[ad foolen, frederike van der leek] constructions (bookzz.org)

CONSTRUCTIONS IN COGNITIVE LINGUISTICS

AMSTERDAM STUDIES IN THE THEORY AND

HISTORY OF LINGUISTIC SCIENCE

General Editor E. F. KONRAD KOERNER

(University of Ottawa)

Series IV - CURRENT ISSUES IN LINGUISTIC THEORY

Advisory Editorial Board

Raimo Anttila (Los Angeles); Lyle Campbell (Christchurch, N.Z.) Sheila Embleton (Toronto); John E. Joseph (Edinburgh)

Manfred Krifka (Austin, Tex.); Hans-Heinrich Lieb (Berlin) E. Wyn Roberts (Vancouver, B.C.); Hans-Jrgen Sasse (Kln)

Volume 178

Ad Fooien and Frederike van der Leek (eds.)

Constructions in Cognitive Linguistics Selected papers from the Fiflh International Cognitive Linguistics Conference,

Amsterdam, 1997

CONSTRUCTIONS IN COGNITIVE LINGUISTICS

SELECTED PAPERS FROM THE FIFTH INTERNATIONAL COGNITIVE LINGUISTICS CONFERENCE

Amsterdam, 1997

Edited by

AD FOOLEN University of Nijmegen

FREDFRIKE VAN DER LEEK University of Amsterdam

JOHN BENJAMINS PUBLISHING COMPANY AMSTERDAM/PHILADELPHIA

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

Library of Congress Cataloging-in-Publication Data

International Cognitive Linguistics Conference (1997 : He Amsterdam) Constructions in cognitive linguistics : selected papers from the Fifth International Cognitive Linguistics Conference, Amsterdam, 1997 / edited by Ad Fooien, Frederike van der Leek.

p. cm. -- (Amsterdam studies in the theory and history of linguistic science. Series IV, Current issues in linguistic theory, ISSN 0304-0763 ; v. 178)

Includes bibliographical references and index. 1. Cognitive grammar-Congresses. I. Fooien, Ad. II. Leek, Frederike van der. III. Title. IV. Series.

P165.I58 2000 415-4IC21 99-089845 ISBN 90 272 3684 4 (Eur.) / 1 55619 955 4 (US) (Hb; alk. paper) CIP

2000 - John Benjamins B.V. No part of this book may be reproduced in any form, by print, photoprint, microfilm, or any other means, without written permission from the publisher.

John Benjamins Publishing Co. P.O.Box 75577 1070 AN Amsterdam The Netherlands John Benjamins North America P.O.Box 27519 Philadelphia PA 19118-0519 USA

CONTENTS

Acknowledgements vii

Editors' Foreword ix

Pragmatic Conditionals Angeliki Athanasiadou & Ren Dirven 1

How Polish Structures Space: Prepositions, Direction Nouns, Case, and Metaphor

Barbara Dancygier 27

Case Meaning and Sequence of Attention: Source Landmarks as Accusative and Dative Objects of the Verb

Robert B. Dewell 47

Fijian Children's Possessive Categories and Constructions Patrick Griffiths 67

Facing up to the Meaning of 'face up to': A Cognitive Semantico-Pragmatic Analysis of an English Verb-Particle Construction

Beate Hampe 81

Gerundive Nominalization: From Type Specification to Grounded Instance

Liesbet Heyvaert 103

A Cognitive Approach to Errors in Case Marking in Japanese Agrammatism: The Priority of the Goal -ni over the Source -kara

Hiroko & Ikuyo Fujita 123

VI CONTENTS

Verbal Aspect and Construal Agota Kochaska 141

How I got myself arrested: Underspecificity in Grammatical Blends as a Source for Constructional Ambiguity

Nili Mandelblit & Gilles Fauconnier 167

Konjunktiv II and Epistemic Modals in German: A Division of Labour

Tanja Mortelmans 191

Subjectivity and Conditionality: The Marking of Speaker Involvement in Modern Greek

Kiki Nikiforidou & Demetra Katis 111

English Imperatives and Passives Hidemitsu Takahashi 239

Lexical Causatives in Thai Kingkarn Thepkanjana 259

Cognitive Models in Transitive Construal in the Japanese Adversative Passive

Eijiro Tsuboi 283

Caused-Motion and the 'Bottom-Up' Role of Grammar Frederike van der Leek 301

Addresses 333

Index 335

Acknowledgements

This volume was realized with the help of a great many people. First of all we would like to thank all authors for their contributions and cooperation. Thanks are also due to the following people who acted as anonymous reviewers: Michel Achard, Willem Botha, Melissa Bowerman, Claudia Brugman, Eve Clark, Herbert Clark, Kenneth Cook, Barbara Dancygier, Bob Dewell, Matthew Dryer, Patrick Duffley, Peter Harder, Joe Hilferty, Bob Kirsner, Ron Langacker, David Lee, Ricardo Maldonado, Yo Matsumoto, Haruko Minegishi Cook, Laura Michaelis, Kiki Nikiforidou, Jan Nuyts, Jan-Ola stman, Maria Polinsky, Klaus-Uwe Panther, Jo Rubba, Masayoshi Shibatani, Simon Slings, Michael Smith, Leon Stassen, Eve Sweetser, John Taylor, Friedrich Ungerer, Johan van der Auwera, Marjolijn Verspoor and Margaret Winters. We thank Anke de Looper of John Benjamins Publishing Company for her professional advice and her patience and, last but not least, Rob van den Berg for his moral support and his expertise in preparing the final versions of the papers.

Nijmegen/Amsterdam November 1999

Ad Fooien & Frederike van der Leek

Editors' Foreword

This volume contains a selection from the proceedings of the 5th International Cognitive Linguistics Conference, which was held at the Free University in Amsterdam, from July 14-19, 1997. The volume is a companion of two others that also contain selected papers of the same conference: Metaphor in Cognitive Linguistics, edited by Raymond W. Gibbs Jr. and Gerard J. Steen (CILT 175) and Discourse Studies in Cognitive Linguistics, edited by Karen van Hoek, Andrej A. Kibrik and Leo Noordman (CILT 176).

The title of the present volume may be assumed to speak, to a certain extent, for itself. As Goldberg (1995:1) observes, "[t]he notion construction has a time-honored place in linguistics", but was temporarily renounced in the Chomsky an Government and Binding framework. In Cognitive Linguistics, it is given pride of place again, though the notion is, on the whole, subject to different interpretations as far as scope and organizational level are concerned. Langacker (1987:409) sees grammatical constructions as the form-meaning "integration of two or more component structures to form a composite expression". The Construction Grammar framework (cf. Fillmore & Kay to appear) employs the much more rigorous notion that a particular form-meaning combination is a construction only if it has some (form and/or meaning) property that "is not strictly predictable from [its] component parts, or from other previously established constructions" (Goldberg 1995:4). Beyond the consensus, then, that constructions constitute recognizable form-meaning patterns, opinions differ.

This is also evident from the papers making up the present volume. They seem to vary significantly in their vision on what constructions are. The research areas they are concerned with, also cover a wide range of different topics. For this reason we have decided to make no attempt to organize the papers thematically; we simply present them in alphabetical order.

In order to give the reader some preliminary idea of what this volume has to offer, we will, however, first outline its contents from various angles. Except for Griffiths' paper, which deals with child language, all the papers are directly concerned with particular constructions as used by mature speakers. A variety of languages is covered; in alphabetical order: English, Fijian, French, German, (Modern) Greek, (Modern) Hebrew, Japanese, Polish and Thai. Two of the papers, by Griffiths and Ihara & Fujita, are experimentally based and approach the constructions dealt with from psychological angles, i.e.

X EDITOR'S FOREWORD

that of acquisition and agrammatism respectively, a widening of the field that is very welcome.

Thematically, we can recognize, very roughly, the following constructional fields of interest: (verbal) Aspect, (Kochanska), Case (Dancygier; Dewell; Griffiths; Ihara & Fujita), Causation (Mandelblit & Fauconnier; Thepkanjana; Van der Leek), Conditionals (Athanasiadou & Dirven; Nikiforidou & Katis), Gerundive Nominis (Heyvaert), Modality (Mortelmans), Particles (Hampe), and Passives (Takahashi; Tsuboi). Various authors furthermore both argue and illustrate the importance of corpus-based data (Athanasiadou & Dirven; Hampe; Mortelmans; Nikiforidou & Katis). As for what we might call 'cognitive mechanisms', we find the notion Construal to be pervasive throughout all the papers, while the following types of mechanisms play a central role in more specific analyses: Blending and Under-specificity (Mandelblit & Fauconnier), Compositionality (and beyond) (Dancygier; Kochanska; Van der Leek), Grounding (Heyvaert; Mortelmans); Cognitive Models/Frames (Takahashi; Tsuboi; Van der Leek); Lexical Alternation (Thepkanjana; Van der Leek), Subjectivity (Mortelmans; Nikiforidou & Katis), and, lastly, Sequence of Attention (Dewell). Obviously, the above classification is a rough cross-specification, not intended as exhaustive; too many specifically cognitive issues crop up in individual papers to be all mentioned here.

Assuming that the above has nevertheless provided the reader with enough of a first impression of what to expect from this volume, we will now turn to a brief characterization of each individual paper.

Athanasiadou & Dirven discuss English pragmatic, non-prototypical conditionals, a class which, in their conception, also comprises epistemic/ logical conditionals. Their paper, which is corpus-based, argues that the four subtypes they distinguish, 'identifying' and 'inferencing' (both logical) and 'discourse' and 'metacommunicative' (both conversational) all share certain basic pragmatic characteristics that are extended in different ways and the paper aims to show that these differences can account for the variety of forms each type can manifest.

Dancygier analyzes the conceptualization of space in Polish, arguing that the way this language construes space is dependent on three subsystems, i.e. direction nouns, prepositions and case. Her overall conclusion is that the spatial expressions in question are constructed compositionally, with each of the three structural subsystems making a consistent meaning contribution of its own.

Dewell discusses the schematic meanings of accusative and dative cases, arguing that these do not simply reflect different roles in the action chain, but,

EDITORS' FOREWORD XI

instead, impose a certain way of construing an event. In particular, he points out a central difference between the above two cases that concerns the notion 'sequence of attention'. The accusative, he argues, typically makes one change focus from the subject to the direct object referent, while atypical variants (with the accusative NP itself evoking a pathway) effect a construal that makes one stay attentive to both at the same time. The dative, on the other hand, has the effect that the NP referent never gains central attention, but is construed as staying, in a sense, outside the direct action chain. In his comparison of English and German, Dewell shows that the structure of the languages involved (absence of a difference between dative and accusative case in English) is a factor that plays its own role in the type of construals that the two languages make available for an event.

Griffiths analyzes children's acquisition data on Fijian possessive constructions. Fijian differentiates between alienable and inalienable possession and has, besides an unmarked possessive marker, special markers for possession of 'food' and 'drink'. The data suggest that acquisition of this type of markers is cognitively driven in that, firstly, possessive meanings get linguistically expressed before the relevant constructions are mastered, secondly, conceptually simpler categories, e.g. the singular marker and the default possessive marker, are learnt prior to the 'food' and 'drink' ones, and, thirdly, the latter type of markers are applied first to prototypical food and drink categories, and only later to more abstract members of the category.

Hampe shows that the existence in English of the phrasal verb 'to face up to' side by side with the simple verb 'to face' is not to be seen as 'wordy speech', but as a multiply motivated construction, showing that the particle combination conveys that the challenge induced by some problem/obstacle is actually met, whereas the simple verb merely expresses that the problem/ obstacle does not go, so to speak, unnoticed. Conceptually speaking, the meaning of the phrasal verb is 'motivated' by the meaning of its parts, though motivation, as Hampe is careful to point out, is not the same as strict compositionality. From a pragmatic perspective, the phrasal verb is motivated by expressivity, and a corpus analysis method is used to argue this latter point.

Heyvaert presents a proposal for refining Langacker's analysis of English gerundive nominis. She argues that the -ing marker in gerundive nominis indicates, differently from the same marker in action nominis, that the complex predication made up of the verb and its object (which she analyzes, pace Langacker (1991), as a clausal type rather than a clausal instance) has the status of nonfinite clausal head of the gerundive construction, a construction which requires a (periphrastic) subject for its instantiation. This analysis enables her to distinguish intrinsically between action nominis, gerundive

Xll EDITORS' FOREWORD

nominis and that-clmse nominis, and thereby account for the semantic differences these three types of nominis exhibit, as well as for the way they differ in their external behaviour.

Ihara & Fujita discover, from an experimental study with three Japanese subjects suffering from agrammatism, that these patients wrongly use the source marker -kara instead of the goal marker -ni in a significant number of cases, and not the other way around. This behaviour, they argue, is in line with certain linguistic phenomena relating to source and goal markers: source markers are marked in comparison to goal markers and source constructions can, diachronically, develop into goal constructions. Both the linguistic phenomena and the agrammatic data can be explained by the cognitive action chain model, in which the goal is in line with the energy flow that originates in an actor, whereas a source conceptualizes a point that is opposite to the direction of the action chain.

Kochaska, dealing with verbal aspect in Polish, restricts herself to accounting for the acceptability and possible interpretations of coordinated VP constructions involving perfective/imperfective VPs, seen from a temporally extended or a temporally momentary Viewing Frame. On the basis of her findings, she argues, with Langacker (1991, inter alia), that the notions of temporal boundedness/unboundedness and internal heterogeneity/homogeneity should be kept distinct; both, she argues, pace Verkuyl (1993), are relevant for verbal aspect. Her general conclusions are that Polish perfective and imperfective VPs are both polysemous, that the meaning of coordinated VPs is constructional rather than purely compositional, and that notions like boundedness are a matter of construal rather than constituting 'objective' facets of the world: an event is, for instance, bounded if we (can) think of it as such, not because it is bounded regardless of how one looks at the situation. An important consequence of this latter viewpoint is that only a truly subjectivist analysis of aspectual phenomena can, according to Kochaska, deal with the facts in a revealing manner.

Mandelblit & Fauconnier argue that a generic conceptual causative schema underlies not only grammatical causative constructions, but passive, middle and reflexive constructions as well. They analyze, to concentrate on the first type, causative syntactic constructions in English, Modern Hebrew (hif'il) and French. In their view, a sentence is the result of blending a conceived event with a schematically meaningful syntactic construction. Because not every element of the conceived event is projected on the construction, possible ambiguities may arise, for example between middle and passive constructions in Modern Hebrew. Because, furthermore, mappings from the conceived event onto the meaningful syntax may differ in systematic ways, the blending

EDITORS' FOREWORD Xll l

mechanism can also account for the difference in mapping responsible for the meanings of Rachel sneezed the napkin off the table on the one hand (here the construction's subject referent is understood to do the sneezing), and of She trotted the horse into the stable on the other (here the trotting is understood as done by the object referent). Language being to a large extent under specified, grammatical blending can thus account for various kinds of constructional ambiguity.

Mortelmans shows, on the basis of original data, that the Konjunktiv II mood in German on its own signals negative epistemic stance (irrealis), that epistemic modals in the Konjunktiv II form, e.g. knnte (cf. Indicative kann), still signal positive epistemic stance (potentiality), be it in a weakened form. However, in conditionals with Konjunktiv II functioning as grounding predication in the protasis, the effect on e.g. knnte is that it now can only express negative epistemic stance (irrealis). Interestingly, the form drfte has gramma-ticalized in a different direction, expressing a highly subjective epistemic evaluation; it is therefore on the whole considered unacceptable if Konjunktiv II has a dominant grounding function. Her conclusion is, therefore, that mood and modals exhibit what she aptly calls 'division of labour'.

Nikiforidou & Katis show, on the basis of authentic material, that the traditional claim that the conditional markers ama, ean and na are free variants of the basic conditional marker an is wrong, and that the three markers under discussion code extra semantic dimensions, in that each involves, in its own way, a "subjective construal ... of the situation depicted in the protasis" (section 5; 'subjective' in Langacker's (1987) sense of 'egocentric construal'). The authors show that the three markers display semantic extensions as well, extensions which seem to point more in the direction of Traugott's (1995) notion of subjectivity by marking highly personal facets like empathy or involvement, thus providing motivated links from the basic egocentric semantics of the markers in question.

As Takahashi points out, imperatives do not, normally speaking, occur in the passive voice in English, cf. *Be helped by Jill; however, there are felicitous examples of passive imperatives as well, cf. Be flattered by what he says, it'll make his day. Takahashi shows that it is conceptual incompatibility that leads to the typical unacceptability of passive imperatives. The (second person) subject of an imperative is, given a prototypical Imperative Event Model, an agent, whereas the subject of a passive, given a prototypical Passive Model, is a patient. Only when used in non-prototypical ways, can the two constructions become compatible. It is this cognitive viewpoint that enables Takahashi, moreover, to explain why in Japanese passive imperatives are hardly (if at all) acceptable: unlike English, the meaning of the Japanese

XlV EDITORS' FOREWORD

imperative is restricted to the prototypical Imperative Event Model, hence passive and imperative constructions are bound to clash in Japanese.

Thepkanjana, analysing the semantics of lexical causative verbs in Thai, makes a major distinction between alternating and non-alternating lexical causatives, and a further subdivision into (a) different types of causatives, (b) agent-oriented causatives and (c) patient-oriented causatives. On the basis of these distinctions, and with the help of Talmy's notion of 'windowing of attention', she neatly accounts for the kind of constructions the verbs in question can, or cannot, occur in, while also making it clear that the notion 'accidental lexical gap' is a misnomer as far as suppletive lexical causatives in Thai are concerned: their distribution is by no means random.

Tsuboi deals with the adversative passive construction in Japanese, a construction which involves the quite fascinating phenomenon of valence increase: particular verb types, which are intransitive when used actively, can passivize provided an extra argument, marked by ablative -ni, is added to the passive construction. Earlier analyses essentially account for this valence increase in random terms: the construction requires an extra-thematic NP. Tsuboi, on the other hand, argues that the adversative passive is motivated by a particular Japanese cultural cognitive model that involves a person's 'sense of responsibility'. This models a person A responsible for something bad happening to a person B, provided A could have prevented this from happening; A is thus seen as having affected adversatively (regardless whether this was intentional or not). The meaning of the passive adversative can only be understood against this ICM, with the --marked actor mapping onto person A, provided, moreover, that the passive subject referent (the focus of attention in the passive) is actually annoyed at the actor in question. This model of responsibility ascription shows itself not only in the passive construction, but also in some active contexts that also involve valence increase. The adversative passive is, thus, not an isolated phenomenon in Japanese, as previous analyses suggested.

Van der Leek, lastly, argues against the analysis of the English caused-motion construction as proposed by Goldberg (1995), who claims, essentially, that this is a construction in the Construction Grammar sense (cf. Fillmore & Kay to appear), in that its syntax prototypically pairs up with a caused-motion sense, and with extended senses in nonprototypical uses of the construction; thanks to this claim, she can argue that verbs have basically only one meaning, and that a verb is licensed in the construction provided its basic meaning is compatible with one of the construction's conventional senses. Van der Leek argues instead that, where mature speakers are concerned, the pure syntax of the construction makes no semantic contribution of its own, and that the

EDITORS' FOREWORD XV

caused-motion pattern (or any of its extensions) gets extracted by the language user as a superordinate conceptual category, a category that, indeed, licenses particular usages without, however, lending them their meaning. She further argues that pure syntactic configurations have no semantic value prior to conceptualization, even though children begin by assuming that there is a direct mapping between their prelinguistic conceptualizations of caused-motion and the meaning of the syntactic configuration in question.

References Fillmore, & P. Kay. To appear. Construction Grammar. Stanford: Center

for the Study of Language and Information. Goldberg, A. 1995. A Construction Grammar Approach to Argument Structure.

Chicago: The University of Chicago Press. Langacker, R. 1987. Foundations of Cognitive Grammar. Vol. I. Theoretical

Perspectives. Stanford: Stanford University Press. 1991. Foundations of Cognitive Grammar. Vol. 2. Descriptive

Application. Stanford: Stanford University Press. Traugott, E. 1995. "Subjectification in Grammaticalization". Subjectivity and

Subjectivization in Language eds. D. Stein & S. Wright, 31-54. Cambridge: Cambridge University Press.

Verkuyl, H. 1993. A theory of aspectuality. Cambridge: Cambridge University Press.

Pragmatic Conditionals

Angeliki Athanasiadou University of Thessaloniki, Greece

Ren Dirven University of Duisburg, Germany

1. Introduction: Types of conditionals * In previous papers on conditionals, we have distinguished between

three major types of conditionals. The first type is the traditionally perceived 'hypothetical conditional' of the type If it rains tomorrow, we'll stay at home, in which the consequent fully depends on the realisation of the antecedent. A second type has largely remained unobserved in English grammars and in the recent literature,1 though it has a very high statistical occurrence in the corpora excerpted.2 It is marked by the absence of modal forms in the consequent, but describes regularly co-occurring sequences of events as in If there is a drought, the eggs remain dormant. Such a sequence of events is somehow also causally related, but not construed as such, i.e. the two events are not construed as simply triggering each other but just as co-occurring. Therefore we have labelled it as 'course-of-events conditional'. The third type is the 'pragmatic conditional' as in If you are hungry, we are having dinner at six, which Fauconnier (1994:121) analyses in terms of a hidden conditional of the type If you are hungry, join us at six. But the real complexity of this pragmatic type of conditional is much greater and we will devote this whole paper to this subtype of a pragmatic conditional amongst various other members of this hybrid category.

Before beginning the analysis of pragmatic conditionals proper we want to point out that both course-of-events conditionals and pragmatic conditionals, contain their own subclass of 'inferencing' conditionals. This raises the question of the fundamental difference between the conceptual domains of course-of-eventness and of pragmaticity. This problem can be best approached by looking into some distinctions that can be made between these two types of conditionals: inferencing course-of-events conditionals as in (1) contrast in various respects with inferencing pragmatic conditionals as in (2):

(1) He looked at his watch; if the soldier was coming, it was nearly time. [COB:243]

(2) If she's divorced, (then) she's been married. [Sweetser 1990:116]

2 ANGELIKI ATHANASIADOU & RENE DIRVEN

The main difference between the two inferencing processes is a question of the nature of the evidence: in (1) the inferencing process is about a really occurring situation, whereas in (2) the inferencing is about the explicitation of a given concept; the mental process in (2) need not involve any actual situation, but may be a purely logical operation in the mind.

In the course-of-events conditional of (1), the course-of-events character follows from the description of an actually recurring event, whereby the soldier's coming is always so punctual that the observer automatically knows what time it is and what to do next; he just checks the time of the situation by looking at his watch. The inferencing process is embedded in an actually occurring situation, so that the inference is based on an observational fact which is linked to earlier occurrences of this situation and consequently linked with a second situation which tends to co-occur with the first. In the pragmatic conditional of (2), there is no observation of situations, but only a logical operation of spelling out entailments of a given statement or concept; the notion of 'being divorced' implies a previous state of 'having been married before'. Here the consecutive aspect of the two events is adequately expressed by the originally temporal, i.e. consecutive use of the adverb then. Historically, this temporal adverb has been grammaticalized as a logical operator, associating temporal succession with logical inference. In (2) the temporal meaning is no longer present, but only the logical 'inference' meaning is given. The two inferencing processes are thus of a different nature: in an observational inference' we make simultaneous ad hoc inferences, applicable to the situational context; it is the logic of simultaneous events. But in a purely 'logical' inference, we make generic inferences, based on the meanings of consecutive events found in our conceptual world; it is the logic of concepts.

But why can the logical inferencing process, as illustrated in (2), be called a 'pragmatic' conditional? This question only arises if one takes the term pragmatic in a narrow sense. Very often the term pragmatic is used only in relation to speech acts, but its original meaning, as specified by Charles Morris (1946), is much wider. Morris made the distinction between syntactics, semantics, and pragmatics, such that each level involves different relations: syntactics specifies the relations between the signs themselves, semantics specifies the relations between signs and the world, and pragmatics specifies the relations between signs and their users. It is in this sense that we want to use the term 'pragmatic' for our present purposes. When speaking of 'pragmatic conditionals', we thus suggest that there is always a special emphasis on the presence of the user of the signs. This is

PRAGMATIC CONDITIONALS

most emphatically felt in 'interpersonal' speech acts,3 such as requests or offers, but it can also be present in speech acts of the ideational type such as inferencing processes of the logical type. Indeed, here the logical inferencing process itself is often stressed by the use of a number of the logicalinference markers, a very frequent one being then.

In an interpersonal speech-act situation the pragmatic aspect is prevalent because the speaker himself as a user is necessarily present, either explicitly or implicitly. The speaker, when using a pragmatic conditional in a discourse situation, performs a speech act, and the if-clause explicates one of the preparatory conditions, i.e. the speaker suggests to the hearer that he believes that the state of affairs in the consequent may become true. Consider sentence (3) below taken from Athanasiadou and Dirven (1996):

(3) If anyone wants me, I'm downstairs.

This issue has led to a great diversity of opinions concerning the categorisation of types of conditionals. On the one hand, we have those such as Sweetser (1990:123), who maintain that logical conditionals as in (2) and conversational conditionals as in (3) are strictly separate categories;4 on the other hand, we have those such as Comrie (1986:81), who group logical and conversational conditionals together. By establishing the superordinate category of pragmatic conditionals we adopt the latter position. Our motivation for doing so is based on the amount of dependency that holds in various degrees for the three main types of conditionals. The causal dependency between antecedent and consequent is absolute in hypothetical conditionals; causal dependency decreases considerably, but remains implicitly given in course-of-events conditionals; and dependency is reduced to a purely logical, i.e. non-causal, relationship in pragmatic conditionals of the inferencing type as found in (2), and to a merely conversational point of relevance in pragmatic conditionals of the discourse type found in (3). But the dependency relation is never totally absent.

The categories and subcategories we propose for discussion in this paper are summarised in Table 1.

This table reflects a taxonomic hierarchy, consisting of the superordinate category 'pragmatic conditional', the two basic-level categories 'logical' and 'conversational' conditionals (where the term 'basic level' is taken in a wider sense than usual), and the four subordinate categories or subcategories.

4 ANGELIKI ATHANASIADOU & REN DIRVEN

Table 1. Categories and subcategories of pragmatic conditionals (PCs)

The paper, then, deals with each of these categories in their turn: - General characteristics of pragmatic conditionals, - Identifying conditionals, - Inferencing conditionals, - Discourse conditionals, - Metacommunicative conditionals.

2. General characteristics of pragmatic conditionals (PCs) As Table 1 shows, pragmatic conditionals comprise four main types:

two of them involve logical processes, i.e. the 'identifying' PC as illustrated in (4) below, and the 'inferencing' PC as depicted in sentence (5). The two other types involve conversational contexts, and comprise the 'discourse' PC as in (6) and the 'metacommunicative' PC as in (7).

(4) If there's one human species that ought to be put out to pasture, it's Presidents and Prime Ministers. [COB: 136]

(5) If the super-organism created by a colony of termites can be compared to an antelope, then the disciplined aggressive columns of the army ants must be reckoned to be the insect equivalent of a beast of prey. [COB:268]

(6) What about the parents demonstrating, if there are no friends? [COB:84]

(7) I've come to offer my congratulations, if that's the right word. [LDC:42]

Before characterising the internal structure of each of these four categories, we must first address more general questions which are relevant

PRAGMATIC CONDITIONALS 5

to the present discussion. First of all, these four types of conditionals belong to two basic-level categories, namely logical conditionals (examples (4) and (5)) and conversational conditionals (examples (6) and (7)). Whereas the logical conditionals involve analytic reasoning processes, and as a result the antecedent can only be preposed to the consequent, conversational conditionals involve speech acts in actual discourse (6) or aspects of the discourse such as metalinguistic references to the linguistic choices made by the speakers (7) and the antecedent tends to be postposed.

Secondly, the separate category of a logical inferencing conditional is not an additional complication or multiplication of categories, since we need the category of 'logical conditionals' anyhow because of the existence of a logical identifying conditional. We can therefore make a distinction between two types of inferencing conditionals, viz. inferencing course-of-events conditionals and inferencing pragmatic conditionals. The latter is much closer to the 'identifying' pragmatic conditional than to the inferencing course-of-events conditional, because both pragmatic logical conditionals express merely reasoning operations, but do not involve any actual observed reality.

Thirdly, the diversity of subcategories within the two basic-level categories of logical and conversational conditionals and the conceptual difference between these two basic-level categories themselves calls for the introduction of a unifying superordinate category, which groups these subcategories together in opposition to the two other superordinate categories of course-of-events and hypothetical conditionals. The term 'pragmatic conditional' is thus employed to refer to a common conceptual link inherent in all of the four subcategories: as will be shown in the next sections, they are all, each in its own way, more strongly either 'speaker-oriented' or 'heareroriented' and thus are apt to serve the interpersonal function of language. In this respect, pragmatic conditionals contrast very strongly with the primordially ideational functions associated with the use of course-of-events and hypothetical conditionals.

Having established these internal commonalities and distinctions in the area of pragmatic conditionals, we will now try to specify the conceptual link between the degree of independency of the antecedent and the consequent, on the one hand, and the functions of the four types of pragmatic conditionals.

How great is the degree of independency between antecedent and consequent in logical pragmatic conditionals? How much smaller is it in comparison with hypothetical conditionals and course-of-events conditionals? As shown in Athanasiadou and Dirven (1997), hypothetical conditionals are


the central or prototypical member of the class of three types since they show the highest level of dependency between antecedent and consequent. In course-of-events conditionals the dependency is still present but less pronounced since the two simultaneous or consecutive events are only seen to co-occur but they are not explicitly linked in any causally or otherwise determined way. Let us examine somewhat more closely the degree of dependency implied in sentence (1) and contrast it to the logical pragmatic conditional in example (5), both of which are repeated here for the sake of convenience:

(1) He looked at his watch; if the soldier was coming, it was nearly time.

(5) If termites are the equivalent of antelopes, army ants are the equivalent of beasts of prey.

The dependency between the two events in (1), viz. the soldier coming and it nearly being time is based on recurring experience, that is to say, on observation of regularly, consecutively occurring events. The dependency between the two events in an inferencing course-of-events conditional like (1) is greater because the construal of this dependency is based on repeatedly observed facts. In an inferencing pragmatic conditional like (5), the dependency between antecedent and consequent is much weaker because the link is based upon a construal of analytic definitions: termites are only comparable to non-carnivorous antelopes from a very special, namely nutritional point of view, and on the basis of that specific parallel we infer a similar comparability between army ants and carnivorous beasts of prey. It is even the case that in (5) the antecedent and the consequent are theoretical constructions in the mind which in themselves have no validity and thus cannot be dependent on one another. In its simplest form, such constructions can be summarised in the formula: if a is b, then is d. It is the parallel between a=b and c=d created by the logical inferencing if-construction that makes some slight form of dependency plausible.

In the scale of dependency of antecedent and consequent as already shown in Athanasiadou and Dirven (1996, 1997) for hypothetical and course-of-events conditionals we can see the following global picture of decreasing dependency for all the types of conditionals:


Table 2. Scale of (in) dependency between antecedent and consequent in conditionals

3. Identifying conditionals The above scale of (in)dependency of conditionals in Table 2 implies

that identifying pragmatic conditionals show a relatively greater dependency degree than inferencing pragmatic conditionals. This has not been demonstrated thus far, so we will first return to example (4) again:

(4) If there's one human species that ought to be put out to pasture, it's Presidents and Prime Ministers.

Such an identifying conditional consists of two parts: an antecedent which functions as an identifying description and a consequent which is the identified one or the identifier. The matching of the entity to be identified and the identifier depends to a large extent on the detailed description in the antecedent. Although there is no causal relationship between the antecedent and the consequent, the two parts are strongly interrelated: the description in the antecedent is made to fit the identifier in the consequent.

But in actual fact, things are more complicated. Although identifying pragmatic conditionals seem to function in order to identify the unknown or


as yet unidentified member of a category, this may be a very misleading interpretation of their communicative function. What the speaker is doing by means of the identifying construction is not only just to accomplish an act of revealing an entity's identity, but rather to create expectations, which is done by a very long description of the entity to be identified:

(8) Now, if any part of the Bible is assuredly the very Word of God speaking through His servant, it is John's Gospel. [BRO:91]

Thus, the identifying pragmatic conditional assumes a strong rhetorical function emphasising the important features of the category to be identified and raising the expectation for the identity of the variable of this category precisely by postponing the naming of it. Preposing of the consequent would therefore not make any sense in identifying pragmatic conditionals.

Our thesis that the main function of the identifying pragmatic conditional is not necessarily that of revealing someone's identity can also be shown by the fact that the construction can be used with a negation thus rather excluding possible candidates than carrying through an identification process:

(9) Luckily both women knew my position and if anyone suffered in their opinion it was not I. [BRO: 104]

The rhetorical use of the identifying pragmatic conditional can even be strengthened by the use of adverbs like ever in the antecedent and the reduction of the consequent to a very short formula, as in the following examples:

(10) If ever a rifle met the needs of the whitetail hunter, this is it. [BRO: 189]

(11) Brains and beauty, high position in both the social and intellectual worlds, athlete, fabled lover if ever the world was any man's oyster it was his. [BRO: 191]

As with most pragmatic phenomena in language, here too we are left again with the rather predictable conclusion that there is no one to one relationship between the syntactic form of this construction and its various pragmatic uses. The syntactic form typically serves an identifying function, but the pragmatic use overrides this syntactic function, as is usually the case


in the pragmatic use of language. This conclusion contains another argument for categorising identifying pragmatic conditionals as a member of the category of pragmatic conditionals and not as a separate category.

Whereas the examples (8) to (11) contain cases of subject identification, the following examples show that in fact any constituent of the sentence of the antecedent may be selected for the identification process. The entity to be identified may be the topic of the sentence (12), the location (13), or the reason why something is the case (14):

(12) If there was one thing Julie couldn't be doing with, it was anonymity. [COB:20]

(13) If there's trouble in a home, it isn't always in the bedroom - it's quite often in the budget. [COB: 19]

(14) If, indeed, we are inclined to deny one particular thing, this is because of the failure to recognize some other aspect of it. [LOB:3, adapted]

Sentence (12) illustrates a colloquial and popular means of using identifying conditionals: it is purely emotional.

Sentence (13), does not seem to provide an identification of what the main marriage trouble might be, but rather an identification of where the trouble might have to be situated, whereby the locatives in the bedroom and in the budget are metonymically used for the sexual aspect of marriage and for its financial management, respectively.

In sentence (14), the identification process is related to a subclause denoting reason. In the identification formula this is because, the pronoun this indeed refers to the whole of the antecedent. So the pragmatic type of identifying conditional in (14) follows the pattern 'if a, then because of b'.

From a cognitive point of view, however, examples (12-14) represent a borderline case of identifying conditionals in the sense that they form a transition from the process of subject identification to a process of emphasising a particular constituent in the sentence. Still in all the examples discussed in this section, the dependency between the antecedent and the consequent is manifestly, though weakly present.

4. Inferencing conditionals Identifying conditionals are used for other communicative purposes

than mere identification. Similarly, inferencing pragmatic conditionals are not only used to make inferences, but also and perhaps even more


importantly, to emphasise the inferencing force of the utterance, which again foregrounds the pragmatic character of the construction. One of the many items emphasising the 'inferencing' force of an inferencing conditional is the adverb then. In origin a temporal adverb, the grammaticised logical then still preserves a strong deictic link with some previous state of affairs. This presupposes an interdependency between antecedent and consequent. If the interdependency is not strong enough, then is ruled out. Thus, it cannot be used in discourse conditionals such as (3). Hence we cannot say *I anyone wants me, then I'm downstairs. In this respect it is worth noting that there is a very strong contrast between inferencing pragmatic conditionals and inferencing course-of-events conditionals. In inferencing pragmatic conditionals the use of then may sometimes be absolutely indispensable as in the following example:

(15) If Arthur Williams was involved in the fraud or the murder, then he too had another identity. [BRO: 114]

This use of the logical operator then can be further clarified by its compatibility with the epistemic auxiliary must, i.e. 'then he too must have had another identity'. Without then it is much more difficult to interpret this inferencing pragmatic conditional since we would not be inclined to stress the pronoun form he and make it refer back to Arthur Williams. In inferencing course-of-events conditionals, on the other hand, such as the soldier example in (1), the inference is drawn without any further focus on the inferencing process as such.5 That may also be the reason why inferencing conditionals with then and without then do not always have the same implications. Let us, for instance, repeat the soldier example without and with then for the sake of easy comparison:

(16) a. He looked at his watch; if the soldier was coming, it was nearly time. b. He looked at his watch; if the soldier was coming, then it was nearly time.

In the course-of-events context of (16a) the conclusion follows from observation: the two events in the antecedent and the consequent are contiguous, in the sense that the one stands metonymically for the other. That's why both are factual: the soldier was coming and this was the sign that it was nearly time. The subject checks the time on his watch to confirm his factual inference. So both the antecedent and the consequent are based


on observational deduction. Here the observational conclusion seems to exclude the use of the logical operator then, which requires a fully different conceptual scene. The speaker of (16b) would not even need to see the soldier coming. Moreover, he could also use the form with must, i.e. then it must nearly be time. Consequently, in (16b) the soldier's coming is much less certain because it is no longer based on an observation of a course of events, but on a logical, truth-finding deduction. Moreover, (16b) now looks like an event only occurring once, whereas (16a) clearly implies a recurring event, which is precisely the basis for the inference. The use of then in (16b) changes the nature of an inferencing course-of-events conditional into that of a logical conditional. The possibility of using then is merely the tip of the iceberg in that it is but one of a whole range of epistemic, inferencing expressions, all of which stress the epistemic or truth-finding nature of the inferencing process and as such stress the pragmatic relationship between 'the sign' and user'.

These epistemic expressions in inferencing conditionals may range from epistemic modal auxiliaries like may, must, should, have to, and paraphrasing constructions for modality such as it is possible, surely, it seems likely, etc. to explicit lexical expressions denoting the truth-finding process, such as (we) conclude, it follows that, is it any wonder, it stands to reason, you must admit, or negative conclusions like it does not follow, it is not necessary, it is strange, I see no reason, etc. Even a certain rhetorical flavour may be present occasionally, especially through the use of rhetorical questions introduced by why shouldn't, why not, how much less, etc.

Of all these possibilities the preferred devices are modal auxiliaries or modal paraphrases:

(17) On the other hand, if there have been no signs of active infection for some time, the murmur may be due to old scars left over from a previous attack. [COB: 106]

(18) If this is so, why is it not also possible that they achieved their original dominance by stealth and cunning and bloodshed.... and not, as they claim, by Heaven's will? [COB:336]

These modal auxiliaries are, more often than not, combined with other inference-focusing devices such as I think it should, should+progressive, there may be some reason, then...should, etc.


(19) If arranging for the girls must take time into account, I think a day or two should be enough to finish our business. [BRO: 112]

(20) If, as Reid says, nearly all his poetry was produced when he was not taking opium, there may be some reason to doubt that he was under its influence in the period from 1896 to 1900. [BRO: 119]

The inference-focusing devices can be even more elaborate and contain a certain degree of redundancy:

(21) Nor is it necessary to hold that if a man wants to get food, he must be tempted to steal. [LOB:4]

(22) It stands to reason that if a horse is too backward to race during his first season in training, he is most unlikely to be sufficiently mature to beat the best of his generation in the late May or early June. [LOB:81]

The phrases in (21-22) are characteristic of the jargon of logic. In more colloquial language the inference-focusing devices tend to be less elaborate, as the following expressions from the Leuven Drama Corpus show. The expressions found here are: I'd say, is it any wonder, I see no reason, it follows, etc.

(23) If I saw a man eating grass I'd say he was hungry. [LDC:31] (24) If he fails to see me in a room, is it any wonder he has difficulty

finding the enemy in the whole of Flanders. [LDC:70] (25) If the sun never sets on your empires, it follows that it never

rises either. [LDC:109]

It turns out that not only the logical adverb then, but several other colloquial devices such as so, or after all, or whole phrases like you must admit, etc. are devices emphasising the act of inferencing.

(26) So there is a God after all, if it is reason. [LDC:93] (27) If you persist in maintaining the almighty Jove aimed a

thunderbolt at me, then, you must admit he had the unusual experience of missing. [LDC:102]


But most clearly, the rhetorical force of an inference-focusing construction is found in an interrogative conditional

(28) If birds can fly, why shouldn't man? [LDC:65] (29) If he is a man, how much less of a man am I? [LDC:72]

The analysis of the many examples in this section was necessary to pave the way to one important conclusion. The linguistic realisation of the process of inferencing includes more than mere inferencing, that is, drawing all sorts of logical consequences. It also includes an awareness of the process of inferencing itself. Indeed, the logical inference type of conditional, as in example (2) If she 's divorced, (then) she 's been married, is rather the exception, not the rule, since it only suggests a minimum of focusing. The rule is rather that the speaker uses numerous inference-focusing devices to refer to the ongoing process of inferencing. In this respect, the speaker is signalling very explicitly that he is engaging upon making the inference and thus these inference-focusing devices are a pragmatic strategy to signal the pragmatic force of the utterance. Of course, such pragmatic self-monitoring is not to be confused with performative acts, but these devices reveal that speakers' interests are far beyond making purely analytic inferences.

The many inference-focusing devices also invite us to use the term 'pragmatic conditionals' in a much wider sense than that of conditional speech-acts, as found in the literature, (e.g. Sweetser 1990:118). Whereas traditionally the concept 'pragmatic' was limited in its application to performing speech acts, presuppositions, hedging, etc., we now propose to use it again in the wider, original as intended by Morris sense of the relationship between the sign and its user, so that it does not only cover the things we do, but also the things we think and say, while performing speech acts or any other communicative acts. Moreover it comprises various sorts of hints at the thought processes behind our speech acts, at the special appeals to the hearer or at comments on the speaker's choice of certain wordings.

5. Discourse conditionals Discourse conditionals link the consequent to some or other, usually

hearer-oriented, pragmatic factor in the conversation. The split between the antecedent and the consequent is extremely great. In English, this split cannot show up in different word orders in antecedent and consequent, as in German and Dutch, but only in the intonation pattern. Thus, discourse con-


(i) no possibility of using then (ii) no hypothetical forms (iii) no change of tenses (see Sweetser 1990; and see (31) below) (iv) intonational 'caesura' (v) no explicit use of performatives (vi) very great ambiguity of possible speech-act forces (see (32) below) (vii) preferred order: antecedent before consequent

Table 3. Characteristics of prototypical discourse conditionals

ditionals show a very important 'caesura' between the antecedent and the consequent. Intonationally, this 'caesura' is shown by the fact that one cannot read or pronounce a discourse conditional (30a) in the same way as a hypothetical conditional as in (30b) below:

(30) a. If you're hungry, there's food in the fridge. b. If you ask him kindly, John will help you.

If one reads the hypothetical conditional in (30b) with a 'caesura' or an intonational break, an odd effect is achieved, but in (30a) it is quite normal. Conversely, (30b) must be read as one intonational unit but if one applies this intonational pattern in the discourse conditional in (30a) the result is very odd.

The discourse conditional is much more a construction sui generis, i.e. a highly idiosyncratic category with very specific characteristics of its own, summarised in Table 3, which can also be seen as a summarising definition of the prototypical discourse conditional.

(31) ?If you were hungry, there was food in the fridge. (32) If anyone wants me, I'm downstairs.

The question mark in front of the sentence in (31) means that this sentence cannot be given the interpretation of a discourse conditional. But in itself the sentence is not impossible; on the contrary, it is a very clear instance of a course-of-events conditional which is to be interpreted as 'Whenever you were hungry, there was always food in the fridge'. In (32) the ambiguity of the possible conversational implicature of the speech act is very great. Possible candidates are 'please tell X', 'don't tell X', 'I don't want to be disturbed', 'they can find me', etc.


5.1. The great variety of discourse conditionals Most discourse conditionals exhibit the following central or

prototypical features of their category: they have a very clear 'caesura', they use present tenses in both antecedent and consequent, they are of the affirmative type and they have the sequence antecedent-consequent. However, in corpus-based examples of real language use we find many other structural possibilities with discourse conditionals, which differ in some interesting respects from the prototypical ones discussed before.

(i) The first discourse conditional ever pointed out is found in Austin's paper "Ifs and cans" (1961:210-12) and its structure has the reversed order of a prototypical discourse conditional, i.e. consequent before antecedent:

(33) There are biscuits on the sideboard if you want them.

Not only is there a reversed order of consequent and antecedent in this example, but there is also absence of a strong 'caesura'. Here in the marked order of consequent and antecedent, this 'caesura' can almost be reduced to zero. This may be due to the fact that with a preposed consequent, as in (33), there is no communicative need for an intonational break, since the //-clause cannot be misunderstood now. But note that here, too, at the end of the antecedent, there is a slightly rising intonation contour, typical of questions or offers. What is even more striking is that instead of a postposed antecedent, one can use an intonation question or any other type of question as in (34):

(34) There are biscuits on the sideboard. You want some?

This possibility of using two independent sentences instead of a discourse conditional is a very strong indication for the very low degree of dependency between the antecedent and consequent in this type of pragmatic conditional.

(ii) A second structural variant found with the discourse conditional is the interrogative form, as in the following example:

(35) If it is found that in recent years the habit of discussing with the chair has increased, may I ask you not to hesitate to say so? [LOB: 131]


The hearer-oriented character of the discourse conditional triggers the use not only of you forms, but also of invitations, requests, etc.

(iii) Another pragmatic form frequently and commonly used to express a discourse conditional is a why not-question in the main clause, as in (36):

(36) a. If he attaches little importance to personal liberty, why not make this known to the world? [BRO:243] b. If anti-Semitism was on trial in Jerusalem, why was it not identified? [BRO:78]

The negated why-question with infinitive in (36a) has an urging force and comes close to a challenge to do something about some negative discoveries. But in (36b) the same negated wh-question is not a challenge to do something about a situation but, since it is in the past tense, it is rather ambiguous between a neutral information question and a reproach on moral grounds. Here the position of the negation adverb not is very significant: it can remain unstressed and then it is an information question, but when it is stressed, it becomes a reproach. But if it is used as a contracted form with the auxiliary (why wasn't it identified?), the contracted form rather suggests the reproach meaning, i.e. 'it should have been identified', as the preferred interpretation.

(iv) An equally common structural realisation of the discourse conditional is an imperative form in the consequent, which is not only postposed as in (37a), but probably more often preposed, as in (37b,c).

(37) a. If it is hallucinations you are after, get drunk. [LDC:96] b. Keep them, if you feel so lost. [LDC:52] c. Prove it if you can.

Such an imperative form in discourse conditionals may express many different speech-act forces, such as letting someone down or cursing (get lost, go to hell), typically found in postposed position (37a), and acts of reassuring (37b), challenging (37c), etc. typically found in preposed positions.

(v) Other structural possibilities are offered by the combinations of different tenses in the antecedent and the consequent. In addition to the prototypical pattern 'present + present' in antecedent and consequent there


are also the unexpected combinations of 'past + present', 'past + past', 'past + past perfect', etc. Most of these tense combinations are motivated by the different times referred to in the antecedent and in the consequent:

(38) Castle thought: that's a bad slip if the telephone call this morning was from the office. [COB:275]

Here, the different use of tenses (present in the consequent + past in the antecedent) is motivated by the different moments when the evaluation occurs (consequent) and when the event evaluated took place (antecedent).

Since speech acts are grounded in the present time of the speech act itself, acts of offering, inviting, challenging, evaluating tend to occur in the present tense, and they do not normally occur with past tense forms referring to past time, as was already shown in the example (31) ?If you were hungry, there was food in the fridge.

6 Metacommunicative conditionals

The function of the if-clause in a discourse conditional is heareroriented: it is mainly to point out under what circumstances the speech-act force expressed in the antecedent is relevant. In this respect a metacommunicative conditional is by its very nature more speaker-oriented and thus differs markedly from discourse conditionals:

(39) a. ...and if I might say so at this stage, that is one of the sentences it may impose on a young male in prison.

[Haegeman 1984:487] b. If I may say so, that's a crazy idea. [Sweetser 1990:118]

Here the link between the antecedent and the consequent is a preparatory link: the if-clause has the function of softening the possible unpleasantness of the statement of the consequent. Besides softening there may be many other metacommunicative functions of pragmatic if-clauses, a very frequent case being that of relativising one's presuppositions in the making of the speech act. Therefore we can define a metacommunicative conditional as the pointing out of some aspects of the global communicative act which need special attention. Consequently, these /-clauses can be seen as comments on various aspects of the communicative act and are therefore metacommunicative in nature.


6.1. Various uses of metacommunicative conditionals Since there are so many different types of communicative acts, of

which speech acts are just one, though very important, subcategory, it may be advisable to make further distinctions between a number of metacommunicative functions which may be fulfilled by conditionals. These functional uses may relate either to the whole of the speech act or to aspects of it. If the whole speech act is concerned we can speak of the metapragmatic use of the conditional. If separate aspects are involved, we can distinguish between a metalinguistic or a restrictive use of the conditional; the //-clause relates either to the 'wording' of the speech act, i.e. the locution used by the speaker or to the presuppositions made by the speaker. In the former case we have a metalinguistic use as this term is traditionally understood in linguistics and in the latter we usually have a 'restrictive' use. In this section we therefore address the following three functions of metacommunicative conditionals:

(i) The metapragmatic use of metacommunicative conditionals, (ii) The metalinguistic use of metacommunicative conditionals, (iii) The restrictive use of metacommunicative conditionals.

(i) The 'metapragmatic * use of metacommunicative conditionals Alongside the stereotypical formula if I may say so in (39b), we find

many other metapragmatic references to the appropriateness or inappropriateness of an utterance in a given social relationship. Thus, the social distance may be too great for the speaker in a lower hierarchical position to perform a blunt speech act of evaluation, and therefore he must 'preface' his evaluation by means of a softening metapragmatic cajoler, as in (40):

(40) Good gracious me, sir, if I may make so bold - it's a bit shocking, isn't it? [LDC:86]

This comment in the if-clause is, in fact, just as in the two examples in (39), a parenthetical construction, which can as easily be left out as not, and which therefore further emphasises the fact that the metacommunicative function is not an essential part of the utterance. This does not mean of course that parenthetical constructions are always optional, e.g. the following metacommunicative conditional, which has the form of a parenthetical construction if you'll excuse us could not be omitted because this would produce an awkward, if not impossible, utterance:


(41) a. Splendid! Now ladies, if you'll excuse us, we have a lot to do. [LDC:54]

b. ** Splendid! Now ladies, we have a lot to do.

Here the if-clause in fact expresses the speech act of asking for the permission to leave and the consequent gives the reason for this.

The metacommunicative conditional may also be used metapragmatically to refer to the communicative intention of the former speaker, as in (42a), or to challenge the hearer to use the correct procedure to find information (42b):

(42) a. I've always been interested in people, sir, if that's what you mean. [LDC:107] b. It's all on the back pages of the paper, just before the sport, if you know where to look. [LDC:16]

(ii) The 'metalinguistic ' use of metacommunicative conditionals Whereas in the above cases the metacommunicative conditional relates

to the speech act as a whole, in the following cases it has a more limited function, and relates only to particular aspects of the speech act, such as to its formal aspects, or more importantly, its functional aspects. Speakers of a language are aware of performing a large number of metalinguistic acts. But, in fact, very few examples occur in the selected corpora:

(43) a. I've come to offer my congratulations, if thas the right word. [LDC:42] b. He was beginning to feel sure that she had recognised him

and wanted to talk but could not summon up the courage - if courage was the word. [LOB:180]

As these examples illustrate, the metalinguistic reference always concerns a single word or expression, and therefore the metalinguistic reference is made by means of an anaphoric pronoun such as that (43a), or by the full word or expression courage, as in (43b). This formal device also shows the enormous difference between the metapragmatic use (42) of the metacommunicative conditional, which can have any possible structural form, and its metalinguistic use which can only have a nominal form such as that in (43), because it must refer to some element in the consequent.


(iii) The 'restrictive' use of metacommunicative conditionals When uttering speech acts the existence of (some of) the referents of

the noun phrases used, or of the events described are presupposed. Metacommunicative conditionals are often used to draw attention to these presuppositions and are typically used either to strengthen the belief of the existence of the referents (assertive use of some or one) or, on the contrary, not to commit oneself to the existence of the referent (non-assertive any) and to restrict the validity of such a belief. These opposite functions are expressed by the parenthetical formula if there is one for the assertive attitude (44), and if there is any for the non-assertive attitude (45):

(44) a. This is a time if there ever was one for parents to show their thoughtfulness and generosity towards each other.

[COB: 117] b. That, I smarted, is a royal rebuff if ever there was one. [BRO: 192]

(45) The shooting-season opens Saturday and the birds'11 be scattered all over the place after that if there's any left. [COB: 165]

In (44a) the main clause implies that now the occasion is present, so the speaker takes an affirmative stand to the existence of such a single referent. This attitude completely changes if one is replaced by any. Then the assertive attitude becomes non-assertive, and the speaker no longer commits himself to the existence of such a referent. In (45) the use of any is the only possibility because the non-assertive suggestion is that there might not be a single bird left.

In addition to these two possibilities of the metacommunicative conditional, its non-assertive use seems to be the only possibility in the elliptic formulaic expression if any, as in (46):

(46) a. Secondly, we must pool and ration our supplies of food, if any. [LOB:64] b. The change, if any, in foreign policy will consist rather of a freshness of approach. [LOB:98]

The elliptical use of if any in these two examples relates to the non-affirmed existence of the referents of the noun phrases supplies of food in (46a), and change in (46b). Structurally, the if any-phrase does not function as an antecedent to the consequent, but it is simply attached to the


noun phrase it modifies, and therefore functions as a small afterthought related to the presupposition made by the use of the definite determiners in the noun phrases.

When the presupposition is related to a locative expression such as here, the elliptic phrase uses the adverbial form anywhere as in (47):

(47) For here if anywhere in contemporary literature is a major effort to counterbalance Existentialism. [BRO: 108]

The use of if anywhere in (47) is ambiguous: it can either restrict the belief in the existence of a given place in contemporary literature referred to by here, or it can also function in a rhetorical way and stress the fact that it is precisely here that Existentialism will be counterbalanced.

This rhetorical function has become the only use of the elliptical expression if anything, but now it is not the existence of the referents or the processes that is focussed upon; on the contrary, the phrase if anything no longer has a reference-modifying function, but rather a sentential function leading to some strengthening effect as in (48):

(48) a. If anything, he had merely become a little more reserved, and much more polite. [LOB: 127] b. The effect of make-believe was, if anything, heightened by the arrival in the room of the German uniforms. [LOB: 126]

Thus the phrase if anything has become an idiomatic expression relating to the sentence as a whole and meaning something like 'the least one can say is that...'. This idiomatic use may also be the explanation why the phrase if anything can be used at the beginning of a sentence, as in (48a) or with an almost adverbial function in mid-position between an auxiliary and a verb as in (48b).

The elliptical use of the metacommunicative conditional occurs in the context of adjectives as in (49):

(49) a. And I suppose now that you've finally grown up, if a little late, you'd go on producing kittens every six months or so.

[BRO:20] b. As usual, delegates were entertained to dinner by the Royal

Yachting Association and a very pleasant affair it was, if a little cramped in its temporary quarters. [LOB:87]


(49) c. Hobhouse was Byron's intimate, if a little stuffy and unimaginative friend from Cambridge. [LOB:88]

Here the elliptical construction is in fact a double elliptical construction: it is (i) syntactically elliptical as in the former cases since there is no subject or verb, but only an adjective and some adjuncts; and (ii) the conjunction if itself is an elliptical form of the concessive conjunction even if. Therefore, the elliptical construction does not relate to a presupposition here, but it expresses a concession, which is always a type of restrictive utterance: the positive characteristic in the consequent, for instance you've finally grown up in (49a), is somewhat restricted by contrasting it with the concession in the antecedent that this achievement happened a little late. So examples (49a, b, c) illustrate a merely concessive construction rather than a metacommunicative conditional and rather fall in a different, though related category of conceptual constructions.6

7. Conclusions The above analyses have shown that although pragmatic conditionals

cover a wide variety of specialized conditional uses, they also have very typical common characteristics as well as specific idiosyncratic ones.

(i) Pragmatic conditionals all have one thing in common. Their antecedents all contribute in their own specific ways to the efficient performance of a communicative act contained in the consequent. In identifying conditionals the speaker uses the identification process in order to create a rhetorical impact on the hearer in his communicative act. In inferencing conditionals the speaker uses all possible types of inference-focusing devices in order to emphasise the speaker's process of inferencing. In discourse conditionals, speakers express the many different conditions under which their speech act forces are relevant for the hearer. And finally, in metacommunicative conditionals, speakers emphasise their own perspective in highlighting some or other important aspects of their communicative acts. In spite of these common aspects of pragmatic conditionals, which consist in the emphasis put on either the speaker's or the hearer's role in the process of communication, their specific functions or uses stand out clearly, too.

(ii) A second common and distinguishing element between the four types of pragmatic conditionals is the relatively great to almost absolute amount of independency of antecedent and consequent. This is a common element in so far as it distinguishes pragmatic conditionals from course-of-events conditionals and hypothetical conditionals, but it also differentiates


between the four subcategories of pragmatic conditionals, i.e. the distinction between identifying and inferencing conditionals on the one hand, and discourse conditionals and metacommunicative conditionals, on the other. This distinction is extremely great, since in the latter two (conversational) conditionals the if-clause can even be omitted. In its turn, the distinction between discourse conditionals and metacommunicative ones is again very great because the former have no formal link with the consequent whereas the latter always have a referential link. Of all the four pragmatic conditionals it is the discourse conditional that represents the most extreme deviation from the prototypical member of the category of conditionals, viz. the hypothetical conditional, since it conveys the highest degree of independence between the antecedent and the consequent from all points of view be it intonational, syntactic, referential or pragmatic ones.7

(iii) Although the two conversational conditionals, i.e. discourse conditionals and metacommunicative conditionals, have some elements in common, the differences between the two have proved to be even greater than originally thought. The present analysis has shown that the biggest difference between the two is precisely what the term 'metacommunicative' suggests: in each metacommunicative conditional there is always a comment on the consequent and the communicative act contained in it. This pragmatic link is syntactically expressed by means of a reference device such as so, in If I may ask so, any in phrases such as if any, if anything, if anywhere, substitution elements such as one in If there is one or anaphoric elements such as that in If that's what you mean. Discourse conditionals contrast very strongly with metacommunicative conditionals in this respect: they do not contain any referential link, but only a relevance link, to the consequent. In fact, as will be remembered, they are characterised by an intonational and syntactic 'caesura'. The 'caesura' which reflects a major conceptual distinction also makes the antecedent almost unnecessary, from a linguistic point of view. Anything that can be left out in the communicative transfer is both linguistically and conceptually less prominent.

(iv) The more general conclusion to be drawn from the analysis of the above structural patterns is that authentic discourse yields many more structural possibilities of any type of conditional, or, more generally, of any type of conceptual structure, than could ever be identified by mere introspection. While analysing these many examples, it was not primarily our purpose to give an exhaustive list of all the types and features of pragmatic conditionals, but in the first place to show that any conceptual category has many more structural realisations than an approach solely based on introspection can guarantee. Probably introspection will or tends to


produce the prototypical cases only, and corpus-based examples will provide the whole range of structural possibilities. Both approaches complement one another, which is a conclusion everyone could probably agree with. Thus one could say that Austin, an introspectivist if ever there was one, put the pragmatic conditional firmly on the agenda. But at the same time, it is only thanks to corpus linguistics that the many diverse subtypes could be brought out. The ultimate purpose of this paper is to take this conclusion one notch higher: neither of the two approaches can do without the other. Without an extensive corpus study, descriptive linguistics would never have managed to single out a class of 'course-of-events conditionals' and similarly, we would never have arrived at the insight of the existence of a superordinate category of 'pragmatic conditionals', comprising a fairly heterogeneous number of subcategories. In fact, we might still be concentrating only on the most prototypical category, i.e. that of 'hypothetical conditionals'.

Notes * We want to thank the two anonymous reviewers for their highly insightful remarks and queries. 1 It is, for instance, not discussed in Comrie (1986), Haegeman (1984), Sweetser (1990), Sweetser (1996), and Dancygier and Sweetser (1996). This latter paper discusses both (our) hypothetical and pragmatic conditionals, but not (our) course-of-events conditionals. This category is vaguely present in Dancygier and Sweetser (1997:123) under the name 'generic conditional' and exemplified by non-corpus-based examples such as If Mary bakes a cake, then she gives a party. 2 The four corpora used here are abbreviated as follows: BRO for Brown Corpus, COB for Cobuild, LDC for Leuven Drama Corpus, and LOB for Lancaster-Oslo-Bergen Corpus. 3 This term is one of a set of three terms, designating Halliday's functions of language use, i.e. the ideational, interpersonal, and textual functions. It is not implied that pragmatic conditionals solely serve the interpersonal function, but that they do so prototypically. The inverse applies to hypothetical and course-of-events conditionals, which mainly serve the ideational function, but allow many other indirect speech acts as well. Sweetser (1990) and Dancygier and Sweetser (1996:88ff.) even refer to hypothetical conditionals as 'content' conditionals, which is an even stronger ideational characterisation. 4 Although it is not explicitly stated, it looks as if Dancygier and Sweetser (1996, 1997) still keep this position. 5 The function of then in conditionals is extensively discussed in Dancygier and Sweetser (1997) and is seen as a deictic reference to a mental space built up by an if-clause, which must therefore precede the consequent. However, this analysis can at best only say when the consequent can take then, not what its presence actually adds to the meaning of the whole sentence. Starting from the iconic principle 'that more linguistic material also implies more meaning', we suggest in the text that the additional function of then is to highlight and to focus on the inferencing process. Thus in (16a) there is an 'inductive leap' from an observation to a conclusion, but in (16b) with the use of then there is a deliberate


inference, which may also be marked by many other syntactic devices, as shown in the further discussion. 6 This view is elaborated in Dirven (1997). 7 This conclusion seems to be somewhat in conflict with the scale of dependency in Table 2 where the metacommunicative conditionals are claimed to have the lowest degree of dependency. However, there is not a real contradiction, since in Table 2 we look at the relationship between antecedent and consequent. Metacommunicative conditionals do not relate to the complete relationship but only to one element in the antecedent. But the link with this selected element is much stronger than the link between the antecedent and the consequent in a discourse conditional.

References Athanasiadou Angeliki and Ren Dirven. 1996. "Typology of if-clauses".

Cognitive Linguistics in the Redwoods ed. Casad Eugene, 609-654. Berlin: Mouton de Gruyter.

. 1997. "Conditionality, hypotheticality, counterfactuality". On Conditionals Again eds. Athanasiadou Angeliki & Ren Dirven, 61-96. Amsterdam & Philadelphia: J. Benjamins.

Austin, J.L. 1961. 'Ifs and cans". Philosophical papers eds. J.O. Urmson & G.J. Warnock, 153-180 (3rd edition, 1979). Oxford: Oxford University Press.

Comrie, Bernard. 1986. "Conditionals: A Typology". On Conditionals eds. E.C. Traugott et al., 77-99. Cambridge: Cambridge University Press.

Dancygier, Barbara & Eve Sweetser. 1996. "Conditionals, Distancing, and Alternative Spaces". Conceptual Structure, Discourse and Language ed. Adele E. Goldberg, 83-98. Stanford, Calif.: CSLI Publications.

. 1997. "Then in Conditional Constructions". Cognitive Linguistics 8:2. 109-136.

Dirven, Ren. 1997. "The conceptual space between conditionals and so-called concessives". Grammar and Text in Synchrony and Diachrony. In Honour of Gottfried Graustein eds. Mechtild Rheinhardt & Wolfgang Thiele, 77-99. Frankfurt: Vervuert / Madrid: Iberoamericana.

Fauconnier, Gilles. 1994. Mental Space. Aspects of meaning construction in natural language. Cambridge: Cambridge University Press.

Haegeman, Liliane, 1984. "Pragmatic Conditionals in English". Folia Linguistica 18.485-502.

Morris, Charles. 1946. Signs, Language and Behavior. New York: Prentice-Hall.

Sweetser, Eve. 1990. From Etymology to Pragmatics. Cambridge: Cambridge University Press.


. 1996. "Mental Spaces and the Grammar of Conditional Constructions". Spaces, Worlds, and Grammars eds. Gilles Fauconnier & Eve Sweetser, 318-333. Chicago & London: The University of Chicago Press.

Traugott, Elizabeth , Alice ter Meulen, Judy Snitzer Reilly & Charles A. Ferguson, eds. 1986. On Conditionals. Cambridge: Cambridge University Press.

How Polish Structures Space Prepositions, Direction Nouns, Case, and Metaphor

Barbara Dancygier Simon Fraser University, Burnaby, Canada

1. Introduction Numerous studies have now shown that spatial construais in different

languages can rely on the cooperation among various subsystems of the language (cf. Talmy 1983, Ameka 1995, Sinha and Kuteva 1994). Even in languages in which prepositions are primary carriers of spatial information, their interaction with other subsystems is an important aspect of the resulting spatial interpretation (cf. Smith 1987, 1993, 1995, Serra-Borneto 1997). In this paper I will present some data from Polish to show how spatial construais result from an interaction of three major subsystems: direction nouns, prepositions, and the case form which is marked on the noun, following the choice of a given preposition.

In Polish, as in a number of languages, prepositions make the most significant contribution to building spatial construais and, not surprisingly, they also display a variety of meanings and uses. At the same time, the main orientation axes ('up/down' and 'front/back') are not expressed by prepositions, but by nouns (which I will call 'direction nouns'). The information on 'direction' or orientation' can be further refined by using direction nouns with prepositions, so that the two subsystems often interact closely.

The third subsystem which contributes to constructing spatial information in Polish is case. On the one hand, some prepositions in Polish are similar to Smith's 'two-way' prepositions, in that the nouns following them can be marked with two different case endings, and that each choice of case results in a different spatial interpretation of the whole expression (cf. Smith 1987, 1993, 1995, Serra-Borneto 1997). Cases themselves can thus be claimed to carry their own meaning, possibly the kind of meaning relevant to spatial interpretations (as regards meaningfulness of case, it has been proposed independently of prepositional contrasts by Smith 1987, Langacker 1991, Janda 1993). On the other hand, at least one case in Polish, the instrumental, can express spatial information independently of prepositions, as well as in interaction with them. It can also be marked on the direction nouns. It seems, then, that, although the three subsystems are mostly seen

28 BARBARA DANCYGIER

in interaction (preposition and case, preposition and a direction noun and case, a direction noun and case), direction nouns and case endings can make their meaningful contributions to spatial interpretation independently of the remaining subsystems.

In Polish, as well as in other Slavic languages, spatial information can also be found in verbal prefixes. The prefixes are often similar to prepositions (first of all in form and sometimes also in meaning), but their primary function is that of providing aspectual information. As was shown in Janda (1986) on the example of some Russian prefixes, the numerous meanings of such prefixes can be related to their spatial source, but their representation of spatial information is not quite straightforward and regular. If they are used to represent spatial information, they do not seem to interact with the remaining subsystems; in fact, the same information is often repeated in a preposition quite like the prefix itself. The analysis of verbal prefixes thus seems to fall beyond the scope of the present paper.

In what follows I will also not attempt to tackle individual prepositions of Polish in any way; rather, I will try to show how broadly defined groups of prepositions interact with the other systems in building spatial construais. In other words, I will argue that spatial construais are motivated compositionally, and that the role of prepositions here is significant, but not unique.

The first section will show the ways in which direction nouns contribute to the construction of spatial meanings. In the next section, I will focus on the role of case marking options, primarily in prepositional phrases, but, in the case of the instrumental, also independently of prepositions. Finally, I will consider some metaphorical mappings motivated by the spatial construais reviewed in the main body of the paper.

2. Direction nouns The spatial nominals that I refer to as 'direction nouns' are used to

describe regions in space or regions of landmarks and usually come in pairs representing the contrasting directions along major orientation axes. For example, nouns such as gra and d ("up" and "down"), or przd and tyl ("front" and "back") are used, with prepositions and appropriate case marking, to indicate meanings such as direction of movement, as in (1), location of objects with respect to the observer, as in (2), or parts of objects, as in (3):

HOW POLISH STRUCTURES SPACE 29

(1) Poszlimy w gr / w d We-walked in up-ACC / in down-ACC

"We walked up/down"

(2) Miasto zostao w tyle City stayed in back-LOC "I left the city behind me"

(3) Kino jest na dole Cinema is on down-LOC

"The cinema is on the ground level/downstairs"

The two orientation axes exemplified above seem to be the most salient in Polish, since the direction nouns marking the contrast are used in the greatest number of 'preposition + noun' combinations, and all of the case-marking options are also used. Another orientational nominal commonly used is the noun rodek ("middle" and "the inside"), which represents regions of landmarks (though typically not regions in space), and is used with a number of prepositions, as in w rodku (in middle-LOC, "in the middle" or "inside") or do rodka (into middle-GEN, "into the middle"). Also the noun bok ("side") has a number of similar uses, even though it is not clear whether rodek and bok make a contrasting pair in the same way as gra and d.

As regards ot

[ad foolen, frederike van der leek] constructions (bookzz.org)

Documents