1987). In personal narratives about confrontations



Frame analysis

with authority figures, Johnstone finds a regular pattern of alternation in the tense of dialogue attributors (words used to introduce a direct quotation): narrators tended to use the historical present (he says) to introduce the speech of authorities (police officers, parents, or anyone of higher status) and the past tense (I said) to introduce the speech of nonauthorities (usually the speakers themselves). The notion of footing, attending as it does to the micro-interactional level, meshes with much of what conversation analysts (among others) have noticed about the intricacies of turn-by-turn sequencing in conversation. When one speaker completes another’s utterance, for example, the original speaker often adds a third part to the sequence, ratifying or rejecting the proposed completion. Antaki, Diaz & Collins (1996) note that in order for a completion to be ratified, it must not only accord with the propositional content of the first speaker’s utterance but must maintain the same participant status; that is, both utterances must have, say, the footing of ‘author’, ‘relayer’, or ‘spokesperson’. Such a sequence (utterance/completion/ratification) may also produce a new footing — which may not be noticed without paying attention to the sequence — of ‘collective author’ (Diaz, Antaki, and Collins 1996). Footing shifts, and thus frame shifts, are a combined matter of form, placement, and content. Clift (1999) demonstrates as much in an investigation of irony in conversation which, by using the concepts of frame and footing, shows irony to be very often affiliative and not, as is often assumed, necessarily hostile or sarcastic. She notes that ironic conversationalists, like journalists (discussed below), act as animators who are not authors or principals and thus “signal a lack of commitment to what they are saying in the very act of saying it” (1999: 532). However, an ironic conversational contribution is framed with a double meaning, an ‘inside’ and an ‘outside’ one, with the speaker ostensibly being detached (animator only) but letting it be known that she is in fact a fully involved participant (animator, author, and principal all in one). Attention to frame and footing can show how identity emerges through talk. Gender, for instance, may be enacted through the alignment — of authority, rapport, or competition — taken to interlocutors (Tannen 1996). Gendered identities may be constructed through what appears to be the same speech act: ‘speaking for another’ (Schiffrin 1993) — speaking on behalf of someone who is present — can convey solidarity (a typically female stance) when it supportively completes the other person’s utterance, but it can be judgmental (a typically male stance) when it contradicts the first speaker. On the other hand, an examination of footing can refute gender stereotypes: M. Goodwin (1998) shows how close analysis of stance during a girls’ game of hopscotch challenges widely held beliefs that girls’ games lack complexity and that all girls shun conflict. Other aspects of identity are created through footing shifts as well. Zentella (1997) shows that New York Puerto Rican children switch between Spanish and English not only for discourse-internal reasons (e.g., mitigating requests or adding clarification) but also to reinforce their identity as members of two worlds and their

81

82

Branca Telles Ribeiro & Susan M. Hoyle

refusal to choose between them. Rampton (1995), too, identifies a group of young people whose language choices from moment to moment are central to their community. Among a multiracial group of London adolescents, patterned switching among Creole, Panjabi, ‘stylized Indian English’, and the local vernacular accomplishes local functions (such as resisting adults, teasing peers, or voicing serious disagreement) but at the same time helps forge a new interethnic community.

7.  Framing in play Framing is essential to any activity, but it is particularly prominent during play. Play often occurs as part of conversation in the form of joking and teasing. Conversational joking of all sorts — teasing, punning, humorous anecdotes, sarcasm, mocking of self and others — relies upon participants’ willingness, indeed eagerness, to engage in extended play frames (Norrick 1993). The creation of a teasing frame (Straehle 1993) is a prime example of one which must be recognized and ratified by all participants, for if an intended tease is interpreted seriously, it will not succeed and miscommunication will result. Such a frame may be signaled through prosody, exaggeration, laughter, unusual pronoun use, or formulaic routines. It is created and sustained through the fine details of footing, with speakers tying their teases formally and sequentially to each other’s utterances. Among children, verbal play is an integral part of imaginary play in which they take on pretend roles. Children hone their framing abilities as they construct elaborate interweavings of literal, possible, and fanciful participation frameworks. Imaginary play gives children the incentive to create a variety of frames using all the resources at their disposal, including their growing mastery of linguistic registers (Hoyle 1993, 1998) and discourse markers (Hoyle 1994). Children at play embed participation frameworks within each other as they shift among their actual identities and pretend identities. In the following example, two nine-year-old boys are playing a game of computer basketball. Josh has been speaking as the sports announcer but then takes on the role of interviewer: Josh: Bird has extended his lead to six points, and we’re gonna have an interview with Larry Bird. Let’s go down to Howard. Howard? Well hi, Larry Bird, how d’ ya feel about your performance Matt: Well if they didn’t call that twenty-four second violation, why did they call that one earlier in the first quarter. [interview continues]



Frame analysis

As announcer, Josh refers to ‘Larry Bird’ (played by Matt) in the third person, thus indicating that he is addressing the (imaginary) audience, not addressing his playmate. When he shifts to speaking as an interviewer, he addresses Larry Bird in the first person. In addition, when he first speaks as interviewer (‘Howard’), Josh marks his question with well, and Matt (as ‘Larry’) does the same with his answer. This discourse marker, which indicates a speaker’s alignment to an interlocutor (Schiffrin 1987), also marks here the speakers’ orientation to a new participation framework. In their play, then, the children sustain imaginary roles by aligning in specific ways to each other’s pretend selves. 8.  Framing and institutional discourse As interactional concepts that capture the workings of face-to-face communication, frame and footing are only starting to be used in discussing broader, more structural, macro-social factors. However, recent work shows that frame analysis can indeed be a powerful analytic tool for understanding the micro-level workings of macro-social factors. Rather than assuming social factors to be self-evident, external, and causative, frame analysis, by focusing on interactional processes, reveals social structures and processes as emerging in talk. It is increasingly noted that institutional talk — in which representatives of medical, legal, academic, or commercial communities talk with their clients, patients, students, or customers — is institutional not because of the setting in which it is produced or because of the pre-existing institutional identities of its producers. Rather, participants themselves construct its nature by displaying their orientation to institutionally relevant activities and identities on a moment-to-moment basis. Much of the argument to this effect focuses on the sequential construction of talk without specifically using the notions of frame or footing. These notions, however, are in fact valuable, as is argued, for instance, by Hutchby (1999) in an analysis of the opening sequences of talk radio shows. He demonstrates that in the first three seconds of talk, speakers take a series of footings as they incrementally achieve ‘frame attunement’ and establish their institutional identities as caller and host. In other recent work, Matoesian (1999) applies the notion of footing to close examination of courtroom interaction and shows how in one trial the defendant is able to take on the identity of ‘expert’ during cross-examination. The notion of footing has helped illuminate the structure of news interviews. Interviews differ from ordinary conversation in that they consist of pre-allocated turn types, the interviewer’s task being to ask questions and the interviewee’s to answer. Routinely, it is found, participants adopt a footing that displays their orientation to this format even when the limits of the format are stretched. For instance, Heritage & Roth (1995) find that

83

84

Branca Telles Ribeiro & Susan M. Hoyle

even when interviewers’ utterances are not purely interrogative in form, interviewees delay their turn until a recognizable question is issued. In addition, interviewers manipulate their footing in order to achieve a ‘neutralistic’ stance (Clayman 1992; Greatbatch 1998) in which they refrain from asserting their own opinions. An interviewer may, for instance, attribute a controversial position to someone not present, so that the interviewer is seen to be merely the animator of objectionable sentiments. Such a neutralistic stance, of course, requires the collaboration of interviewees, which is usually accorded. Sometimes, however, such collaboration is not forthcoming, and it is these unusual instances that point up the regularity with which tacit collaboration does occur. Neutralism is seen in other kinds of institutional interaction as well. In informal court proceedings (Atkinson 1992), arbitrators display their neutrality by actively refraining, on a turn-by-turn basis, from both affiliation and disaffiliation with witnesses. Family mediators, too (Greatbatch & Dingwall 1999), refrain from expressing their own opinions, even while subtly advocating certain actions and discouraging others. As in the case of interviews and arbitration, mediators and disputants collaborate to uphold the mediator’s neutralism, for it is necessarily a joint production. 8.1  Framing and education Skillful teachers can facilitate children’s adoption of a range of participant roles. O’Connor & Michaels (1996) show how some teachers provide children with access to academically valued roles used in solving complex problems, such as ‘hypothesizer’, ‘evidence provider’, or ‘maker of distinctions’. These teachers may reformulate or ‘revoice’ students’ contributions to the solution of a problem, but in doing so the adult merely re-animates (while making more concise or coherent) the child’s utterance. The child remains the principal, the one who gets credit for the idea. McCreedy (1998) contrasts lessons during which children merely animate their teachers’ scripts (by providing a single word or phrase in response to elicitation) with lessons that afford students opportunities to author their own oral academic texts. Adger (1998), too, examines how children may be given opportunities to practice an academically valued stance. Among elementary-school students who are native speakers of a vernacular dialect (African American Vernacular English), shifting into standard English marks a prototypically academic stance, especially in a classroom in which the vernacular is not stigmatized. By reserving the standard dialect for their most formal register, the children use it to take an “authoritative footing” that is, “speaking as one who expects to be believed, by virtue of one’s academic knowledge” (Adger 1998: 151). 8.2  Framing and medicine Examples from medicine (Candlin 1997; Cicourel 1992; Hydén & Mishler 1999) and psychiatry (Ribeiro 1994) show that that doctor-patient communication should be



Frame analysis

studied as “meaningful talk […] an essential and critical component of clinical practice” (Mishler 1984:  8). In doctor-patient talk, institutional frames prevail over other possible frames, with talk about medical concerns overshadowing more personal talk. However, frame dynamics can modify the typical asymmetrical doctor/patient relationship. Ribeiro’s (1996) analysis of a psychiatric interview shows how institutional frames (mostly proposed by the doctor) and personal frames (mostly proposed by the patient) constitute talk and interaction. These two contexts have different sequential rules, different constraints on what talk should be about, and different underlying assumptions about who controls the ongoing activity. While the institutional framings are asymmetrical, a more balanced, conversational interaction is attempted in the personal framings. By proposing and accomplishing personal reframings, the patient displays her social identities and regains control over them. Here a woman-to-woman relationship fleetingly emerges in the foreground with the roles of patient and doctor in the background. Such shifts have implications for the larger context of the interview, altering the initial distance between the doctor and the patient. Most of all, these changes represent a gradual transition from patient to person, with each personal reframing staging an expansion of self. Erickson (1999) examines a case presentation in which a medical intern reports on a patient’s symptoms and medical history to an experienced physician — the sort of interaction through which novices learn how to engage in collegial discussion. These two doctors establish a footing of collegiality by following one another’s lead about when to switch into an informal register and when to use items from the specialized, formal register of medicine. However, at certain uncomfortable moments during the conversation, the intern conspicuously avoids following the preceptor’s implicit invitation to speak informally. The older doctor is white, the intern an upper-middle-class African American, and the patient about whom they are speaking a lower-class African American. By declining to speak informally about drug use, the young doctor may well be demonstrating (rather than saying outright) that he (unlike a stereotypical lower-class black man) has no firsthand experience of drug use. For his part, the older doctor seems to be offering collegiality, not attributing a stigmatized identity to the intern. The interactants’ footing is disrupted, thus, by problematic assumptions about co-membership (Erickson & Shultz 1982). A final example from medical discourse: Bredmar & Linell (1999), like Tannen & Wallat ([1987] 1993), examine the complexity of framing in encounters between medical personnel and lay people: midwives and expectant mothers. A midwife must balance two tasks, a conversational one and a medical one, and hence two interpretive frames: she must reassure her patient while checking the progress of the pregnancy and taking action if she notes a problem. The midwife strives simultaneously to affiliate with the patient and allay her fears but also to act as a medical professional with the authority to direct the encounter. The tension between the two frames is evident at

85

86

Branca Telles Ribeiro & Susan M. Hoyle

points when a midwife notices a problematic symptom; her speech, for instance, may be filled with hesitations, pauses, or other hitches. Analyses of data from a variety of institutional settings, then, reveal both the specialized and often conflictual nature of interaction in those settings and the wide applicability of frame analysis.

9.  Perspectives for future research Our discussion has tried to capture major theoretical developments and applications to data. Future research may result in refining methodology and analytic tools. For example, researchers might investigate how analytic tools such as footing (Goffman 1981), voice (Bakhtin 1981), positioning (Harré & van Langenhove 1991), and identity (Mishler 1999; Sarbin 2000) are related. These concepts are well established in the social sciences and capture the transient local shifts in everyday talk. However, each may shed a slightly different light on our ‘performed social attributes’ (Erickson & Shultz 1982). Another area for future research may be child language acquisition. Through their framing practices, children construct themselves as particular kinds of social actors. Young children of many speech communities are introduced to the framing of everyday life by participating in teasing (Eisenberg 1986; Miller 1986; Schieffelin 1986). By school age, children are adept at manipulating framing and constructing complex embedded participation frameworks — for instance, embedding stories within arguments or gossip activities (M. Goodwin 1990), embedding imaginary characters and activities within a literal situation (Hoyle 1993), or slipping easily between serious and playful opposition (M. Goodwin 1998). Bilingual children of school age additionally can frame a switch between playful dramatization and negotiation of the play by code-switching (Halmari & Smith 1994). But there has been little systematic investigation of the development of the framing capacity from the youngest children up until the school years. Studies of interaction in computer-mediated communication (Cherny 1999; Davis & Brewer 1997; Herring 1996) may offer new data for comparative analysis of framing in oral and written communication. Computer-mediated communication, such as email, chat, and electronic discussion lists, is a genre that shares some characteristics, such as spontaneity and informality, with face-to-face oral discourse, but on the other hand is one in which participants are separated in space and (often) time, as in traditional written communication. Interaction using the computer medium can result in cohesive communities (Baym 1995; Hamilton 1998; Paolillo 1999) in which individuality and solidarity are constructed. Future investigation of the stances taken by members toward one another and toward the content of their messages would, it seems, benefit from a frame analysis.



Frame analysis

References Adger, C.T. (1998). Register shifting with dialect resources in instructional discourse. In S. Hoyle & C. Adger (eds.): 151–169. Antaki, C., F. Diaz & A.F. Collins (1996). Keeping your footing: Conversational completion in three-part sequences. Journal of Pragmatics 25: 151–71. Atkinson, J.M. (1992). Displaying neutrality: Formal aspects of informal court proceedings. In P. Drew & J. Heritage (eds.): 199–211. Atkinson, J.M. & J. Heritage (eds.) (1984). Structures of social action: Studies in conversation analysis. Cambridge University Press. Bakhtin, M.M. (1981). The dialogic imagination. University of Texas Press. Bartlett, F.C. (1932). Remembering. Cambridge University Press. Bateson, G. ([1954] 1972). A theory of play and fantasy. In Steps to an ecology of mind: 117–193. Ballantine. Baym, N. (1995). The performance of humor in computer-mediated communication. Journal of Computer-Mediated Communication 1(2). Available: http://www.ascusc.org/jcmc/vol1/issue2/ baym.html. Bobrow, D.G. & A.M. Collins (1975). Some principles of memory schemata. In Bobrow & Collins (eds.): 131–149. Bobrow, D.G. & D.A. Norman (eds.) (1975). Representation and understanding. Academic Press. Bredmar, M. & P. Linell (1999). Reconfirming normality: The constitution of reassurance in talks between midwives and expectant mothers. In S. Sarangi & C. Roberts (eds.): 237–270. Candlin, S. (1997). Towards excellence in nursing. An analysis of the discourse of nurses and patients in the context of health assessments. Ph.D. Diss. University of Lancaster. Cherny, L. (1999). Conversation and community: Discourse in a social MUD. Cambridge University Press. Cicourel, A.V. (1992). The interpenetration of communicative contexts: Examples from medical encounters. In A. Duranti & C. Goodwin (eds.): 291–310. Clayman, S.E. (1992). Footing in the achievement of neutrality: The case of news-interview discourse. In P. Drew & J. Heritage (eds.): 163–198. Clift, R. (1999). Irony in conversation. Language in Society 28: 523–53. Davis, B.H. & J.P. Brewer. (1997). Electronic discourse: Linguistic individuals in virtual space. State University of New York Press. Diaz, F., C. Antaki & A.F. Collins (1996). Using completion to formulate a statement collectively. Journal of Pragmatics 26: 525–542. Dorr-Bremme, D.W. (1990). Contextualization cues in the classroom: Discourse regulation and social control functions. Language in Society 19: 379–402. Drew, P. & J. Heritage (eds.) (1992). Talk at work: Interaction in institutional settings. Cambridge University Press. Duranti, A. & C. Goodwin (eds.) (1992). Rethinking context: Language as an interactive phenomenon. Cambridge University Press. Eisenberg, A. (1986). Teasing: Verbal play in two Mexicano homes. In B.B. Schieffelin & E. Ochs (eds.): 182–197. Erickson, F. (1996). Going for the zone: The social and cognitive ecology of teacher-student interaction in classroom conversations. In D. Hicks (Ed.): 29–62. ——— (1999). Appropriation of voice and presentation of self as a fellow physician: Aspects of a discourse of apprenticeship in medicine. In S. Sarangi & C. Roberts (eds.): 109–143.

87

88

Branca Telles Ribeiro & Susan M. Hoyle Erickson, F. & J. Shultz (1981). When is a context? Some issues and methods in the analysis of social competence. In J. Green & C. Wallat (eds.), Ethnography and language in educational settings: 147–160. Ablex. ——— (1982). The counselor as gatekeeper: Social interaction in interviews. Academic Press. Fillmore, C.J. (1976). The need for a frame semantics within linguistics. Statistical methods in linguistics 12: 5–29. Goffman, E. (1974). Frame analysis. Harper & Row. ——— (1981). Footing. In Forms of talk: 124–159. University of Pennsylvania Press. Goodwin, C. (1981). Conversational organization: Interaction between speakers and hearers. Academic Press. ——— (1995). Co-constructing meaning in conversations with an aphasic man. Research on Language and Social Interaction 28: 233–60. Goodwin, C. & A. Duranti (1992). Rethinking context: An introduction. In A. Duranti & C. Goodwin (eds.): 1–42. Goodwin, M.H. (1985). The serious side of jump rope: Conversational practices and social organization in the frame of play. Journal of American Folklore 98: 315–30. ——— (1990). He-said-she-said: Talk as social organization among black children. Indiana University Press. ——— (1995). Co-construction in girls’ hopscotch. Research on Language and Social Interaction 28: 261–81. ——— (1998). Games of stance: Conflict and footing in hopscotch. In S.M. Hoyle & C.T. Adger (eds.): 23–46. Greatbatch, D. (1998). Conversation analysis: Neutralism in British news interviews. In A. Bell & P. Garrett (eds.), Approaches to media discourse: 163–185. Blackwell. Greatbatch, D. & R. Dingwall (1999). Professional neutralism in family mediation. In S. Sarangi & C. Roberts (eds.): 271–292. Gumperz, J.J. (1982). Discourse strategies. Cambridge University Press. ——— (1992). Contextualization and understanding. In A. Duranti & C. Goodwin (eds.): 229–52. Halmari, H. & W. Smith (1994). Code-switching and register shift: Evidence from Finnish-English child bilingual conversation. Journal of Pragmatics 21: 427–45. Hamilton, H.E. (1998). Reported speech and survivor identity in on-line bone marrow transplantation narratives. Journal of Sociolinguistics 2: 53–67. Harré, R. & L. Van Langenhove (1991). Varieties of positioning. Journal for the Theory of Social Behaviour 21: 393–407. Heritage, J. & A.L. Roth (1995). Grammar and institution: Questions and questioning in the broadcast news interview. Research on Language and Social Interaction 28: 1–60. Herring, S.C. (ed.) (1996). Computer-mediated communication: Linguistic, social and cross-cultural perspectives. John Benjamins. Hicks, D. (ed.) (1996). Discourse, learning, and schooling. Cambridge University Press. Holland, D. & N. Quinn (eds.) (1987). Cultural models in language and thought. Cambridge University Press. Hoyle, S.M. (1993). Participation frameworks in sportscasting play: Imaginary and literal footings. In D. Tannen (ed.): 114–45. ——— (1994). Children’s use of discourse markers in the creation of imaginary participation frameworks. Discourse Processes 17: 447–64. ——— (1998). Register and footing in role play. In S.M. Hoyle & C.T. Adger (eds.): 47–67. Hoyle, S.M. & C.T. Adger (eds.) (1998). Kids Talk: Strategic language use in later childhood. Oxford University Press.



Frame analysis

Hutchby, I. (1999). Frame attunement and footing in the organisation of talk radio openings. Journal of Sociolinguistics 3: 41–63. Hydén, L.-C. & E.G. Mishler (1999). Language and Medicine. Annual Review of Applied Linguistics 19: 174–192. Cambridge University Press. Irvine, J.T. (1996). Shadow conversations: The indeterminacy of participant roles. In M. Silverstein & G. Urban (eds.), Natural histories of discourse: 131–59. University of Chicago Press. James, W. (1950). Principles of psychology. Dover Publications. Johnstone, B. (1987). “He says … so I said”: Verb tense alternation and narrative depictions of authority in American English. Linguistics 25: 33–52. Kendon, A. (1990). Conducting interaction: Patterns of behavior in focused encounters. Cambridge University Press. Levinson, S. (1988). Putting linguistics on a proper footing: Explorations in Goffman’s concepts of participation. In P. Drew & A. Wootton (eds.), Erving Goffman: Exploring the interaction order: 161–227. Polity Press. Matoesian, G. (1999). The grammaticalization of participant roles in the constitution of expert identity. Language in Society 28: 491–521. Mccreedy, L. (1998). The effect of role and footing on students’ oral academic language. In S.M. Hoyle & C.T. Adger (eds.): 170–186. Miller, P. (1986). Teasing as language socialization and verbal play in a white working-class community. In B.B. Schieffelin & E. Ochs (eds.): 199–212. Minsky, M. (1975). A framework for representing knowledge. In P.H. Winston (ed.), The psychology of computer vision: 211–277. McGraw-Hill. Mishler, E.G. (1984). The discourse of medicine: Dialectics of medical interviews. Ablex. ——— (1999). Storylines: Craftartists’ narratives of identity. Harvard University Press. Morgan, M. (1996). Conversational signifying: Grammar and indirectness among African American women. In E. Ochs, E.A. Schegloff & S.A. Thompson (eds.), Interaction and grammar: 405–434. Cambridge University Press. Nelson, K. (1986). Event knowledge and cognitive development. In K. Nelson (ed.), Event knowledge: Structure and function in development: 1–19. Erlbaum. Norrick, N.R. (1993). Conversational joking: Humor in everyday talk. Indiana University Press. O’Connor, M.C. & S. Michaels (1996). Shifting participant frameworks: Orchestrating thinking practices in group discussion. In D. Hicks (ed.): 63–103. Paolillo, J. (1999). The virtual speech community: Social network and language variation on IRC [internet relay chat]. Journal of Computer-Mediated Communication 4(4). Available: http://www.ascusc.org/jcmc/vol4/issue4/paolillo.html. Petruck, M.R.L. (1995). Frame semantics and the lexicon. In M. Shibatani & S. Thompson (eds.), Essays in semantics and pragmatics: 279–296. John Benjamins. Rampton, B. (1995). Crossing: Language and ethnicity among adolescents. Longman. Ribeiro, B.T. (1993). Framing in psychotic discourse. In D. Tannen (ed.): 77–113. ——— (1994). Coherence in psychotic discourse. Oxford University Press. ——— (1996). Conflict talk in a psychiatric discharge interview: Struggling between personal and official footings. In C.R. Caldas-Coulthard & M. Coulthard (eds.), Texts and practices: Readings in critical discourse analysis: 179–193. Routledge. Rumelhart, D.E. (1975). Notes on a schema for stories. In D.G. Bobrow & A.M. Collins (eds.): 211–236. ——— (1980). Schemata: The building blocks of cognition. In R.J. Spiro, B.C. Bruce & W.F. Brewer (eds.), Theoretical issues in reading comprehension: Perspectives from cognitive psychology, linguistics, artificial intelligence, and education: 33–58. Erlbaum.

89

90

Branca Telles Ribeiro & Susan M. Hoyle Sarangi, S. & C. Roberts (eds.) (1999). Talk, work and institutional order: Discourse in medical, mediation and management settings. Mouton de Gruyter. Sarbin, T.R. (2000). The Poetics of Identity. Paper presented at meeting From talk to identity: methodological issues and problems, Clark University. Schank, R.C. & R.P. Abelson (1977). Scripts, plans, goals, and understanding: An inquiry into human knowledge structures. Erlbaum. Schegloff, E.A. (1999). What next?: Language and social interaction study at the century’s turn. Research on Language and Social Interaction 32: 141–148. Schieffelin, B.B. (1986). Teasing and shaming in Kaluli children’s interactions. In B.B. Schieffelin & E. Ochs (eds.): 165–181. Schieffelin, B.B. & E. Ochs (eds.) (1986). Language socialization across cultures. Cambridge University Press. Schiffrin, D. (1987). Discourse markers. Cambridge University Press. ——— (1993). “Speaking for another” in sociolinguistic interviews: Alignments, identities, and frames. In D. Tannen (ed.): 231–63. Straehle, C.A. (1993). “Samuel?” “Yes, dear?”: Teasing and conversational rapport. In D. Tannen (ed.): 210–230. Tannen, D. ([1979] 1993). What’s in a frame? Surface evidence for underlying expectations. In D. Tannen (ed.), 14–56. ——— (1996). The sex-class linked framing of talk at work. In D. Tannen (ed.), Gender and discourse: 195–221. Oxford University Press. Tannen, D. (ed.) (1993). Framing in Discourse. Oxford University Press. Tannen, D. & C. Wallat ([1987] 1993). Interactive frames and knowledge schemas in interaction: Examples from a medical examination/interview. In D. Tannen (ed.): 57–76. Van Dijk, T.A. (1980). Macrostructures: An interdisciplinary study of global structures in discourse, interaction, and cognition. Erlbaum. Zentella, A.C. (1997). Growing up bilingual: Puerto Rican children in New York. Blackwell.

Functional Discourse Grammar: pragmatic aspects Mike Hannay & Kees Hengeveld Vrije Universiteit Amsterdam/University of Amsterdam

1.  Introduction This chapter introduces Functional Discourse Grammar,1 focusing on the way in which this model is capable of accounting for the grammatical encoding of pragmatic distinctions and for the typological variation found in this area of grammar. An outline of the model is given in Section 2, which shows, among other things, that Discourse Acts rather than sentences are taken as the basic units of analysis, and that the grammar is allowed to interact with a specification of the context in which it is used. Within the grammar there are four levels of analysis, one of which is pragmatic in nature. Section  3 presents this level of analysis in more detail. Sections 4 and 5 then present the way in which the model deals with several phenomena that are of interest to pragmaticians. Section 4 is dedicated to a number of aspects related to Discourse Acts, such as the encoding of illocution and of rhetorical relations, while Section 5 deals with aspects related to ascription and reference, such as the encoding of the straightforwardness of ascription, the specificity of reference, and the encoding of information structure. The chapter ends with a conclusion in Section 6. 2.  Outline of the model2 2.1  FDG and verbal interaction As shown in Figure 1, FDG is conceived of as the Grammatical Component of an overall model of verbal interaction in which it is linked to a Conceptual Component, an Output Component and a Contextual Component. These three non-grammatical components interact in various ways with the Grammatical Component, more specifically with the operations of Formulation and Encoding. Formulation concerns the rules that determine what constitute valid underlying pragmatic and semantic representations in

1.  A full description of FDG may be found in Hengeveld & Mackenzie (2008). Technical terms specific to FDG are presented in title case throughout the text. 2.  This section is largely based on Hengeveld & Mackenzie (forthc.).

Mike Hannay & Kees Hengeveld

a language. Encoding concerns the rules that convert these pragmatic and semantic representations into morphosyntactic and phonological ones. Conceptual Component

Grammatical Component

92

Formulation Contextual Component

Encoding

Output Component Figure 1.  FDG as part of a wider theory of verbal interaction

The Conceptual Component is responsible for the development of both a communicative intention relevant for the current speech event and the associated conceptualizations with respect to relevant extra-linguistic events, and is thus the driving force behind the Grammatical Component as a whole. The Output Component generates acoustic or signed expressions on the basis of information provided by the Grammatical Component. The Contextual Component contains a description of the content and form of the preceding discourse, of the actual perceivable setting in which the speech event takes place, and of the social relationships between Participants. This type of information is relevant to many grammatical processes, such as narrative chaining, reflexives, and passives. 2.2  The architecture of FDG The general architecture of FDG itself, in relation to the components that flank it, may now be represented as in Figure 2, in which the Grammatical Component is presented in the centre, the Conceptual Component at the top, the Output Component at the bottom, and the Contextual Component to the right. A distinguishing feature of FDG shown in Figure 2 is its rigorous top-down architecture: FDG starts with the speaker’s intention and works down to articulation. This is motivated by the assumption that a model of grammar will be more effective the more its organization resembles language processing in the individual. Psycholinguistic studies (e.g. Levelt 1989) clearly show that language production is indeed a top-down



Functional Discourse Grammar: pragmatic aspects

Conceptual Component

Frames Lexemes Interpersonal and Representational Operators

Formulation

Interpersonal Level

Templates Grammatical morphemes Morphosyntactic Operators

Templates Suppletive forms Phonological Operators

Morphosyntactic Encoding

Morphosyntactic Level

Phonological Encoding

Contextual Component

Grammatical Component

Representational Level

Phonological Level

Articulation Output Component

Output

Figure 2.  General layout of FDG

process. The implementation of FDG reflects this process and is organized accordingly. This does not mean, however, that FDG is a model of the speaker: FDG is a theory about grammar, but one that tries to reflect psycholinguistic evidence in its basic architecture. The top-down organization of the grammar has far-reaching consequences at all levels of analysis, as we will show in Section 4. In Figure 2 ovals contain operations, boxes contain the primitives used in operations, and rectangles contain the levels of representation produced by operations. We will discuss all of these in more detail in Section 4, and limit ourselves here to describing the general top-down process on the basis of a simple example, given in (1), produced in a context in which the Addressee wants to enter a field that hosts a bull:

(1) There’s a bull in the field!

93

94

Mike Hannay & Kees Hengeveld

In the prelinguistic Conceptual Component, what is relevant is a communicative intention (issuing a warning) and the corresponding mental representation (of the event causing danger). The operation of Formulation translates these conceptual units into pragmatic and semantic representations at the Interpersonal and the Representational Levels, respectively. Warnings are not a separate illocutionary category in English, but the Speaker solves this problem by selecting a Declarative Illocution combined with an Emphatic operator at the Interpersonal Level. The entity causing danger is furthermore characterized as a Focal Topic at this Level. At the Representational Level the Speaker chooses to designate the entity causing danger as part of a locative predication frame. The configurations at the Interpersonal and Representational Levels are translated into a morphosyntactic structure at the Morphosyntactic Level through the operation of Morphosyntactic Encoding. In (1) this involves, for instance, the word order characteristic of existentials, the insertion of dummy there, etc. Similarly, the structures at the Interpersonal, Representational, and Morphosyntactic Levels are translated into a phonological structure at the Phonological Level. In this case, for instance, the selection of the declarative illocution combined with an emphatic operator is responsible for the overall intonation contour with a high fall on the Focal Topic bull. By organizing the Grammatical Component in the way illustrated here, FDG takes the functional approach to language to its logical extreme: within the top-down organization of the grammar, pragmatics governs semantics, pragmatics and semantics govern morphosyntax, and pragmatics, semantics, and morphosyntax govern phonology. The Phonological Level of representation is the input to the operation of Articulation, which contains the phonetic rules necessary for an adequate utterance. Articulation takes place in the Output Component, outside the grammar proper. The various levels of representation within the grammar feed into the Contextual Component, thus enabling subsequent reference to the various kinds of entity relevant at each of these levels once they are introduced into the discourse. The Contextual Component feeds into the operations of Formulation and Encoding, so that, for instance, the availability of antecedents may influence the composition of (subsequent) Discourse Acts. 2.3  Levels and layers Each of the levels of representation distinguished within the Grammatical Component in Figure 2 is structured in its own way. What all the levels have in common is that they have a hierarchically ordered layered organization. In its maximal form the general structure of layers within levels is as follows:

(2) (π V1: [head (V1)]: [Σ (V1)])Φ

Here V1 represents the variable of the relevant layer, which is restricted by a (possibly complex) head that takes the variable as its argument, and may be further restricted by a



Functional Discourse Grammar: pragmatic aspects

modifier Σ that also takes the variable as its argument. The layer may be specified by an operator π and carry a function Φ. Heads and modifiers represent lexical strategies, while operators and functions represent grammatical strategies. The difference between operators and functions is that the latter are relational, holding between the entire unit and other units at the same layer, while the former are not, applying only to the unit itself. Not all relations between units are hierarchical. In those cases in which units together form a non-hierarchical (equipollent) configuration, they are enclosed between square brackets, as exemplified in (2). The levels differ as regards the nature of the distinctions that are relevant to each, but what they all have in common is that only those distinctions are provided that are actually reflected in the grammar of the language involved. This also applies to the Interpersonal Level, to which we will direct our attention now. At this level, too, only those distinctions are represented that have a grammatical (morphosyntactic, prosodic) impact. FDG is therefore not to be seen as a discourse-analytical model, but as a tool for grammatical analysis sensitive to the communicative functions of language. 3.  The interpersonal level The Interpersonal Level captures all distinctions of Formulation that pertain to the interaction between Speaker and Addressee. These cover, at the higher layers, rhetorical notions of the overall structuring of discourse, to the extent that they are reflected in linguistic form, and at the lower layers, the pragmatic distinctions that reflect how Speakers mould their messages taking account of their assumptions about the Addressee’s state of knowledge. The hierarchical structure arises through the application of an appropriate set of frames from those available to the Speaker. The following shows the hierarchical relationships that apply at the Interpersonal Level, with each layer exhibiting the structure given in (2): (3) (π M1: [   (π A1:  [ (π F1: ILL (F1): Σ (F1)) (π P1: ... (P1): Σ (P1))S (π P2: ... (P2): Σ (P2))A (π C1:  [ (π T1: [...] (T1): Σ (T1))Φ (π R1: [...] (R1): Σ (R1))Φ ] (C1): Σ (C1))Φ ] (A1): Σ (A1))Φ ] (M1): Σ (M1))

Move Discourse Act Illocution Speaker Addressee Communicated Content Ascriptive Subact Referential Subact Communicated Content Discourse Act Move

The Move (M1) is the largest unit of interaction relevant to grammatical analysis. It may be defined as an autonomous contribution to the ongoing interaction: it either

95

96

Mike Hannay & Kees Hengeveld

calls for a reaction, or is itself a reaction. A Move contains one or more Discourse Acts (A1), which may be defined, following Kroon (1995: 65), as “the smallest identifiable units of communicative behaviour”. Kroon (1995: 65) notes further that Discourse Acts, in contrast to Moves, “do not necessarily further the communication in terms of approaching a conversational goal”. A Discourse Act is characterized by the fact that it has an Illocution (F1), which captures the lexical and formal properties of that Discourse Act that can be attributed to its conventionalized interpersonal use in achieving a communicative intention. An Illocution specifies a relation between a Speaker (P1)S, an Addressee (P2)A, and a Communicated Content (C1), which contains the totality of what the Speaker wishes to evoke in his/her communication with the Addressee. In actional terms it essentially comprises what Searle (1969) calls the “representational act” and corresponds to the choices the Speaker makes in order to evoke a picture of the external world he/she wants to talk about. In building up this picture the Speaker typically executes one or more Subacts of Reference (R1) and of Ascription (T1). An Ascriptive Subact is an attempt by the Speaker to evoke an attribute, while a Referential Subact is an attempt by the Speaker to evoke a referent.

4.  Discourse Acts and the relations between them 4.1  Introduction Discourse Acts are fundamental units of analysis in Functional Discourse Grammar. In the top-down approach advocated in this model Discourse Acts may map onto various kinds of morphosyntactic unit: clauses, phrases, or single words. Everything else being equal, a Speaker need not express more of his/her communicative intention than he/she thinks is required to understand it. This is what we mean when we say that FDG is a discourse grammar rather than a sentence grammar. The structure of a Discourse Act, disregarding the internal structure of its component parts, is given in (4):

(4) (π A1: head (A1): Σ (A1))Φ

This representation shows that a Discourse Act may be modified by operators (π), such as the Irony operator, and modifiers (Σ), such as style adverbs like briefly. Discourse Acts may also carry a Function (Φ) in relation to other Discourse Acts, which is what we will concentrate on in Section 4.2. The head position of a discourse act is filled by a frame that maximally contains positions for the Speaker (PS), the Addressee (PA), and the Communicated Content (C1). Not all of these elements need to be present. Compare the following examples:





Functional Discourse Grammar: pragmatic aspects

(5) Ouch! (6) Congratulations (on winning the race). (7) I am hereby informing you that I wish to resign.

An Expressive Discourse Act such as (5) does not require the presence of an Addressee and does not contain a Communicated Content either. An Interactive Discourse Act such as (6) requires the presence of a Speaker and an Addressee, and optionally contains a Communicated Content. Contentive Discourse Acts, such as (7), require the presence of a Speaker, an Addressee, and a Communicated Content. These three types of Discourse Act may be represented as in (8): (8) a. (A1: [(F1) (P1)S] (A1)) b. (A1: [(F1) (P1)S (P2)A {(C1)Φ}] (A1)) c. (A1: [(F1) (P1)S (P2)A (C1)Φ] (A1))

Expressive Discourse Acts Interactive Discourse Acts Contentive Discourse Acts

A defining feature of Discourse Acts is that they have an Illocution (F1). We will go into the treatment of illocutionary values in FDG in Section 4.3. 4.2  Rhetorical functions A Move may consist of a number of distinct Discourse Acts. Typically, one Discourse Act is communicatively central, and thus has a nuclear status. Other Discourse Acts either have the same communicative status as the Nuclear Discourse Act, in which case the Move will contain multiple nuclei, or otherwise they enter into a dependency relation with either the nucleus or another subsidiary Discourse Act. The nature of the dependency may be specified by the assignment of a Rhetorical Function. Consider the following move, which consists of two Discourse Acts:

(9) Watch out, because there will be trick questions in the exam.

Here, the Discourse Act performed by means of the because clause does not provide a reason for watching out; rather, it is to be seen as a motivation for the Speaker’s decision to issue the warning.3 Accordingly, (9) may be analyzed as in (10), with the Rhetorical Function Motivation being assigned to the Subsidiary Discourse Act:4 (10) (MI: [(AI: –watch out– (AI)) (AJ: –there will be trick questions in the exam– (AJ))Motiv] (MI))

3.  Note that the Function Motivation is defined in FDG in speaker-related terms, in contrast to Rhetorical Structure Theory, where it is addressee-related (cf. Mann et al. 1992: 70). 4.  This is a representational convention; the rhetorical relation actually holds between the two acts (cf. Hengeveld & Mackenzie 2008: 306).

97

98

Mike Hannay & Kees Hengeveld

The choice of subordinate clause and conjunction in (9) is directly determined by the Motivation relation. Now compare (9) with (11): (11) There will be trick questions in the exam, so watch out.

Even though the first act in (11) is realized as an independent main clause, it is again the imperative clause which realizes the Nuclear Discourse Act, just like in (9). In this case the Nuclear Discourse Act is introduced by so, which indicates its nuclear status, and (11) will accordingly be analyzed as in (12): (12) (MI: [(AI: –there will be trick questions in the exam– (AI))Motiv (AJ: –watch out– (AJ))] (MI))

The dependency relations in (9) and (11) are thus identical, but the alternation in form results from the ordering of the two Acts. The relative order of Discourse Acts within a Move is seen in FDG as an outcome of the Formulation process rather than of the Encoding process. On the basis of contextual considerations and communicative goals, the Speaker adopts a particular communicative strategy for performing the intended Move and selects a frame in which Discourse Acts are combined and, when they are dependent, carry a Rhetorical Function. These frames may contain Subsidiary Acts which either precede, interrupt, or follow the Nuclear Discourse Act. Pre-nuclear Discourse Acts are orientational in nature, while interruptive or post-nuclear Discourse Acts are in some sense elaborative. Thus, Rhetorical Functions do not trigger a certain order, since Discourse Acts are already ordered within their frame. This is in contrast with the workings of Pragmatic Functions relevant in the Communicative Content layer: in particular languages and under certain circumstances, Pragmatic Functions may trigger the placement of a constituent in a particular position in the clause (see Section 5.2 below). In order to account for the Speaker’s strategic choices within the process of Formulation, more would have to be known about what specific features of the output of the Conceptual Component have an effect on the Formulation process, and how. This would allow a degree of formalization of the Formulation process, leading to a further enrichment of FDG as a discourse-sensitive grammar. Another feature of the system of Rhetorical Function assignment worth noting is that, as currently formulated, only those dependency relationships between Discourse Acts that have an effect on their formal manifestation are captured by means of Rhetorical Functions (Hengeveld & Mackenzie 2008: 53). This situation may be contrasted with that in for instance Rhetorical Structure Theory (cf. Mann 2005; Taboada & Mann 2006), which recognizes a set of multinuclear relations such as ‘list’ and ‘neutral contrast’ that are normally not formally expressed. The major difference between the two theories lies, of course, in the status of the rhetorical relations. Because RST aims to explain the coherence of texts, each unit of text is assumed to be related by some specific



Functional Discourse Grammar: pragmatic aspects

rhetorical relation, which is the analyst’s attempt to reconstruct the author’s rhetorical intention. FDG, by contrast, only makes use of Rhetorical Functions inasmuch as they can account for specific features of linguistic expressions. Accordingly, if there were languages where equipollent rhetorical relations such as ‘list’ were linguistically coded, it would be necessary to define such Rhetorical Functions for the grammar of the languages in question. So far we have looked at Subsidiary Discourse Acts which are dependent on other complete Discourse Acts. However, Subsidiary Discourse Acts may stand in a relationship with part of another Discourse Act. This is the case with for instance left and rightdislocation constructions. Consider (13) and (14): (13) John, he’s a real wizard. (14) He’s a real wizard, John.

In (13), the left-dislocated element John, which constitutes a Subsidiary Discourse Act that would typically be expressed as a separate intonation and punctuation unit, serves to provide the Addressee with an idea of one of the referents that the Speaker wishes to talk about within the ensuing Nuclear Discourse Act. Similarly in (6), John can be seen as expressing a distinct Subsidiary Discourse Act, but in this case serves to clarify an aspect of the Nuclear Discourse Act. Whereas in Functional Grammar (Dik 1997: 310) the relation between dislocation constructions and their host clause was seen as a matter of Pragmatic Function, FDG analyzes them as involving Rhetorical Functions, as they are typically realized as independent intonation units and can carry their own illocutionary force. (13) and (14) are analyzed as (15) and (16), respectively, involving the Rhetorical Functions of Orientation and Clarification (Hengeveld & Mackenzie 2008: 55): (15) (MI: [(AI: –John– (AI))Orient (AJ: –he’s a real wizard– (AJ))] (MI)) (16) (MI: [(AI: –he’s a real wizard– (AI)) (AJ: –John– (AJ))Clar ] (MI))

Although such constructions typically involve a noun phrase, a similar function can be performed by a prepositional phrase functioning as a framing adverbial, as in (17): (17) In the animal kingdom, moths usually have many pairs of wings?

This is an example of a set question format on a television quiz show. The expression In the animal kingdom does not serve to identify the domain in which moths have a certain number of wings; rather, it is an instruction to the contestant to activate a specific area of his/her store of knowledge, with the aim of helping the contestant to answer quickly. Consequently, here too the sentence is best analyzed as consisting of two Discourse Acts, an Orientational Discourse Act followed by the Nuclear Discourse Act. As noted in Section 3, by distinguishing between the Interpersonal and Representational Levels, and by distinguishing at the Interpersonal Level between both

99

100 Mike Hannay & Kees Hengeveld

rhetorical and information-structure relations, FDG can account in an insightful way for features of linguistic expressions which pertain to the Speaker’s management of the interaction with the Addressee and offers a powerful tool for the analysis of these features. This analysis of structural phenomena in terms of rhetorical relations rather than semantic or information-structure relations is by no means an easy task and requires extensive knowledge of the language concerned. We give two examples of the complexity of the task facing the investigator. The first involves initial subordinate clauses, which in spoken and written English are often presented as separate intonation and punctuation units, respectively. However, their relative independence does not necessarily mean that they will typically constitute a separate Discourse Act. Hengeveld & Mackenzie (2008: 54–55) discuss the case of concessive clauses, which like because clauses can entertain different kinds of semantic and rhetorical relation with their hosts. Concessive clauses like that in (18), represented in (19), may indeed be seen to constitute a separate Discourse Act, since the optional addition of a performative predicate shows that such clauses may have illocutionary force: (18) The work was fairly easy, although (I concede that) it took me longer than expected. (19) (MI: [(AI: –the work was fairly easy– (AI)) (AJ: –it took me longer than expected– (AJ))Conc] (MI))

Hengeveld & Mackenzie (2008: 55) then argue that the situation is different when the subordinate clause precedes the host clause, as in (20), which they assume constitutes one Discourse Act, with the relation between the subordinate clause and the main clause being captured at the Representational Level by a Semantic Function Concession, rather than by a Rhetorical Function at the Interpersonal Level: (20) Although it took me longer than expected, the work was fairly easy.

It is indeed generally assumed that subordination involves a neutralizing of illocutionary force, and this is most strongly present where subordinate clauses occur in an orientational context rather than an elaborational one (cf. Matthiessen & Thompson 1988: 315). However, examples can also be found in which the initial concession clauses can indeed contain a predicate functioning performatively, as witnessed by the following Internet examples: (21) While I will concede that these thoughts are somewhat simplistic, I firmly believe that they do provide a more relevant philosophical framework for the development of future generations of Agriculture and Agri-Food Policy. (22) David, although I happily admit that I work for the Vienna Symphonic Library as a freelancing developer and engineer (I’m not employed by them), I strongly suggest to ask those questions in the VSL’s own forum.

In these examples, the use of will in (21) and happily in (22) clearly indicates the performative use of the verbs concerned. This suggests that even initial subordinate clauses may realize separate Discourse Acts; accordingly, (20) may be analyzed as follows:



Functional Discourse Grammar: pragmatic aspects

(23) (MI: [(AI: –it took me longer than expected– (AI))Conc (AJ: –the work was fairly easy– (AJ))] (MI))

A second example of the complex nature of an analysis in terms of Rhetorical Functions relates to the interpretation of certain phenomena which might be seen as either pragmatic – relating to the informational status of elements within the Discourse Act, at the level of the Communicated Content – or rhetorical, and hence relating to the relations between Discourse Acts. Dik (1997b: 85ff) discusses what he calls “orientation + clause” constructions, arguing in the case of (24), from Imbabura Quechua (Cole 1982), that the suffix -ka may be interpreted as marking a clause-external Orientation relation rather than a clause-internal Topic function: (24)

Ñuka chay punlla-pi chaya-shka-ka sumaj-mi ka-rka. I that day-on arrive-nmlz-top beautiful-val be-pst ‘Given my arriving on that day, it was beautiful.’ →‘The day on which I arrived was beautiful.’

Hengeveld & Mackenzie (2008: 57) counter this analysis, pointing out that it is clear that only one Discourse Act is involved. Consequently, the -ka suffix is reanalyzed as a marker of the pragmatic status of ñuka chay punllapi chayashka ‘my arriving that day’, rather than as a marker of its rhetorical status. The problematic nature of such cases lies first of all in the closeness of the conceptual relationship between Topichood on the one hand, and Orientation on the other. The function of the Orientational Act is to “orient the Addressee to the Speaker’s communicative intentions” (Hengeveld & Mackenzie 2008: 55), while the function of a Topic is to relate the Communicated Content to existing information in the Contextual Component (Hengeveld & Mackenzie 2008: 92). In other words, both have a grounding function: ‘this is what the utterance is about’. When the orientational element is not only intonationally distinct but also nominal in form, it is attractive to analyze such an element as constituting a separate Discourse Act, but when such an element is propositional, there are competing analyses: a Propositioncombining and an Act-combining analysis. Another complicating factor may be the way the structures of certain languages are described by linguists. In this regard, de Vries (2006) argues that some so-called Topic markers in Papuan languages may in fact function as discourse markers, which in FDG terms would realize a Rhetorical rather than a Pragmatic Function. He cites the following example of clause chaining in Korowai (2006: 826): (25)

Wa-gol ülme-tél-e-kha-fè nokhu-gol. that-pig kill-3.pl.real-tr-sub-top our-pig ‘Given that they killed the pig, it is our pig.’ or ‘As for killing the pig, it is our pig.’

101

102 Mike Hannay & Kees Hengeveld

The topic analysis for the clitic fè in such cases is based on the description that the information provides the relevant discourse domain for what the next part of the utterance is about. However, de Vries analyzes this use in the context of the strategy of thematization which is characteristic of the narrative discourse of Papuan speech communities. He argues that translations of such constructions involving for instance relative clauses (i.e. ‘the pig that they killed is our pig’) promote a Topic analysis but do not do sufficient justice to the thematization strategy. Indeed, such translations may distort the grammatical description. By contrast, the alternative translation (‘as for killing the pig, it is our pig’) reflects the discourse strategy employed by the Speaker. All in all, the specification of Rhetorical Functions provides an insightful means of accounting for the formal expression of the relations between Discourse Acts. However, a full understanding of which formal features of linguistic expressions can best be accounted for in rhetorical rather than semantic or information-structural terms requires in-depth analysis of intonational and illocutionary phenomena, while at the same time taking heed of language-specific discourse practices. 4.3  Illocution As stated above, Discourse Acts are characterized by the fact that they have an Illocution. The Illocution of a Discourse Act captures the (lexical and grammatical) properties of that Discourse Act that can be attributed to its conventionalized interpersonal use in achieving a communicative intention. Communicative intentions include such things as calling for attention, asserting, ordering, questioning, warning, requesting, etc., which may map onto Illocutions such as Vocative, Declarative, Imperative, etc. There is no one-to-one relation between a specific communicative intention and an Illocution, neither within a language, nor cross-linguistically, as languages may differ significantly in the extent to which they make use of linguistic means to differentiate between communicative intentions. The Illocution is represented by means of the variable F. Its maximal representation is given in (26): (26) (π F1: head (F1): Σ (F1))

The Illocution may have various types of head and may be specified by operators (π), such as the Reinforcing and Mitigating operators, and modifiers (Σ), such as illocutionary adverbs like frankly. The head position of the Illocution may be filled by lexical material or may be abstract. Examples (27a) and (27b) differ in this respect: (27) a. I am hereby informing you that I wish to resign. b. I wish to resign.



Functional Discourse Grammar: pragmatic aspects

The head of their Illocution would be represented as in (28a) and (28b), respectively: (28) a. (FI: informV (FI)) b. (FJ: decl (FJ))

In (28a) the performative verb inform is inserted directly into the F-slot, giving rise to the performative use of this verb. In (28b) the abstract illocutionary predicate decl is inserted into the F-slot, triggering the prosodic and morphosyntactic devices that characterize the utterance as conventionally associated with assertions. Languages may differ widely from one another in the range of lexical elements that can be used performatively, as in (28a), and in the range of abstract illocutions that can be distinguished on the basis of distinct prosodic and morphosyntactic behaviour, as in (28b). A study of cross-linguistic variation in the lexical domain is Verschueren (1989), which studies the occurrence and distribution of “Basic Linguistic Action Verbs” (BLAVs) in a stratified sample of 81 languages. The study shows that there are languages that have just one BLAV, meaning something like ‘to use language’, ‘to express linguistically’ (Verschueren 1989: 31), while other more specific meanings are expressed by non-basic verbs or constructions. The highest number of BLAVs attested is twelve, including such specific meanings as ‘greet’, ‘answer’, and ‘name’. The intermediate types between 1 and 12 BLAVs show interesting patterns, a very early split being the one between saying and asking a question, which normally precedes the occurrence of BLAVs denoting Interactive Discourse Acts such as greeting and thanking. As regards abstract Illocutions, the following have been observed in one or more languages: Declarative (decl): the Speaker informs the Addressee of the Propositional Content evoked by the Communicated Content Interrogative (inter): the Speaker requests the Addressee’s response to the Propositional Content evoked by the Communicated Content Mirative (mir): the Speaker expresses his/her surprise about the Propositional Content evoked by the Communicated Content Optative (opt): the Speaker indicates to the Addressee his/her wish that the positive situation evoked by the Communicated Content should come about Imprecative (impr): the Speaker indicates to the Addressee his/her wish that the negative situation evoked by the Communicated Content should come about Imperative (imper): the Speaker directs the Addressee to carry out the action evoked by the Communicated Content Prohibitive (proh): the Speaker forbids the Addressee to carry out the action evoked by the Communicated Content

103

104 Mike Hannay & Kees Hengeveld

Hortative (hort): the Speaker encourages him/herself or an Addressee together with him/herself to carry out the action evoked by the Communicated Content Dishortative (dishort): the Speaker discourages him/herself or an Addressee together with him/herself from carrying out the action evoked by the Communicated Content Admonitive (admon): the Speaker warns the Addressee against the situation evoked by the Communicated Content Cautionary (caut): the Speaker warns the speaker to avoid the situation evoked by the Communicated Content Commissive (comm): the Speaker commits him/herself to the future realization of the positive situation evoked by the Communicated Content Minative (min): the Speaker commits him/herself to the future realization of the negative situation evoked by the Communicated Content Permissive (perm): the Speaker grants permission to the Addressee to realize the situation evoked by the Communicated Content Supplicative (suppl): the Speaker asks permission of the Addressee to realize the situation evoked by the Communicated Content One study of cross-linguistic variation in the availability of abstract Illocutions is Hengeveld et al. (2007), in which it is shown that a primary split in all languages is the one between propositional Illocutions (those that have to do with exchanging information, such as decl, inter, and mir) on the one hand, and behavioural Illocutions (those that have to do with influencing behaviour, such as imper, hort, and admon) on the other. This means that there are languages with a single propositional Illocution, in which there is no formal distinction between constructions used for assertions and questions. Such a language is, for instance, Acoma. In this language, the way to ask questions is to state ignorance or doubt. An example is (29) (Miller 1965: 123): (29)

C-aityata dyâiyaani. 2.dub-gather piñons ‘I guess you are gathering piñons.’ ‘Are you gathering piñons?’

On the other hand, there are languages with a single behavioural Illocution, most often called Imperative, though also serving to execute other kinds of speech act, such as warnings and curses. It is more common, however, to have a variety of behavioural Illocutions, the most common ones being imper, hort, and proh, as in the following examples from Macushi (Abbott 1991: 50):



Functional Discourse Grammar: pragmatic aspects

(30) Apo’ era’ma-ta. fire get-imper ‘Go get firewood.’ (31) Epîrema-n-pai’-nîkon. pray-1.incl-hort-coll ‘Let’s all pray.’ (32) K-ena-i. proh-fall-proh ‘Don’t fall.’

Apart from performative verbs, languages may use specific particles to fill the illocutionary slot, which may then give expression to highly specific interpersonal meanings. Examples from English are particles such as hey (drawing attention), ouch (expressing pain), hello (greeting), and many more. Two components in the overall organization of FDG (see Figure 2) are fundamental in accounting for the differences between languages as regards the availability of lexical and abstract Illocutions: (i) the Lexicon, more specifically the set of (Interpersonal) Primitives within it; (ii) the operation of Formulation. Both lexical and abstract Illocutions are stored in the Lexicon contained within the grammar. The operation of Formulation then determines how communicative intentions, developed in the Conceptual Component by the Speaker of a certain language, map onto the various illocutionary expressions available in that language.

5.  Subacts and the relations between them 5.1  Introduction Several authors have emphasized that reference is actional in nature. Thus, Lyons (1977: 177) states that “... the speaker ... invests the expression with reference by the act of referring”. In FDG referring is similarly first and foremost interpreted as an act of the Speaker, but ascription is considered to be actional in the same way as well. In exactly the same way as with referring, the choice of lexical material, and the amount offered, are the result of decisions of the Speaker. Hence Ascriptive Subacts (T) and Referential Subacts (R) are the basic units on the basis of which Communicated Contents at the Interpersonal Level are organized. This approach is supported by the fact that a single semantic unit may be used referentially, as in (33), or ascriptively, as in (34): (33) A teacher came in. (34) He is a teacher.

105

106 Mike Hannay & Kees Hengeveld

The Speaker chooses configurations of Subacts partially in view of his/her estimate of the state of knowledge of the Addressee, and may select particular strategies to adapt his/her message to make it more suitable to be integrated into this knowledge. These information-structural strategies are dealt with in Section 5.2. Individual Subacts may also be characterized in relation to the state of knowledge of Speaker and Addressee, and with respect to their adequacy in the light of the Speaker’s communicative aims, a topic we will address in Section 5.3. 5.2  Pragmatic functions The Communicated Content, as “the totality of what the Speaker wishes to evoke in his/her communication with the Addressee” (Hengeveld & Mackenzie 2008: 87), comprises one or more Subacts of Reference or Ascription. Each Communicated Content is shaped according to the Speaker’s communicative aims and in particular to his/her assessment of the Addressee’s communicative needs in the given setting. FDG captures this by assuming that for any given language a speaker chooses, in the process of Formulation, a specific configuration of Subacts called a Content Frame. Each Content Frame is characterized by a combination of Referential and Ascriptive Subacts, one or more of which may carry a Pragmatic Function. The Speaker’s adoption of a particular Content Frame in the process of Formulation lays out a blueprint for the communicative shape of a Discourse Act. A Subact carrying a Pragmatic Function represents an individual component of the Speaker’s adopted communicative strategy which he/she wishes to pay special attention to in order to help the Addressee interpret the informational status of that Subact in the given communicative setting. FDG recognizes Pragmatic Functions along three different dimensions: Focus – Background Topic – Comment Contrast – Overlap

The first element on each dimension represents the Pragmatic Function itself. Focus relates to information which the Speaker wishes to mark in terms of its newness in the communicative situation; as such it typically reflects what the Speaker wants the Addressee to see as the point of the utterance. Topic, which is clearly postulated as belonging to a separate dimension from Focus, relates to the Speaker’s desire to mark what he/she sees as the communicative starting point for the utterance. Finally, Contrast relates to the Speaker’s decision to highlight a difference between one or more elements of the Communicated Content and elements of the same or another Communicated Content. Information which belongs to the



Functional Discourse Grammar: pragmatic aspects

Background, the Comment, and the Overlap will typically not be given special treatment in a language. Hengeveld & Mackenzie (2008: 101) offer the following examples of types of Content Frame containing Subacts with Topic and Focus status (note that SA stands for Subact): (35) Thetic [(SA)N]foc (36) Categorical [(SA)TOP (SA)N (SA)foc] (37) Presentative [(SA)N (SA)topfoc]

Thetic statements as a whole contain only new information, and however many Subacts are involved, together they will be marked as in Focus. By contrast, Categorical statements typically involve presupposed information of some kind, and will often, though not necessarily, have elements which are marked not only for Focus but also for Topic. Finally, Presentative constructions are characterized in FDG at the Interpersonal Level as combining Topic and Focus functions. Each function – Focus, Topic, and Contrast – will now be detailed in turn. Focus The Focus function indicates the Speaker’s decision to give special treatment to information which is new in the given setting. There are four frequently attested means for signaling Focus function (cf. Dik 1997a: 327): –– –– –– ––

Focus markers Special constituent order Focus constructions Prosodic prominence

An example of a language with a specific Focus marker is Wambon. In the following example from de Vries (1985: 172), the focus marker -nde is first used in a question to signal the gap in the Speaker’s knowledge, and then in the answer to signal the piece of information which fills that gap: (38) a.

Jakhove kenonop-nde takhim-gende? 3.pl what-foc buy-3.pl.pres.final ‘What do they buy?’

b. Ndu-nde takhim-gende Sago-foc buy-3.pl.pres.final ‘They buy sago.’

107

108 Mike Hannay & Kees Hengeveld

In Turkish and Hungarian, there is a fixed position for the Focus constituent in a clause, immediately before the verb: (39) Turkish (Kornfilt 1997: 190) Kitab-ı Ali-ye Hasan ver-di. book-acc Ali-dat Hasan give-pst ‘It was Hasan who gave the book to Ali.’ (40) Hungarian (Kenesei et al. 1998: 166) A vendégek tegnap érkeztek a szállodá-ba. def guests yesterday arrived def hotel-loc ‘It was yesterday that the guests arrived at the hotel.’

In Slavic languages Focus is preferably placed in clause-final position, for instance in Bulgarian (Stanchev 1997): (41) Az vchera v magazina kupix edna kniga. 1.sg yesterday in shop.def buy.pst.1.sg indef book ‘I yesterday bought (a book)foc in the shop.’

With regard to special Focus constructions, an often noted example is the English it-cleft construction: (42) a. Who ate my cake? b. It was the dog that did it.

Verbal predicates are less happy in this cleft construction, but the pseudo-cleft and identifying constructions allow a focused predicate to appear in either initial or final position: (43) a. b. c.

What you want to do is disappear. Disappear is what I want to do. ??It’s disappear that I want to do.

(44) a. b.

The only thing I want to do now is sleep. *It is sleep that I want to do now.

Finally, Focus is perhaps most strongly associated with prosodic prominence. The use of prosody is most strongly apparent in languages where the word order is not in the first instance determined by pragmatic ordering, but by syntactic ordering. In these examples from English, the capitalized words carry the nuclear stress: (45) John PREFERS coffee, but he does not insist on it. (46) JOHN prefers coffee, not Fred. (47) John prefers COFFEE, not tea.

Functional Discourse Grammar: pragmatic aspects 109



Topic We turn now to the Topic function. As with Focus, certain languages have Topic markers. In Oromo, for example, subjects, but not objects, may be marked for Topic, and if they do then that also triggers gender agreement. If the subject does not have Topic function, then gender agreement does not apply:   Oromo [Clamons et al.1996, quoted in Bolkestein 1998: 204] (48) a.

Intal-tii-n magalaa dhuf-t-e. girl-f.sbj-top market come-f-prf ‘The girl came to the market.’

b. Intala takk-a magalaa dhuf-e. girl indef-sbj market come-prf ‘A girl came to the market.’

Certain languages also have fixed positions for the Topic constituent. An example is Mandarin Chinese (van den Berg 1989: 38), where clauses are constructed on the basis of Topic-first: (49) Wǒmen chī miàn. we eat noodles ‘We are eating noodles.’

Certain non-subject elements in a verb-second language like Dutch might also be analyzable as having Topic function when placed in initial position: (50) Dat weet ik. That know I ‘I know.’

Contrast Finally, we turn to Contrast, which is often seen as a special kind of Focus because both notions involve saliency. However, FDG places it on a separate dimension, as it can co-occur with both Topic and Focus. Kham (Watters 2002: 183–185) has a special contrast marker te: (51) Ao po:-lә te tam ja:h-si-u li-zya. this place-loc contr wheat put-detrans-nmlz be-cont ‘In this place, as opposed to others, wheat has been sown.’

In Bulgarian (Stanchev 1997), Focus constituents appear clause-finally, as we noted above, but contrasted information appears in pre-verbal position (though before clitics): (52) Az kola-ta vchera ya prodadox, a ne dnes. 1.sg car-def yesterday 3.sg.acc sell.pst.1.sg and neg today ‘It was yesterday that I sold the car, and not today.’

110

Mike Hannay & Kees Hengeveld

As the previous example suggests, it appears that languages may also exhibit strategies for marking combinations of Pragmatic Functions. For instance, Korean has a special Contrastive Topic marker (see Lee 1999), illustrated by the following example (Pultr 1960: 224): (53) Na-nɯn morɯ-mnida. 1.sg-contr.top not.know-hon ‘I don’t know (but others may).’

The combination of Topic and Focus is seen as characteristic of Presentative constructions, which introduce a new Topic into the discourse. Consider this example from French: (54) Il est arrivé trois trains. it aux.pres.3.sg arrive. ptcp.sg.m three trains ‘There arrived three trains.’

In this construction the single argument appears in clause-final position and there is no agreement on the verb. The position of the constituent expressing the presented entity can be directly related to the combination of the Topic and Focus functions. We would like to round off this section by referring to certain problematic aspects of linguistic analysis when it comes to assessing the coding of pragmatic status. We already saw in Section 4.2 that it may be difficult to determine whether certain phenomena should be analyzed as rhetorical or as pragmatic in nature. Another aspect is what actually constitutes the coding of information status. First of all consider this stretch of text analyzed by Moya Guijarro (2005: 212): (55) Our little farm [new topic] is a combined small-farm park and nature trail, designed to fascinate children, parents and grandparents equally […]. The incubator and Hatchery [subtopic] houses fertilized eggs being kept warm until they [known topic] hatch […]. The shop [subtopic] sells gifts and all kinds of delicious, local country fare from Stilton to strawberry jam. The Old Cowshed Tea Room [subtopic] will refresh and replenish. Lots of information inside the buildings [known topic] will help you understand more about the animals.

Moya Guijarro shows how different kinds of topicality (see the elements in square brackets) are involved in creating coherence in a text, and how different degrees of accessibility of individual referents may affect the form of the referential expression and may do so in particular ways in certain genres. However, in cases like this there is no actual coding of Topichood in the grammar itself. The informational status of elements can also have an influence on grammatical choices without pragmatic status having to be coded. Here is an example from Latin (Bolkestein 1985, quoted in Dik 1997a: 320) involving two alternative constructions for the same three-place predicate:



(56) a.

Functional Discourse Grammar: pragmatic aspects

Mensam aqua aspergit. table.acc water.abl sprinkle.3.sg.pres ‘He sprinkles the table with water.’

(56) b. Aquam mensae aspergit. water.acc table.dat sprinkle.3sg.pres ‘He sprinkles water onto the table.’

In the first example, mensa is in the accusative, whereas in the second example aqua has accusative case. Bolkestein shows that which element will be presented in the accusative here is strongly determined by what is the discourse topic at that point in the discourse, or what is to be introduced as a new discourse topic. In the first example, the table will tend to be topical; in the second, the water will be topical. However, there is no formal marking of Topichood, by means of either clause position or case form. There are thus many linguistic choices that are influenced by pragmatic factors which are recorded in the Contextual Component, and which have an effect on the Formulation process in terms of, for example, the choice of predication frame and referential strength. Nevertheless, in FDG the underlying representation that results from Formulation will not carry a Pragmatic Function in such cases. 5.3  Ascription and Reference Ascriptive and Referential Subacts may be represented as in (57) and (58): (57) (π T1: head (T1): Σ (T1)) (58) (π R1: head (R1): Σ (R1))

In most cases Subacts have no lexical heads, since the lexical material they contain is introduced at the Representational Level. There are, however, certain lexical elements that can be introduced directly into the head slot of a Referential Subact. These elements include proper names and pronouns used deictically. Proper names can be used as uniquely referring elements that have no meaning in the strict sense, while pronouns used deictically derive their reference from the speech situation in which they are used. Both classes of elements can thus be said to function uniquely at the Interpersonal Level. While most Subacts are thus headless, mapping onto lexical material available at the Representational Level, they can be modified in various ways, either by grammatical (π) or by lexical (Σ) means. We will consider the various types of modification in what follows. The most important operators applying to the R-variable have to do with the identifiability of the referent, as assessed by the Speaker. FDG distinguishes between two aspects of identifiability. The first concerns the Speaker’s assumptions about the identifiability of the referent for the Addressee: this is reflected in the operators {+id, –id}

111

112

Mike Hannay & Kees Hengeveld

for identifiable and non-identifiable, respectively. The second concerns the Speaker’s indication of the identifiability of the referent for her/himself: this will be reflected in the operators {+s, –s} for specific and non-specific, respectively. The two sets of identifiability operators can be combined in the following ways: – {+id, +s} applies in all cases in which the referent is assumed to be identifiable for both speech act participants, as in (59): (59) I met the teacher today.

– {–id, +s} applies when the referent is identifiable for the Speaker but assumed to be non-identifiable for the Addressee: (60) I am looking for a (certain) teacher.

– {–id, –s} applies when the referent is identifiable to neither Speaker nor Addressee: (61) I am looking for someone to help me prepare for my exams.

– {+id, –s} applies when the referent is non-identifiable for the Speaker but assumed to be identifiable for the Addressee. This is the situation typically reflected in content questions, such as: (62) Who do you think can help me prepare for my exams?

This four-way distinction is helpful in accounting for a wide range of linguistic facts. We will give just two examples here. In Spanish, case-marking of human Undergoers is sensitive to specificity: all specific human Undergoers are introduced by the preposition a, while all non-specific human Undergoers lack this preposition. Thus, we get oppositions like: (63) a.

La anciana ama a los niño-s. def.f old.woman(f) love.pres.3.sg u def.m.pl child(m)-pl ‘The old woman loves the children.’

b. La anciana ama los niño-s. def.f old.woman(f) love.pres.3.sg def.m.pl child(m)-pl ‘The old woman loves children (in general).’ (64) a.

Busc-o a un-a mujer. look.for-1.sg.pres u indef-f woman(f) ‘I am looking for a certain woman.’

b. Busc-o un-a mujer. look.for-1.sg.pres indef-f woman(f) ‘I am looking for a wife.’

In both the definite (63) and the indefinite (64) domain, Spanish thus systematically distinguishes between specific and non-specific reference. The relevance of this distinction



Functional Discourse Grammar: pragmatic aspects

for Spanish grammar furthermore shows up in the fact that the verb in relative clauses in non-specific NPs is in the subjunctive, while it is in the indicative in relative clauses in specific NPs. Bininj Gun-Wok (Evans 2003) provides another example of how specificity may play a pervasive role in the organization of grammar. Interrogative and indefinite pronouns share the feature ‘non-specific’ in the analysis given above. This fits in nicely with the formal properties of Bininj Gun-Wok, a language in which interrogative and indefinite pronouns form a single class of elements, labeled “ignoratives” by Evans (2003: 273). In this language it is the Illocution (Interrogative versus Declarative) that leads to the intended interpretation of the ignorative. Thus, njale is variously glossed as ‘what’ or ‘something’. What the interrogative and indefinite uses of ignoratives such as njale share in terms of FDG is the presence of an operator for non-specificity at the layer of the Referential Subact. The most important operators at the layer of the Ascriptive Act concern the indications on the Speaker’s part that the Subact only approximates his/her actual communicative intentions. This involves introducing a hedge to this effect. Consider the following examples (see Hengeveld & Keizer forthc.): (65) Don’t you get the feeling that she’s living there in that house, and the rest of it’s sort of derelict or totally deserted. (66) He like wrestles with me and just stuff that a big brother would do. (67) The wall is yellowish.

The elements sort of in (65), like in (66), and -ish in (67) all indicate that the Speaker thinks the Ascriptive Subact is realized by means that are not fully adequate to express the intended meaning. They are thus called approximative expressions that lead to non-straightforward Ascription. The opposite of approximation is exactness. Some languages do not mark approximation but rather the opposite. Such a language is Leti (van Engelenhoven 2004: 160): (68) a.

N-tivl=e. b. 3.sg-flap=exct ‘It flaps.’

N-tivla. 3.sg-flap ‘It (sort of) flaps.’

The presence of the exactness morpheme in (68a) indicates that the Speaker considers the Ascription fully adequate, while its absence in (68b) indicates that the Ascription is approximate. In FDG the distinctions that we have outlined in this section are captured by operators at the Subact layer to the extent that they are expressed by grammatical means, and by modifiers to the extent that they are expressed by lexical means. Operators and modifiers of these types are stored as part of the set of Interpersonal Primitives within the FDG Lexicon, and inserted into underlying structure by the Formulator.

113

114

Mike Hannay & Kees Hengeveld

6.  Conclusion In this chapter we have outlined the basic characteristics of FDG, focusing on those aspects of this model that are of particular interest to those working in the field of pragmatics. FDG is a functionally oriented model of grammar, which means that it uses pragmatic notions only to the extent that languages encode the distinctions corresponding to these notions, be it morphosyntactically or phonologically. This perspective, when applied to a sufficiently varied sample of languages, provides new insights within the field of pragmatics, to the extent that it shows how the grammars of languages may be used to uncover pragmatic distinctions that language communities have found sufficiently crucial so as to encode them in their linguistic system.

Abbreviations used 1 3 abl acc aux coll cont contr dat def detrans dub exct f final foc hon hort imper incl

first person third person ablative accusative auxiliary collective continuative contrast dative definite detransitivizer dubitative exact feminine final verb form focus honorific hortative imperative inclusive

indef loc m neg nmlz pl pres prf proh pst ptcp real sbj sg sub top tr u val

indefinite locative masculine negation nominalization plural present perfective prohibitive past participle realis subject singular subordinator topic transitive undergoer validator

References Abbott, M. (1991). Macushi. In: D.C. Derbyshire & G.K. Pullum (eds), Handbook of Amazonian languages, Volume 3: 23–160. Mouton de Gruyter. Berg, M. van den (1989). Modern Standaard Chinees: Een functionele grammatica. Coutinho.



Functional Discourse Grammar: pragmatic aspects

Bolkestein, A.M. (1985). Discourse and case marking: Three-place predicates in Latin. In: C. Touratier (ed.), Syntaxe et Latin: 191–225. Université de Provence. ——— (1998). What to do with Topic and Focus? In: M. Hannay & A.M. Bolkestein (eds), Functional Grammar and verbal interaction: 193–214. Benjamins. Clamons, R., A.E. Mulkern, G.A. Sanders & N.J. Stenson (1996). The limits of formal analysis: Pragmatic motivation in Oromo grammar. Paper presented at the Conference on Formalism and Functionalism in Linguistics, Milwaukee. Cole, P. (1982). Imbabura Quechua. North-Holland. Dik, S.C. (1997a). The theory of Functional Grammar. Part 1: The structure of the clause. Edited by Kees Hengeveld. Mouton de Gruyter. ———(1997b). The theory of Functional Grammar. Part 2: Complex and derived constructions. Edited by Kees Hengeveld. Mouton de Gruyter. Engelenhoven, A. van (2004). Leti, a language of Soutwest Maluku (Verhandelingen van het Koninklijk Instituut voor Taal-, Land- en Volkenkunde 211). KITLV Press. Evans, N.D. (2003). Bininj Gun-Wok: A pan-dialectal grammar of Mayali, Kunwinjku and Kune. 2 vols. Australian National University. Hengeveld, K. & M.E. Keizer, M (forthc.). Non-straightforward communication. Hengeveld, K. & J.L. Mackenzie (2008). Functional Discourse Grammar: A typologically-based theory of language structure. Oxford University Press. ——— (2009). Functional Discourse Grammar. In: Bernd Heine & Heiko Narrog (eds), The Oxford Handbook of Linguistic Analysis. Oxford University Press. Hengeveld, K., E. Nazareth Bechara, R. Gomes Camacho, A. Regina Guerra, T. Peres de Oliveira, E. Penhavel, E. Goreti Pezatti, L. Santana, E.R.F. de Souza & M.L. de Sousa Teixeira (2007). Basic illocutions in the native languages of Brazil. In: M.M. Dall’Aglio Hattnher & K. Hengeveld (eds), Advances in Functional Discourse Grammar. Alfa – Revista de Lingüística 51: 73–90. Kenesei, I., R.M. Vago & A. Fenyvesi (1998). Hungarian. Routledge. Kornfilt, J. (1997). Turkish. Routledge. Kroon, C. (1995). Discourse particles in Latin (Amsterdam Studies in Classical Philology 4). Gieben. Lee, C. (1999). Topic, contrastive topic and focus: What’s on our minds. http://www.mind.sccs.chukyo-u. ac.jp/jcss/ICCS/99/olp/pt1/pt1.htm. Levelt, W.J.M. (1989). Speaking. MIT Press. Lyons, J. (1977). Semantics. 2 vols. Cambridge University Press. Mann, W.C. (2005). RST web site. http://www.sfu.ca/rst Mann, W.C., C.M.I.M. Matthiessen & S.A. Thompson (1992). Rhetorical Structure Theory and text analysis. In: W.C. Mann & S.A. Thompson (eds), Discourse description: Diverse linguistic analyses of a fund-raising text: 39–78. Benjamins. Matthiessen, C.M.I.M. & S.A. Thompson (1988). The structure of discourse and ‘subordination’. In: J. Haiman & S.A. Thompson (eds), Clause combining in discourse and grammar: 275–329. Benjamins. Miller, W.R. (1965). Acoma grammar and texts (University of California Publications in Linguistics 40). University of California Press. Moya Guijarro, A.J. (2005). The assignment of topical status in FDG: A textual analysis. In: J.L. Mackenzie & M.Á. Gómez-González (eds), Studies in Functional Discourse Grammar: 195–226. Peter Lang. Pultr, A. (1960). Lehrbuch der koreanischen Sprache. Max Niemeyer. Searle, J.R. (1969). Speech acts. Cambridge University Press. Stanchev, S.B. (1997). Pragmatic functions and special sentence position in Bulgarian. In: J.H. Connolly, R.M. Vismans, C.S. Butler & R.A. Gatward (eds), Discourse and pragmatics in Functional Grammar: 121–135. Mouton de Gruyter.

115

116

Mike Hannay & Kees Hengeveld Taboada, M. & W.C. Mann (2006). Rhetorical Structure Theory: Looking back and moving ahead. Discourse Studies 8: 423–459. Verschueren, J. (1989). Language on language: Toward metapragmatic universals. IPrA Papers in Pragmatics 3: 5–144. Vries, L. de (2006). Areal pragmatics of New Guinea: Thematization, distribution and recapulative linkage in Papuan narratives. Journal of Pragmatics 38: 811–828. Watters, D.E. (2002). A grammar of Kham (Cambridge Grammatical Descriptions). Cambridge University Press.

Generative semantics James D. McCawley

The term ‘generative semantics’ (GS) is an informal designation for the school of syntactic, semantic, and pragmatic research that was prominent from the late 1960s through the mid-1970s and whose best-known practitioners were George Lakoff, James D. McCawley, Paul M. Postal, and John Robert Ross.

1.  The history of generative semantics The appearance of Chomsky’s Aspects of the theory of syntax (1965) initiated a striking burst of activity in syntax, in large part because Aspects made it respectable to integrate semantic research with syntactic research. The GS research tradition originated as part of that burst of activity. The earliest recognizably GS works were attempts to fill gaps and rectify defects in the Aspects conception of syntax while preserving most of its framework. The most notable example of this is George Lakoff ’s (1965) widely circulated Ph.D. thesis (eventually published in 1970) in which Lakoff worked out a theory of lexical exceptions to transformations such as was needed to prevent the transformations posited in Aspects from applying in cases where they should not. Lakoff realized that his approach to exceptions could be used not only e.g. to account for the differences among try, expect and, believe with respect to Equi-NP-deletion but also to eliminate many category distinctions from deep structures (e.g. adjectives could be taken as occupying the same deep structure positions as verbs and difering from them in allowing application of a transformation that inserted copula be) and to develop analyses in which semantically complex predicate elements such as causative verbs were decomposed in deep structures into semantically simpler elements that occupied the predicate positions of a syntactically complex structure, as in a derivation that related the surface form Floyd broke the glass to a deep structure of the form Floyd CAUSE ((the glass BROKEN) BECOME). A recognizable GS research community came into being in the 1966–1967 US academic year, when Lakoff (at Harvard) and Ross (at MIT) taught courses in which they developed a large body of syntactic analyses in terms of deep structures that were fairly remote from the corresponding surface structures and fairly close to what were arguably the semantic structures of the sentences in question; these courses (which

118

James D. McCawley

took place during a year when Chomsky was on leave and out of contact with current developments in Cambridge MA) stimulated many students at the two institutions to do research in the evolving GS framework. GS first achieved prominence as a clearly defined alternative to standard transformational grammar at the conference on language universals and linguistic theory (held at the University of Texas at Austin in March 1967) whose proceedings were published as Bach & Harms (1968). Of the two GS contributions to that volume (Bach 1968; McCawley 1968), it is Bach’s that is the more unequivocally generative semantic. Only in the postscript to McCawley’s paper, added subsequently to the conference, are the issues taken up in the body of the paper dealt with in a framework that rejects the distinctions between syntactic deep structure and semantic structure and between syntactic transformation and semantic interpretation rule. Lakoff and Ross, though not presenting papers, also played a major part in the conference, both in the discussion of the papers and in informal lectures that they gave after the end of the conference proper. The as yet quite small GS research community increased considerably in size during the 1968 Linguistic Institute (held in June-August 1968 at the University of Illinois at Champaign-Urbana), at which many students attended courses in which Lakoff, McCawley, and Ross presented a large body of GS syntactic analysis and devoted considerable attention to justifying specific semantic structures and integrating them with their syntactic analyses. An unusually large number of GS papers were presented at the 1969 annual conference of the Chicago Linguistic Society, which dwarfed all previous CLS conferences in terms of numbers of participants, submissions, and papers accepted. For the next few years, the CLS conference was the venue at which the largest amount of GS research was presented. After his return from his 1966–67 sabbatical, Chomsky devoted large segments of his classes to arguments against a number of the GS analyses and claims, disavowing his earlier identification of deep structure as the one syntactic level that is directly relevant to semantics (“the syntactic component of a grammar must specify, for each sentence, a deep structure that determines its semantic interpretation and a surface structure that determines its phonetic interpretation”, Chomsky 1965: 16) and beginning to work out the interpretivist/lexicalist approach to syntax of Chomsky (1972). From the late 1960s through the 1970s, GS-ists were one side in what Newmeyer (1980) has dubbed ‘the linguistic wars’: widely publicized and often heated disputes with the Chomskyan wing of generative grammar. This Chomskyan wing was itself in a period of rapid flux, as Chomsky, Jackendoff, Emonds and others developed an ‘interpretivist’ version of generative grammar, in which the domain of syntax was sharply reduced, conceptions of semantic structure were developed that made it appear very unlike syntactic structure, and both deep structure and surface structure were taken to figure directly in semantic interpretation. GS critiques of early interpretivist work



Generative semantics

influenced the evolving interpretivist tradition in that many consequences of the interpretive stance that GS-ists had pointed out as reductiones ad absurdum of interpretivism were then happily embraced by interpretivists. (One such consequence was that a transformation deleting repeated VPs would have to be rejected in favor of deep structures containing zero elements plus a semantic interpretation rule that sometimes allowed the zero elements to be given an interpretation). The disputes were often over matters that had little connection with the more fundamental differences between the two sides. For example, as Pullum & Wilson (1977) pointed out, an analysis of auxiliary verbs as main verbs taking underlying sentential complements is not inherently any less compatible with the central tenets of the interpretive school than with those of GS, but such analyses were uniformly attacked by the former and defended by the latter. The size of GS community and the amount of research undertaken within a GS framework continued to increase in the early 1970s, but my the mid 1970s the community began to dwindle in size and by the end of the decade no real GS community existed any longer. The reasons for the rapid demise of a GS community are a matter of controversy. The catastrophic decline in identification with a GS approach to syntax and semantics was not a straightforward matter of GS-ists coming off the worse in the battles of ‘the linguistic wars’, since the two sides had about equal success in picking holes in the other’s analyses and encountered equally ignominious failures in providing substance to their claims. As factors responsible for the reversal of the rapid growth of the GS community in the late 1960s, Newmeyer has suggested: (i) the fact that some prominent GS-ists trained relatively few syntacticians in the 1970s; (ii) the fact that a number of erstwhile GS-ists and many neophyte syntacticians became converts to other approaches to syntax that were developing in the early 1970s, particularly Montague grammar and relational grammar; (iii) distaste by some linguists for the casual and at times flippant, vulgar, and dadaist style that GS-ists often affected; and (iv) the willingness of leading GS-ists to allow the factual domain for which they held themselves and their theories accountable to expand beyond what many linguists regarded as reasonable limits. Items (iii) and (iv) in the above list were implicated in a consideration that has been less remarked on, namely that Chomsky and many of his followers affected the role of defenders of virtue, while GS-ists avoided any claims to virtue. Newmeyer’s (1980: 167) statement that “Generative semanticists simply gave up on attempting to explain grammatical phenomena” is historically inaccurate (see McCawley 1980), but draws credibility from the fact that GS-ists rarely used the word explain and its derivatives in presenting their explanations of linguistic phenomena, while interpretive semantics used those words with high frequency. The interpretive side was aided in its attempt to don the mantle of virtue by a series of mathematical results achieved by Stanley Peters

119

120 James D. McCawley

and Robert Ritchie during the very period when ‘the linguistic wars’ were raging, having to do with ‘generative capacity’ in a technical sense. Peters & Ritchie (1971) gave a definition of ‘possible transformational grammar’ to which at least the more mathematically minded generative grammarians did not object and proved that every ‘recursively enumerable’ set (i.e. every set whose membership one can specify in even a week sense) is generated by some transformational grammar. Thus, if one is doing ‘generative grammar’ in the technical sense of Chomsky (1957), that is, if one takes a language to be a set a sentences and a grammar to be a system of rules that specifies what that set of sentences is, the claim that all languages can be described by transformational grammars is nearly vacuous. As long as there is any way of specifying what sentences make up a given language, one is guaranteed that there will be a transformational grammar that will do that job. The Peters & Ritchie results were relevant to issues in linguistics only insofar as linguists were doing ‘generative grammar’ in the classical technical sense, i.e. specifying the membership of sets of sentences. While most transformational grammarians in the 1960s and 1970s claimed to be generative grammarians, the proportion who were doing generative grammar in that sense decreased steadily over the course of those decades. The Peters & Ritchie results, when they first appeared, were relevant to the preaching more than to the practice of most self-proclaimed ‘generative grammarians’, whose concern was with facts that had semantic as well as syntactic aspects — the various example sentences were given acceptability judgements relative to assumed understandings of them, and those who pointed out ‘extraneous’ alternative readings on which an author’s ‘*-ed’ examples were acceptable, were treated with derision by interpretivists and generativists alike. Because of the large disparity between the preaching and the practice of transformational grammarians, the Peters & Ritchie results served to popularize the truism that syntactic theories were virtuous to the extent that they were ‘restrictive’ and, since no analog to Peters & Ritchie had been proved for semantic theories (if one can even imagine what such an analog would be), served to legitimize the comparison of syntactic theories without regard to the semantic theories they are combined with. If the syntactic part of theory X is more restrictive than the syntactic part of theory Y, then X could be portrayed as more virtuous than Y, even if ‘syntax’ in Y takes in both what in X counts as ‘syntax’ and what counts as ‘semantics’ in X as well as what counted as ‘syntax’. The Peters & Ritchie results thus encouraged many linguists both to narrow their conception of ‘syntax’, thereby enabling them to produce ever more restrictive theories of syntax, while admitting less and less restrictive conceptions of semantic structure and of its relation to syntax, since only in syntax and not in semantics was there a strong moral demand for restrictiveness. Interpretative semantics accordingly condemned many GS analyses on grounds of insufficient restrictiveness while accepting what often



Generative semantics

were essentially identical analyses in which the objectionable syntactic rules counted instead as semantic rules (see McCawley 1975: Sections 3.2.4.4. and 4.8.).

2.  Tenets of GS The expression ‘generative semantics’ first appeared in a memorandum by George Lakoff that was written and circulated in 1963 but not published until 1976. The term gives undue prominence to an issue that is intelligible only against the background of an assumption that was made by virtually all generative grammarians in the 1960s but has gradually faded from all the lines of development of generative grammar: The assumption that there is a single level of linguistic structure for which there is a system of ‘generative’ rules, in the sense of rules specifying directly what structures are wellformed on that level (as in Chomsky’s Aspects, in which the ‘base rules’ specified directly what deep structures the language allowed), with all other levels of structure being related to the privileged level through ‘interpretive’ rules of some sort (transformations, in the case of the relationship between deep structure and surface structure; semantic interpretation rules, in the case of the relationship between deep structure and semantic structure) and thus well-formedness on the various other levels being specified only indirectly. The word ‘base’ embodied a metaphor that ascribed ‘fundamental’ status to the privileged level of structure: structures on other levels were held to owe their existence to the ‘base structure’ to which the interpretative rules interpreted related them. The issue memorialized in the name ‘generative semantics’ is that of whether the privileged level of linguistic structure is a level of semantic rather than syntactic structure. While the issue after which GS is named did in fact figure importantly in early disputes between GS and more orthodox generative grammar, a number of other issues of greater substance played more central roles in the disputes and in GS research. Since the views of the various GS-ists evolved rapidly and took in issues on which there was substantial disagreement among GS-ists, one cannot simply give a list of tenets that serve to characterize a GS viewpoint. Nonetheless, there are a number of points on which most GS-ists held views that were substantially at odds with those of the interpretivist branch of generative grammar, of which the following are some of the most important. 2.1  Against deep structure Whereas interpretivists insisted on a sharp distinction between syntactic structure and semantic structure (some of them disputing that such a linguistic level as semantic

121

122

James D. McCawley

structure could be justified — “I cannot really believe that anyone would be willing to take a stand on this issue with any conviction, given the present state of descriptive semantics”, Chomsky 1972: 137 — and others such as Jackendoff 1972 setting up a scheme for semantic structure that was formally very different from their scheme for syntactic structure), GS-ists insisted that there was no level of syntactic deep structure as distinct from semantic structure, and that the derivation that linked a semantic structure to a corresponding surface structure could not in general be divided into a semantic part and a syntactic part non-arbitrarily. GS-ists argued that syntax and semantics cannot be separated both by arguing against the conception of ‘deep structure’ that allegedly served as an interface between syntactic and semantic components of grammar (as in Lakoff ’s 1968 arguments that no single structural level has all the properties that supposedly characterized Chomsky’s level of ‘deep structure’) and by exhibiting syntactic reflexes of what were arguably constituents of semantic structures (e.g. the propositional functions commonly posited as the scopes of quantifiers in such sentences as Most linguists admire many philosophers can be the antecedents of pronouns). By the early 1970s, GS-ists were in a position to claim that virtually every aspect of logical semantics had some role in grammar. McCawley (1968) offered an argument against Chomskyan deep structure modeled after Halle’s (1959) argument against a phonemic level in phonology: the assumption of such a level prevents one from giving a unitary account of the distribution and interpretation of respective and respectively. To GS-ists, Chomsky’s segregation of syntax from semantics was an illusion achieved by building into syntactic structures ad-hoc substitutes for pieces of information that are more plausibly regarded as semantic, as in his treatment of selectional restrictions in terms of features such as ‘Animate’ and ‘Male’ that also figured in agreement rules. McCawley (1968) argues that selectional restriction can involve any semantic information whatever, that to determine whether a constituent satisfies a selectional restriction it is necessary to identify what sort of entity it denotes and not merely what features its head bears (e.g. the restriction on the object of subtract is that it has to denotes a number, which cannot be determined just by examining its head N: Subtract what you have computed/*baked from 87.2), and that ‘referential indices’ are not merely linguistic markers but correspond to purported referents and figure in set-theoretic structures. GS-ists rejected the highly unstructured representations of Katz & Fodor (1963), arguing that semantic structures, like syntactic structures, are labeled trees, but with semantic rather than lexical units as the atoms of those structures. Considerable attention was devoted by GS-ists (see esp. Lakoff 1972; McCawley 1972, 1981) to developing a conception of semantic structure that could be related in a systematic way to surface syntactic forms and which also was an appropriate structural level to serve as input to logical rules of inference and truth conditions.



Generative semantics

2.2  Derivational constraints Gs-ists rejected the prevalent metaphor of linguistic rules as steps in a procedure for assembling linguistic structures, adopting instead a conception of rules as derivational constraints, i.e. as imposing conditions that the structures making up a derivation have to meet. For example, an interrogative Wh-movement transformation was conceived of as requiring that if there is an interrogative S containing and coindexed with a Wh-expression in one stage of a derivation, there must be a subsequent stage of the derivation differing from the first one in that the Wh-expression is an adjunct of the interrogative S and the position of the Wh-expression in the first structure is empty. The notion of derivational constraint started out as a ‘notational variant’ of transformations, in that any relation between consecutive stages of a derivation that could be described in terms of one could be restated in terms of the other. However, as with most instances of ‘notational variants’, it opened up different possibilities for further development and made different things ‘conceptually problematic’ in the sense of Laudan (1976). For example, global rules (i.e. constraints on non-consecutive stages of derivations) are non-problematic under the ‘derivational constraint’ conception of rules but problematic under the ‘sentence factory’ metaphor, while rule ordering, which is coherent with the ‘sentence factory’, is problematic in a derivational constraint framework. 2.3  Context and acceptability GS-ists rejected the then prevalent conception of a sentence being grammatical or ungrammatical in and of itself. They rather spoke of the grammatically of a sentence only relative to an assumed meaning and a class of contexts. While there was no consensus among GS-ists on many other issues relating to ‘grammaticality’, GS-ists were receptive to views that deviated from the more orthodox conception, and individual GS-ists rejected the notion of ‘grammaticality’ as distinct from ‘acceptability’ (being careful not to confuse ‘acceptable’ with ‘accepted’: a sentence that a person has accepted is not necessarily acceptable for him), extended grammaticality and/or acceptability to units other than sentences (e.g. to discourses), or took acceptability and/or grammaticality to have multiple degrees and/or dimensions rather than being strictly binary. 2.4  Pragmatics integrated in semantics GS-ists generally held a broad conception of what consitutes semantics and accordingly did not draw a clear distinction between semantics and pragmatics. Accordingly, GS-ists often assimilated pragmatics to semantics (cf. the way that

123

124 James D. McCawley

morphology was assimilated to syntax in orthodox transformational grammar until the mid-1970s), as where Gordon & Lakoff (1971) described indirect speech acts in terms of ‘conversational postulates’ that treated the logical form of the conveyed act (e.g. the request for help that is conveyed by the interrogative Can you help me?) as being logically entailed by the context plus logical form of the literal speech act, with the sincereity conditions for a given illocutionary act determining the possibilities for performing it indirectly (e.g. Can you help me? can function as a request because a request presupposes that the addressee can comply). 2.5  The status of logic GS-ists argued that the syntactic categories of the deepest syntactic structures were logcial categories (such as ‘proposition’, ‘predicate’, and ‘argument of predicate’) and that logical categories played roles in numerous syntactic phenomena. They took those conslusions as establishing that syntactic categories simply were logical categories, an unwarranted leap that reflects an unfortunate assumption that GS-ists shared with interpretivists, namely that an expression belonged to a syntactic category solely in virtue of what it corresponded to in ‘base structures’ and that categories thus did not change in the course of derivations. (The widespread misconception that GS-ists’ analyses of nominalizations such as the destruction of the bridge required derivations in which transformations changed the category of an element rests on an equivocation between two conceptions of ‘category’ and a sloppy use of ‘change’). Gs-ists thus attempted to reduce the notion of syntactic category to logical category plus exception features and ignored the possibility that, besides logical category, additional factors (such as the part of speech of the head) might influence a unit’s syntactic behavior. It is of course possible to combine the other ideas commonly accepted by GS-ists with a conception of syntactic categories as combinations of syntactically relevant factors, with both logical category and part of speech of the head of a constituent being among the factors that affect its syntactic behavior (McCawley 1982, 1988). The fact that interpretive semantics was packaged together with a conception of syntactic categories that did recognize two such factors (namely the early version of X-bar syntax in which the ‘X’ had to be a part of speech, unlike more recent versions of X-bar syntax in which that restriction has been rejected) enabled the successes of X-bar syntax (see e.g. Jackendoff 1977) to count as successes for the whole ‘interpretive semantic’ package, even where the ‘interpretive’ aspect of the package was irrelevant. (Ironically, though Chomsky introduced X-bar syntax as part of an account of nominalizations, the successes of X-bar syntax have been in areas having nothing to do with nominalization; the only work to substantially broaden our understanding of nominalizations in the 1970s is Levi 1978, a GS work in which X-bar syntax plays no role).



Generative semantics

2.6  ‘Transformations’ In claiming that a single system of derivational constraints related semantic structures to surface structures, GS-ists committed themselves to accounting in syntactic terms for the relationship between semantically complex lexical items and semantic structures whose atomic parts were the semantic units making up the meanings of those items. Partial ‘decompositions’ of certain kinds of lexical items had in fact been treated by Lakoff (1965) in terms of transformations that applied to structures containing an ‘abstract’ causative or inchoative verb and combined the V with the V of its complement. CS-ists subsequently replaced Lakoff ’s separate causative and inchoative transformations by a considerably more general transformation of predicate-raising, which combined the predicate element of a complement S with that of its hist S, yielding a combination of semantic elements that could in principle correspond to a lexical item of the language in question, though only some of those combinations would correspond to actual lexical items of the language, e.g. Become (Not Alive) corresponded to English die, but Become (Not Obnoxious) did not correspond to any lexical item. More generally, GS-ists recognized prelexical transformations: both transformations that combined elements into complexes that represented possible lexical items (predicate-raising, as well as transformations that incorporated purpose and other adverbial expressions into the predicates of their host Ss) and prelexical applications of ‘ordinary’ transformations, as in the prelexical application of a passive transformation in the derivation of German adverbs such as polizeilich that express an underlying subject. By treating both lexical insertion and prelexical transformations as being simply transformations and thus applying in conformity with the cyclic principle, GS-ists (e.g. Binnick 1971) were able to account for the parallelism between idiomatic expressions with come and with bring (come/bring about, around, to…): Predicate raising, applying to a structure in which the complement of Cause had undergone lexical insertion of an expression with come, could yield a combination of semantic and morphological elements such as Cause-come, which lexical insertion could then replace by bring, yielding bring plus the remainder of the come expression. However, the claim that prelexical transformations conformed to the cyclic principle remained a matter of dispute (cf. Newmeyer 1976; Dowty 1979). Since GS-ists were thus committed to analyses in which superficially simple sentences such as Floyd broke the glass had underlying structures involving multiple embedded Ss, much of the attention of both adherents and opponents of GS was devoted to exploring the consequences of anayses involving such embedded Ss. Many cases were found in which adverbials appeared to modify not surface constituents of their host Ss but clauses of their semantic structures, as in He opened the door again, in the interpretation in which again modifies ‘The door is open’, or He almost killed her, in the interpretation in which almost modifies ‘She is not alive’ (i.e. ‘He brought

125

126 James D. McCawley

here close to death’). However, as Fodor (1970) pointed out, there are also cases in which an adverbial cannot modify a clause contained in the meaning of a word, e.g. John killed Mary on Friday can’t mean ‘John caused (Mary die on Friday)’. Neither side in the dispute achieved a satisfactory account of which cases worked the one way and which cases worked the other way. See, though, McCawley (1978) and the response by Horn (1978) for arguments that many of the differences in interpretation between simple lexical items and their supposed paraphrases are consequences of Grice’s maxim of manner. 2.7  The ‘base’ GS-ists generally espoused the ‘universal base hypothesis’, roughly the hypothesis that all languages allowed thesame set of semantic structures. The term reflects the origins of GS in the Aspects tradition, in which the rules specifying what deep structures a language allowed were called its ‘base rules’. When GS-ists used the term ‘base rules’ in their own work (before many of them came to reject the metaphor contained in the term, with its implication that the level in question is ‘fundamental’), it referred to rules for what semantic structures are possible. 3.  Pragmatics in GS While GS-ists quickly developed an interest in many questions of the sorts that occupy pragmaticists, it was not until the GS community was already in a period of decline that many GS-ists came to do pragmatics as pragmatics rather than to assimilate it to semantics. Accordingly, the body of research that is both unequivocally generative semantic and unequivocally pragmatic is relatively small. Still, the GS literature (especially the CLS volumes of the early to mid-1970s — see reprints in Schiller et al. 1978, and books such as Cole & Morgan 1975 and Rogers et al. 1975) includes papers devoted to a broad range of pragmatic issues, such as the meanings of a number of English ‘interjections’ (James 1972, 1973), conversational strategies in which the intentional use of ambiguous sentences enables a speaker to shift conversational risks to his interlocutor (Weiser 1974, 1975), the linguistic implications of the division of labor among different dimensions of politeness (R. Lakoff 1973), ways in which the choice between reflexive and simple pronouns reflects whose viewpoint is being taken (Cantrall 1974), ways in which the choice of tenses depends not only on time relations but on the role of speech act participants in what is said (R. Lakoff 1970), and how the division of labor among linguistic forms reflects which if any of the speech participants a given piece of information ‘belongs to’ (Forman 1974). Even GS works in which the author did not specifically do pragmatics usually concerned themselves with issues that had



Generative semantics

pragmatic aspects, as in G. Lakoff ’s (1970) demonstration that even such supposed purely grammatical phenomena as pronoun choice and agreement reflect inferences from background knowledge and semantic structure and that accordingly judgments of acceptability made in a pragmatic vacuum are not generally even possible.

References Bach, E. (1968). Nouns and noun phrases. In E. Bach & R.T. Harms (eds.): 90–124. Bach, E. & R.T. Harms (eds.) (1968). Universals in linguistic theory. Holt, Rinehart & Winston. Binnick, R.I. (1971). ‘Bring’ and ‘Come’. Linguistic Inquiry 2: 260–265. Cantrall, W. (1974). Viewpoint, reflexives, and the nature of the noun phrase. Mouton. Chomsky, N. (1957). Syntactic structures. Mouton. ——— (1965). Aspects of the theory of syntax. MIT Press. ——— (1972). Semantic interpretation in generative grammar. Mouton. Cole, P. & J.L. Morgan (eds.) (1975). Syntax and semantics vol. 3. Academic Press. Davidson, D. & G. Harman (eds.) (1972). Semantics of natural language. Reidel. Dowty, D. (1979). Lexical insertion in Montague grammar. Reidel. Fodor, J.A. (1970). Three reasons for not deriving ‘kill’ from ‘cause to die’. Linguistic Inquiry 1: 429–438. Forman, D. (1974). The speaker knows best principle. Papers from the 10th regional meeting of the Chicago Linguistic Society: 162–177. Gordon, D. & G. Lakoff (1971). Conversational postulates. Papers from the 7th regional meeting of the Chicago Linguistic Society: 63–84. Halle, M. (1959). The sound pattern of Russian. Mouton. Horn, L.R. (1978). Lexical incorporation, implicature, and the least effort hypothesis. Papers from the 14th regional meeting of the Chicago Linguistic Society, Parasession on the lexicon: 196–209. Jackendoff, R.S. (1972). Semantic interpretation in generative grammar. MIT Press. ——— (1977). X-bar syntax. MIT Press. James, D. (1972). Some aspects of the syntax and semantics of interjections. Papers from the 8th regional meeting of the Chicago Linguistic Society: 162–172. ——— (1973). Another look at, say, some grammatical constraints on, oh, interjections and hesitations. Papers from the 9th regional meeting of the Chicago Linguistic Society: 42–51. Katz, J.J. & J.A. Fodor (1963). The structure of a semantic theory. Language 39: 170–210. Lakoff, G. (1963). Towards generative semantics. In J. McCawley (ed.) (1976).: 43–61. ——— (1965). On the nature of syntactic irregularity. Ph.D. Diss., Indiana University. ——— (1968). Instrumental adverbs and the concept of deep structure. Foundations of Language 4: 4–29. ——— (1970). Presupposition and relative well-formedness. In D. Steinberg & L. Jakobovits (eds.), Semantics: 329–340. Cambridge University Press. ——— (1972). Linguistics and natural logic. In D. Davidson & G. Harman (eds.): 545–565. Lakoff, R.T. (1970). Tense and its relation to participants. Language 46: 838–849. ——— (1973). The logic of politeness. Papers from the 9th regional meeting of the Chicago Linguistic Society: 292–305. Laudan, L. (1976). Progress and its problems. University of California Press. Levi, J.N. (1978). The syntax and semantics of complex nominals. Academic Press. McCawley, J.D. (1968). The role of semantics in a grammar. In E. Bach & R.T. Harms (eds.): 124–170.

127

128 James D. McCawley ——— (1972). A program for logic. In D. Davidson & G. Harman (eds.): 498–544. ——— (1975). Review of Chomsky 1972. Studies in English Linguistics 3: 209–311. ——— (1978). Conversational implicature and the lexicon. In P. Cole (ed.), Syntax and Semantics, vol. 9: 243–259. Academic Press. ——— (1980). Review of Newmeyer 1980. Linguistics 18: 911–930. ——— (1981). Everything that linguists have always wanted to know about logic. University of Chicago Press. ——— (1982). The nonexistence of syntactic categories. In J.D. McCawley, Thirty million theories of grammar: 176–203. University of Chicago Press. ——— (1988). The syntactic phenomena of English. University of Chicago Press. ——— (ed.) (1976). Notes from the linguistic underground. Academic Press. Newmeyer, F.J. (1976). The prelexical nature of predicate raising. In M. Shibatani (ed.), Syntax and Semantics, vol. 6: 131–164. Academic Press. ——— (1980). Linguistic theory in America. Academic Press. Peters, S. & R. Ritchie (1971). On restricting the base component of transformational grammars. Information and Control 18: 483–501. Pullum, G.K. & D. Wilson (1970). Autonomous syntax and the analysis of auxiliaries. Language 53: 741–788. Rogers, A., B. Wall & J. Murphy (eds.) (1975). Proceedings of the Texas conference on performatives, presuppositions, and implicatures. Center for Applied Linguistics. Schiller, E. et al. (eds.) (1988). The best of CLS. Chicago Linguistic Society. Weiser, A. (1974). Deliberate ambiguity. Papers from the 10th regional meeting of the Chicago Linguistic Society: 723–731. ——— (1975). How not to answer a question. Papers from the 11th regional meeting of the Chicago Linguistic Society: 649–660.

Iconicity Elżbieta Tabakowska Jagiellonian University of Kraków

1.  Introduction Iconicity, which may be informally defined as correspondence between form and meaning, was not much favoured by the 20th century structural linguistics. Ever since the appearance of Cours de linguistique generale (1916), de Saussure’s fundamental claim that linguistic signs are arbitrary has reigned supreme in mainstream structural theories, as has his postulate of autonomy of the synchronic study of language. While de Saussure’s basic assumption that language is an instrument of communication has remained the cornerstone of many contemporary theories of language and language use, the arbitrariness dogma, as well as the Saussurean thesis that the state of a language at a given time (i.e. synchrony) can never be explained by reference to its past (i.e. diachrony), have now been seriously questioned. De Saussure claimed that limitations of arbitrariness are revealed in complex linguistic structures that display varying levels of what he defined as ‘relative motivation’, but in his theory of linguistic signs iconicity as one particular type of motivation was considered as a fairly marginal phenomenon. However, in their search for the motivated sign the followers of structuralism — notably students of literature and, towards the end of the 20th century, linguists of the cognitivist persuasion — have been accumulating convincing evidence of the ubiquity, the great variety and the intricacy of verbal iconicity. More recent contributions in the fields of theoretical and applied linguistics, pragmatics, discourse analysis and literary studies show that iconicity, which the framework of Saussurean or classical structural linguistics restricted to onomatopoeia and sound symbolism, can no longer be dismissed as an issue of little importance by anyone dealing with the nature and functioning of language. Today most students of language and its uses seem to share the conviction that iconicity is omnipresent in language and indeed constitutes a “prerequisite of mutual understanding in communication in general” (Nöth 2000: 25).

2.  History Iconicity is a semiotic notion; it has been recurring in philosophical thought about language — in various disguises — since antiquity. In Plato’s Cratylus Hermogenes

130 Elżbieta Tabakowska

adresses Socrates asking him the question whether the relationship between form and meaning should be considered as natural (physei) or conventional (ethei). Looking for an answer, modern semioticians were inspired by the writings of Charles Sanders Peirce (1902 [1955]). Peirce postulated a tripartite typology of signs, dividing them into icons (signs that bear certain similarity to the objects to which they refer, i.e. directly resemble the object, e.g. “a lead-pencil streak representing a geometrical line”), indices (signs connected to the objects they stand for by spatial and/or temporal links, e.g. a sound of knocking on the door as a sign of somebody’s presence and desire to enter) and symbols (signs arbitrarily associated with their objects “by virtue of a law”, i.e. by convention, e.g. “any utterance of speech which signifies what it does only by virtue of its being understood to have that signification”) (Examples taken from Peirce 1955: 104). Peirce’s semiotic categories were applied to linguistic signs: icons were defined as instances of linguistic isomorphism (e.g. Tok Pisin wilwil [= wheel-wheel] for ‘bicycle’), indices were described as deictic expressions (e.g. this one), and symbols comprised de Saussure’s ‘arbitrary signs’. When adapted by linguists, Peirce’s typology often became reduced to the binary opposition between iconicity (or motivation) and arbitrariness, or between ‘natural’ and ‘conventional’ signs. However, some authors had been pointing to the fact that the reduction is indeed an unwarranted simplification, since it is the joint impact of all the three modes of representation that accounts for the speaker’s use of language (cf. e.g. Seiler 1989). Peirce himself did not consider his three classes of signs as strictly dusjunct categories, claiming that in actual utterances they tend to overlap, although in varying proportions. In agreement with this claim, it was also pointed out that diachronic development of language shows a dynamic shift of class membership, with the development of symbols from icons being the main tendency (cf. e.g. Seiler 1989: 166). A plea for iconicity may be found, albeit implicitly, in the writings of the representatives of American anthropological linguistics, notably Edward Sapir and Benjamin Lee Whorf. What is today known as the Sapir-Whorf hypothesis originated from Whorf ’s conviction (inspired by the earlier work of Sapir, who was his teacher and master) that language determines thought, or — in its weaker version, known as the linguistic relativity principle, that “users of markedly different grammars are pointed by their grammars towards different types of observation, and hence […] must arrive at somewhat different views of the world” (Whorf 1940 [1956]: 221). Using examples from several indigenous languages of America, especially an Amerindian language Hopi, Whorf argued that in different languages (e.g. Hopi and English) different lexical and grammatical categories reveal distinct notional attitudes of the speakers. One of Whorf ’s much quoted examples refers to the notion of physical quantity. While English grammaticalizes the distiniction between count and mass nouns (many girls vs. much salt), Hopi has only count nouns. The relation between grammar (form) and



Iconicity

conceptualization (meaning) that pertains to the difference may be seen as iconic, and according to Whorf it makes the notional opposition between countability and uncountability “instantly acceptable” to speakers of English but not to those of Hopi. In more recent thought, the type of isomorphism postulated by Whorf may be retraced in the writings of linguists of ‘neo-Whorfian’ persuasion. For instance, in her seminal paper of 1995 Anna Wierzbicka argues that (at least in some languages) the parts-ofspeech membership of words is clearly non-arbitrary; in particular, she claims that the difference between grammatical structures expressing or describing emotions in Polish and English iconically reflects a corresponding (cultural) difference in the conceptualization of emotions: with their culture-induced penchant for spontaneity and warmth, speakers of Polish think of emotions in terms of processes (hence the prevalence of emotion verbs in their language), while the less spontaneous and outgoing Anglo-Saxons conceptualize emotions as states (and therefore refer to them mostly by means of adjectives) (Wierzbicka 1995). In the same ‘neo-Whorfian’ vein one of the theorists of iconicity, Paul Bouissac, claims that the dynamic relationship between our perception of objects in extralinguistic reality and linguistic expressions is a two-way-traffic: the former influences the iconic shape of the latter, but language, and especially the language of literature, can also influence perception (Bouissac 2005). Among the first linguists who explicitly undermined the Saussurean dogma of arbitrariness was Roman Jakobson: his seminal essay “Quest for the essence of language” (1965) is a plea for iconicity as a non-arbitrary property of language, and it offers the earliest analysis of what was to become the most quoted ritual example of iconicity on the level of syntax: Julius Ceasar’s dictum veni, vidi, vici. Of course the issue itelf was not new; for instance, in his book on the history of German syntax (1932) Otto Behaghel formulated his ‘first law’, which has since become recognized as one of the fundamental principles of iconicity: das geistig eng Zusammengehörige auch eng zusammengestellt wird (“what belongs together mentally, will also be placed together within the syntactic structure”; 1932:4, cf. 3.2. below). One of the first scholars whose contribution helped to revive the old interest in iconicity was Ivan Fónagy, whose work on the interplay of iconicity and convention in live speech goes back to the 1950’s. More recently, the relationships between the imitative and the conventional in spoken language were investigated with reference to phonological systems of natural languages, and with the help of modern electronic equipment. Predictably, it was revealed that auditory iconicity is language-specific, i.e. restricted by a given phonological system (Masuda 2003). Another forerunner in the field was Dwight Bolinger, whose work on iconic relationships in prosody came as a part of his overall “challenge to the fallacy of meaningless”, which he considered a fundamental sin of the generative theories of language as proposed in the late 1970’s (Bolinger 1977: ix). Bolinger’s plea for the recognition of motivated relations between form and meaning was carried over to the level of syntax. For instance, his remarks

131

132

Elżbieta Tabakowska

on the function in discourse of repetition, deletion of pronominalization of nominal phrases can well serve as an augury of much later, pragmatically oriented research (cf. e.g. Östman 1989; Schönefeld 2005). The direction for new scholarly pursuits was in fact set in the 1980’s by John Haiman, author of two seminal articles on iconicity (1980, 1983) and editor of a groundbreaking volume called Iconicity in Syntax (1985), which appeared at a time when the mainstream linguistic paradigm of Chomskyan generative grammars was not particularly favourable to the idea: not new, but “somewhat unfashionable” (Haiman 1985a: 3). During the decade which passed from the publication of that book to the appearance of another major landmark, a collection on articles edited by Raffaele Simone (1995), iconicity established its position in both linguistic and literary studies. Simone called his book Iconicity in Language, thus implying a widened scope of the phenomenon, covering all levels of linguistic structure and all aspects of language use. In the introduction, he states that the renewed interest of students of language in iconicity comes as an inevitable result of their realization that linguistics needs “a semiotic foundation” (1995: viii). Indeed, this widening of scope is symptomatic of the evolution of the revived interest in the topic as such. As was said above, in modern linguistics iconicity was traditionally recognized on the sound level as onomatopoeia. More general semantic implications of sound symbolism were dismissed as non-scientific by the scientism that underlay the American structural linguistics of Bloomfield and his followers. Although iconicity was admitted to syntax in the 1980’s, its ‘grammatical’ instantiations were still largely limited to the level of sentence, thus reinforcing the deeply rooted conviction that, as in all other areas of linguistic study, (theoretical) linguistics should be kept out of larger stretches of language, and in particular, distanced off from literary study. The situation changed with the advance of text grammars and discourse analysis. Till then, what kept the topic alive was the theory of literary mimesis (dating back to antiquity), which dealt with iconicity disguised as rhetorical schemas, tropes and figures. The well-established recognition that poetry involves raising arbitrary signs to the status of natural ones surfaced in the 1970’s as a theory of literary iconicity, first developed by J.M. Lotman (1970). Lotman’s claim that when used in poetry verbal signs become models of their contents and thus become semanticized, was in fact a sythesis of earlier research on iconicity as a fundamental literary device. The last twenty five years witnessed some crucial developments in linguistics. Among changes most relevant to the issue of iconicity one should mention the transcendence of the sancrosanct limits of a sentence as the largest unit of consideration, which brought the linguists’ interests closer to the pursuits of literary students. The same may be said about the increased emphasis that various theories of language have been putting on creativity as the main mechanism of language use and development. More crucially, following the ‘two revolutions in linguistics’ — the Chomskyan revolution



Iconicity

of 1970’s and the cognitive revolution of 1980s — there is a focus on the cognitive processes involved in language origins, development and use. For Chomsky and his followers investigation of language was tantamount to investigation of the cognizant mind; for the proponents of cognitive linguistics to investigate language means to investigate the cognitive processes themselves: from perception through data processing to actual expression. In the last decade of the 20th century research on iconicity placed it within the area of common interest shared by linguistics and literary study. This turn was perhaps best summarized by the title of the first international and interdisciplinary symposium on iconicity which was held in 1997 in Zürich and sigificantly called ‘Iconicity in Language and Literature’. In 2005, the fifth iconicity symposium, which was held in Kraków, sanctioned much further widening of the scope: linguists, literary scholars, translators, experts in nonverbal communication, musicians and representatives of visual and fine arts reached a general consensus, whereby iconicity was proclaimed to constitute one of the most fundamental factors underlying all systems od communication.

3.  Iconicity we live by: The state of the art 3.1  Iconicity as interpretation The extreme naturalist view that ‘nature’ created real and ‘correct’ links between linguistic signs and the things that those signs stand for — the moot point of Plato’s Cratylus — has long been discarded as the “natural language fantasy” (Fischer & Nänny 1999: xv). Iconicity is defined as a similarity between form and meaning as conceived by a human mind. ‘Similarity’ is a relation and as such it has to be recognized by the user of language; in Peirce’s triadic model this recognition becomes the function of the ‘interpretant’, who makes sense of the sign by relating the signifier (or Peirce’s ‘representamen’) to the signifiant (Peirce’s ‘object’). Thus, in this view, iconicity is inherently subjective, as it concerns an interpretant’s perception of reality rather than the (objective) reality as such; it pertains to language users’ cognitive models of the world. However, perception is not man’s passive response to the surrounding reality, but rather a constructive process, conditioned by both internal and external factors. The relation of similarity may — but need not — be imposed upon reality by a cognizant mind, which creates its conceptual structures. The assumption that the human mind is endowed with such a world-creating capacity underlies our present understanding of iconicity. On the one hand, it is also the cornerstone of the cognitive theory of metaphor, which constitutes the supreme manifestation of this capacity (cf. 3.3. below); on the other, it has brought about a shift of focus — from iconic signs that are found in the

133

134 Elżbieta Tabakowska

message towards the process of their reception. (cf. e.g. the discussion of iconization in a theoretical paper by Johansen, 2003). New assumptions call for a redefinition of old ideas: the emergence and the use of iconic signs is seen as conditioned by a wide bio-cultural context of human cognitive processes. Icons are culture- and therefore also language-specific, and as such they cannot be considered as either ‘universal’ or ‘natural’ (as opposed to ‘conventional’) (cf. e.g. Wierzbicka 1995). What is more, it is the most archaic, primitive signs, sometimes referred to as emotive, that are truly natural and universal: grunts, whistles, interjections etc. are innate (rather than culturally transmitted), common to all human beings and, in all probability, also to our biological ancestors. And it is precisely those natural and universal signs that are essentially non-iconic, as their “vocal pattern bears no formal relation to the referent” (Sadowski 2000: 71). “[I]conicity is not just characteristic of an earlier, more primitive stage of language but it plays a role whenever a speaker’s (or a writer’s) expressivity is at issue” (Nänny & Fischer 1999: xx). And expressivity is creative by definition: it is a novel interpretation of reality in “a speaker’s (or a writer’s)” search for new similarities and resemblances. On the other hand, the study of some typological issues relevant to the topic of iconicity suggests that the fundamental principles of iconicity may well belong among the universals of language (cf. e.g. Givón 1988, 1995; Haiman 1995; Östman 1989). 3.2  Principles of iconicity Iconicity manifests itself in three basic principles, known as the principle of sequentiality, the principle of proximity (distance) and the principle of quantity. The first principle refers to what has been known as ordo naturalis; it says that “language can be iconic of, and directly isomorphic with, subjects that can be linearised into temporal sequences” (Enkvist 1990: 176). Nils Erik Enkvist called this type of iconicity “experiential”; in literary tradition it has been known under the name of discourse mimesis. When confronted with utterances like (1), which is one of Enkvist’s well-known examples, (1) a. John and Susie got married and had a baby b. John and Susie had a baby and got married,

“we assume that the linear ordering of accounts of events in the text in fact corresponds to their temporal sequence in the world described” (Enkvist 1990: 172). A crucial factor in determining text strategies, this type of iconicity makes it possible to explain sentence ordering (e.g. adverbial fronting) in three basic types of texts: action oriented, with a cookery book as the category prototype:

(2) In an iron pot over high flame put 4 tablespoons butter,

location oriented, whose prototype is a guidebook:

(3) At the end of the entrance avenue to the right is the Turbe of Murad II



Iconicity

and time oriented, typically represented by a chronicle:

(4) In the twelfth century … the chapel was turned into Charlemagne’s shrine. In 1165 the body of the newly canonized saint was transferred to a casket of solid gold. … Today the imperial chapel at Aachen is ranked among the foremost wonders of romanesque art.

(Examples (1) – (3) after Enkvist 1990. Example (4) from Norman Davies, Europe, Oxford Univ. Press 1996, 305 – 6)

In Ronald Langacker’s cognitive grammar experiential iconicity is taken to motivate structures known as ‘nested locatives’:

(5) The rake is in the yard by the back fence near the gate (Example from Langacker 1988: 71).

In literary studies, the principle is applied to the structure of narratives: a story in which events are related in a given sequential order is naturally taken to be directly isomorphic with its subject matter. In other words, the chronology of narration mimes the purported sequence of events in the fictional world (Leech & Short 1981). The second principle of iconicity is a modernized version of the Behaghel’s ‘first law’ (cf. above); it states that things that are conceptualized as being close conceptually tend to be put close together in linguistic expressions. In morphology, it accounts for instance for the fact that in nouns markers of number usually precede markers of grammatical case, which expresses the more indirect of the two relations. On the level of syntax, it provides explanation for word order variations, a classical example being the oft-quoted phrase

(5) the famous delicious Italian pepperoni pizza (Example from Radden 1991: 5),

where the growing distance between the head noun and its modifiers mimes the growing notional distance between the relevant concepts: while ‘pepperoni’ is an inherent property (or a criterial attribute), of the pizza, ‘famous’ and ‘delicious’ refer to subjective evaluation and judgement, which are prone to change. The last principle of iconicity, that of quantity, accounts for the tendency to maintain the relation of direct proportion between the amount of form and the amount of meaning: more form tends to carry more meaning while less form is usually associated with less meaning. Linguistic evidence abounds; quantitative iconicity is found on all levels of linguistic structure, from the level of sound (iconic prolongation of vowels, e.g. a looong story) and morphology (reduplication as a marker of the plural number) through syntax (the rhetorical strategy of repetition, e.g. a long, long story) to verbal politeness (cf. a model experimental study in Őstman 1989), where the length of the utterance has been shown to mime the metaphorical distance that the speaker respectfully keeps from the listener, as in:

(6) Would you be so kind as to close the door?

135

136 Elżbieta Tabakowska

Some scholars point to the iconic function of ‘zero’, the extremum value situated at the opposite end of the quantity scale. For instance, in Langacker’s grammar nonexpression of the observer’s position, as in (7) a. The hill gently rises from the bank of the river b. The hill gently falls to the bank of the river (Example from Langacker 1988: 87)

is an icon of the subjective construal of the scene, whereby the oberver himself is not part of the scene observed. With reference to literary texts, the iconic function of asyndetic syntax and ellipsis was dicussed by one of the pioneers in the field, Wolfgang Müller (1999). In one of his wide-ranging discussions of the origins of language and grammar Talmy Givón argues convincingly for the ‘extremely iconic’ character of rules in protogrammars. Giving testimony to the principles of iconicity at work, he claims that these rules were ‘cognitively transparent’ and ‘non-arbitrary’. Isomorphic to the sequential ordering of events, proto-grammatical structures are believed to have revealed direct iconic relationships between sentence stress and the predictability of information, between linguistic proximity and semantic relevance, between reduced (or zero) expressions and the amount of attention paid by the speaker to the information conveyed (Givón 1995: 406–7). In Givón’s work linguistic thought about iconicity acquires an important diachronic dimension. 3.3  Types of iconicity On the cognitive plane, the generally accepted dichotomous taxonomy of iconicity involves the distinction between imagic and diagrammatic iconicity. Imagic iconicity is concrete and direct; it is defined as consisting in a one-to-one relation between the sign and the object, or in Saussurean terms, between the signifier and the signified. Even though the postulate of a direct one-to-one correspondence might seem problematic (cf. the discussion of the Peircean ‘interpretant’ in 3.1. above), imagic iconicity is considered to be directly perceptual. It goes beyond the limits of ‘language proper’ and may be revealed through vision rather than through the oral/aural channel of verbal communication, i.e. be visual rather than verbal (e.g. the iconic value of carmen figuratum or gestures known as emblems; cf. 4. below). Because of the postulated directness of the iconic relationship, imagic iconicity is sometimes also referred to as primary. In contrast, diagrammatic iconicity is often called secondary, as it is abstract and relational; it consists in a correspondence between a structural pattern and its meaning. The iconic link exists “between the (horizontal) relation(s) on the level of the signifier and the (horizontal) relations on the level of the signified” (Fischer & Nänny 1999: xvii). In 1989 Seiler claimed that “the relevant research on iconicity is still in its taxonomic phase” (Seiler 1989: 165); more than a decade later, terminological confusion in the



Iconicity

field shows that the claim has not yet lost its validity. Neat divisions and straight demarcation lines are proving unrealistic. Scholars of differing theoretical persuasions and practical interests propose various definitions and introduce a profusion of terms. For instance, within diagrammatic iconicity a further subdivision was proposed, allotting to diagrams such properties that basically pertain to imagic relationships: a differentiation is made between inter- and intralinguistic iconicity. The former involves diagrams which link the linguitic sign with an object outside language, and threfore is semiotic par excellence. The latter is revealed through internal relationships within the system itself. In literary studies, an analogous differentiation is made: the two types are called, respectively, first- (or representational) and seconddegree (or intrasystemic) iconicity (terms introduced by Johansen 1996). While the first degree iconicity “mirrors the order of the events [that the signs] refer to in the real world”, “[i]ntra-linguistic iconicity involves a lower degree of iconicity because there is no mirroring of any physical or conceptual structure in the outside world”; “instead there is a relation of similarity between linguistic signs, based on the fact that the concepts they refer to show some similarity” (1999: xxv–v; emphasis in the original). To use once again the classical veni, vidi, vici example, first-degree iconicity will be seen to involve the correspondence between the order of the three signs and the order of the events to which they refer. On the other hand, second-order iconicity, which the dictum also manifests, pertains not to temporal ordering, but rather to the linguistic similarity of the three signs (i.e. their phonetic shape). Recently Winfred Nöth proposed yet another distinction: that between endophoric and exophoric iconicity. To borrow the Nänny and Fischer 1999 title, the former type refers to “form miming meaning”, and the latter — to “form miming form” (Nöth 2000). Nöth’s objections to Johansen’s terms is that they are misleading: what is needed is a distinction between “two modes of reference and not between degrees of iconicity”. Moreover, “endophoric iconicity is intratextual and not intrasystemic” (Nöth 2000: 22). Basically a refinement of the earlier classification, Nöth’s distinction was indeed found to have a practical utility by many scholars — of both literary and linguistric persuasion (cf. articles in Fischer & Nänny 2000; Műller & Fischer 2003; Maeder, Fischer & Herlofsky 2005). But apart from its applicability, the proposal for a refinement of earlier typologies marks an important turn in the theory of iconicity in language. Nöth’s seminal paper points to two crucial aspects of the adapted view: the scalar nature of the phenomenon and its inherently subjective character. Recalling Peirce’s notion of hypoicon (“a degenerate, i.e. an imperfect, or derived, kind of icon”), he claims that “[t]here is thus a scale of iconicity from hypoiconicity, ranging from hypoicons which share only few features with their objects, but at the end of this scale, the genuine icon is a mere abstraction” (Nöth 2000: 19). This is in agreement with the general conviction that the issue concerns various degrees of iconic motivation rather than direct linguistic isomorphism (Haiman 1980; cf. also Östman 1989: 158–9). The second postulate

137

138 Elżbieta Tabakowska

comes as an implication of Nöth’s claim that iconicity pertains to individual texts rather than to the system of language and results from his appeal to the type of iconicity (originally described by Perice) which underlies a successful act of communication: that involved “in the parallelism between the speaker’s and the hearer’s interpretations of the [transmitted] signs”, and which is only possible through reference to the latter’s “past experience” (Nöth 2000: 26). In some theoretical works diagrammatic iconicity is further subdivided into structural diagammatic iconicity and semantic diagrammatic iconicity. The former comes as the result of the three priniciples of iconicity listed in 3.2. above; it was illustrated by examples (1)–(7). The latter arises through processes of metaphorical extension. Metaphor is by definition an instance of diagrammatic iconicity, as it consists in the recognition of a similarity between two different (conceptual) domains. Thus, for instance, the linear sequence of signs within an utterance may be used as an iconic diagram to convey the notion of succession of events in time (and thus constitutes a case of first-degree, or structural, or exophoric diagrammatic iconicity, based upon the principle of sequentiality), but it may also signify such metaphorical notions as continuity and duration, growth or decay, physical or abstract motion. Then the diagram reveals semantic iconicity. Ultimately, such a semantic diagram (or, in terms of George Lakoff ’s theory, a conceptual metaphor; cf. Lakoff 1987) may become an icon of the overall meaning of a literary text (notably a poem, where metaphorical crerativity is usually most strongly manifested). Again, the concept itself has been recurring for a long time in literary theory, which saw literary narrative as an image of a ‘real’ story and emphasized the iconic function of various narrative techniques. Students of literature pointed to the psychological dimension, or the world-creating power, of semantic diagrammatic iconicity, claiming that in literary narrative the “syntatic order represents the order in which things spontaneously arise in the consciousness of the author” (Leech & Short, 1981: 236). The inherently ‘interdisciplinary’ (i.e. ‘the linguistic’ and ‘the literary’) character of semantic diagrammatic iconicity found its full expression in the work of George Lakoff and Mark Turner (1989), where Lakoff ’s cognitive theory of metaphor was applied to a full-fledged linguistic analysis of a wide array of poetic metaphors. All taxonomies, although necessary from the point of view of a theorist, are ultimately a model-theoretic fiction. In the real use of language the borderline between imagic and diagrammatic iconicity is often blurred; it may be argued that in the much-quoted dictum veni, vidi, vici each of the three verbs is an iconic image of the corresponding action (brief and complete), while the entire expression consititutes an instance of diagrammatic iconicity. A modern travesty of the ancient dictum, a journalist’s report from the accident that happended to a well-known Polish sportsman when he went dancing to a discotheque, preserves the diagrammatic but not the imagic iconicity, following the ‘form-miming-form’ formula:





Iconicity

(8) Veni, saltavi, procidi.

[(I) arrived, (I) danced, (I) fell down] (From Gazeta Wyborcza, October 12th, 2001)

Similarly, in the theory of language developed by Langacker the basic postulate for the symbolic character of grammar may be easily rephrased as a merger of structural and semantic (or metaphorical) diagrammatic iconicity. The semantic contrast claimed to exist between (9a) and (9b). (9) a. Bill sent a walrus to Joyce. b. Bill sent Joyce a walrus.

results from the fact that grammar “embodies conventional imagery”, i.e. it does justice to diagrammatic iconicity. The structural iconicity (in (9a) “the morpheme to specifically designates the path followed by the walrus”, while in (9b) “the juxtaposition of two unmarked nominals — Joyce and a walrus — after the verb symbolizes a possessive relationship between the first nominal and the second”) has a semantic value: (9b) “lends added prominence to the configuration that results when the walrus completes its trajectory, namely that which finds it in Joyce’s possession” (Langacker 1991: 13–14). A structural diagram becomes a (grammatical) metaphor. Complex relationships between iconicity and metaphor have constituted an object of vital interest, with cognitive theories providing an important source of inspiration for both literature- and language-oriented analyses. Recent studies cover some new areas of research, sample topics ranging from the application of Lakoff ’s theory of metaphor to iconism in signed poetry (Herlofsky 2003) to the explanation of the dynamic interplay of the two in Japanese haiku in terms of Fauconnier’s theory of blending (Hiraga 2003). As was said earlier, Peirce’s own taxonomy of signs was never meant to imply discreteness of categories. In Peirce’s ‘complete sign’ symbolic, iconic and indexical elements are blended. All three are necessarily fused — both in grammar (i.e. in the system) and in literature (i.e. in individual texts). Synchronic analyses reveal and explain the fusion. Diachronic investigation helps discover the working of dynamic cognitive mechanisms that underlie the rites of passage — from image to diagram, from index to icon to symbol. Therefore the pursuit of iconicity requires departure form the Saussurean principle of the autonomy of synchrony. 3.4  Areas of research In the diachronic perspective, research on iconicity focuses upon the issue of iconic foundations of language (cf. above, reference to Givón 1995) and the evolution of linguistic signs from icons to symbols, whereby forms gradually lose their iconic motivation (Fónagy 2001). In historical linguistics the analogous process has long been known

139

140 Elżbieta Tabakowska

as grammaticalization, or semantic bleaching, whereby full-fledged lexical items or grammatical structures become reduced to functional grammatical markers. The transformation of icons into symbols is considered a fundamental mechanism of language change. The postulate, put foreward in groundbreaking works by Haiman (1980, 1983), fostered the functional approach to the issue. In a diachronic perspective, the principles of iconicity become dynamic forces at work rather than static categories. In recognition of the dynamics, Seiler postulates an ensuing significant change of terminology: he replaces Peirce’s indices, icons and symbols with, respectively, indicativity, iconicity and predicativity (Seiler 1989: 167). The diachronic perspective proves equally promising for students of literature. For instance, the German literary scholar Werner Wolf convincingly argues that the chronological passage from non-iconic, essentially static, landscape descriptions in early English fiction (16th – 18th century) to the dynamized (sequentially) iconic descriptions of later literary epochs can be explained against the general historical and literary background of the time (Wolf 2000). Over the last fifteen years numerous case studies were offered, taking varying theoretical stances and demonstrating various aspects of iconicity at work. They help bridge the gulf separating linguistics from literary study and cover a wide area of language, communication and culture, from the level of speech sounds to non-verbal (especially visual) communication. The focus is mainly on the ubiquity and the multiple functions of iconicity, which is shown to be manifested to a lesser degree in language as a system than in literary (or poetic) texts in particular. It must thus be seen as a continuum, running from fully conventionalized lexicon and morphosyntax to highly creative rhetorical structures. But all manifestations are of equal scholarly interest. Like Molière’s Monsieur Jourdain, we all communicate by means of (at least partially) iconic signs; although we do not realize the fact, “there are forces at work here which we are not normally aware of ” (Norrman 2000: 91). Indeed, iconicity becomes most consciously employed and acknowledged in what has been traditionally defined as literary texts, thus providing a definitional criterion to differentiate between the ‘ordinary’ and the ‘poetic’ language use. But on the other hand, since “literary iconicity is based on a linguistic potential” (Nänny & Fischer 1999: xxvi), it is obvious that research on iconicity should be undertaken jointly by linguistic and literary students. While the application of the linguistic theory of iconicity to literary discourse is a relatively recent development, modern linguistics has seen a renewed interest in some old, long discarded issues. One of the retrieved areas of study is the broad area of sound symbolism. Going beyond the traditional borders of lexical investigations, text linguistics and discourse analysis broadened their field to include such new aspects of language use as, for instance, iconic reflections of the speaker’s emotions in syntax (hence the new notion of emotional syntax; cf. e.g. Lecercle 2000).



Iconicity

In the field of pragmatics, the topic of iconicity has come to be discussed in a broad socio-cultural context. A wider perspective reveals, for instance, the existence of mutual relationships between linguistic or textual forms on the one hand and the development of more general cultural conventions on the other. Looking for the iconic relationship between form and function in language Östman (1989) gives convincing experimental evidence to substantiate a well-known claim that, at least in some languages, the degree of (linguistic) politeness is proportional to the length of the speaker’s utterance. A significant step towards the realization of the program set by Haiman (1985), his work also reveals Haiman’s principle of ‘conflicting motivations’ at work; in the actual language use, iconicity competes with economy, and speakers ultimately choose a compromise that ensures optimal communicative effects. In other words, although iconicity is omnipresent and most varied, it may be overriden by non-iconic principles of communication. In recent research — both in linguistics and in literature — the issue of competing motivations (e.g. the conflict between economy and expressivity, between the requirements of the principle of sequentiality and the principle of end focus, or between ordo naturalis and the mental ordering within cognitive structures) has come to the fore. Investigation into the phenomenon of iconicity itself as well as of the communicative strategies that override iconicity help explain the overall psychological organization of cognition. The human power of imitation is being revealed as a survival strategy and a prerequisite of mutual understanding. The investigation of mutual relations between the cognitive domains of the human mind discloses what is perhaps the highest level of diagrammatic semantic iconicity (kown in the literature as automorphism): a ‘correspondence beween two or more parts of the same system’. For some scholars (e.g. Greenberg 1985), the fact that time, space and discourse deixis are mapped onto the same set of demonstrative words (e.g. the same form may signal the first person, closeness to the speaker and immediate future) is a signal that “these conceptual domains are thought of in the same way” (Haiman 1985a: 4; cf. also Sidnell 1998). It is then natural that in its search for the explanation of cognitive mechanisms scholars should reach not only for the most sophisticated forms of lnguage use (viz. fiction and poetry) but also go beyond verbal communication. In the field of visual iconicity, recent developments include the systematic study of the semantics of gestures. The interest in signed languages gave rise to studies (e.g. Herlofsky 2003, 2005) that reveal the fundamentally iconic character of signed communication. Recent research on the interrelationships between gestures and linguistic signs, inspired by the pioneering work of David McNeill (McNeill 1992), shows that gestures that accompany speech do more than provide mere ‘accompaniment’: they complement spoken messages by supplying additional information, usually in highly iconic and/or metaphorical mode (Antas & Załazińska 2004).

141

142 Elżbieta Tabakowska

4.  Perspectives Since, as was said above, “literary iconicity is based on a linguistic potential”, and since there is no quantitative difference between ‘poetic language’ (the traditional subject of literary study) and the ‘ordinary language’ (traditionally reserved for linguists), future research on iconicity will probably continue to extend its scope, with linguists and literary scholars working together to mutual benefit. Their ultimate common goal is the explanation of man’s cognitive processes. This has always been the goal of of transformational-generative theories (which ocassionally tackle the topic of iconicity, e.g. Newmayer 1992), but it was the arrival of the cognitive theory of language and grammar (as developed by Langacker and his followers) that recognized man’s cognitive abilities in the wide context of perception, information processing and background knowledge about the world. The growing interest in ‘the human factor’, which marks an overall turn towards subjectivity as an element of theoretical paradigms that underlie many of today’s scholarly pursuits, not only brings together theoretical linguistics and literary studies, but also opens new vistas for those who are working in neighbouring fields. The interdisciplinary impact for future research on language functions and use is obvious, and recent developments in the field of pragmatics, especially investigation into discourse organization, look promising. Recent research on visual iconicity makes use of theories and models developed in the domain of visual arts, significantly revealing “the role played by perceptual conventions and cultural codes in processes of iconic signification” (Műller & Fischer 2003: 8). The new focus upon the connections between various types of iconicity on the one hand and culture-specific representations of man’s environment on the other inevitably ties the research on iconicity with cultural studies. In turn, the realization of the role that cultural diversity plays in the reception of various manifestations of iconicity brings in the problem of equivalence in interlingual translation of iconic signs (Tabakowska 2003). This particular aspect has a wider dimension: that of intersemiotic translation, tackled upon in works dealing with what has already been called ‘intermedial iconicity’ (cf. Wolf 2003; Tabakowska 2005). Last but not least, there is the growing interest in the systematic investigation of gestures, which “may tell us more about the iconic foundations of spoken languages and the way iconicity has evolved in them” (Müller & Fischer 2003: 8). The focus on the dynamic, cognitive aspects of iconicity calls for cooperation of philosophers of language, psychologists and cognitive scientists. As a result, the study of iconicity has been developing into a genuinely interdisciplinary enterprise. Access to modern technological equipment and large language corpora makes it all the more worthwhile, increasing its factual and scientific validity. All this, however, is a two-way-traffic: a better understanding of the mechanisms of iconicity means a better understanding of the workings of the human mind, and



Iconicity

therefore it will certainly prove profitable for psychologists, copywriters, translators, language teachers and speech therapists alike. It means a better understanding of a wide array of phenomena — from purely theoretical issues (e.g. grammaticalization in the development of language or the notorious question of formal language universals), through practical applications (e.g. to language teaching or literary criticism) to such seemingly non-theoretical topics as advertisment or business negotiations. In the realm of language, the first steps have been made; most importantly, the underlying assumption has now been accepted by all: “Diagrammatic and metaphorical iconicity is manifest on all levels — phonological, morphological, lexical, syntactic, conceptual, pragmatic — in synchronic and diachronic dimensions, and in ordinary and poetic language” (Hiraga 1994: 19). Its overwhelming presence in other systems of communication is now revealing itself with clarity that also demands general recognition.

References Antas, J. & A. Załazińska (2004). The mental body. Gestures as signs of familiarised concepts. In B. Lewandowska-Tomaszczyk & A. Kwiatkowska (eds.), Imagery in Language: 567–84. Peter Lang. Beghagel, O. (1932). Deutsche Syntax: eine gesichtliche Darstellung. Carl Winters Universitätsbuchhandlung. Bolinger, D. (1977). Meaning ad Form, Longman. Bouissac, P. (2005). Iconicity or iconization? Probing the dynamic interface between language and perception. In C. Maeder, O. Fischer & W.J. Herlofsky (eds.): 15–37. Enkvist, N.E. (1990). Discourse Comprehension, Text Strategies and Style, Aumla 73. Fischer, O. & M. Nänny (1999). “Iconicity as a creative force in language use”. In M. Nänny & O. Fischer (eds.) (1999). xv – xxxvi. Fischer, O. & M. Nänny (eds.) (2000). The motivated sign. Benjamins. Fónagy, I. (1980). La métaphore en phonétique. Didier. ——— (2001). Languages within Language. An Evolutive Approach. Benjamins. Givón, T. (1988). Mind Code and Context: Essays in Pragmatics, Erlbaum. ——— (1995). Functionalism and grammar. Benjamins. Greenberg, J.H. (1985). Some iconic relationships among place, time and discourse deixis. In J. Haiman (ed.): 271–288. Haiman, J. (1980). The iconicity of grammar. Language 56: 515–540. ——— (1983). Iconic and Economic Motivation. Language 59: 781–819. ——— (1985). Introduction. In J. Haiman (ed.): 2–7. Haiman, J. (ed.) (1985). Iconicity in syntax. Benjamins. Herlofsky, W.J. (2003). You See Is What You Get: Iconicity and metaphor in the visual language of written and signed poetry: A cognitive poetic approach. In W.G. Műller & O. Fischer (eds.): 41–62. Hinton, L., J. Nichols & J.J. Ohala (eds.) (1994). Sound symbolism. Cambridge Univ. Press. Hiraga, M.K. (1994). Diagrams and metaphors: iconic aspects in laguage. Journal of Pragmatics 22. 5–21. ——— (2003). How metaphor and iconicity are entwined in poetry: A case in Haiku. In W.G. Műller & O. Fischer (eds.): 317–335.

143

144 Elżbieta Tabakowska Jakobson, R. (1971 [1965]). Quest for the essence of language. In Selected writings, vol. II: 345–359. Mouton. Johansen, J.D. (1996). Iconicity in literature. Semiotica 110. 37–55. ——— (2003). Iconizing literature. In W.G. Műller & O. Fischer (eds.): 379–410. Kefer, M. & J. Van der Auwera (eds.) (1989). Universals of Language. Belgian Journal of Linguistics 4. Lakoff, G. (1987). Women, Fire and Dangerous Things: What Categories Reveal about the Mind. The Univ. Of Chicago Press. Lakoff, G. & M. Turner (1989). More than Cool Reason: A Field Guide to Poetic Metaphor. Chicago Univ. Press. Langacker, R.W. (1988). A view of linguistic semantics. In: B. Rudzka-Ostyn (ed.), Topics in Cognitive Linguistics: 49–90. Benjamins. ——— (1991). Concept, Image and Smbol. The Cgnitive Bsis of Gammar. Mouton de Gryuter. Lecercle, J.-J. (2000). On Markov chains and upholstery buttons: ‘Moi, madame, votre chien…’. In O. Fischer & M. Nänny (eds.): 289–302. Leech, G.N. & M. Short (1981). Iconicity: the imitation principle. In Style in Fiction : 233–43. Lotman, J. (1970). Die Struktur literarischer Texte. Fink. Masuda, K. (2003). What imitates birdcalls? Two experiments on birdcalls and their linguistic representations. In W.G. Műller & O. Fischer (eds.): 77–102. Maeder, C., O. Fischer & W.J. Herlofsky (eds.) (2005). Outside-In — Inside-Out. Benjamins. McNeill, D. (1992). Hand and Mind. What Gestures Reveal about Thought. The Univ. of Chicago Press. Müller, W.G. (1999). The iconic use of syntax in fiction. In M. Nänny & O. Fischer (eds.): 393–408. Mülller, W.G. & O. Fischer (eds.) (2003). From Sign to Signing. Benjamins. Nänny, M. & O. Fischer (2000). Introduction. In O. Fischer & M. Nänny (eds.): 1–16. Nänny, M. & O. Fischer (eds.) (1999). Form Miming Meaning. Benjamins. Newmeyer, F.J. (1992). Iconicity and generative grammar. Language 68(4): 756–96. Norrman, R. (2000). On natural motivation in metaphors. In: In O. Fischer & M. Nänny (eds.): 89–108. Nöth, W. (2000). Semiotic foundations of iconicity in language ad literature. In In O. Fischer & M. Nänny (eds.): 17–28. Östman, J.-O. (1989). Testing Iconicity: Sentence Structure and Politeness. In M. Kefer & J.V der Auwera (eds.): 145–162. Peirce, C.S. (1955 [1902]). Logic and semiotic: theory of signs. In J. Buchler (ed.), Philosophical Writings: 98–119. Dover. Radden, G. (1991). The Cognitive Approach to Natural Language. Series A paper No.300. Linguistic Agency, Univ. of Duisburg. Sadowski, P. (2000). The iconicity of English gl-words. In In O. Fischer & M. Nänny (eds.): 69–88. Schonefeld, D. (2005). Frozen locutions — frozen dimensions: LEFT and RIGHT in English, German and Russian. In C. Maeder, O. Fischer & W.J. Herlofsky (eds.). Seiler. H. (1989). Iconicity in functional perspective. In M. Kefer & J. der Auwera (eds.) : 165–172. Sidnell, J. (1998). Deixis. In Handbook of Pragmatics. R. Simone (ed.) (1995). Iconicity in Language. Benjamins. Tabakowska, E. (2003). Iconicity in literary translation. In W.G. Műller & O. Fischer (eds.): 361–376. ——— (2005). Iconicity as a function of point of view. In In C. Maeder, O. Fischer & W.J. Herlofsky (eds.): 375–387. Tabakowska, E., Ch. Ljunberg & O. Fischer (eds.) Insistent Images. Benjamins. Tabakowska, E. (ed.) (2006). Ikoniczność znaku. Słowo-przedmiot-obraz-gest. Universitas.



Iconicity

Wierzbicka, A. (1995). Adjectives vs. verbs: The iconicity of parts-of-speech membership. In E. Landsberg (ed.): 223–45. Whorf, B.L. (1956 [1940]). Language, Thought and Reality: Selected Writings of Benjamin Lee Whorf. MIT Press. Wolf, W. (2000). Emergence of iconicity in landscape descriptions. In: O. Fischer & M. Nänny (eds.): 323–350. ——— (2003). Intermedial iconicity in fiction: Tema con variazioni. In W.G. Műller & O. Fischer (eds.): 339–360.

145

Information structure Jeanette K. Gundel & Thorstein Fretheim University of Minnesota/University of Trondheim

1.  Introduction Grammars of natural languages offer speakers a variety of options for expressing the same basic informational content. To take a simple example from English, the sentences in (1a)–(1j) can all be used to convey the information that a particular dog chased a particular cat. (1)

a. b. c. d. e. f. g. h. i. j.

The dog chased a cat. The cat, the dog chased. There was a dog that chased a cat. The cat was chased by a dog. A dog chased the cat. It was the dog that chased a cat. What the dog chased was a cat. The dog chased the cat. (As for) the cat, the dog chased it. The dog chased it, the cat.

And so on. The possibilities in (1) can be further multiplied by placing primary stress on different parts of the sentence. It is generally assumed that these different structural and intonational options reflect different ways that a speaker can structure information depending on (a) what she intends the utterance to be primarily about (as distinct from the new information being asserted, questioned, etc. about that entity), and (b) what she assumes the addressee already knows or believes and/or is attending to. In relevance theory terms (Sperber & Wilson 1986/95), the different forms may be said to assist the addressee in identifying the context in which the utterance is to be interpreted, thus reducing the amount of processing effort. The term ‘information structure’ has been used to refer to the different ways that information expressed in an utterance can be structured based on criteria such as these. The term has been used to refer to variation both at the level of meaning and at the level of morphosyntactic and prosodic form used to express that meaning. Within the western grammatical tradition, research on information structural notions can be traced back to at least the second half of the 19th century, specifically to



Information structure

the work of German linguists von der Gabelenz (1868) and Paul (1880). These authors used the term psychological subject for the concept that is now variously called topic, ground, or theme; and they use the term psychological predicate for what is now more commonly called comment, focus, or rheme. Work of the Czech linguist Mathesius in the 1920s (e.g. Mathesius 1928) initiated a rich and highly influential tradition of research in this area within the Prague School that continues to the present day (see Daneš 1974; Firbas 1966; Sgall, Hajičová & Benešová 1973; Sgall, Hajičová & Panevová 1986 inter alia). Also influential has been the seminal work of Halliday (1967), Chafe (1976) and, within the generative tradition, Chomsky (1971), Gundel (1974), Jackendoff (1972), Kuno (1972, 1976), Kuroda (1965, 1972), and Reinhart (1981) inter alia. More recent work will be cited below. 2.  What is information structure? Information structural concepts have been widely associated with the division between given and new information in a sentence. There has been disagreement and confusion, however, regarding the exact nature of this association. Some of the confusion has resulted from conflating two types of givenness/newness. Following Gundel (1988, 1999a,b) we refer to these as referential givenness/newness and relational givenness/newness. 2.1  Referential givenness/newness Referential givenness/newness involves a relation between a linguistic expression and a corresponding non-linguistic entity in the speaker/hearer’s mind, the discourse (model), or some real or possible world, depending on the analyst’s assumptions about where the referents or corresponding meanings of these linguistic expressions are assumed to reside. Some representative examples of referential givenness concepts include existential presupposition (e.g. Strawson 1964), various senses of referentiality and specificity (e.g. Fodor & Sag 1982), the familiarity condition on definite descriptions (e.g. Heim 1982), the activation and identifiability statuses of Chafe (1994) and Lambrecht (1994), the hearer-old/new and discourse old/new statuses of Prince (1992), the levels of accessibility of Ariel (1988), and the cognitive statuses of Gundel, Hedberg & Zacharski (1993). For example, the cognitive statuses on the Givenness Hierarchy below represent referential givenness statuses that an entity mentioned in a sentence may have in the mind of the addressee. in focus > activated > familiar > uniquely identifiable > referential > type identifiable Figure 1.  The Givenness Hierarchy (Gundel et al. 1993)

147

148 Jeanette K. Gundel & Thorstein Fretheim

Gundel et al. propose that human languages have determiners and pronouns which conventionally signal such statuses as part of their lexical meaning. A speaker’s assumptions about the status of some entity in the mind of the addressee thus constrains the linguistic forms that may be approporiately used in referring to that entity and, conversely, the form of a nominal expression facilitates understanding by constraining possible interpretations. For example, the conceptual information directly encoded in the phrase these primitive reptiles in (2) is that the intended referent is some group of primitive reptiles. However, the content words in the phrase alone do not uniquely determine which primitive reptiles are being referred to.

(2) A restudy of pareiasaurs reveals that these primitive reptiles are the nearest relatives of turtles. (M.S.Y. Lee, The origin of the Turtle Body Plan. Science 1993: 1649).

Since the proximal demonstrative determiner this/these codes the cognitive status ‘activated’, it restricts possible interpretations to pareiasaurs, as these are the only activated plural entity at the point when the phrase is encountered. The reference can thus be successfully resolved even without prior knowledge that pareiasaurs are primitive reptiles. Referential givenness concepts have also been shown to play a role in such linguistic phenomena as sentence accent and word order. Greater degrees of referential givenness are associated with weaker stress (cf. Chafe 1976, 1994; Lambrecht 1994, Ladd 1978 inter alia) and also with right attachment/dislocation (Gundel 1988; Ziv & Grosz 1994; Vallduví 1992). 2.2  Relational givenness/newness — Topic-focus structure Relational givenness/newness involves a partition of the semantic-conceptual representation of a sentence into two complementary parts, X and Y, where X is what the sentence is about and Y is what is predicated about X. X is given in relation to Y in the sense that it is independent and outside the scope of what is predicated in Y. Y is new in relation to X in the sense that it is new information that is asserted, questioned, etc. about X. Relational givenness/newness thus reflects how the informational content of a particular event or state of affairs expressed by a sentence is represented and how its truth value is to be assessed. For example, (1a) above may be used to describe a particular event of a dog chasing a cat from the perspective of the dog, where the dog is what the sentence is about and what the speaker is predicating about the dog is that it is chasing the cat. The same event may be described from the perspective of the cat, as in (1d) or (1j), where the cat is what the sentence is about and what the speaker is predicating about the cat is that the dog chased it. A somewhat more complex situation exists when, for example, the speaker assumes the addressee knows that the dog chased something, this entity being what the sentence is about, and the main predication is identification of this entity as the dog, as in (1g).



Information structure

Terms used to denote relational givenness/newness concepts include psychological subject and predicate (e.g. in the works of van der Gabelenz and Paul mentioned above), presupposition-focus (e.g. Chomsky 1971; Jackendoff 1972), topic-comment (e.g. Gundel 1974), theme-rheme (e.g. Vallduví 1992), and topicpredicate (Erteschik-Shir 1997). We will use the term ‘topic’ for what has variously been called psychological/logical subject, theme, and ground; and we will use the term ‘information focus’ for what has variously been called psychological/logical predicate, rheme, and comment.1 Referential givenness/newness and relational givenness/newness are logically independent, as seen in the following examples (from Gundel 1980 and 1985, respectively). (3) A. Who called? B. Pat said SHE2 called. (4) A. Did you order the chicken or the pork? B. It was the PORK that I ordered.

If SHE in (3) is used to refer to Pat, it is referentially given in virtually every possible sense. The intended referent is presupposed, specific, referential, familiar, activated, in focus, identifiable, hearer old, and discourse old. But the subject of the embedded sentence is at the same time relationally new, and therefore receives a focal accent here. It instantiates the variable in the relationally given, topical part of the sentence, x called, thus yielding the new information expressed in (3). Similarly, in (4), the pork is referentially given. Its cognitive status would be at least activated, possibly even in focus, since it was mentioned in the immediately preceding sentence. But it is new in relation to the topic of (4), what B ordered. The two kinds of givenness/newness also differ in other respects. While relational givenness/newness is necessarily a property of linguistic representations, referential givenness-newness is not specifically linguistic at all. Thus, one can just as easily characterize a visual or non-linguistic auditory stimulus, for example a house or a tune, as familiar or not, in focus or not, and even specific or not. By contrast, the topic-focus partition can only apply to linguistic expressions, specifically sentences or utterances and their interpretations. Corresponding to this essential difference is the fact that referential givenness statuses, e.g. familiar or in focus, are uniquely determined by the knowledge and attention state of the addressee at a given point in the discourse. The speaker has no

1.  The relational notion of information focus (as complement of topic) should not be confused with the referential notion ‘in focus’, which refers to the cognitive status of a discourse referent. See Gundel (1999a) for further discussion of different senses of the term ‘focus’. 2.  Uppercase letters here and elsewhere in the paper indicate location of a prominent pitch accent.

149

150 Jeanette K. Gundel & Thorstein Fretheim

choice in the matter.3 Relational givenness notions like topic, on the other hand, may be constrained or influenced by the discourse context (as all aspects of meaning are in some sense), but they are not uniquely determined by it. As the Czech linguist Peter Sgall pointed out a number of years ago, a sentence like There was a soccer game last night could be followed by Poland beat Sweden or by Sweden was beaten by Poland. The latter two sentences could each have an all-focus interpretation, where the whole sentence is a comment on some topic not overtly represented in the sentence, possibly established by the preceding utterance. However, it is also possible in exactly the same discourse context to interpret the first of these sentences as a comment about Poland and the second as a comment about Sweden. Which of these possible interpretations is the intended one depends on the interests and perspective of the speaker. There is still some controversy about the referential givenness properties of topics. Some authors (e.g. Gundel 1985, 1988) have proposed that topics must be familiar, citing the association of topics with definiteness in support of this claim. Others (e.g. Reinhart 1981; Prince 1985) propose that the familiarity condition on topics is too strong, since indefinites can sometimes appear in structural positions typically reserved for topics.4 It is generally agreed, however, that topics must be at least referential. There must be an individuated entity for the utterance, sentence or proposition to be about, and in order for truth value to be assessed in relation to that entity. 3.  How do languages express information structure? 3.1  Information structure and sentence intonation The association between prosody and information structure has been shown to hold in a variety of typologically and genetically diverse languages, and is widely believed to be universal.5 The most consistent way of marking information structure across languages is through placement of a prominent pitch accent within the comment/focus. In English, for example, information focus is coded by what Bolinger (1961) and Jackendoff (1972) call an A accent (the simplex H* tone of Pierrehumbert 1980). Elements within the prosodic domain of the H* accent are

3.  The speaker does of course choose what she wants to refer to, or whether she wants to refer at all; but once this choice is made, the referential givenness status of this choice is predetermined by the hearer’s knowledge and attention state at the given point in the discourse. 4.  See Gundel (1999b) for further discussion of this controversy. 5.  Gundel (1988) notes, however, that one of the languages in the sample she surveyed, Hixkaryana (Derbyshire 1979, cited in Dooley 1982), was reported not to use prosody to mark focus.



Information structure

interpreted as part of the information focus and elements outside that domain are interpreted as part of the topic.6 Lambrecht (1994: 133) provides an especially compelling example of the role of prosody in marking information structure. He notes that most people, when asked to interpret a written sentence like (5) in the absence of any contextual cues, would assign a generic interpretation, where the topic and focus coincide with the grammatical subject and predicate respectively.

(5) Nazis tear down antiwar posters.

One might imagine a context, for example, where (5) is uttered during a discussion about Nazis, where Nazis is the topic and what is predicated about Nazis (the information focus) is that they tear down antiwar posters. Another likely interpretation, which Lambrecht doesn’t consider here, is one where the whole sentence is focus, for example as a newspaper heading, where the topic is simply what happened today. Both these interpretations would be consistent with an H* accent on the direct object (ANTIWAR posters), the default (wide focus) accentual pattern that people normally assume when presented with written sentences in isolation. In fact this sentence was “written with a felt pen across a poster protesting the war in Central America. The poster had been partly ripped down from the wall it had been glued onto” (Lambrecht 1994: 133). Provided with this additional contextual information, the interpretation of the sentence changes, as does the accentual pattern we assign to it. The prominent H* pitch accent now shifts to the subject and the interpretation is: people who tear down antiwar posters are Nazis. The projection of information focus to higher constituents results in information structural ambiguities. Thus, a sentence like (6b), with an H* pitch accent on the direct object, is an appropriate answer to the question in (6a) because it has a possible interpretation where the information focus includes only the direct object. (6) a. What did the dog chase? b. The dog chased the CAT. c. The DOG chased the cat.

But the same sentence also has an interpretation where the information focus is the VP chased the cat as well as an interpretation where the whole sentence is focus, for example as an answer to What happened? This latter interpretation corresponds to

6.  See Steedman (2000) for more detailed discussion. The identification of topic with material outside the domain of focus only holds if topic and focus are complementary relational categories, as we assume here. This position is not shared by all authors. For example, Büring (1999) considers what he calls S-topic to be only a part of non-focal material. Others define topic positionally, for example as the first element in the sentence (Halliday 1967), independent of its focal status.

151

152

Jeanette K. Gundel & Thorstein Fretheim

what Marty (1918) calls a thetic judgment (see also Kuroda 1972), and what Schmerling (1976) calls an ‘all-news’ sentence. However, (6c), with an H* pitch accent on the subject, the dog, would not be appropriate in any of these contexts since it has a so-called ‘narrow focus’ interpretation, where the information focus includes only the subject.7 A new or contrastive topic typically receives some sort of prosodic prominence as well, resulting in sentences with two prosodic peaks, as in (7b). (7) a. What about Fred, what did he do? b. FRED ate the BEANS.

In languages like English and German, the type of pitch accent that marks information focus is distinct from the one that marks a contrastive topic. For example, as an answer to (7a), (7b) would have an H* accent on ‘beans’, and what Bolinger and Jackendoff call a B accent (Pierrehumbert’s complex L + H* tone) on ‘Fred’. But in response to a question like What about the beans, who ate them?, ‘Fred’ would have an H* accent and ‘beans’ would have an L + H* accent.8 However, not all languages distinguish prosodically between information focus and a contrastive or newly introduced topic. Thus, according to Vallduví & Vilkuna (1998: 89), information focus (their ‘rheme’) and contrast (their ‘kontrast’) are “associated with a single high tone accent” in Finnish; and the distinction between the two is coded syntactically rather than prosodically. Similarly, Fretheim (1987, 1992a, 1992b, 2001) argues that there is no particular pitch contour that encodes topic or focus in Norwegian. When a Norwegian unit contains two fundamental frequency maxima for maximum prosodic prominence, either one of them could be the information focus. Thus (8), with a prosodically prominent subject as well as a prosodically prominent direct object, could be a statement about Fred or a statement about the beans. There is no intonational phenomenon in Norwegian that enables the hearer to uniquely identify topic and focus in an utterance of (8). This must be determined by pragmatic inference alone. (8) FRED spiste BØNNENE Fred ate the beans ‘Fred ate the beans’

7.  Focal accent on the subject can, however, project to the whole sentence with certain intransitive predicates, as in ‘all-news’ sentences like The DOOR’s open, Her UNCLE died, My CAR broke down, all of which would be appropriate responses to What happened? or What’s wrong?, where the whole sentence is focus and the topic is not overtly expressed at all. See Ladd (1978); Schmerling (1976); Selkirk (1984), Zacharski (1993) for more detailed discussion. 8.  See Hetland (2003) for a more thorough overview and discussion of the distinction between contrastive topic and information focus.



Information structure

Similarly, the Norwegian sentence in (9) produced with the highest degree of prosodic prominence on de bildene (‘those pictures’) and on etterpå (‘afterwards’) means either (9) ‘Looking at those pictures’ (topic) ‘is something you must postpone till some later time’ (focus), or (b) ‘Afterwards’ (topic) ‘you have to take a look at those pictures’ (focus). (9) Du må se på de BILDENE ETTERPÅ. you must look at those pictures afterwards ‘You have to look at those pictures afterwards.’

3.2  Information structure and morphosyntax All human languages appear to have different syntactic constructions for describing the same event or state of affairs, depending on information structural concepts such as those discussed above. The structure most widely and consistently associated with topic marking across language is one where a constituent referring to the topic of the sentence is adjoined to the left or right of a full sentence comment/focus. Such prototypical topic-comment constructions, exemplified in (1i,j), are presumably found in all human languages, and are relatively unmarked structures in so-called topic-prominent languages like Chinese and Japanese (Li & Thompson 1976). In some of these languages, for example Japanese and Korean, topic or focus may also be marked morphologically (see Kuno 1972; Kuroda 1965; Bak 1977; Gundel 1988; Lee 1999 inter alia). However, languages may differ in the consistency with which information structure is directly mapped onto linguistic form. For example, as noted in Chao (1968), the grammatical predicate of a sentence and what he calls the ‘logical predicate’ (our information focus) do not always coincide in English. Chao illustrates this point with the following exchange between a guide (A) and a tourist (B). (10) A: We are now passing the oldest winery in the region. B: Why?

The source of the humour here is that the English sentence uttered by the guide has two possible interpretations. On one interpretation, the main predicate asserted by the sentence (Chao’s logical predicate) coincides with the grammatical predicate, i.e. are now passing the oldest winery in the region. On the other interpretation, the logical predicate includes only the direct object. The tourist (B) seems to be questioning the first interpretation (we are now passing the oldest winery in the region), but it is the second interpretation that the guide actually intended to convey (what we are now passing is the oldest winery in the region). Chao notes (1968: 78) that the humour would be absent in Chinese because “[i]n general, if in a sentence of the form S-V-O the object O is the logical predicate, it is often recast in the form S-V de shO ‘what S V’s

153

154 Jeanette K. Gundel & Thorstein Fretheim

is O’, thus putting O in the center of the predicate”. In this case, the guide’s intended message would be expressed in Chinese by a cleft construction, which more literally translates as What we are passing now is the oldest winery in the region. Thus, while the different syntactic constructions are available in both languages, the distinction is more consistently exploited in Chinese for mapping information structure directly onto syntactic structure. Similar differences are found even between very closely related languages. For example, while ‘it’-clefts such as those in (1f) exist in all Germanic languages, they are much more frequently used in Norwegian and Swedish than in English (Gundel 2002; Johannsson 2001). Gundel (2002) proposes that this difference is due to a stronger tendency to map information structure directly onto syntactic structure in the Scandinavian languages. Specifically, Scandinavian languages appear to exhibit a much stronger tendency to keep information focus (psychological predicate) out of subject position and to syntactically subordinate presupposed material by encoding it in the cleft clause. Sentence structures like (1b), where an argument of the verb is displaced from its canonical position and appears in sentence initial position seem to be particularly well suited to encoding information structure. But even here, there is rarely a simple mapping between the information structure and syntax. For example, the sentence initial constituent in an example like (1b) may refer either to the topic (e.g. as an answer to What about the cat? Did the dog chase the cat?) or to the information focus (e.g. as an answer to What did the dog chase?). Corresponding to this distinction, the sentenceinitial phrase would also have two different pitch accents in English, but, as noted above, this would not be the case in Finnish or Norwegian, for example. In either case, non-canonical placement of constituents in sentence initial position is not in itself uniquely associated with a single information structural role. Birner & Ward (1998: 95) argue that preposing in English is associated with the more general function of marking the preposed constituent as representing “information standing in a contextually licensed partially ordered set relationship with information invoked in or inferrable from the prior context”. This contextually determined function is stated solely in terms of referential givenness, and is thus independent of the topic-focus distinction. Similarly, the clefted constituent, for example the dog in an example like (1f) typically codes the information structural focus while the open proposition expressed by the cleft clause (x chased the cat in 1f) is presupposed and topical.9 (1f) would thus be an

9.  The equation of presupposition and topic again depends on an analysis such as the one we are assuming here that views topic and focus as complementary relational categories. The equation does not require that the topic be construed as an open proposition rather than an entity (see Gundel 1985). But see Lambrecht (1994) for a different view of the relation between topic and presupposition.



Information structure

appropriate response to Who chased the cat? In English this interpretation would have a prominent H* pitch accent on the clefted constituent. However, not all clefts have only a single prominent pitch accent on the clefted constituent. The accent associated with information focus may also fall within the cleft clause. Hedberg (1990, 2000) argues that in such cases the information focus is on the whole sentence, i.e. these are all news sentences where the information focus includes both the cleft clause and the clefted constituent, as in (11). (11) [Beginning of a newspaper article] It was just about 50 years ago that Henry Ford gave us the weekend. On September 25, 1926, in a somewhat shocking move for that time, he decided to establish a 40-hour work week, giving his employees two days off instead of one. (Philadelphia Bulletin, cited in Prince 1978)

4.  The grammar-pragmatics interface In her classic paper on topic, Reinhart (1981) writes that topics “are a pragmatic phenomenon which is specifically linguistic”. While human languages differ in the manner and extent to which information structural concepts such as topic, focus and various degrees of referential givenness are directly and unambiguously encoded by linguistic form (syntax, prosody, morphology, or some combination of these), all human languages appear to have some means of coding such concepts and categories. Information structure is thus constrained, and in this sense partly determined, by linguistic form across languages. In addition, differences in information structure alone sometimes correlate with profound differences in meaning, with corresponding truth-conditional effects. It is not surprising then that many accounts of information structure have built these concepts into the grammar, as part of the syntax and/or semantics (interpreted by the phonology in the case of prosody) or as a separate information structural component (cf. for example, Lambrecht 1994). The expression and interpretation of information structure cannot be completely reduced to general principles governing human interaction or to other cognitive/ pragmatic abilities that are independent of language, as proposed by Sperber & Wilson (1986/95). At the same time, however, it is evident that not all the phenomena associated with information structure can be directly attributed to the grammar. Topic and focus are pragmatically relevant categories with clear pragmatic effects, including the appropriateness/inappropriateness of sentences with different possibilities for topic-focus interpretation in different discourse contexts. Indeed, the attempt to explain a speaker’s ability to choose among various morphosyntactic and prosodic options and the corresponding ability of speakers to judge sentences with different topic-focus structure

155

156 Jeanette K. Gundel & Thorstein Fretheim

as more or less felicitous in different contexts, has been one of the primary motivations for introducing these categories into linguistic analysis and theory. Contrary to what is sometimes assumed, however, the fact that topic and focus have pragmatic effects does not in itself make them essentially pragmatic. The failure to clearly distinguish between properties of topic and focus that are grammar-driven and those that are purely pragmatic is especially evident in attempts at topic and/or focus identification, which typically involve taking a sentence or part of a sentence and testing its appropriateness in a particular discourse context. Such tests often fail to uniquely identify the information structure of a given sentence, even in the simplest cases. Thus, the fact that the sentence in (12b) would be an appropriate response to the wh-question in (12a) shows that (12b) has a possible topic-focus structure where the topic is Jane or what Jane is doing and the focus/comment is that she is walking her dog. (12)

a. b. c. d. e.

What’s Jane doing? Jane’s walking the DOG. As for Jane, she’s walking the DOG. JANE’s walking the dog. I love dogs.

The fact that someone could report an utterance of (12b) (in any discourse context) as Someone said about Jane that she’s walking the dog (see Reinhart 1981) would provide further evidence for this analysis, as would the fact that (12b) is an appropriate response to What about Jane? (see Gundel 1974) But none of these tests necessarily shows that Jane must be analyzed as the topic of (12b). Even in this discourse context, (12b) could have an all-focus (thetic) interpretation. The failure of such tests to provide a fool-proof procedure for identifying topics has led some authors to question the linguistic relevance of this concept (cf. Prince 1998). But such tests were in fact never intended to serve as necessary conditions for topic or focus. At best, they can help to determine when a particular information structural analysis is possible. Pragmatic tests can’t be used for identifying linguistic categories, because pragmatics is not deterministic. One place where the linguistic context often seems to determine a single topicfocus structure is in question-answer pairs, which is why these provide one of the more reliable contextual tests for relational givenness/newness concepts. Thus, (12b) is judged to be an appropriate answer to the question in (12a) because the location of the prominent pitch accent is consistent with an interpretation where the topic is Jane or what Jane is doing and walking the dog is part of the information focus. But (12d), where the location of prominent pitch accent requires an interpretation where the topic is who is walking the dog is not an appropriate response to (12a). The fact



Information structure

that the judgments here are sensitive to linguistic context has no doubt contributed to the widely held view that topic and focus are pragmatic concepts. However, as noted in Gundel (1999b), questions constrain other aspects of the semantic-conceptual content of an appropriate answer as well. All aspects of the meaning of a sentence have pragmatic effects in the sense that they contribute to a relevant context for interpretation. This much is determined by general principles that govern language production and understanding (Sperber & Wilson 1986/95). Thus, (12e) is no more appropriate as an answer to (12a) than (12d) would be, though the exact reason for the inappropriateness is different. The fact that location of the prominent pitch accent has pragmatic effects does not itself warrant the conclusion that pitch accent codes a pragmatic concept, any more so than it would follow that the difference in meaning between (12b) and (12e) is pragmatic because the two sentences would be appropriate in different linguistic contexts. Assuming a relevance-theoretic pragmatics, Gundel (1999b) proposes that information structure is an essential component of the semantic/conceptual representation associated with natural language sentences by the grammar, as it is basic to the information processing function of language. This representation and the expressed proposition which is an ‘enrichment’ of it, is a topic-focus structure, where the topic is what the sentence is about and the comment/focus is the main predication about the topic. Topic-focus structure is exploited at the grammarpragmatics interface, where information expressed in the proposition is assessed in order to derive ‘contextual effects’, assessment being carried out relative to the topic. Within this framework, it is possible to reconcile the different positions concerning referential properties of topics noted in Section 1 above. A semantic/conceptual representation will be well-formed provided that the topic is referential, and thus capable of combining with a predicate to form a full proposition. This much is determined by the grammar. It follows from what speakers know about the way sentence forms are paired with possible meanings in their language. Utterances with non-familiar topics may fail to yield adequate contextual effects, since assessment can only be carried out if the processor already has a mental representation of the topic. The result is that such utterances are often pragmatically deviant, even if they are grammatically well-formed. Thus, while the referentiality condition on topics is a semantic, grammar-based, restriction, the stronger familiarity condition on topics is pragmatic and relevance-based; it applies at the grammar-pragmatics (conceptualintentional) interface. The important question for future research then is not whether information structure is basically grammatical or pragmatic, but which information structural concepts and properties are purely linguistic, i.e. grammar-driven, and which are derivable from more general pragmatic principles that govern language production and understanding.

157

158 Jeanette K. Gundel & Thorstein Fretheim

References Ariel, M. (1988). Referring and accessibility. Journal of Linguistics 24: 67–87. Bak, S.-Y. (1977). Topicalization in Korean. University of Hawaii Working Papers in Linguistics 9(2): 63–88. Birner, B.J. & G. Ward (1998). Information status and noncanonical word order in English. John Benjamins. Bolinger, D. (1961). Contrastive accent and contrastive stress. Language 37: 87–96. Bosch, P. & R. Van Der Sandt (eds.) (1999). Focus in natural language processing. Cambridge University Press. Büring, D. (1999). Topic. In P. Bosch & R. van der Sandt (eds.): 142–165. Chafe, W.L. (1976). Givenness, definiteness, contrastiveness, subjects, topics, and point of view. In C.N. Li (ed.): 25–56. ——— (1994). Discourse, Consciousness, and Time. Chicago University Press. Chao, Y.R. (1968). A grammar of spoken Chinese. University of California Press. Chomsky, N. (1971). Deep structure, surface structure and semantic interpretation. In D. Steinberg & L. Jakobovits (eds.), Semantics, an Interdisciplinary Reader in Linguistics, Philosophy and Psychology: 183–216. Cambridge University Press. Daneš, F. (1974). Papers on functional sentence perspective. Mouton. Derbyshire, D.C. (1979). Hixkaryana Syntax. Ph.D. Diss. University of London. Dooley, R.A. (1982). Options in the pragmatic structuring of Guarani sentences. Language 58(2): 307–31. Erteschik-Shir, N. (1997). The Dynamics of Focus Structure. Cambridge University Press. Firbas, J. (1966). On defining the theme in functional sentence analysis. In F. Daneš et al. (eds.), Travaux linguistiques de Prague Vol. 1: 267–280. University of Alabama Press. Fodor, J.D. & I. Sag (1982). Referential and quantificational indefinites. Linguistics and philosophy 5: 355–398. Fretheim, T. (1987). Pragmatics and intonation. In J. Verschueren & M. Bertuccelli-Papi (eds.), The Pragmatic Perspective: 395–420. John Benjamins. ——— (1992a). Grammatically underdetermined theme-rheme articulation. ROLIG no. 49, Roskilde University Center. ——— (1992b). Themehood, rhemehood and Norwegian focus structure. Folia Linguistica XXVI/1–2: 111–150. ——— (2001). The interaction of right-dislocated pronominals and intonational phrasing in Norwegian. In W.A. van Dommelen & T. Fretheim (eds.), Nordic Prosody: Proceedings of the VIIIth Conference. 61–76. Peter Lang. von der Gabelenz, G. (1868). Ideen zur einer vergleichenden Syntax: Wort-und Satzstellung. Zeitschrift für Völkerpsychologie und Sprachwissenschaft 6: 376–384. Gundel, J.K. (1974). The Role of Topic and Comment in Linguistic Theory. Ph.D. Diss., University of Texas at Austin. (Published by Garland, 1989.) ——— (1980). Zero NP-anaphora in Russian: a case of topic-prominence. In Proceedings from the 16th Meeting of the Chicago Linguistic Society. Parasession on Anaphora: 139–146. ——— (1985). Shared knowledge and topicality. Journal of Pragmatics 9: 83–107. ——— (1988). Universals of topic-comment structure In M. Hammond, E. Moravczik & J. Wirth (eds.), Studies in syntactic typology: 209–239. John Benjamins. ——— (1999a). On different kinds of focus. In P. Bosch & R. van der Sandt (eds.): 293–305.



Information structure

——— (1999b). Topic, focus and the grammar pragmatics interface. In J. Alexander, N. Han & M. Minnick (eds.), Proceedings of the 23rd Annual Penn Linguistics Colloquium. Penn Working Papers in Linguistics 6(1): 185–200. ——— (2002). Information structure and the use of cleft sentences in English and Norwegian. In B. Behrens, C. Fabricius-Hansen, H. Hasselgaard & S. Johansson (eds.), Information structure in a cross-linguistic perspective: 113–128. Rodopi. Gundel, J.K., N. Hedberg & R. Zacharski (1993). Cognitive status and the form of referring expressions in discourse. Language 69: 274–307. Halliday, M.A.K. (1967). Notes on transitivity and theme in English. PartII. Journal of Linguistics 3: 199–244. Hedberg, N. (1990). Discourse pragmatics and cleft sentences in English. University of Minnesota dissertation. ——— (2000). The referential status of clefts. Language 76: 891–920. Heim, I.R. (1982). The semantics of definite and indefinite noun phrases. Ph.D. Diss., University of Massachusetts. Hetland, J. (2003). Contrast, the fall-rise accent, and information focus. In J. Hetland & V. Molnar (eds.), Structures of focus and grammatical relations. Niemeyer. Jackendoff, R. (1972). Semantic interpretation in generative grammar. MIT Press. Johannsson, M. (2001). Clefts in contrast. A contrastive study of English and Swedish texts and translations. Linguistics 39: 547–582. Kuno, S. (1972). Functional sentence perspective. Linguistic Inquiry 3(3): 269–320. ——— (1976). Subject, theme and the speaker’s empathy. In C.N. Li (ed.): 417–455. Academic Press. Kuroda, S.Y. (1965). Generative grammatical studies in the Japanese language. MIT Diss. ——— (1972). The categorical and the thetic judgement: evidence from Japanese syntax. Foundations of Language 9: 153–185. Ladd, D.R. (1978). The structure of intonational meaning: evidence from English. Indiana University Press. Lambrecht, K. (1994). Information structure and sentence form: topic, focus and the mental representation of discourse referents. Cambridge University Press. Lee, C. (1999). A locus of the interface. Evidence from Korean and English. In K. Turner (ed.), The semantics/pragmatics interface from different points of view: 317–342. Elsevier. Li, C.N. (ed.) (1976). Subject and Topic. Academic Press. Li, C.N. & S.A. Thompson (1976). Subject and topic: a new typology of language. In C.N. Li (ed.): 457–490. Mathesius, V. (1928). On linguistic characterology with illustrations from Modern English. Actes du Premier Congrès International de Linguistes à La Haye: 56–63. (Reprinted in J. Vachek (ed.). (1964). A Prague School reader in linguistics: 59–67. Indiana University Press). Paul, H. (1880). Prinzipien der Sprachgeschichte. Niemeyer. Pierrehumbert, J. (1980). The Phonology and Phonetics of English Intonation. Ph.D. Diss., MIT. Prince, E.F. (1985). Fancy syntax and shared knowledge. Journal of Pragmatics 9(1): 65–81. ——— (1998). On the limits of syntax, with reference to left-dislocation and topicalization. In P. Cullicover & L. McNally (eds.), The limits of syntax: 261–302. Academic Press. Reinhart, T. (1981). Pragmatics and linguistics. An analysis of sentence topics. Philosophica 27: 53–94. Schmerling, S.F. (1976). Aspects of English sentence stress. University of Texas Press. Selkirk, E.O. (1984). Phonology and syntax: the relation between sound and structure. MIT Press. Sgall, P., E. Hajičová & E. Benešová (1973). Topic, focus, and generative semantics. Scriptor Verlag.

159

160 Jeanette K. Gundel & Thorstein Fretheim Sgall, P., E. Hajičová & J. Panevová (1986). The meaning of the sentence in its semantic and pragmatic aspects. Reidel. Sperber, D. & D. Wilson (1986/95). Relevance: Communication and Cognition. Blackwell. Strawson, P.F. (1964). Identifying reference and truth values. Theoria 3: 96–118. Steedman, M. (2000). Information structure and the syntax-phonology interface. Linguistic Inquiry 31(4): 64–689. Vallduví, E. (1992). The informational component. Garland. Vallduví, E. & M. Vilkuna (1998). On rheme and kontrast. In P. Cullicover & L. McNally (eds.), The limits of syntax: 79–108. Academic Press. Zacharski, R. (1993). A discourse pragmatics model of English Accent. Ph.D. Diss., University of Minnesota. Ziv, Y. & B. Grosz (1994). Right dislocation and attentional state. Proceedings of the 9th annual conference and of the workshop on discourse: 184–199. The Israeli Association for Theoretical Linguistics.

Mental spaces Todd Oakley Case Western Reserve University

1.  Meanings are not “in” the words themselves If you visit the Frick Gallery in New York City and you wander into the Living Hall and address the fireplace from the center of the room, examining the Hans Holbein portraits of Sir Thomas More on your left and Sir Thomas Cromwell on your right, you are likely to get the rather odd sensation that Cromwell is watching More. This was certainly my experience, and its intersubjective nature is born out in comments I overheard during a visit. One person remarked “Cromwell is staring at More”, to which a listener paraphrased the same sentiment to their straggling companion, “He’s looking at him”. Learning that Henry Clay Frick himself collected each of these paintings and arranged them in just this way nearly a hundred years ago, I made the following assertion to my companion: “Frick has Cromwell staring at More.” Each of these utterances appears to its speakers as fairly innocuous. None of them, I predict, will be quoted as epigrams of some literary merit in the way one might quote lines from Oscar Wilde’s The Picture of Dorian Gray: How sad it is! I shall grow old, and horrid, and dreadful. But this picture will remain always young. It will never be older than this particular day of June. [...] If it was only the other way! If it was I who were to be always young, and the picture that were to grow old!

Furthermore, if one assumes the perspective of many philosophers from Frege, Russell, and Quine to Kripke, Montague, and Lewis, and of linguists like Katz, Partee, and Keenan, then the above utterances are utterly untrue and devoid of meaning, for each of them takes the view that meaning and truth stem from the underlying structural features of sentences, that the meanings of words are “in” the words themselves. Paintings cannot stare at other paintings, nor can it be the case that a third person can cause a painting to stare at another painting, just as it is not true or meaningful to predicate ‘youth’ of a picture, as Wilde seems to be doing in his gothic novel. It is indeed possible to treat these utterances as deviations from the norm, as semantic and pragmatic anomalies to be consigned to the periphery of usage, and having no significant effect on our models of language and grammar. But doing so leads to several problems, not the least being that such utterances are far from atypical. What is more, it overlooks the very real possibility that meaning does not to any significant degree fall under the jurisdiction of truth conditions, and ignores the radical but no

162 Todd Oakley

less plausible possibility that meanings do not reside in words at all. Meaning and reference are not a part of language itself, or of its grammars in any formal sense, nor is there some hidden layer of linguistic representation that one can uncover. Meaning and reference are the function of interconnected mental spaces, set up as we think, talk, and interact. These mental spaces are representational in nature, but their representational structure follows the dynamic features of how we conceptualize scenes and situations in the natural, social, and symbolic worlds. Without mental spaces, we have no meaning. They are not the products of grammar: grammar consists, instead, of the regular products or formulae for setting up, managing, interconnecting, and navigating through sets of mental spaces. Thus, it is perfectly natural to speak of Cromwell staring at More or of Frick making Cromwell stare at More, with the establishment of a connection between scenes and scenarios governed by representations of persons, and scenes and scenarios governed by the historical facts of real historical beings: the represented. It is furthermore the most natural thing in the world for us to use references typically applied to the domain of the represented as a means of accessing the domain of the representational, and vice versa. Wilde exploits a robust pragmatic connection between representation and represented, such that non-artistic actions from the represented world shape the features of static figures of the representations themselves: an ontological impossibility but an epistemological stroke of genius. This, then, is an article about mental spaces, the underlying cognitive theory behind them, and that theory’s broad influence in Cognitive Linguistics, a school of linguistics that identifies knowledge of language with its use and the vast range of cognitive and cultural resources users tap into when constructing meaning. In these pages, we will learn about the basic architecture of the mental spaces model, as articulated by Gilles Fauconnier and his students and collaborators, through an exploration of how mental spaces theory — first developed as a way of offering solutions to puzzles of ambiguity and reference that have stymied several generations of philosophers, logicians, and some linguists — evolved into a more generalized theory of meaning construction, capable of modeling a range of phenomena relevant to the study of semantics, pragmatics, and discourse and rhetoric. Semantic and pragmatic phenomena discussed in this article include role versus value readings, referential opacity, depictive reference, presupposition, conditionals, counterfactuals and analogical counterfactuals, and deixis. We will conclude this article by applying a mental spaces theory to model meaning construction in written discourse.

2.  What are mental spaces? Open any recent textbook on Cognitive Linguistics, examine the table of contents, and you will find several entries on mental spaces. If you flip to the first section of the



Mental spaces

textbook, you will soon find that a mental space is an important organizing principle for conceptual structure. And since Cognitive Linguists see language and grammar as conceptualization, such a principle must be at the very heart of the enterprise. So, what is a mental space? If we attend to the words of Fauconnier, the locus genus of mental spaces theory, mental spaces are “partial structures that proliferate as we think and talk allowing for fine-grained partitioning of our discourse and knowledge structures” (1997: 11). We learn two things from this definition. First, that a cognitive model of language has to be discourse-based, and, second, that such a model must be sufficiently dynamic to account for the more or less continuous proliferation of these structures. Thus, mental spaces are partial models of present, past, future, possible, impossible, or otherwise imagined states of affairs understood by the cognizer. They are not models of the world; they are dynamic models of the moment-by-moment understanding of states of affairs. As dynamic models of cognitive life, mental spaces “contain” elements that include roles, values, and relations recruited from various semantic frames and defined by Fillmore (1982: 111) as “a system of concepts related in such a way that to understand any one of them you have to understand the whole structure in which it fits”. Mental spaces construct scenes that may integrate several frames in one space or connect elements in one space with a specific organizing frame to another space with a different organizing frame. Let us begin by considering utterances (1)–(3):

(1) Cromwell is staring at More. (2) He is staring at More! (3) Frick has Cromwell staring at More.

Analysis of these utterances demonstrates two basic principles of mental spaces theory: Identity and Access. Fauconnier presents the identity (ID) principle thus: “If two objects (in the most general sense), a and b, are linked by a pragmatic function F(b=F(a)), a description of a, da, may be used to identify its counterpart b” (1994: 3). In a later work, he uses a similar formula for presenting the access principle: “If two elements a and b are linked by a connector F (b=F(a)), the element b can be identified by naming, describing or pointing to its counterpart a” (Fauconnier 1997: 41). Utterances (1) and (2) become meaningful when the same two mental spaces, FG (Frick Gallery) and TE (Tudor England), operate in accordance with these principles, but in subtly different ways. For (1), language users construct a version of space TE with the elements, c and m, which stand for the names Cromwell and More, respectively. In this historical space, these two elements are understood to be important political figures during the reign of Henry VIII, and thus this version of the space is framed by background knowledge of Tudor England (at least for those participants listening to or reading the curatorial commentary). This space is the domain of the represented.

163

164 Todd Oakley

But the present participle staring frames the content of space FG, in which two iconic representations of human figures, elements c′ and m′, are perceived as if one is doing something to the other that can be characterized as staring or surveilling. This space is the domain of representation, but it is also the one that is structured by the immediate perceptual, mental, and interpersonal context of the speaker; thus, the descriptions Speaker’s Reality Space, Origin, and Base Space common within mental spaces nomenclature can be applied to it. As diagrammed in Figure 1a, the mental space configuration relies on an open pragmatic connector (depicted by an arrow) running from elements c and m in TE to their counterparts in FG, a pragmatic relation licensed by the commonplace connection between representation and represented. Notice that in (1), however, the access principle operates from TE to FG, using the surnames to identify the figures in Holbein’s two portraits. By way of contrast, suppose the speaker had uttered (4) instead:

(4) This figure is staring at that other one.

In this case, we have a mental space configuration as depicted in Figure 1b: a single mental space, FG, organized by the same semantic frame for staring and posing but with no discernable cross-space mappings to space TE: elements x and y in FG imply only the status of two-dimensional human figures without explicit historical counterparts.1 The integration of the scene stays within one mental space, with the arrow depicting an agent-patient relation developing therein. The speaker focuses attention on the figures as representations in a present aesthetic and dramatic configuration, leaving their historical identities “off stage”. Utterance (2) can have the same mental space configuration as (1), with the single difference that Cromwell’s identity is already established as common knowledge and accessible via pronominal reference, but the more interesting possibility is that it exemplifies asymmetrical identity. That is, the speaker may know the identity of Sir Thomas More but may not yet know the identity of Sir Thomas Cromwell. In this construal, the speaker notices the aesthetic effect of the arrangement and knows the counterpart of y in TE but does not know the counterpart of x. In short, element x in FG represents a figure of a person, for we can assume the speaker integrates information from the organizing frame of artistic portraiture in this space, but the speaker does not have access to its counterpart in TE. Thus, the access principle applies only to element m in TE and not to c. Figure 1c diagrams this instance of asymmetrical access.

1.  The speaker could very well know the identity of the figures. The point is that such knowledge is not part of the semantics of this particular expression.



Mental spaces Tudor England

Stare m, c

a.

c m

c2 m2 FG

b.

c = Cromwell m = More x, y = portraits FG = Frick Gallery TE = Tudor England = pragmatic connector

TE

Stare y, x x y FG Tudor England

Stare y, x

c. x y FG

c m TE

Figure 1.  Mental-space diagrams for utterances (1)–(3)

Utterance (3) evidences what Brugman (1996: 37–45) calls “depictive sentences”. These sentences depict a state of affairs holding in a space other than the present Reality (or Origin) Space and which the initial NP brings about or is otherwise responsible for. In this case, the present Reality Space R corresponds to what I have been calling FG, Cromwell’s “act of staring” witnessed by the speaker and his interlocutors. This is a space set in the past starring the historical figure Henry Clay Frick, who acquired and arranged these portraits in this way. The speaker is then attributing an intention to Frick, namely that he is the one responsible for the arrangement producing this odd effect. Brugman provides a set of examples of such depictive sentences that cover intentions, predictions, cause-effect relationships, and coercions, among others.

3.  Role and value in reference One important distinction necessary for understanding natural language is that of role vs. value in reference, for role and value readings are often key pieces in the construction of meaning, especially concerning matters of social ontology. In fact, it is virtually impossible to think of how we can live in advanced societies without the mental flexibility of distinguishing between a role and its value. Consider the following attributions to the

165

166 Todd Oakley

President of the United States, some of which only make sense as role or as value readings, while others are ambiguous:

(5) The president has been commander-in-chief since 1789. (6) The president has been commander-in-chief since January 20, 2001. (7) The president supports faith-based initiatives and opposes embryonic stem-cell research. (8) The president needs at least 50 million dollars in campaign contributions to be re-elected.

Sentence (5) only makes sense within a mental space where the institutional role and the requisite background knowledge are active regardless of the value filling it, whereas with (6) both role and value remain equally salient. Sentence (7), in contrast, builds a mental space with a particular person (value) who harbors specific ideological commitments and policy priorities not attributable to the role itself. In terms of mental spaces, we say that the mental-space connection is “open” in (5) and “closed” in (7). Distinct from the previous expressions, sentence (8) requires the construction of a mental space with a similar role focus as in (5), though with the historical constraint that it cannot apply to any occupants of the position before the advent of mass media, and under economic conditions unthinkable in earlier times. Thus, (8) implies that George W. Bush and Barack Obama need this many millions but certainly not George Washington, Thomas Jefferson, or Abraham Lincoln. 4.  Other features of mental spaces theory 4.1  Elements, relations, frames Meaning arises when roles and values interact as elements and relations within mental spaces and across spaces, which, in turn, are organized according to semantic frames, as discussed above. This section will examine these items and their pragmatic connectors and trans-spatial operators. Consider the sentences in (9):

(9) (a) Thomas More is hanging on the wall. (b) It was painted in 1527 by Hans Holbein, the Younger. (c) He refused to sign the Act of Supremacy that declared Henry VIII Supreme Head of the Church of England.

The usual meaning of (9a) entails two mental spaces, TE for the historical figure, Thomas More, and P for the representation of More in the two-dimensional canvas. Space TE connects to space P via a representational connector, such that the element y in P is identified with the element x in TE. In TE, x refers to an historical figure in Tudor England. In P, y functions as the figure of the painting, bringing with it the background knowledge associated with the frame for Portraiture. It includes a subject, a painter, and a painting. The use of the neuter pronoun it in (9b) activates space P,



Mental spaces

for purposes of adding element z and its value, Hans Holbein. In P, element z brings element y into existence, and the meaning of y resonates as an historical figure to the extent that it can be identified with an element x in TE. Sentence (9c) adds content to space TE by specifying actions of x as they relate to a new element w, Henry VIII, and the political framings prompted by the verbs refuse and declare.2 In this space, elements x and w are in an antagonistic relationship that has little relevance to the content of space P, as presently construed in the discourse. Once the elements, roles, and relations are set up in each mental space and accessed via the pragmatic connector that links an image to a real person, the grammar can prompt for adding content in either space, as it does in this case. Other pragmatic connections include metonymy, analogy, and similarity, and as Fauconnier demonstrates (1994: xx), some of these operate only in one direction. For instance, sentence (10), (10) The steak and eggs wants his check,

makes sense only in a mental space in which the organizing frame applies to restaurants, diners, and other dining establishments, where customers are identified by the food they order. But this metonymic link tends to work in one direction, as is demonstrated in (11): (11) *The steak and eggs left without paying. It was inedible.

Once the metonymic connection of food/customer is established, it becomes cognitively difficult to apply the literal anaphoric referent in the next clause. 4.2  Space builders Mental spaces are not to be identified as purely linguistic entities. They are scenes and scenarios or facets of such scenes and scenarios. One function of the grammar is to cue language users and guide them in the construction and conceptualization of these scenes and scenarios. Specific words and phrases act as “space builders” for new mental spaces or as reanimations of a previously constructed space. Temporal and spatial adverbial phrases, such as in 1929, when I was twelve, or in France, and so on, as well as epistemic verbs such as think, believe, and feel are cues for setting, maintaining, or refocusing attention on a specific mental space. For example: (12) In 1918, the lady with gray hair was blonde. (13) That picture is staring across the room.

2.  It is, of course, possible to read the third-person pronoun in (8c) as referring to Hans Holbein, in which case the counterpart of element z in P (z′) would stand in an antagonistic relation to w in space TE.

167

168 Todd Oakley

Sentence (12) uses the temporal adverbial to identify the same person with an attribute in a past space M no longer obtaining in the Reality Space R, while sentence (13) uses the demonstrative pronoun as a means of building a mental space M. It is the figure in the painting that is staring across the room. Once again, background knowledge pertaining to portraits and portraiture permits language users to recruit relevant semantic frames. 5.  Spaces and the problems of reference, ambiguity, and presupposition The initial development of mental spaces theory was to account for the configurations of knowledge and representation associated with general puzzles in language and meaning as they relate to sentence logic. Referential opacity, presupposition, and pragmatic ambiguity exemplify three such puzzles that model-theoretic semantics and logic have strived to explain through formal theories but which turn out to be natural consequences of minds capable of setting up networks of mental spaces and optimizing the content therein. 5.1  Referential opacity Fundamental to reference is a certain point of view, as descriptions can be given from the speaker’s point of view or from some other point of view. A problem in theories of reference, point of view is handled straightforwardly in mental spaces theory. All one has to do is distinguish between a mental space attributed to Speaker’s Reality, R, and some other point of view, M, as in sentences (12–13). For example, sentences (14) and (15) make the same assertion and are both true in the world of Sophocles’ play, but (14) takes the point of view of Oedipus while (15) takes the point of view of the speaker, someone with full knowledge of the situation: (14) Oedipus married Jocasta. (15) Oedipus married his mother.

Thus, sentence (15) builds information from space R, and the reference of mother is transparent, whereas sentence (14) accesses information exclusive to space M, and thus the identification of Jocasta as Oedipus’ mother is not accessible in M. In (14), the referential relation in R is “opaque” in M. Referential opacity is really a matter of setting up mental spaces with contrasting knowledge and beliefs. In both spaces, Jocasta and Oedipus are married, and in both spaces Jocasta is identified as Oedipus’ wife, but in space R and only in space R is Jocasta also identified as his mother. In short, sentences (14) and (15) set up similar mental spaces, with similar elements and relations, and are thus optimized in this regard, except for one crucial piece of information that changes the entire nature of the events depicted in R and M.



Mental spaces 169

5.2  Pragmatic ambiguity As said, mental spaces are models or facets of a cognizer’s world, not the world itself (pace “possible worlds”), and as such space configurations are semantically and pragmatically developed at the same time. Semantics, as Fauconnier argues (1997: 71), provides the instructions for space construction, while pragmatics elicits background knowledge necessary to channel interpretations in one direction or another. Hence, many of the linguistic structures we use to construct spaces are in fact ambiguous, because the semantics leaves open many specific routes of conceptualization that can only be disambiguated as discourse proceeds. As an example, consider this headline from a recent newspaper: (16) Black Women in the US are Shrinking, Data Finds (York Daily Record, December 26, 2008, 11A).

The semantics of this headline prompts for the construction of a mental space that has three possible configurations. In each space, element x corresponds to Black Women In the US, but the pragmatic scales (Fauconnier 1975) of this shrinking presents us with two possible spatial dimensions: (a) height or (b) width, and two possible temporal dimensions: (a) over the course of singular lifetimes or (b) across populations. For instance, it is possible to construe the situation in which the data find that individual black women are shrinking in height as they get older, or, by contrast, to construe a situation in which the data find that individual black women are getting skinnier over time; equally possible is the interpretation that the population of adult black women in the present is skinnier than the population of adult black women (or women in general) a generation earlier, or that the population of adult black women in the present is shorter than the population of adult black women (or women in general) a generation earlier. The last interpretation is indeed the correct one in this context, but we home in on this version of the mental space configuration only after the discourse proceeds. The main point is that the same elements can be framed and construed along multiple possible dimensions and relations, for which the grammar and sentence logic only specify partially. 5.3  Presupposition and optimization One of the thorniest problems in pragmatics has to do with presupposition phenomena. As Fauconnier points out, there is a widespread consensus about which grammatical devices involve the projection of presupposed knowledge and information: definite descriptions, clefts, aspectuals, factives, etc. (1994: 100). Despite this consensus on the correlation of presupposition phenomena and grammatical structures, there is little consensus on the definition of presupposition itself, and how it differs from “uncontroversial assertion”, nor are there adequate semantic and pragmatic accounts of presupposition. Fauconnier demonstrates that a great range of presupposition

170 Todd Oakley

phenomena can be accounted for in a straightforward way within mental spaces theory, thus recasting the “problem of projection” in terms of the internal logic and structure of mental spaces themselves. More specifically, Fauconnier offers a unified semantic account of presupposition, namely that all presuppositions are able to float “from space to space under optimization strategies” (1994: 100). By this he means that mental spaces tend to build off one another and preserve information from one to the other, unless explicitly forbidden from doing so. When a “daughter” space M is set up from its “parent” R, such that M implicitly maximizes similarity with R, and elements in R have counterparts in M, then relations holding in R hold in M, and background assumptions in R also hold in M. Spaces are optimized in this way unless there are explicit stipulations to the contrary (Fauconnier 1994: 91). We will look at two procedures, presupposition float and transfer: (17) Todd believes it’s snowing and hopes it will stop soon.

We have three spaces, Speaker’s Reality, Todd’s Belief Space M, and Todd’s Hope Space H. The belief that it is snowing obtains in space M; what is more, for the content of space H to make any sense it is required that Todd’s belief floats to space H, for Todd’s hope that the snow stops presupposes that it is currently snowing. However, the presupposition that it is snowing does not automatically float into space R, Speaker’s Reality, for the speaker is only expressing what Todd believes and not what she or he thinks is really happening. In fact, one can infer that it may not be snowing in reality because otherwise the speaker would likely not create a belief space, since there would be no need to contrast belief with reality. In contrast, sentence (18) presents a situation in which the presupposition ‘it is snowing’ floats into Speaker’s Reality Space: (18) Todd hopes it will stop snowing soon.

Sentences (17) and (18) demonstrate the basic semantics of presupposition, which is that presuppositional knowledge from one space floats into other related spaces in the network, via optimization, until it meets its opposite. In (17), its opposite is the potential contrast between Todd’s belief space and the speaker’s conception of reality, whereas in (18) no such opposition exists; thus (19) is anomalous, and in fact would be easier for English speakers to parse if the two clauses were joined by a disjunctive marker, but: (19) Todd hopes it will stop snowing soon. It is not snowing.

In mental spaces theory, much of the presupposition phenomena can be accounted for by this general optimization strategy of floating. Another related phenomenon is presupposition transfer, as exemplified in (20):



Mental spaces

(20) Keynes thinks that only a massive influx of government spending in 2009 will stimulate the economy and avert massive unemployment.

This is an example of presupposition transfer, in which the economic theory of John Maynard Keynes in space M transfers to the present Speaker’s Reality Space. The space built by Keynes thinks requires a transfer of the consequences from M to R, even though it is factually untrue that Keynes ever thought one way or another about the present economic crisis. Sentences such as (20) are examples of presupposition transfer through space optimization — transfer of a consequence in the mental space of Keynes’ economic theory to R. In this instance, Keynes’ general theory of employment, interest, and money proscribes specific actions of government spending that have macroeconomic consequences in the present day.

6. Mental spaces and perspective in conditionals, counterfactuals, and deixis Many familiar with mental spaces theory value it for its seemingly straightforward way of handling conditional, counterfactual and analogical counterfactual expressions, as well as deixis — all modes of reasoning and understanding not easily modeled or explained in formal semantics. An additional value of mental spaces theory is that it accounts for the semantics of tense, as tense not only marks “temporal distance” but also the “epistemic distance” between speaker and event; that is, the degree of likelihood that the situation will obtain. In English, the past tense signals a degree of epistemic distance from Speaker’s Reality. Conditionals and counterfactuals are two grammatical structures whose tense structure dictates specific mental space configurations. Deixis, a pervasive phenomenon of language, involves grammatical structures that signal the perspective to be adopted by the discourse participants, whether it be that of the interlocutors or of some other entity, fictional or historical. 6.1  Conditionals and counterfactuals Conditional and counterfactual reasoning is long considered a fundamental semantic and pragmatic problem, which has also been recognized as pervasive by philosophers such as Goodman (1947) but has been difficult for model-theoretic semanticists to accommodate. Fauconnier cautions that we should not regard conditionals and counterfactuals as truth functions (i.e., entailments from an alternative set of premises) but as analogies, or as projections of elements and relations in one mental space to another (cf. Sweetser 1996). Human cognition “dwells” in possibility, in imagining scenes and

171

172

Todd Oakley

scenarios with diverse viewpoints, epistemic stances, and ontological conditions. Consider sentences (21a–c), each of which expresses the same possible state of affairs, but their contrasting tense markers indicate different epistemic stances on a continuum from conditionality to counterfactuality: (21) a. If More has no conscience and then takes the Oath of Supremacy, he will keep his head. b. If More had no conscience and then took the Oath of Supremacy, he would keep his head. c. If More had had no conscience and then had taken the Oath of Supremacy, he would have kept his head.

Spoken by a contemporary of More, sentence (21a) signals a consonance in space R, Speaker’s Reality, and the conditional space M, the situation with the consequence of More staying alive. In other words, the speaker uses a grammar of conditionals in which a positive stance is taken toward the possibility of the situation in M actually coming about. Of course, (21a) can be nothing but counterfactual from a contemporaneous perspective, but in such cases a mental spaces approach would nevertheless claim that the speaker is taking a positive stance toward the situation in M. One could, for instance, imagine a situation of actors and directors discussing the motivations of More and his political enemies during rehearsals of A Man For All Seasons, where the director adopts a stance of historical possibility in order to get the actor to envision the set of possibilities open to the character to be staged. In this case, the relationship between R and M is still counterfactual but the dominant stance is taken from the perspective of M rather than R. Spoken by English historians, sentences (21b) and (21c) adopt increasingly negative stances toward the possibility of the situation in M, with (21b) adopting a negative stance toward the premise of the protasis and (21c) adopting a negative stance toward the premise of the protasis as well as the consequence of the apodosis. In contrast to the above expressions, (22) presents us with a clear case of counterfactuality, for the premise of the protasis is ontologically impossible in principle: (22) If I were you, I’d be ashamed.

Lakoff (1996: 91–123) offers extensive analyses of similar constructions, arguing that the counterfactual scenario entails the construction of a mental space that substitutes the subjectivity of the person being referred to with that of the speaker, such that the intentionality and attitude of the person in M matches the intentionality of the speaker in R, even thought respective selves of each remain distinct. 6.2  Deictic expressions An intriguing and challenging aspect of language analysis is accounting for the reference of nominals in discourse, particularly personal pronouns and deictics. Deictic expressions are canonically associated with placing the speaker and hearer in some spatial



Mental spaces

and temporal relation relative to the current discourse situation, where I refers to the speaker, you to the hearer, here to a spatial configuration proximal to the current discourse situation and there to a spatial configuration distal to the current situation; the same for the temporal adverbs now and then, and the demonstratives this, that, these, and those. But it is well known that deictic elements are not always used to identify items in the immediate discourse situation, and, what is more, forms typically associated with being proximal to the current discourse space can be used to refer to objects and relations in some temporally and spatially dislocated situation. A theory of mental spaces accounts for these uses quite easily, as discussed at length in a study by Rubba (1996: 227–261). Consider this excerpt from an interview she conducted with a local San Diegan: (23) There’s a part of southeast San Diego where you do go down, you see all these Vietnamese theaters and everything in Vietnamese and when I see that I just kind of feel, well, I don’t belong in this place, this is where the Vietnamese people are, I don’t belong here. (Rubba 1996: 227)

The use of the proximal deictics this place and here occurs through a network of mental spaces distinct from R, the origin space of Speaker’s Reality, consisting of the interviewer and interviewee in La Jolla, CA. From space R, the speaker sets up a space for South East San Diego, space VN, organized by the ethnic frame of ‘Vietnamese Neighborhood’ — a place distal from the speaker at coding time. The conjunction when functions as a space builder for a situation embedded in VN, or VN’, a “shifted space” of reported thought. In VN’, but not in VN or R, the speaker is now traveling through this neighborhood. Furthermore, the perception verb see is in the present indicative, suggesting that this situation is iterative and generic — any non-Vietnamese San Diegan would witness the same situation. With the introduction of the verb feel, space VN’ focuses attention on the attitude of the speaker as she travels through this neighborhood. It is space VN’ that sanctions the use of proximal deictics in this segment of the discourse, as the focus of attention is now on the speaker’s attitude within VN’ imputed on anyone with a similar cultural and linguistic background, as if they too were passing through and feeling like an interloper. In other words, the mental space VN’ licenses the use of proximal deictics from the perspective of the value of I in that space. Had the speaker used the distal deictics, the effect would be to maintain the perspective of R rather than VN’. Because the speaker was projecting herself into that mental space, it is perfectly acceptable and intuitive for the speaker to adopt the perspective of the I in space VN’ rather than in R. We find the same construal operations in cases of reported speech, in which the value of I in R is not the same as that of I in the shifted space. For instance, it could be the case that the feelings attributed to I in (23) were those of someone the speaker knows, or fellow San Diegans not of Vietnamese descent. The upshot of Rubba’s account is that deictic centers are framed by the same mental

173

174 Todd Oakley

space operations of accessing, spreading, and viewpoint that govern other semantic and pragmatic phenomena discussed above. 7.  Mental spaces and discourse management So far we have focused almost exclusively on single sentences and discrete utterances, providing just enough contextual background to model the complexity and flexibility of the semantic and pragmatic aspects of natural language use, but recall that mental spaces theory has always pertained to discourse representations and operations. While Fauconnier has not conducted exhaustive discourse analysis, many of his students and colleagues have extended the work into the study of pieces of discourse, written and verbal (see Oakley 1998; Rubba 1996, and Sanders & Redeker 1996). Mental spaces are said to manage the flow of information and reference in a discourse, rendering it coherent and cohesive. Research by Cutrer (1994), Dinsmore (1991), and Takubo (1993) has provided us with some early inspiration for tracking the proliferation of mental spaces in discourse. The multiple grammatical cues for structuring our understanding operate within discourse, as all significant language use happens in discourse; therefore, a theory of meaning construction in natural language has to account for the management of information at this level of granularity. To that end, mental spaces theory develops three basic kinds of space associated with the online construction of meaning in discourse: Base, Viewpoint, and Focus. A mental space serves as a Base when it identifies the initial situation and its focal participants. The Base Space (sometimes but not uniformly equated with Origin or Speaker’s Reality) is, in Fauconnier’s words, the mental space of first mention, “a starting point for a construction to which it is always possible to return” (1997: 49). The Viewpoint is the space from which other mental spaces are accessed, structured, and created, while the Focus is the space which is currently being internally structured, and to which discourse participants are currently attending. It must be emphasized that Base, Viewpoint, and Focus are not mutually exclusive designations within a mental space network, for a single mental space can function as all three, as any two of these, or only as one. A brief illustration of the notions of Base, Viewpoint, and Focus will help flesh out the finer points of mental spaces theory as a theory of discourse. Consider this paragraph taken from a medical case study, published in the periodical Hospital Practice titled “The Suicidal Patient”. Readers are introduced in the previous paragraph to a sixty-three-year-old retired prison guard with a history of bipolar disorder, seizures, and chronic alcoholism, who was prescribed antidepressants and anxiety inhibitors by several physicians without consistent psychological care. This paragraph underscores the severity of the situation.



Mental spaces

The attending physician relates the events that brought the patient into the emergency room and into his care: (24) [1] By the time I saw the patient, it was clear that this hit-or-miss approach was not working. [2] The man was unable to stop drinking and had increasingly cut off contact with the outside world. [3] His wife reported that he remained at home, in a depressed and agitated state, for days at a time. [4] On the morning in question, he had waited until she left for work and then ingested approximately a month’s supply of antiseizure and mood-stabilizing medications including phenytoin, valproate, and paroxetine. [5] Had his wife not sensed that something was wrong and returned to check on him, the suicide would have been completed.

Understanding this paragraph entails the construction of five distinct mental spaces, as depicted in Figure 2. Shut-In h

Examine p, h d p k

h w

Base

M h2 w2

Leave w Wait h Ingest h, m

d = physician p = patient k = next of kin w = wife h = husband m = medications = pragmatic connector

m

AS (alpha) kills h, h h4

h3 w3

Sense w Return w Find h Alive h

AS (beta) Counterfactual Figure 2.  Mental-space network for “The Suicidal Patient” (excerpt)

First is the Base Space, initially containing two participants identified via their roles as physician and patient, d and p. This space is governed by the semantic frame of an examination (diagramed as Examine p,h in the adjacent box, according to conventions established in Fauconnier 1994 & 1997). Sentence 3 adds the third participant in the role of “next of kin”, k, to the Base as the source of the ensuing narrative taking place in the other spaces. This space serves as the Viewpoint for the construction of space M, which aligns with the appearance of the relative clause in

175

176 Todd Oakley

sentence 3. Now in Focus, we learn that p is depressed and has become a shut-in. The next-of-kin role in the Base changes to the “wife”, w, in M, and thus the “patient” also changes to “husband”, h, through semantic entailment. (Notice that the identity mappings from Base to M depict the preservation of the value across different role assignments.) The Focus shifts from the examination room to the home, a domestic drama for which the wife role gains narrative prominence. Sentence 4 builds a new space from the Viewpoint of M, with the fronted adverbial phrase as a stock space builder for narrating an event about to unfold within this domestic scene. Space AS (alpha), Attempted Suicide, is now in Focus with actions and counteractions of the husband and wife unfolding in this space, and with the end result being a “failed” suicide attempt. It is sentence 5, then, that functions as the rhetorical crescendo of this story based on counterfactual reasoning. It does so by constructing two alternative mental spaces. The writer introduces a dependent clause that functions as a protasis optimizing the content of the Viewpoint Space AS (alpha) save for one critical event, so that in space AS (beta), now the Focus of attention, the wife does not return home. With this critical piece of the narrative missing, the writer then draws a logical conclusion in the apodosis: the fifth and final mental space in this network. This is the Counterfactual Space in which the husband’s suicide attempt succeeds, a hypothetical scenario that is contrary-to-fact but eminently plausible. Cued grammatically by the modal verb complex would have, AS (alpha) is established as the Viewpoint and attention is paid to the causal relationship between these two spaces now in Focus. Much more can be said about the relationship between morpho-syntax, semantics, pragmatics, and discourse in the unfolding of the conceptual structures for this particular text, but a broad sketch of the types of mental space involved and the connections between them should suffice to illustrate the nature of a mental spaces approach to discourse: when we construct meaning in language, we typically construct a set of interrelated mental spaces for which dynamic contents pertaining to values, roles, relations, events, actions, and situations can be preserved or altered as the need arises, using the content of one space as the Viewpoint for constructing other spaces that come into and go out of Focus. In this case, we begin with the clinical setting, then focus attention on the domestic spaces relevant to the narrative, the emotional tenor of which reaches a crescendo in the final sentence, which posits the hypothetical situation of a successful suicide, a situation that is also taken to be counterfactual relative to the Base Space. It is possible in principle, but factually false. This contrast between what is the case (hypostasis) and what could be the case (hypothesis) provides much of the underlying drama of the situation. Physicians need to read about suicide attempts in order to be more effective in preventing them. The strong implication from this narrative is that the protagonist will try it again (and likely succeed) without better and more consistent treatment.



Mental spaces

One noticeable gap in the present account is the lack of an explicit set of features for modeling this rhetorical situation, for the Base Space applies to the characters in the narrative and not to the ontologically grounded individuals writing and reading the text, or at least it is not always clear in this theory how we are to treat these elements. Others have proposed specific mechanisms for modeling facets of the rhetorical situation in mental spaces theory itself when analyzing extended stretches of discourse. Brandt & Brandt (2005) call it Semiotic Space, while Oakley & Coulson (2008) and Oakley & Kaufer (2008) call it Grounding, a nomenclature inspired by both Langacker (1999) and Clark (1996). Whatever the small differences between these accounts may be, they each posit a Base endowed with a specific situational relevance, enclosed in a specific setting or discourse. The discourse ground, or common ground, includes common beliefs, needs, interests, and concerns shared between reader and writer, speaker and listener. Much contemporary work in mental spaces theory involves the analysis of discourse in its spoken and written manifestations (see Oakley & Hougaard 2008). 8.  Conclusion This article has presented an overview of several main features of mental spaces theory, focusing on the principles of identity, access, and projection. The fundamental tenet is that an approach to meaning in terms of mental spaces attempts to model the cognizer’s understanding of the world, not the world itself. Mental spaces theory, then, grows out of several concerns in philosophy of language and logic (particularly possible-world semantics), but differs from these approaches in that it treats meaning not as a matter of truth conditions but as a matter of conceptualization. Language provides efficient means of setting up mental spaces and is thus an indispensable source of evidence for how human beings conceptualize meaningful situations. Meaning is not in the words themselves. Grammar provides efficient cues for setting up mental space configurations. Mental spaces theory deals especially elegantly with a variety of semantic and pragmatic phenomena, from reference, presupposition, semantic and pragmatic ambiguity, and the role/value distinction in language and thought, to conditionals, counterfactuals, deixis, and the management of information and attention in discourse. By no means is it the only framework for treating these phenomena, and by no means can it capture everything about them, but it does have the advantage of dealing with these phenomena from a single, unified perspective without having to remake the theory for every new phenomenon. Readers of this article will note that I have steered clear of a later manifestation of mental spaces theory, known as conceptual blending theory, even though much of the material covered here calls for a corresponding treatment, from staring portraits to counterfactuals.

177

178 Todd Oakley

References Brandt, L. & P.A. Brandt (2005). Making sense of a blend: A cognitive semiotic approach to metaphor. In F.J. Ruiz de Mendoza Ibáñez (ed.), Annual Review of Cognitive Linguistics 3: 216–249. John Benjamins. Brugman, C. (1996). Mental spaces, constructional meaning, and pragmatic ambiguity. In G. Fauconnier & E. Sweetser (eds.), Spaces, worlds, and grammar: 29–56. University of Chicago Press. Clark, H. (1996). Using language. Cambridge University Press. Cutrer, M. (1994). Time and tense in narratives and everyday language. Ph.D. Dissertation, University of California, San Diego. Dinsmore, J. (1991). Partitioned representations. Kluwer. Fauconnier, G. (1975). Pragmatic scales. Linguistic Inquiry 6: 357–375. Fauconnier, G. (1994 [1985]). Mental spaces: Aspects of meaning construction in natural language. Cambridge University Press. ———(1997). Mappings in thought and language. Cambridge University Press. Fauconnier, G. & E. Sweetser (eds.) (1996). Spaces, worlds, and grammar. University of Chicago Press. Fauconnier, G. & M. Turner (2002). The way we think: Conceptual blending and the mind’s hidden complexities. Basic Books. Fillmore, C. (1982). Frame semantics. In Linguistic Society of Korea (ed.), Linguistics in the morning calm: 111–137. Hanshin. Goodman, N. (1947). The problem of counterfactual conditionals. Journal of Philosophy 44: 113–128. Lakoff, G. (1996). I’m sorry I’m not myself today. In G. Fauconnier & E. Sweetser (eds.), Spaces, worlds, and grammar: 91–123. University of Chicago Press. Langacker, R.W. (1999). Grammar and conceptualization. Mouton de Gruyter. Oakley, T. (1998). Conceptual blending, narrative discourse, and rhetoric. Cognitive Linguistics 9: 320–360. Oakley, T. & S. Coulson (2008). Connecting the dots: Mental spaces and metaphoric language in discourse. In T. Oakley & A. Hougaard (eds.), Mental spaces in discourse and interaction: 27–50. John Benjamins. Oakley, T. & A. Hougaard (eds.) (2008). Mental spaces in discourse and interaction. John Benjamins. Oakley, T. & D. Kaufer. (2008). Designing clinical experiences with words: Three layers of analysis in clinical case studies. In T. Oakley & A. Hougaard (eds.), Mental spaces in discourse and interaction: 149–177. John Benjamins. Rubba, J. (1996). Alternate grounds in the interpretation of deictic expressions. In G. Fauconnier & E. Sweetser (eds.), Spaces, worlds, and grammar: 227–261. University of Chicago Press. Sanders, J. & G. Redeker (1996). Perspective and the representation of speech and thought in narrative discourse. In G. Fauconnier & E. Sweetser (eds.), Spaces, worlds, and grammar: 290–317. University of Chicago Press. Sweetser, E. (1996). Mental spaces and the grammar of conditional constructions. In G. Fauconnier & E. Sweetser (eds.), Spaces, worlds, and grammar: 318–333. University of Chicago Press. Takubo, Y. (1993). Wanwakanri riron kara mita nihongo no hanjijitu (Discourse management analysis of Japanese counterfactuals). In T. Masuoka (ed.), Nihongo no Joken Hyogen (Conditionals in Japanese): 169–183. Kurosio Shuppan.

Modality Ferenc Kiefer Hungarian Academy of Sciences

1.  Introduction Modality is a central notion in logic and it has attracted considerable attention in linguistics. Logical definitions of modality are based on the notions of necessity and possibility and modal notions must all be traceable to the logical notions of necessity and possibility. Logical treatments of modality are restricted to properties of propositions and exclude all nonpropositional features of sentences. In linguistics, normally a broader view of modality is espoused. Three major approaches can be distinguished. (i) Modality is related to necessity and possibility, it is used to relativize the validity of propositions to a set of possible worlds. On this view, modality is not necessarily propositional, it may also include nonpropositional aspects of the sentence. (ii) Any modification of a proposition comes under the heading of modality. According to this view, volitional, emotive, evaluative modifications, too, belong to modality, in spite of the fact that these modifications are not related to necessity and possibility. (iii) Modality is what the speaker is doing with a proposition. This notion of modality includes (i) and (ii); in addition, it also covers illocution, in particular, the speech acts of imposing obligation and granting permission. On this approach, there is no dividing line between semantics and pragmatics, or, at least much of what has traditionally been considered to be pragmatics belongs to semantics. In the present article we are going to adopt the first one of these three approaches. This choice presupposes a division of labor between semantics and pragmatics. The dividing line between semantics and pragmatics will be drawn in the following fashion. Roughly speaking, semantics deals with those aspects of meaning which are predictable on the basis of lexical meaning and sentence structure. Pragmatics, on the other hand, relates linguistic structure to contextual phenomena, which include time, location, social setting and participants’ roles, on the one hand, and the interlocutors’ strategies, plans, goals and intentions, on the other. 2.  Modality in logic The notion of modality originates in classical logic and it has been associated with necessity and possibility from the very outset. A proposition is necessarily true if its

180 Ferenc Kiefer

truth is established, or guaranteed, by the meaning of the sentences which express them. Such a proposition is true in all possible worlds, it is not grounded in, and cannot be modified by experience. Examples (1a–c) express necessarily true propositions. (1) a. Bachelors are not married. b. A triangle has three angles. c. All girls are female.

Necessity and possibility are interrelated and their relationship can be expressed in terms of negation. If p is necessarily true, then its negation, ¬p, cannot possibly be true; and if p is possibly true, then its negation is not necesssarily true. If the necessity operator is denoted by ☐ and the possibility operator by ◊, then these interrelationships may be expressed as follows. (2) a. ☐p = ¬◊¬p b. ◊p = ¬☐¬p

Various implications can be defined as holding between logically necessary or logically possible propositions and the unmodalized propositions embedded in them: (3) a. ☐p → p b. p → ◊p

If a proposition p is necessarily true then p may legitimately be asserted. And if p is the case, it follows that the weaker proposition, ‘p is possibly true’, too, is true. Until recently, modal logic has been concerned almost exclusively with what is referred to as alethic modality, i.e. with the necessary or contingent truth of propositions. (The term ‘alethic’ is derived from the Greek word meaning “true”.) In addition to alethic modality, two other kinds of necessity and possibility are recognized and formalized by logicians: epistemic and deontic. Epistemic modality is based on knowledge rather than on the content of the proposition. (The term ‘epistemic’ is derived from the Greek word meaning “knowledge”.) Correspondingly, the alethic necessity operator ☐ is replaced by Kap meaning ‘person a knows that p’, and the alethic possibility operator ◊ by Pap meaning ‘it is possible, for all that a knows, that p’ (Hintikka 1962: 3–15). In epistemic logic the reference to person a is normally waived, and the epistemic operators ‘K’ and ‘P’ are interpreted as in (4a,b). (4) a. In light of what is known, it is necessarily the case that p. b. In light of what is known, it is possibly the case that p.

In these formulations no reference is made to the speaker or to the actual drawing of inferences, but only to the evidence which determines the epistemic necessity and possibility. This evidence is thus treated as something objective. (5a,b) are examples of epistemic necessity.



Modality

(5) a. The temperature of the water must be 100°C. b. The sun must rise a few hours earlier in that part of the world.

The implications (3a,b) are equally valid in epistemic logic. Deontic logic is concerned with the logic of obligation and permission. (The term ‘deontic’ is derived from the Greek deon meaning “what is binding”.) The sentence (6a) expresses obligation and the sentence (6b) permission. (6) a. John must leave now. b. John may leave now.

Example (6a) can be paraphrased as (7a) and (6b) as (7b). (7) a. John is obliged to leave now. b. John is permitted to leave now.

In deontic logic obligation is represented by the operator O and permission by the operator P. The two notions are interrelated in the same way as the corresponding alethic operators. (8) a. Op = ¬P¬p b. Pp   = ¬O¬p

‘p must obligatorily be the case’ means that ‘it is not permitted that p not be the case’ and ‘p is permitted’ means that ‘it is not obligatory that p not be the case’. However, the implications corresponding to (3a,b) do not hold in deontic logic: (9) a. Op ↛ p b. p ↛ Pp

In other words, if p is obligatory, it does not follow that p is the case, and if p is the case it does not follow that p is permitted. Epistemic modality is much closer to alethic modality than is deontic modality. Alethic and epistemic modality have to do with the truth of propositions, deontic modality is concerned with the necessity or possibility of acts performed by morally responsible agents. This implies that there is a close connexion between deontic modality and futurity: the acts to be performed refer to a future world-state. Moreover, obligation and permission derive from a source or cause (an institution, an authority, moral or legal principles). The truth of propositions, on the other hand, is not dependent on such sources. 3.  Necessity and possibility in linguistics Alethic modality plays a rather marginal role in linguistics but both epistemic and deontic modality are important notions, though they are interpreted somewhat differently in linguistics.

181

182 Ferenc Kiefer

3.1  Epistemic modality In logic, as we saw, necessity may be defined in terms of the negation of possibility, and possibility in terms of the negation of necessity, that is, neither necessity nor possibility is considered to be more basic. In natural language, on the other hand, epistemic possibility seems to be more basic than epistemic necessity. Notice that, while (10a) is quite possible (‘It is not possible that it is not raining’), (10b) normally would not be taken to express epistemic necessity (‘It is not necessarily the case that it is not raining’, Lyons 1977: 801). (10) a. It can’t not be raining. b. It mustn’t not be raining.

Moreover, there are quite a few modal adverbs, which express epistemic possibility while there seems to be only one modal adverb, namely necessarily, which expresses necessity. Typically, modal adverbs do not belong to the proposition: they are not part of what the speaker asserts, rather they express the speaker’s attitude toward the state of affairs described by the proposition. This can be shown by various tests: (a) negation, (b) question formation, (c) embedding under the verb know, (d) conjunction, (e) if-clauses. The following sentences are all ill-formed (provided, of course, if the that-clause is not read as embedding quotation). (11) a. b. c. d. e.

*Bill is not probably at home. *It is not true that Bill is probably at home. *Is Bill probably at home? *I know that Bill is probably at home. *Bill is probably and perhaps at home. *If Bill is probably at home, I’ll give him a ring.

Any proposition can be negated or queried, propositions can be known, they can be coordinated (if p and q are propositions, then p & q, too, is a proposition), and propositions can occur in if-clauses. Consequently, the ungrammaticality of (11a–e) proves that ‘Bill is probably at home’ cannot be a proposition. Since ‘Bill is at home’ is undoubtedly a proposition, the non-propositionality of ‘Bill is probably at home’ can only be due to the presence of the modal adverbial ‘probably’. Epistemic modality is very often subjective rather than objective (Lyons 1977: 797–800). In fact, whenever a modalized sentence is not propositional (which can be tested, as in (11a–e)), it expresses subjective modality. Consequently, modal adverbials indicate subjective modality. The modals can, may and must are ambiguous between objective and subjective epistemic modality; in most cases, however, the two readings can be kept apart on the basis of the context only. It would seem that the modal receives the objective reading, if stressed, and the subjective reading, if unstressed. For example,



Modality

(12) a. It MUST be raining in London. b. It must be RAINing in London.

Example (12a), in which must is stressed, means that, given what is known, it cannot be the case that it is not raining in London. (12b), in which raining carries stress, on the other hand, is more or less synonymous with ‘It is perhaps raining in London’. In English the difference between the two readings may not always be clear-cut but there are languages in which the distinction between objective and subjective epistemic modality is fully grammaticalized. In Hungarian, for example, possibility is expressed by the suffix -hat/-het, which is attached to the verb stem: játsz+hat ‘can play’, énekel+het ‘can sing’. Focus is positionally defined: either the verb itself is focussed or the focussed constituent occupies the position immediately preceding the verb. Consider now the following two sentences. (13) a.

Pisti JÁTSZhat a kertben. Steve play-can the garden-in ‘Steve can be playing in the garden’

b. Pisti a KERTben játszthat. Steve the garden-in play-can ‘Steve may be playing in the garden’

On the epistemic reading, sentence (13a) (with stress on játszthat ‘play-can’) can only be interpreted objectively: in view of what is known, it is not excluded that Steve is playing in the garden (but he may be doing something else as well). Sentence (13b) (with stress on the locative kertben ‘garden-in’), on the other hand, has a subjective reading only: the speaker infers from what he knows that it is very likely the case that Steve is playing in the garden. (13a) is propositional while (13b) is not (since, for example, it cannot be negated, Kiefer 1986). It is worth noting that, if interpreted subjectively, must and similarly its Hungarian equivalent kell does not express necessity but rather a high degree of probability. Subjective modality is inferential: the speaker infers from what he knows that such and such a state of affairs is likely to hold. Consequently, subjective epistemic modality cannot be glossed as (4a,b), rather it means something like (14). (14) The speaker infers with a certain degree of certainty that p is the case.

Subjective epistemic modality thus expresses different degrees of commitment to factuality and it thus relates to evidentiality (cf. Section 4). Note that (13b), but not (13a), can be the answer to the question ‘Where is Steve?’. In contrast, (13a) can only be interpreted as an answer to the question ‘Can Steve be playing in the garden?’. In asking such a question one is considering the possible courses of events and a simple ‘yes’ would be considered to be an appropriate answer.

183

184 Ferenc Kiefer

The question (13a) is prompted by a discussion of Steve’s possible whereabouts. This is not the case with (13b) where a more informative answer is expected. Normally, one is not simply inquiring about the existing possibilities but about the state of affairs which holds or which is most likely to hold. Sentence (13a), in which stress falls on the verb, is undoubtedly more marked than (13b), in which stress is carried by the constituent immediately preceding the verb, which is the usual position of ‘focus’ in Hungarian. In other words, sentence (13b) is a neutral (indicated by a neutral stress pattern) whereas (13a) is not a neutral sentence (it exhibits a marked stress pattern). This suggests that subjective epistemic modality is more basic than objective epistemic modality. That this may be so is also evidenced by the large number of subjective modal adverbs and particles: ‘perhaps’, ‘probably’, ‘certainly’, ‘surely’, ‘likely’, ‘no doubt’, etc. To summarize, epistemic modality in linguistics seems to be different from epistemic logic in at least two respects: (a) in languages such as English, epistemic modality is possibility-based, (b) a distinction must be made between subjective and objective epistemic modality; subjective epistemic modality seems to be more basic (in the sense of ‘unmarked’) than objective epistemic modality. However, subjective epistemic modality is not propositional hence it cannot be accounted for by means of epistemic logic. Whether this holds for all languages is an empirical question that is “unanswarable on the basis of the evidence available at present” (Lyons 1977: 802). 3.2  Deontic modality It is possible to make a distinction between a subjective and an objective reading of the deontic modalities as well. The sentence (15) may not only be paraphrased by (16a) but also by (16b) (Lyons 1977: 791–793). (15) You must leave. (16) a. You are obliged to leave. b. I (the speaker) hereby oblige you to leave.

Example (16a) is constative whereas (16b) is performative. Only the latter involves the speaker. That is, subjective deontic modality appears as a performative utterance whereas objective deontic modality is expressed by means of a constative utterance. The immediate source of the obligation in (16b) is the speaker who has the authority to impose an obligation on the hearer. The account of performative utterances such as (16b) belongs, however, to pragmatics and will be treated in more detail in Section 7.3. As pointed out in Section 2, deontic modality derives from some source or cause. This source may be some person or institution, and it may be some more or less explicitly formulated body of moral or legal principles. Philosophers, in their discussion of deontic modality, have been mainly concerned with the notions of moral obligation, duty and right conduct. But it should be made clear that the notion of



Modality

obligation is culture-dependent, it may be variously categorized and differentiated in terms of courses and sanctions in different cultures. Different kinds of deontic modality can then be distinguished by specifying the source or cause of obligation (Lyons 1977: 823–825). Let us denote the source of obligation by S; deontic necessity and deontic possibility can now be formulated in the following fashion. (17) a. In view of S it is necessarily the case that p. b. In view of S it is possibly the case that p.

It has been suggested that deontic modality derives from the desiderative and instrumental function of language (Lyons 1977: 826–840). The imperative is the grammaticalized means to issue orders in many languages. On the other hand, granting permission is, in general, not grammaticalized since it is not reflected in the morphology of verbs (i.e. there is no ‘permissive mood’). Moreover, permission can be analyzed in terms of the absence of prohibition but hardly the other way around. This seems to indicate that deontic modality, unlike epistemic modality, is necessity-based rather than possibilitybased (Lyons 1977: 839–840). In linguistics, a clear distinction must be made between the statement of a deontic necessity or of a deontic possibility (between deontic propositions) and the issuing of an order (imposing an obligation) and granting a permission. The former belong to semantics (and, if propositional, are describable in terms of formal logic), the latter are part of speech act theory and must thus be treated in pragmatics. Note that the formulations in (17a,b) do not differ essentially from the ones in (4a,b). Both are statements about necessity and possibility, they differ, however, with respect to the respective basis of the modal notions. 3.3  Some further types of modality In addition to epistemic and deontic modality some further types of modality may be distinguished (Kratzer 1978; Kiefer 1983). Quite often a distinction is made between dispositional, circumstantial and boulomaic modality. (18a) illustrates dispositional necessity and (18b) dispositional possibility. (18) a. Bill must sneeze. b. Bill can sneeze.

The corresponding meanings can be paraphrased as follows. (19) a. In view of Bill’s dispositions it is necessary for him to sneeze. b. In view of Bill’s dispositions it is possible for him to sneeze.

In other words, the source of dispositional modality consists of the various types of dispositions of the person denoted by the subject noun.

185

186 Ferenc Kiefer

A state of affairs can also be possible or necessary in view of certain circumstances, as in (20a,b). (20) a. Bill can relax in his summer house. b. In the mountains pitched roofs must be built.

The corresponding paraphrases are given in (21a,b). (21) a. In Bill’s summer house the circumstances are such that it is possible for him to relax. b. In the mountains the circumstances are such that it is necessary to build pitched roofs.

Sentences (20a,b) exemplify what may be referred to as circumstantial modality. Finally, boulomaic modality has to do with someone’s wishes. Compare (22) a. Charles must be our leader. b. Charles may be our leader.

By uttering the sentence (22a) the speaker’s intention may be to express the fact that in view of his wishes it is necessary that Charles be the leader. Correspondingly, (22b) may be used to express the fact that in view of the speaker’s wishes it is not excluded that Charles be the leader. That is, boulomaic necessity and possibility may be paraphrased as in (23a,b). (23) a. In view of the speaker’s wishes it is necessary that Charles be the leader. b. In view of the speaker’s wishes it is possible that Charles be the leader.

All these modalities share the property with deontic modality that they do not obey the rule (3a). That is, if p is made necessary by someone’s dispositions, by outer circumstances or by someone’s wishes, it does not follow that p is the case. (3b), on the other hand, remains valid. Furthermore, the above-mentioned modalities all derive from some source, as was the case with deontic modality. In the case of dispositional modality, the source is someone’s dispositions; in the case of circumstantial modality, the circumstances of the outer world; finally, in the case of boulomaic modality, the source is provided by someone’s wishes. Consequently, the main dividing line seems to lie between epistemic modality and all the other modalities. The difference between the two types of modality manifests itself in a number of syntactic and semantic properties, as we will see later. The nonepistemic modalities (sometimes including ability) are often referred to as root modality. The distinction between epistemic and root modals was originally proposed as a syntactic one, epistemic modals modifying the whole sentence whereas root modals the predicate phrase. Some researchers use the term ‘dynamic modality’ instead of root modality, other prefer to restrict the former to events that are not conditioned deontically (e.g. circumstantial, dispositional and boulomaic modality) noting that both dynamic and deontic modality are distinct from



Modality

epistemic modality in that they are modalities of events, while the latter is the modality of popositions. Ability, too, would come under the heading of dynamic modality (Palmer 1979: 3–4). But as we shall see later on, ability, though related to modality, is not a modal notion. Formally, epistemic necessity can be construed as logical entailment and epistemic possibility as logical compatibility (Kratzer 1978). The relevant definitions can be found in (24), (25) and (26). (24) p is epistemically necessary if it is entailed by the given epistemic background, i.e. by the set of propositions which represent the speaker’s knowledge activated by the given speech situation. (25) p is epistemically possible if it is compatible with the given epistemic background, i.e. with the set of propositions which represent the speaker’s knowledge activated by the given speech situation. (26) p is compatible with a set of propositions if this set does not contain any proposition p* such that p and p*, or p and p′, where p′ is logically entailed by p, are in contradiction.

Epistemic necessity as logical entailment from an epistemic background is exemplified in (27) (the Examples (27) through (30) are taken from Kratzer 1981: 65–66, also discussed in Öhlschäger 1989: 136–138). (27) I will become mayor. I will become mayor only if I go to the pub regularly. I must go to the pub regularly.

Note that ‘must’ can be omitted in such cases, as shown in (28). (28) I will become mayor. I will become mayor only if I go to the pub regularly. I go to the pub regularly.

Examples (27) and (28) are equivalent, the only difference between them concerns the force of the conclusion. In other words, ‘must’ is not part of the conclusion: its only function is to indicate the force of the conclusion. The situation is different with deontic necessity. Consider (29) (Öhlschläger 1989: 138). (29) Bill’s task is to play the piano whenever there is a party in the house. There will be a party in the house tonight. Bill must play the piano tonight.

In this case, ‘must’ cannot be omitted, it is part of the conclusion. That is, (30) is not a valid conclusion. (30) Bill plays the piano tonight.

187

188 Ferenc Kiefer

This property follows automatically from the invalid implication (8a), repeated here under (31). (31) Op ↛ p

The same holds true for other types of root modality. Consider, for example, (32). (32) I want to become mayor. I will become mayor only if I go to the pub regularly. I must go to the pub regularly.

Once again, ‘must’ is part of the conclusion and does not express logical entailment: (33) is not a valid conclusion. I go to the pub regularly.

In such cases the conclusion is arrived at from the premises (the background) by means of practical inferencing. The general scheme of practical inferences underlying deontic necessity is this (von Wright 1974): (34) a. Speaker wants to bring about the state of affairs p; b. Speaker thinks that s/he can bring about the state of affairs p only if s/he performs act A; consequently c. Speaker sets about to perform act A.

The first premise describes the speaker’s intention, the second one his cognitive attitude, and the conclusion the ensuing act. The transition from intention and knowledge to an act cannot be accommodated within a formal system. Consequently, an important semantic difference between epistemic and root modality is that, whereas the former can be described by logical entailment and logical compatibility, respectively, the latter must be accounted for by means of practical inferences. 3.4  The linguistic tradition Linguists often identify modality with the meaning of syntactically or morphologically defined ‘modal’ expressions. It is possible to say, for example, that modality is whatever is expressed by the set of modal auxiliaries in languages such as English or German. Modality in this case refers to the meanings of modal auxiliaries. However, there are several problems with such a proposal. One of the problems is connected with the fact that in whichever way the set of modals is defined, not all modals will share all the defining properties of this set. For example, the following differences have been noted, among other things, between modals and full verbs in English: (a) modals do not allow for do-support, (b) they undergo inversion with the subject in questions, and (c) they occur with the enclitic not. These properties do not apply consistently to the

Modality 189



‘semi-modals’ need and dare. Moreover, the properties mentioned under (a)–(c) are idiosyncratic properties of these verbs in English. It is particularly difficult to distinguish semantically between modal and nonmodal verbs. This is shown by the fact that English uses auxiliary verbs or other verbs to express the meaning conveyed by modals in German (Calbert 1975: 3–4). Compare, for example, (35a–c) and (36a–c). (35) a. The house is to be sold. b. He wants to come. c. The film is said to be good. (36) a. Das Haus muss verkauft werden. b. Er will kommen. c. Der Film soll gut sein.

Another view of modality identifies it with the speaker’s evaluation of a state of affairs (Bally 1932, 1942). Each sentence consists of two parts: of what is said (the ‘dictum’) and of how it is said (the ‘modus’). The modus can be expressed in a large number of ways. (37)

a. b. c. d.

I think it is raining. It is probably raining. It must be raining. I hope it will be raining.

In the above sentences the dictum can be identified as ‘It is raining’, everything else has to do with modality (the modus). Modality can thus be defined as the speaker’s cognitive, emotive or volitive attitude toward a state of affairs. The main task of the linguist (according to Bally) is to separate the modus from the dictum. Whenever the modus is expressed by a higher predicate, this task is relatively simple. The problem of separating the modus from the dictum becomes more complicated, however, if the expression of the modus is made part of a simple sentence. In such cases, the modus may appear in various forms: as an adverbial (‘It is probably raining’), as a modal verb (‘It must be raining’), or as mark on the verb (mood, tense), which may be quite complex (‘He would never have left us’). According to this view, then, all cognitive, emotive, volitional modifications of a propositional content count as expressions of modality. This view of modality is certainly too broad and has to be narrowed down appropriately, as we shall see Section 5. In principle, it is also possible to define modality as any modification of a proposition. In this case also negation would fall under the scope of modality. On the basis of the definition of modality to be proposed in Section 5, however, we will not consider negation a modal category. Moreover, if modality is conceived as ‘any modification of a proposition’, then ‘temporal modalities’, too, would count as a subcategory of

190 Ferenc Kiefer

modality: ‘It was yesterday that John visited Bill’, ‘It is often the case that John visits Bill’. Once again, however, by adopting a more restrictive definition of modality, ‘temporal modalities’ do not come under the heading of modality.

4.  Evidentials Evidentials are markers that indicate the source and the reliability of the speaker’s knowledge. They may also be characterized as expressing the kind of evidence a person has for making factual claims (Anderson 1986). A first distinction is generally made between direct evidence and indirect evidence (Bybee 1985), i.e. whether the source of the speaker’s information is of a primary or a secondary nature. Direct or attested evidence may be marked as involving the visual sense, the auditory sense, and/or one of the other three senses. There are two types of indirect evidence: reported and inferred. Reported evidence may be specifically marked as second-hand or thirdhand (hearsay), or as part of the oral literature (folklore); and inferred evidence may be specifically marked as involving either observable evidence (results) or a mental construct only (reasoning) (Willett 1988: 56–57). In Kashaya (a Hokan language), for example, direct visual evidence is marked by -y(a), auditory evidence by (V)n(na), and general sensory evidence by -w(a). Unspecified direct evidence is also separately marked by yow(a) or -miy(a), used in personal narratives of recent or remote origin, respectively. The suffix -do marks all reported evidence and inference from experienced results is marked by -q(a) (Oswalt 1964). The relevant examples can be found in (38a–f). (38) a.

momaay ran in EV ‘(I just saw that) he ran in’

b. mu haya cahnonnam ?dog sound EV response   ‘(That’s why) we hear the dog(s) barking’ c.

s’ihtayacma cahnow bird pl subj sound EV ‘Birds sing’

d. men s’i yi?ci?timiy thus do pl hab neg EV ‘They never used to do that in the old days’ e.

qacuhse hqamac’ke?domta grass game play fut assertion EV response ‘I was told they will play the grass game’



f.

Modality

cuhnii mu?aq ‘Bread has been cooked (I can smell)’

In Tuyuca, a Tucanoan language, evidentiality is the criterion that determines which one of a complex set of paradigms for person/tense suffixes will be used. Two direct (visual and auditory) and three indirect (second-hand, inference from results, other inference) sources of evidence are distinguished, yielding five corresponding person/ number paradigms. Visual evidence is marked as in (39a), auditory evidence as in (39b), second-hand evidence as in (39c), inference from results as in (39d), and other types of inference as in (39e) (Willett 1988: 72–93). (39) a.

diiga apewi ‘He played soccer (I saw him play)’

b. diiga apeti ‘He played soccer (I heard, but did not see, him)’ c.

diiga apeyigɨ ‘He played soccer (someone else told me)’

d. diiga apeyi ‘He played soccer (I see evidence for it)’ e.

diiga apehiyi ‘He played soccer (it’s reasonable to assume so)’

However, evidential markers do not only occur in ‘exotic’ languages. It is known, for example, that in both Georgian and Turkish, the aorist implies that the speaker has direct evidence, and one of the uses of the perfect covers both kinds of indirect evidence. In Turkish the unconjugated past tense suffix -dı indicates direct evidence and the suffix -mıš indirect evidence. Compare (40) a.

Ahmet geldı. ‘Ahmet has come (I saw/heard it)’

b. Ahmet gelmıš. ‘Ahmet must have come’

German has no fewer than three devices to indicate what is said (Palmer 1986: 71–79). First, the subjunctive may be used to indicate that what is said is not part of the speaker’s own statement. For example, (41) Er habe sich von ihm bedroht gefühlt. ‘He felt himself to be threatened by him’

where the subjunctive form habe indicates the speaker’s distance. Secondly, the modal verb sollen can be used to mean ‘it is said that…’, e.g. (42) Er soll steinreich sein. ‘He is said to be extremely rich’

191

192 Ferenc Kiefer

Thirdly, the modal wollen can be used to indicate what the subject of the sentence claims or says, e.g. (43) Er will eine Mosquito abgeschossen haben. ‘He claims to have shot down a Mosquito (plane)’

This means that in German modality is partly based on necessity and possibility, and partly on evidentiality. The source of evidence indicates the speaker’s commitment to the truth of what he is saying. Attested evidence is more or less equivalent with ‘I know for sure’, reported evidence with ‘it is possible’ and inferred evidence with ‘it is likely the case’. It can be argued that modality can be based on possibility and necessity, on the one hand, and on evidentiality, on the other. This gives rise to two different modal systems and languages may vary according to which of these two systems they exploit. English seems to be a language which has only possibility/necessity based modality, Kashaya and Tuyuca indicate modality by means of evidentials, Turkish and German have a mixed system (Palmer 1986: 51–57).

5.  A possible synthesis Informally, modality can be characterized by locutions such as ‘envisaging several possible courses of events’ or ‘considering the possibility of things being otherwise’. Things might be, or might have been, other than they actually are, or were. To conceive of a state of affairs being otherwise is to conceive of its being true or real in some nonactual world(s), or true or real in some state of the actual world at a point in time other than the present moment (Perkins 1983: 6–7; Kiefer 1987: 89–90; Kiefer 1992: 2515). The essence of modality consists in the relativization of the validity of a proposition to a set of possible worlds. Talk about possible worlds can thus be construed as talk about ways in which people could conceive the world to be different. Modality in logic, as we saw, is based on the concepts of possibility and necessity. The person who utters the sentence ‘It is possible that Bill finished his paper yesterday’ does not normally know for sure whether Bill finished his paper yesterday, though it is not excluded, this is one of the possibilities. Possibility thus means that the proposition in question is true in at least one possible world. On the other hand, a sentence such as ‘It is certain that Bill finished his paper yesterday’ means that whichever of the possible worlds comes to be realized, Bill finished his paper yesterday. That is, Bill finished his paper in all possible worlds. In order to account for subjective epistemic modality in terms of possible worlds the notion of graded possibility is called for (Kratzer 1981: 46–51). The speaker does not know what the real world is like, but he can draw conclusions from the growing



Modality

evidence available to him. At any time, this evidence is compatible with a set of worlds which could be the real world. This view of modality also covers evidentiality. Attested evidence is equivalent with ‘it is certain’, hearsay, second and third hand evidence represent various degrees of possibility. Modality as the relativization of the validity of a proposition to a set of possible worlds excludes, however, a number of things which are often mentioned in connection with modality. For example, from the three types of modifications (cognitive, emotive, volitional) only some cognitive modifications count as expressions of modality. In expressions such as ‘fearing that p’, ‘be astonished that p’, ‘hoping that p’, ‘wanting p’ the higher predicate does not restrict the validity of the proposition p hence emotive predicates do not come under the heading of modality. Factivity, too, is alien to modality. Thus, the factive evaluative predicates such as ‘it is good’, ‘it is bad’, ‘it is amazing’, ‘it is phantastic’ express the speaker’s qualifications of a state of affairs. Since the primary function of a factive predicate is to comment on, or evaluate, an aspect of the world that is, rather than of some world that might be or might have been, evaluations do not come within the scope of modality. Negation is not a modal category either. In order to paraphrase ‘It is not true that Bill is sick’ the speaker would have to say that the proposition ‘Bill is sick’ has to be evaluated in those possible worlds in which this proposition is false (Perkins 1983: 47–48). Or, to put it differently, the negation of an event would have to be characterized in terms of the occurrence of the event being made relative to its nonoccurrence, which is a strange consequence. As to illocutionary verbs, these verbs refer to an act and are not used to relativize the validity of a proposition. Consider the utterance ‘I assert that all men are mortal’. What does it mean to say that something is compatible with what I assert? There is no meaningful answer to this question. Illocution is alien to modality (Kiefer 1987: 87). The distinction between modality and modulation reflects the same point (Halliday 1970: 336–338). The same holds true for perlocution. The effects that an act may bring about in the addressee cannot affect the act itself, i.e. the proposition which describes the act is independent of the eventual consequences of the particular act. We must also exclude from the realm of modality the so-called ‘modal particles’ in German (and similar languages, Palmer 1986: 45–46). Some of these particles are listed with English glosses in (44). (44)

a. b. c. d. e.

ja doch denn schon wohl

‘truly, why, don’t you see, you know’ ‘after all, though, just, truly, surely…’ ‘evidently, as is well known, as I learn…’ ‘never fear, no doubt, surely, as a matter of course’ ‘indeed’

193

194 Ferenc Kiefer

These particles carry various pragmatic functions, they are often comments on a state of affairs but they do not relativize the validity of the proposition. This is also true of the particle vielleicht ‘perhaps’ in its function as a pragmatic particle (and not as a sentence adverbial). The difference between the two meanings can clearly be seen in (45a,b). (45) a.

Das war vielleicht ein Durcheinander! ‘What a mess’

b. Er ist vielleicht noch zu Hause. ‘He is perhaps still at home’

The particle vielleicht in (45a) expresses astonishment, in (45b) it is a modal adverbial (Thurmair 1989: 192–195). Similar things are true of the particle wohl, which can function either as a sentence adverbial or as a pragmatic particle. It can be replaced by vermutlich ‘presumably’ in the former case only (Thurmair 1989: 139–145). On the reading of ‘presumably’ wohl is, of course, modal. In sum, then, modal expressions include a. b. c. d. e. f. g.

modal auxiliaries such as can, may, must, will, shall; modal adverbs such as certainly, maybe, perhaps, probably, supposedly; modal adjectives such as possibly, necessary, sure, probable, certain; modal nouns such as possibility, necessity, probability, likelihood; modal verbs such as believe, think, presume, suppose, suspect; mood; some evidentials (particles, affixes).

6.  Syntactic treatments of modality The syntactic description of modals is normally restricted to modal verbs. The hypothesis that modal auxiliaries should be analyzed as main verbs was put forward as early as at the rise of generative semantics (Ross 1969). Several arguments can be advanced in favor of the treatment of modal auxiliaries as main verbs. Thus, for German it was pointed out that genuine auxiliaries (as, for example, haben ‘have’ and sein ‘be’) do not exhibit scope ambiguities with negation, whereas modal auxiliaries do (Öhlschläger 1989: 80–98). For example (46) Fritz hat nicht geschlafen. ‘Fritz has not slept’

can adequately be paraphrased as (47). (47) Es ist nicht der Fall, dass Fritz geschlafen hat. ‘It is not the case that Fritz has slept’



Modality

On the other hand, scope ambiguities are quite normal with modal verbs. For example, (48) Fritz will/möchte nicht kommen. ‘Fritz does not want to come’

can either be interpreted as (49a) or as (49b). (49) a. It is not the case that Fritz wants to come. b. Fritz wants not to come.

In other words, the modals will and möchte have both sentence and constituent scope. The same holds true of ordinary full verbs. (50) Fritz rät nicht zu kommen. ‘Fritz does not advise (me) to come’

which can be interpreted as either (51a) or (51b). (51) a.

Es ist nicht der Fall, dass Fritz zu kommen rät. ‘It is not the case that Fritz advises (me) to come’

b. Fritz rät folgendes: nicht zu kommen. ‘Fritz advises (me) this: not to come’

The same seems to hold true for adverbial modification. Once again, in the case of genuine auxiliaries there is no scope ambiguity with temporal adverbials such as today, yesterday, now, etc., whereas scope ambiguities may arise with modal verbs. For example, (52) Fritz darf jetzt kommen. ‘Fritz may come now’

may be paraphrased as either (53a) or as (53b). (53) a. What Fritz is allowed to do is to come now. b. It is now the case that Fritz is allowed to come.

The problem is, however, that though scope may be a good test in German it does not seem to work in other languages. Thus, for example, in English can, could, need seem to have only a wide scope reading whereas might and must only a narrow scope reading. Moreover, the root modal may has a wide scope reading whereas epistemic may a narrow scope reading (Miller — Kwilosz 1981). As to scope the only valid generalization seems to be that subjective epistemic operators always have wide scope. Since epistemic and nonepistemic modals have different semantic properties, syntacticians have long since cherished the idea of associating the two types of modality with two different syntactic structures. There is some indication to consider epistemic modals of type S/S, i.e. of being sentence operators and root modals of type VP/VP, i.e. as operators whose scope is restricted to the predicate phrase. However, this does

195

196 Ferenc Kiefer

not seem to work, as pointed out above. According to one of the more recent proposals, nonepistemic modals should be considered control verbs and epistemic modals raising verbs (Stechow & Sternefeld 1988: 428). That is, the German verb können ‘know, can, may’, if used nonepistemically, is described as a predicate of a matrix clause under which an infinite sentence with an empty subject is embedded. The empty subject must be coreferent with the subject of the main clause, that is, the latter controls the former; for example, (54) Fritz kann gut kochen. ‘Fritz can cook well’

which is assigned the structure (55). (55) [Fritzi [CP Proi gut kochen] kann]

In the case of epistemic können, the underlying structure is different. The sentence (56) Fritz kann gut gekocht haben. ‘Fritz must have been a good cook’

is analyzed as (57) [e [CPFritz gut kochen] kann]

Once again, the modal embeds an infinite clause (a CP) but the subject of the main clause is empty (e). The subject of the embedded clause, Fritz, gets raised into this position by a raising rule. The above syntactic account explains certain differences between epistemic and nonepistemic modals. For example, it accounts for the ‘transitive-like’ behavior of root modals, which contrasts with the ‘intransitive-like’ behavior of epistemic modals. Compare (58) Ottokar muss singen, und das musst du auch. ‘Ottokar must sing and you must do it, too’ (59) *Ottokar muss Krebs haben, und das musst du auch. ‘Ottokar must have cancer and you must have it, too’

While (58) is fully grammatical, (59) is unacceptable (Ross 1969: 85). It would seem, however, that the view that the two types of modals can systematically be associated with two different syntactic structures cannot be maintained. Though some modals in German can reasonably be analyzed as raising verbs (dürfen, können, mögen, müssen, and sollen) and others (wollen, möchte) as control verbs, the syntactic analysis seems to be quite independent of their modal meaning (Öhlschläger 1989: 54–131). In other words, both types of modals can be interpreted as epistemic or as root modals: thus the distinction ‘epistemic modal — root modal’ does not correlate with the distinction ‘raising verb — control verb’.



Modality

In sum, then, it would seem that syntactic differences do not correlate systematically with semantic differences. Neither scopal properties nor the opposition ‘raising verb — control verb’ can be used to provide a systematic syntactic foundation for the relevant semantic distinctions.

7.  Modality and pragmatics Traditionally, research on the pragmatics of modality has centered around two problems: on the ambiguity of ‘possible’ and on the illocutionary meaning of modal verbs. More recent topics include the relationship between ability and possibility and the role of conversational implicatures in the grammaticalization of modality. 7.1  Two readings of  ‘possible’ The first problem concerns the relationship between the two senses of possibility. It has been known since at least Aristotle that there are two senses of ‘may’: on one interpretation ‘may’ is an entailment of ‘must’ and does not allow the interpretation ‘may not’. On the weaker interpretation, on the other hand, ‘may’ admits both possibilities, as in (60). (60) John may be there and he may not be there.

The relation between the two senses of ‘may’ is usually accounted for in terms of a scalar Quantity implicature. The argument runs as follows. The assertion of the lower, less informative statement ‘◊p’ instead of the higher, more informative statement ‘☐p’, conversationally implicates the negation of the higher statement, i.e. ‘¬☐p’. When ‘p’ is possible yet not necessary, ‘¬p’ is possible too. Hence the weaker sense of possibility is (the result of) an implicature (Horn 1989: 211–216; van der Auwera 1996). Consequently, pragmatic means (aspects of the speech event) are called for in order to account for the relationship between the two senses of ‘may’. 7.2  The illocutionary meaning of modal verbs The second more traditional research area is the inquiry into the ‘illocutionary act potential’ of modal expressions. Such an investigation requires, in general, a discourse analytic approach. It has been observed, for example, that can/could is often used in a quasi-imperative manner, to suggest a course of action to the addressee. The instruction can be made more polite by using could (Quirk et al. 1985: 222). (61) a. You can sit here until I get back. b. You children could help me move these chairs.

197

198 Ferenc Kiefer

By turning the statement into a question, the speaker changes the suggestion into a rather polite request: (62) Can/could you (please) check these figures?

The modal will may be used, among other things, to express intention, willingness and insistence, as exemplified in (63a–c) (Quirk et al. 1985: 229). (63) a. I’ll write as soon as I can. b. I’ll do it, if you like. c. If you will go out without your overcoat, what can you expect?

In German some modal formulae are typical of certain discourse types: du kannst ‘you can’ is used for suggestion or advice, du musst ‘you must’ for the explanation of a rule, man kann ‘one can’ for the explanation of the utilization of utensils, ich wollte sagen ‘I wanted to say’ for calling for attention, soll ich? for an offer of help, etc. (Wunderlich 1981: 15). In all these cases we have to do with conventionalized formulae. Modal verbs can be classified according to the place they occupy in the action process. Consider, for example, the sentences in (64) (Wunderlich 1981: 46–48). (64) Robert mag/möchte/soll/will/wird/kann/darf/muss/braucht nicht das Zimmer anstreichen. ‘Robert may/would like to/has to/wants to/can/is allowed to/must/need not paint the room’

The whole action process consists of various phases: motivation, goal, planning, decision, realization/execution, achievement/result. Each phase can be associated with an assessment or evaluation. Some modal verbs are used to refer to the various phases of the action process, others express the speaker’s evaluation. For example, mögen ‘may’ indicates preference for the action at hand; möchten ‘would like, want’ indicates the speaker’s motivation or the need of the action; werden indicates the fact that the decision concerning the action has already been taken, the execution is impending. A detailed analysis of the use of modals in action processes along these lines using class room interactions can be found in Redder (1984). Modality occurs quite often in bound utterances (Kiefer 1995) carrying various discursive functions, as exemplified by the following Hungarian utterances. (65) Azt én nem tudhatom. that I not know+can+I ‘How should I know?’ (66) Mondhatom finom alak vagy! say+can+I fine figure are+you ‘You’re a nice lot you are!’



Modality 199

(67) Én mondhatom! I say+can+I ‘I can assure you’ (68) Gondolhatjátok, se ebéd, se semmi. think+can+you neither lunch nor nothing ‘Just think, there was neither lunch nor anything else’ (69) Hogy gondolhatsz arra, hogy ezt megteszem? how think+can+you that+on that that will+do+I ‘How do you expect me to do it?’

The verbs which occur most often in such expressions are tud ‘know’, mond ‘say’ and gondol ‘think’. The verb forms contain the possibility suffix -hat ‘can’ but none of the utterances (65)–(69) expresses possibility, as becomes clear from the English equivalents. Rather, they carry various illocutionary and discurse functions: distance, detachment (65), ironic reproach (66), emphasis (67), fulfillment of an expectation (68), indirect refusal (69). Quite often modal expressions are used in the service of politeness strategies. In particular, they seem to play an important role in the strategy termed ‘negative politeness’, which is defined as ‘redressive action addressed to the addressee’s negative face: his want to have freedom of action unhindered and his attention unimpeded’ (Brown & Levinson, 1978: 134). For example, (70a) is more polite than (70b) because it allows the addressee a wider range of options (Perkins 1983: 118). (70) a. You may stop writing now. b. You must stop writing now.

7.3  Deontic speech acts It was pointed out in 3.2. that obligation and permission are pragmatic rather than semantic notions. In some recent studies on modality a distinction is being made between ‘agent-oriented modality’ and ‘speaker-oriented modality’ (Bybee 1985: 165–169; Bybee, Perkins & Pagliuca 1994: 177–179). The former encompasses all modal meanings that predicate conditions on an agent with regard to the completion of an action referred to by the main predicate such as obligation, desire, ability, and permission; the latter, on the other hand, represents speech acts through which a speaker attempts to move an addressee to action. Accordingly, (71) is an expression of speaker-oriented modality and (72) an expression of agent-oriented modality. (71) All students must obtain the consent of the Dean of the faculty concerned before entering for examination. (72) You can start the revels now.

200 Ferenc Kiefer

It would seem, however, that both (71) and (72) are ‘ambiguous’: they either state certain conditions on the agent or they are used to perform certain speech acts (granting permission and imposing an obligation, respectively). The same holds true of utterances such as (73) and (74). (73) You may park your car here. (74) You must pay your rent tomorrow.

Both (73) and (74) may be interpreted as either expressions of agent-oriented or of speaker-oriented modality. (73) can be used to state a possibility and (74) to state a necessity. For example, by uttering (73), the speaker may want to inform somebody about parking possibilities. In this case, the utterance of (73) does not mean permission since the speaker has no authority to perform such an act. On the other hand, (73), if uttered by a policeman, may mean permission. The terms ‘agent-oriented modality’ and ‘speaker-oriented modality’ refer to two types of speech acts and are not two different modal notions. Both speech act types are based on deontic modality. It should be kept in mind that speech acts do not come under the heading of modality. More generally, illocutionary verbs refer to an act and are not used to relativize the validity of a state-of-affairs. Cf. ‘I assert that all men are mortal’. What does it mean to say that something is compatible with what I assert? There is no meaningful answer to this question: illocution is alien to the notion of modality. The same holds true for perlocution. The effects that an act may bring about in the addressee cannot affect the act itself, i.e. the proposition which describes the act is independent of the eventual consequences of this very act. In sum, then, the act of asserting, asking, permitting, ordering, or promising is alien to modality. As we saw, deontic possibility and deontic necessity are semantic notions whereas imposing obligation and granting permission belong to pragmatics. From the point of view of semantics, (72) and (73) express deontic possibility, which can be paraphrased as ‘the deontic background allows the inference that p is not precluded’. If additional conditions are met (i.e. the speaker has the authority to grant permission, the act in question has not yet been performed, the addressee is in the position to perform the act, etc.), (72) and (73) may be used to grant permission. Similarly, semantically (71) and (74) express deontic necessity, that is, given a deontic background, practical inferencing tells us that (71) or (74) is the only possibility available. And, once again, if additional conditions are met, (71) and (74) can be used to issue an order. 7.4  Ability and possibility Ability is not a modal notion but it is related to modality via dispositional or circumstantial modality (Kiefer 1988: 409–412). The utterance



Modality

(75) Mary can sing.

may express, among other things, dispositional possibility. (75) pragmatically presupposes (76). (76) Mary is able to sing.

On the other hand, (76) conversationally implicates (75): if Mary is able to sing, then there must be occasions when her dispositions are such that she can sing. Similar things hold true of circumstantial modality as well. Note that (75) may also be interpreted as expressing circumstantial possibility. And, once again, the circumstantial possibility expressed by (75) pragmatically presupposes Mary’s ability to sing and Mary’s ability to sing conversationally implicates circumstantial possibility. In some languages the relatedness of ability and dispositional/circumstantial possibility manifests itself in the use of the same lexical resources. In Hungarian, for example, the auxiliary tud ‘know how to, be able to’ (as a full verb tud means ‘know’) is usually used to express ability, but it can also be used to indicate dispositional or circumstantial possibility. It never occurs in expressions of deontic, boulomaic or epistemic modality. The same seems to be true of the Belgian French verb savoir ‘know’. The close link between ability and circumstantial/dispositional possibility and the lack of such a link between ability and deontic/boulomaic possibility seems to indicate that two types of root modality have to be distinguished: deontic/boulomaic, on the one hand, and circumstantial/dispositional modality, on the other. This is further corroborated by the fact that the speech act potential of circumstantial or dispositional modality is not different from that of ability whereas deontic as well as boulomaic modality have their own speech act potential such as order and permission, wishes and desires (Kiefer 1997). 7.5  Modality and grammaticalization It has been argued that the development of the different meanings of can in English and of similar modal expressions of possibility in other languages shows the following general scheme (Bybee, Perkins & Pagliuca 1994: 176–242). (77) ability → root possibility → epistemic possibility

The meaning changes involved at each stage can be described as the loss of one feature of meaning: (78) a.

mental ability: mental enabling conditions exist in an agent for the completion of the predicate situation;

b. physical ability: enabling conditions exist in an agent for the completion of the predicate situation;

201

202 Ferenc Kiefer

c.

root possibility: enabling conditions exist for the completion of the predicate situation;

d. epistemic possibility: the modal takes scope over the whole proposition.

According to some researchers an important mechanism for the above semantic generalization is metaphorical extension, in which the concrete meaning of an expression is applied to a more abstract context where some of its (original) cooccurrence restrictions no longer hold. The continued use of the expression in a wider context leads to a gradual loss of the more specific meaning (Bybee & Pagliuca 1985; Bybee, Perkins & Pagliuca 1994). One can, however, view these meaning changes as being a consequence of the gradual conventionalization of implicatures as well (Dahl 1985). Both explanations may be adequate, in fact, they seem to complement each other. As pointed out above, ability conversationally implicates root possibility: if x is able to do y, then it is possible — given certain circumstances — that x does y. The relationship between ability and possibility can in no way be construed in terms of semantic implication: ‘x is able to do y’ does not entail ‘it is possible that x does y’. The two propositions may be true under different conditions. Yet another view of the development of modality attempts to explain the semantic change in terms of “sociophysical concepts of forces and barriers” (Sweetser 1990: 52). The may of permission, for example, is understood in terms of a potential but absent barrier, obligative must in terms of a compelling force directing the subject towards an act. The epistemic meanings of these modals can be derived from the tendency to experience the physical, social, and epistemic worlds in a partially similar way. This similarity allows the mapping of sociophysical potentiality onto the world of reasoning. As testified by Hungarian and Belgian French the verb meaning ability can also be used to express circumstantial and dispositional possibility. That is, the implicature ability Æ root possibility may develop into the semantics of root possibility. Finally, root possibility can be construed as practical inference from a given background (cf. 3.3.), which may develop into logical compatibility, giving rise to epistemic possibility. A sentence such as (79) may — in a context in which someone is estimating an arrival time — conversationally implicate (80), which expresses epistemic possibility (Bybee, Perkins & Pagliuca 1994: 197–198). (79) It can take me up to four hours to get there. (80) It may take me up to four hours to get there.

We thus get the following scheme: (81) ability with implicated root possibility → root possibility based on practical inferences Æ epistemic possibility through the conventionalization of the practical inferences underlying root possibility



Modality 203

By taking into account the distinction between circumstantial/dispositional and deontic/ boulomaic possibility we arrive at the following scheme: (82) ability → circumstantial/dispositional possibility → deontic/boulomaic possibility → epistemic possibility

The development of epistemic probability (subjective epistemic meaning) cannot be explained in this way. The main reason for this is that the obligation sense and the probability sense may occur in mutually exclusive contexts and that the obligation sense does not implicate the probability sense (Traugott 1989). In the future, must has only an obligation reading: (83) The letter must arrive sometime next week.

In present and past sentences, however, must has only an inferential reading. (84) a. The letter must be in the mail. b. The letter must have been in the mail.

The sentence (85) can have two readings: an epistemic reading with a stative interpretation of the main verb and an obligation reading ‘he must be made to understand’. (85) He must understand what we want.

Notice that the obligation sense does not implicate the epistemic; in fact, they imply two distinct states-of-affairs, one in which the subject probably does understand and one in which he does not.

8.  Prospects Current research on modality centers around the following topics:

a. Evidentials Evidential particles or affixes as expressions of modality are a relatively recent discovery. As we saw above, evidentials are markers that indicate the source and the reliability of the speaker’s knowledge. Attested evidence is more or less equivalent with ‘I know for sure’, reported evidence with ‘it is possible’ and inferred evidence with ‘it is likely’. It can be argued that modality can be based either on possibility and necessity, or on evidentiality. This gives rise to two different, but interrelated, modal systems, and languages may vary according to which of these possibilities they exploit. English seems to be a language which has only possibility/necessity based modality, Tuyuca (Tucanoan, Colombia and Brazil) indicates modality by means of evidentials, German has a mixed system. However, even modal verbs in languages such as English, German, French,

204 Ferenc Kiefer

which traditionally were considered to express possibility and necessity, may in certain contexts function as evidentials. Thus, the dividing line between evidentials and expressions of possibility and necessity is not as clear-cut as was first thought. Much descriptive work has still to be done in order to obtain a clearer picture of the ways in which evidentials function in language and the traditional analyses of possibility/ necessity based modality have to be reconsidered in the light of evidentiality.

b. Subjective modality Subjective modality plays an overwhelming role in natural language. Modal adverbs express normally subjective modality and even modal auxiliaries are used subjectively in the majority of cases. Yet we don’t have any adequate description of subjective modality at our disposal. Some linguists even doubt the epistemological validity of the distinction between objective and subjective modality, at least in the case of modal verbs. It has become clear, however, that this distinction is grammaticalized in some languages. Moreover, modal adverbials are always used to express speakers’ attitudes. Granted that the distinction between subjective and objective modality is viable, two important questions have still to be answered: (i) What is the relationship, if any, between subjective modality and evidentiality?; (ii) How can an adequate description of subjective modality be achieved?

c. The development of modality The developmental scheme ability → root modality → epistemic modality seems to be well motivated. Much of the details remains, however, unclear. For example, can independent motivation be found for the split of root modality into circumstantial/ dispositional and deontic/boulomaic modality? Do cross-linguistic data support the claim that ability develops first into circumstantial/dispositional modality? Are practical inferences or conversational implicatures the major factor in the development of modality? Since only objective epistemic modality can figure in the developmental scheme and can thus be related to ability, how can the development of subjective epistemic modality (e.g. probability) be explained? All these questions must be relegated to subsequent research.

d. Tense and modality Though time is not a modal category on the view of modality accepted in the present article, tense is intricately related to modality. Since we cannot have knowledge, but only beliefs, about future world-states, “what purports to be a statement describing a future event or state of affairs is therefore, of necessity, a subjectively modalized utterance: a prediction rather than a statement” (Lyons 1977: 815). This explains



Modality 205

the development of expressions of the future in a number of languages into a purely modal category. In German, for example, the so-called future auxiliary werden has been shown to behave exactly like genuine modal auxiliaries such as müssen ‘must’ and können ‘can’ (Vater 1975). Past Tense forms do not necessarily have past time reference either. Thus, for example, tense does not refer to time in the following sentences. (86) a. I thought I’d come with you, if you don’t mind. b. If I said that, you would hit me.

Example (86a) reflects the tentativeness of the speaker, in (86b) the past tense form is embedded into a counterfactual context. It would seem that whenever a tense form lacks time reference it may express modality. It is, however, unclear under which conditions this close relationship between tense and modality holds. Moreover, there are some clear indications that certain tense forms may carry a modal value even if they express temporal reference. This is yet another question which needs to be clarified.

e. The pragmatics of modality The studies on the pragmatics of modality seem to concentrate on the interactional and discursive functions of modality. Some modal expressions may be considered formulaic (e.g. May I?, Could you (please)…?), they carry by convention a certain illocutionary force. Notions such as ‘willingness’, ‘intention’, ‘insistence’ (e.g. in the case of the English modal will) receive a distinctive role in the description of the use of modals. As we noted above, the action process can be split up into various phases: motivation, goal, planing, decision, execution, result. Each phase can be associated with an assessment or evaluation. Some modal verbs refer to various phases of the action process, others express the speaker’s evaluation. But only a few languages have modal verbs. How are then these various functions expressed in languages which lack such verbs? Much current work is devoted to this kind of comparative study.

f. The acquisition of modality Acquisitional studies seem to support the developmental scheme (76). Children’s modal expressions have been reported to develop also from deontic modality to epistemic modality. Studies of the acquisition of modality in English (Wells 1985) and other languages (Stephany 1986; Aksu-Koç 1988) have suggested that deontic modality is universally acquired earlier than epistemic modality. Specifically, the agent-oriented meaning of desire, e.g. ‘I want to/wanna’, is the earliest indirect request to develop for English speaking children, and ‘can/can’t’ also develops early to express ability and permission. In contrast, epistemic meanings related to degrees of certainty about

206 Ferenc Kiefer

propositions (e.g. possibility, probability, certainty) are developed much later in children’s language. The development of epistemic modality occurs in large part between 3 and 5 years (Stephany 1986). It remains to be seen whether the more elaborate developmental scheme in (82) can also be supported by acquisitional data. It is equally an open question at what age a difference is being made between circumstantial and dispositional modality.

References Aksu-Koç, A. (1988). The acquisition of Aspect and Modality: The Case of Past Reference in Turkish. Cambridge University Press. Anderson, L. (1986). Evidentials, paths of change, and mental maps: typologically regular asymmetries. In W. Chafe & J. Nichols (eds.), Evidentiality: The Coding of Epistemology in Language. Ablex. Bally, C. (1932). Linguistique générale et linguistique française. E. Leroux. ——— (1942). Syntaxe de la modalité explicite. Cahiers Ferdinand de Saussure II: 3–13. Brown, P. & S. Levinson (1978). Universals in language usage: politeness phenomena. In E.N. Goody (ed.), Questions and Politeness: 56–289. Cambridge University Press. Bybee, J. (1985). Morphology: A Study of the Relation Between Meaning and Form. John Benjamins. Bybee, J. & W. Pagliuca (1985). Cross-linguistic comparison and the development of grammatical meaning. In J. Fisiak (ed.), Historical Semantics, Historical Word Formation: 59–83. Mouton. ——— (1994). The Evolution of Grammar. Tense, Aspect, and Modality in the Languages of the World. The University of Chicago Press. Calbert, J. (1975). Toward the semantics of modality. In J.P. Calbert & H. Vater (eds.): 2–70. Calbert, J.P. & H. Vater (eds.) (1975). Aspekte der Modalität. Gunter Narr. Dahl, Ö. (1985). Tense and Aspect Systems. Blackwell. Givón, T. (1993). English Grammar. A Function-Based Introduction. John Benjamins. Halliday, M.A.K. (1970). Functional diversity of language, as seen from a consideration of modality and mood in English. Foundations of Language 6: 322–361. Hintikka, J. (1962). Knowledge and Belief. Cornell University Press. Horn, L.R. (1989). A Natural History of Negation. The University of Chicago Press. Kiefer, F. (1983). What is possible in Hungarian? Acta Linguistica Hungarica 33: 149–187. ——— (1986). Epistemic possibility and focus. In W. Abraham & S. de Meij (eds.), Topic, Focus, and Configurationality: 161–179. John Benjamins. ——— (1987). On defining modality. Folia Linguistica XXI(1): 67–94. ——— (1988). Ability and possibility. The Hungarian verb tud ‘to be able to’. Studies in Language 12(2): 393–423. ——— (1992). Modality. International Encyclopedia of Linguistics: 2515–2520. Oxford University Press. ——— (1995). Bound utterances. Language Sciences 18 (2): 575–587. ——— (1997). Modality and pragmatics. Folia Linguistica XXXI(3–4): 241–253. Kratzer, A. (1978). Semantik der Rede. Kontexttheorie — Modalwörter — Konditionalsätze. Scriptor. ——— (1981). The notional category of modality. In H.-J. Eikmeyer & H. Rieser (eds.), Words, Worlds, and Contexts. New Approaches to Word Semantics: 38–74. de Gruyter. Lyons, J. (1977). Semantics. Cambridge University Press.



Modality 207

Miller, G.A. & D.M. Kwilosz (1981). Interaction of modality and negation in English. In A.K. Joshi, B.L.Webber, I.A. Sag (eds.), Elements of Discourse Understanding: 201–216. Cambridge University Press. Oswalt, R. (1964). Kashaya Texts. University of California Press. Öhlschläger, G. (1989). Zur Syntax und Semantik der Modalverben im Deutschen. Linguistische Arbeiten 144. Niemeyer. Palmer, F.R. (1979). Modality and the English Modals. Longman. ——— (1986). Mood and Modality. Cambridge University Press. Perkins, M.R. (1983). Modal Expressions in English. Pinter. Quirk, R., S. Greenbaum, G. Leech & J. Svartvik (1985). A Comprehensive Grammar of the English Language. Longman. Redder, A. (1984). Modalverben im Unterrichtsdiskurs. Pragmatik der Modalverben am Beispiel eines institutionellen Diskurses. Niemeyer. Von Stechow, A. & W. Sternefeld (1988). Bausteine syntaktischen Wissens. Ein Lehrbuch der generativen Grammatik. Westdeutscher Verlag. Stephany, U. (1986). Modality. In P. Fletcher & M. Garman (eds.), Language Acquisition: 375–400. Cambridge University Press. Sweetser, E. (1990). From Etymology to Pragmatics: Metaphorical and Cultural Aspects of Semantic Structure. Cambridge University Press. Thurmair, M. (1989). Modalpartikeln und ihre Kombinationen. Niemeyer. Traugott, E.C. (1989). On the rise of epistemic meaning: An example of subjectification in semantic change. Language 65(1): 31–55. Van Der Auwera, J. (1996). Modality: the three-layered scalar square. Journal of Semantics 13: 181–195. Vater, H. (1975). Werden als Modalverb. In J.P. Calbert & H. Vater (eds.): 71–148. Wells, G. (1985). Language Development in the Preschool Years. Cambridge University Press. Willett, T. (1988). A cross-linguistic survey of the grammaticalization of evidentiality. Studies in Language 12(1): 51–97. Von Wright, G.H. (1971). Explanation and Understanding. Cornell University Press. Wunderlich, D. (1981). Modalverben im Diskurs und im System. In I. Rosengren (ed.), Sprache und Pragmatik: 11–53. Gleerup.

Negation Matti Miestamo University of Helsinki

In propositional logic negation is a connective that changes the truth value of a proposition. If p is true, then ¬p is false, and vice versa. As many sections in this article will show, things are not that straightforward in natural language. Philosophers have been interested in negation since ancient times. Within linguistics negation has been approached from various points of view, and a large number of topics related to negation is discussed in the literature. The central aspects of the most important issues will be covered here. Section 1 discusses the scope of negation, Section 2 takes up some issues pertaining to the markedness of negation, Section 3 looks at how negation is expressed in the world’s languages, the topic of Section 4 is negative polarity, negation and scalarity is treated in Section 5, metalinguistic negation in Section 6, negative transport in Section 7, some diachronic issues in Section 8, and finally Section 9 addresses the acquisition of negation. Before going into questions of scope, it is perhaps appropriate to mention two central publications in the field: Jespersen (1917) can be considered the classical work on negation in linguistics, and Horn (1989), covering a wide range of issues in the semantics and pragmatics of negation, is indispensable to anybody doing research on the subject.1

1.  Scope of negation A long philosophical tradition holds that negation is ambiguous between an internal and an external reading. I will use the well-known “King of France” examples to illustrate the difference. In internal negation (2a), the subject (or topic) of the sentence is interpreted as being outside the scope of negation, whereas in external negation (2b) it is included in the scope.

(1) The King of France is bald. (2) The King of France is not bald.

.  The 2001 reissue of Horn’s book contains a bibliography update and a summary of recents developments in the field.



Negation 209

a. ∃x(Kx & ∀y(Ky ⊃ y = x) & ¬Bx) [The King of France is not-bald] b. ¬∃x(Kx & ∀y(Ky ⊃ y = x) & Bx) [not(the King of France is bald)] (Horn 1989: 364)

Internal negation preserves the presuppositions of the corresponding affirmative. In the case of the French monarch, both (1) and (2a) presuppose the existence of the referent of the subject. External negation, on the contrary, denies the presuppositions. Thus (2b) does not presuppose the existence of the King of France. External negation is rare in natural language. Negation typically does not affect the presuppositions of a sentence. In fact, a useful test to see whether something is a presupposition is its constancy under negation. In classical propositional logic there are only two possible truth values: every proposition is either true or false. An utterance of (2) in 2006, when France is a republic, would be nonsensical on the internal reading, and a multivalued logic, a system that allows for a third truth value (i.e. neuter, neither true nor false, undetermined), would be needed to account for it. On the external reading the utterance of (2) would not pose a problem for a two-valued logic, either. The law of the excluded middle says that everything must be either true or false, formally (p ∨ ¬p). The law of contradiction says that two opposites, e.g. a proposition and its negation, cannot be true at the same time, formally ¬(p & ¬p). These laws help to clarify the distinction between contrary and contradictory negation. Contradictory opposition is governed by both laws, but contrary opposition only by the law of contradiction. Contrary opposites can thus be simultaneously false, but not simultaneously true. Contradictory opposites, on the other hand, cannot be simultaneously true nor simultaneously false. Pairs of antonyms such as odd/even or married/unmarried and pairs of affirmative vs. negative sentences involving ordinary sentential negation are contradictory opposites. Typical contrary opposites are pairs of antonyms such as warm/cold or friendly/unfriendly. Von Wright (1959) introduces the distinction between strong and weak negation. Strong negation is an affirmation and a denial at the same time. It affirms a negative predicate of a subject (x is not-a). Weak negation is merely a denial (x is not a). The laws of contradiction and excluded middle are both valid for weak negation, but for strong negation only the law of contradiction applies. Sentential negation takes the whole sentence in its scope, whereas constituent negation only applies to a part of the sentence. Jespersen (1917) uses the terms nexal and special negation: in nexal negation the negative operator belongs to the combination of two ideas, whereas in special negation it belongs to only one idea. Klima (1964) proposes the following criteria for identifying sentential negation in English: instances of sentence negation are those structures that permit the occurence of the either-clause, the negative appositive tag not even and the question

210 Matti Miestamo

tag without not.2 Furthermore, Klima suggests that strong sentence negation can be distinguished from weak sentence negation following the criterion of occurence with a neither-clause (of course strong and weak sentence negation are to be kept distinct from Von Wright’s weak and strong negation). According to these criteria, (3) contains strong sentence negation (4) contains weak sentence negation and (5) does not contain sentence negation at all.

(3) She doesn’t like running, {either./not even in the forest./does she?/and neither does he.}



(4) Scarcely anybody accepts suggestions, {either./not even writers./do people?/*and neither do writers.}



(5) She is unhappy, {*either./*not even with him./*is she?/*and neither is he.}

These criteria are specific for English, but similar language specific criteria can be and have been established for other languages as well. In Tottie’s (1977) study different speakers show different degrees of negativity for different negative elements; it is argued that there can be no strict division between strong and weak sentence negation, and these negative elements should be placed on a continuum from strong to weak negativity: not>never>hardly>little>few>seldom. Some have tried to identify sentential negation by paraphrasing it as It is not the case that p. It is, however, rare for the topic of a sentence to be in the scope of negation – as was already noted above, external negation is rare in natural language. Givón (1984: 326) notes that in actual (English) texts, there are usually no cases where a definite subject would fall under the scope of negation, and the rare cases where a subject is negated use special subject-negative forms. Payne (1985) argues that negation is semantically placed at the border of old and new information. Sentential negation can be given the following “performative paraphrase”: I say of X that it is not true that Y, where X contains the contextually bound elements, i.e. old information. Thus the sentence John is not running could be paraphrased as I say of John that it is not true that he is running. Finally, it should be noted that negation tends to be attracted to focused elements, and different ways of indicating focus (focus markers, special focus constructions, prosody etc.) can be used to restrict the scope of negation to a part of a sentence. 2.  Markedness of negation This section addresses a number of differences, or asymmetries, between affirmation and negation. Negation is marked vis-à-vis affirmation both functionally (in cognitive, 2.  Klima (1964) is a thorough investigation of English negation in an early generative perspective; for a more recent generative account, see Haegeman (1995).



Negation

semantic and pragmatic terms) and formally (in morphosyntactic terms). This section first discusses aspects of the functional asymmetry between affirmation and negation and the criteria of morphosyntactic markedness are taken up in the end of the section; evidence for the morphosyntactic markedness of negation will also be seen in Section 3, where the expression of negation is discussed from a cross-linguistic point of view. Negation is conceptually marked with respect to affirmation. It is a mental process added by language users, and there is thus more semantic content in a negative sentence than in the corresponding affirmative. Psycholinguistic experiments have shown effects of this complexity (see Clark 1974). Negatives are harder to understand than affirmatives, which applies to both inherently negative terms (remember/forget, present/absent), where the positive terms are processed faster than the negative ones, and to explicitly negative structures. True affirmative and false affirmative sentences are both processed faster than their negative counterparts, true affirmatives being the fastest. But it is interesting to note that false negatives tend to be easier to process than true negatives. Different explanations have been suggested for this result, e.g. Clark’s (1974) true model. Horn (1989, following Wason 1965) argues that true negatives are actually more complex than false ones, because they involve the denial of a falsehood (i.e. double negation) whereas false negatives involve the denial of a fact. Negatives differ from affirmatives in terms of the stativity/dynamicity of the states of affairs they report. Affirmatives can report both stative and dynamic states of affairs, but the states of affairs that negatives refer to are stative. Consider the following examples:

(6) (7) (8) (9)

Chris knows the song. Chris does not know the song. Chris drank the coffee. Chris did not drink the coffee. (Miestamo 2005: 196)

With stative predicates, both affirmation (6) and negation (7) are naturally stative. With dynamic predicates affirmatives are dynamic – there is change in the state of the universe in (8) – but negatives are stative – (9) refers to a state of affairs with no change in the universe (see e.g. Givón 1978: 103–108 and Miestamo 2005: 196–197 for more on the stativity of negation). Negative sentences are prototypically used a denials. They are typically uttered in contexts where the corresponding affirmative is somehow present. Tottie (1991) classifies negatives in discourse-functional terms as follows: i. rejections (including refusals) ii. denials: a. explicit b. implicit (Tottie 1991: 22) Rejections reject suggestions and denials deny assertions. As to the distinction between explicit and implicit denials, the former deny something that has been explicitly asserted

211

212

Matti Miestamo

and the latter “something which might merely have been expected, or which can be contextually inferred but which has not been asserted” (Tottie 1991: 21). In Tottie’s conversation data, rejections constitute only a small minority of all uses of negatives – all the rest are denials, implicit denials being clearly more common than explicit ones. As Givón (1978) points out, (10) uttered out of the blue would be odd, but it is a perfectly plausible utterance if the pregnancy of the speaker’s wife has already been discussed or alluded to. (10) Oh, my wife is not pregnant. (Givón 1978: 80)

Givón (1978, 1984, 2001) argues that negatives are discourse-presuppositionally more marked than their affirmative counterparts. When ¬p is uttered, p is present in the context as backgrounded information: “Negative assertions are typically made on the tacit assumption that the hearer either has heard about, believes in, is likely to take for granted, or is at least familiar with the corresponding affirmative” (Givón 2001: 370–371). In Clark’s (1974) terms, negatives involve the supposition of their affirmative counterparts. Events, e.g. the one reported by (8) above, occur at specific points in time, but there is an infinite number of moments when a given event does not take place and when its negation, e.g. (9), is thus true. From a perceptual point of view, as Givón (1978: 103–108) argues, an event (change in the state of the universe, reported by an affirmative sentence) can be associated with figure, and the absence of an event (no change in the state of the universe, reported by a negative sentence) can be associated with ground. In communication, the information value of figure is high, whereas the ground has low information value – there is less need to communicate anything about the inert background. Negatives need a special context to be plausibly used in discourse, and this context is provided by the supposition of the corresponding affirmative, which gives higher information value for negatives. The context thus involves a reversal of figure and ground. Leech (1983) introduces the maxim of negative uniformativeness. Negative sentences are typically less informative than their affirmative counterparts. It is more informative to say (11) than (12). (11) Abraham Lincoln was shot by John Wilkes Booth. (12) Abraham Lincoln was not shot by Ivan Mazeppa. (Leech 1983: 100)

The maxim of negative uniformativeness, together with the maxim of quantity, says that a negative sentence should be avoided if a positive one can be used instead. It also implies that when negative sentences are used, there is a special purpose for doing so: They are used in contexts where they are not less informative than positives, i.e. when the supposition of a positive is present in the context. Leech argues that the maxim of negative uninformativeness thus provides an explanation for why negative



Negation

propositions are usually denials of positive propositions present in the context. Leech has been criticized for proliferating maxims beyond necessity. According to (Brown & Levinson 1987: 51) the maxim of negative uninformativeness, too, could be derived from other maxims. For Horn (1989: 198–201) the pragmatic markedness of negation results from the interplay of the maxims of quantity and relation. A well known phenomenon connected with the markedness of negation is the tendency for indefinite NPs occurring in the scope of negation to be non-referential. The following English examples illustrate the situation. In the definite cases (13) and (15), the NPs are referential. The indefinite NP can get either a referential or a non-referential reading in the affirmative (14), but in the scope of negation the indefinite NP gets a non-referential reading (16).

(13) (14) (15) (16)

John read the book. John read a book. John didn’t read the book. John didn’t read a book. (Givón 1978: 71)

As (17) shows, this is not an absolute restriction. Examples like this are extremely rare in actual discourse and alternative structures are preferred for expressing the same information. (17) Well, she didn’t read a book that was put on the required list, and as a result she failed her exam. (Givón 1978: 72)

Givón (1978, 1984: 333) explains the tendency for non-referentiality of indefinite NPs under negation by the discourse context of negation: as negatives are used in contexts where the corresponding affirmative is supposed, referential participants are definite in negatives – negatives do not introduce new participants into the discourse. Unlike in English, in some languages this pragmatic effect has been conventionalized in grammar. Finnish objects, for example, take partitive case under negation, whereas in the affirmative there is a choice between accusative and partitive objects. In French, objects marked by indefinite articles in affirmative contexts receive partitive marking by de in the scope of negation. The markedness of negation also shows in the lexicon. Antonyms are positive/ negative pairs where negativity can be overtly signalled (happy/unhappy) or it can be implicit (happy/sad). Again, the negative term is the marked one. The positive term is conceived of as neutral. Long can be used to designate the whole scale of length. It is natural to ask How long is the rope? but it would be a marked use to ask How short is the rope? It has also been noted that when pairs of antonyms are coordinated, the positive term tends to come first in the so-called fixed binominals or freezes (e.g. positive or negative, all or none, many or few, plus or minus, win or lose, pro and con, tall and short, good and bad). A cognitive motivation for these markedness patterns is that the positive

213

214 Matti Miestamo

term is perceptually salient since it represents the presence of the property on the scale, whereas the negative term represents its absence. The functional markedness of negation vis-vis affirmation is reflected in formal (morphosyntactic) markedness as well. In the world’s languages, according to Greenberg (1966: 50), “[t]he negative always receives overt expression while the positive usually has zero expression”; there are indeed extremely few languages where no overt marker of negation is found, and in none of these do we find overt expression of affirmation (see Miestamo 2005: 121). Negation also satisfies the frequency criterion for markedness by occurring less frequently in discourse than affirmation. Similarly, negation appears as the marked member of the affirmative vs. negative opposition according to the behavioral potential criteria (Croft 2003: 95–99; see also Greenberg 1966): less grammatical distinctions are made in the negative than in the affirmative and negation can be embedded in fewer contexts than affirmation. The functional asymmetry between affirmation and negation discussed in this section is thus reflected in morphosyntactic asymmetry as well. The next section will, among other things, address the cross-linguistically common phenomenon of neutralization of grammatical categories in negatives, which is a clear indication of the morphosyntactic markedness of negation.

3.  The expression of negation in the world’s languages This section looks at negation from a typological point of view. Its main concern is clausal negation, with a special focus on standard negation – the basic strategies that languages use for negating declarative verbal main clauses. Some other aspects of negation, e.g. the negation of indefinite pronouns, will receive some space as well. According to Payne (1985) standard negation can be expressed by negative particles, negative verbs or in the morphology of the verb (and marginally also by negative nouns, although the examples he gives are not instances of standard negation). He also briefly mentions some “secondary modifications”, i.e. ways in which negatives may differ from affirmatives structurally. Dahl (1979) makes a basic distinction between syntactic and morphological negation. Morphological negation is most typically affixal. Syntactic negation can be realized by negative particles or negative verbs, in addition to which there can be some changes in the lexical verb and a special finite element (FE) can be added in the negative construction.3

3.  Dahl introduces the term finite element (FE) as a more abstract term for what is basically the finite verb of a clause. The FE added in negatives is usually an auxiliary verb.



Negation

More recent cross-linguistic studies of clausal negation have paid attention to the structure of negative clauses more globally, and observed the functional effects of the changes that negation can bring to the structure of a clause. Honda (1996) identifies three basic types of expression of negation according to the status of the FE of the negative clause: (1) the FE of the negative is not different from the FE of the corresponding affirmative, (2) there is a special (non-negative) FE added in the negative, and (3) the negative marker itself is the FE of the negative. Forest (1993) distinguishes two main types of negatives: “recusative” and “suspensive-reassertive” (négation récusative, négation suspensive-réassertive). To put it simply, in the former type the negative differs from the corresponding affirmative only by the presence of negative marking, whereas in the latter the morphosyntactic marking of some semantic domains differs from their marking in affirmatives. Both authors discuss ways in which negatives differ from affirmatives in the world’s languages, e.g. neutralization of tense-aspect-mood (TAM) categories and obligatory non-indicative marking in negatives. Miestamo (2005) distinguishes two basic types of negative structures – symmetric and asymmetric negation – according to whether or not negatives differ structurally from affirmatives in addition to the presence of negative markers. This distinction is observed from the point of view of constructions on the one hand and paradigms on the other. Symmetric negative constructions do not differ from the corresponding affirmative in any other way than by the presence of (a) negative marker(s) (e.g. in Washo 18), whereas in asymmetric constructions further differences – asymmetries – are observed (e.g. in Finnish 19). In symmetric paradigms the correspondences between the members of the paradigms used in affirmatives and negatives are one-to-one (e.g. in Italian 20), whereas in asymmetric paradigms such one-to-one correspondence does not obtain (e.g. in Maung 21 and in Burmese 22). Grammatical distinctions are often neutralized in paradigmatic asymmetry. (18) Washo (Jacobsen 1964: 603, 604–605)4 a.

le-ímeʔ-hu-i b. 1-drink- pl.incl-impf ‘We are drinking.’

le-ímeʔ-é·s-hu-i 1-drink- neg-pl.incl-impf ‘We are not drinking.’

4.  Abbreviations used in the glosses: 1 – first person, 2 – second person, 3 – third person, act – actual, cng – connegative, impf – imperfective, incl – inclusive, irr – irrealis, neg – negation, npst – nonpast, perf – perfect, pl – plural, pot – potential, sg – singular.

215

216 Matti Miestamo

(19) Finnish a.

laula-n b. sing- 1sg ‘I sing.’

e-n laula neg-1sg sing. cng ‘I do not sing.’

(20) Italian a. cantare ‘to sing’, present b. affirmative negative 1sg canto non canto 2sg canti non canti 3sg canta non canta 1pl cantiamo non cantiamo 2pl cantate non cantate 3pl cantano non cantano

cantare ‘to sing’, future affirmative negative canterò non canterò canterai non canterai canterà non canterà canteremo non canteremo canterete non canterete canteranno non canteranno

(21) Maung (Capell & Hinch 1970: 67) a.

ŋi-udba b. 1sg.3-put ‘I put.’

ni-udba-ji c. 1sg.3-put- irr.npst ‘I can put.’

marig ni-udba-ji neg 1sg.3-put- irr.npst ‘I do/shall not put.’

(22) Burmese (Cornyn 1944: 12–13) a.

θwâ-dé b. go- act ‘goes, went’

θwâ-mé c. go- pot ‘will go’

θwâ-bí d. ma-θwâ-bû go- perf neg-go- neg ‘has gone’ ‘does/did/will not go, has not gone’

Cross-cutting the constructional–paradigmatic distinction, asymmetric negation is divided into subtypes according to the nature of the asymmetry. In subtype A/Fin, the finiteness of the lexical verb is reduced or lost in the negative, and a new FE is usually added; in the Finnish SN construction (19) the negative verb e- appears as the FE of the clause taking person inflection and the lexical verb appears in the non-finite connegative form. In subtype A/NonReal, negatives are obligatorily marked for a category that refers to non-realized states of affairs; in Maung negation is marked with marig and the construction is symmetric (21b,c), but there is paradigmatic asymmetry since negatives obligatorily use the irrealis form of the verb and the distinction between realis and irrealis (21a,b) is lost in the negative (21c). There is a marginal subtype, A/Emph, characterized by the presence of marking that denotes emphasis in non-negatives (for lack of space not exemplified here). Finally, in subtype A/Cat negatives, the marking of



Negation

grammatical categories differs from their marking in affirmatives in other ways, the most commonly affected categories being TAM and person-number-gender; in Burmese (22) the affirmative makes a distinction between actual, potential and perfect, the negative construction is asymmetric since the suffixal part of the discontinuous negative marker replaces the TAM markers, and there is also paradigmatic asymmetry since these TAM distinctions are then lost. The following principles of explanation are proposed for the typology in Miestamo (2005, cf. Itkonen 2005): Symmetric negation copies the linguistic structure of the affirmative and is thus language-internally analogous to the corresponding affirmative, ultimately motivated by pressure for system cohesion. Asymmetric negatives copy aspects of the functional asymmetry between affirmation and negation (see Section 2 above) and are thus language-externally analogous to these functional-level asymmetries. Different subtypes of asymmetric negation have conventionalized different aspects of the functional asymmetry in their grammars: A/Fin is motivated by the stativity of negation, A/NonReal by the semantic connection between negation and other conceptualizations of the non-realized, and type A/Emph and those type A/Cat structures where grammatical distinctions are lost are both motivated by the prototypical discourse context of negatives in different ways; see Miestamo (2005: 195–235) for more discussion. It may also be noted that negatives are typically less transitive than affirmatives according to the criteria proposed by Hopper & Thompson (1980). Negation itself figures among the criteria of low transitivity and many of the other criteria are also satisfied by negatives: non-action and irrealis, as well as the non-affectedness and nonindividuation of the object. This is very much in accordance with the classical view of transitivity as transfer of energy from agent to patient – in negatives, where the action does not happen (or happens less completely), there is no (or less) transfer. Negation itself can be marked differently in connection with different categories. Even within standard negation, different negative constructions can be found for example in connection with different tense-aspect categories. In clausal negation more generally, the most typical environments for negative constructions to differ from standard negation are imperatives and existentials (see Kahrel 1996: 70–71). Van der Auwera & Lejeune (2005) found out that roughly two thirds of their sample languages used special negative marking in prohibitives (negative imperatives), i.e. negative marking not used in declarative negatives.5 Van der Auwera (2006) examines diachronic developments behind these cases and proposes an explanation of the

5.  In fact, (western) Europe is the only area in the world where dedicated prohibitives are a minority. Note that Haspelmath et al. (eds. 2005) contains maps concerning other aspects of negation as well.

217

218 Matti Miestamo

preference for dedicated prohibitives based on the radically different speech act status of prohibition vs. (the more frequent) declarative negation. Turning now to word order, the position of negative markers is rarely determined in relation to the whole clause, i.e languages tend to place their clausal negators close to the finite verb rather than in clause-initial or clause-final position; this is a clear reflection of the fact that external negation is rare in natural language (see Section 1 above). Jespersen (1917: 5) introduced the Neg-first principle, according to which negative markers tend to be placed before the elements they negate. In standard negation, the element negated is the finite verbal element of the sentence, and according to Dryer (1992) there is indeed a tendency for negative particles to be placed preverbally, regardless of basic word order. Basic word order does play a role for negative auxiliaries, which tend to be placed before the lexical verb in VO-languages and after it in OV-languages. Kahrel (1996) distinguishes the following types of constructions used for the negation of indefinite pronouns: (1) standard negation with ordinary (positive) indefinite, (2) standard negation with special indefinite, (3) inherently negative indefinite pronoun without standard negation, and (4) inherently negative indefinite pronoun with standard negation; there is a fifth type, usually found in languages with no indefinite pronouns, where the equivalent function is expressed by an existential construction. The first four types are also identified in Dahl (1979) and Bernini & Ramat (1992). Kahrel’s typology works in isolation, but it is often more interesting to see how negative indefinite pronouns are related to other indefinite pronouns. Haspelmath (1997) proposes an implicational map for the functions/uses of indefinite pronouns. The functions range from specific known to direct negation and free-choice items. An indefinite pronoun in a given language serves only adjacent functions on the map. Negativeness being a property of the whole sentence, it is not easy to say categorically that an indefinite is negative or non-negative, but its different functions can be placed on the implicational map. The only important parameter of negative indefinite constructions that is not accounted for by the implicational map is the co-occurrence of negative indefinites with standard negation. There are three main types: NV-NI, where the negative indefinite always co-occurs with standard negation; V-NI, where standard negation and the negative indefinite never co-occur; and (N)V-NI, where negative indefinites sometimes do and sometimes do not co-occur with standard negation.6 There are of course some interdependencies between the implicational map and co-occurrence with standard negation. It seems clear, for example, that an indefinite

6.  In this third type, standard negation is present usually when the negative indefinite is postverbal. This is understandable via the Neg-first principle: it is communicatively effective to signal negativeness as early in the sentence as possible.



Negation

whose functions are at the positive end of the implicational map could not get a negative reading without standard negation.7 4.  Negative polarity items A lot of research has been done on polarity items in recent years. Van der Wouden (1997: 61) proposes the following descriptive definitions of positive and negative polarity items: “Positive polarity items (PPIs) are expressions that cannot felicitously appear in negative contexts.” “Negative polarity items (NPIs) are expressions that can only appear felicitously in negative contexts.” NPIs include elements such as English any, yet or a red cent, and PPIs are elements like some and already. The behaviour of these elements is illustrated in (23)–(27).

(23) (24) (25) (26) (27)

The president signed {some/*any} of the papers yesterday. The president didn’t sign {*some/any} of the papers yesterday. I’ve had coffee {already/*yet}. I haven’t had coffee {*already/yet}. Chris {*gave/didn’t give} a red cent to the Tsunami Relief Fund.

Polarity items are found in various syntactic and semantic classes. The fact that NPIs are a much larger, more productive and more frequently occurring class than PPIs can be considered an exception to the markedness of negation (cf. Section 2 above). Van der Wouden (1997) examines different attempts to classify NPIs in the literature. He concludes that it is not fruitful to classify them on etymological or syntactic grounds. No simple semantic classification can be established either, but certain tendencies can be found. Most languages have NPIs that denote a minimal amount. Similarly, negatively polar verbal idioms are found in most languages. The rhetorical figures of understatement and litotes are found in many languages and these are a frequent source for NPIs. Many languages also have indefinites functioning as NPIs. Furthermore, many intensifiers are either NPIs or PPIs. In studies dealing with negative polarity the crucial question is what licenses NPIs; what are the negative contexts that NPIs occur in? Many different answers – syntactic, semantic and pragmatic – have been proposed to the licensing question in the literature. The syntactic approaches are based on the presence of a negative operator in the sentence, and the notion of c-command is central to most of them.

7.  In this context it is also worth noting that the so-called double negation, the pattern exemplified by e.g. Italian and Spanish, is by far more common in the languages of the world than the “logical” Standard English type, obeying the law of double negation.

219

220 Matti Miestamo

According to the semantic approach proposed in Ladusaw (1980), NPI licensers are downward-entailing (monotone-decreasing, downward-monotonic) expressions. Downward-entailing expressions allow one to infer from sets to subsets.

(28) (29) (30) (31)

Chris does not drink coffee. Chris does not drink espresso. Chris drinks little coffee. Chris drinks little espresso.

We can infer (29) from (28) and (31) from (30); not and little are downward-entailing and they are both NPI licensers. The approach to polarity licensing based on downward monotonicity is further elaborated in Van der Wouden (1997), where polarity is not seen as an isolated phenomenon in language, but polarity licensing (as well as some other phenomena connected with negation) is treated as a subcase of collocational behaviour. Furthermore, several contributions in Horn & Kato (2000) and Hoeksema et al. (2001) report results from the intensive work that has been done on negative polarity in recent years.

5.  Negation and scalarity Ordinary sentential negation typically results in contradictory oppositions: Mary is singing/Mary is not singing, whereas constituent negation of the type happy/unhappy gives contrary opposites (see Section 1 above). When applied to scalar predicates, sentential negation shows some interesting effects. Fauconnier (1975) observes that negation has the effect of reversing the ordering of elements on a scale. (32) John had four cups of coffee. (33) John didn’t have four cups of coffee.

The sentence in (32) implies that John had three cups, that he had two cups etc. In the negative (33), the implications go to the opposite direction: John did not have five or six cups etc. In fact, already Jespersen (1924: 325–326) noted that negation when applied to certain (scalar) predicates means ‘less than’. Not good implies inferior but not excellent. The same applies to numerals: Example (34) means the hill is lower than 200 feet. (34) The hill is not two hundred feet high. (Jespersen 1924: 326)

However, Jespersen notes that the same expression can exceptionally mean ‘more than’, as is shown by the Examples (35) and (36). This requires a special prosody in the utterance.



Negation

(35) [It is] not lukewarm, but really hot. (36) His income is not two hundred a year, but at least three hundred. (Jespersen 1924: 326)

Horn (1989: 266–267) argues that these effects are not brought about by some special property of negation, but are rather due to the nature of the scalar predicates. The scalar operators are lower-bounded by their literal meaning and upper-bounded by a conversational implicature based on the maxim of quantity. This context-dependent implicature gives two readings for sentences containing these scalar operators: ‘at least P’ and ‘exactly P’. Negation, on the normal reading, contradicts the literal meaning, yielding ‘not (at least) P’, i.e. ‘less than P’. Thus in (34) negation applies to the literal (pre-upper-bounded) value of the scalar predicate yielding the reading ‘not (at least) 200 feet’, i.e. ‘less than 200 feet’. In the marked cases (35) and (36), on the other hand, negation applies to the ‘exactly’-reading produced by the conversational implicature. Thus (35) does not mean ‘It is not (at least) lukewarm’ but rather ‘It is not (exactly) lukewarm’.8 These examples can be analyzed as instances of metalinguistic negation, the topic of the next section. 6.  Metalinguistic negation The term metalinguistic negation originates from Ducrot (1972). It is used for negations where what is negated is not the content of the proposition but rather the way it is expressed. Metalinguistic negation has received a lot of attention since Horn’s seminal article (Horn 1985). This section will first present Horn’s view of the phenomenon and then mention some alternative approaches and comments inspired by Horn’s writings. For Horn metalinguistic negation is a device for objecting to a previous utterance on any grounds whatever, including the conventional or conversational implicata it potentially induces, its morphology, its style or register, or its phonetic realization. (Horn 1989: 363)

In opposition to the unmarked descriptive negation, which is internal and truth-functional, the marked metalinguistic negation is external and non-truth-functional. As discussed in Section 1 above, on the internal reading it would be hard to assign a truth value to The King of France is not bald. The continuation there is no King of France in (38) forces an external presupposition-cancelling reading – this is a metalinguistic negation denying the assertability of (37) on the grounds that there is no King of France.

8.  For alternative views of scalarity, see for example Carston’s (1998a) Relevance Theoretic and van Kuppevelt’s (1996) topic-based approaches.

221

222 Matti Miestamo

(37) The King of France is bald. (38) The King of France is not bald – there is no King of France.

The grounds for objecting to an utterance can be of many kinds. The objection can be due to presupposition-failure, as is the case with the King of France. It can be due to denial of implicatures, as in (39) and (40), where what is objected to is an upperbounded conversational implicature – that he does not have more than three children in (39) and that coffee is merely liked, but not loved, in (40). (39) He doesn’t have three children, he has four. (40) Around here we don’t like coffee – we love it. (Horn 1989: 382)

The grounds for rejecting an utterance can be its phonetic realization, as in (41), or the stylistic effect of the utterance, as in (42). In these cases it is very clear that the objection has nothing to do with the content of the proposition. (41) He didn’t call the [pólis], he called the [polís] (42) Phydeaux didn’t shit the rug, he soiled the carpet. (Horn 1989: 371)

Metalinguistic negation also involves a special intonational pattern, as was already noticed by Jespersen in connection with the scalar cases (see previous section). In (43) the word manage has a special fall-rise intonation contour. (43) John didn’t manage to solve the problem – it was quite easy for him to solve. (Horn 1985: 130)

One characteristic property of metalinguistic negation is that PPIs, rather than NPIs, are found in its scope. The affirmative sentence (44) contains the PPI some and (45) includes a descriptive negation with the NPI any. In (46) the negation is metalinguistic, and the PPI appears instead of the NPI. (44) John managed to solve some problems (45) John didn’t manage to solve any problems (46) John didn’t manage to solve {some/*any} problems – they were quite easy for him to do. (Horn 1985: 130)

This is understandable given that in the metalinguistic case the propositional content is not negated – in (46) it is not denied that problems were solved by John. Horn (1985, 1989) argues for the pragmatic ambiguity of the negation operator: it can be a descriptive truth-functional operator, taking a proposition p into a proposition not-p, or a metalinguistic operator which can be glossed ‘I object to u’, where u is crucially a linguistic utterance rather than an abstract proposition. (1985: 136)

This has been both agreed with and criticized. Van der Sandt (1993) does not treat metalinguistic uses of negation differently from other denials and does not consider the marked use of negation a pragmatic phenomenon, but gives a semantic analysis instead. Carston (1996) emphasizes the echoic use of material in the scope of negation



Negation

and argues against the ambiguity of the negative operator; the two uses are attributed to the capacity of using language to represent states of affairs or other representations (including other utterances). For Geurts (1998) Horn’s metalinguistic negation is not a unified phenomenon but involves different mechanisms of denial. Burton-Roberts (1989) agrees with Horn on the nature of metalinguistic negation, but emphasizes the fact that all cases of metalinguistic negation involve a contradiction, and it is the contradiction that triggers the need for a special pragmatic interpretation: to say that one does not like coffee and then to go on by saying that one loves it is a contradiction; to say that one does not have three children and assert that one has four is a contradiction. When one operates in a presuppositional semantics, a contradiction can be seen in the presupposition-denial cases as well.9 7.  Negative transport Negative transport (Neg(ative)-raising or not-hopping) refers to the phenomenon by which a negative in a higher clause is interpreted as the negation of an embeddedclause predicate. In (47) the negative is in the higher clause and in (48) in the embedded one, but on the negative transport reading, (47) does not negate the thinking but the coming, and the sentences are (nearly) synonymous. (47) I don’t think he has come. (48) I think he has not come. (Jespersen 1917: 53)

It has been noted that different verbal predicates behave in different ways as to whether they permit negative transport or not. As seen in (47) and (48), think allows negative transport but predicates such as claim and hope do not seem to do so: (49) has a clearly different meaning from (50), and (51) from (52). Of course, it also varies from dialect to dialect and language to language which predicates allow negative transport and which do not.

(49) (50) (51) (52)

I don’t claim he has come. I claim he has not come. I don’t hope he has come. I hope he has not come.

Horn (1978, 1989) suggests the following explanation. The different predicates can be arranged on a scale according to the strength of subjective certainty (for epistemic predicates) or strength of obligation (for deontic predicates). The predicates that allow negative transport have mid-scalar values on the scale (believe, think, be likely, seem, should, want), whereas predicates that do not allow negative transport have either

9.  Discussion on the nature of metalinguistic negation continues in Carston (1998b, 1999) and Burton-Roberts (1999).

223

224 Matti Miestamo

weak (be able, be possible, may, can, permit, be allowed) or strong (know, be certain, must, cause, order) values on the scale. When negated, predicates at different points on the scale behave differently: a. The (contradictory) negation of a weak scalar value (e.g. possible, allow) will be a strong value on the corresponding negative scale (impossible, forbid). b. The negation of a strong scalar value (e.g. certain, have to) will be a weak value on the corresponding negative scale (not certain, do(es)n’t have to). c. The (contradictory) negation of a mid-scalar value (e.g. likely, advisable) will be an intermediate value on the corresponding negative scale (e.g. not likely, not advisable). (Horn 1978: 195) The mid-scale predicates, thus, do not radically change their position on the scale when negated. This allows for the (partial) synonymy between sentences where the negation is in the upper clause and sentences where it is in the embedded clause. The issue of negative transport has been recently addressed by Tovena (2001). 8.  Negation in diachrony Negative markers are often ancient elements whose origins cannot be reached. The Indo-European *ne was a negative element already as far back in history as a protolanguage can be reconstructed. In some languages negative markers have grammaticalized from verbs with negative meanings, e.g. ‘fail’, ‘lack’, ‘refuse’, ‘decline’ or ‘avoid’, see e.g. Givón (2001: 267–268). Other possible sources are elements that reinforce negation and negative existentials, both of which will now be dealt with in more detail. In the development known as Jespersen’s cycle, elements that were introduced into negative clauses to reinforce negation are reanalysed as negative markers. I will illustrate this development with the best known example, viz. French negation. In Latin the standard negation marker was non (53). (53) Latin (Jespersen 1917: 7) non dic-o neg say-1 sg ‘I do not say.’

In Old French this negator had become phonetically weaker, it had reduced to ne (54). (54) Old French (Jespersen 1917: 7) Jeo ne di 1 sg neg say ‘I do not say.’



Negation

To fulfil its communicative function, it was frequently reinforced by words such as pas ‘step’, goutte ‘drop’, point ‘point’, or mie ‘crumb’. These words appeared in the direct object position after the verb and were originally found in contexts where they were motivated by their semantic content. Thus for example pas appeared in sentences that had to do with walking and goutte in contexts that involved liquids. These elements started to grammaticalize and lose their semantic content, and were then generalized to other contexts as well. Finally, pas ousted out the other elements (although point still survives as a marginal variant). It became an obligatory part of the negative construction and negation was expressed with a double particle appearing on both sides of the verb ne…pas. This is the situation in Modern Standard French (55). (55) Modern Standard French Je ne dis 1 sg neg say ‘I do not say.’

pas neg

Modern Spoken French has gone further; preverbal ne has disappeared, and standard negation is expressed with the postverbal particle pas (56). (56) Modern Spoken French Je dis pas 1 sg say neg ‘I do not say.’

As has been noted above, negative particles tend to be preverbal, and the postverbal negative particle in French thus seems typologically odd. This exceptional word order can be easily understood in the diachronic perspective. Some other European languages, e.g. English, German and Swedish, have gone through a similar development and their postverbal negators can be explained in the same way (for the history of English negation, see also Mazzon 2004). If the Neg-first principle is effective, we might expect the development to go through the whole cycle and postverbal particles to change their position becoming preverbal. This has indeed happened in most French Creoles.10 According to Ramat & Bernini (1990: 30), Germanic (especially Scandinavian) postverbal negators are moving towards preverbal position. Croft (1991) proposes a hypothesis of a cyclical development of negative existential predicates. He finds three distinct types in the languages of the world: A in which the ordinary existential predicate is negated by the verbal negator, B in which there is

10.  This view of course presupposes that we take French creoles to be historical continuations of French in the relevant sense.

225

226 Matti Miestamo

a negative existential predicate different from the verbal negator, and C in which the negative existential is identical to the verbal negator. Some languages show variation between two different types: A~B, B~C and C~A. Croft argues that these cases of variation can be interpreted as ongoing change from one type to another. The principles of grammaticalization theory suggest that the directionality of change is A>B>C>A. This is what Croft calls the negative existential cycle. Thus, one further origin of standard negators is the reanalysis of negative existentials as verbal negators (B>C).

9.  The acquisition of negation As any well-behaved marked category, negation is more difficult for a child to learn than its unmarked counterpart. Negative sentences appear in children’s speech later than affirmative ones, and as far as pairs of antonyms are concerned, the negative term is harder to acquire than its positive counterpart (see Clark & Clark 1977: 513). According to Clark & Clark (1977: 348), the first expressions of negation are gestures that may be combined with single words, or a negative word may be used alone. Some children may mark single-word or two-word utterances as negative by using a different intonation pattern. Klima & Bellugi (1966) identify three stages in the acquisition of English negation. At the first stage negation is sentence-external; a negative (not or no) is combined with a proposition by placing it at the beginning (or less commonly at the end) of a sentence. At the second stage negatives start to be incorporated into the sentence and negatives such as can’t and don’t are used in addition to not. At this stage the adult synthetic forms, e.g. don’t, are unanalysed wholes for the child. At stage three children master the essentials of the adult system of English negation; they no longer use sentence-initial (or final) negative markers and don’t is now analysed as the combination of do and n’t. Bloom (1970, 1991: 144–145) questions the sentence-initial first stage by arguing that Klima and Bellugi’s data were not correctly analysed. When negation occurs at the beginning of a sentence, it is most often a negative sentence with the sentence subject omitted, otherwise it is anaphoric or emphatic. De Villiers & de Villiers (1985: 82) conclude that sentence-initial negatives do not constitute a universal first step, although individual children may adopt such a strategy. Cross-linguistically speaking the sentence-external first step is not valid, either. Lieven (1997: 230–231) points out that there are languages in which children use a clearly sentence-internal negation strategy right from the beginning. From the point of view of function, it has been noted that different semantic and pragmatic functions of negatives appear at different points in language development. According to Bloom (1970) the earliest negative sentences are expressions



Negation

of nonexistence. The categories of rejection and denial are acquired later (in this order). Thus the order of acquisition is: nonexistence > rejection > denial (‘>’ meaning ‘is acquired before’). For more recent categorizations see for example Pea (1980) and Choi (1988). It has been argued that the order reflects the level of cognitive difficulty of these categories. However, some variation in the order of acquisition has been found for individual children and for different languages. The cross-linguistic variation can at least partly be explained by whether or not each negative meaning is expressed by a different negative element in a given language, while some of the differences between individual children can be explained by differences in the input. Therefore, as Lieven (1997: 230) argues, the level of cognitive difficulty cannot be the only factor underlying the similarities and differences in the order of acquisition of negative functions.

References Bernini, G. & P. Ramat (1992). La frase negativa nelle lingue d’Europa. Il Mulino. Bloom, L. (1970). Language development. M.I.T. Press. ——— (1991). Language development from two to three. Cambridge University Press. Brown, P. & S.C. Levinson (1987). Politeness. Cambridge University Press. Burton-Roberts, N. (1989). On Horn’s dilemma: presupposition and negation. Journal of linguistics 25: 95–125. ——— (1999). Presupposition-cancellation and metalinguistic negation: a reply to Carston. Journal of linguistics 35: 347–364. Capell, A. & H.E. Hinch (1970). Maung Grammar. Mouton. Carston, R. (1996). Metalinguistic negation and echoic use. Journal of pragmatics 25: 309–330. ——— (1998a) Informativeness, relevance and scalar implicature. In R. Carston & S. Uchida (eds.), Relevance theory. Applications and implications: 179–236. John Benjamins. ——— (1998b). Negation, ‘presupposition’ and the semantics/pragmatics distinction. Journal of linguistics 34: 309–350. ——— (1999). Negation, ‘presupposition’ and metapresentation: a response to Noel Burton-Roberts. Journal of linguistics 35: 365–389. Choi, S. (1988). The semantic development of negation: a cross-linguistic longitudinal study. Journal of child language 15(3): 517–531. Clark, H. (1974). Semantics and comprehension. In T.A. Sebeok (ed.), Current trends in linguistics, vol. 12, Linguistics and adjacent arts and sciences: 1291–1428. Mouton. Clark, E. & H. Clark (1977). Psychology and language. Harcourt Brace Jovanovich. Cornyn, W. (1944). Outline of Burmese Grammar. Linguistic Society of America. Croft, W. (1991). The evolution of negation. Journal of linguistics 27: 1–27. ——— (2003). Typology and universals. Second edition. Cambridge University Press. Dahl, Ö. (1979). Typology of sentence negation. Linguistics 17 (1/2): 79–106. DE Villiers, J.G. & P.A. De Villiers (1985). The acquisition of English. In D. Slobin (ed.), The cross-linguistic study of language acquisition, vol. 1: 24–140. Erlbaum. Dryer, M.S. (1992). The Greenbergian word order correlations. Language 68: 81–138.

227

228 Matti Miestamo Ducrot, O. (1972). Dire et ne pas dire. Hermann. Fauconnier, G. (1975). Pragmatic scales and logical structure. Linguistic Inquiry VI: 353–375. Forest, R. (1993). Négations: essai de syntaxe et de typologie linguistique. Klincksieck. Geurts, B. (1998). The mechanisms of denial. Language 74: 274–307. Givón, T. (1978). Negation in language: pragmatics, function, ontology. In P. Cole (ed.), Syntax and semantics vol. 9. Pragmatics: 69–112. Academic Press. ——— (1984). Syntax, a functional typological introduction. John Benjamins. ——— (2001). Syntax, an introduction. Vol. I. John Benjamins. Greenberg, J. (1966). Language universals. Mouton. Haegeman, L. (1995). The syntax of negation. Cambridge University Press. Haspelmath, M. (1997). Indefinite pronouns. Clarendon Press. Haspelmath, M., M. Dryer, D. Gil & B. Comrie (eds.) (2005). World Atlas of Language Structures. Oxford University Press. Hoeksema, J., H. Rullmann, V. Sánchez-Valencia & T. Van Der Wouden (eds.) (2001). Perspectives on Negation and Polarity Items. John Benjamins. Honda, I. (1996). Negation: a cross-linguistic study. UMI Dissertation service. Hopper, P.J. & S.A. Thompson (1980). Transitivity in grammar and discourse. Language 56: 251–299. Horn, L. (1978). Remarks on neg-raising In P. Cole (ed.), Syntax and semantics vol. 9. Pragmatics: 129–220. Academic Press. ——— (1985). Metalinguistic negation and pragmatic ambiguity. Language 61: 121–174. ——— (1989). A natural history of negation. The University of Chicago Press. [reprint with supplemental bibliography, CSLI Publications, 2001] Horn, L. & Y. Kato (eds.) (2000). Negation and polarity: syntactic and semantic perspectives. Oxford University Press. Itkonen, Esa (2005). Analogy as structure and process. Benjamins. Jacobsen, W. (1964). A grammar of the Washo language. Ph.D. Diss., University of California, Berkeley. Jespersen, O. (1917). Negation in English and other languages. Høst. ——— (1924). The philosophy of grammar. Allen & Unwin. Kahrel, P. (1996). Aspects of negation. Ph.D. Diss., University of Amsterdam. Klima, E. (1964). Negation in English In J.A. Fodor & J.J. Katz (eds.), The structure of language: 246–323. Prentice Hall. Klima, E. & U. Bellugi (1966). Syntactic regularities in the speech of children In J. Lyons & R.J. Wales (eds.). Psycholinguistics papers: 183–208. Edinburgh University Press. Ladusaw, W.A. (1980). Polarity sensitivity as inherent scope relations. Garland Press. Leech, G.N. (1983). Principles of pragmatics. Longman. Lieven, E. (1997). Variation in a cross-linguistic context in D. Slobin (ed.), The cross-linguistic study of language acquisition, vol. 5. Expanding the context: 199–264. Erlbaum. Mazzon, G. (2004). A history of English negation. Pearson. Miestamo, M. (2005). Standard negation: the negation of declarative verbal main clauses in a typological perspective. Mouton de Gruyter. Payne, J.R. (1985). Negation. In T. Shopen (ed.), Language typology and syntactic description, vol. I: 197–242. Cambridge University Press. Pea, R. (1980). The development of negation in early child language. In D.R. Olson (ed.), The social foundations of language and thought: 156–186. Norton. Ramat, P. & G. Bernini (1990). Area influence versus typological drift in Western Europe: the case of negation. In J. Bechert, G. Bernini & C. Buridant (eds.), Toward a typology of European languages: 25–46. Mouton de Gruyter.



Negation 229

Tottie, G. (1977). Fuzzy negation in English and Swedish. Almqvist & Wiksell. ——— (1991). Negation in English speech and writing: a study in variation. Academic Press. Tovena, L. (2001). Neg-raising: negation as failure. In Hoeksema et al. (eds.): 331–356. Van Der Auwera, J. & L. Lejeune, with V. Goussev (2005). The prohibitive. In Haspelmath et al. (eds): 290–293. Van Der Auwera, J. (2006). Why languages prefer prohibitives. Wai guo yu-Journal of foreign languages 161: 2–25. Van Der Sandt, R.A. (1993). Denial. CLS 27(2): 331–344. Van Der Wouden, T. (1997). Negative contexts: collocation, polarity and multiple negation. Routledge. Van Kuppevelt, J. (1996). Inferring from topics, scalar implicatures as topic-dependent inferences. Linguistics and philosophy 19: 393–443. Von Wright, G.H. (1959). On the logic of negation. Societas scientiarum Fennica: Commentationes physico-mathematicae XXII 4. Wason, P.C. (1965). The contexts of plausible denial. Journal of verbal learning and verbal behavior 4: 7–11.

Prague school Petr Sgall Charles University, Prague

1.  Historical overview In the Prague school, one of the first and main centers of European structural linguistics, principles of a synchronic theory of language were established as substantial starting points not only for a theory of language systems, especially of phonology, but also for what may be understood today as having preceded modern sociolinguistics and text linguistics. Moreover, the systemic character of the development of language was first characterized in this context. Prague functionalism is of importance both for analyzing language as a system linking expressions and meanings and for pragmatic studies of the interactive features of communication. It differs from other trends of structural linguistics by concentrating on semantic functions of language units and on pragmatic aspects of the use of language and of its varieties. Therefore the achievements of its classical period, although not stated in the shape of a fully explicit formal framework, still offer insights fruitful for inquiries into pragmatically oriented linguistic studies. Sentence structure itself has been understood as comprising a patterning conditioned by the interactively based role of the sentence in context, in discourse. Even though the Prague School tradition was interrupted by the war events and later severely hampered by the ruling totalitarian ideology, studies have continued. Empirical and theoretical frameworks handling certain pragmatic aspects of discourse patterns have been elaborated. Along with a broad range of results in various synchronic and diachronic domains of general linguistics, of Slavistics, of English studies and so on, a specific approach to the formal description of language has been formulated. The main sources of the Prague school approach came from trends of linguistic thinking (around the beginning of this century) countering the earlier prevailing Neogrammarian views. The most important factor was Saussure’s program for a synchronic account of language as a system of signs (in the bilateral sense of ‘sign’ as a pair of signified and signifying), langue, distinguished from the use of language in discourse, parole. In 1911, V. Mathesius, a specialist in English at Charles University, Prague, analyzed issues of synchronic variation. He found inspiration with O. Jespersen, F. Brunot, W. Horn and A. Marty, and also with Polish and Russian linguists. Ideas of the Moscow school, as well as the insight into intrinsic connections between linguistics and poetics as studied by the Russian Formalists, were brought to Prague by R. Jakobson at the



Prague school

beginning of the 1920’s. In 1926 Mathesius, together with Jakobson and with several other young specialists, founded the Prague Linguistic Circle, which presented its Theses (1929) to the First Congress of Slavic Philologues. As was soon recognized by the international linguistic community, the School was successful in submitting a new theory, phonology, describing the sound patterns of languages as specific subsystems viewed from a functional angle. A synthetic treatment of this subfield was presented by Trubetzkoy (1939). Mathesius (1929, 1939) devoted a concentrated effort to a structural description of the topic-focus articulation. Jakobson (1929) opened new horizons in disclosing the systematic character of language development (determined by its goal-oriented character, not just causally), the invariant nature of semantic functions of morphemes (1936) and the universality of laws having the form of implication (1958), which were later intensively exploited by J. Greenberg and others. R. Jakobson and J. Mukařovský presented a many-sided analysis of the interface between linguistics and poetics. Jakobson’s (1960) views of the functions of language (and/or of discourse), based on K. Bühler’s approach, are still of great importance; see Figure 1. context reffered to REFERENTIAL adresser EMOTIVE

message POETIC

adressee CONATIVE

contact PHATIC code METALINGUAL Figure 1.

Karcevskij (1929) pointed out that vagueness is a necessary property of the meaning of linguistic units, which denote the ever changing and thus unlimited phenomena of reality as perceived by humans. Issues of stylistics, of the standard language and of other varieties of national languages were analyzed by Havránek (1929, 1932, 1938). B. Trnka’s writings started the phonological analysis of English and were of great importance for the understanding of the role of structural linguistics in the thought pattern of the first half of this century. The development of the school can be followed in various collections of articles published in the West.1 After 1989, thanks to the initiative of O. Leška, the activities

.  See especially Vachek (1964, 1966, 1983) and Luelsdorff et al. (1994). Crucial moments in the history of the Circle are recalled and analyzed by Toman (1991). Holenstein (1975) presents a deepened

231

232 Petr Sgall

of the circle (stopped first in 1939 and then again at the beginning of the 1950’s) have been resumed. Also the series Travaux du Cercle linguistique de Prague is being published again (with a parallel English title, Prague Linguistic Circle Papers).

2.  Main concepts and fields of research Among the important results achieved by the Prague School, only those which have or offer an immediate impact on studies in pragmatics can be mentioned here. 2.1  Jakobson’s (1941, 1960, 1964) analysis of the teleonomic (‘means-ends’, goaloriented) character of language as a tool serving a function and adapted to the corresponding needs of human communication was further discussed by Shapiro (1991), Leška (1991) and others. One of the manifestations of this feature of language can be seen in the anthropocentric character of such basic areas as the lexicon, or the core of syntax, cf. Skalička (1962). Inspired by the teleonomic approach and by V. Mathesius’ ‘characterology of languages’, Skalička (1935, 1979) came back to the tradition of W. von Humboldt, G. von der Gabelentz, E. Sapir and others and elaborated a structuralist theory of typology, weakening Jakobson’s implicational laws in the probabilistic sense. He understood the types as ideal extremes, not fully realized in real languages, in each of which properties of different types are combined with each other. Even so, the predictive power of his typology can well be compared with that of the word order based approach, since the clusters of cognate properties comprise phenomena from all levels of language. The basic properties of types of languages can be understood as pertaining to a fundamental dichotomy of language structure, namely to the opposition of lexicon and grammar (see Sgall 1986). While lexical values always are expressed by strings of phonemes, grammatical values can be expressed basically in one of the following ways, each of which is favorable to the cluster of properties constituting a type: (i) morphs having the shape of words (analysis or ‘isolation’), (ii) morphs having the shape of affixes (agglutination), (iii) alternations at the end of lexical morphs, i.e. endings (inflection), (iv) or in the inner parts of lexical morphs (introflexion), or (v) order of lexical morphs (‘polysynthesis’).

analysis of the life and work of R. Jakobson. A volume of continuations of the classical School was edited by Tobin (1988). Also the trends initiated by C.H. van Schooneveld and by A. Martinet, and such personalities as E.M. Uhlenbeck or K. Heger have substantially enriched the Prague tradition.



Prague school

2.2  A possibility of finding an appropriate position for Jakobson’s ‘semantic universals’ in an overall description of language, including the sentence structure, was examined by Pauliny (1943), who understood the functions of cases as being primarily linked with the valency or ‘intention’ of the verb, i.e. of its optional and obligatory complementations. Most important for the fundamentals of structural syntax have been the writings of Tesnière (1959), representing the first systematic elaboration of dependency (valency-based) syntax. A similar approach was applied to Czech by Šmilauer (1947).2 Dokulil & Daneš (1958) characterized the level of ‘sentence meaning’ (one of the first versions of an underlying level of syntax) as a linguistic patterning of the layer of ontological content (using recent terms, one can speak of the interface level of sentence structure as opposed to semantic-pragmatic interpretation). Basic issues of dependency syntax were discussed from a closely related viewpoint (stressing the notions connected with markedness, and also the syntactic roles of cases) by Kuryłowicz (1949). 2.3  One of the pragmatically based aspects of the structure of the sentence, namely its division into topic and focus (theme and comment or rheme), based on ‘given’ and ‘new’ information, was discussed by V. Mathesius. Firbas (1957, 1992) added a third element to the dichotomy (the transition, consisting primarily in the temporal and modal elements of the verb) and introduced the scale of communicative dynamism as going from theme proper (typically at the beginning of the sentence) to rheme proper (at its end). Daneš (1974) pointed out how the ‘thematic progressions’ relate the topicfocus dichotomy of an utterance to its position in a discourse (the theme often being coreferential with the rheme or with the theme of the preceding utterance). It was then possible to show that the topic-focus articulation, although anchored in the pragmatic layer (in the position of the utterance in the discourse), intrinsically belongs to the structure of the sentence (seen as an abstract unit of the language) and is relevant for the semantics of the sentence in the narrow sense (see Sgall 1967). 2.4  The development of the Circle’s views on stylistics started with Havránek’s (1929, 1932) formulations on ‘functional languages’ (comparable to what more recently has been called ‘sublanguages’) and (individuating) styles. Mathesius (1942: 35f) characterized style as (i) typical of an individual language, (ii) of an author, and (iii) of a kind of discourse. .  Šmilauer was one of the linguists opposing the circle in the polemical discussion on the norm of Standard Czech; he was one of the ‘purists’, against whose prescriptive doctrine Havránek (1932) and other structuralists duly protested, showing that the stylistic richness of the language should be applied without narrowing inhibitions, with various lexical and grammatical items displaying functions of their own.

233

234 Petr Sgall

The anaphoric and coreferential relationships in the discourse are based on the degrees of salience of the elements of the stock of information proper to the hearer (according to the speaker’s assumptions), as discussed by Hajičová (in Sgall et al. 1986, Chapter 3). This pragmatic factor is relevant for anaphoric relations in the text, coreference, which have to be delimited from grammatical coreference (see Panevová 1986). The intrinsic closeness of stylistics and of the functional stratification of language is clearly to be seen in the complex situation of Czech. This fact (the description of which partly still suffers from residues of a prescriptive attitude) has been discussed on several occasions. Sgall et al. (1992) also handle some of the relevant sociolinguistic issues, including such pragmatic aspects as the speakers’ attitudes towards non-standard items in everyday speech. 2.5   The attention given in the Prague School to issues of pragmatics presents a challenge for the ideas of the School to be applied within the modern methodological trend of formal linguistics. In this way a framework could be formulated that would describe language not only as an abstract system of sentence patterns internalized by an ideal speaker, but rather as an interactive system displaying a wide range of variation, presenting options for a development controlled by the requirements of communication. A starting point of the elaboration of such a complex framework can be seen in the functional generative description as developed step by step in Sgall et al. (1969), Panevová (1974/1975), Hajičová (1972, 1992) and elsewhere, with a comprehensive view presented in Sgall et al. (1986). The framework (presented in its formal shape by Petkevič 1987) has been designed so as to handle sentence structure in its anthropocentric aspects, i.e. based on syntactic dependency (the core of which, at least in most European languages, is the human-like pattern of actor and action) and comprising the topic-focus articulation, i.e. specifying sentences not just as abstract objects, but as anchored in interactive context (which opens a way to understanding them as operations on the hearer’s states of mind). Much grammatical information is included into the lexical data, especially in the valency frames, where the kinds of complementation (Actor, Addressee, Patient or Objective, Source, Locative, several modifications of Direction, Time, Condition, and so on) are specified. Dependency allows for a rather natural specification of the ‘free’ modifications (adjuncts). It also offers a basis for a relatively economical description, since the sentence representation having the shape of a dependency tree (with the verb as its root) contains only nodes corresponding to lexical items proper, rather than to nonterminals and function morphemes. Since the functional generative description is more oriented towards the interactive nature of language and requires just a simple set of descriptional devices (structures close to rooted trees, or to strings with indexed brackets), it constitutes a challenge for the Chomskyan universal grammar, presenting a starting point for discussions on the general prerequisites of language acquisition.



Prague school

These representations of sentences are suitable as input also for those aspects of semantic interpretation which account for the scopes of operators. As was stated by E. Hajičová (1972; see also Sgall et al. 1986, Chapter 3), the topic-focus articulation is relevant for the scopes and in some cases it determines whether a presupposition is triggered (especially by a definite noun group in the topic) or not (by such a group in the focus). In the prototypical case the scope of negation is identical with the focus of the sentence; cf. such examples as The king of France is (not) bald versus Yesterday Prague was (not) visited by the king of France (in the latter sentence the presupposition of the existence of such a person is absent), or Our victory was (not) caused by Jim vs. Jim caused (didn’t cause) our victory (only in the former sentence the victory is being presupposed). 3.  Prague functionalism and pragmatics 3.1   The concept of function has been contrasted both with ‘means’ and ‘form’. Thus the functions of the units of language are understood both in the syntactic sense (with elementary units functioning in complex wholes) and in the semantic sense (with units of both kinds seen in relation to their meanings). The latter view, opposing function to form, is of immediate importance for pragmatic studies, since the patterning of natural language does not directly reflect the trichotomy of syntax, semantics and pragmatics. Especially the two last domains do not correspond to two distinct components of language, although in some fields it is possible at least in principle to distinguish between semantic and pragmatic functions of linguistic items. While some of the Praguians claim, e.g. that synonymy, i.e. identity of meaning, is never absolute, it may be required to distinguish between a semantic and a stylistic value: such nouns as stay and sojourn share their meaning, but differ stylistically (in the bookish value of the latter); even the stylistic difference may be absent (cf. the synonymy of the endings of different morphological paradigms; or that of the nouns noun and substantive). Opposing function to form, Saussure’s signified to signifying, offers a basis for understanding language as a set of levels, ordered either from meaning to expression (close to the speaker’s view, stressed by V. Mathesius), or in the opposite direction. Thus the levels of phonetics, phonemics, morphophonemics, morphemics, surface syntax (Dokulil’s ‘grammatical structure of the sentence’) and underlying syntax (meaning) were established.3

.  Some of the pairs of adjacent levels always have been discussed as constituting less substantial oppositions than others. The Prague school standpoint differs from Chomsky’s in stressing phonology and assigning to morphophonemics only a secondary (intermediate) role. As to syntax, in both these trends a development can be found from stressing surface to deposing it.

235

236 Petr Sgall

If natural language is duly understood in its interactive character, then the topicfocus articulation can be seen as an aspect of sentence structure. This makes it possible to account for the fact that this pragmatically based articulation is relevant for truth conditions (or for situationally anchored semantics), especially as concerns the scopes of operators. Moreover, the semantics of utterances (sentence tokens) can only be specified (be it in the shape of Carnapian propositions or in another way) with the inclusion of the assignment of reference, since the truth or falsity of an occurrence of, e.g. The table is too old depends on which table is being referred to. Thus, the semiotic claim that pragmatics includes semantics can only be understood as a working abstraction; as concerns natural language, pragmatic factors (such as those pertinent for reference assignment, i.e. especially the degrees of salience) are relevant even for the basic layer of semantics. 3.2   Each of the levels established according to the opposition of function and form displays its own syntax; its more elementary units have certain functions in more complex wholes. Typical of the Prague views, then, is a series such as: distinctive feature – phoneme – morph(eme) – word – sentence (– discourse)

The pragmatic dimension is present with the last two notions in this scale, and it has been claimed that the relationship between sentence and discourse differs from the other oppositions involved. It is not an oppositon between a constituent part and a whole, but rather between language and communication. As we have seen above, already in the classical period of the School, language was viewed as an ingredient of communication, i.e. in a fundamentally pragmatic understanding. This has been a fruitful basis for approaches to style, to discourse patterns and to the social and functional stratification of language. In this context, the teleonomic understanding finds a natural substantiation: language and its development are conditioned by the function of language in communication. The task then is to explain how the development of the intersubjective system is influenced by individual intentions of the speakers, who want to be both understandable (avoiding too much ambiguity) and economical (reducing redundance). Jakobson’s (1941) ‘therapeutic’ changes restore the balance in the phonological and other subsystems, so that language can cope with the needs of communication. It remains to find means of identifying the boundaries of these needs: if language behaves as a goal-oriented system, what is its regulating mechanism?

References Daneš, F. (1974). Functional sentence perspective and the organization of the text. In F. Daneš (ed.), Papers on functional sentence perspective: 106–128. Academia.



Prague school

Dokulil, M. & F. Daneš (1958). K tzv. významové a mluvnické stavbě věty. [On the so-called semantic and grammatical structure of the sentence.] In O vědeckém poznání soudobých jazyků: 231–246. Academia. Firbas, J. (1957). Some thoughts on the function of word-order in Old English and Modern English. In Sborník prací filosofické fakulty brněnské university A5: 723–100. ——— (1992). Functional sentence perspective in written and spoken communication. Cambridge University Press. Hajičová, E. (1972). Some remarks on presuppositions. Prague Bulletin of Mathematical Linguistics 17: 11–23. ——— (1992). A challenge for universal grammar. XVth International Congress of Linguists, Plenary session texts: 50–61. University Laval. Havránek, B. (1929). Influence de la fonction de la langue littéraire sur la structure phonologique et grammaticale du tchèque littéraire. Travaux du Cercle linguistique de Prague 1: 106–120. ——— (1932). Úkoly spisovného jazyka a jeho kultura. [The tasks of the Standard language and its culture.] In Spisovná čeština a jazyková kultura. Melantrich. ——— (1938). Zum Problem der Norm in der heutigen Sprachwissenschaft und Sprachkultur. Actes du IVe Congrès international des linguistes: 151–156. Holenstein, E. (1975). Roman Jakobsons phänomenologischer Strukturalismus. Suhrkamp. Jakobson, R. (1929). Remarques sur l’évolution phonologique du russe comparée à celle des autres langues slaves. Travaux du Cercle linguistique de Prague 2. ——— (1936). Beitrag zur allgemeinen Kasuslehre. Travaux du Cercle linguistique de Prague 6: 240–288. ——— (1941). Kindersprache, Aphasie und allgemeine Lautgesetze. Suhrkamp. ———(1958). Typological studies and their contribution to historical comparative linguistics. Proceedings of the Eighth International Congress of Linguists: 17–35. Oslo University Press. ——— (1960). Closing statement: Linguistics and poetics. In T.A. Sebeok (ed.), Style in language: 350–377. MIT Press. ——— (1964). Efforts towards a means-ends model of language in interwar Continental linguistics. In J. Vachek (ed.): 481–485. Karcevskij, S. (1929). Du dualisme asymétrique du signe linguistique. Travaux du Cercle linguistique de Prague 1: 88–93. Kuryłowicz, J. (1949). Le problème du classement des cas. Biuletyn Polskiego towarzystwa językoznawczego 9: 20–43. Leška, O. (1991). Phenomenological inspiration of Prague structuralism. Proceedings of Linguistics and Phonetics ‘90: 97–99. Charles University. Luelsdorff, P., J. panevová & P.Sgall (eds.) (1994). Praguiana 1945–1990. John Benjamins. Mathesius, V. (1929). Zur Satzperspektive im modernen Englisch. Archiv für das Studium der neueren Sprachen und Literaturen 155: 202–210. ——— (1939). O tak zvaném aktuálním členění větném. [On information-bearing structure of the sentence.] Slovo a slovesnost 5: 171–174. ——— (1942). Řeč a sloh. [Speech and style.] In B. Havránek & J. Mukařovský (eds.), Čtení o jazyce a poesii: 10–100. Družstevní práce. Panevová, J. (1974/1975). On verbal frames in functional generative description. Prague Bulletin of Mathematical Linguistics 22: 3–40/23: 17–52. ——— (1986). The Czech infinitive in the functions of Objective and the rules of coreference. In J.L. Mey (ed.), Language and discourse: 123–142. John Benjamins. Pauliny, E. (1943). Štruktúra slovenského slovesa. [The structure of the Slovak verb.] Bratislava.

237

238 Petr Sgall Petkevič (1987). A new dependency based specification of underlying representations of sentences. Theoretical Linguistics 14: 143–172. Sgall, P. (1967). Functional sentence perspective in a generative description. Prague Studies in Mathematical Linguistics 2: 203–225. ———(1986). Classical typology and modern linguistics. Folia Linguistica 20: 15–28. ——— (ed.) (1984). Contributions to functional syntax, semantics, and language comprehension. John Benjamins/Academia. Sgall, P., L. Nebeský, A. Goralčĺková & E. Hajičová (1969). A functional approach to syntax in generative description of language. American Elsevier. Sgall, P., E. Hajičová & J. Panevová (1986). The meaning of the sentence in its semantic and pragmatic aspects. Reidel/Academia. Sgall, P., J. Hronek, A. Stich & J. Horecký (1992). Variation in language. John Benjamins. Shapiro, M. (1991). The sense of change. Indiana University Press. Skalička, V. (1935). Zur ungarischen Grammatik. Charles University. ——— (1962). Das Wesen der Morphologie und der Syntax. Acta Universitatis Carolinae, Slavica Pragensia 4: 123–127. ——— (1979). Typologische Studien. Viehweg. Šmilauer, V. (1947). Novočeská skladba. [The syntax of Modern Czech.] Mikota. Tesnière, L. (1959). Eléments de syntaxe structurale. Klincksieck. Thèses (1929). Thèses présentées au Premier Congrès des philologues slaves. Travaux du Cercle linguistique de Prague 1: 5–29. Tobin, Y. (ed.) (1988). The Prague school and its legacy in linguistics, literature, semiotics, folklore, and the arts. John Benjamins. Toman, J. (1991). Prague School in the cultural context. Chicago University Press. Trubetzkoy, N. (1939). Grundzüge der Phonologie. Travaux du Cercle linguistique de Prague 7. Vachek, J. (1966). The linguistic school of Prague. Indiana University Press. ———(ed.) (1964). A Prague School reader in linguistics. Indiana University Press. Vachek, J. (ed.) (1983). Praguiana. John Benjamins/Academia.

Role and Reference Grammar Robert D. Van Valin, Jr. University of Duesseldorf & University at Buffalo

1.  Introduction Role and Reference Grammar (RRG, Van Valin 1993a, 2005, 2008; Van Valin & LaPolla 1997; Kailuweit, et al. 2008) may be termed a ‘structural-functionalist theory of grammar’. This locates it on a range of perspectives from extreme formalist at one end to radical functionalist at the other. RRG falls between these two extremes, differing markedly from each. In contrast to the extreme formalist view, RRG views language as a system of communicative social action, and consequently, analyzing the communicative functions of morphosyntactic structures has a vital role in grammatical description and theory from this perspective. Language is a system, and grammar is a system in the traditional structuralist sense. What differentiates the RRG conception of grammar from the standard formalist one is the view that grammatical structure can only be understood and explained with reference to its semantic and communicative functions. Syntax is not autonomous; rather it is viewed as relatively motivated by semantic and pragmatic factors. In terms of the abstract paradigmatic and syntagmatic relations that define a structural system, RRG deals not only with relations of cooccurrence and combination in strictly formal terms but also with semantic and pragmatic cooccurrence and combinatory relations. Hence RRG may be properly designated as a structural-functionalist theory, rather than purely formalist or purely functionalist. 2.  Historical background RRG grew out of an attempt to answer two fundamental questions: 1. what would linguistic theory look like if it were based on the analysis of Lakhota, Tagalog and Dyirbal, rather than on the analysis of English?, and 2. how can the interaction of syntax, semantics and pragmatics in different grammatical systems best be captured and explained? These questions reflect issues that were prominent in the mid-1970s in some strands of American linguistics. Dixon’s grammar of Dyirbal and Schachter & Otanes’ grammar of Tagalog had been published in 1972, and the implications of these languages for linguistic theories were just being recognized. Furthermore, the Prague School and Hallidayan ideas regarding the role of discourse-pragmatics in grammar were being

240 Robert D. Van Valin, Jr.

explored from a number of different perspectives. Many of the typological issues, e.g. the universality of the notion of ‘subject’, and theoretical issues, e.g. the relation between ‘subject’ and ‘topic’ in grammatical systems, were central in the initial conceptualization of RRG, and this is reflected in the early work on the theory (Foley & Van Valin 1977; Van Valin 1977a, 1977b, 1980, 1981; Van Valin & Foley 1980). The theory from which RRG is most directly descended is Fillmore’s (1968) case grammar. As in Fillmore’s model, there is a semantic representation employing semantic case roles which is mapped into the syntactic surface structure, without any intervening level of syntactic representation. The details of the mapping differ substantially, however, and one prime difference is that in RRG discourse-pragmatic factors may play a role in the mapping. The RRG theory of clause structure follows the insight of Fillmore’s division of the clause into ‘modality’ and ‘proposition’ by treating predicates, arguments and their modifiers distinctly from grammatical categories such as tense, aspect, modality and mood. Finally, RRG, like Fillmore, does not assume grammatical relations to be universal. 3.  Central concepts 3.1  Clause structure RRG rejects the standard formats for representing clause structure (grammatical relations, X-bar syntax), because they are not universal and hence necessarily impose aspects of structure on at least some languages where it is not appropriate. The RRG conception of clause structure (originally proposed in Foley & Van Valin 1984 and further refined in Van Valin 1993a and Van Valin & Lapolla 1997), the layered structure of the clause (LSC), is made up of the nucleus, which contains the predicate(s), the core, which contains the nucleus plus the arguments of the predicate(s), and the peripheries, which contains adjunct modifiers of the nucleus, core and clause. These aspects of the LSC are universal; in addition, some languages have a pre-core slot (PrCS), which is the position of WH-words in languages like English and Icelandic, and a left-detached position (LDP), which is the position of the pre-clausal element in a left-dislocation construction.1 Each of the major layers (nucleus, core, clause) is modified by one or more operators which include grammatical categories such as tense, aspect, modality and evidentiality. In the formal representation of the LSC (proposed in Johnson 1987), operators are represented in a distinct projection of the clause from the predicates and arguments. This is presented in Figure 1.

1.  There is also a post-core slot (Shimojo 1995) and a right-detached position is some languages (Van Valin & LaPolla 1997).



Role and Reference Grammar SENTENCE

SENTENCE (LDP)

LDP

CLAUSE

(PrCS) ARG

CORE

CLAUSE CORE

PrCS

(PERIPHERY)

(ARG) NUCLEUS

ARG

NUC

(XP)

V

PP/Adv

NUCLEUS

Aspect

NUCLEUS/CORE

ARG

PRED

PRED XP XP XP

PERIPHERY

ADV

NP

NP

V

Yesterday, what did John

Directionals

CORE

Modality

CORE

Negation (internal)

CLAUSE

Status

CLAUSE

Tense

CLAUSE

Evidentials

CLAUSE

Illocutionary Force

give

PP

PP

to Mary in the library?

NUCLEUS CORE TNS IF

CLAUSE CLAUSE SENTENCE

SENTENCE

Figure 1.

Representations such as these should be viewed as syntactic templates, the inventory of which in a language constitutes an important component of its grammar; it may be termed the syntactic inventory and complements the lexicon. The LSC applies equally to fixed word-order and free word-order languages, to head-marking and dependentmarking languages, to languages with and without grammatical relations.2 The three central components of the LSC also turn out to be the three fundamental building blocks of complex sentences in human language. The unmarked pattern for the construction of complex sentences involves combining nuclei with nuclei, cores with cores, or clauses with clauses. These are called levels of juncture in RRG, i.e. nuclear juncture, core juncture and clausal juncture. Examples of nuclear, core and clausal juncture from English are given in (1).3

2.  It is assumed that noun phrases and adpositional phrases have a comparable layered structure; see Van Valin (1993a, Section 1.7), Van Valin & La Polla (1997, §2.3). 3.  While it is alien to the typological, universalist character of RRG to give examples only from English, the severe space constraints imposed on this article make this necessary as a space-saving strategy.

241

242 Robert D. Van Valin, Jr.

(1) a. Fred pushed open the door. [two nuclei, ‘push’ and ‘open’, in a single core] b. Fred persuaded Bill to open the door. [two cores, ‘Fred persuaded Bill’ and ‘Bill open the door’, in a single clause] c. Fred talked to Mary, and she agreed with his suggestion. [two clauses in a single sentence]

Of equal importance in the RRG theory of complex sentences is the set of possible syntactic and semantic relations between the units in a juncture.4 The syntactic relations between units are called nexus relations in RRG. Traditionally, only two basic nexus relations are recognized, coordination and subordination, but RRG, following Olson’s (1981) analysis of clause linkage in Barai (a Papuan language), posits three nexus types: coordination, subordination, and cosubordination, which is essentially dependent coordination. Subordination and cosubordination are illustrated in (2) with examples of clausal juncture from English; (1c) is an example of clausal coordination. (2) a. Max called Sue, because he was going to be late for the party. [Subordination] b. Having called Sue, Max left for the party. [Cosubordination]

The three levels of juncture combine with the three nexus types to generate nine possible complex sentence types. Not all of them are instantiated in every language, and the types found in a language may be realized by more than one formal construction type. The nine juncture-nexus types may be ordered into a hierarchy in terms of the tightness of the syntactic link between the units (see the hierarchy in Figure 2 below). 3.2  Semantic structure The heart of the RRG approach is the system of lexical representation and semantic roles. The system of lexical representation in Van Valin (2005) is based on the scheme for lexical decomposition proposed in Dowty (1979), which is in turn based on Vendler’s (1967) classification of verbs into states, activities, achievements and accomplishments. Examples of each class and their formal representation are given in (3). (3) a. b. c. d. e. f. g. h.

State: The lamp is broken. broken′ (lamp) Achievement: The lamp shattered. INGR shattered′ (lamp) Accomplishment: The lamp melted. BECOME melted′ (lamp) Activity: The lamp shone brightly. do′ (lamp, [shine′ (lamp)]) Semelfactive: The lamp flashed. SEML do′ (lamp, [flash′ (lamp)]) Active accomplishment: The lamp rolled to the door. do′ (lamp, [roll′ (lamp)]) & INGR be-at′ (door, lamp) Casusative state: The lamp frightens the child. [do′ (lamp, Ø)] CAUSE [feel′ (child, [afraid′])] Causative achievement: The child shattered the lamp. [do′ (child, Ø)] CAUSE [INGR shattered′ (lamp)]

4.  The semantic relations will be discussed in 3.2. below.



Role and Reference Grammar

i. j. k. l.

Causative accomplishment: The heat melted the lamp. [do′ (heat, Ø)] CAUSE [BECOME melted′ (lamp)] Causative activity: The child rolled the lamp. [do′ (child, Ø)] CAUSE [do′ (lamp, [roll′ (lamp)])] Causative semelfactive: The child flashed the lamp. [do′ (child, Ø)] CAUSE [SEML do′ (lamp, [flash′ (lamp)])] Causative active accomplishment: The child rolled the lamp to the door. [do′ (child, Ø)] CAUSE [do′ (lamp, [roll′ (lamp)]) & INGR be-at′ (door, lamp)

A crucial component of this system is a set of syntactic and semantic tests for determining the class membership of a verb in a particular sentence, since the class of the verb determines its lexical representation or logical structure [LS] (see Van Valin 1993a: 35). Examination of the verbal systems of a number of languages had led to the conclusion that this set of distinctions is one of the fundamental organizing principles of verbal systems in human language.5 The RRG theory of semantic roles is rather different from that of other theories, in that it posits two types or tiers of semantic roles. The first are specific thematic relations, the traditional (since Fillmore 1968 and Gruber 1965) notions of agent, theme, patient, experiencer, etc. The second are generalized semantic roles called semantic macroroles; they were introduced in Van Valin (1977b) and have no analog in other theories. Following the ideas of Gruber (1965) and Jackendoff (1976), RRG derives thematic relations from argument positions in LSs such as those in (3); they are in effect labels for LS argument positions. Because the LS of a verb is determined by syntactic and semantic tests which make no reference to thematic relations, the assignment of thematic relations to a verb is independently motivated in RRG. The second type of semantic role plays a crucial role in the theory; macroroles act as the primary interface between the LS and syntactic representations. There are only two macroroles, actor and undergoer corresponding to the two primary arguments in a prototypical transitive relation. They are called ‘macroroles’ because each subsumes a number of specific thematic relations; the relationship between the two tiers is captured in the Actor-Undergoer Hierarchy in (4). (4)





ACTOR

Arg of DO [

UNDERGOER

1st arg of 1st arg of 2nd arg of Arg of state do′ (x,… pred′ (x,y) pred′ (x,y) pred′ (x) ’ = increasing markedness of realization of argument as macrorole]

5.  For further development of this decompositional system, see Van Valin & Wilkins (1993), Wilkins & Van Valin (1993).

243

244 Robert D. Van Valin, Jr.

Given the LS of a verb, the most agent-like argument will be actor, the most patientlike undergoer, in the default case. Macroroles are not equivalent to grammatical relations; this is illustrated in (5). (5)

a. b. c. d.

The boy [SUBJ, ACTOR] ate the sandwich [OBJ, UNDERGOER]. The sandwich [SUBJ, UNDERGOER] was eaten by the boy [ACTOR]. The girl [SUBJ, ACTOR] ran down the stairs. The girl [SUBJ, UNDERGOER] got sick.

The exact role of thematic relations and macroroles in the mapping (or linking) between semantic and syntactic representations will be sketched in 3.4. and summarized in Figure 4. As mentioned in the previous section, an important component of the theory of complex sentences is the semantic relations that obtain between units in a juncture. These include causality, psych-action, direct perception, cognition, propositional attitude, conditional, and varieties of temporal sequence. These may be ordered into a hierarchy in terms of whether the units in the juncture express facets of a single event, state or action or distinct events, states or actions. This semantic hierarchy interacts with the syntactic hierarchy of juncture-nexus types as follows: there is an iconic relation between the semantics and syntax of clause linkage, such that the tightness of Strongest Nuclear Cosubordination Nuclear Subordination Daughter Peripheral

Nuclear Coordination Core Cosubordination Core Subordination Daughter Peripheral

Core Coordination Clausal Cosubordination Clausal Subordination Daughter Peripheral

Clausal Coordination Sentential Subordination Sentential Coordination Weakest Syntactic Relations Figure 2.

Closest Causative [1] Phase Manner Motion Position Means Psych-Action Purposive Jussive Causative [2] Direct Perception Indirect perception Propositional Attitude Cognition Indirect Discourse Direct Discourse Circumstances Reason Conditional Concessive Simultaneous Actions Sequential Actions Situation-Situation: Unspecified Loosest Semantic Relations



Role and Reference Grammar

the syntactic linkage directly reflects the semantic integration of the units in the linkage. This is expressed in the Interclausal Relations Hierarchy in Figure 2. 3.3  Focus structure The issue of the distribution of information in clauses and sentences was not addressed in Foley & Van Valin (1984), and in Van Valin (1993a) Lambrecht’s (1986, 1987, 1994) theory of focus structure is integrated into RRG. Focus structure is the grammatical system which serves to indicate the scope of the assertion in an utterance in contrast to the pragmatic presupposition, and it is vital to the RRG analysis of many grammatical phenomena. The focus structure of an utterance is represented in a projection of the clause distinct from the operator and constituent projections; this is exemplified in Figure 3. Operator Projection

Constituent Projection

SENTENCE

SENTENCE

IF

CLAUSE

CLAUSE

TNS

CLAUSE

PCS

PERIPHERY

CORE

CORE

ARG NUC ARG

NUC

PRED

V

NP What

NP did

ARG

John

V

ADV NP

give Mary

ARG NUC ARG

yesterday ADV

S A Focus Structure Projection Figure 3.

3.4  Grammatical relations and linking In the earliest work on RRG it was argued that grammatical relations like subject and direct object are not universal and cannot be taken as the basis for adequate grammatical

245

246 Robert D. Van Valin, Jr.

theories. In place of these notions, RRG employs the notion of privileged syntactic argument [PSA], which is a construction-specific relation and is defined as a restricted neutralization of semantic roles and pragmatic functions for syntactic purposes. The other arguments in a clause are characterized as direct or oblique core arguments; there is nothing in RRG corresponding to direct or indirect object. The linking system relating semantic and syntactic representations is summarized in Figure 4. Syntactic functions like pivot and direct core argument (which are structurally instantiated in the LSC) represent the syntactic pole of the system, while LSs represent the semantic pole. The linking between LSs and macroroles is universal, and cross-linguistic variation, e.g. accusative vs. ergative syntactic systems, is located in the mapping between macroroles and syntactic functions. Direct Core Arguments

Oblique Core Arguments Languagespecific

SYNTACTIC FUNCTIONS: PSA

Privileged Syntactic Argument [PSA] Selection: Highest ranking MR = default (e.g. English) Lowest ranking MR = default (e.g. Dyirbal) SEMANTIC MACROROLES: ACTOR Arg of DO

1st arg of do′ (x,…

1st arg of pred′ (x,y)

2nd arg of pred′ (x,y)

UNDERGOER Arg of state pred′ (x)

Argument Positions in LOGICAL STRUCTURE

Verb Class

Logical Structure

Universal

Transitivity = No. of Macroroles Transitive = 2 Intransitive = 1 Atransitive = 0

predicate′ (x) or (x, y) STATE do′ (x, [predicate′ (x) or (x, y)]) ACTIVITY INGR predicate′ (x) or (x, y) ACHIEVEMENT SEML predicate′ (x) or (x, y) SEMELFACTIVE ACCOMPLISHMENT BECOME predicate′ (x) or (x, y) ACTIVE ACCOMPLISHMENT do′ (x, [predicate1′ (x, (y))]) & INGR predicate2′ (z, x) or (y) α CAUSE β, where α, β are LSs of any type CAUSATIVE

Figure 4.

One of the questions which RRG asks is, when there is an option as to which arguments can be linked to pivot, what factors can affect the choice? It turns out that the answer to this question has important typological ramifications, for some languages permit discourse-pragmatic factors to play a role, whereas others do not. This distinction is



Role and Reference Grammar

expressed in the RRG typology of ‘role-dominated’ (no discourse-pragmatic influence on PSA selection) vs. ‘reference-dominated’ (possible discourse-pragmatic influence on PSA selection) languages, and in the contrast between semantic pivots vs. pragmatic pivots (which are found only in reference-dominated languages).

4.  Some implications of RRG RRG illustrates one possible answer to the questions stated at the beginning of Section 2, and it shows that it is possible to have a rigorous, typologically-sensitive grammatical theory which takes semantics and pragmatics as central features. RRG’s contribution to pragmatics is not in terms of developing any new pragmatic concepts but rather in terms of showing how pragmatics pervades grammar and is involved in many aspects of it. This is summarized in Figure 5. Influence on choice of template Syntactic Inventory

LSC Syntactic Template

Focus Structure

Discourse Model

Pragmatic Pivot [in some languages]

Referent-1: Activated Referent-2: Accessible Referent-3: Inactive etc.

Linking Algorithm

h al c xic e l n eo

[do′ (x, …)] CAUSE [BECOME pred′ (y, x)]

es oic

c

Lexicon

uen infl

Figure 5.

RRG has not only attempted to deal with the issues that have interested typologists and functionalists, e.g. universality of grammatical relations, but also with some of the leading questions raised by formal theories, e.g. extraction restrictions (subjacency). Recent work (Van Valin 1993a, 1993b) has proposed a functional account of subjacency which relies crucially on interactions among the linking system, focus structure, and syntactic structure. In addition, RRG has been shown to provide a potentially explanatory framework for the study of language acquisition and child language (Van Valin 1991, 1992, 1998; Van Valin & LaPolla 1997; Rispoli 1991a, 1991b; Weist 2002) and for language processing (Bornkessel, Schlesewsky & Van Valin 2004; Guest 2008; Van Valin 2006).

247

248 Robert D. Van Valin, Jr.

References Bornkessel, I., Schlesewsky, M. & Van Valin, R.D. Jr. (2004). Syntactic templates and linking mechanisms: A new approach to grammatical function asymmetries. Poster presented at 17th CUNY Conference on Human Sentence Processing. [available at linguistics.buffalo.edu/research/rrg.html] Dowty, D. (1979). Word meaning and Montague grammar. Kluwer. Fillmore, C. (1968). The case for case. In E. Bach & R. Harms (eds.), Universals in linguistic theory: 1–88. Holt, Rinehart & Winston. Foley, W. & R. Van Valin Jr. (1977). On the viability of the notion of ‘subject’ in universal grammar. Proceedings of the 3rd annual meeting of the Berkeley Linguistics Society: 293–320. ——— (1984). Functional syntax and universal grammar. Cambridge University Press. Gruber, J. (1965). Studies in lexical relations. Ph.D. Diss., MIT. Guest, E. (2008). Parsing for Role and Reference Grammar. In R. Van Valin (ed.), 435–454. Jackendoff, R. (1976). Toward an explanatory semantic representation. Linguistic Inquiry 7: 89–150. Johnson, M. (1987). A new approach to clause structure in role and reference grammar. Davis Working Papers in Linguistics 2: 55–59. Kailuweit, R., Staudinger, E., Wiemer, B., & Matasović, R. (eds.) (2008). New Applications of Role and Reference Grammar: Diachrony, Grammaticalization, Romance Languages. Cambridge Scholar Press. Lambrecht, K. (1986). Topic, focus and the grammar of spoken French. Ph.D. Diss., University of California. ——— (1987). Sentence focus, information structure, and the thetic-categorial distinction. Proceedings of the 13th annual meeting of the Berkeley Linguistics Society: 366–382. ——— (1994). Information structure and sentence form. Cambridge University Press. Olson, M. (1981). Barai clause junctures. Ph.D., Diss. Australian National University. Rispoli, M. (1991a). The acquisition of verb subcategorization in a functionalist framework. First Language 11: 41–63. Shimojo, M. (1995). Focus structure and morphosyntax. In Japanese: wa and ga, and word order flexibility, Ph.D. dissertation, SUNY at Buffalo [avaliable at linguistics.buffalo.edu/research/rrg.html] ——— (1991b). The mosaic acquisition of grammatical relations. Journal of Child Language 18: 517–52. Van Valin Jr., R. (1977a). Ergativity and the universality of subjects. Papers from the 13th regional meeting of the Chicago Linguistic Society: 689–706. ——— (1977b). Aspects of Lakhota syntax. Ph.D. Diss., University of California. ——— (1980). On the distribution of passive and antipassive constructions in universal grammar. Lingua 50: 303–27. ——— (1981). Grammatical relations in ergative languages. Studies in Language 5: 361–94. ——— (1991). Functionalist linguistic theory and language acquisition. First Language 11: 7–40. ——— (1992). Extraction restrictions, competing theories and the argument from the poverty of the stimulus. In The reality of linguistic rules. John Benjamins. ——— (1993a). A synopsis of role and reference grammar. In R. Van Valin (ed.), Advances in role and reference grammar: 1–164. John Benjamins. ——— (1993b). The interaction of pragmatics and syntax. Paper, Fifth Biennial Symposium of the Dept. of Linguistics, Rice University. Van Valin, R.D., Jr (1998). The acquistions of WH-questions and the mechanisms of language acquistion. In M. Tomasello (ed.), The new psychology of language: Cognitive and functional approaches to language structure, 221–249. Lawrence Erlbaum.



Role and Reference Grammar 249

——— (2005). Exploring the syntax-semantics interface. Cambridge University Press. ——— (2006). Semantic macroroles and language processing. In I. Bornkessel et al. (eds.), Semantic role universals and argument linking: Theoretical, typological and psycholinguistic perspectives, 263–302. Mouton de Gruyter. Van Valin, R.D., Jr. (ed.) (2008). Investigations of the syntax-semantics-pragmatics interface. John Benjamins. Van Valin, R.D., Jr. & R.J. LaPolla (1997). Syntax: structure, meaning & function. Cambridge University Press. Van Valin Jr., R. & W. Foley (1980). Role and reference grammar. In E. Moravcsik & J. Wirth (eds.), Syntax and semantics, vol. 13: 329–52. Academic Press. Van Valin Jr., R. & D. Wilkins (1993). Predicting syntactic structure from semantic representations. In Van Valin (ed.), Advances in role and reference grammar: 499–534. John Benjamins. Vendler, Z. (1967). Linguistics in philosophy. Cornell University Press. Wesit, R.M. (2002). The first language acquistion of tense and aspect: A review. In R. Salaberry & Y. Shirai (Eds.), Tense-aspect morphology in L2 acquistion, 21–78. John Benjamins. Wilkins, D. & R. Van Valin Jr. (1993). The case for a case reopened. SUNY Buffalo Center for Cognitive Science Technical Report 93(2).

Semantics vs. pragmatics Ken Turner University of Brighton

1.  Fregean beginnings In 1897, Gottlob Frege stated an ambition and made a number of observations about the obstacles to fulfilling that ambition. The passage is worth quoting at length: we have to make a distinction between the thoughts that are expressed and those which the speaker leads others to take as true although he does not express them. If a commander conceals his weakness from the enemy by making his troops keep changing their uniforms, he is not telling a lie; for he is not expressing any thoughts, although his actions are calculated to induce thoughts in others. And we find the same thing in the case of speech itself, as when one gives a special tone to the voice or chooses special words. If someone announces the news of a death in a sad tone of voice without actually being sad, the thought expressed is still true even if the sad tone is assumed in order to create a false impression. And we can substitute words like ‘ah’ and ‘unfortunately’ for such a tone of voice without altering the thought. Naturally things are different if certain actions are specifically agreed on as a means of communicating something. In language common usage takes the place of such agreements. Of course borderline cases can arise because language changes. Something that was not originally employed as a means of expressing a thought may eventually come to do this because it has constantly been used in cases of the same kind. A thought which to begin with was only suggested by an expression may come to be explicitly asserted by it. And in the period in between different interpretations will be possible. But the distinction itself is not obliterated by such fluctuations in language. In the present context the only essential thing for us is that a different thought does not correspond to every difference in the words used, and that we have a means of deciding what is and what is not part of the thought, even though, with language constantly developing, it may at times be difficult to apply. (Frege 1897: 140–1)

And so the distinction between what is now called, in a more modern idiom, semantics and pragmatics was introduced. 2.  From then until now The history of this distinction since Frege has not been particularly auspicious. On the one hand the legitimacy of the distinction has been denied by a vociferous philosophical



Semantics vs. pragmatics

community which takes its lead from remarks by the later Wittgenstein and which to this day proclaims that meaning is use (see Segerdahl 1996 for a recent installment of this story). Dummett has spoken of this Wittgensteinian position at some length and this discussion concludes with a wholesale condemnation: Naturally, so grotesquely false a methodology could not be consistently adhered to by intelligent people. (Dummett 1975: 444–5)

On the other hand, those who have acknowledged the legitimacy of the distinction have drawn it in many non-compatible and non-equivalent ways. The following list, from Lyons (1987), is not complete:1 Semantics has to do with (i) meaning, (ii) competence, (iii) the conventional, (iv) rules, (v) the truth-conditional, (vi) the literal, (vii) sentences and (viii) context-independence; pragmatics has to do with (i′) use, (ii′) performance, (iii′) the non-conventional, (iv′) tendencies, principles, maxims and/or strategies, (v′) the non-truth-conditional, (vi′) the non-literal, (vii′) utterances and (viii′) context-dependence. Lyons concludes, in a somewhat defeatist vein, that “current presentations of the distinction between semantics and pragmatics tend to be riddled with inconsistencies and unjustified assumptions” (Lyons 1987: 155).2 It is possible to be more optimistic, however. One should not draw his conclusion, with its implicature that the conceptual and empirical territory is too difficult to be subject to adequate theoretical survey, before examing the method of dualism on which the above lists rest. Consider three examples, none of which are discussed by Lyons. First, Kamp (1979) draws attention to a widespread but rarely articulated assumption about the general form that a comprehensive theory of language should take. He calls this assumption SSP and formulates it as follows: a. The concept, or concepts, characterized by the recursive component of the theory belong(s) to semantics; b. Moreover, no notion to which this component refers belongs to pragmatics. and he goes on to say that there are no a priori grounds to favour a theory which embraces SSP over one that violates it. This is an early statement against the strict separation of domains: it should not be read as an endorsement of the ‘meaning is use’ conflationist

1.  A more comprehensive list can be found in Nemo & Cadiot (1997a,b). They notice 35 different ways in which this distinction is discussed in the literature. 2.  Defeatism in the face of the complexities of navigating between meaning and use, whilst not endemic, is certainly a common enough attitude amongst those who have attempted to chart these waters: see the last two paragraphs of Allwood (1981) and Davis (1987) and Fant (1990).

251

252 Ken Turner

slogan but as an early advertisement that some notions that belong to pragmatics may intrude into and perhaps constitute those notions that belong to semantics. Second, Posner (1980), in what is essentially a methodological discussion, sets out a number of strategies for the description of communication. The first he calls the monism of meaning and the second the monism of use. These strategies occupy extreme positions on the spectrum of possible positions. Although they are different in every way these positions share that assumption that each can give a complete explanation of how language functions. More interesting than these extremes is the possibility of intermediate positions. The recognition of intermediate positions introduces the necessity of estimating how much of the content of a given utterance must be traced back to its literal meaning and how much is to be found in its context-dependence. From these procedural issues Posner concludes that there are two broad spectra of intermediate positions: one occupied by the meaning-maximalists who attempt to deduce as much as possible, but, crucially, not all, from the literal meanings of expressions; the other occupied by the meaning-minimalists who attempt to deduce as much as possible, but, crucially, not all, from the pragmatic principles controlling the uses of expressions. The issues between the maximalists and minimalists were once partially ventilated (Cohen 1971, 1977, 1986; Walker 1975). Whilst conclusions on these questions are sometimes more a matter of theoretical consistency, methodological strategy and, sometimes, individual temperament than of fact, it is, I believe, true to say that the minimalists have, in some sense, carried the day in that they have motivated the larger research community. Third, Dascal (1981) reaches the same conclusion but by different means. His extremes he calls radical literalism and radical contextualism and his intermediate positions he calls moderate literalism and moderate contextualism. He says, siding with his equivalent to meaning-minimalism: The obvious strategy to cope with the arguments of radical and moderate contextualists, in so far as they call attention for hitherto unrecognized forms of contextual dependency, is to adopt … ‘moderate literalism’. Instead of considering literal meanings as somewhat complete entities, identical to or capable to determine truthconditions and significance, such a position would view meanings as ‘schematic’, incomplete entities, specifying conditions, guiding principles and other means through which their ‘gaps’ might be appropriately filled by contextual information. The study of these abstract entities would properly belong to semantics, whereas the study of the ways in which contextual information would be embedded in them, after being duly selected, would belong to pragmatics. But, in view of the interaction between them … each one of them should be developped, from the outset, with the other ‘in mind’. (Dascal 1981: 175)

This kind of advice — encouraging the move away from the simple dichotomies mentioned earlier and a picture where semantics is subject to pragmatic addition and a move towards a picture which highlights the interaction of semantics and pragmatics — has



Semantics vs. pragmatics

been taken very seriously in recent years. These contemporary perspectives will be discussed in the next sections. The general message that is sent by the Kamp-Posner-Dascal and related manifestoes has been implemented by two very large, but only marginally overlapping, research communities. It is within one, or the other, of these communities that significant progress has been made in recent years and it is within either or both that considerable further advances can be anticipated in the future. 3.  Current manoeuvres: (Neo-(Post-))Gricean pragmatics In 1967, Paul Grice gave the William James Lectures at the University of Harvard. Some of these individual lectures eventually trickled into print (Grice 1968, 1969, 1975, 1978) but the continuous suite, no doubt as a consequence of the author’s recognition of the many imperfections, was only published more than 20 years after their original delivery (Grice 1989). It is well-known that what he called his ‘tottering steps’ (Grice 1989: 4) towards a systematic philosophical theory of language included the postulation of the Co-operative Principle and its associated conversational maxims. These maxims constituted a part of a systematic philosophical theory of language which was predicated upon the assumption that the primary purpose of conversation is the ‘maximally effective exchange of information’ (Grice 1989: 28). This (simplifying) assumption is not always acknowledged in subsequent discussions and some have mistaken Grice’s tottering steps towards a theory as confident strides within a theory (e.g., Sarangi & Slembrouck 1992; Attardo 1997 among others). Where this simplifying assumption has been recognised, attempts to generalise the maxims have usually followed. Thus, Martinich (1984) loosens the maxims so that they accommodate more speech act types than truth-bearing declaratives; Clyne (1994: 194–5) attempts to format the maxims so that they can accommodate intercultural communication; and Green (1996: 90–1) puts them in a form that makes them usable in contexts relevant to the artificially intelligent. Martinich, Clyne and Green (and several others) all respect the Kantian format within which Grice’s initial proposals were originally presented. It remains unclear whether their proposed more general maxims, inflected for different discourse contexts, escape or inherit the following criticism: Grice’s original framework is clearly at best incomplete and at worst inadequate beyond repair to the task of predicting sets of nonlogical inferences which are actually drawn from a given utterance in a given context. (Horn 1988: 130)

What is clear, however, is that criticisms such as Horn’s have stimulated a new industry of pragmatic theory production and that these new theories are much more attentive than were their predecessors to their responsibilities for effecting an adequate and elegant interface with semantic theories.

253

254 Ken Turner

3.1  Relevance Theory There is, first, Relevance Theory. Wilson & Sperber (1990: 45) give the following characterization of the Principle of Relevance: Every act of inferential communication creates a presumption of optimal relevance.

They add that to communicate is to imply that the utterance used is worth the addressee’s attention and that any utterance addressed to an audience automatically conveys a presumption of its own relevance. They call this fact the principle of relevance (Wilson & Sperber 1992: 68). They continue: The principle of relevance differs from every other principle, maxim, convention or presumption proposed in modern pragmatics in that it is not something that people have to know, let alone learn, in order to communicate effectively; it is not something that they obey or might disobey: it is an exceptionless generalization about human communicative behaviour. (Wilson & Sperber 1992: 68)

This sentence is, in fact, one of Wilson & Sperber’s favourites and it occurs, repeated word for word, in numerous of their writings (e.g. Sperber & Wilson 1986: 160; Wilson & Sperber 1988: 140). It is rather hard to assess the merits of claims made within Relevance Theory because there are several question marks over its internal consistency. Some of these questions have been raised by O’Neill (1988), Levinson (1989), Roberts (1991) and especially Bird (1994). Although Sperber & Wilson (1995) have attempted to answer these criticisms, doubts remain about the theory’s explanatory and predictive potential and indeed its empirical falsifiability. 3.2  The Least Effort Hypothesis There is, second, what might be called the Least Effort Hypothesis. Horn (1984: 13) gives the following characterization of what he calls the Q- and R-Principles: The Q Principle (Hearer-based): Make your contribution sufficient. Say as much as you can (given R). The R Principle (Speaker-based): Make your contribution necessary. Say no more than you must (given Q).

and he suggests that the whole Gricean mechanism for pragmatic inference can be derived from the dialectic interaction between the Q- and R-Principles in the following way: The Division of Pragmatic Labour: The use of a marked (relatively complex and/or prolix) expression when a corresponding unmarked (simpler, less ‘effortful’) alternative expression is available tends to be interpreted as conveying a marked message (one which the unmarked alternative would not or could not have conveyed).



Semantics vs. pragmatics

In other words, the R-Principle generally takes precedence until the use of a contrastive linguistic form induces a Q-Implicature to the non-applicability of the pertinent R-Implicature. Horn (1989: 194) is confident about the explanatory potential of these principles but it is rather hard to assess the merits of claims made with the assistance of the Least Effort Hypothesis because there is a question mark over its internal consistency. One question has been raised by Levinson (1987a: 73) who notices that Horn’s R-Principle is ambiguous between an interpretation that enjoins a speaker to prefer semantically general expressions to semantically specific ones and an alternative interpretation that enjoins a speaker to prefer ‘shorter’ expressions (with fewer units of speech production) to ‘longer’ ones. There are, therefore, grounds for remaining cautious about the explanatory and predictive adequacy of this hypothesis and indeed about its empirical falsifiability. 3.3  The Q-, I- and M-Principles Hypothesis There is, third, the position defended by Levinson (1987a). He outlines the following mechanism which rests upon the distinction between what each principle enjoins the speaker to do and what it licenses the hearer to think: The Q-Principle Speaker’s Maxim Do not provide a statement that is informationally weaker than your knowledge of the world allows, unless providing a stronger statement would contravene the I-Principle.

Recipient’s Corollary Take it that the speaker made the strongest statement consistent with what he or she knows.

The I-Principle Speaker’s Maxim (The Maxim of Minimization) Say as little as necessary. That is, produce the minimal linguistic information sufficient to achieve your communicational ends (bearing the Q-Principle in mind).

Recipient’s Corollary (The Maxim of Enrichment) Amplify the informational content of the speaker’s utterance, by finding the most specific interpretation, up to what you judge to be the speaker’s intended point. Specifically (a) assume that stereotypical relations obtain between referents or events, unless (i) that is inconsistent with what has been taken for granted, or (ii) the speaker has broken the Maxim of Minimization by choosing a prolix expression; (b) assume the existence or actuality of what a sentence is ‘about’ if that is consistent with what is taken for granted; (c) avoid interpretations that multiply entities referred to; specifically, prefer co-referential readings of reduced noun phrases.

255

256 Ken Turner

The M-Principle Speaker’s Maxim

Recipient’s Corollary

Do not use a prolix, obscure or marked expression without reason.

If the speaker used a prolix or marked expression, he or she did not mean the same as he or she would have had he or she used the unmarked expression — specifically, he or she was trying to avoid the stereotypical associations and I-implicatures of the unmarked expression.

The Q-Principle is of restricted scope. It operates only in clearly defined contrast sets, of which Horn Scales are prototypical, and it enriches utterance meaning by inducing the negation of a stronger possible proposition. (1)–(3) illustrate the Q-Principle:

(1) (2) (3)

It is warm. I believe Sam is away. I saw an animal there.

Q-implicates Q-implicates Q-implicates

It is not hot. I don’t know Sam is away. I don’t know what kind of animal exactly it was.

The I-Principle enables many different kinds of informationally ampliative conversational inferences. It induces a stronger interpretation from the use of a less specific, semantically general expression. (4)–(6) illustrate the I-Principle:

(4)



(5)



(6)

He turned the key and the engine started. Harry and Sue bought a piano. John came in and he sat down.

I-implicates I-implicates I-implicates

He turned the key and then the engine started. Harry and Sue bought a piano together, not one each. Johni came in and hei sat down.

The M-Principle assumes prototypicality and signals that departures from unmarked expressions implicate that something other than I-implicatures should be drawn. (7)–(10) illustrate the M-Principle and contrast it with the normal running of the I-Principle:

(7) (7′)



(8) (8′)

(9) (9′) (10) (10′)

John could solve the problem… John had the ability to solve the problem… Peter stopped the bus… Peter caused the bus to stop…

I-implicates M-implicates

…and he did. …but he didn’t.

I-implicates M-implicates

Sam moved the car… Sam caused the car to move… He’d like a drink… He’d like a beverage…

I-implicates M-implicates I-implicates M-implicates

…in the normal way. …in a non-normal way, by, for example, shooting the radiator out. …by driving it. …by pushing it. …an alcoholic drink. …a non-alcoholic drink.



Semantics vs. pragmatics

There is much more to say about the nature and operation of these principles and their interaction with an assumed truth-conditional or model-theoretic semantics. Levinson has, in fact, been developing a sophisticated neo-Gricean position for several years (Atlas & Levinson 1981; Levinson 1987b, 1991, 1995) and his book on generalized conversational implicature (Levinson 2000) makes available the full account. 3.4  Pragmatic intrusion These pragmatic theories have placed the notion of pragmatic intrusion high on the theoretical agenda. Carston (1988) is a early employer of this notion. She considers sentences such as (11)–(17):

(11) (12) (13) (14) (15) (16) (17)

He ran to the edge of the cliff and jumped. I went to the exhibition and ran into John. She took the gun, walked into the garden and killed her mother. I had a holiday in Austria and did some cross-country skiing. Mary was in the kitchen and she was listening to the radio. He fell into a deep sleep and dreamed he was flying. We investigated the problem and it was far more complex than expected.

and argues that they have interpretations which are pragmatically derived enrichments and completions of the linguistically given logical forms. This argument enables her to draw the following conclusion: It seems … that we must distinguish two kinds of semantics, linguistic and truth conditional, the former naturally figuring in a theory of utterance meaning, the latter taking as its domain propositional forms, whether of utterances or unspoken thoughts. Linguistic semantics IS autonomous with respect to pragmatics; it provides the input to pragmatic processes and the two together make propositional forms which are the input to a truth-conditional semantics. Once this distinction is made, the compulsion to treat all pragmatically derived meaning as implicature subsides; there is no reason why pragmatics cannot contribute to the explicature, the truth-conditional content of the utterance. (Carston 1988: 176)

Bach (1994a,b) is also sensitive to the possibility of pragmatic intrusion and he comes to an almost identical conclusion. He considers sentences such (18)–(22):

(18) (19) (20) (21) (22)

You are not going to die. I haven’t eaten. Steel isn’t strong enough. Willie almost robbed a bank. Everybody went to Paris.

and argues that “pragmatic processes of the same sorts as those involved in conversational implicature … come into play prior to the working out of implicatures” (Bach1994b: 269).

257

258 Ken Turner

This discussion is still in its infancy. The new picture is one in which there are pragmatic processes, variously called completion, expansion and enrichment, that develop linguistic meanings into propositional forms but all the important terminology in this picture remains formally undefined. Carston (1988: 169), in fact, entertains the eventuality that “It might seem that we need to define the notion development of a logical form more closely” but once entertained, she dismisses this definition as “unnecessary”. One question of importance that does emerge, however, is that, if Carston and Bach are right and there is what might be called a pre-semantic pragmatics and a postsemantic pragmatics then it is necessary to be able to distinguish the contributions of these two kinds of pragmatics. One needs to be able to distinguish the information provided by processes of completion, expansion and enrichment, however these processes become more adequately defined, from the information provided by some version of the (Neo-(Post-))Gricean maxims. Carston suggests that the Minimalist Principle can be used as an instrument to distinguish the two: Minimalist Principle: A pragmatically determined aspect of meaning is part of what is said if and only if its determination is necessary for the utterance to express a complete proposition.

This principle entails what she calls the Linguistic Direction Principle: Linguistic Direction Principle: To every pragmatically determined aspect of meaning that is part of what is said, there corresponds a slot in the meaning of the sentence which must be filled for the utterance to be truth-evaluable.

She also suggests a criterion of functional independence of explicatures which “ensures that what counts as an explicature isn’t arbitrarily confined to linguistic sense plus reference assignment and disambiguation” (Carston 1988: 136). But Récanati (1989, 1993) provides arguments against Carston’s diagnostics for distinguishing conversational implicatures and the pragmatic constituents of semantic representation and logical form. He suggests the Availability Principle as an instrument to distinguish the two: Availability Principle: In deciding whether a pragmatically determined aspect of utterance meaning is part of what is said, that is, in making a decision concerning what is said, we should always try to preserve our pre-theoretic intuitions on the matter. (Récanati 1993: 248)

It is not possible to endorse this principle without much more further discussion and without a full demonstration that rebuts the following plausible claim: “We do not come equipped with a meter that reliably distinguishes between semantic and pragmatic implications” (Richard 1990: 123).

Semantics vs. pragmatics 259



This matter has been the subject of experimental test. Gibbs & Moise (1997) argue, on the basis of experimental results, that speakers and hearers can pretheoretically distinguish between the information that makes up the pragmatic constitituents of what is said and the information that constitutes the conversationally implicated and that these results force a change from the Gricean view of speaker meaning, illustrated in Figure 1, to a revised view of speaker meaning, illustrated in Figure 2. Speaker Meaning What is Said

What is Implicated

Generalized Implicatures

Particularized Implicatures

Figure 1.  The Gricean view of speaker meaning Speaker Meaning What is Said Minimal

What is Implicated

Enriched

Figure 2.  The revised view of speaker meaning

These results, and the revisions that are based on them, have, however, been disputed by Nicolle & Clark (1999). They ran a series of experiments similar to the Gibbs & Moise experiments but found that there was no evidence that speakers and hearers had reliable intuitions about pragmatically determined aspects of semantic content. The Availability Principle is, therefore, not only intuitively implausible but perhaps also empirically disconfirmed.

4.  Current manoeuvres: (Neo-(Post-))Kaplanean semantics In 1977, David Kaplan read a paper at a Symposium on Demonstratives at the March meetings of the Pacific Division of the American Philosophical Association. Some of this paper eventually trickled into print (Kaplan 1978a,b) but the whole paper, no doubt as a consequence of the author’s recognition of the many imperfections, was only published more than 10 years after its original delivery (Kaplan 1989a).

260 Ken Turner

The position that Kaplan attempted to defend was one that noticed the ambivalence in the Fregean notion of Sinn and which sought to remove that ambivalence. The ambivalence is widely acknowledged (see Platts 1997: 114–5, for a perspicuous discussion). Kaplan’s remedy is to introduce a distinction between ‘content’ and ‘character’: The content of a sentence in a given context is what has traditionally been called a proposition. (Kaplan 1989a: 500) The character of an expression is set by linguistic conventions and, in turn, determines the content of the expression in every context. Because character is what is set by linguistic conventions it is natural to think of it as meaning in the sense of what is known by the competent language user. (Kaplan 1989a: 505)

The consequence of this distinction is that semantics now becomes two-tiered and the notion of Sinn that Frege employed to solve the informativeness of identity statements problem is moved from content to character. The resulting theory is called the Theory of Direct Reference but Kaplan anticipates and heads off a possible mis-understanding: The ‘direct’ of ‘direct reference’ means unmediated by any propositional component, not unmediated simpliciter. The directly referential term goes directly to its referent, directly in the sense that it does not first pass through the proposition. Whatever rules, procedures, or mechanisms there are that govern the search for the referent, they are irrelevant to the propositional component, to content. When the individual is determined … it is loaded into the proposition. It is this that makes the referent prior to the propositional component, and it is this that reverses the arrow from propositional component to individual in the Direct Reference Picture. (Kaplan 1989b: 569)

The architecture of this theory will call to mind certain mechanisms that are employed by the (Neo-(Post-))Griceans. Kaplan himself is noncommital about the division of theoretical labour: A[n] … interesting question is whether to call the theory of these mechanisms semantics or pragmatics. The central role of the notion context of use in determining content might incline one to say that the theory of character is semantics, and the theory of content is pragmatics. But truth is a property of contents, and one wouldn’t want to be caught advocating a pragmatic theory of truth. The problem is that on my analysis, the mechanisms of direct reference operate before the familiar semantical notions of truth and denotation come into play. If I continue to think, as Carnap taught me, that the overall theory of a language should be constructed with syntax at the base, semantics built upon that, and pragmatics built upon semantics, I am faced with a dilemma. The mechanisms of direct reference certainly are not postsemantical. But equally surely they are not syntactical. Thus I put them in the bottom layer of semantics. … Whether semantics or



Semantics vs. pragmatics

pragmatics, it is important to emphasize that there are two roads from singular terms to individuals. The road through what is said, through the propositional component, through content. And the direct road, outside of what is said, outside content. Both roads belong to the rules of the language, and not to the vagaries of individual difference among language users. Both connect language to the world. (Kaplan 1989b: 575–576)

The two-tiered architecture of this theory is also apparent in Perry (1993) — for whom the notion of character is called ‘role’ and that of content is called ‘value’ — and in Récanati (1993). Perry is a little less noncommittal than Kaplan about the need to theorize the role — in the non-technical sense — of pragmatics: he says, apropos an analysis of belief reports, that “the move to unarticulated constituents emphasizes the importance of pragmatic facts about language to the study of what seem like purely semantic issues. In order to express claims, we exploit a tremendous variety of facts, conventions, and circumstances, of which the meanings and referents of our terms form just a part. So it is a mistake to relegate pragmatics to matters of felicity and implicature. In the case of belief reports, it is central to understanding content and truth” (Perry 1993: 278). But he does not go further and theorize the role of pragmatics himself. There is no Logic of the Essential Indexical as there is a Logic of Demonstratives. Récanati, on the other hand, has proposed and developed what is perhaps the most elaborate two-tiered theory of the interface. He distinguishes between (i) linguistic meaning (a.k.a. character (Kaplan) and role (Perry)), (ii) what is said (a.k.a. content (Kaplan) and value (Perry)) and (iii) what is implicated and he introduces two primary pragmatic processes to bridge the gap between (i) and (ii). The first process is strengthening: it takes a proposition as input and yields as output another proposition which entails the original. Thus (22) is strengthened into (23) and (24) into (25):

(22) (23) (24) (25)

It’ll take us some time to get there. It’ll take us a relatively long time to get there. I have had breakfast. I have had breakfast this morning.

The second primary pragmatic process is expansion: it takes a proposition as input and yields as output another proposition which does not necessary entail the original. Thus (26) is expanded into (27) and (28) into (29):

(26) (27) (28) (29)

I have nothing to wear. I have nothing suitable to wear. The door is open. The door in this room is open.

Once the primary pragmatic process have converted (i) into (ii), then the familiar Gricean mechanisms kick in in the usual way.

261

262 Ken Turner

This discussion is a significant advance over the Bach-Carston account mentioned in Section 3.4 above. The notion of strengthening is here well-defined: it is disciplined by entailments. Récanati acknowledges, however, that there is still much work left to do on the less disciplined process of expansion (cf. Récanati 1994). The Kaplan-Perry-Récanati picture is, to repeat, a picture of direct reference. This is a picture which, to underscore, acknowledges that there are Fregean modes of presentation but it denies that these modes of presentation are propositional constituents. The strategy involves putting these modes in another, deeper, level of semantics and postulating processes to convert this deeper level into the level where the correct truth values are assigned. Recently, the question has been explicitly asked of the compatibility of this family of theories with a Gricean pragmatics. The interface between syntax and semantics, it will be recalled, is disciplined by the rule-to-rule hypothesis and something like the principle of compositionality but the interface between semantics and pragmatics has not, until now, been disciplined by similar mechanisms. Recently, Lumsden (1996) has proposed that this interface be disciplined by the Principle of Helpfulness: Principle of Helpfulness: We expect the semantics to be capable of being helpful to the hearer in working out the speaker’s communicative intentions.

and he has attempted to demonstrate that at least in the case of proper names directly referential semantics is fully compatible with a (broadly) Gricean pragmatics. In its present form, though, the principle is rather vague. Further research will confirm or disconfirm its more general application. Two further points deserve to be mentioned. First, just as there were reservations about Frege’s apparent conflation of sense and meaning, so have there recently emerged similar reservations about Kaplan’s conflation of character and meaning. Braun (1995, 1996) notices that the informal characterization of character that Kaplan gives does not correspond to the formal definition he provides in the Logic of Demonstratives. This observation enables him to argue that, whichever definition is taken as the appropriate device, there is no extensional function on contexts that does the job of capturing the relevant linguistic meaning. Braun concludes by postulating a three-tiered theory, underwritten by a formal logic, consisting of (i) linguistic meaning, (ii) character and (iii) content.3 Second, Salmon (1986) also takes modes of presentation out of the proposition but instead of embedding them deeper, Kaplan-Perry-Récanati-Braun style, into another level of semantics he takes them out of the semantics altogether, calls them ‘guises’ and places them within the pragmatics of the expressions involved. He

3.  I mention, but do not discuss, a recent attempt to put suitably revised modes of presentation back into the proposition (Bezuidenhout 1993, 1996a; cf. Bezuidenhout 1996b; Grimberg 1996).



Semantics vs. pragmatics

acknowledges that much in his account remains philosophically unsatisfactory4 but it is possible that the kind of problems that originally provoked Frege to distinguish between sense and reference find an elegant and economical solution by shifting the burden of explanation from the semantics to an interaction of the semantics and pragmatics.

5.  Postscript: The logical basis of the semantics-pragmatics interface (Neo-(Post-))Gricean positions assume a first-order logic as the basis of their semantics and their principal task is to account for the alleged differences in meaning and use between that logic’s connectives and quantifiers and their counterparts in a natural language like English. (Neo-(Post-))Kaplanean positions, on the other hand, seek to defend quantified modal logic as the basis of their semantics and their principal task is to account for the alleged identity of an object as it travels through a suite of identity transforming possible worlds. The two families of positions have certain elements of contact and compatibility and we have seen in previous sections the development of a pragmatics that was designed principally on the basis of the first but has more recently been brought into service to interface with the second. Future research will reveal more adequately and accurately the character of the contact and compatibility of the systems discussed in this entry.5 6.  Conclusion This entry began with Gottlob Frege in 1897. In the ninety years until 1987, when Lyons made his pessimistic remark, it seemed that the amount of work put into the investigation of (the interaction between) these two domains was inversely proportional 4.  The account is loose enough for some to ask whether ‘guises’ are not really a terminological variant of Fregean ‘senses’. See Branquinho (1990). 5.  Atlas & Levinson (1981) believe that the story has much further to go: “There has been a regrettable temptation to adopt a logical primitivism when theorizing about implicature. The canonical languages of our logical theories are constructed to achieve pellucidity, but a certain measure of complexity is compatible with … a satisfactory use of truth-conditional semantics within a pragmatic theory. … Classical semantics is inadequately explanatory, in either its extensional or intensional varieties.” (p. 50 and p. 57). For an argument to the effect that it is neither first-order quantificational logic with identity nor quantified modal logic but rather intuitionistic logic that provides the most adequate basis for the semantics of natural language see Dummett (1991).

263

264 Ken Turner

to the clarity achieved. The last few years, however, have been very different and there have emerged two extremely creative intellectual communities whose research has begun the production of numerous new questions for investigation.6 These questions include, but are not exhausted by, the following: 1. How many different kinds of semantics does the evidence make it necessary to recognize? 2. What is the character of the theory for each of these different kinds of semantics? 3. How many different kinds of pragmatics does the evidence make it necessary to recognize? 4. What is the character of the theory for each of these different kinds of pragmatics? 5. How do the theories of these (different kinds of) semantics interface with each other? 6. How do the theories of these (different kinds of) pragmatics interface with each other? 7. How do the theories of these (different kinds of) semantics interface with the theories of these (different kinds of) pragmatics? 8. What is the cross-linguistic applicability of the answers to 1–7 above? 9. What is the psychological validity of the answers to 1–7 above? 10. What are the implications for theories of language change of the answers to 1–7 above? The answers to these questions ought to scotch any simple-minded belief, induced by the method of dualism, that a neat bold line can be drawn between the semantics and pragmatics of linguistic forms and more sophisticated theories of the interface can be anticipated in the future. The answers to these questions ought also to prevent the disagreeable hostilities that one sometimes sees in certain parts of the literature (see Davis 1987). The answers to 1–10, and to other questions besides, will be made available in a peace mission, with a mandate to encourage the wholesale reinvestigation of the semantics-pragmatics interface(s) in the spirit of enlightened and benign cooperation, that is currently under way (see Turner 1999; von Heusinger & Turner 2006). The anticipated trajectory of these discussions over the next ten years leave one optimistic that Lyons-Allwood-Davis type remarks will not be repeated in the years to come.

6.  There is, of course, research conducted outside the mainstream of these two communities: in recent formal linguistics see, for example Krifka (1995) and de Swart (1996) and in recent philosophy see, for example, Hintikka & Halonen (1995) and Bezuidenhout (1997). Jaszczolt & Turner (1996, 2003) attempts to get a contrastive angle on the issues.



Semantics vs. pragmatics 265

References Allwood, J. (1981). On the distinctions between semantics and pragmatics. In W. Klein & W. Levelt (eds.), Crossing the Boundaries in Linguistics: 177–189. Reidel Publishing Company. Atlas, J.D. & S.C. Levinson (1981). It-clefts, informativeness, and logical form: Radical pragmatics (revised standard version). In P. Cole (ed.), Radical Pragmatics: 1–61. Academic Press. Attardo, S. (1997). Competition and cooperation: Beyond Gricean pragmatics. Pragmatics & Cognition 5: 21–50. Bach, K. (1994a). Conversational impliciture. Mind and Language 9: 124–162. ——— (1994b). Semantic slack: What is said and more. In S.L. Tsohatzidis (ed.): 267–291. Bezuidenhout, A. (1993). Demonstrative modes of presentation. Communication and Cognition 26: 17–36. ——— (1996a). Pragmatics and singular reference. Mind and Language 11: 133–159. ——— (1996b). The truth-conditional relevance of de re modes of presentation: A reply to Grimberg. Mind and Language 11: 427–432. ——— (1997). Pragmatics, semantic underdetermination and the referential/attributive distinction. Mind 106: 375–409. Bird, G.H. (1994). Relevance theory and speech acts. In S. Tsohatzidis (ed.): 292–311. Branquinho, J. (1990). Are Salmon’s ‘guises’ disguised Fregean senses? Analysis 50: 19–24. Braun, D. (1995). What is character? Journal of Philosophical Logic 24: 227–240. ——— (1996). Demonstratives and their linguistic meanings. Noûs 30: 145–173. Carston, R. (1988). Implicature, explicature, and truth-theoretic semantics. In R. Kempson (ed.): 155–181. Clyne, M. (1994). Inter-cultural Communication at Work: Cultural Values in Discourse. Cambridge University Press. Cohen, J.L. (1971). Some remarks on Grice’s views about the logical particles of natural language. In Y. Bar-Hillel (ed.), Pragmatics of Natural Languages: 50–68. Reidel Publishing Company. ——— (1977). Can the conversationalist hypothesis be defended? Philosophical Studies 31: 81–90. ——— (1986). How is conceptual innovation possible? Erkenntnis 25: 221–238. Cole, P. (ed.) (1978). Syntax and Semantics 9: Pragmatics. Academic Press. Dascal, M. (1981). Contextualism. In H. Parret, M. Sbisà & J. Verschueren (eds.), Possibilities and Limitations of Pragmatics: 153–177. John Benjamins. Davis, S. (1987). The distinction between pragmatics and semantics. In J. Verschueren & M. Bertuccelli-Papi (eds.): 685–693. John Benjamins. De Swart, B.V.H. (1996). Meaning and use of not … until. Journal of Semantics 13: 221–263. Dummett, M. (1975). Can analytical philosophy be systematic, and ought it to be? [Republished in his 1978 Truth and Other Enigmas. London] ——— (1991). The Logical Basis of Metaphysics. Duckworth. Fant, L. (1990). On the relevance of the semantics/pragmatics distinction. In L. Lundquist & L. Schack Rasmussen (eds.), Pragmatics and its Manifestations in Language: 16–40. Handelshøjskolens Forlag. Frege, G. (1897). Logic. In H. Hermes et al. (eds.) (1979). Gottlob Frege: Posthumous Writings: 126–151. Blackwell. Gibbs, R. & J. Moise (1997). Pragmatics in understanding what is said. Cognition 62: 51–74. Green, G.M. (1996). Pragmatics and Natural Language Understanding. Lawrence Erlbaum Associates. Grice, P. (1968). Utterer’s meaning, sentence meaning, and word meaning. Foundations of Language 4: 225–42.

266 Ken Turner ——— (1969). Utterer’s meaning and intentions. Philosophical Review 78: 147–77. ——— (1975). Logic and conversation. In P. Cole & J. Morgan (eds.), Syntax and Semantics 3: Speech Acts: 41–58. Academic Press. ——— (1978). Further notes on logic and conversation. In P. Cole (ed.): 113–27. Academic Press. ——— (1989). Studies in the Way of Words. Harvard University Press. Grimberg, M.L. (1996). Pragmatically determined aspects of what is said: A reply to Bezuidenhout. Mind and Language 11: 415–426. von Heusinger, K. & K. Turner (eds.) (2006). Where Semantics Meets Pragmatics. Elsevier. Hintikka, K. & I. Halonen (1995). Semantics and pragmatics for why-questions. Journal of Philosophy 92: 636–657. Horn, L.R. (1984). Toward a new taxonomy for pragmatic inference: Q-based and R-based implicature. In D. Schiffrin (ed.), Meaning, Form, and Use in Context: Linguistic Applications: 11–42. Georgetown University Press. ——— (1988). Pragmatic theory. In F.J. Newmeyer (ed.), Linguistics: The Cambridge Survey: 113–145. Cambridge University Press. ——— (1989). A Natural History of Negation. University of Chicago Press. Jaszczolt, K. & K. Turner (eds.) (1996). Contrastive Semantics and Pragmatics. Pergamon Press. Jaszczolt, K. & K. Turner (eds.) (2003). Meaning Through Language Contrast. Benjamins. Kamp, H. (1979). Semantics versus pragmatics. In F. Guenthner & S.J. Schmidt (eds.), Formal Semantics and Pragmatics for Natural Languages: 255–287. Reidel Publishing Company. Kaplan, D. (1978a). DTHAT. In P. Cole (ed.): 221–243. ——— (1978b). On the logic of demonstratives. Journal of Philosophical Logic 8: 81–98. ——— (1989a). Demonstratives: An essay on the semantics, logic, metaphysics, and epistemology of demonstratives and other indexicals. In J. Almog, J. Perry & H.K. Wettstein (eds.), Themes from Kaplan: 481–563. Oxford University Press. ——— (1989b). Afterthoughts. In J. Almog, J. Perry & H.K. Wettstein, (eds.), Themes from Kaplan: 565–614. Oxford University Press. Kempson, R.M. (ed.) (1988). Mental Representations: The Interface between Language and Reality. Cambridge University Press. Krifka, M. (1995). The semantics and pragmatics of polarity items. Linguistic Analysis 25: 209–257. Levinson, S.C. (1987a). Minimization and conversational inference. In J. Verschueren & M. Bertuccelli-Papi (eds.): 61–129. ——— (1987b). Pragmatics and the grammar of anaphora: A partial pragmatic reduction of binding and control phenomena. Journal of Linguistics 23: 379–434. ——— (1989). A review of Relevance. Journal of Linguistics 25: 455–472. ——— (1991). Pragmatic reduction of the binding conditions revisited. Journal of Linguistics 27: 107–161. ——— (1995). Three levels of meaning. In F.R. Palmer (ed.), Grammar and Meaning: 90–115. Cambridge University Press. Levinson, S.C. (2000). Presumptive Meanings: The Theory of Generalized Conversational Implicature. MIT Press. Lumsden, D. (1996). How well does direct reference semantics fit with pragmatics? Philosophical Papers 25: 139–148. Lyons, J. (1987) Semantics. In J. Lyons et al. (eds.), New Horizons in Linguistics 2: 152–178. Penguin. Martinich, A.P. (1984). Communication and Reference. Walter de Gruyter. Nemo, F. & P. Cadiot (1997a). Un problème insoluble? Revue de Sémantique et Pragmatique 1: 15–22.



Semantics vs. pragmatics

——— (1997b). Un problème insoluble? (2e partie). Revue de Sémantique et Pragmatique 2: 9–40. Nicolle, S. & B. Clark (1999). Experimental pragmatics and what is said: A response to Gibbs and Moise. Cognition 66: 337–354. O’Neill, J. (1988). Relevance and pragmatic inference. Theoretical Linguistics 15: 241–261. Perry, J. (1993). The Problem of the Essential Indexical and Other Essays. Oxford University Press. Platts, M.B. (1997). Ways of Meaning: An Introduction to a Philosophy of Language. MIT Press. Posner, R. (1980). Semantics and pragmatics of sentence connectives in natural language. In J.R. Searle, F. Kiefer & M. Bierwisch (eds.), Speech Act Theory and Pragmatics: 169–203. Reidel. Récanati, F. (1989). The pragmatics of what is said. Mind and Language 4: 295–329. ——— (1993). Direct Reference: From Language to Thought. Blackwell. ——— (1994). Contextualism and anti-contextualism in the philosophy of language. In S. Tsohatzidis (ed.): 156–166. Richard, M. (1990) Propositional Attitudes: An Essay on Thoughts and How We Ascribe Them. Cambridge University Press. Roberts, L.D. (1991). Relevance as an explanation of communication. Linguistics and Philosophy: 453–472. Salmon, N. (1986). Frege’s Puzzle. MIT Press. Sarangi, S.K. & S. Slembrouck (1992). Non-cooperation in communication: A reassessment of Gricean pragmatics. Journal of Pragmatics 17: 117–154. Segerdahl, P. (1996). Language Use: A Philosophical Investigation into the Basic Notions of Pragmatics. Macmillan Press Ltd. Sperber, D. & D. Wilson (1986). Loose talk. Proceedings of the Aristotelian Society LXXXVI: 153–171. ——— (1995). Relevance: Communication and Cognition (2nd edition). Blackwell. Tsohatzidis, S.L. (ed.) (1994). Foundations of Speech Act Theory: Philosophical and Linguistic Perspectives. Routledge. Turner, K. (ed.) (1999). The Semantics-Pragmatics Interface from Different Points of View. Elsevier Science. Verschueren, J. & M. Bertuccelli-Papi (eds.), The Pragmatic Perspective. John Benjamins. Walker, R.C.S. (1975). Conversational implicatures. In S. Blackburn (ed.), Meaning, Reference and Necessity: 133–181. Cambridge University Press. Wilson, D. & D. Sperber (1988). Representation and relevance. In R. Kempson (ed.): 133–153. ——— (1990). Outline of Relevance theory. Hermes 5: 35–56. ——— (1992). On verbal irony. Lingua 87: 53–76.

267

Tense and aspect Robert I. Binnick

1.  The semantics of markers of tense and/or aspect 1.1  Tense All languages contain grammatical markers of tense and/or aspect (MTAs). Linguists who analyze tense and aspect from a functional point of view may define the meanings of MTAs in cognitive terms (Janssen 1996) or in terms of their discourse functions (Benveniste 1959; Weinrich 1964; Waugh 1975, 1991; Hopper 1979, 1982), and do not necessarily view their use and interpretation as dependent in the first instance on their semantics, as do the majority of linguists studying tense and aspect, who adopt a formal point of view from which tense and aspect are seen as basically referential categories. Consequently, the assumptions of this formal, referential approach to tense and aspect underlie most of the following discussion, though in certain areas (Section 2) functional analyses have played a central role. In a referential theory, tense is taken to be deictic in the sense that a tense denotes the temporal relation of the time E of an eventuality — event or state — expressed in an utterance to the time S of the speech act of producing that utterance. However, only deictic tenses such as the present and past relate E directly to the speech act time S. Anaphoric tenses such as the pluperfect and the conditional relate E directly to an interval of time called the reference point R, and only indirectly to S. In perfect tenses, such as the pluperfect, E precedes R, while in prospective tenses, e.g. the conditional, E follows R. 1.2  Aspect Aspect, also called grammatical or viewpoint aspect, intuitively concerns how an eventuality is viewed, whether as a completed whole (1a) or in progress (1b). (1) a. Sue had lunch with Bill. b. Sue was having lunch with Bill.

Aspect is explicitly marked by MTAs, which may combine aspect with tense, as in the case of the French simple past (2a) and imperfect (2b) tenses, both past in tense but contrasting in aspect, respectively perfective and imperfective.



Tense and aspect 269

(2) a. Jeanne lut le journal. ‘Jeanne read the newspaper.’ b. (Quand je suis entré,) Jeanne lisait le journal. ‘(When I entered,) Jeanne {was reading, used to read} the newspaper.’

For scholars who view the basic function of MTAs as referential, aspect denotes the relation between E and its temporal frame of reference, which many scholars identify with the reference point R. Thus the perfective aspect of the French simple past tense (2a) indicates that the eventuality is complete within its frame, that is, E is a subinterval of R, while the imperfective aspect of the imperfect tense (2b) indicates that it is not. The progressive aspect (1b) is a subtype of the imperfective in which R is a subinterval of E. Many scholars also admit perfect and prospective aspects, in which E respectively precedes and follows R. If MTAs mark both tense and aspect, it is possible to define the deictic tenses as those whose MTAs incorporate perfective aspect. Anaphoric tenses are then either those whose MTAs are imperfective or in which E is disjoint from R, and thus either perfect or prospective in aspect. 1.3  Aktionsart Aktionsart ‘kind of action’, also called lexical or situation aspect, is not a grammatical category. It concerns the temporal structures of types of situations. The Aktionsarten differ in three features, dynamicity, telicity, and durativity. States such as being ill or having left home are non-dynamic and uniform, while processes such as building a house may be dynamic and have different phases. Activities such as running or falling are non-telic processes; they do not contain an inherent point of culmination, and can be said to stop, but not to finish. Events such as running a mile or falling down are telic processes with inherent culminations, and are either accomplishments like running a mile which have duration or achievements like spotting a coin on the pavement which lack it. The period before an eventuality begins is its preliminary phase and that after it ends, its result state. The stage from the beginning of a process to the point at which it either culminates or merely terminates is its preparatory phase. An activity expression such as run refers to a preparatory phase, an achievement expression like notice a coin on the pavement to a culmination point, while an accomplishment expression such as build a house refers to both. An alternative to the temporal definition of aspect is one in terms of phasic structure. The perfective aspect includes both preparatory phase and culmination (if any) of an eventuality, while the imperfective and progressive include only the preparatory phase. The perfect and prospective aspects represent the result state and preliminary phase respectively.

270 Robert I. Binnick

Aspect and Aktionsart interact, and aspect may trigger ‘coercion’ of an Aktionsart into a different type, as when the progressive (3) and perfect (4) convert processes into states (Steedman 1997).

(3) Mr. Blandings is building his dream house. (4) Mr. Blandings has built his dream house.

Coercion receives a natural explanation within the phasic structure theory of aspect; for example, the progressive includes only the preparatory phase and so the eventuality is represented as similar to a state in lacking both an inherent bound and an internal phasic structure. 1.4  Underspecification and the pragmatics of tense and aspect Natural language expressions underdetermine the properties and relations they predicate. They do not encode propositions, but aid or direct the hearer in the process of recognizing the speaker’s communicative intention (Carston 1998: 18) and serve to filter out impossible readings (Wilson & Sperber 1998: 25). It is not always desirable to make temporal information explicit (5a is at least as coherent a text as 5b), though sometimes, as in the case of (6a), ambiguity or vagueness — the events in the second clause of (6b) could precede or follow that in the first clause — is avoided through explicit language (Lascarides & Oberlander 1992: 168).

(5) a. Max entered the office. John greeted him with a smile. He showed Max to the seat in front of his desk and offered him a cup of coffee. b. Max entered the office. Then, John greeted him with a smile. After that, he showed Max to the seat in front of his desk. He then offered Max a cup of coffee. (6) a. Jon switched off the heating. Then, Judy came in and said the room was too hot. b. Jon switched off the heating. Judy came in and said the room was too hot.

To fully interpret a spoken discourse or written text, a considerable amount of content must be filled in (Carston 1998). Pragmatic factors, including cotext (the discourse preceding, or the text surrounding, the material to be interpreted) and context, play a crucial role in enriching the bare-bones content provided by the semantics. (In functional theories of tense and aspect, pragmatics may in fact be seen to underlie the semantics.) The reference point R is pragmatically determined. The past tense (7) is normally understood as definite, referring to a certain point in time (Partee 1973: 602).

(7) I didn’t turn off the stove.

But definiteness is not part of the meaning of the past tense and must be inferred, and pragmatic factors determine to which time interval the marker refers.



Tense and aspect

Neither the relative sequence of events (the default interpretation of a sequence of past tenses is iconic, the order of the eventualities described being that of the clauses, as in 8), nor the length of the interval between successive eventualities (e.g., whether immediate as in 9 or non-immediate as in 10), nor any causal relationship holding between clauses (9, 10) is normally linguistically encoded; all must be inferred. (8) John entered the office of the president. The president greeted him. (9) John dropped the glass. It broke. (10) They planted an acorn. It grew.

Wilson & Sperber (1998: 11f.) propose that the source of such interpretations is the interaction of contextual information with the principle of optimal relevance (Sperber & Wilson 1986). For example, an indefinite reading for (11a) is precluded by the knowledge that the speaker must have had breakfast at some time in their life; a reading consistent with optimal contextual relevance must be one in which the speaker has just, or not long ago, had breakfast. On the other hand, there is no such reasonable expectation of the speaker’s having been to Tibet, so (11b) most readily receives an indefinite, existential reading. (11) a. I have had breakfast. b. I have been to Tibet.

Precisely because under normal conditions discourse material is presupposed to be relevant, when information is not explicitly given, it is ‘filled in’; the interpretive process enriches or augments the linguistic content of the discourse. In (12) the material in brackets is not made explicit, but is implicit (Carston 1998). (12)

a. b. c. d.

He took off his boots and [then] got into bed. She gave him a push and [consequently] he fell over the edge. Writing my essay will take [an appreciable amount of] time. He hasn’t had any lunch [today].

Pragmatics also plays a crucial role in MTA choice. Not only is it possible to present the same eventuality using different aspects (1, 2), but with different tenses, too (13, 14). MTA choice, like interpretation, is underspecified by the semantics and depends on pragmatic considerations, including the genre of the discourse (Section 2.1) and the function of the clause in the discourse (Section 2.3). (1) a. Sue had lunch with Bill. b. Sue was having lunch with Bill. (2) a. Jeanne lut le journal. ‘Jeanne read the newspaper.’ b. Jeanne lisait le journal. ‘Jeanne {was reading, used to read} the newspaper.’ (13) Columbus was well aware that the earth {is, was} round. (14) a. A man walks into a bar and says to the bartender,…. b. A man walked into a bar and said to the bartender,….

271

272 Robert I. Binnick

2.  Discourse functions in MTA choice 2.1  Genre Both MTA choice and interpretation are sensitive to the purpose of the utterance, and hence to the genre of the discourse containing it. MTA use is conventionalized in various genres. Narratives (14b, above) are normally recounted in the past tense (Weinrich 1964: 38, 43), while jokes (14a) generally make use of the present tense (Fleischman 1991: 87). Plato (Republic, book 3) distinguishes mimesis, the imitation of reality in genres such as drama, from diegesis, the narrative recounting of events. Benveniste (1959: 70) and Weinrich (1964: Ch. 2) too contrast narration with genres of ‘discourse’ or ‘commentary’ such as drama, criticism, and philosophy, in which the speech act participants are involved with, and not detached from, the eventualities recounted. Sperber & Wilson (1986) similarly contrast the descriptive use of language with the interpretive, which involves a secondary representation by a narrating consciousness. Benveniste, Weinrich and Sperber & Wilson all note the association of diegesis with past tenses (past, conditional, pluperfect, past prospective), and contrasting genres with non-past tenses (present, future, present perfect, present prospective). Sthioul argues (1998: 200) that all tenses allow both descriptive (narrative) and interpretative (discoursal, commentative) uses, and proposes that it is not genre as such that affects MTA choice and interpretation, but rather focalization, the choice of a perspective from which to recount the eventuality (Section 2.2). MTA use depends as well on the specific function of the clause in the discourse (Section 2.3). 2.2  Focalization In diegetic genres there is a perspective or point of view from which the event is narrated, and MTA use is sensitive to this focalization. Narration may be non-focalized, or it may adopt the perspective of an internal focalizer, a character within the narrative, or of an external focalizer outside of it (Fleischman 1990: Ch. 7). In (13), is represents the present time perspective of the narrator, while was reflects the past time perspective of the protagonist Columbus. (13) Columbus was well aware that the earth {is, was} round.

The perspective of the focalizing agent has been identified with the reference point R. Indirect discourse involves an internal focalization in which R is the time of the perception, cognition, or speech act of a protagonist. Thus (15c) uses the pluperfect (had eaten) because the event of eating is past relative to the viewpoint of Mme. Dupont, which is itself



Tense and aspect

anterior to the time of the speech act. Had the non-focalized perspective been selected instead, a past tense would have resulted, as in (15c′) (Lascarides & Asher 1993: 2). (15)

a. b. c. c′. d.

The telephone rang. It was Mme Dupont. Her husband had eaten too many oysters for lunch. Her husband ate too many oysters for lunch. The doctor recommended a change in lifestyle.

A flashback (as in (16)) represents a secondary narrative line, dependent on a past perspective (Kamp & Reyle 1993:  594). Such shifted narratives allow a normally backgrounding tense such as the pluperfect to be used as a foregrounding, narrative tense (Section 2.3). (16) Fred arrived at 10. He had got up at 5; he had taken a long shower, had got dressed and had eaten a leisurely breakfast. He had left the house at 6:30.

Focalization may affect aspect as well as tense. It has been observed that in (17a) the imperfective était represents the omniscient point of view of the narrator, while in (17b) the perfective fut represents that of Napoleon (Fleischman 1990: 217). (17) a. Soudain, joyeux, il dit: Grouchy! — C’était Blücher. ‘Suddenly, joyfully, he [Napoleon] said, “Grouchy!” — It was Blücher.’ (He thought it was Grouchy, but it was Blücher.) b. Il dit: Grouchy! — Ce fut Blücher. ‘He said, “Grouchy!” — It was Blücher.’ (He became aware that it was Blücher.)

Temporal advance in narration is normally associated with the perfective aspect, but internal focalization may trigger exceptional advance with the imperfective aspect, as in (18a), in which the imperfective était ‘was’ represents the viewpoint of Max: he becomes aware of the darkness of the room after he enters it (Moeschler 1998: 3) and (18b), in which the light dazzles Jean after he flips the switch (Kamp & Rohrer 1983: 259). (18) a. Max entra dans la pièce. Elle était complètement noire. ‘Max entered the room. It was pitch dark.’ b. Jean tourna l’interrupteur. La lumière éclatante l’éblouissait. ‘Jean flipped (perfective) the switch. The brilliant light dazzled (imperfective) him.’

By default, an utterance is interpreted as “descriptive, representing a non-focalized state of affairs” (Sthioul 1998:  200). If such a reading is inconsistent or the context favors an alternative reading, the utterance is interpreted as involving an attributed thought or sensation, that is, as focalized (‘interpretive’). In this case the addressee tries to find a focalizer in the discourse in question, and only if that is impossible is an external ‘source of perception’ constructed in a process of accommodation.

273

274 Robert I. Binnick

2.3  Function Choice of an MTA is sensitive to the discourse functions the clause containing the MTA is to perform. There are a variety of types of functions operating at different levels of the linguistic system including referential, textual, expressive and metalinguistic (Fleischman 1990, 1991). The textual function is to create and maintain the coherence, both internal and external, of the discourse, signaling such relations as those of topic and focus, as well as foregrounding material (placing it in the main sequence of narrative events) or backgrounding the material (Hopper 1982; Fleischman 1990: Ch. 6). The foreground of a narrative is expressed using MTAs of perfective aspect, while the background uses imperfective aspect (Weinrich 1964; Hopper 1979, 1982; Kamp 1979). There is a correlation between aspect, Aktionsart, and grounding. The foregrounded material normally consists of bounded activities and events expressed using telic expressions, while the background material consists usually of unbounded states and processes, expressed with atelic expressions. The expressive function is to convey personal attitudes towards the discourse; diegetic sentences lack the immediacy of mimesis and offer a distanced, objective perspective on the events they describe (Fleischman 1991: 81ff.). Commentative, discoursal tenses such as the present fulfill an expressive function in narrative discourse precisely because they are characterized by involvement of the speech act participant(s) in the discourse (Fleischman 1990: 115). The metalinguistic function signals a particular style, register, genre, or type of discourse. For example, the simple past tense of French is formal and restricted to writing, and thus occurs in diegesis but not in mimesis. The French complex past is however a mimetic tense, and Camus achieves a special effect by employing it in his novel L’étranger (The Stranger) as the tense of the (first-person) narration (Fleischman 1990: 254).

3.  Discourse coherence in the interpretation of MTAs 3.1  Discourse coherence Discourses are not merely strings of clauses, but display a hierarchical structure of sub-units which cohere both globally (between units), and locally (within a unit) (Grosz et al. 1995). Discourse coherence and temporal coherence go hand in hand. Discourse structure affects the interpretation of MTAs and MTAs in turn contribute to structuring discourse, principally on the local level, and specifically between pairs of clauses. (19a) is rendered temporally coherent by its set of MTAs, whereas (19b) is incoherent.



Tense and aspect

(19) a. John was distraught. He had witnessed a terribe accident. b. John was distraught. He has witnessed a terrible accident.

Centering Theory (Grosz et al. 1995) assumes that there are three levels of discourse coherence: linguistic, intentional, and attentional. On the linguistic level, coherence consists of the proper anchoring of the tenses. In theories such as Discourse Representation Theory (Kamp & Reyle 1993) and Centering Theory, anchoring means the binding of temporal anaphors by antecedents (Section  3.2), much as pronouns are bound (Kehler 1994). In theories like Segmented Discourse Representation Theory (Lascarides & Asher 1993), anchoring consists in attaching each clause to some segment of the discourse (Sections 3.3, 3.4). (19b) is incoherent because the first clause cannot serve as anchor for the second one. On the intentional level, coherence means that each clause serves some function and hence is related by a rhetorical relation (Section 3.3) such as elaboration or explanation to its anchor (Hobbs 1979; Lascarides & Asher 1993). In (19a), for example, the second clause serves as an explanation for the first. There is, however, no rhetorical relation that allows the temporal pattern of (19b); (19b) is incoherent because its second clause cannot serve any discourse purpose relative to the first clause. On the attentional level, coherence means that each clause belongs to a thread (Section 3.4) defined by a common topic or focus (Hitzeman et al. 1995). On this level, (19b) is incoherent because its first clause is about John’s state of distress, but the second clause cannot be interpreted either as being about, or related to, that eventuality. In terms of either sentence generation or interpretation, the three levels interact simultaneously. The clauses of (19b) cannot be interpreted as having a common topic because they cannot be interpreted as linked by a rhetorical relation, given their unrelated R points. But often enough the temporal relation defined by the MTAs depends on the rhetorical relation; (20) and (21) have the same MTAs but different temporal relations because of their different rhetorical relations (Lascarides & Asher 1993). (20) Max stood up. John greeted him. (temporal advance in narration: John’s greeting follows Max’s standing up) (21) Max fell. John pushed him. (temporal regression in explanation: John’s pushing precedes Max’s falling)

Since MTAs contribute to the interpretation of discourse and their own interpretation depends on the cotext as well as the speech act participants’ world knowledge and knowledge of language, the speaker, in generating a sentence, must take these various factors into consideration when choosing an MTA for a clause. 3.2  The linguistic level Based on the observation that tenses act like bound anaphors, many scholars have identified discourse coherence with the binding of temporal anaphors by antecedents

275

276 Robert I. Binnick

in the discourse, whether tenses (22a), temporal adverbials (22b), or nominal expressions (22c). (22) a. John went home early. He took the subway. (The tense of took is bound by that of went.) b. Yesterday Susan was quite hungry. (The tense of was is bound by the adverbial yesterday.) c. The war years were difficult ones for Tom’s family. They missed him a lot. (The tense of missed is bound by the expression the war years.)

The initial utterance in a discourse receives its binding, if any, from just such an adverbial or nominal expression, but in the absence of such, as in the first sentence of Wolfgang Borchert’s short story ‘Die drei dunklen Könige’ (23), if an indefinite reading is impossible, the interpreter must assume some temporal frame of reference; an anchor time must be accomodated, that is, assumed (Steedman 1997). (23) Er tappte durch die dunkle Vorstadt. ‘He groped his way through the dark suburb.’

In a discourse, an adverbial such as an hour later (24a) or a nominal such as his successor (24b) may shift the reference point R. (24) a. Sue ate the cereal even though it tasted a bit odd. An hour later she was bitterly regretting that decision. b. The king was a very learned man. His successor was not well-educated, however.

In the absence of such expressions, MTAs are bound by other MTAs. Time advances in a narrative sequence because the R point of the MTA of each clause is temporally located just after that of the immediately preceding one, and so temporal anaphors, unlike nominal ones, are not identical with their antecedents (Partee 1984). Narrative advance is sensitive to the nature of the eventuality recounted. Normally events trigger advance, and states do not (Kamp 1979; Partee 1984; Hinrichs 1986). In (25a), Sue held the newspaper up just after Tom came in, but in (25b), Sue was already holding it. (25) a. Tom came in. Sue held up the newspaper. b. Tom came in. Sue was holding up the newspaper.

Whether an MTA expresses an event or a state depends on the inherent Aktionsart of the predicate complex the MTA governs, in combination with the aspect of the MTA (Moens & Steedman 1987, 1988; Boogaart 1999:  §3.3). Since events are temporal entities representing a change of state, one explanation for narrative advance is that the R point of a clause is the time of the result state of an event described in the immediately preceding clause (Moens & Steedman 1988). Supporting this is the dependence of binding on the structure of events. The perfective aspect, unlike the perfect and progressive, introduces the entire structure of the event into the discourse: the time of a succeeding sentence or clause may be



Tense and aspect

placed in any part of the event and the succeeding event may occur within the result state (26a) or the preparatory phase (26b) of the preceding event (Webber 1988: 67). On this theory, in the former case discourse time advances because the result state of an event follows the time of the event proper, and in the latter it regresses because the preparatory phase precedes the culmination of the event. (26) a. John went into the flower shop. He picked out three red roses, two white ones and one pale pink. b. John bought Mary some flowers. He picked out three red roses, two white ones and one pale pink.

The time of the event as a whole may also serve as binder. Thus in (27), the state of John’s having twisted his ankle is bound by the time of his going to the hospital and thus is simultaneous with it. (27) John went to the hospital. He had twisted his ankle on a patch of ice.

Where more than one temporal structure is available as binder, ambiguity results, as in (28). This could be read as simultaneous (Jane played the piano while Bill sang), taking the totality of the time Bill sang as binding the MTA of the second clause; or it could be read as successive, taking only the result state of Bill’s singing the song as its R point. (28) Bill sang a song. Jane played the piano.

An activity sentence may be interpreted as either a bounded event with a result state or as purely a preparatory phase. In the latter case (29) a successive reading for the last two sentences is impossible, but in the former, the same type of ambiguity appears as that exhibited by (28). (29) Bill and Jane found things to do while waiting. Bill read Hamlet. Jane did a crossword puzzle.

Since states have a homogeneous phasic structure, the time the state holds is taken as a whole to be the anchor time for a state (30a) or event (30b) expression following in the discourse. (30) a. Sue was unhappy. Max however was ecstatic. (The states temporally overlap: Max was ecstatic at the time Sue was unhappy.) b. Sue was unhappy. Max decided to help her. (The event occurs within the time of the state: Max decided to help Sue while she was unhappy.)

Narration and the consequent narrative advance is the default value for sequences of MTAs, particularly perfective past tenses in diegetic genres. But other rhetorical relations, which may define different temporal relationships between successive MTAs (as in 21), are possible (Section 3.3). (21) Max fell. John pushed him.

277

278 Robert I. Binnick

3.3  The intentional level Rhetorical relations, also called discourse or coherence relations (Hobbs 1979; Polanyi 1985; Thompson & Mann 1987; Mann & Thompson 1988; Scha & Polanyi 1988) bind segments of text into larger structures by subordinating or coordinating discourse segments. Among the subordinating rhetorical relations are explanation (21), elaboration (26b), background (31a), and result (31b). Narration (20) and list (28) and are coordinating relations. (20) Max stood up. John greeted him. (a sequence of events) (21) Max fell. John pushed him. (the second clause states the cause of the event in the first) (26) b. John bought Mary some flowers. He picked out three red roses, two white ones and one pale pink. (the second clause adds more specific content to that in the first clause) (28) Bill sang a song. Jane played the piano. (related, but not necessarily temporally ordered, events or facts) (31) a. The troll swung his club. It was huge and looked more like a log than a stick. (the second clause is background information to the event in the first clause.) b. Tom severed the power cable. The lights went out. (the event recounted in the first clause results in that in the second.)

The sundry discourse relations can be categorized in terms of the various temporal relations they define (Hitzeman et al. 1995: 5). When a ‘satellite’ clause b is attached to a previous ‘nucleus’ clause a by a rhetorical relation, a may precede b (result), follow it (explanation), just precede it (narration), coincide with it (elaboration) or overlap it (background). In the case of a list, the members have no defined temporal relation (as in 28, above), though they may have preferred readings of simultaneity in special contexts (32). (32) a. Dean Martin sang the songs. Jerry Lewis made the faces. b. Sue performed the surgery. Jane acted as her assistant.

Neither the rhetorical relation nor the temporal relation it defines can be read off of sentential syntax and compositional semantics alone. (20) and (21) have the same syntax, but the events described bear dissimilar temporal relations. For this reason Lascarides & Asher (1993) argue that temporal relations follow rhetorical relations. (20) Max stood up. John greeted him. (21) Max fell. John pushed him.

Because Lascarides, Asher, and Oberlander define temporal relations, and hence anchoring, in terms of rhetorical relations rather than in terms of anaphoresis (binding



Tense and aspect

of an anaphor by its antecedent), they treat all tenses as deictic, introducing reference times into a logical form only where there is a relational temporal adverbial such as afterwards, two minutes later, the week before (Lascarides 1992). As we have seen, the relation of narration involves an iconic sequence, as in (8) (Lascarides & Asher 1991: 1). The other relations impose different temporal relations (Lascarides & Asher 1993).

(8) John entered the office of the president. The president greeted him.

The relation of background is one in which a state serves as a ‘backdrop’ or circumstance under which an event occurs (31a), so naturally the state (the club’s being huge and looking like a log) overlaps the event it accompanies (the troll swinging the club). Similarly, in explanation (21) the eventuality in b is a cause of (or otherwise, an explanation for) that in a, and it is necessarily the case that the eventuality in a must not precede that in b. On the other hand, in result (31b) the event expressed in clause a causes the event or state expressed in clause b. If clauses a and b form an example of result, then the event expressed by a must precede the occurrence expressed by b. (21) Max fell. John pushed him. (31) a. The troll swung his club. It was huge and looked more like a log than a stick. b. Tom severed the power cable. The lights went out.

In elaboration (26b) the eventuality expressed in the second clause b is part of the preparatory phase of the eventuality expressed by the first clause a. If b elaborates a, then the eventuality in a necessarily does not precede that in b. (26) b. John bought Mary some flowers. He picked out three red roses, two white ones and one pale pink.

More than one relation can obtain at the same time, as in (33) (Lascarides & Asher 1991: 55). Here Max’s sickness describes the circumstances under which he took the aspirin and hence constitutes background, while at the same time it forms an explanation for why he took the aspirin. (33) Max took an aspirin. He was sick.

Wilson & Sperber (1998: 1) refer to the determination of the temporal relation between clauses as the sequence problem. For Wilson and Sperber the solution of the sequence problem falls out from those of what they call the interval problem (determining the length of the interval between the eventualities described) and the cause-consequence problem (determining the causal relation, if any, between the eventualities). In effect, this is roughly equivalent to Lascarides and Asher’s view that the temporal relation between clauses follows from the rhetorical relation attaching one to the other. The question then is how to determine which rhetorical relation, if any, holds between successive clauses. To begin with, the interpretation must be consistent with

279

280 Robert I. Binnick

the semantics of the clauses, both their MTAs and any expressions with temporal content, such as connectives (and, because), adverbials (later, yesterday), and temporal substantives (previous, successor). The inference of the rhetorical relation holding between a pair of clauses depends on knowledge of the world and on knowledge of language. (21) receives the preferred reading it does because it is common knowledge that pushing may cause falling, but pushing is unlikely to occur as a result of, or consequent to, falling. Interpretation also depends on principles of logic, since relations between clauses can be mediated by the implicit consequences of explicit clauses, as in (34a), the coherence of which depends on the knowledge that ice cream is a sweet. (Cf. the incoherent 34b.) (21) Max fell. John pushed him. (34) a. Sue said Max had eaten the ice cream. Max never eats sweets, though. b. Sue said Max had eaten the ice cream. Max never eats meat, though.

The process of interpretation is not deterministic; the knowledge base and logic do not necessarily point to a unique interpretation, but merely serve to filter out certain readings as impossible, while evaluating some of the remaining, possible readings as preferred or most likely. Lascarides & Asher (1993: 6) model the logic of interpretation in what they call Discourse and Commonsense Entailment, which is not intended as a psychologically plausible theory of human discourse processing. In it the knowledge base is modeled as a set of formalized pragmatic rules (axioms and meaning postulates) (Lascarides & Oberlander 1993:  13). Some of the rules thus formalized express world knowledge (e.g. pushing precedes falling; causes precede their effects), others linguistic knowledge. In addition, a defeasible logic (one in which rules have exceptions, and conflicts may occur between the rules) operates on this knowledge base to determine the rhetorical relation, and hence the temporal relation, of the discourse segment in question to the point in the discourse it attaches to (Lascarides & Asher 1993; Lascarides & Oberlander 1993). Among the patterns of logical inference incorporated in the logic is defeasible modus ponens. From the premises (i) if p, then normally q and (ii) p, one nonmonotonically (further premises can alter the conclusion) obtains the conclusion (iii) q: Defeasible modus ponens plays a role, for example, in interpreting (20). If a clause b is attached to a clause a by a rhetorical relation, then normally they form a narration (if p, then normally q); b is attached to a (p); therefore they form a narration (q). (20) Max stood up. John greeted him.

If two defeasible laws result in conflicting conclusions, and the antecedent to one of the laws entails the antecedent of the other, the inference is drawn based on the entailing antecedent. This is called the Penguin Principle (Lascarides & Asher 1993), from the



Tense and aspect

following fact: from the premises that (i) penguins are birds, (ii) penguins don’t normally fly, (iii) birds normally fly, and (iv) Tweety is a penguin we deduce that Tweety doesn’t fly, because Tweety’s being a penguin entails, and hence is more specific than, Tweety’s being a bird. If the principle that causes precede effects is more specific (as a special case) than the principle of iconic ordering in a narrative sequence (as the default case), then the Penguin Principle accounts for the preferred interpretation of (21) as a non-narrative. The Penguin Principle may be taken as a special case of Wilson & Sperber’s principle of optimal relevance. (21) Max fell. John pushed him.

Where the antecedents of defeasible rules aren’t related, however, the Penguin Principle is of no help, and indeed, no inference is possible. This is captured by the principle called the Nixon Diamond from the fact that no conclusion can be drawn as to whether Nixon is a pacifist or not from the set of premises (i) Quakers are normally pacifists, (ii) Republicans are normally non-pacifists, (iii) Nixon is a Quaker, and (iv) Nixon is a Republican. The approach of Lascarides and her associates is computationally intense and inefficient (Hitzeman et al. 1995), but the computational load on the interpreter is considerably reduced by the powerful mutual constraints rhetorical relations, tense and aspect impose on one another. A state can elaborate another state (35a) or event (35b), but an event can only elaborate another event (36a), not a state (36b); contrast (36c), where Mary’s building a dog house has been converted into a state by use of the progressive. (35) a. Mary was tired. She was exhausted. b. Mary built a dog house. It was a labor of love. (36) a. Mary built a dog house. She used two tons of bricks. b. Mary was {tired, working hard}. ?She built a dog house. c. Mary was {tired, working hard}. She was building a dog house.

Explicit cues also considerably assist the process of interpretation. They constrain or even indicate the temporal relation between clauses (in 37 because indicates temporal regression), and override any defaults (as in 38, where the adverbial the previous Thursday overrides the default value of simultaneity of the clauses, yielding a reading in which the second eventuality precedes the first). (37) John entered the room because Mary stood up. (38) John was in Boston. The previous Thursday he was in Detroit.

The role of explicit cue terms forms an important part of several theories, and requires some mechanism for determining the relative weighting of logic, the knowledge-base, and explicit markers in contributing to inferences of rhetorical and temporal relations.

281

282 Robert I. Binnick

The interplay of explicit, linguistic elements such as temporal adverbials and connectives on the one hand and contextual components on the other forms precisely the subject of Moeschler’s (1998, 2000) theory of directional inferences. Moeschler (2000: 4f.) denies that temporal directionality is tied either to tense or to aspect, pointing out that the perfective aspect does not always advance time (39), while the imperfective can sometimes trigger an advance, as in (18a). Nor is it tied to tense, as shown by (20) and (21). (18) a. Max entra dans la pièce. Elle était complètement noire. ‘Max entered the room. It was pitch dark.’ (20) Max stood up. John greeted him. (21) Max fell. John pushed him. (39) Bianca chanta le récitatif et Igor l’accompagna au piano. ‘Bianca sang the recitative and Igor accompanied her on the piano.’

Connectives make the temporal direction explicit. And causes an advance (40), because (41) a regress (Moeschler 2000: 6). (40) Max poussa Jean et il tomba. ‘Max pushed Jean and he fell.’ (41) Jean tomba parce que Max le poussa. ‘Jean fell because Max pushed him.’

Based on distinctions made by Sperber & Wilson (1986), Moeschler (1998, 2000) distinguishes contextual and linguistic, and procedural and conceptual, information, and argues that MTAs encode procedural information while expressions such as connectives encode conceptual information. Explicit linguistic elements, MTAs and cue words such as connectives and adverbials, bear directional features of advance (e.g. and and the simple past tense) or regress (because and the pluperfect tense). He distinguishes two types of directional features, strong and weak. The direction of time is given by strong features. To be activated, a feature needs to be validated by a strong feature of the same direction. If no strong feature is borne by any linguistic expression, the temporal direction is inferred from context. But when there is a conflict between a contextual hypothesis and the feature borne by a linguistic expression, the former annuls the latter. Thus there is a hierarchy of strength: contextual information > procedural information > conceptual information (Moeschler 1998, 2000). Tenses bear weak directional features, connectives strong ones. (41), for example, is interpreted as involving temporal regress because the connective parce que has a strong feature of regress while the simple past tense has only a weak feature of advance. (41) Jean tomba parce que Max le poussa. ‘Jean fell because Max pushed him.’

Contextual hypotheses can block directional inferences (Moeschler 2000). In (39), the simple past tenses and the conjunction et ‘and’ both bear forward inferences and



Tense and aspect

should therefore produce a forward inference such as a narration reading. But the normal interpretation of (39) is that the two played together, the forward directional feature being cancelled by the contextual hypothesis (42) and the lexical rule (43). (39) Bianca chanta le récitatif et Igor l’accompagna au piano. ‘Bianca sang the recitative and Igor accompanied her on the piano.’ (42) In a concert, a song and its accompaniment on the piano are simultaneous events. (43) 〈〈x accompany y on the piano〉 IMPLIES 〈x and y play music together and at the same time〉〉

Moeschler proposes to handle cases of elaboration such as (44) in like fashion. The contextual hypothesis (45) blocks the weak forward inference feature of the second clause, confirmed by the explicit element le premier jour. (44) L’année dernière Jean escalada le Cervin. Le premier jour, il monta jusqu’à la cabane H. Il y passa la nuit. Ensuite il attaqua la face nord. ‘Last year Jean scaled the Matterhorn. The first day, he ascended to cabin H. He passed the night there. Then he attempted the north face.’ (45) To scale a summit in the Alps is a complex event composed of sub-events which are temporally ordered.

Cases of temporal regression (e.g. 46) are also explained by contextual hypotheses (48); (48) is based on the conceptual rule (47) and constitutes an implicit premise. (46) Socrates died of poisoning. He drank hemlock. (47) 〈〈x drink hemlock〉 CAUSES 〈x die of poisoning〉〉 (48) If Socrates drinks hemlock, then Socrates dies of poisoning.

Whether specific features such as Moeschler posits actually enter into the process of interpretation or not, it does seem that some weighting must be assigned, explicitly or implicitly, to the various elements that enter into the interpretation of MTAs and the clauses containing them. On the intentional level a clause is attached to some other clause in the discourse by one or more rhetorical relations. The process of interpretation, or alternatively that of discourse generation, involves not only determining what that relation is, but which discourse segment the clause attaches to. Here the intentional and attentional levels (Section 3.4) work together; rhetorical relations serve to structure discourse on the global as well as the local level, and discourse structure in turn plays a crucial role in determining the discourse segments to which the clause in question may, and that one segment to which it in fact does, attach to. 3.4  The attentional level Each segment of a discourse can be said to be about something, to have a topic. The structure of a discourse correlates with this topicality. Subordinate structures change

283

284 Robert I. Binnick

the topic, as when an extended flashback like (16) provides background or one or more sentences constitute an explanation. (16) Fred arrived at 10. He had got up at 5; he had taken a long shower, had got dressed and had eaten a leisurely breakfast. He had left the house at 6:30.

A thread then may be defined as a sequence of clauses sharing a topic. A thread may, however, embed a related subsidiary thread. Because the clauses of a thread are attached by rhetorical relations, they cohere not only thematically but temporally. Thus a thread is also associated with a chain of related times. When a clause is added to a discourse by a speaker (or is interpreted by a hearer or reader) it may continue the current thread of discourse and so exemplify coordinate structuring, as the second clause in (49), or it may create a subordinate thread, exemplifying subordinate structuring (50). In this case either the subordinate thread may be further continued, as in (51a), or the superordinate thread resumed (51b). Or yet a third thread may be created, as in (51c). (49) Sue came home and relaxed. (50) Sue came home. She had had a rough day at work. (51) a. Then she had had to fight bad traffic on the freeway. b. She changed out of her workclothes and sat down in front of the T. V. c. Her boss had yelled at her for something that was not her fault.

Sometimes a discourse segment is ambiguously structured (Webber 1988:  79). The second sentence of (52) could be coordinate with the first and form a continuation of the thread (we refers to the narrator and Frank), or initiate a subordinate thread (we refers to the narrator and Ira). (52) I told Frank about my meeting with Ira. We talked about ordering a Butterfly.

Subordinate structuring is characteristic of a change of reference point. The ambiguity of (52) arises because the two clauses share the past tense. Had the second clause contained a pluperfect or conditional tense (53), the preferred reading would have been an R-shifted, subordinating one. (53) I told Frank about my meeting with Ira. We {had talked, would talk} about ordering a Butterfly.

Subordination may be explicitly marked by ‘cue’ words or phrases such as connectives like because (54a) and prepositions like because of (54b). (54) a. Jane left because Max insulted her. b. Jane left because of Max’s having insulted her.



Tense and aspect

Certain rhetorical functions subordinate material as well, as we saw in Section  3.3. The implicit explanatory relation between the clauses in (21) subordinates the second clause to the first. (21) Max fell. John pushed him.

Temporal centering theory (Kameyama et al. 1993: 71) assumes that each temporal operator in a sentence introduces a discourse referent time into the discourse context. This set of times constitutes the potential future reference times, the ‘temporal forwardlooking centers’. Each sentence contains a unique temporal ‘backward-looking center’, the most salient temporal entity at that point in the discourse. In the transition from one sentence to the next, the same backward-looking center may be retained; a member of the set of forward-looking centers may be established as the new backward-looking center; an old backward-looking center, not in the set of forward-looking centers, may be resumed; or there may be a transition to an attentional state with no backward-looking center. For example (Kameyama et al. 1993:  73), the initial utterance (55a) has no backward-looking center, but the reference time (r1) of the time (t1) of went over provides a forward-looking center. Because (55b) shares the reference time, that time r1 is the backward-looking center of (55b). (55b) also establishes a set of forwardlooking centers, r2 = r1 (for had) and r3 corresponding to the t3 of stopped. (55c) retains r1 as backward-looking center; it continues the thread of (55a–b). It establishes a new forward-looking center, namely r4 = r1. But (55c′) establishes r3 as a new backward-looking center. (55)

a. b. c. c′.

John went over (t1) to Mary’s house. On the way, he had (t2) stopped (t3) by the flower shop for some roses. Unfortunately, they failed (t4) to cheer her up. He picked out (t4′) five red ones, three white ones, and one pale pink.

When a subordinate structure is created, the current temporal focus (set of temporal centers) is stored in a stack for potential later resumption. Thus retention of the center corresponds to maintenance of the thread and resumption of a center to hopping in the discourse structure. Temporal centering assumes that there is one thread that the discourse is currently following, and that there is a preference for a new utterance to continue a thread which is linguistically or semantically related to it, with a preference for continuing the current thread rather than switching to another one (Hitzeman et al. 1995: 4). To properly interpret an MTA the interpreter has to know how it is anchored. In Segmented Discourse Representation Theory this means knowing how its clause is attached, that is, by what rhetorical relation and to what attachment site (Lascarides & Asher 1993). The class of open or available sites is constrained in this theory by two

285

286 Robert I. Binnick

constraints: the right-frontier constraint and the no-resumption constraint (Lascarides & Asher 1993, following Scha & Polanyi 1988). The former says that the ‘open’ (available) sites for attachment are those on the right-hand side in a structural diagram of discourse. The latter says that an explanation or elaboration that occurs earlier than the previous clause cannot be resumed. These constraints are readily explicable in Centering Theory terms: the rightfrontier constraint captures the fact that new material cannot be added to the middle of a thread, but only at the end (the right-hand side of diagrams corresponds to the ends of subordinate threads such as narrative lines and extended explanations or elaborations), and the constraint on resumption that a completed thread which has been ‘popped’ out of the stack, along with its subsidiary threads, no longer exists and hence is no longer available to be added to. In the case where the end of more than one thread is ‘open’ (available), the interpreter seeks to render the discourse coherent on all levels by finding an anchor for the clause in question, a rhetorical relation holding between the clause and its anchor, and a common topic or focus. In generation, the speaker seeks to construct a coherent discourse by, amongst other things, selecting an MTA leading to a coherent interpretation. A theory of discourse structure defines the potential attachment sites (threads), but the actual point of attachment, or preferred point of attachment in interpretation, depends on relevance and the possibility of finding a rhetorical relation between the clause in question and its attachment site. Where a discourse is coherent, such a relation can be found. Where that relation is lacking, incoherence results, as in (56); Tom’s never leaving the house is simply irrelevant to Sue’s claim about Max’s eating ice cream, and vice-versa. (56) Sue said Max had eaten the ice cream. Tom never leaves the house, though.

To be sure, a context that would render (56) coherent could be reconstructed via a process of accommodation. For example, Tom’s leaving the house might be a prerequisite in some way for Max’s eating the ice cream; perhaps the ice cream was outside Max and Tom’s house and Max never goes anywhere without Tom. In contrast, (57a, b), where there is obvious mutual relevance between the clauses, are clearly coherent. (57) a. Sue said Max had eaten the ice cream. She’s a notorious liar, though. b. Sue said Max had eaten the ice cream. Max never eats sweets, though.

The interpretation of MTAs in context, like the choice of MTAs in discourse generation, depends on the linguistic information contained in the cotext, contextual information (including knowledge of the world and knowledge of language), and the logic of inference. The process of inferring contextually preferred readings for MTAs is a complex one which has only just begun seriously to be explored in the last decade and a half, and which no one theory can as yet fully model or explain.



Tense and aspect

References Benveniste, E. (1959). Les relations de temps dans le verbe français. Bulletin de la Société de Linguistique de Paris 54: 69–82. Boogaart, R. (1999). Aspect and temporal ordering: A contrastive analysis of Dutch and English. Holland Academic Graphics. Carston, R. (1998). The semantics/pragmatics distinction: a view from relevance theory. Working Papers in Linguistics 10, University College London. Fleischman, S. (1990). Tense and narrativity: From medieval performance to modern fiction. University of Texas Press. ———(1991). Toward a theory of tense-aspect in narrative discourse. In J. Gvozdanovic & T. Janssen (eds.), The function of tense in texts: 75–97. North-Holland. Grosz, B.J., A.K. Joshi, & S. Weinstein (1995). Centering: A framework for modelling the local coherence of discourse. Computational Linguistics 21: 203–25. Hinrichs, E. (1986). Temporal anaphora in discourses in English. Linguistics and Philosophy 9: 63–82. Hitzeman, J., M. Moens & C. Glover (1995). Algorithms for analyzing the temporal structure of discourse. In Proceedings of the 7th Conference of the European Chapter of the Association for Computational Linguistics: 253–60. University College, Dublin. Hobbs, J.R. (1979). Coherence and coreference. Cognitive Science 3: 67–90. Hopper, P.J. (1979). Aspect and foregrounding in discourse. In T. Givon (ed.), Discourse and syntax: 213–41. Academic Press. ——— (1982). Aspect between discourse and grammar. In P.J. Hopper (ed.), Tense/Aspect: Between Semantics and Pragmatics: 3–18. John Benjamins. Janssen, T.A. (1996). Deictic and anaphoric referencing of tenses. In W. De Mulder, L. Tasmowski-De Ryck & C. Vetters (eds.), Anaphores temporelles et (in-)cohérence: 79–107. Rodopi. Kameyama, M., R. Passonneau & M. Poesio (1993). Temporal centering. In Proceedings of the 31st Annual Meeting of the Association for Computational Linguistics: 70–77. Ohio State University Kamp, H. (1979). Events, instants, and temporal reference. In R. Bäuerle, U. Egli & A. von Stechow (eds.), Semantics from different points of view: 376–417. Springer Verlag. Kamp, H. & U. Reyle (1993). From discourse to logic. Kluwer. Kamp, H. & C. Rohrer (1983). Tense in Texts. In R. Bäuerle, C. Schwarze & A. von Stechow (eds.), Meaning, use, and interpretation of language: 250–69. Walter de Gruyter. Kehler, A. (1994). Temporal relations: Reference or discourse coherence? In Proceedings of the 32nd Conference of the Association for Computational Linguistics (Student Session): 319–321. New Mexico State University Lascarides, A. (1992). Knowledge, causality and temporal representation. Linguistics 30: 941–73. Lascarides, A. & N. Asher (1991). Discourse relations and defeasible knowledge. In Proceedings of the 29th Annual Meeting of the Association for Computational Linguistics: 55–62. University of California. ———(1993). Temporal interpretation, discourse relations and common sense entailment. Linguistics and Philosophy 16: 437–93. Lascarides, A. & J. Oberlander (1992). Abducing temporal discourse. In R. Dale et al. (eds.), Aspects of automated natural language generation: 167–182. Springer Verlag. ———(1993). Temporal coherence and defeasible knowledge. Theoretical Linguistics 19: 1–37. Mann, W.C. & S.A. Thompson (1988). Rhetorical structure theory: Toward a functional theory of text organization. Text 8: 243–281.

287

288 Robert I. Binnick Moens, M. & M. Steedman (1987). Temporal ontology in natural language. In Proceedings of the 25th Annual Meeting of the Association for Computational Linguistics: 1–7. Stanford University ——— (1988). Temporal ontology and temporal reference. Journal of Computational Linguistics 14: 15–28. Moeschler, J. (1998). Directional inferences and the conceptual/procedural encoding distinction. In Proceedings ot the Relevance Theory Workshop: 3–8. University of Luton. ——— (2000). L’ordre temporel dans le discours: le modèle des inférences directionelles. In A. Carlier, V. Lagae & C. Benninger (eds.), Passé et parfait: 1–11. Rodopi. Partee, B.H. (1973). Some structural analogies between tenses and pronouns. The Journal of Philosophy 70: 601–09. ——— (1984). Nominal and temporal anaphora. Linguistics and Philosophy 7: 243–86. Polanyi, L. (1985). A theory of discourse structure and discourse coherence. In W. Eilfort, P. Kroeber & K. Peterson (eds.), Papers from the General Session at the 21st Regional Meeting of the Chicago Linguistics Society: 25–27. Chicago Linguistics Society. Scha, R. & L. Polanyi (1988). An augmented context free grammar. In D. Vargha (ed.), Coling Budapest: Proceedings of the 12th International Conference on Computational Linguistics: 573–77. John von Neumann Society for Computing Sciences. Sperber, D. & D. Wilson (1986). Relevance: Communication and cognition. Blackwell. (2nd ed., 1995.) Steedman, M. (1997). Temporality. In J. Van Benthem & A. ter Meulen (eds.), Handbook of Logic and Linguistics: 895–938. Elsevier North Holland. Sthioul, B. (1998). Temps verbaux et point de vue. In J. Moeschler (ed.), Le Temps des événements: pragmatique de la référence temporelle: 197–220. Éditions Kimé. Thompson, S.A. & W. Mann (1987). Rhetorical structure theory: A framework for the analysis of texts. IPRA Papers in Pragmatics 1: 79–105. Waugh, L.R. (1975). A semantic analysis of the French tense system. Orbis 24: 436–85. ——— (1991). Tense-aspect and hierarchy of meanings: Pragmatic, textual, modal, discourse, expressive, referential. In L. Waugh & S. Rudy (eds.), New vistas in grammar: 242–59. John Benjamins. Webber, B.L. (1988). Tense as discourse anaphor. Computational Linguistics 14: 61–73. Weinrich, H. (1964). Tempus: Besprochene und erzählte Welt. W. Kohlhammer. (2nd ed., 1971.) Wilson, D. & D. Sperber (1998). Pragmatics and time. In R. Carston & S. Uchida (eds.), Relevance Theory: Applications and implications: 1–22. John Benjamins.

Word order Mirjam Fried Czech Academy of Sciences

The one-dimensional, linear nature of linguistic expression represents a fundamental property of human language and yet, it is evidently at odds not only with the fact that any particular linearization likely expresses a piece of conceptually complex reality, but also that it does so by using a hierarchically organized, multi-dimensional grammatical system. Moreover, languages display strikingly different linearization patterns, while at the same time showing equally striking regularities. Issues of word order and word order variation thus can, and have been, studied from various perspectives, with various research goals in mind, and on the basis of various types of empirical evidence. Overall, word order research has been shaped by three major, sometimes overlapping but not always interrelated, approaches to language analysis each of which focuses on a different dimension of linearization issues: word order as a syntactic phenomenon; as an expression of pragmatic relations in discourse; and as a representational issue in linguistic theory. In addition, word order phenomena have always attracted attention of historical linguists, as a specific area within the study of language change and variation. One area in linearization research that is only beginning to emerge is the study of word order in child language acquisition.

1.  Syntactic typology The typologically oriented research has its systematic beginning in Greenberg’s (1963) pioneering work on word order universals, which gave birth to a distinct area of typological studies, namely, word order typology. Its main focus is on the order of sentence constituents, but it is also concerned with the order of morphemes within words. At the clause level, the bulk of systematic research has been devoted to the relative order of the verb and its complements in basic transitive sentences. The relevant categories are expressed in terms of S(ubject), O(bject), and V(erb) and languages are said to fall into six distinct word order types depending on which of the six logically possible orders (SOV, SVO, VSO, VOS, OVS, OSV) appears to be the predominant order in a given language. On the basis of his data sample, Greenberg also proposed that we can expect correlations between certain clause-level orders and the order of elements below the

290 Mirjam Fried

clause level, both within syntactic phrases (e.g. the relative order of the adjective, the genitive phrase, or a determiner in a noun phrase) and within morphologically complex words (e.g. suffixation vs. prefixation). Greenberg’s correlations are expressed in the form of very carefully formulated unidirectional implications, such as ‘languages with dominant VSO order are always prepositional’ (universal 3) or ‘if a language has dominant SOV order and the genitive follows the governing noun, then the adjective likewise follows the noun’ (universal 5). This line of research, refining and further developing Greenbergian universals, is most prominently associated with the work of Lehmann (1973), Vennemann (1972), Hawkins (1983, 1994), and Dryer (1992). A major contribution of this approach consists in capturing cross-linguistic generalizations about syntactic word order patterns. However, the typology by itself does not explain why certain orders are disproportionately less frequent across the world’s languages (OSV) than others (SVO, SOV, VSO). Nor is it concerned with investigating what functions might be served by variable word order or what cognitive mechanisms might be at work in accounting for the distribution of word order patterns. Consequently, syntactic typology leaves a large number of languages unaccounted for, dismissing them as typologically irrelevant. Specifically, if a language falls into the poorly defined category of a ‘free word order’ language (roughly, a language in which all six permutations in ordering S, V, and O occur in natural discourse), it mostly stays outside of typological investigation with respect to its linearization habits. A somewhat more inclusive view is advocated by Dryer (1989), who argues that such languages still offer typologically valuable material since even high degrees of ordering flexibility at the clausal level often co-occur with relatively fixed linear patterns at the phrasal level. In addition, it is not self-evident that the very notion of ‘basic’ word order is particularly useful or informative, although the assumption that a basic word order can be identified for a given language has been embraced by many typologists (Lehman 1973; Vennemann 1972; Nichols 1986; Dryer 1988; several papers in Siewierska 1997). Some researchers acknowledge problems with identifying what exactly should count as basic (e.g. Mallinson & Blake 1981) but the possibility that we may not be able to determine for a given language a basic word order in terms of V, S, and O is seen as an exception (Hawkins 1983:14).

2.  Pragmatic functions of word order A number of functional typologists have questioned the universal status of the notion of basic (syntactic) word order. Thompson (1978), in her work on Modern English, was among the first to include functional considerations in a typological study, and



Word order

there has been a growing body of typological literature presenting evidence that if we wish to understand the phenomenon of word order flexibility cross-linguistically, we must pay close attention both to the cognitive dimension of linearization patterns and to the functions served by word order in communication (e.g. Givón 1981, 1983, 1988; Payne 1987, 1990; Siewierska 1988; Harold 1995; Vilkuna 1997). Several scholars argue explicitly that the notion of basic word order is of no typological value if it is based on purely structural categories (notably Mithun 1987; Hale 1992). A very useful overview of the main arguments in this debate can be found in the introduction to Payne’s 1992 book on variable word order, in which she lays out the foundations of functionally and cognitively informed typological research. The question of motivating word order variations has been always one of the main concerns of functional linguistics, which can be described — broadly — as an approach interested in the relationship between linguistic form and the pragmatic functions it serves in speaker–hearer interaction in actual discourse, whether written or spoken. Among the first proponents of a systematic study of the discourse-functional basis of word order variation are Prague School linguists, starting with Mathesius’ (1915) observation that a given utterance can be analyzed as a bipartite structure in which one part establishes what the sentence is about (its ‘theme’ in subsequently established terminology) and the other what the speaker says about that theme. This fundamental insight that sentence structure (not just larger text) ought to be analyzed as a reflection of the speaker’s communicative strategies in structuring information flow has been actively pursued by many other scholars, both within Prague School and outside (Bolinger 1952 can be credited with introducing such an approach into American linguistics), and has developed into a rich area of study. The Prague School tradition is associated with two fundamental notions: communicative dynamism and the theme-rheme articulation. The theme-rheme distinction is defined in terms of informational categories which express ‘functional sentence perspective’ (FSP): theme is the relatively uninformative part of an utterance, while rheme is what pushes communication forward and is not (readily) recoverable from the preceding discourse. These notions are understood as relational entities that are distributed along a continuum called ‘communicative dynamism’ (CD); CD is defined as a scale that captures the degrees to which particular sentence elements contribute to the development of communication (Firbas 1966: 240, 1987). Thus theme is an element carrying the lowest degree of CD, while rheme carries the highest degree; importantly, both theme and rheme may consist of more than one sentence element. Possible instantiations of CD, known as ‘theme progressions’ have been proposed by Daneš (1974). With respect to the relationship between FSP and word order, Firbas maintains that while it is common and, cognitively speaking, expected that messages might be structured as progressions from the least informative to the most informative elements

291

292 Mirjam Fried

(theme-rheme order), such an order by no means represents the only possibility, and thematic status should not be automatically and universally equated with sentence-initial position even in a language in which the pragmatically most neutral linearization pattern consistently obeys the principle of CD, i.e. is essentially of the theme-rheme type (Firbas 1966: 240, 1992: 117ff.). Prague School scholars have always argued that word order in any language, ‘free’ or fixed, may depend on a combination of factors, including syntactic, semantic, prosodic, and pragmatic (e.g. Daneš 1964; Firbas 1966; Sgall, Hajičová & Panevová 1986). The idea that the theme should not be automatically tied to a particular position in a sentence is in marked contrast to Halliday’s (1967) theme; while his approach to information structure is directly inspired by the Prague School framework, he explicitly links themes with sentence-initial placement. The notion of theme thus may mean different things to different linguists, and similar problems of using identical terms referring to notionally and definitionally distinct entities (e.g., topic and focus) surfaces in other theories as well, thus often obscuring an already complicated domain of analysis; informative surveys of the terminological issues can be found especially in Vallduví 1992; Lambrecht 1994; and briefly also in the introductory chapter in Downing & Noonan 1995. The association between the sentence-initial position of an element and its lowinformation status is also central to the distinction between topic and comment (esp. Gundel 1988), which is based more on the notion of aboutness: comment is understood as that portion of a sentence which adds information about the topic. The distinction is strictly bipartite, dividing sentences into sentence-initial topic and non-initial comment, without much sense of internal structure as a possibility for either category. This is thus still different from accounts that work with the distinction topic-focus. One conception of the topic-focus contrast is associated with another strand of the Prague School tradition, represented by the work of Hajičová and her associates, who recast the idea of CD in terms of contextual boundedness, using a specific indexing formalism to capture the gradation (e.g. Hajičová 1984, 1993) and arguing for a finer-grained analysis of the information structure, particularly the focus portion. This issue is exemplified by the sentence Beavers build DAMS (Sgall, Hajičová & Panevová 1986: 57), with intonation peak on dams, which out of context is three-way ambiguous with respect to its information structure: (a) beavers is the topic, the rest of the sentence is focus, (b) beavers build is already in the hearer’s consciousness and hence topical, while only dams is focal, or (c) the whole sentence is contextually unbound and therefore topicless. This basic insight is echoed also in the explicitly articulated typology of focus realizations in Lambrecht’s 1994 theory of information structure, where three types of focus are argued for: predicate focus (corresponding to the (a) interpretation above), argument focus (the (b) interpretation), and sentence-focus (c).



Word order

3.  Cognitive correlates of theme/rheme notions Once it is established that word order may reflect different ways of structuring the flow of information, new research questions emerge. Typologically speaking, for example, the issue of establishing neutral word order can be reevaluated in terms of syntactic versus pragmatic word order as marking a particular language type. Systematic attempts at incorporating pragmatic roles into typological generalizations started with the studies of the relationship between topic and subject and their position in a sentence, as a ‘new typology’ (Li & Thompson 1976, and a number of other influential papers in Li 1976). This entails that we need to establish criteria for determining a pragmatically neutral/ unmarked order as a typologically relevant category (e.g. Payne 1990, 1992). And that, in turn, is at least partly dependent on establishing the cognitive correlates of word order, which is becoming a particularly lively and fruitful area of study. A host of proposals have been put forth about where and how to search for those correlates. While it is now widely recognized that word order reflects segmentation of discourse for the purposes of specific communicative needs, there is much less agreement on what cognitive principles underlie information structure and why languages can differ quite radically from one another in the way their speakers choose to linearize sentences as well as larger stretches of discourse. At the heart of the debate is what is known, roughly, as the old–new distinction, which appeals to the efficiency of information processing as a major motivating factor. Intuitively, old can be associated with the thematic or topical elements, new with the rhematic or focal elements, and the general consensus is that speakers are likely to structure information from what is known to what is unknown (or from ‘less important’ to ‘more important’ in Behaghel’s 1932: 4 terms). The fundamental questions that have always surrounded these intuitive notions and remain under vigorous debate are (i) what exactly is meant by old and new and (ii) why should speakers of any given language use a constituent order reflecting the new–old configuration rather than the presumably more helpful and cognitively plausible old–new structure. The uneasiness with defining old (‘given’ in Chafe 1976) and new stems from the commonly observed fact that for an element to be focal/rhematic, it may very well be the case that its referent is present in the immediately preceding discourse, and thus ‘old’. Firbas (1966: 242, 1987) emphasizes that rhematic need not be new in the absolute sense of ‘absent in preceding context’, but should be also understood as relative to a given piece of discourse. Firbas himself does not develop this observation into more precise theoretical constructs, but there have been attempts in recent decades to approach the basic old–new dichotomy in systematic ways. Several linguists who are engaged in incorporating pragmatic relations into grammatical descriptions operate with a finer-grained taxonomy of notions that capture the sometimes subtle variations in the old–new configurations (e.g. Prince 1992, Birner & Ward 1998). The result are

293

294 Mirjam Fried

distinctions such as hearer–old vs. discourse–old and hearer–new vs. discourse–new, which allow for a much more refined analysis of information packaging as reflected in linearization patterns. An even more detailed typology is proposed by Östman & Virtanen (1999:  107), who try to define and categorize these concepts as operating both at the level of individual sentences and at the level of larger discourse and point out that the two need not coincide. Some researchers refer to these notions in terms of cognitive states in information processing, such as relative degree of activation and accessibility (Hajičová 1993; Lambrecht 1994), salience (Hajičová 1984), or focal attention in relation to old (Tomlin 1995) and high memorability in relation to new (Jarvella 1979); the latter two are based on linguistic experiments measuring the role of attention and recall, respectively. The experimental work also relates to the second question posed above, namely, what are possible explanations for the old–new linearization as compared to the new–old alternative. Tomlin’s experiment, for example, corroborates the general concensus that ‘focally attended’ elements (essentially overlapping with topicality) tend to be expressed sentence-initially. Yet, there are languages that run against this expected, natural order and use a rheme-first pattern as the pragmatically neutral structure (Mithun 1987; Tomlin & Rhodes 1992). Even within generally theme-first languages there are particular sentence patterns in which the unmarked order of pragmatic relations is reversed, putting the most informative element sentence-initially, and not simply for contrasting purposes. Mithun (1987) offers the concept of newsworthiness as a cognitively plausible justification for such linearization, suggesting that the initial position in a sentence may be used for marking the fact that the speaker considers the new information particularly important from the hearer’s perspective (rather than tying that position to the hearer’s presumed shared knowledge or to the speaker’s privileged status of the one who chooses what the utterance will be about). This interpretation of the cognitive salience of the initial position seems supported experimentally by Jarvella’s findings about speakers’ recall patterns: the highest memorability appears to be associated with both sentencefinal and sentence-initial positions, thus making both edges of a sentence cognitively salient and therefore potentially useful for marking special information status. In addition, Mithun’s (1995) study of the rheme-first patterning in the Iroquois languages documents that the sentence-initial position may attract informationally prominent (i.e. rhematic) material also for its inherent prosodic prominence.

4.  Word order in grammatical descriptions and linguistic theory Studying word order from the point of view of grammatical theory revolves around the question of what syntactic devices, including word order, are used to convey differences



Word order 295

in information structure and how word order variations could be best represented as part of a grammatical system. However, the scope and formulation of the research questions as well as methods of finding satisfying answers depend largely on the theoretical perspective one chooses to apply. The issue of pragmatically motivated word order and its status in grammar has been particularly salient in the analysis of ‘free’ word order languages; this interest is reflected in a number of monographs devoted to grammatical descriptions of specific languages, from various theoretical points of view, such as Erguvanlı (1984) on Turkish, Yokoyama (1986) on Russian, Kiss (1987) on Hungarian, Vilkuna (1989) on Finnish. But within the last two decades, the relationship between information structure and word order has become the subject of more systematic attention also in languages which do not possess the full flexibility of ‘free’ word order languages, thus reinforcing the functionalist view that no order is really ‘free’. One particular area of interest seems to center on verb-initial languages that exhibit varying degrees of flexibility in ordering major constituents (e.g. Payne 1990; Longacre 1995; Borsley & Roberts 1996; Carnie & Guilfoyle 2000). However, there has been steadily growing interest in investigating the discourse basis of word order even in languages that are normally classified as ‘fixed’ word order types, or as having ‘syntactic’ rather than ‘pragmatic’ word order, such as English. This expansion beyond ‘free’ word order languages has also shifted the way in which various syntactic patterns involving constituent order have been analyzed, such as dislocations (Those kids, I never liked them), fronting/preposing (Those kids I never liked), clefting (It was Tom that Nancy missed), heavy-NP shift (I consider foolish anyone who leaves his doors unlocked), etc. An illuminating analysis of the pragmatic basis of some of these patterns is offered in Birner & Ward’s (1998) detailed investigation of several constructions in English as attested in written discourse (with some corroborating data from a few other European languages) and in Wasow (2000), who explores the significance of ‘weight’ as a factor in word order variation in English. A systematic treatment of many of these patterns in English and French can be found especially in Lambrecht (e.g. 1986, 1987, 1988), who makes it most explicit that the mental representations of word order phenomena in any language cannot be dissociated from the functional/semantic dimension but instead, need to be analyzed as conventional pairings of form and function, i.e. ‘constructions’ in the sense of Construction Grammar (cf. Fillmore 1989). A shift toward a constructional analysis of word order variation has been emerging in the work of various other linguists as well (Hajičová 1993; Birner & Ward 1998; Currie 2000). Not all of these scholars attribute the same theoretical status to the notion of construction in the way it is applied by Lambrecht within Construction Grammar, but they all are based on the insight that word order cannot be studied independently of its pragmatic functions because neither is necessarily predictable from the other. The inherent connection between the syntactic

296 Mirjam Fried

and functional domains is also one of the core tenets of Functional Grammar, a grammatical model that has paid a very systematic attention to the effects of larger discourse on linearization patterns both at the clause level and within syntactic phrases (Dik 1983, 1989; Rijkhoff 1986; Nuyts & de Schutter 1987; Siewierska 1988). Word order variability has been receiving increasing attention in formal syntax as well (e.g. Corver & van Riemsdijk 1994; Kiss 1995). Word order studies within that tradition are based on the fundamental assumption, shared with syntactic typology, that every language has a basic syntactic word order. Variations in linearization are, then, seen as superficial, language-specific patterns that are derived from the underlying basic order by particular syntactic rules, known as ‘scrambling’. In effect, word order variation in the formal theory is treated as a underlying syntactic property of sentences and remains in sharp contrast to analyses which are concerned with word order patterning as attested in actual usage and as arising from grouping words according to specific communicative needs. 5.  Diachronic perspective Changes in word order represent “probably the most drastic and complex category of syntactic changes” (Li 1977: xii), and as the existing literature indicates, we are far from having satisfying explanations for what exactly triggers major shifts in word order and what mechanisms are involved, whether the changes are conceived of as resulting from shifts in communication strategies and reinterpretations, or as involving reorganization of cognitive representations. The search for explanations tends to reflect division in focus (or emphasis) along the same lines as the theoretical approaches discussed above. Studies within syntactic typology are interested predominantly in the shifts in the basic syntactic orders (e.g. Lehman 1974; Li 1975; several papers in Li 1977), while functionally and cognitively oriented typology explores the effects of pragmatic functions on changes in linearization patterns beyond the order of S, V, and O (e.g. Li & Thompson 1974; Vincent 1976; Harris & Campbell 1995). One of the first areas of focus was especially the relationship between subjects and topics and its manifestations in word order changes (e.g. Vennemann 1974; Lehmann 1976). Formal generative theories advocate the view that change in word order represents a reorganization of the syntactic system itself and is tied directly to language acquisition, although no systematic studies of word order acquisition are available as yet. A comprehensive and influential statement of the generative position can be found in Lightfoot (1979, 1991). Finally, a constructional approach, i.e. framing word order change in terms of formfunction relationships, has gained some currency in more recent research (cf. several papers in Sornicola, Poppe & Shisha-Halevy 2000).



Word order

6.  Concluding remarks The current state of our understanding of the relationship between word order, information structure, and the cognitive principles that may motivate the interaction between the two can be summarized as follows. At the most general level, word order patterning can be motivated by grammatical, pragmatic, or cognitive principles — typically by a combination of these. In addition, even within a single layer of motivations, there may be several a priori equally justifiable principles that need not work in unison, but instead, present speakers with multiple options in selecting specific subsets of principles as the dominant ones (Haiman 1985:  239ff.; cf. also Halliday 1985, although his approach to conflict resolution is not particularly concerned with the cognitive dimension). Within the competing cognitive principles proposed so far, some are more speaker-centered (reliance on the shared stock of knowledge, choice of a starting point), some are more hearer-centered (newsworthiness as an assessment of hearer’s interest; reliance on accessibility in hearer’s memory), and some are motivated by iconicity, such as temporal sequencing of events (e.g. Hopper 1979; Longacre 1990). But as text linguists point out (e.g. Gasparov 1978; Longacre 1980), the choice of a specific linearization pattern in a sentence may also depend on the nature of the text in which it occurs (e.g. expository narrative vs. spontaneous dialogue). It remains an open question how the potential conflicts are resolved and whether or not they form coherent networks of combinations, both within individual languages and cross-linguistically.

References Behaghel, O. (1932). Deutsche Syntax, Vol. IV. Carl Winter. Birner, B.J. & G. Ward (1998). Information status and non-canonical word order in English. John Benjamins. Bolinger, D. (1952). Linear modification. Publications of the Modern Languages Association 67: 1117–1144. Borsley, R. & I. Roberts (eds.) (1996). The syntax of the Celtic languages. Cambridge University Press. Carnie, A. & E. Guilfoyle (2000). The syntax of verb-initial languages. Oxford University Press. Chafe, W. (1976). Givenness, contrastiveness, definiteness, subject, topics, and point of view. In C. Li (ed.): 25–56. Corver, N. & H. Van Riemsdijk (1994). Studies on scrambling: movement and non-movement approaches to free word-order phenomena. Mouton de Gruyter. Currie, O. (2000). Word order stability and change from a socio-linguistic perspective: the case of Early Modern Welsh. In R. Sornicola et al. (eds.): 203–230. Daneš, F. (1964). A three-level approach to syntax. Travaux linguistique de Prague 1: 225–240. ——— (1974). Functional sentence perspective and the organization of the text. In F. Daneš (ed.), Papers on functional sentence perspective: 106–128. Academia.

297

298 Mirjam Fried Dik, S.C. (ed.) (1983). Advances in Functional Grammar. Foris Publications. ——— (1989). The theory of Functional Grammar, Part I. Foris Publications. Downing, P. & M. Noonan (eds.) (1995). Word order in discourse. John Benjamins. Dryer, M. (1988). Object–verb order and adjective–noun order: dispelling a myth. Lingua 74: 77–109. ——— (1989). Discourse-governed word order and word order typology. Belgian Journal of Linguistics 4: 69–90. ——— (1992). The Greenbergian word order correlations. Language 68: 81–138. Erguvanlı, E. (1984). The function of word order in Turkish grammar. University of California Press. Fillmore, C. (1989). Grammatical Construction Theory and the familiar dichotomies. In R. Dietrich & C.F. Graumann (eds.), Language processing in social context: 17–38. North-Holland/Elsevier. Firbas, J. (1966). On defining the theme in functional sentence perspective. Travaux linguistiques de Prague 2: 267–280. ——— (1987). The theme in functional sentence perspective. In R. Dirven & V. Fried (eds.), Functionalism in linguistics: 137–156. John Benjamins. ——— (1992). Functional sentence perspective in written and spoken communication. Cambridge University Press. Gasparov, B.M. (1978). Ustnaja reč kak semiotičeskij objekt. [Oral speech as a semiotic object.] Semantika nominacii i semiotika ustnoj reči. Acta et commentationes Universitatis Tartuensis 442: 63–112. Givón, T. (1981). Typology and functional domains. Studies in language 5: 136–193. ——— (1988). The pragmatics of word order: predictability, importance and attention. In M. Hammond et al. (eds.): 243–284. Givón, T. (ed.) (1983). Topic continuity in discourse: a quantitative cross-language study. John Benjamins. Greenberg, J. (1963). Some universals of grammar with particular reference to the order of meaningful elements. In J. Greenberg (ed.), Universals of Language: 58–90. MIT Press. Gundel, J. (1988). Universals of topic-comment structure. In M. Hammond et al. (eds.): 209–239. Haiman (1985). Natural syntax: iconicity and erosion. Cambridge University Press. Hajičová, E. (1984). Topic and focus. In P. Sgall (ed.), Contributions to functional syntax. Semantics and language comprehension: 189–202. John Benjamins. ——— (1993). Issues of sentence structure and discourse patterns. Prague, Charles University Hale, K. (1992). Basic word order in two "free word order" languages. In D.L. Payne (ed.): 63–82. Halliday, M.A.K. (1967). Notes on transitivity and theme in English, Part II. Journal of Linguistics 3: 199–244. ——— (1985). An introduction to functional grammar. Edward Arnold. Hammond, M., E. Moravcsik & J. Wirth (eds.), Studies in syntactic typology. John Benjamins. Harold, B.B. (1995). Subject–verb word order and the function of early position. In P. Downing & M. Noonan (eds.): 137–162. Harris, A.C. & L. Campbell (1995). Historical syntax in cross-linguistic perspective. Cambridge University Press. Hawkins, J.A. (1983). Word Order Universals. Academic Press. ——— (1994). A performance theory of order and constituency. Cambridge University Press. Hopper, P. (1979). Aspect and foregrounding in discourse. In T. Givon (ed.), Discourse and syntax: 213–241. Academic Press. Jarvella, R.J. (1979). Immediate memory and discourse processing. In G.H. Bower (ed.), Psychology of learning and motivation: 379–421. Academic Press. Kiss, K.É. (1987). Configurationality in Hungarian. Reidel.



Word order 299

——— (1995). Discourse configurational languages. Oxford University Press. Lambrecht, K. (1986). Pragmatically motivated syntax: presentational cleft constructions in spoken French. Proceedings from the Twenty-Second Meeting of the Chicago Linguistic Society: 115–126. Chicago Linguistic Society. ——— (1987). On the status of SVO sentences in French discourse. In R. Tomlin (ed.): 217–261. ——— (1988). Presentational cleft constructions in spoken French. In J. Haiman & S. Thompson (eds.), Clause combining in grammar and discourse: 135–180. John Benjamins. ——— (1994). Information structure and sentence form. Cambridge University Press. Lehman, W. (1973). A structural principle of language and its implications. Language 49: 47–66. ——— (1974). Proto-Indo-European Syntax. University of Texas Press. ——— (1976). From topic to subject in Indo-European. In C. Li (ed.): 445–456. Li, C.N. (ed.) (1975). Word order and word order change. University of Texas Press. ——— (ed.) (1976). Subject and topic. Academic Press. ——— (ed.) (1977). Mechanisms of syntactic change. University of Texas Press. Li, C.N. & S. Thompson (1974). Historical change of word order: a case study of Chinese in its implications. In J.M. Anderson & C. Jones (eds.), Historical linguistics, vol. I: 199–217. North Holland. ——— (1976). Subject and topic: a new typology of language. In C. Li (ed.): 457–490. Lightfoot, D. (1979). Principles of diachronic syntax. Cambridge University Press. ——— (1991). How to set parameters: arguments from language change. MIT Press. Longacre, R.E. (1980). Discourse typology and in relation to language typology. In S. Allén (ed.) Text Processing: 457–486. Almqvist & Wiksell International. ——— (1990). Storyline concerns and word order typology in East and West Africa. UCLA. ——— (1995). Left shifts in strongly VSO languages. In P. Downing & M. Noonan (eds.): 331–354. Mallinson, G. & B.J. Blake (1981). Language typology. John Benjamins. Mathesius, V. (1915). O passivu v moderní angličtině. Sborník filologický 5: 198–220. Mithun, M. (1987). Is basic word order universal? In R. Tomlin (ed.): 281–328. ——— (1995). Morphological and prosodic forces shaping word order. In P. Downing & M. Noonan (eds.): 387–424. Nichols, J. (1986). Head-marking and dependent-marking grammar. Language 62: 56–119. Nuyts, J. & G. De Schutter (eds.) (1987). Getting one's words into line: on word order and Functional Grammar. Foris Publications. Östman, J.-O. & T. Virtanen (1999). Theme, comment, and newness as figures in information structure. In K. van Hoek, A.A. Kibrik & L. Noordman (eds.), Discourse studies in Cognitive Linguistics: 91–110. John Benjamins. Payne, D.L. (1987). Information structuring in Papago narrative discourse. Language 63: 783–804. ——— (1990). Pragmatics of word order: typological dimensions of verb initial languages. Mouton de Gruyter. Payne, D.L. (ed.) (1992). Pragmatics of word order flexibility. John Benjamins. Prince, E.F. (1992). The ZPG letter: subjects, definiteness, and information status. In S. Thompson & W. Mann (eds.), Discourse description: diverse analyses of a fundraising text: 295–325. John Benjamins. Rijkhoff, J. (1986). Word order universals revisited: the principle of Head Proximity. Belgian Journal of Linguistics 1: 95–125. Sgall, P., E. Hajičová & J. Panevová (1986). The meaning of the sentence in its semantic and pragmatic aspects. Reidel. Siewierska, A. (1988). Word order rules. Croom Helm.

300 Mirjam Fried Siewierska, A. (ed.) (1997). Constituent order in the languages of Europe. Mouton de Gruyter. Sornicola, R., E. Poppe & A. Shisha-Halevy (2000). Stability, variation and change of word-order patterns over time. John Benjamins. Thompson, S. (1978). Modern English from a typological point of view: some implications of the function of word order. Linguistische Berichte 54: 19–35. Tomlin, R.S. (1995). Focal attention, voice, and word order: an experimental cross-linguistic study. In P. Downing & M. Noonan (eds.): 517–554. Tomlin, R.S. (ed.), Coherence and grounding in discourse. John Benjamins. Tomlin, R.S. & R. Rhodes (1979). And introduction to information distribution in Ojibwa. Papers from the Fifteenth Regional Meeting of the Chicago Linguistic Society: 307–321. Chicago Linguistic Society. Vallduví, E. (1992). The informational component. Garland Publishing. Vennemann, T. (1972). Analogy in generative grammar, the origin of word order. In L. Heilmann (ed.), Proceedings of the Eleventh International Congress of Linguistics, Vol. 2: 79–83. ——— (1974). Topics, subjects and word order: from SXV to SVX via TVX. In J.M. Anderson & C. Jones (eds.), Historical linguistics, vol. I: 339–376. North Holland. Vilkuna, M. (1989). Free word order in Finnish: its syntax and discourse functions. Helsinki: SKS. ——— (1997). Word order in European Uralic. In A. Siewierska (ed.): 173–234. Vincent, N. (1976). Perceptual factors and word order change in Latin. In M. Harris (ed.), Romance syntax: 54–68. University of Salford. Wasow, T. (2000). Postverbal behavior. CSLI Publications. Yokoyama, O.T. (1986). Discourse and word order. John Benjamins.

Index

A a priori grammar,  70 A Priori Grammar attitude,  67 Abbott, B.,  60, 104 Abbott, M.,  104 ability and possibility,  200–201 abstract illocutions,  103–104 access principle,  163 acquisition of negation,  226–227 Adger, C.T.,  84 adjunct control,  38–40 affirmatives,  211 agent-oriented modality,  199–200 Aksu-Koç, A.,  205 Aktionsart,  269–270 alethic modality,  180–181 Amerindian language,  130 Anderson, L.,  59 Antaki, C.,  81 Antas, J.,  141 antonyms,  213 approximative expressions,  113 arbitrary control,  34, 38, 40–43 arbitrary control vs. pragmatic control,  41 arbitrary signs,  130 architecture of FDG,  92–94 Ariel, M.,  9, 56–57, 147 ascriptive and referential subacts,  111–114 ascriptive subacts,  105 Asher, N.,  273, 275, 278–280, 285–287 aspect,  268–269 asymmetric negation,  215–216 asymmetric negation, subtypes of,  216–217 Atkins, S.,  16 Atkinson, J.M.,  77, 84 attachment rule,  40 attentional level,  283–286 attested evidence,  192, 203 availability principle,  258

B Bach, E.,  118, 257–258, 262 Bak, S-Y.,  153 Bakhtin, M.,  86 Bally, C.,  189 Barrie, M.,  38 Bartlett, F.C.,  75 base rules,  126 base space,  174 Basic Linguistic Action Verbs,  103 Bateson, G.,  6, 75–76 Baym, N.,  86 Becker, A.L.,  67, 71 Behaghel’s first law,  135 behavioural illocution,  104 Behne, D.,  48 Bellugi, U.,  228 Benešová, E.,  147 Benveniste, E.,  268, 272 Berg, M. van den,  109, 114 Bernini, G.,  218, 225 Binnick, R.I.,  10, 125, 268, 270, 272, 274, 276, 278, 280, 282, 284, 286, 288 Birner, B.,  55, 60, 154, 293, 295 Blake, B.J.,  290 Bloom, L.,  226 Bobrow, D.G.,  74 Bolinger, D.,  131, 143, 150, 152, 291 Bolkestein, A.M.,  109–111 Boogaart, R.,  276 Bornkessel, I.,  247 Borsley, R.,  295 Bouissac, P.,  131 boulomaic modality,  186 Brandt, L.,  177 Brandt, P.A.,  177 Bredmar, M.,  85 Bresnan, J.,  35 Brewer, J.P.,  86 Brisard, F.,  1–4, 6, 8, 10, 12, 14 Brown, P.,  199, 213

Brown, R.,  61 Brugman, C.,  165 Bultinck, B.,  3 Büring, D.,  151 Burton-Roberts, N.,  223 Bybee, J.,  26–27, 30, 66–67, 190, 199, 201–202 C Cadiot, P.,  251 Calbert, J.P.,  189 Campbell, L.,  66, 296 Candlin, S.,  84 Cantrall, W.,  126 Capell, A.,  216 Carnie, A.,  295 Carston, R.,  221–223, 257–258, 262, 270–271 centering theory,  275 Chafe, W.,  52–56, 59, 66, 147–148, 293 Chao, Y.R.,  153 Charolles, M.,  53 Cherny, L.,  86 Chesterman, A.,  50, 60 child framing practices,  82, 84 child language acquisition,  86 Choi, S.,  227 Chomsky, N.,  4–5, 21, 35, 67, 117–122, 124, 132–133, 147, 149, 234–235 Christophersen, P.,  53 Cicourel, A.V.,  84 circumstantial modality,  186 Clamons, R.,  109 Clancy, P.M.,  57 Clark, E.,  226 Clark, H.H.,  52–54, 177, 211–212, 226, 259 clause structure,  240–242 Clayman, S.E.,  84 Clift, R.,  81 co-constructions,  69–70 coercion,  28

302 Index Cognitive Construction Grammar,  16 cognitive constructs,  74–75 cognitive correlates of theme/ rheme notions,  293–294 coherence,  78, 275 coherence attribute,  25 Cole, P.,  101, 126 Collins, A.F.,  81 communicated content,  106 communicative dynamism,  291 complement control,  35–6 complement control, kinds of,  37–38 computer-mediated communication,  86 Comrie, B.,  59 conceptual blending theory,  177 conceptual component,  92 conditionals and counterfactuals,  171–172 configurational/lexicalist approach,  36 connectives,  282 construction grammar,  7, 11, 16–17 constructional analysis,  20–21 constructional approaches,  16, 19, 21, 24, 26–27 constructional method,  24 constructions in grammaticalization,  26–28 content frame,  106–107 contentive discourse acts,  97 context and acceptability of GS,  123 contextual component,  92 contextual hypothesis,  282–283 contextualization cues,  77 contrast marker,  109 contrast,  106, 109–111 contrastive topic marker,  110 control,  33 control in adjective phrases,  43–45 control in noun phrases,  43–45 control relations,  37, 46–47 control shift,  36–37, 48 controller types,  33 conventional discourse,  21–25 conventional implicatures,  29 conventional pragmatics,  21–25 conventional/constructional pragmatics,  18, 26

conversation analysis,  24 Cornyn, W.,  216 Corver, N.,  296 Coulson, V.,  177 Couper-Kuhlen, E.,  24, 71 Croft, W.,  7, 16–17, 22, 27, 60, 214, 225–226 cross-linguistic variation,  103–104 Cruse, D.A.,  7, 17, 22 Culicover, P.,  36 Currie, O.,  295 Cutrer, M.,  174 D Dahl, Ö.,  202, 214, 218 Dancygier, B.,  25 Daneš, F.,  147, 233, 291–292 Dasinger, L.K.,  61 Davies, W.D.,  35, 135 Davis, B.H.,  86, 251, 264 De Villiers, P.A.,  226 DE Villiers., J.G.,  226 definite expressions types,  55–59 definite expressions,  51 definite noun phrases,  50–51, 54–56 definite reference,  51–52 definiteness and grammar,  59–60 definiteness effect,  59–60 definiteness marking,  60–61 definiteness,  50 deictic expressions,  172–174 denials,  211–212 deontic logic,  181 deontic modality,  184–185, 205 deontic necessity,  200 deontic possibility,  200 deontic speech acts,  199–200 deontic/boulomaic possibility,  201 depictive sentences,  165 Derbyshire, D.C.,  150 derivational constraints and GS,  122 development of definiteness,  61–62 diachronic perspective,  296 diagrammatic iconicity,  137 Diaz, F.,  81 Dik, S.C.,  4–5, 33, 99, 101, 107, 110, 296 Dingwall, R.,  84 Dinsmore, J.,  174

direct or attested evidence,  190 discourse acts,  96–97 discourse coherence,  274–275 discourse functions,  272–274 discourse representation theory,  275 discourses,  274 dislocations and clefts,  71 dispositional modality,  185 dispositional/circumstantial possibility,  201–202 distinguishing keyings,  76 Dokulil, M.,  233, 235 Donnellan, K.,  9, 51–52 Dooley, R.A.,  150 Dorr-Bremme, D.W.,  78 Downing, P.,  58, 292 downward-entailing expressions,  220 Dowty, D.,  125, 242 Dryer, M.,  218, 290 Du Bois, J.W Ducrot, O.,  20, 221 Duranti, A.,  77 dynamic modality,  186 E Eisenberg, A.,  86 elements, relations, frames,  166–167 Emergence of Grammar attitude,  67, 70 emergent grammar,  66 emergent grammar within linguistics,  70–72 encoding idiom,  22 endophoric and exophoric iconicity,  137 Engelenhoven, A.,  113 English,  203 Enkvist, N.E.,  134–135 epistemic modality,  180–181, 182–184 epistemic modal-root modal,  196 epistemic modals,  195–196 epistemic necessity,  187 epistemic possibility,  187, 202 epistemic probability,  203 Epstein, R.,  54–55, 60 Erickson, F.,  77, 80, 85–86 Erteschik-Shir, N.,  149 Evans, N.D.,  113 evidential particles/affixes,  203

Index 303

evidentiality,  191–193 evidentials,  190–192, 203–204 existential constructions,  59 expansion,  261 experiential iconicity,  134–135 explicit cues,  281 explicit type-shifting,  28 expression of negation,  214–219 expressive discourse act,  97 expressive function,  274 external negation,  208–209 F factivity,  193 Faingold, E.,  55 Farkas, D.,  37 Fauconnier, G.,  8, 20, 139, 162–163, 167, 169–171, 174–175, 220 FDG see functional discourse grammar FDG and verbal interaction,  91–92 F-formation,  80 Fillmore, C.J.,  6, 11, 16–19, 22–24, 27, 29, 75, 163, 240, 243, 295 Fischer, O.,  133–134, 136–137, 140, 142 flashback,  273 Fleischman, S.,  272–274 focalization,  272–273 focus,  106, 174 focus constituent,  108 focus constructions,  108 focus function and subacts,  107–111 focus marker,  107 focus structure,  245 Fodor, J.A.,  122, 126 Foley, W.,  240, 245 Fónagy, I.,  131, 139 footing shifts,  81 Ford, C.E.,  58 Forest, R.,  215 Forman, D.,  126 form-meaning pairing,  7, 19 form-miming-form formula,  138–139 Fox, B.,  24, 57–58, 66 frame analysis,  74, 76 frame and context,  77–79 frame and footing,  79, 81 frame and topic,  78

frame lamination,  76 frame semantics,  16 frame shifts,  81 frame-attunement,  80 frames,  6, 74–77 framing and institutional discourse,  83–84 framing and medicine,  84–86 framing and nonverbal communication,  80 framing in play,  82–83 framing in talk,  80–82 Fraurud, K.,  53 Frege, G.,  3, 50, 161, 250, 260, 262–263 Fregean beginnings,  250 Fretheim, T.,  4, 11, 146, 148, 150, 152, 154, 156, 158, 160 Fried, M.,  11, 16, 24–25, 28, 289–290, 292, 294, 296 function,  274 functional discourse grammar,  5, 91 functional sentence perspective,  291 functionalist,  4 G Gasparov, B.M.,  297 generative semantics,  117 generative semantics, history of,  117–121 generativism,  4 genre,  272 German,  195, 203 gestures,  141 Geurts, B.,  223 Ghomeshi, J.,  59 given and new information,  233 Givón, T.,  12, 18, 58, 66–67, 71, 134, 136, 139, 210–213, 224, 291 Goffman, E.,  6, 76, 79–80, 86 Goldberg, A.,  16, 21, 27 Goodluck, H.,  48 Goodman, N.,  171 Gordon, D.,  124 grammar,  66 grammar-pragmatics interface,  155–157 grammatical relations and linking,  245–247 grammaticization,  13, 26, 66 Greatbatch, D.,  84

Greenberg, J.H.,  11, 141, 214, 231, 289 Grice, H.P.,  253 Gricean pragmatics,  253 Gries, S. Th.,  7, 21, 27 Grinder, J.T.,  46 Grosz, B.J.,  274–275, 148 Gruber, J.,  243 GS see generative semantics GS and deep structure,  121–122 GS research community,  117–119 Guest, E.,  247 Guilfoyle, E.,  295 Gundel, J.K.,  4, 11, 53–54, 56–57, 146–150, 152–154, 156–158, 292 H Haegeman, L.,  210 Haiman J.,  12, 67, 132, 134, 137, 140–141, 297 Hajičová, E.,  147, 234–235, 292, 294–295 Hale, K.,  291 Halliday, M.A.K.,  4, 53, 147, 151, 193, 297 Halmari, H.,  86 Halonen, I.,  264 Hamilton, H.E.,  86 Hanks, P.,  29 Hannay M.,  5, 60, 91–92, 94, 96, 98, 100, 102, 104, 106, 108, 110, 112, 114 Harms, R.T.,  118 Harold, B.B.,  22, 291 Harré, R.,  86 Harris, A.C.,  296 Hasan, R.,  53, 108 Haspelmath, M.,  217–218 Haviland, S.E.,  53–54 Havránek, B.,  231, 233 Hawkins, J.A.,  50, 52–53, 290 Hedberg, N.,  56–57, 147, 155 Heim, I.R.,  52–53, 147 Heine, B.,  8, 66 Helasvuo, M.-L.,  13, 66, 68–70, 72 Hellman, C.,  54 Hengeveld, K.,  5, 91, 94, 96–102, 106–110, 112–114 Heritage, J.,  77, 83 Herlofsky, W.J.,  137, 139, 141 Herring, S.C.,  86 Hetland, J.,  152

304 Index Hickmann, M.,  61 Hilpert, M.,  27 Hinch, H.E.,  216 Hinrichs, E.,  276 Hintikka, J.,  180, 264 Hiraga, M.K.,  139, 143 Hitzeman, J.,  275, 278, 281, 285 Hjelmslev, L.,  51 Hobbs, J.R.,  275, 278 Hoeksema, J.,  220 Holenstein, E.,  231 Holland, D.,  75 Honda, I.,  215 Hopper, P.,  13, 26, 66–68, 70–72, 217, 268, 274, 297 Horn, L.,  10, 20, 126, 197, 208–209, 211, 213, 220–224, 230, 253–256 Hornstein, N.,  35 Hougaard, A.,  177 Hoyle, S.M.,  6, 74, 76, 78, 80, 82, 84, 86 Huang, C-T.J.,  60 Hutchby, C.T.,  83 Hydén, L.-C.,  84 hypoicon,  137

indirect discourse,  272 indirect evidence,  190 indirect speech acts,  10, 124 individual subacts,  106 inferred evidence,  190 infinitival relatives,  44–45 information structure,  146–147 information structure and morphosyntax,  153–155 information structure and sentence intonation,  150–153 intentional level,  278–283 inter- and intralinguistic iconicity,  137 interactional linguistics,  71 interactive discourse acts,  97, 103 interactive frames,  76 intermedial iconicity,  142 internal negation,  208–209 interpersonal level of FDG,  95–96 interviews,  83 I-principle,  255–256 Irvine, J.T.,  79

I iconicity,  129, 133 iconicity as interpretation,  133–4 Iconicity in Language,  132 Iconicity in Syntax,  132 iconicity, history of,  129–33 iconicity, principles of,  134–136 iconicity, research on,  139–141 iconicity, types of,  136–9 icons,  130, 134 identity principle,  163 ignoratives,  113 illocution,  102–105 illocution of discourse act,  102 illocutionary act potential,  10, 197 illocutionary force,  10, 100 illocutionary meaning of modal verbs,  197–199 illocutionary verbs,  193, 200 imagic iconicity,  136 implications of RRG,  247 implicit typeshifting,  28 indefinite NPs,  213 indices,  130 indirect control,  45–46

J Jackendoff, R.,  12, 21, 36, 118, 122, 124, 147, 149, 150, 152, 243 Jacobsen, W.,  215 Jakobson, R.,  4, 12, 131, 230–232 James, D.,  117–118, 120, 122, 124, 126 Janssen, T.A.,  268 Jarvella, R.J.,  294 Jasperson, R.,  66 Jaszczolt, K.,  264 Jespersen, O.,  4, 53, 59, 67, 208–209, 218, 220–224, 230 Johannsson, M.,  154 Johansen, J.D.,  134, 137 Johnson, M.,  240 Johnstone, B.,  80 Joseph, B.D juncture in RRG,  241 K Kadmon, N.,  55 Kahrel, P.,  217–218 Kailuweit, R.,  239 Kameyama, M.,  285 Kamp, H.,  251, 253, 273–276

Kaplan, D.,  259–262 Kaplanean semantics,  259–263 Karcevskij, S.,  231 Kärkkäinen, E.  68 Karmiloff-Smith, A.,  61 Kato, Y.,  220 Katz, J.J.,  122, 161 Kaufer, D.,  177 Kawasaki, N.,  38 Kay, P.,  16–18, 21–22, 24, 27 Kehler, A.,  275 Keizer, M.E.,  113 Kenesei, I.,  108 Kibrik, A.A.,  58 Kiefer, F.,  179–180, 182–186, 188, 190, 192–194, 196, 198, 200–202, 204, 206 Kiss, K.É.,  295–296 Kleiber, G.,  54 Klima, E.,  209–210, 226 knowledge schemas,  76 Kornfilt, J.,  108 Kratzer, A.,  185, 187, 192 Krifka, M.,  264 Kroon, C.,  96 Kuno, S.,  147, 153 Kuroda, S.Y.,  147, 152–153 Kuryłowicz, J.,  233 Kwilosz, D.M.  195 L Ladd, D.R.,  148, 152 Ladusaw, W.A.,  220 Lakoff, G.,  3, 8, 11, 27, 117–118, 121–122, 124–126, 138, 172 Lambrecht, K.,  71, 147–148, 151, 154–155, 292, 294–295 Landau, I.,  34–36, 47 Langacker, R.,  22, 26–27, 135–136, 139, 142, 177 LaPolla, R.J.,  239–240, 247 Lascarides, A.,  270, 273, 275, 278–281, 285–286 Laudan, L.,  123 Laury, R.,  9, 50, 52, 54–56, 58–60, 62, 64 layered structure of clause,  240 least effort hypothesis,  254–255 Lecercle, J.-J.,  140 Lee, C.,  110, 130, 140, 153 Leech, G.,  5, 135, 138, 212–213 left-detached position,  240 Lehmann, W.,  26, 290, 296

Index 305

Lejeune, L.,  217 Leška, O.,  231–232 Leti,  113 levels and layers of FDG,  94–95 Levelt, W.J.M.,  6 Levi, J.N.,  124 Levinson, S.,  9, 79, 199, 213, 254–255, 257, 263 lexicon,  105 lexicon and grammar,  232 Li, C.N.,  59, 53, 293, 296 Lieven, E.,  226–227 Lightfoot, D.,  26, 296 linguistic direction principle,  258 linguistic level,  275–277 linguistic relativity principle,  130 linguistic signs,  129 linguistic tradition,  188–190 logic of conversation,  10 logical positivism,  2 Luelsdorff, P.,  231 Lumsden, D.,  262 Lyngfelt, B.,  12, 33–36, 38–40, 42–44, 46, 48 Lyons, C.,  50, 60–61, 105, 182, 184–185, 204, 251, 263–264 M Mackenzie, J.L.,  91, 97–101, 106–107 Maeder, C.,  137 Mallinson, G.,  290 Mann, W.C.,  26, 97, 98 Manzini, M.R.,  35 markedness of negation,  210–214 Marshall, C.R.,  52 Martinich, A.P.,  253 Master, P.,  62, 130, 226 Masuda, K.,  131 Mathesius, V.,  4, 11, 147, 230–233, 235, 291 Matoesian, G.,  83 Matthiessen, C.M.I.M.,  100 maxim of negative uniformativeness,  212–213 Mazzon, G.,  225 McCawley, J.D.,  3, 4, 117–122, 124, 126 Mccreedy, L.,  84 McNeill, D.,  141

meaning and reference,  162 Meillet, A.,  26 mental spaces,  162–165 mental spaces and discourse management,  174–177 mental spaces theory,  9, 168 metalinguistic function,  274 metalinguistic negation,  221–223 metaphorical extension,  202 Michaelis, L.,  16, 19, 20, 28, 31 Miestamo, M.,  10, 208, 210–212, 214–218, 220, 222, 224, 226, 228 Miller, W.R.,  86, 105, 195 Milsark, G.L.,  60 Minimal Distance Principle (MDP),  36 minimalist principle,  258 Minsky, M.,  74 Mishler, E.G.,  84–86 Mithun, M.,  59, 66, 291, 294 modal adverbs,  182, 204 modal auxiliaries,  194, 204 modal expressions,  194, 199 modal particles,  193 modal verbs,  198, 205 modality,  188–189 modality and grammaticalization,  201–203 modality and pragmatics,  197–203 modality approaches,  179 modality in logic,  179–181 modality types,  185–188 modality, acquisition of,  205–206 modality, development of,  204 Moens, M.,  276 Moeschler, J.,  273, 282–283 Montague, R.,  2, 4, 119, 161 Moon, R.,  29 Morgan, J.L.,  21–22, 79, 126 morphological negation,  214 Moya Guijarro, A.J.,  110 M-principle,  256 MTAs,  276 Mulac, A.,  68 Müller, W.G.,  136–137, 142 N Nänny, M.,  133–134, 136–137, 140

narration,  272, 277 negation,  10, 193, 208 negation and scalarity,  220–221 negation in diachrony,  224–226 negative indefinite pronouns,  218 negative markers,  224 negative polarity items,  219–220 negative politeness,  199 negative transport,  223–224 negatives vs. affirmatives,  215, 217 negatives,  211–212 Nelson, K.,  75 Nemo, F.,  251 nested locatives,  135 neutralism,  84 Newmeyer, F.J.,  118–119, 125 nexal negation,  209 nexus relations,  242 nexus types,  242 Nichols, J.,  290 Nikiforidou, K.,  7, 13, 18, 20, 22, 24, 26–28, 30 nominal expression,  276 nonepistemic modals,  186, 195–196 non-obligatory control,  34 Noonan, M.,  292 Norman, D.A.,  135 Norrick, N.R.,  82 Norrman, R.,  140 Nöth, W.,  129, 137–138 nuclear discourse act,  98–99 Nuyts, J.,  296 O O’Connor, C.,  16–17, 22, 84 Oakley, T.,  8, 161–162, 164, 166, 168, 170, 172, 174, 176–177 objective deontic modality,  184 objective epistemic modality,  184 objective modality,  204 object-oriented verbs,  37 obligatory control,  34 Öhlschläger, G.,  187, 194, 196 old-new distinction,  293 Ono, T.,  66 orientational discourse act,  99, 101 Östman, J.-O.,  16, 23–25, 28, 68, 132, 134–135, 137, 141, 294

306 Index Oswalt, R.,  190 output component,  92 P Pagliuca, W.,  199, 201–202 Palmer, F.R.,  187, 191–193 Panevová, J.,  147, 234, 292 Paolillo, J.,  86 Partee, B.H.,  161, 270, 276 Paul, H.,  13, 117, 131, 147, 149, 253 Pauliny, E.,  233 Payne, D.L.,  210, 214, 291, 293, 295 Pea, R.,  227 Peirce, C.S.,  130 penguin principle,  280–281 Perkins, M.R.,  192–193, 199, 202, 210 perlocution,  193 Petkevič.,  234 Petruck, M.R.L.,  6, 75 Pierrehumbert, J.,  150 pitch accent,  152 Polanyi, L.,  278, 286 Poppe, E.,  296 positive polarity items,  219 positive term,  213 pragmatic analysis,  16–17 pragmatic control,  38–39 pragmatic functions of word order,  290–292 pragmatic intrusion,  256–259 pragmatic particle,  194 pragmatic process,  261 pragmatics and semantics,  123–124 pragmatics in GS,  126–7 pragmatics issues,  234 pragmatics of constructions,  17 pragmatics of grammar,  17–21 pragmatics of modality,  205 pragmatics of tense and aspect,  270–271 pragmatics role in MTA choice,  271 Prague functionalism,  230 Prague functionalism and pragmatics,  235–236 Prague School,  4 pre-core slot,  240 predicative complements,  38 preferred clause structure,  71 prelexical transformations,  125

prepositional verb complements,  38 prepositions,  67 presupposition and optimization,  169–171 presupposition transfer,  170–171 Prince, E.,  52–54, 56, 147, 150, 155–156, 293 principle of proximity,  134 principle of quantity,  134 principle of relevance,  254 principle of sequentiality,  134 PRO,  33 propositional illocutions,  104 prosodic prominence,  108 prosody,  151 proximal deictics,  173 Pullum, G.K.,  119 Pultr, A.,  110 Q Q-principle,  254–256 quantitative iconicity,  135 Quinn, N.,  75 Quirk, R.,  197–198 R Radden, G.,  135 Radical Construction Grammar,  16 raising verb—control verb,  196 Ramat, P.,  218, 225 Rampton, B.,  82 Redder, A.,  198 Redeker, G.,  174 referential choice,  57–58 referential function,  274 referential givenness-newness,  147–149 referential opacity,  168–169 referential subacts,  105 referential theory,  268 referring expressions,  57 Reinhart, T.,  147, 150, 155–156 rejections,  211–212 relational givenness/newness,  148–150 relevance theory,  254 reported evidence,  190 Reyle, U.,  273, 275 rhetorical functions of discourse acts,  97–102 rhetorical relations,  278

Rhodes, R.,  294 Ribeiro, B.T.,  6, 74, 76, 78–80, 82, 84, 86, 88–90 Rijkhoff, J.,  296 Rispoli, M.,  247 Roberts, C.,  223, 254, 295 Rogers, A.,  126 Rohrer, C.,  273 role and reference grammar,  6, 239–240, 261 role and value in reference,  165–166 root modality,  186 Rosenbaum, P.,  36 Roth, A.L.,  83 routinization and emergence of grammar,  67–70 R-principles,  254–255 RRG see role and reference grammar Rubba, J.,  173–174 Rumelhart, D.E.,  74–75 Russell, B.,  50–51, 161 S Sadowski, P.,  134 Sag, I.,  35–37, 147 Sanders, G.A.,  130, 174 Sapir-Whorf hypothesis,  130 Sarangi, S.,  253 Sarbin, T.R.,  86 scalarity,  20 Scandinavian languages,  154 Scha, R.,  278, 286 Schank, R.C.,  74 Scheibman, J.,  68 schemas,  74–75 Schieffelin, B.B.,  86 Schiffrin, D.,  78, 81, 83 Schiller, E. et al.,  126 Schlesewsky, M.,  247 Schmerling, S.F.,  152 Schonefeld, D.,  132 script,  74–75 segmented discourse representation theory,  285 Seiler. H.,  130, 136, 140 Selkirk, E.O.,  152 Selting, M.,  71 semantic diagrammatic iconicity,  138 semantic macroroles,  246 semantic roles,  243

Index 307

semantic structure,  242–245 semantic/lexicalist approach,  36 semantics vs. pragmatics,  250 semi-schematic patterns,  17 sentential negation,  209–210, 220 Sgall, P.,  147, 150, 230, 232–236, 292 Shapiro, M.,  232 Shimojo, M.,  240 Shisha-Halevy, A.,  296 Short, M.,  135, 138, 164, 168, 213, 276 Shultz, J.,  77, 85–86 Sidnell, J.,  141 Siewierska, A.,  290–291, 296 Sinclair, J.,  7, 29, 67–68 Skalička, V.,  232 Šmilauer, V.,  233 Smith, W.,  83, 51–52, 61, 86 social semiotics,  5 Sorjonen, M.-L.,  69 Sornicola, R.,  296 space builders,  167–168 speaker meaning,  259 speaker-oriented modality,  199–200 special negation,  209 speech act types,  200 Sperber, D.,  146, 154–155, 157, 270–272, 279 Stanchev, S.B.,  108–109 standard negation,  214, 218 status of logic,  124 Steedman, M.,  151, 270, 276 Stefanowitsch , A.,  7, 21–22, 27 Stephany, U.,  205–206 Sternefeld, W.,  196 Sthioul, B.,  272–273 Straehle, C.A.,  82 Strawson, P.F.,  51–52, 147 strong negation,  209 structural diagammatic iconicity,  138 structural-functionalist theory of grammar,  239 stylistics,  233 subacts,  105–106 subacts and pragmatic functions,  106–107 subjectification,  26 subjective deontic modality,  184 subjective epistemic modality,  183

subjective modality,  183, 204 subject-oriented verbs,  37 subordination,  284 subsidiary discourse acts,  99 sundry discourse relations,  278 super-equi,  46–47 Sweetser, E.,  8, 25, 171, 202 symbols,  130 symmetric negation,  215, 217 syntactic constructions,  69 syntactic functions,  246 syntactic inventory,  241 syntactic negation,  214 syntactic treatments of modality,  194–197 syntactic typology,  289–290

topic markers,  109 topicality kinds,  110–111 topic-focus articulation,  234–236 topic-focus structure,  157 Tottie, G.,  210–212 tough constructions,  42 transformations,  125–6 transitive construction,  69–70 Traugott, E.,  26–28, 66, 203 Trubetzkoy, N.,  231 Turner, K.,  3, 138, 250, 252, 254, 256, 258, 260, 262, 264, 266 Tuyuca (Tucanoan language),  191, 203 two-tiered theory,  261

T Tabakowska, E.,  12, 129, 130, 132, 134, 136, 138, 140, 142, 144 Taboada, M.,  98 Takubo, Y.,  174 Tannen, D.,  74–76, 78, 81, 85 temporal adverbials,  276, 282 temporal centering theory,  285 temporal regression,  283 temporal relations,  278 tenets of GS,  121 tense and modality,  204–205 tense,  268 Tesnière, L.,  233 textual function,  274 ‘that-deletion’,  68 theme-rheme articulation,  291 theories of grammar,  2–9 theory of discourse structure,  286 theory of meaning construction,  174 Thompson, S.,  13, 24, 59, 66–68, 71, 100, 153, 159, 217, 278, 290, 293, 296 Thurmair, M.,  194 Tobin, Y.,  232 Toman, J.,  231 Tomasello, M.,  16 Tomlin, R.S.,  294 topic,  78, 106 topic analysis,  102 topic and focus,  110, 155–156, 233 topic constituent,  109 topic function,  109

U universal base hypothesis,  126 utterances,  163–165, 276 V Vallduví, E.,  148–149, 152, 292 Van Der Auwera, J.,  197, 217 Van Der Sandt, R.,  222 Van Der Wouden, T.,  219–220 Van Dijk, T.A.,  75 Van Kuppevelt, J.,  221 Van Langenhove, L.,  86 Van Riemsdijk, H.,  296 Van Valin, R.,  5, 239–243, 245–247 Vater, H.,  205 Vendler, Z.,  242 Vennemann, T.,  290, 296 verbs,  199 Verhagen, A.,  25 Verschueren, J.,  63, 103, 116 viewpoint,  174 Vilkuna, M.,  60, 152, 291, 295 Vincent, N.,  296 Virtanen, T.,  294 visual iconicity,  141–142 von der Gabelenz, G.,  147 von Heusinger, K.,  264 Von Wright, G.H.,  188, 207, 209, 210 Vries, L.,  101–102, 107 W Wallat, C.,  76, 78, 85 Ward, G.,  154, 295 Wason, P.C.,  211

308 Index Wasow, T.,  295 Watters, D.E.,  109 Waugh, L.R.,  268 weak negation,  209 Webber, B.L.,  277, 284 Weinrich, H.,  268, 272, 274 Weiser, A.,  126 Wells, G.,  205 Wh-expression,  123 Whorf, B.L.,  130–131 Wierzbicka, A.,  131, 134

Wilkins, D.,  243 Willett, T.,  190–191 Williams, E.,  34, 38 Wilson, D.,  10, 44, 119, 146, 155, 157, 254, 270–272, 279, 281–282 Wolf, W.,  140, 142 word order, changes in,  296 word order in grammatical descriptions and linguistic theory,  294–296

word order variability,  296 Wunderlich, D.,  198 Y Yokoyama, O.T.,  295 Z Zacharski, R.,  56–57, 147, 152 Załazińska, A.,  141 Zentella, A.,  81 Ziv, Y.,  148

In the series Handbook of Pragmatics Highlights the following titles have been published thus far or are scheduled for publication: 10 Sbisà, Marina, Jan-Ola Östman and Jef Verschueren (eds.): Philosophical Perspectives. ca. 250 pp. Forthcoming 9 Östman, Jan-Ola and Jef Verschueren (eds.): Pragmatics in Practice. ca. 250 pp. Forthcoming 8 Ledin, Per, Jan-Ola Östman and Jef Verschueren (eds.): Discursive Pragmatics. ca. 250 pp. Forthcoming 7 Jaspers, Jürgen, Jan-Ola Östman and Jef Verschueren (eds.): Society and Language Use. ca. 250 pp. Forthcoming 6 Fried, Mirjam, Jan-Ola Östman and Jef Verschueren (eds.): Variation and Change. Pragmatic perspectives. ca. 250 pp. Forthcoming 5 Brisard, Frank, Jan-Ola Östman and Jef Verschueren (eds.): Grammar, Meaning and Pragmatics. 2009. xiii, 308 pp. 4 D’hondt, Sigurd, Jan-Ola Östman and Jef Verschueren (eds.): The Pragmatics of Interaction. xiii, 254 pp. + index. Expected September 2009 3 Sandra, Dominiek, Jan-Ola Östman and Jef Verschueren (eds.): Cognition and Pragmatics. ca. 250 pp. Forthcoming 2 Senft, Gunter, Jan-Ola Östman and Jef Verschueren (eds.): Culture and Language Use. 2009. xiii, 280 pp. 1 Verschueren, Jef and Jan-Ola Östman (eds.): Key Notions for Pragmatics. 2009. xiii, 253 pp.