Cause - Condition - Concession - Contrast: Cognitive and Discourse Perspectives [Reprint 2021 ed.] 9783110219043, 9783110166903

Citation preview

Cause - Condition - Concession - Contrast

W DE G

Topics in English Linguistics 33

Editors

Bernd Kortmann Elizabeth Closs Traugott

Mouton de Gruyter Berlin · New York

Cause - Condition - Concession - Contrast Cognitive and Discourse Perspectives

Edited by

Elizabeth Couper-Kuhlen Bernd Kortmann

W DE

G

Mouton de Gruyter Berlin' New York 2000

Mouton de Gruyter (formerly Mouton, The Hague) is a Division of Walter de Gruyter GmbH & Co. KG, Berlin.

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

Library of Congress Cataloging-in-Publication-Data Cause, condition, concession, contrast : cognitive and discourse perspectives I edited by Elizabeth Couper-Kuhlen, Bernd Kortmann. p. cm. - (Topics in English linguistics; 33) Includes bibliographical references and index. ISBN 3-11-016690-9 (alk. paper) 1. Grammar, Comparative and general-Clauses. 2. Grammar, Comparative and general-Syntax. 3. Cognitive grammar. 4. Discourse analysis. I. Couper-Kuhlen, Elizabeth. II. Kortmann, Bernd, 1960III. Series. P297.C382000 415-dc21 00-033563

Die Deutsche Bibliothek - Cataloging-in-Publication-Data Cause - condition - concession - contrast : cognitive and discourse perspectives I ed. by Elizabeth Couper-Kuhlen ; Bernd Kortmann. - Berlin ; New York : Mouton de Gruyter, 2000 (Topics in English linguistics; 33) ISBN 3-11-016690-9

© Copyright 2000 by Walter de Gruyter GmbH & Co. KG, 0-10785 Berlin All rights reserv~d, including those of translation into foreign languages. No part of this book may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopy, recording, or any information storage and retrieval system, without permission in writing from the publisher. Printing: Werner Hildebrand, Berlin - Binding: Liideritz &. Bauer GmbH, Berlin. Cover design: Christopher Schneider, Berlin. Printed in Germany.

Contents

1

Introduction Elizabeth Couper-Kuhlen and Bernd Kortmann

I. Cause The relevance of causality Paul Georg Meyer

9

On the processing of causal relations Leo G. M Noordman and Femke de Blijzer

35

Domains of use or subjectivity? The distribution of three Dutch causal connectives explained Henk Pander Maat and Ted Sanders

57

Causal relations in spoken "discourse: Asyndetic constructions as a means for giving reasons Christine Gohl

83

II. Condition Constructions with if, since, and epistemic stance, and clause order Barbara Dancygier and Eve Sweetser

because:

Causality,

III

On affirmative and negative complex conditional connectives Estrella Montolio

143

Pre- and post-positioning of wenn-clauses In spoken and written German Peter Auer

173

Counterfactual reasoning and desirability Noriko McCawley Akatsuka and Susan Strauss

205

vi

Contents

III. Contrast Adversative connectors on distinct levels of discourse: A re-examination of Eve Sweetser's three-level approach Ewald Lang

235

Viewpoints and polysemy: Linking adversative and causal meanings of discourse markers l~Cott A. Schwenter

257

The treatment of contrasts in interaction Cecilia E. Ford

283

IV. Concession Concessives on different semantic levels: A typological perspective Mily Crevels

313

Causal and concessive clauses: Formal and semantic relations Ekkehard Konig and Peter Siemund

341

Concession implies causality, though in some other space Arie Verhagen

361

Concessive patterns in conversation Elizabeth Couper-Kuhlen and Sandra A. Thompson

381

"that's true, although not really, but still": Expressing concession in spoken English Dagmar Barth

411

From concessive connector to discourse marker: The use of obwohl in everyday German interaction Susanne Giinthner

439

Index

469

Authors' addresses

473

Introduction

The theme of this book-endearingly referred to by its editors and contributors as the CCCC or four-Cs volume-is a set of relations, conceptual in nature but instantiated linguistically, which can be said to hold typically between clauses or sequences of clauses in discourse. Most, if not all, of our contributors will undoubtedly agree that each of these relations can be realized or marked by different linguistic means, e.g. by adverbials, particles, coordinating and subordinating conjunctions, word order (see, for instance, the chapters by Barth, Dancygier/Sweetser, Gohl, Montolio and Pander Maat/Sanders, which compare and contrast different lexical connectives in the fields of causality, conditionality and concessivity). Some of our contributors will undoubtedly maintain that the C-relations can even hold in the absence of specific lexical or syntactic markers (see, for instance, the chapters by Crevels, Gohl and Meyer). Yet many of our contributors may disagree on whether these relations are fundamentally semantic (see, e.g., the contribution by Konig/Siemund) or fundamentally rhetorical/interactional in nature (see, e.g., the chapters by Barth, Couper-KuhlenfThompson, Gohl and Meyer). Those who think of the C-relations as semantic will be concerned to represent their meaning (or the meanings of their markers) in a contextindependent, perhaps even a formal fashion, whereas those who see them as basically rhetorical or interactive will address their interpretation (or the interpretation of their markers) in specific-although perhaps generalizable-contexts. Yet in this divergence-or rather, diversity-of opinion we see one of the strong points of our endeavor. Indeed, the novelty of the present volume lies not only in the cutting-edge research which it presents but also in the fact that it embodies work at the frontier of two very different approaches to language-cognitive linguistics and discourse or interactional linguistics. By bringing these two traditions together in one volume, we hope to initiate a dialogue in which the respective bodies of work can be evaluated for their relevance to one another. The contributions collected here have been grouped roughly into sections according to C-relation in the order: cause, condition, contrast, concession. However, since some chapters explicitly address the relationship between more than one relation, the section boundaries are by no means rigid. In fact, this permeability is a reflection of deep-lying

2

Elizabeth Couper-Kuhlen and Bernd Kortmann

affinities and wahlverwandtschaften between the C-relations, which make themselves apparent in both cognition and discourse. Cause, condition and concession have long been known to have a special relationship with one another as adverbial relations of circumstance (cf e.g. Halliday 1985, Thompson/Longacre 1985, Kortmann 1997). At times they have even been conceptualized in terms of one another, a conditional relation being seen e.g. as a hypothetical variant of a causal relation, a concessive relation as an inoperant cause (Harris 1988, Konig 1986, 1988). The conceptual similarity between cause, condition and concession is also reflected in the fact that languages may encode them in the same way (Kortmann 1997). All three relations lend themselves to expression via syntactic subordination; moreover, single subordinators may be polysemous between cause, condition or concession (e.g., if can express both condition and cause, for both cause and concession; see the chapter by Konig/Siemund in this volume). Dancygier/Sweetser (this volume) explore how this functional overlap can be accounted for with respect to the connectives if, since and because. The relation of contrast, on the other hand, is typically expressed by syntactic coordination and not unanimously thought of as an adverbial circumstantial relation (Halliday 1985, Thompson/Longacre 1985). Yet especially when it is considered from a pragmatic and/or discourse perspective, contrast begins to have more in common with concession (see e.g. Rudolph 1996) and even with cause and certain kinds of condition. Sweetser (1990), for instance, applies her three-domain model not only to causal, conditional and concessive sentences but also to adversative sentences with the coordinator but (see Lang, this volume). Moreover, as several of the discourse-based papers in this collection argue, contrast is centrally implicated, e.g., in counterfactual conditionality (Akatsuka/Strauss) and in concession (Barth, CouperKuhlen/ Thompson). Contrast, specifically adversativity, is furthermore involved in the discourse-marker use of Spanish si, a canonical conditional conjunction (Schwenter) and in the discourse-marker use of German obwohl, a canonical concessive conjunction (Giinthner). Finally, contrast enters into wahlverwandtschaften with circumstantial C-relations, in particular with causality, in that justifications recurrently follow contrasts in discourse, as Ford (this volume) shows. These findings suggest then that the affinity between contrast on the one hand, and cause, condition and concession on the other, has been underestimated in the past. For this reason contrast has been included as a fourth C-relation here.

Introduction

3

Despite the division of our collected papers into sections, there areas might be expected-recurrent themes which cut across the C-relations. One of the motifs in the cognitively oriented papers, for instance, is Sweetser's (1990) seminal work on domains or levels of interpretation. Noordmanlde Blijzer (this volume) adopt her distinction between content and epistemic level and show how it will help account for differing degrees of cognitive complexity in causal sentences. Other papers are more critical of Sweetser's model: Lang (this volume) takes issue with her claim that content, epistemic and speech-act levels of interpretation can be determined independently of syntactic structure and, like Crevels (this volume), fmds it necessary to supplement Sweetser's three domains with a fourth, textual level. Given a fourth level, Crevels fmds the model useful for investigating the formal means which typologically different languages deploy in each domain. Pander Maat/Sanders (this volume), on the other hand, reject Sweetser's model as a means of accounting for the distribution and interpretation of specific causal markers in Dutch and propose in its stead a notion of subjectivity. Another recurrent motif in several of the chapters collected here is Fauconnier's theory of Mental Spaces (1985, 1997). Dancygier/Sweetser (this volume) show that important distinctions can be drawn based on the ways in which causal and conditional conjunctions participate in the configuration of mental spaces. Verhagen (this volume) uses a mentalspace model to account, e.g., for the relationship between concession and (negated) causality, with which Konig/Siemund (this volume) are also concerned. Yet AkatsukaiStrauss (this volume) are critical of a mentalspace account, because it neglects what they see as an inherent dimension of counterfactual conditionality, namely that speakers are expressing a stance towards the events in question as desirable or undesirable. AkatsukaiStrauss advocate a subjective dimension in the analysis of counterfactual conditionality and in this sense are in agreement with Pander Maat/Sanders, who see subjectivity as responsible for the behavior of specific causal connectives in Dutch discourse. It is perhaps worth noting that both AkatsukaiStrauss and Pander Maat/Sanders take a corpus-based approach to C-constructions and for this reason belong more in the discourse than the cognitive tradition. Significantly, it is in relying on real data rather than on introspection that these two studies independently establish a need-above and beyond mental-space or domain considerations-for taking the speaker's (or subject of consciousness's) stance into account in the analysis of C-constructions, especially for counterfactual conditionality and causality. In this sense they are initiating the kind of dialogue between cognitive and discourse

4

Elizabeth Couper-Kuhlen and Bernd Kortmann

approaches to the C-relations which the editors have envisaged (cf. also the chapters by Meyer and Konig/Siemund in this respect). A second notion to be found in both cognitive and discourse-oriented papers collected here is polyphony (Verhagen), or multiple viewpoints (Schwenter). For Verhagen, the interpretation of concessive and negated causal sentences as well as of all sentences involving epistemic Crelations calls for the construction of two nearly similar mental spaces in Fauconnier's theory-or, equivalently, for the assumption of polyphony in Ducrot's sense of the term (1984, 1996). Schwenter also mentions Ducrot's theory of polyphony as one way of conceptualizing viewpoints and linguistic structure. He opts, however, for Roulet's (1984) distinction between one viewpoint (monological) and two viewpoints (dialogical), both of which can be found with either one speaker or two. Schwenter finds that dialogicality is not only conceptually present in adversative constructions but also (to a lesser degree) in 'exhortative' and epistemic causals (i.e., ones which accompany exhortative speech acts and inferential conclusions). The notion of polyphony or multiple viewpoints then appears to be relevant to both cognitive and discourse-based understandings of the C-relations. And it is clearly implicated in the interactional model of concession which Couper-KuhlenfThompson (this volume) propose, where multiple viewpoints often go hand in hand with multiple speakers-although, as Barth (this volume) shows, they need not do so. Related to the dimension of multiple speakers is of course the question of whether a piece of discourse-specifically, a C-construction-is spoken or written. This is a recurrent motif in several of the discourseoriented papers collected here. Aside from questions of overall frequency of use (see, for instance, Altenberg 1984, 1986 or Ford 1993), medial considerations are relevant for C-constructions in at least two other ways. One of these is with respect to the order of clauses, which Auer's chapter in this volume investigates with respect to wenn-clauses in German (cf. English 'when'/'if -clauses). A second issue relating to C-relations in spoken and written discourse is the question of whether mediality has an influence on the choice of C-marker. Montolio (this volume) argues that, as in English, complex conditional constructions in Spanish such as a condici6n de que ('provided that') and a menos que ('unless') express a more restricted conditional relation than si 'if and are reserved for formal written registers. And Barth (this volume) shows that the set of concessive markers as well as the relative frequency of parataxis as opposed to hypotaxis is significantly different when spoken English is compared to written English.

Introduction

5

A further recurrent motif in the papers dealing with spoken discourse is the use of C-connectives as discourse markers. This is an aspect of Crelations which has been neglected in cognitively oriented studies. In fact, the development of discourse markers from adverbial subordinators of cause, condition, contrast and concession has only recently come to the attention of discourse and interactional linguists (see e.g. Gohl/Giinthner 1999 for discussion of a causal discourse marker in German, and Lenk 1997 for discussion of contrastive discourse markers in English). In the present volume it is the chapters by Schwenter on Spanish si and Giinthner on German obwohl which deal with C-relations as discourse markers. Whereas the obwohl phenomenon is comparable to paratactic uses of although in English, the adversative and causal use of si has no real equivalent with English if. But the evidence in both cases casts doubt on developmental claims made in the literature to the effect that conditional markers develop into concessive conditionals and from there into concessives, while concessives do not develop any further (Konig 1986, 1988, Kortmann 1997). A line of development from conditionality to adversativity and (paratactic) causality as with si, or from concessivity to adversativity as with obwohl has gone unnoticed until now. Once again, it is not insignificant that such insights have come from corpusbased studies of spoken language. In fact, as Auer (this volume) points out, a full understanding of spoken-language constructions can only be achieved through corpus-based investigation. And this is another recurrent theme in virtually all the discourse-oriented papers collected here. The interactional linguistic papers in this volume (Auer, Barth, Ford, Gohl, Giinthner, Couper-Kuhlenffhompson) go one step further with respect to corpus-based language study. They take it as axiomatic that spoken language is first and foremost a tool for social action and that in order to be understood fully it must be examined in its original habitat, i.e. in everyday interaction. Viewed from this perspective, the C-relations can be thought of as ways of carrying out social actions-causality being instrumental in providing justifications or accounts (Gohl), contrast and concession in negotiating agreement/disagreement (Barth, CouperKuhlenffhompson, Ford). C-constructions in tum can be conceptualized as linguistic resources or practices for carrying out the actions in question, e.g. for conceding a point or correcting a prior claim (Giinthner). In an interactional perspective, C-constructions are seen as especially appropriate for or adapted to situated interactional needs and tasks: a German wenn-clause in post-position-as Auer shows-is not merely a positional variant of a pre-posed 'one (and thereby freely

6

Elizabeth Couper-Kuhlen and Bernd Kortmann

interchangeable with it) but a tool 'designed' for or tailored to specific contexts of usage. A fmal motif-relevant to both cognitive and discourse-oriented studies-is the relationship between asyndetic, paratactic and hypotactic means of expressing C-relations. Although none of our chapters looks at C-connectives diachronically, the implicational hierarchy for concession that Crevels (this volume) offers in her typological study, namely Content > Epistemic > lliocutionary > Textual, in conjunction with her marking hypotheses has clear implications for language development. She hypothesizes that the higher the semantic level, the more likely a concessive relation is to be realized asyndetically. Moreover, if a concessive relation is realized syndetically, the higher the semantic level at which it is realized, the more likely it is that q rather than p will be marked-where q-marking corresponds to parataxis and p-marking to hypotaxis. The implicational hierarchy above corresponds, of course, to what are thought to be increasing levels of cognitive complexity (see also Noordmanlde Blijzer, this volume), whereas the cline asyndesisparataxis-hypotaxis corresponds to the unidirectionality hypothesis of grammaticalization theory (HopperfIraugott 1993). Yet evidence from spoken language does not fully conform to these hypotheses: asyndesis is widespread, even for content-domain causality (Gohl, this volume) and parataxis is more frequent than hypotaxis for the expression of (epistemic) concessive relations (Barth, this volume). Furthermore, as Schwenter's and Giinthner's chapters show, discourse markers which originate from adverbial subordinators create parataxis out of hypotaxis and thus reverse the unidirectional cline. In order to resolve such contradictions we anticipate that more dialogue between cognitive and discourse-oriented approaches to the C-relations will be necessary in the future. Meanwhile the reader is invited to appreciate the state-of-the-art research documented here for cause, condition, contrast and concession. In line with the title of the series, the present collection focuses on English; yet it does not eschew the world beyond. Indeed, the contributions by Noordmanlde Blijzer and Pander Maat/Sanders (Dutch), by Auer, Gohl, Giinthner and Lang (German), by AkatsukaiStrauss (Korean, Japanese) and by Montolio and Schwenter (Spanish) serve, each in their own way, as a valuable corrective on what would otherwise be a perhaps excessively Anglocentric perspective. In this sense the volume presents a picture of the C-relations which is enriched by cross-linguistic research.

Introduction

7

The editors would like to thank Verena Haser, Lieselotte Anderwald and Manfred Krug for careful comments on earlier versions of the contributions, Susanne Wagner for compiling the index, and Sabine Conrad for checking all final versions for conformity to the publisher's stylesheet. A word of heartfelt thanks also to Melitta Cocan for the excellent job she has done in preparing the camera-ready copy.

Elizabeth Couper-Kuhlen, Konstanz

Bernd Kortmann, Freiburg

References Altenberg, Bengt 1984 Causal linking in spoken and written English. Studia Linguistica 38: 20-69. Altenberg, Bengt 1986 Contrastive linking in spoken and written English. In: G. Tottie and I. Baecklund (eds.), English in Speech and Writing. A Symposium, 13-40. Stockholm: Almqvist & Wiksell. Ducrot, Oswald 1984 Le dire et Ie dit. Paris: Minuit. Ducrot, Oswald 1996 Slovenian lectures/Conferences slovenes: Argumentative semantics/Semantique argumentative. Ljubljana, ISH. Fauconnier, Gilles 1985 Mental Spaces: Roles and strategies. Cambridge, MA: MIT Press. Fauconnier, Gilles 1997 Cognitive Mappings for Language and Thought. Cambridge: Cambridge University Press. Ford, Cecilia E. 1993 Grammar in Interaction. Cambridge: Cambridge University Press. Gohl, Christine and Susanne Gtinthner 1999 Grammatikalisierung von 'wei!' als Diskursmarker in der gesprochenen Sprache. Zeitschriftfur Sprachwissenschaft 18: 39-75. Halliday, M. A. K. 1985 An Introduction to Functional Grammar. London: Edward Arnold. Harris, Martin 1988 Concessive clauses in English and Romance. In: 1. Haiman and S. A. Thompson (eds.), Clause Combining in Grammar and Discourse, 71-99. Amsterdam: Benjamins. Hopper, Paul and Elizabeth C. Traugott 1993 Grammaticalization. Cambridge: Cambridge University Press.

8

Elizabeth Couper-Kuhlen and Bernd Kortmann

Konig, Ekkehard 1986 Conditionals, concessive conditionals and concessives: Areas of contrast, overlap and neutralization. In: E. C. Traugott, A. ter Meulen, 1. S. Reilly and C. Ferguson (eds.), On Conditionals, 229246. Cambridge: Cambridge University Press. Konig, Ekkehard 1988 Concessive connectives and concessive sentences: Crosslinguistic regularities and pragmatic principles. In: 1. A. Hawkins (ed.), Explaining Language Universals, 145-166. New York: Basil Blackwell. Kortmann, Bernd 1997 Adverbial Subordination. A typology and history of adverbial subordinators based on European languages. Berlin: Mouton de Gruyter. Lenk, Uta 1997 Marking Discourse Coherence. Functions of discourse markers in spoken English. Tiibingen: Narr. Roulet, Eddy 1984 Speech acts, discourse structure, and pragmatic connectives. Journal ofPragmatics 8: 31-47. Rudolph, Elizabeth 1996 Contrast: Adversative and concessive relations and their expressions in English, German, Spanish, Portuguese on sentence and text level. Berlin: de Gruyter. Sweetser, Eve 1989 From Etymology to Pragmatics. Metaphorical and cultural aspects ofsemantic structure. Cambridge: Cambridge University Press. Thompson, Sandra A. and Robert E. Longacre 1985 Adverbial clauses. In: T. Shopen (ed.), Language Typology and Syntactic Description, Vol. 2 Complex constructions, 171-234. Cambridge: Cambridge University Press.

The relevance of causalityPaul Georg Meyer

"In der Natur gibt es keine Ursache und keine Wirkung". (Mach 41901: 513) "The law of causality, I believe, like much that passes muster among philosophers, is a relic of a bygone age, surviving, like the monarchy, only because it is erroneously supposed to do no harm." (Russell [1912] 1921: 180) "The most fundamental fact of the world" (KamIah 1991, title) "The cement of the universe" (Mackie 1974, title) This article discusses interrelations between relevance and causality from different angles: first, it is shown that the very notion of causality is subject to a relevance constraint. In communicating about causes people tend to seek out one or at most a limited number of possible causes from a possibly infinite number of logically admissible conditions. The criterion for selecting this limited number is obviously relevance. This interrelation between causality and relevance is analysed by way of a critical application of relevance theory, with the result that the conditions of relevance of causal statements are more similar to felicity conditions of speech acts than to a cognitive relevance principle. Second, causality is described as a major device for creating relevance in discourse. The relevance of causality in discourse cannot easily be overstated. Nevertheless, causality is only one among several principles of text organisation, though obviously the most relevant one.

Causality is an "elusive" concept (Ziv 1988: 543). We all seem to know a causal relation when we see one, but as soon as we try to explain why it is a causal relation, or to communicate what it is precisely that we have seen, we run into problems. Like time and space, causality is taken for granted by everybody and fully understood, so it seems, by nobody. And, what is most disconcerting to scientifically-minded people, "knowledge of causes is possible without a satisfactory grasp of what is involved in causation" (Anscombe [1971] 1975: 67). Scientists have

10

Paul Georg Meyer

consequently sought to eschew the notion in their considerations altogether (see mottoes). Philosophers, on the other hand, partly drawing on scientific results (Reichenbach 1991), have stressed its crucial importance for the functioning of the world (see mottoes). Almost a hundred years after Mach's and Russell's verdicts, philosophers, psychologists, social and cognitive anthropologists, and linguists are still wondering about causality in their different ways. The reason why the "law of causality", or at least the notion of causality, is still going strong, is independent of its status in nature or science. Causality is simply an indispensable principle in the organisation of people's everyday lives, both on the individual and the micro-group level, and, a fortiori, also on the level of societies, states, and international relations. When I clean the blades of my electric lawn-mower with my fingers, I rely on the law of causality, which tells me that once the plug is pulled from the socket, the engine cannot start and make the blades rotate again. When I am late for a meeting, I will try to find a cause l for my being late, and include it in the account that will presumably be expected from me; and my account will presumably be much more acceptable if I am able to give a satisfactory explanation for my being late. When people claim money from an insurance company, their claim presupposes the law of causality, which means that there was a cause for the damage and depending on what the cause was, the insurance company will or will not pay. When the UN sets up a tribunal to indict and convict war criminals, the legitimacy of the procedures hinges on the law of causality, on the belief that certain people are responsible for certain crimes, that is, that their actions were causally involved in these crimes in a crucial way. The law of causality has not been allowed to survive because people believed it could "do no harm" (Russell [1912] 41921: 180, cf motto). It has been allowed to survive because without a notion of causality, most of our material-technical, micro- and macro-social, economic, legal, and international life and institutions would not make sense and would break down. 2 It will be shown presently that these remarks have more meaning in the context of this article than the usual conventional introductory assertions about the outstanding importance of an article's subject matter (Swales 1981). The first reason for this is that the very topic of this article is the relevance of causality. Second, the examples adduced above to show the relevance of causality in an intuitive sense also show that, occasional scientific necrologues on the concept notwithstanding, causal-

The relevance ofcausality

11

ity seems to have its locus in the social sciences and humanities. Everyday action, group cohesion on all levels, economic transactions of all kinds, notions of obligation and responsibility, law as the formalisation of the latter notions, and many other topics of social sciences and humanities crucially depend on causality. We may thus take the above examples as a first approximation to a concept of causality that is located in a social rather than in a physical world. In other words, the universe of which causality is the cement, the world in which causality is the most fundamental fact, is primarily a socially constructed universe, a socially constituted world. This article does not undertake to solve the problem of causality. Its objective is much more modest. Based on observations made in the philosophical, scientific, psychological, social scientific, and linguistic literature, I will defend a set of linguistically inspired theses concerning the relationship between causality and relevance (cf Meyer 1983: 125126), putting it into the context of the present discussion and of an emerging research programme. I will begin with a (very) brief history of the concept of causality (Section 1). I will then show how the history of the concept, the way it has been discussed in the literature, and the problems it has raised, seem to indicate a certain constraint on the notion of cause that is best described by a notion of relevance (Section 2). Next I will ask whether this notion of relevance is explicable in terms of the presently most influential pragmatic theory of relevance (Sperber and Wilson 1995 2) (Section 3), and discuss problems arising from such an attempt. Finally, I will argue for a basically discursive explication of the notion of causality in the framework of a coherence relations approach, establishing causality as one of several primitive universal principles that govern the construction of coherence in discourse (Section 4).

1. A brief history of causality Talking of causality under a unitary label misleadingly suggests that the label covers a unitary, homogeneous phenomenon. Breul (1997: 81-112) has recently summarised the develoJ?ment of the notion of causality from Aristotle via Hume into the 20 century. Among Aristotle's four different kinds of causality (Metaphysics 5, 2, 1013a24ff), there is only one'· (known as causa efficiens from the Latin translation) that is clearly identifiable with a modern common-sense notion of causality. In the

12

Paul Georg Meyer

scientific and philosophical tradition of post-Renaissance Europe (e.g. Hume), another notion of causality is brought forward, emphasising the necessity and lawlikeness of the causal relation. Finally, the notion of causality in that tradition dissolved into the mathematical notion of function, as exemplified in Mach (1901) and Russell ([1912] 1921). This notion is certainly far removed from everyday intuitive concepts of causality. Strawson (1985) argued against a law-based notion of causality to cover everyday cases of causation: A man, say, falls down a flight of stone steps as he begins the descent. The steps are slippery and the man's mind is elsewhere. This is a sufficient explanation of his fall. But of course not every preoccupied man falls down every flight of slippery steps he descends. There is absolutely no question of our formulating or envisaging exceptionless laws, ... , to cover all such cases. (Strawson 1985: 131)

That is, all rational and enlightened people will assume that the man's falling down the stairs is fully in accordance with some presumed laws of nature and, in principle, fully explainable in such terms. And yet, in ordinary language we do not expect such full "scientific" accounts to be given, and indeed it is highly doubtful whether such full causal accounts of singular events are useful, desirable or, for that matter, feasible. For all practical purposes, we are content with much less than a listing of all the conditions, regularities, probabilities, and laws of nature that account for the event in question. A satisfactory explanation, in fact, is not one that lists all this. A satisfactory explanation is concise and concentrates on one or just a few causes. The problem that scientists and philosophers have had with the notion of causality emerging from this observation is that they cannot fmd a unique logical or mechanical or otherwise physical characterisation of a state of affairs that happens to be intuitively singled out as the "cause" of a certain event. Whatever is named as "the cause" of something, does not seem to differ in any significant way from other factors present in the situation in question. There simply seems to be no logical or scientific basis for calling something a cause. Logical analyses of the notion of cause have operated with necessary and sufficient conditions. The two intuitively most plausible and (presumably for that reason) most influential defmitions have been Mackie's (1965) "INUS condition" and the "counterfactual conditional", originally suggested by Hume and reintroduced into the modern debate by

The relevance ofcausality

13

David Lewis (1973). As we shall see in explicating what an INUS condition is, the two are actually quite close to each other. According to Mackie (1965) causes are best defined as "INUS" conditions. An INUS condition is "an insufficient but necessary part of a condition which itself is unnecessary but sufficient for the result. " (Mackie 1965: 16). Thus, to take Strawson's example quoted above, both the slippery state of the stairs and the man's absentmindedness are causes of the man's fall because they are • insufficient: Slippery stairs or absentmindedness alone do not cause a man to fall down a staircase, and would not have done so in that particular case. • necessary: Had the man been more careful or had the stairs not been slippery, the man would not have fallen. • part: The two causes mentioned are only part of the set of conditions causally involved in the man's fall. Thus, e.g., the man's trying to descend the stairs in the first place is another condition on the man's fall (that could also be called a cause on that account because it is an insufficient but necessary part of the set of conditions leading to the accident). The set of such conditions taken as a whole is • unnecessary for the result because the same effect might also have been achieved by other causes, in an altogether different set of conditions: Thus, e.g., the man might have fallen down the stairs because somebody maliciously pushed him, or because he was drunk and the steps were worn. • sufficient (in that particular case) for the result because, after all, the result was brought about. According to Hume and David Lewis (1973) c is the cause of an event e if and only if the counterfactual statement: "If c had not been true, e would not have happened" is true. In other words, a cause is a 'conditio sine qua non'. The attentive reader will have noticed that in explicating the notion of INUS condition above, we repeatedly had to resort to counterfactual formulations: if the man had been more careful or if the steps had not been slippery, he would not have fallen. Absentmindedness and the slippery state of the steps alone would not have brought about the accident. The same accident could have been brought about by altogether different causes, etc. It is thus clear that the INUS definition of causality is only a more formalised and elaborate variant of the counterfactual formulation of that definition.

14

Paul Georg Meyer

The problems with both these defmitions are manifold (Sosa 1975: 34), but it is not my purpose to discuss them here. There is just one point that I want to make: Mackie refers to a set of conditions that is unnecessary, but sufficient for the effect. The factor singled out as "the" cause is not the whole set of conditions, but only an insufficient but necessary part of this set. The problem that this definition raises is that it is not very useful for an application of the concept of causality in the analysis of everyday communication. The definition covers a number of conditions that we would not call causes in a given situation. That is, Mackie describes a necessary but not sufficient condition for calling an event a cause. He describes a semantic precondition for a proposition to serve as a causal explanation. Real events always have a large number of necessary and sufficient conditions. Arguing that one circumstance was the cause of a given event means foregrounding one insufficient and necessary condition and stating it, and allocating all the other conditions that jointly were unnecessary but sufficient to bring about the effect to the background. Thus, the very problem of Mackie's definition is an indication of the basically pragmatic character of any reasonable notion of causality that would be applicable in the analysis of discourse. It is a situation that logical positivists and context-free semanticists cannot easily cope with, but it is not unfamiliar to pragmatically oriented linguists: reality is always understated and underdetermined by the propositional content of utterances. The meaning of utterances always goes beyond what was said. There is always a set of meanings that were meant but left unsaid. And there is always a background of knowledge that is taken for granted, that is not problematised, that is simply not at issue. If we believe Searle (1983: 141), this background contains infinitely many propositions. And indeed, it is precisely this characteristic of the set of sufficient conditions for a certain event that has led early modem philosophers (such as Hume) to confine the notion of causality to "lawful" instances and has driven 19th and 20th century scientists and philosophers to despair.

2. What makes a cause a cause? Let me try to illustrate this with Strawson's example (in fact, any example of a causal relation would do). What is the set of conditions that is jointly sufficient (though unnecessary) to bring about "the man's" fall (for simplicity's sake let's call him Jones)? The two conditions men-

The relevance ofcausality

15

tioned by Strawson are in the foreground, of course: the steps were slippery, and Jones was absent-minded. Apart from these, I have already mentioned one further condition: the fact that Jones did try to descend the stairs at the very time when the first two conditions were already given. It seems that these three conditions, each of them independent of the two others, were jointly sufficient for Jones' fall. But, as Strawson (1985: 131) pointed out, "not every preoccupied man falls down every flight of slippery steps he descends." So there must have been further conditions that were necessary to make the set of conditions sufficient for Jones' fall to occur: maybe if Jones had been wearing different shoes (say, with non-slip rubber soles), he would not have fallen. Or, if there had been a railing for him to hold on to, he would not have fallen. Or, if his ankles had not been so stiff that day, he would not have fallen. Or, perhaps, if he hadn't been drinking, in addition to his absentmindedness. Or ... On the other hand, we might also begin to question the most obvious and plausible causes that were given: granted that the steps were slippery, were they slippery enough for a man to slip? Granted that Jones was absent-minded that day, was he absent-minded enough not to notice the slippery state of the steps? That is, even the most readily accepted explanations for an event are not undebatable. There is always a possibility that the most plausible explanation is erroneous. Suppose that Jones is absent-minded most of the time and the steps are always slippery and he walks down these steps several times a day without slipping. It is obvious that in this case we would still accept the explanations given by Strawson to a certain extent, although we would have evidence for the fact that the explanation is not sufficient. Even though we would be willing to grant that there must have been other "causes" for Jones' fall, the explanations "slippery steps" and "Jones' absentmindedness" are just too good to be discarded. What this discussion is supposed to show is: as scientists and philosophers found out long ago, there is no such thing as "the" cause or "set of causes" of an event. Causality emerges as a discursive phenomenon. Note that all the "causes" mentioned above could be adduced in some kind of dispute over the "causes" of Jones' fall. But it depends precisely on the kind of dispute as to which of the causes will be adduced by whom. Suppose Jones' wife had always been of the opinion that that staircase is dangerous and had always been warning Jones to use it (supposing there was an alternative available). In this case, the cause of Jones'

16

Paul Georg Meyer

fall that would be accepted by his wife would be the fact that Jones descended the stairs in the first place. ("Did you have to use that dangerous flight of stairs? I told you one day you'd break your neck on it".) Or that Jones descended the stairs while they were slippery ("Why couldn't you wait until the steps had dried?") (Supposing, e.g., that they were slippery because Jones' wife had just scrubbed them).3 For his wife, in that case, the slippery state of the stairs would be taken for granted, just like her husband's absentmindedness, which she has learnt to put up with in long years of marriage. Now take the case of an insurance company. Depending, of course, on what kind of risk had been the object of the insurance, all kinds of questions might be raised. Who was responsible for the slippery state of the steps? Why did the flight of stairs have no railing? Was Mr. Jones' absent-mindedness due to alcohol or drugs? What was the purpose of his descent down the stairs? Was it on his way to work, for example? Was Mr. Jones' stiflhess in the ankles due to an accident or some illness? Note that the insurance company's questions open up a much wider background by considering what might provisionally be called second- or third-order causes. The reason for this is that an insurance company will tend to follow up causal chains until somebody is found who can be blamed for an accident. Thus, for an insurance company or a lawyer, the cause of Mr. Jones' fall might be the fact that the owner of the house failed to replace a broken railing or to provide some kind of slip protection measures for the dangerous stone steps, or the fact that Mr. Jones' doctor prescribed some drug for him that affected his capacity to keep his balance to a sufficient extent to cause, in connection with the slippery state of the steps and Jones' absentmindedness, his fall down the stairs. That is, when considering the set of conditions jointly sufficient to bring about an event, it is not sufficient to look at "first-order" causes only. In any case, it is a question of focus and perspective what is regarded as "first-order cause". Take Mr. Jones' absentmindedness as an example. Strawson said it "caused" Mr. Jones' fall. But looking more closely at the actual event, we may be certain that innumerable other "causes" intervened between Mr. Jones' absentmindedness and his fall. Absentmindedness cannot immediately "cause" a fall. Before the fall there must have been a slip. And it is difficult to ascertain at which point in the causal chain leading to Mr. Jones' fall absentmindedness is supposed to come in. Did absentmindedness cause him to slip, and the slip cause him to fall? Or did he slip first, and did absentmindedness prevent

The relevance ofcausality

17

him from keeping his balance after the slip? Or did he slip and lose his balance inevitably, and did absentmindedness prevent him from holding on to something, which might have prevented his fall? Or was absentmindedness a limiting factor of all of Mr. Jones' movements that day, and did it have disastrous effects all through the causal chain leading to the accident? Even if this was a real case, I doubt that all this detail could be ascertained. It should be clear, however, that a distinction between first-order and second- or third-order causes is impossible and has no systematic status. What we regard as a cause is seldom the "immediate" cause (in the sense that no intervening event between cause and effect can be found) because our observations are seldom so fine-grained that "immediate" causes would be noticed. Let us suppose that theoretically we come to the conclusion that some basic muscular movement of Mr. Jones immediately prior to his fall must have been the "immediate" cause. Presumably neither Mr. Jones nor a by-standing observer would have been able to observe and identify this "immediate" cause. Presumably this "immediate" cause was the inevitable result of some prior events, conditions, and maybe of Mr. Jones' decisions, and these will most likely be singled out as "the cause". We tend to look at events and causal chains not from a mechanistic perspective, but from a perspective that allows us to see and represent points of potential intervention (cf Von Wright 1973). We look for conditions that could have been different. And usually conditions can be different if people behave differently or change conditions deliberately. Mr. Jones might not have used that dangerous staircase at all that day. Mr. Jones could have been more careful. The steps might not have been slippery (somebody could have done something about that). So the insurance company, by opening up an infinite space of conditions jointly sufficient for the accident, at the same time limits the search to those states of affairs that are accessible to human perception and intervention. Let us now look at Mr. Jones himself How would he explain his fall (supposing he survived it)? Again it would depend on a number of contingent factors what would be a satisfactory explanation of the accident for Mr. Jones himself So Mr. Jones might (most likely) blame the slippery stairs, but it should be clear that he can only do so under certain conditions. Suppose that Mr. Jones has been using these stairs for years, they have always been slippery, he has never had an accident and has never complained about the state of the steps. In that case it would be somewhat difficult for Mr. Jones to blame the slippery steps. If he chose

18

Paul Georg Meyer

to do so, it would mean a "change of policy", so to speak. After years of putting up with the slippery stairs, he would suddenly blame them for an accident. Under a benevolent interpretation, this would mean his fall has taught him a lesson, and he probably will have to do something about the slipperiness of the stairs in the near future. Mr. Jones could also blame his own absentmindedness, if he is willing to make such a facedamaging admission. What is more likely (assuming Mr. Jones to be vain like most people) is that he will blame some other condition that does not figure in Strawson's account at all. He could resort to some other explanation that would be more face-saving. "I was a bit stiff in the ankles that moming" would do perfectly. What this lengthy exposition of potential explanations for a fictitious accident is meant to show is that explanations are not objective, based on facts existing independently. In fact they are guided and informed by people's interests. What explanations we are willing to give and willing to accept as a "cause" is not so much a matter of fact, but in the best case open to negotiation. Establishing a cause of an event is not so much a fact-finding procedure, but rather a process of social interaction in which diverging interests have to accommodated, from which different consequences (even obligations) may arise (cf Hilton and Erb 1996: 303). It will thus be appropriate to look at the notion of causal explanation in order to clarify the notion of cause. This will be done in the next section. It should, however, be emphasised that the above account does not imply that "the truth" is at the mercy of subjective interests. All the explanations mentioned above that could be adduced for Mr. Jones' fall might be, strictly speaking, true. Truth is a semantic concept, and semantically, all the conditions mentioned above as possible causes of Mr. Jones' accident are INUS conditions: they are insufficient but necessary parts of the whole set of conditions leading to the accident. They differ not in terms of truthfulness but in terms of relevance in certain discursive contexts and for certain people.

3. The relevance of causality What does it mean for an explanation to be relevant in a given context? The notion of relevance plays a crucial role in several theories of pragmatics (Grice 1975; Hom 1984; Sperber and Wilson 1986, 1995). But before I relate the previous discussion to these theories, I would like to

The relevance ofcausality

19

explicate in my own terms what I mean by "relevance" here. What makes an explanation relevant? A good explanation raises an issue that was either - outside the "normal course of events", unpredictable, and out of control. This explains the occasional meteorite examples in the literature (Stegmuller [1969] 1983: 911). Or it was - within control of people, that is, something that could have been otherwise, a condition that was amenable to change, but still to a certain extent unexpected and outside the normal course of events. So when the slippery steps are adduced as an explanation of Jones' fall, it is implicated that perhaps they should not have been slippery, that something should be done to prevent people from slipping there. If someone takes the slippery stairs for granted as a fact of life, then Jones' absentmindedness or his using the steps in the first place will be foregrounded. Conditions that are absolutely unquestionable will never be mentioned as causes. Thus, the fact that Mr. Jones' body has weight, or, in particular, that its specific weight is higher than the specific weight of the atmosphere above the stairs and accordingly obeys the law of gravity, would certainly not be accepted as an explanation for Jones' fall except in circumstances (e.g., science fiction) where weight and gravity are not to be taken for granted. Neither would the fact that Mr. Jones was born be accepted as an explanation for his fall, although his birth is an INUS condition on his fall, and if he hadn't been born, he couldn't have fallen. The closest equivalent to this informal notion of relevance for causal explanations that I could fmd does not come from Gricean and neoGricean pragmatics, but from speech-act theory. This should not be too surprising if we accept that explanation or explaining is a speech act in its own right. Like all illocutionary acts, it could have among its felicity conditions a formulation reminiscent of Searle's classical formulation (Searle 1969: 59)4

(1)

Afelicitous explanation will adduce as a cause for a given event E a condition C that is an INUS condition ofE in Mackie's (1965) sense, where C is a proposition that would not have been true in the normal course ofevents, or the truth of C is not obvious to the hearer H.

20

Paul Georg Meyer

This formulation, as in all cases where Searle uses formulations of this kind, has its problems, however. In particular, it is difficult to determine what a "normal course of events" would be in each case, and what precisely is "obvious" to the hearer. Our discussion above has shown something else. There is, strictly speaking, no such thing as a "normal course of events"; and it is not obvious what could be called "obvious to speaker or hearer"; but there is, for each context of situation, a background (Searle 1983: 141). Propositions which belong to the background are regarded as unquestionable, presupposed, imputed and taken for granted. The background is never thematised except when communication breaks down or illocutionary acts fail. What is part of the background in this sense cannot felicitously be invoked as an explanation (like in most everyday contexts the fact that bodies have weight). We may thus stipulate:

(2)

A felicitous explanation X will adduce as a cause for a given event E a condition C that is an [NUS condition in Mackie's (1965) sense and that is not part of the background (in the sense of Searle 1983) for the description ofE orfor X

This formulation, I think, is quite analogous to what should replace the corresponding felicity conditions in Searlean descriptions of illocutionary acts. A background fact cannot be felicitously made the content of an assertion; it would be a typical case of a trivial assertion. Since explanations are assertions, the above formulation provides only a special case of such a felicity condition. Nor can background facts be made the content of a directive: when you order a table in a restaurant, you don't say you want chairs around it as well (unless it is a type of restaurant where people usually sit or lie on cushions on the floor). And making part of the background the content of a promise ("A happily married man who promises his wife he will not desert her in the next week") "is likely to provide more anxiety than comfort", as Searle (1969: 59) remarked. Now this classical Searlean felicity condition, invoking obviousness and the normal course of events, on most types of illocutionary acts, has usually been equated with Grice's maxim of Relation ("Be relevant", 1975: 46-47), Horn's R-principle (1984) and Sperber and Wilson's principle of RelevanceS (1986, 1995). It is doubtful to me whether this is really the case. True, the Searlean felicity condition is approximately consistent with colloquial meanings of the term relevance. But Grice's

The relevance ofcausality

21

maxim is a maxim of Relation, not of relevance (although its summary uses the term). Grice's maxim of Relation has to do with the requirement that utterances should be related to the previous discourse and the context of situation. In this way, it guarantees coherence of discourse. But this is not what Searle had in mind when he referred to "the normal course of events". Hom's R-principle (1984), however, comes closer to the Searlean notion because it incorporates half of the Gricean Quantity maxim as well. In Gricean terms, utterly trivial statements or explanations violate the Quantity maxim in that they give too little information. In Hom's conception, this submaxim is combined with Grice's Relation maxim to yield a complex R-principle, constituting a lower bound on how much information should be given in an utterance. Sperber and Wilson's (1986) conception of Relevance, although it seems no more than a radicalisation of Hom's R versus Q duality, constitutes yet another concept of 'relevance' because it includes Q- and Rbased considerations from the Hornian model in one formula. Relevance is no longer one of several principles, but the only principle guiding ostensive-inferential communication (Sperber and Wilson 1986: 155158). The R versus Q duality is converted into a quotient of "cognitive effects" divided by processing effort. This means (counter to everyday intuitions about relevance) that high processing costs of a message (measurable in length, complexity, difficulty, amount of noise, etc.) reduce the Relevance of that message. Applying this conception to explanations, it is true that lengthy, complicated, difficult explanations given in noisy surroundings or through a fragile channel are certainly less acceptable to hearers. But they may still be good, felicitous, relevant explanations in terms of the above thesis (2). This is not meant as an objection to Sperber and Wilson's theory-it just points out a difference in conception. But there is another problem that arises in connection with Sperber and Wilson's Relevance theory when equating relevance as discussed here with their Relevance: The Relevance of an event, Sperber and Wilson say (1986: 156-157), correlates positively with the effect of that event on a given context of situation. What, now, is an effect? Doesn't this notion presuppose a notion of causality? The event whose Relevance is to be measured is seen as a cause having some effect on a situation. An event's Relevance is gauged in terms of the relative effect of that event, seen as a cause. And it is a cause that could be adduced in a felicitous causal explanation as defined

22

Paul Georg Meyer

in (2) above. It is a prerequisite for successful communication that an utterance is not just an INUS condition for its contextual effect. It is the very issue discussed by Sperber and Wilson that utterances should be relevant causes of their contextual effects. If other INUS conditions of that same effect (such as the Hearer's knowledge of the language used, absence of noise drowning out the utterance, etc.) were more relevant in a given situation, then the utterance itself would lose its relevance. The problem is: combining Relevance theory and the relevance-based notion of causality put forward here leads us into a vicious circle: we explicate causality in terms of relevance only to find that relevance is explicated in terms of causality! I can presently see two ways out of this dilemma, without being sure which of the two is to be preferred, and without being sure whether they are actually mutually exclusive: The first solution is less clear to me in all its detailed theoretical implications, but there is some evidence in the literature in favour of it. This solution would be to declare causality a semantic and cognitive primitive (cf Ziv 1993: 21). This idea is actually an old one in linguistics, first put forward formally, as far as I know, by postulating an "atomic" semantic predicate CAUSE in generative semantics (McCawley 1968). The "linker" because is also contained in Wierzbicka's (1991: 8, 1996: 70) lists of semantic primitives; though not in the original one (1972: 16). Lakoff's (1982: 163) idealised cognitive model (ICM) of causation adds an interesting aspect to the discussion. Lakoff rejects the notion of primitive in the traditional sense (1987: 279-280),6 replacing it by his own notions of basic-level concept and image schema as the nearest equivalents. Causation is obviously a basic-level concept in Lakoff's terms. The interesting point is that basic level concepts have some characteristics of traditional primitives in that they are "not put together by fully productive principles of semantic composition" (ibd.). But they do or may have internal structure. In this they differ from traditional primitives. This internal structure is spelled out by Lakoff in idealised cognitive models. The elements of Lakoff's model of causation do not mention lawfulness or necessary and sufficient conditions at all. Similar to Von Wright (1973), Lakoff regards agency and the transfer of energy as essential features of prototypical causation. It should be noted that it is a model of causation, not of causality, causation being understood as the nominalisation of an agentive verb. It is plausible that prototypical causation is agentive; and it is quite conceivable that the abstract principle

The relevance ofcausality

23

of causality as used in discursive reasoning should be derived from agentive causation. In this case, the cognitive experience of causality as an abstract principle would be regarded as a notion that is gained by abstraction from prototypical agentive causation. There is also some justification to be found for the primitive analysis, or rather non-analysis, of causality in the philosophical literature: Anscombe ([1971] 1975) seems to advocate a view, summarised by Sosa (1975: 4), that "causation is what it is and nothing else-and that there is no analysis of causation that essentially involves conditionality or lawfulness". And Sosa continues: "So far as I know no one has published a successful analysis of causation by reference to conditionality or lawfulness" (1975: 5). Note that Sosa, too, speaks of "causation" rather than of"causality". If it is true that causality is a primitive concept or very close to a primitive concept of causation, we encounter a familiar phenomenon that is often observable when we try to analyse a primitive concept. Take the example of see. In accordance with Wierzbicka (1996: 78-79) I would argue that it is a primitive concept. An obvious classical definition of see along the lines of genus proximum and differentia specifica would of course be possible in terms of "visual perception" (Searle 1983: 61; Alm-Arvius 1993: 17). I argued in Meyer (1997: 121-122) that visual, and presumably perception, too, cannot be defmed without reference to the primitive concept see. It might just be that "analysing" causality in terms of necessary and sufficient conditions and in terms of relevance or lawfulness is a similar fallacy to "analysing" see in terms of "visual perception". It is quite conceivable that necessary and sufficient conditions are conceptually secondary to causality; material implication as defmed by means of truth tables in logic is an abstract notion far removed from the conceptual lives of ordinary people. It is at least as far removed from everyday concepts of if-then as an INUS condition is from everyday causality. And, as we have seen above, relevance is possibly a notion that cannot do without causality because it has to do with causal impact. As this solution is compatible with Sperber and Wilson's account of relevance and communication, and as Sperber and Wilson's account is grounded in cognitive categories, it would mean that presumably causality is the cement of a cognitively appropriated world, on which both perception and communication are based. The other solution would be to develop a "causality-free" conception of relevance. Maybe the formulation in (2) above is a step in that direc-

24

Paul Georg Meyer

tion. It would be a step "back" towards Grice, Hom, and Searle, away from Sperber and Wilson. It would mean that relevance, and, in consequence, causality, are categories of communication and interaction, not of cognition (Edwards and Potter 1993: 24). For the time being, I am inclined to pursue the second line of reasoning, interpreting causality as a discursive-interactional rather than a cognitive phenomenon. There are two reasons for this: first, this line is more in accordance with my own previous research on discourse (Meyer 1975, 1983, 1996) and the role of causality in it. Second, it seems to be the tum that psychological attribution research is taking at the moment (Edwards and Potter 1993), so that further linguistic research in that direction opens up promising channels for an interdisciplinary exchange of ideas. As a textlinguist who has been workin~ in a coherence relations framework (though under a different name), I have always been convinced that causality has a major role to play in the explication of text constitution and text coherence. But until recently I had tended to believe that causality was a principle "out there", in the world, being referred to and being made use of in discourse. To see research in science, logic, psychology and social sciences converge on the idea that causality is primarily located in discourse does not come as a complete surprise, though, given the discursiveconstructionist tum in social sciences in general (Woolgar 1988; Latour 1987; Bazerman 1988) which has left none of the sacred traditional notions of epistemology untouched. Expressions such as fact, truth, nature or world cannot be used naIvely anymore after taking notice of that discussion (in which they invariably tum up in "scare quotes", and are ridiculed more often than seriously discussed). Nevertheless, this "discursive turn" of the causality discussion sets new tasks for discourse analysts as well. We cannot shift responsibility for the notion of causality to scientists, psychologists, or philosophers any more, but find ourselves as the branch of knowledge most immediately competent for a characterisation of that notion. While I remain sceptical concerning social constructionism in general (Meyer 1997: 6774), I have allowed myself to be convinced by the overwhelming evidence in the case of causality. It remains an open question, though, which is the more appropriate approach, a cognitive or a discursive one. My decision for a discursive approach is preliminary and inspired by research-practical considerations.

The relevance ofcausality

25

4. Causality in discourse So far I have considered the notion of causality as a concept external to linguistics and found it subjected to relevance constraints. This led me to adopt the perspective of causality as a discourse phenomenon in a much more radical way than I had hitherto done. Let me now start from the other end and see what can be said from inside linguistics about interrelations between causality and relevance, in particular from a textlinguistic perspective. Seen from this perspective, causality is not only subjected to relevance constraints, but also creates relevance. This can best be seen from the numerous examples, discussed controversially in the literature, which show that asyndetic clause connections are often interpreted as causal connections, be they juxtaposing as in (3) and (4), coordinating as in (5), or subordinating as the participle construction in (6) (the examples are taken from a popularised social science text):8

(3)

People badly want a demonstration, a dramatisation of justice. They need to have defined and reinforced for them what is right and what is wrong.

(4)

Penal methods by themselves will not put an end to crime. Even at their fiercest, as in nineteenth-century England, ( ..) they did not succeed.

(5)

The participants remain anonymous and there is no one upon whom the authorities can pin the offence.

(6)

The system is more adaptable than the inquisitorial one, allowing for practical changes in treatment methods to be introduced easily during the sentence.

It seems that causality is such a basic principle that very often there is no need to draw special attention to it, and the causal relation between the two clauses is a matter of implicature. Even the direction of causality is left to the reader to infer in such cases. While in (3) to (6), the second clause gives an explanation of some kind relating to the first clause, the causal relationship is reversed in (7). The second clause states a consequence of the fact reported in the first clause:

26

(7)

Paul Georg Meyer

In all the countries concerned the population is growing quickly. All will be faced sooner or later with the spectre of overpopulation.

The precise character of the explanations in (3) to (6) has also been the subject of much debate, but is of little concern for the present contribution (see Dancygier and Sweetser; Pander Maat and Sanders [this volume]). Kortmann (1991: 118-121) has suggested a scale of informativeness for adverbial relations. Causally based adverbial relations (concession, condition, instrument, and cause) are to be found above a dividing line in the middle of Kortmann's scale which separates more informative from less informative adverbial relations, contrast being the only noncausal relation figuring in the top group. Kortmann uses this scale in explaining the semantic interpretation of free adjuncts and absolutes, i.e., participle constructions in English. It could also be used for a ranking of coherence relations. It is obvious, e.g., that a causal relation is more informative than a temporal one because a causal relation implies a temporal relation and only adds further information to this. A standard analysis of examples (3) to (7) within the coherence relations approach would be to find specific coherence relations such as "explanation" or "justification" to describe the relation between the two clauses. Against such analyses much scepticism has been voiced, e.g. recently by Dahl (1995: 259): Causality in discourse is a rather fluid phenomenon whose relations with what we would want to call rhetorical structure are of an indirect character. The view I have argued against-that causality in discourse may be reduced to a small set of rhetorical relations9-has to my knowledge never been stated explicitly, but many treatments of discourse structure certainly give the impression that such a reduction is possible.

I am not quite sure what is meant by "reduction" in the above quotation. I agree with Dahl that the functions of causality in discourse are manifold (see examples (3) to (7) above). But this is no reason why they should not be capable of being reduced to one principle. There is one thing that can be gained by an introduction of a principle of causality into a coherence relations approach: the complexity brought about by the "fluid" character of a fairly large number of different causally-based coherence relations which are not easily delimited from each other may be "reduced" by "reducing" them to a principle of causality. As was also

The relevance ofcausality

27

observed by Ziv (1993: 181), causality could be regarded as one of a small number of basic universal reasoning principles that readers and hearers use in interpreting clause sequences in texts. I think that such principles of text organisation should have the following properties: a. They stem from basic needs arising in human communication. b. They therefore constitute reasons why we should need more than one clause to say what we want to say, that is, motives to utter a text rather than a single clause. c. They represent ways in which clauses, sentences, or larger text chunks can be relevant to each other. d. They are not derivable from a further common principle. Causality fulfils all the above criteria to a high degree: a. It is a basic need in human discourse to explain, to justify, to reason about causes, conditions and consequences. The close affinity between the primitive concept of causation and human action is no coincidence. People want to know about causes, reasons and consequences because they need to act. b. The most natural discourse strategy of explanation or stating consequences is to add another clause. c. The common use of asyndetic clause connections to explicate causal relations in discourse shows that causality creates relevance on its own. To apply criterion d. we need to know what the other principles are. In my own work (Meyer 1983, 1996) I have identified five principles so far that fulfil the above conditions: Topic, Time, Clarification, Causality, and Persuasion. 10 Among these, Causality seems to be the most "relevant" one, both in terms of text frequency, and in number and diversity of coherence relations derivable from them. But the other principles can also be shown to be important organising principles in texts that are not derivable from the principle of Causality. When we speak or write about a certain topic, then usually there is more to be said about this topic than can be accommodated in one clause. Often several topics have to be related to each other. This is how the different forms of topic development arise which in tum give rise to a number of different coherence relations such as "constant topic", "topic progression", "split topic", etc. (cf. Danes 1970). When we narrate a story, there is more than one event to narrate, and these events are primarily and minimally related temporally. This is how time comes in as an organising principle of texts.

28

Paul Georg Meyer

Furthermore, when we describe something, there is often the need to clarify what we are referring to, that is, to spell out in more detail what we wish to say about our topic, to compare, to give examples, to paraphrase. All this cannot be done in one clause and thus, specific coherence relations such as elaboration and contrast arise. And it is clear that very often persuasion, that is, the need to convince the hearers rather than just make them comprehend, can be the motivation for adding another clause, too, giving rise to coherence relations such as "qualification" or "evidence". If what we said above is true, our five principles also represent five ways in which clauses or sentences can be relevant to each other in the Gricean sense: they represent five different ways to fulfil the maxim of Relation in coherent discourse. But the five principles are not on an equal footing here. 11 Kortmann's scale can be applied to them, yielding different degrees of relevance, depending on the way in which a clause relates to the rest of the discourse. The lowest degree of relevance that two sentences can have to each other is brought about by their being related through their topic or topics. Very often this does not suffice for text coherence to be brought about. When a topic is split up into subtopics, or when the text proceeds from one topic to the next, this is only admissible insofar as the subtopics or the newly introduced topics are relevant in terms of some global purpose of the text, or, to put it in Grice's terms: contributions to a coherent discourse must be "appropriate to the immediate needs at each stage of the transaction" (Grice 1975: 47). It is obvious that clarification and persuasion devices are highly relevant, even in Sperber and Wilson's sense, to the purposes of the communication of which they are part. Clarification strategies aim at lowering the processing costs of hearers or readers: they make the text more easily comprehensible. But also in Grice's sense, it may be said that clarification may be the way in which one clause is relevant to another in discourse, serving the "immediate needs", at this "stage of the transaction". Persuasion devices are meant to heighten the probability of a contextual effect of the message on the reader / hearer. They sometimes explicitly invoke relevance as an argument to impress the reader or hearer. In Gricean terms, persuasion devices are applied at a certain stage of a communicative transaction where scepticism is apparent on the part of the hearer, or anticipated by a writer on the part of the reader.

The relevance ofcausality

29

When we narrate events that are in some temporal relation to each other, these events must be also relevant to each other. This is why narrated event sequences are so often causally connected as well. Two events that are temporally related to each other do not necessarily make a good story. But a causal chain of events is much more likely to be accepted as suiting the "immediate needs" at this stage of the narration: its narration may explain why something happened or point out consequences of a certain event. More often than not, the question "What happened next?" aims at learning about the consequences of a certain event rather than about some causally unrelated event that happened to happen next. And informing about the causes of a certain event fulfils a basic need in human communication (see above). Thus, causality seems to be the most relevant of the five principles mentioned. Explicating a causal relationship in discourse is rarely questioned or regarded as pointless. This is no real surprise given that, as we saw in the last section, causality is constrained by relevance considerations. Causality is thus much more intimately connected with relevance than mere temporal sequence (cf Kortmann's scale, 1991: 121).

5. Summary In this paper I have discussed interrelations between causality and relevance. I have tried to show from three different perspectives that there are good reasons to regard causality as a primarily discursive phenomenon. 1. The relevance of a causal statement is not ascertainable objectively, but only in a discourse in which human interests and social obligations play a role. 2. The relevance of a causal statement is best described in a speech-act framework. Its conditions of relevance are more similar to felicity conditions of speech acts than to a cognitive relevance principle. 3. The relevance of causality in discourse cannot easily be overstated. Nevertheless, causality is only one among several principles of text organisation, though obviously the most relevant one. If this is reasonable (and this article, I hope, contains arguments to show that it is), then social-psychological causal attribution research (Edwards and Potter 1993) becomes more important for a characterisation of the concept of causality than logical analyses. This is all the more compelling for linguists, like myself, working in a functionally

30

Paul Georg Meyer

oriented typological framework. Thirty years of experience with this framework have shown that what is likely to emerge as an interesting typological parameter for such a framework, or as a non-trivial semantic-pragmatic universal, is more likely to be found in everyday linguistic usage, everyday reasoning, everyday argument, than in logically purified abstractions of these. Social scientists working on causal attribution have long begun to look to linguistics for answers to some of their questions (Edwards and Potter 1993; Hilton, Jaspars and Clarke 1990; Hilton and Erb 1996). It is time that linguists began to understand these questions and provide more satisfactory answers.

Notes *

1.

2.

3. 4. 5. 6.

This paper is dedicated to Ekkehard Konig on the occasion of his 60 th birthday. I wish to thank the editors, Elizabeth Couper-Kuhlen and Bernd Kortmann, and, in particular, Verena Haser and Manfred Kmg (University of Freiburg, Germany) for extensive (and partly devastating) comments on an earlier draft of this paper. Although I failed to be convinced by some of their arguments, they certainly were an invaluable help in reformulating and supplementing my thoughts and in many cases provided me with further insights into the problems concerned. Some people might say that this would be a reason rather than a cause. I do not wish to enter into the causes-reasons debate here, but if there is a distinction to be made, it would have to be between causes for events and reasons for human actions (Beckermann 1977). Being late is not an action, so whatever causes my being late would be a cause rather than a reason. It is perhaps interesting to note, in this connection, that the German word for 'reality', Wirklichkeit, is etymologically related to the verb wirken 'take effect' which in tum is etymologically related to English work). Literally, Wirklichkeit could be translated as 'a coherent whole of things that take effect'. I apologise for the somewhat stereotypic character of this whole example. I could try to give a causal explanation for this, but I seriously doubt the potential relevance of such an explanation. The formulation for promises (Searle 1969: 59), e.g., is: "It is not obvious to bothS and H that S will do A in the normal course of events". To distinguish Sperber and Wilson's concept of Relevance from others, in particular the colloquial notion of relevance, I will capitalise the word whenever referring to Sperber and Wilson's principle. I wish to thank Verena Haser and Manfred Kmg (University of Freiburg) for pointing this out to me.

The relevance ofcausality

31

7.

Other approaches with a similar set of basic assumptions, are Grimes (1975); Mann and Thompson (1988); Hobbs (1983); Graustein and Thiele (1979) and many others. 8. Klare, Hugh 1., Stress violence and crime, in: Mayne, Richard (ed.): Europe tomorrow, London 1972,48-63. 9. Dahl uses Mann and Thompson's (1988) terminology, where "rhetorical relation" corresponds to what is now called "coherence relation" in most of the literature. 10. I use capital letters to indicate that I am talking about principles within a framework that are not necessarily identical with their colloquial counterparts. 11. Grice himself (1975: 46) speaks about "different kinds and focuses of relevance" whose exact nature is not made very clear. The maxim of relevance is the least explicit one in Grice's account. Maybe the five principles discussed here could contribute to a clarification.

References Alm-Arvius, Christina 1993 The English verb see: a study in multiple meaning. (Gothenburg studies in English 64.) Goteborg: University. Anscombe, G.E.M. 1971 Causality and determination. Cambridge: Cambridge University : Press. 1975 Causality and determination. In: Ernest Sosa (ed.), Causation and conditionals, 63-81. Oxford: Oxford University Press. Bazerman, Charles 1988 Shaping written knowledge: the genre and activity ofthe experimental article in science. Madison, WI: University of Wisconsin Press. Beckermann, Ansgar Grande und Ursachen: zum vermeintlichen grundsiitzlichen Unter1977 schied zwischen mentalen Handlungserkliirungen und wissenschaftlich-kausalen Erkliirungen. Kronberg: Scriptor. Breul, Carsten 1997 Grammatik und Bedeutung der kausalen Satzverbiinde. (Linguistische Arbeiten 368.) Tubingen: Niemeyer. Dahl, Osten 1995 Causality in discourse. In: Gert Rickheit and Christopher Habel (eds.), Focus and coherence in discourse processing, 251-260. (Research in text theory 22.) Berlin: de Gruyter. Danes, Frantisek 1970 Zur linguistischen Analyse der Textstruktur. Folia linguistica 4: 7279. Edwards, Derek and Jonathan Potter 1993 Language and causation: a discursive action model of description and attribution. Psychological review 100: 23-41.

32

Paul Georg Meyer

Graustein, Gottfried and Wolfgang Thiele 1979 Towards an analysis of English texts. Zeitschrift fur Anglistik und Amerikanistik 27: 62-74. Grice, Herbert Paul 1975 Logic and conversation. Syntax and semantics 3: 41-58. Grimes, Joseph 1975 The thread 0.[ discourse. (Janua linguarum. Series minor 207.) The Hague: Mouton. Hilton, Denis 1., Joseph M.F Jaspars and David D.Clarke 1990 Pragmatic conditional reasoning: context and content effects on the interpretation of causal assertions. Journal of pragmatics 14: 791-812. Hilton, Denis 1. and Hans-Peter Erb 1996 Mental models and causal explanation: judgements ofprobable cause and explanatory relevance. Thinking and reasoning 2: 273-308. Hobbs, Jerry R. 1983 Why is discourse coherent? In: Fritz Neubauer (ed.), Coherence in natural-language texts, 29-70. Hamburg: Buske. Horn, Laurence H. 1984 Toward a new taxonomy for pragmatic inference: Q-based and R-based implicature. In: Deborah Schiffrin (ed.), Meaning, form and use in context: linguistic applications, 11-42. Washington: Georgetown University Press. KamIah, Andreas 1991 The causal relation as the most fundamental fact of the world. Comments on Hans Reichenbach's paper: the space problem in the new quantum mechanics. Erkenntnis 35: 49-60. Kortmann, Bernd 1991 Free adjuncts and absolutes in English: problems of control and interpretation. London: Routledge. Lakoff, George 1982 Categories: an essay in cognitive linguistics. In: Linguistic Society of Korea (ed.), Linguistics in the morning calm: selected papers from SICOL-1981, 139-193. Seoul: Hanshin. 1987 Women, fire and dangerous things: what categories reveal about the mind. Chicago: Chicago University Press. Latour, Bruno 1987 Science in action: how to follow scientists and engineers through society. Milton Keynes: Open University Press. Lewis, David 1973 Causation. Journal ofphilosophy 70: 556-567. Mach, Ernst 1901 Die Mechanik in ihrer Entwicklung: historisch-kritisch dargestellt, 4th ed. Leipzig: Brockhaus. Mackie, John Leslie 1965 Causes and conditions. American philosophical quarterly 2: 245-264. 1974 The cement ofthe universe: A study ofcausation. Oxford: Clarendon.

The relevance ofcausality

1975

33

Causes and conditions. In: Ernest Sosa (ed.), Causation and conditionals, 15-38. Mann, William C. and Sandra Thompson 1988 Rhetorical structure theory: toward a functional theory of text organization. Text 8: 243-281. McCawley, James D. 1968 Lexical insertion in a transformational grammar without deep structure. Papers from the Regional Meeting of the Chicago Linguistic Society 4: 71-80. Meyer, Paul Georg 1975 Satzverknupfungsrelationen. Ein Interpretationsmode II fur situationsunabhtingige Texte. (Tubinger Beitrage zur Linguistik. 61.) Tubingen: Narr. 1983 Sprachliches Handeln ohne Sprechsituation: Studien zur theoretischen und empirischen Konstitution von illokutiven Funktionen in 'situationslosen' Texten. (Linguistische Arbeiten. 135.) Tubingen: Niemeyer. 1996 Textfunktion-Textstruktur-Textanalyse: Zur Linguistik des Sachtextes. In: Wolfgang Bomer and Klaus Vogel (eds.), Texte im Fremdsprachenerwerb: Verstehen und Produzieren, 21-44. (Tubinger Beitrage zur Linguistik 418.) Tubingen: Narr. 1997 Coming to know: studies in the lexical semantics and pragmatics of academic English. (Forum fur Fachsprachenforschung 35.) Tubingen: Narr. Reichenbach, Hans 1991 The space problem in the new quantum mechanics. Erkenntnis 35: 29-47. Russell, Bertrand 1912 On the notion of cause. Proceedings of the Aristotelian Society 13: 1-26. 1921 On the notion of cause. In: Mysticism and logic and other essays, 180-208. London: Longman. Searle, John Rogers 1969 Speech acts: an essay in the philosophy of language. Cambridge: Cambridge University Press. 1983 Intentionality: an essay in the philosophy ofmind. Cambridge: Cambridge University Press. Sosa, Ernest 1975 Introduction. In: Ernest Sosa (ed.), Causation and conditionals, 1-14. Oxford: Oxford University Press. Sosa, Ernest (ed.) 1975 Causation and conditionals. Oxford: Oxford University Press. Sperber, Dan and Deirdre Wilson 1986 Relevance: communication and cognition. Oxford: Blackwell. 1995 Relevance: communication and cognition. 2. ed. Oxford: Blackwell.

34

Paul Georg Meyer

Stegmtiller, Wolfgang 1983 Probleme und Resultate der Wissenschaftstheorie und Analytischen Philosophie. Bd. 1: Erkltirung Begrundung Kausalittit. 2., verb. und erw. Aufl. Berlin: Springer. Strawson, P.F. 1985 Causation and explanation. In: Bruce Vermazen and Merril B. Hintikka (eds.), Essays on Davidson: actions and events, 115-135. Oxford: Clarendon. Swales, John 1981 Aspects of article introductions. (Aston ESP research reports 1.) Birmingham: University of Aston in Birmingham / Language Studies Unit. Von Wright, Georg H. 1973 On the logic and epistemology of the causal relation. In: P. Suppes et al. (eds.), Logic, methodology and philosophy of science, 293-312. Amsterdam: North Holland. 1975 On the logic and epistemology of the causal relation. In: Ernest Sosa (ed.), Causation and conditionals, 95-113. Oxford: Oxford University Press. Wierzbicka, Anna 1972 Semantic primitives. (Linguistische Forschungen 22.) Frankfurt: Athenaum. 1991 Cross-cultural pragmatics: the semantics of human interaction. (Trends in linguistics: studies and monographs 53.) Berlin: Mouton de Gruyter. 1996 Semantics: primes and universals. Oxford: Oxford University Press. Woolgar, Steve 1988 Science: the very idea. ChichesterlNew York: Tavistock/Methuen. Ziv, Yael On the rationality of 'relevance' and the relevance of 'rationality'. 1988 Journal ofpragmatics 12: 535-545. Causality in language and in discourse. Journal ofliterary semantics 1993 22: 177-185. Conjunction and causality: pragmatics and the lexicon. Pragmatics 1993 and language learning monograph series 4: 17-27.

On the processing of causal relations Leo G. M Noordman and Femke de Blijzer

This study deals with how different kinds of causal sentences are understood. The central hypothesis is that sentences that more directly reflect a causal relation in the world and in the cognitive representation of this situation are easier to understand than other causal sentences. This hypothesis is tested in reading-time experiments: longer reading times reflect a greater complexity in the processing of the underlying infonnation. Three factors are discussed that detennine how directly a causal sentence matches the cognitive representation of causality: content vs. epistemic relations, linear order of cause and effect in the sentence, and conceptual order of cause and effect. It is demonstrated that each of these factors makes an independent contribution to the complexity of understanding causal sentences. In addition, it is demonstrated that these factors have an effect independently of a fourth factor, namely the causal constraint between cause and effect.

1. Causality in cognition Causality is an important ordering principle of human perception and human experience, and thus a central category in human cognition. It is fundamental both to the representation of human knowledge and to other cognitive processes like predicting, explaining, and comprehending. Our claim is that sentences that more directly reflect causality are easier to understand than other sentences. We should therefore be explicit about what we mean by understanding, and about how we conceive the cognitive representation of causality. From a psycholinguistic point of view, we consider the process of understanding as one in which the reader constructs a representation of the information in the discourse. An important aspect of that representation is that it is coherent, since the consecutive sentences in a discourse are related to each other. Accordingly, the representation contains relations between the sentences, or rather between the information units that are expressed by the sentences. But that representation is also a representation of something else, i. e., of the world. Therefore, the representation has relations with the world. The representation may be related to the world in terms of a number of concepts, such as truth~n which semantic theories have in general concentrated-possibility, and

36

Leo Noordman and Femke de Blijzer

plausibility. The distinction between relations in the representation and relations of the representation to (a model of) the world is related to the distinction made by Guenthner (1989) between D-relations (for discourse) and T-relations (for truth). A similar distinction is made in theories about psychological representations (Johnson-Laird 1983). In this conception, understanding implies that the reader not only constructs a coherent representation but also that the reader implicitly evaluates whether the information in the discourse corresponds to the world, for example, whether it is true or false, plausible or possible. Accordingly, understanding causal sentences requires matching the information in the clauses with a cognitive representation of the world, in this case with a cognitive representation of the causality in the world. This leads to the question of how causality is cognitively represented. We will not develop an elaborate theory of the representation of causality. We only assume that causal relations are represented as ordered causeeffect pairs where the cause temporally precedes the effect. The representation of causal relations originates in our experience in the world. We observe co-occurrences of events in the world. On the basis of these cooccurrences we deduce causal relationships. We have a strong tendency to interpret sequences of events in terms of causal relations, even when there is no real causality involved (Michotte 1954). A child learns that when she drops a glass, the effect is that it breaks into pieces. A child also learns to use causal schemes in reasoning processes: if the glass is broken, it probably has fallen. The experience of causal relations between events in the world is fundamental to the conceptual representation of causality. Given the fact that causal relations derive from our experience of the world, where causes temporally precede effects, we assume that causal relations in our representation of the world reflect this experience and that they are represented as cause-effect pairs with the cause preceding the effect.

2.

Relevant factors in the processing of causal sentences

2.1.

Content vs. epistemic relations

Causal sentences can be characterized with respect to a number of factors, which presumably affect their processing. One factor is the distinction between content relations and epistemic relations (Sweetser 1990), sometimes also identified as semantic vs. pragmatic relations (Sanders, Spooren, and Noordman 1992). A sentence that expresses a causal content

On the processing ofcausal relations

37

relation describes a real-world causality between two events or states in the world. An example is (1): (1)

Because John worked hard, he passed the exam.

An epistemic causal relation expresses a conclusion by the writer/speaker that is based on a causal relation in the world. An example is (2): (2)

Because John passed the exam, he must have worked hard

This sentence expresses that the speaker's knowledge of John's passing causes the conclusion that he worked hard. A relation is a content relation if the two clauses are related because of their propositional content. A relation is an epistemic relation if the clauses are related by the speaker's reasomng. What can we expect with respect to the difference in processing between content and epistemic causals? Earlier, it was said that understanding requires matching the information in the sentence with a model of the world. Since a content causal sentence directly reflects a state of affairs in the world, it may be easier to process than an epistemic causal sentence, which does not directly express real-world causality. One can argue that an epistemic causal expresses a real-world causality in a derived way. In fact, an epistemic relation reflects a line of reasoning that is allowed by the cooccurrence of events or situations in the world. The justification of that reasoning is the contingency of events in the world. So, in this sense, an epistemic relation is based on an underlying content relation. In understanding an epistemic relation, the reader has to check the possibility of the underlying content relation in the world. Understanding an epistemic relation implies understanding the underlying content relation. We will illustrate this with two examples. In sentence (2), the epistemic conclusion is justified by the underlying content relation. The reason that one may infer from "John passed the exam" to "he worked hard" is that ifyou work hard, you pass the exam. In (3), there is the same underlying content relation: if you work hard, you pass the exam.

(3)

John worked hard, so he must have passed the exam.

The content relation is used in an act of concluding, which is expressed by the connective so.

38

Leo Noordman and Femke de Blijzer

According to this analysis, content relations reflect experience in the real world and epistemic relations are derived from these content relations. Epistemic relations express real-world causality in a less direct way. One might predict that processing an epistemic relation requires more time than processing a content relation. There is some support in the literature for this prediction. Noordman (1979) asked readers to read conditional sentences that expressed content and epistemic causal relations. An example of the first type is: "If John is ill, he is not going to work". An example of the second type is: "If John is not going to work, he is ill". The readers had to press a response button as soon as they understood the sentence. The time was measured from the moment the sentence was presented until the participant pressed a button. Sentences expressing a content causal relation were processed 286 msec faster than sentences expressing an epistemic causal relation. So, sentences that express the causal relation more directly are processed faster than sentences that express the causal relation in a less direct way. Similar results have been obtained by Traxler et ale (1997) for the understanding of causal sentences and diagnostic sentences. In their experiment, these two types of sentences corresponded to content and epistemic causals. The interpretation by Traxler et ale is that diagnostic statements need a more complex mental representation than causal statements; diagnostic statements require the representation of an embedding phrase like: "someone believes that. ..". These data show that content and epistemic causal sentences are processed in a different way. A possible interpretation is that understanding causal relations that describe events in the real world is easier than understanding causal relations that express a reasoning process. But we may not yet derive this conclusion. When we scrutinize the sentences that are used in the experiment by Noordman, it is clear that the difference between content and epistemic relations is not the only difference between the sentences. There are other factors that might play a role in the processing of these causal sentences. One factor is what we call the conceptual order of the clauses. Another factor is the linear order of the clauses. We will first discuss these factors: conceptual order in Section 2.2. and linear order in Section 2.3. Finally, we will discuss a factor that will be called causal constraint (Section 2.4.). In Section 3, we will present a reading-time experiment in which these factors are investigated.

On the processing ofcausal relations

2.2.

39

Conceptual order

Consider sentences (1) and (4):

(1)

Because John worked hard, he passed the exam.

(4)

John passed the exam, because he worked hard

Both sentences express a cause-effect relation. Conceptually, the direction in which the causality is expressed is from cause to effect. In the conceptualisation underlying sentences (1) and (4), the effect is derived from the cause. We will refer to this order as the conceptual cause-effect order. There is a correspondence between the antecedent (John worked hard) and the consequent (he passed the exam) in these sentences on the one hand, and the cause and effect in the real world on the other hand. What conceptually is the cause (working hard) is expressed as the antecedent in the sentence; what conceptually is the effect (passing the exam) is expressed as the consequent in the sentence. In these causal sentences, we define the antecedent as the starting point in the causal deduction that is expressed by the sentence and the consequent as the end point. In a content sentence, the antecedent is the cause and the consequent is the effect, as in (1). In an epistemic sentence, the antecedent is the argument and the consequent is the claim, as in (2). Sentences that have the conceptual cause-effect order exhibit what we call conceptual correspondence: they indicate that on the basis of the cause the effect is derived. This order corresponds to the order ofthe events in the world. Sentence (2) expresses a effect-cause relation. (2)

Because John passed the exam, he must have worked hard

Conceptually, the cause is derived from the effect. The effect is considered as a sign for the cause. (Therefore it was identified as an epistemic relation in the first place). Sentence (2) has the conceptual effect-cause order. There is an incongruence between what is cause and effect in the real world and what is expressed in the sentence as antecedent and consequent. It should be noted that the distinction between conceptual cause-effect order and conceptual effect-cause order is not the same as the distinction between content and epistemic relations. In fact, sentence (3) expresses an epistemic relation, but the effect is derived from the cause.

40

(3)

Leo Noordman and Femke de Bli.Jzer

John worked hard, so he must have passed the exam.

Therefore, sentence (3) has the conceptual cause-effect order. On the other hand, the two distinctions are not completely independent of each other, as is shown in Table 1. Content relations only occur in the conceptual causeeffect order, because content relations directly reflect real-world causality. Epistemic relations can have either conceptual cause-effect order or effectcause order. Table 1.

linear order content relation

epistemic relation

Examples of content and epistemic relations, in different conceptual and linear orders. conceptual order cause-effect effect-cause cause-effect effect-cause cause-effect effec~cause Because John worked hard, he passed the exam. John worked hard, so he must have passed the exam.

John passed the exam, because he worked hard.

-

John must have passed the exam, since he worked hard.

John must have worked hard, since he passed the exam.

John passed the exam, so he must have worked hard.

Since sentences with the conceptual cause-effect order reflect causality in the real world more directly than sentences with the conceptual effectcause order, and since we assume that the more directly a sentence matches a situation in the world, the easier it is to process, we predict that sentences with cause-effect order (such as (1)) are easier to process than sentences with effect-cause order (such as (2)). We prefer deriving effects from causes to deriving causes from effects. The basis for this assumption is our sensorimotor experience in the world. We learn the notion of causality and we learn causal relations by acting in the world, by observing that causes precede effects. If the causeeffect order is conceptually more fundamental than the effect-cause order, it seems likely that we prefer to reason from cause to effect instead of from effect to cause. Support for the assumption that conceptual cause-effect order is more fundamental than effect-cause order is obtained by Tversky and Kahneman (1982). They found that reasoning from cause to effect is easier than from effect to cause, even in situations in which the likelihood of the cause given the effect is the same as the likelihood of the effect given the cause.

On the processing ofcausal relations

41

Tversky and Kahneman conclude from their studies that people "infer effects from causes with greater confidence than causes from effects" (page 118). People make use of cause-effect schemas, and causal inferences that follow the normal cause-effect sequence are easier to make than diagnostic inferences that reverse this sequence.

2.3.

Linear order

The conceptual order of the clauses should be distinguished from the linear order. Both sentences (1) and (4), repeated below, express the conceptual order of cause-effect.

(1)

Because John worked hard, he passed the exam

(4)

John passed the exam, because he worked hard

The sentences differ in the linear order of the clauses. In (1), the first clause expresses the cause and the second clause the effect. In (4), the first clause is the effect and the second clause is the cause. Sentence (1) thus has an order which is iconic with the world; sentence (4) is non-iconic. What can we predict with respect to the processing of causal sentences? An obvious prediction is that an iconic order facilitates processing. The reason is similar to the one given in the section on conceptual order. If the order of the clauses corresponds to the order of causality in the world (or more specifically, to a model of causality), the matching process between language and knowledge is easier (and thus the understanding of the sentence). However, on the basis of the literature one might make a different prediction. Magliano et al. (1993) argue that inferences are made only about causes and not about effects. They argue that readers make inferences about causes because these explain the current situation (and the sentence that expresses this situation). The inference forms a backward link. Forward inferences tend not to be made, because, at any point in a text, there are numerous inferences one can make about the possible effects of the situation expressed in the current sentence. Lexical decision data did indeed support the conclusion that inferences about causal antecedents are made but inferences about effects are not. However, in this experiment the causal constraints were rather weak (.30 and .26). We predict that in a situation in which the cause strongly determines the effect and the effect

42

Leo Noordman and Femke de Blijzer

determines the cause, iconicity will facilitate processing of a causal sentence. This prediction rests on the assumption that the cause-effect order reflects the structure of our causal knowledge. These predictions were tested by Noordman, Vonk and Meyer Viol (Noordman and Vonk 1998; Meyer Viol 1984). In the experimental texts, a causal relation was expressed in two different orders. The following fragment is an example: "In order to earn some money, John was cutting out weeds in his mother's garden. It was a tough job, because the stinging-nettles were a metre high. After two hours, he took a short break. The job was half finished. The sun stood high in the sky and it was sweltering hot. He wiped away the sweat from his forehead with his hand. He had touched the stinging-nettles with his hand. His hand itched terribly."

In this fragment the cause sentence ("He had touched the stinging-nettles") preceded the' effect sentence ("His hand itched terribly"). In the other version of the same fragment, the cause sentence and the effect sentence were reversed ("His hand itched terribly. He had touched the stinging-nettles"). It should be noted that the causal relations in both conditions have conceptual cause-effect order. If we make the causality of the relation explicit, we obtain the sequences: "because he had touched the stinging-nettles, his hand itched terribly", and "his hand itched terribly, because he had touched the stinging-nettles". The materials in this experiment were constructed in such a way that, first, the effects could very well be predicted on the basis of the causes and, second, that the causes could very well be predicted on the basis of the effects. This was achieved by two separate pilot studies to test and improve the materials. The experiment showed that a "cause sentence speeds up the processing of a subsequent consequence sentence, but a consequence sentence does not speed up the processing of a subsequent cause sentence... What this experiment demonstrates is a kind of iconicity between cognitive structure and language" (Noordman and Vonk 1998: 205). If the linear order of the clauses is iconic with our model of causality in the world, processing the sentence is speeded up. This has been demonstrated in texts in which the effect is highly predictable given the cause, and the cause is highly predictable given the effect. In the experiment of Section 3, reported more completely in de Blijzer (1999), predictability will no longer be held constant, but will be an independent variable. Summarizing, we have made three distinctions: content vs. epistemic relations, conceptual order, and linear order. In principle, these three distinctions yield a 2x2x2 matrix of eight different kinds of causal sentences

On the processing ofcausal relations

43

(see Table 1). As has already been noted, two cells are empty. A content relation always has cause-effect order. If the conceptual order is effectcause, the relation is always epistemic. In Section 2.1, we discussed an experiment that compared two kinds of sentences that differ on all three factors. What we wish to do in the remainder of this paper is to tIy and find evidence for the factors separately. Also, we would like to further investigate a fourth factor, the predictability of causes and effects.

2.4.

Causal constraint: the predictability ofcauses and effects

A fourth factor that may affect the processing of causal sentences is the predictability of causes and effects in causal relations. How likely is the effect given the cause and how likely is the cause given the effect? In particular, if a speaker draws a conclusion on the basis of an epistemic relation, it is crucial that there is a solid ground for drawing that conclusion and that there are few possible alternatives for the conclusion. More specifically, when the derivation is from cause to effect, it is crucial that there are few conditions that prevent the cause from having an effect. Following Cummins et al. (1991), we will call such conditions "disabling conditions". When the derivation is from effect to cause, it is crucial that there are few other causes. We will call these causes "alternative causes". In the previous experiment these factors were held constant. Sentences (5) and (6) illustrate that disabling conditions affect the plausibility of a sentence that expresses a derivation from cause to effect (i.e., where the effect is derived from the cause).

(5)

The woman must have got a sunstroke, since she was in the sun for a long time.

(6)

The girl must have got wet, since she jumped into the river.

In (5) it is quite likely that there are disabling conditions which could prevent the effect (e.g., wearing a hat, sitting in front of a ventilator, throwing water over your head once in a while or only being in the sun when it is not too hot). So when a person has been in the sun for a long time, it is not at all certain that the effect occurs and that this person does indeed get a sunstroke. This line of reasoning, expressed by sentence (5), is rather implausible. In (6), however, it is rather unlikely that·there are disabling conditions that could prevent the effect. If someone jumps into the river, it is

44

Leo Noordman and Femke de Blijzer

highly likely that this person gets wet. Therefore, when we know the cause we can easily draw the conclusion that the effect must have taken place. When we derive the cause from the effect, the plausibility of a sentence is affected by alternative causes, as is illustrated by sentences (7) and (8). (7)

The woman got a sunstroke, so she must have been in the sun for a long time.

(8)

The girl got wet, so she must have jumped into the river.

In sentence (7) it is unlikely that there are alternative causes for a sun-

stroke. Therefore, if someone has a sunstroke, we can easily draw a conclusion about the cause: this person has been in the sun for a long time. This line of reasoning is very plausible. In sentence (8), however, it is likely that there are alternative causes for being wet (e.g., rain, snow, taking a shower or a bath, a bucket of water). So, when a person is wet, it is not at all certain that this person has jumped into the river. The line of reasoning is not very plausible. It should be noted that these observations do not regard the acceptability of the sentences as linguistic entities (they all are acceptable), but the plausibility of the sentences as descriptions of the world. 2.4.1. Causal constraint in logical deduction

Cummins (1995) and Cummins et al. (1991) have shown that disabling conditions and alternative causes affect logical reasoning. For example, people have difficulty with the following deduction: (9)

If you lie

in the sun for a long time, you get a sunstroke. Mary has been lying in the sun for a long time. So Mary will get a sunstroke.

Relatively many subjects judged such a conclusion incorrect, although it is a logically valid deduction. The interpretation is that there are many disabling conditions. On the other hand, if there are few alternative causes, relatively many subjects made the logically incorrect deduction from effect to cause: (10)

If you lie in the sun for a long time, you get a sunstroke. Mary got a sunstroke. So Mary must have been lying in the sun for a long time.

On the processing ofcausal relations

45

Many subjects judged such a conclusion correct, because there are few alternative causes. Cummins et al. showed that disabling conditions and alternative causes affect judgments on correct and incorrect conclusions. But in this study we are not interested in true/false judgments concerning logical conclusions. We are interested in the understanding of causal sentences. The question then is whether disabling conditions and alternative causes also affect the understanding of causal sentences. Earlier we argued that language understanding implies an implicit evaluation of the sentence on the basis of a model of the world. Therefore, it is quite likely that the understanding of causal sentences is sensitive to factors that affect the validity of conclusions, and therefore that the understanding of causal sentences is sensitive to disabling conditions and alternative causes. For this reason, we incorporated causal constraint, operationalized in terms of disabling conditions and alternative causes, as a variable in the experiment to be reported in Section 3.

2.4.2. Causal scenarios Sentences (5) to (8) illustrate two causal scenarios. The scenario underlying (5) and (7)-if one lies in the sun for a long time, one gets a sunstroke-has many disabling conditions (DC) and few alternative causes (AC); the scenario underlying (6) and (8)-if one jumps into the river, one gets wet-has few disabling conditions and many alternative causes. These two kinds of scenarios will be used to study the role of causal constraint. We collected a set of 71 causal scenarios. In order to find appropriate sentences for the experiment, we presented these scenarios to twenty participants in two judgment tasks. In one task, participants had to assume that the causal event occurred and estimate on a four-point scale how certain the effect was. For example: Suppose that the girl jumps into the river. How certain is it that she will get wet? The higher the score, the higher the causal constraint, the smaller the impact of disabling conditions. The second task was just the reverse. Participants had to assume that the effect event had occurred and to estimate on a fourpoint scale how certain the cause event was. For example, suppose the girl has got wet, how certain is it that she jumped into the river? The higher the score, the higher the causal constraint, the smaller the impact of alternative causes. For each scenario, the mean DC (disabling conditions) and AC (alternative causes) scores over the twenty participants were computed as well as the mean difference between the AC scores and the DC scores

46

Leo Noordman and Femke de Blijzer

(AC-DC). The 24 scenarios with the highest negative AC-DC difference scores and the 24 scenarios with the highest positive AC-DC difference scores were selected for the experiment. These groups of scenarios constituted the two categories of the experimental materials: high on disabling conditions and low on alternative causes on the one hand, and low on disabling conditions and high on alternative causes on the other hand.

3. Experimental study: the processing of causal sentences The aim of this experiment was to find out whether the processing of content relations differs from that of epistemic relations and whether the processing of relations that have cause-effect conceptual order differs from relations that have effect-cause conceptual order. Linear order was kept constant: in all cases the first clause mentioned the effect and the second clause the cause. The experimental conditions in this experiment are illustrated in Table 2. The experiment was conducted in Dutch, but the table presents English translations as well. There is only one type of content sentence: a sentence type that expresses a cause-effect conceptual order. There were three types of epistemic sentences. The fact that a sentence expresses an epistemic relation was marked by the modal must have (in Dutch zal we/), or by the conjunctions so or since (in Dutch: dus and want). In sentence type 3, the epistemic character was indicated only by the conjunction so. The four sentence types were applied to the two categories of scenarios.

On the processing ofcausal relations Table 2.

47

The experimental conditions according to type of relation (content vs. epistemic), conceptual order (cause-effect vs. effect-cause) and causal constraint (scenarios low on disabling conditions/high on alternative causes vs. high on disabling conditions/low on alternative causes).

Causal constraint: low disabling conditions / high alternative causes Relation Conceptual type order 1 content cause-effect Het meisje is nat geworden, omdat ze in de rivier is gesprongen. The girl got wet, because she jumped into the river. 2 epistemic cause-effect Het meisje zal weI nat zijn geworden, want ze is in de rivier gesprongen. The girl must have got wet, since she jumped into the nver. 3 epistemic effect-cause Het meisje is nat geworden, dus ze is in de rivier gesprongen. The girl got wet, so she jumped into the river. 4 epistemic effect-cause Het meisje is nat geworden, dus ze zal weI in de rivier zijn gesprongen. The girl got wet, so she must have jumped into the nver. Causal constraint: high disabling conditions / low alternative causes Relation Conceptual type order 1 content cause-effect De vrouw heeft een zonnesteek, omdat ze lang in de zon heeft gezeten. The woman got a sunstroke, because she was in the sun for a long time. 2 epistemic cause-effect De vrouw zal weI een zonnesteek hebben, want ze heeft lang in de zon gezeten. The woman must have got a sunstroke, since she was in the sun for a long time. 3 epistemic effect-cause De vrouw heeft een zonnesteek, dus ze heeft lang in de zon gezeten. The woman got a sunstroke, so she was in the sun for a long time. 4 epistemic effect-cause De vrouw heeft een zonnesteek, dus ze zal wei lang in de zon hebben gezeten. The woman got a sunstroke, so she must have been in the sun for a long time.

The causal relations were expressed by sentences without context. That is a reasonable first step to find out the extent to which different factors play a role in understanding, e.g., the difference between content and epistemic relations and conceptual order. Further research will be required, for ex-

48

Leo Noordman and Femke de Blijzer

ample, to find out whether the context can overrule the effects of these factors. What kind of reading task is appropriate in this situation? We are interested in understanding causal relations. Understanding requires that the sentence is related to a model of the world. Accordingly, when readers understand a sentence, they should be able to evaluate whether the causal relation expressed by the sentence is a possible causal relation in the world. In the experiment, that was the task the readers were instructed to perform: Is a causal relation possible between the two clauses? Since this task requires that there are also impossible causal relations, so-called filler items are incorporated in the experiment that expressed impossible relations (see Section 3.3.)

3.1.

Procedure

The sentences were presented using a moving window paradigm. In this paradigm, the sentences are presented in three parts: the first clause, the conjunction and the second clause. Only one part is visible at a time; in the other parts, the letters are replaced by hyphens. In the first presentation, only the letters of the first part are visible; the letters of the other parts are hyphens. As soon as the participant has pressed a button, the reading time for the first part is registered and the hyphens of the second part are replaced by letters. In this way, the three parts were read consecutively and it was possible for the reading times for the three parts to be measured separately. The reading times for the third part are especially interesting, because at that moment and only then can the causal relation be understood and does it become clear what kind of relation was expressed by the sentence. Consequently, differences in processing time are expected to occur in the third part. In this experiment, the third part is always the causal clause.

3.2.

Predictions

The four types of sentences (see Table 2) define the four conditions in this experiment. The predictions for the reading times for the identical second clauses (excluding the conjunction) in conditions 1, 2, and 3 are as follows. On the basis of conceptual order, the reading times for conditions 1 and 2 are expected to be shorter than for condition 3. The reading times for

On the processing ofcausal relations

49

the content condition 1 should be shorter than for the epistemic condition 2. This leads to a predicted increase in reading times for the conditions 1 to 2 to 3. The second clause in condition 4 is two words longer than in the other conditions, so these reading times cannot be compared to the reading times for the other second clauses, but the total sentence in condition 4 has the same length (in terms of words and letters) as in condition 2. Based on conceptual order, the reading time for sentences in condition 2 is predicted to be shorter than in condition 4. It is likely that these predictions are qualified by the causal constraint: disabling conditions may make the cause-effect order more difficult, and alternative causes may make the effect-cause order more difficult. More specifically, for the scenarios that have low DC scores and high AC scores, the low DC scores will make the cause-effect items easier, and the high AC scores will make the effect-cause order more difficult. For the scenarios that have high DC scores and low AC scores, the high DC scores will make the cause-effect items more difficult and the low AC scores will make the effect-cause items easier. So the difference between conditions 1 and 2 on the one hand and 3 on the other hand will be greater for the scenarios with low DC scores and high AC scores than for the scenarios with high DC scores and low AC scores. In other words, an interaction is predicted between conceptual order and the two categories of causal constraint. No effect of causal constraint is expected on the comparison between content and epistemic sentences (condition 1 versus 2), because both sentence types have the conceptual cause-effect order. The effect of causal constraint can be investigated in an additional way. The hypothesis is that the cause-effect order becomes more difficult when the DC scores are higher. Consequently, there should be a positive correlation between the DC scores and the reading times for the causal clauses of conditions 1 and 2. The effect-cause order is supposed to become more difficult when the AC scores are higher. Consequently, there should be a positive correlation between the AC scores and the reading times for the causal clauses of conditions 3 and 4.

3.3.

Materials

As was explained in Section 2.4.2., the materials in the experiment consisted of two groups of 24 scenarios each: the group of scenarios with high DC scores and low AC scores, identified as high DC/lowAC and the group of scenarios with low DC scores and high AC scores, identified as low

50

Leo Noordman and Femke de Blijzer

DC/high AC. For each scenario, four sentences were formulated corresponding to the four experimental conditions. Four lists of materials were composed, by pairing each of the 48 scenarios with a different experimental condition in each list, so as to balance the lists with respect to the scenarios and experimental conditions. Although sentence length does not affect the way the hypotheses are tested, it is advisable to keep the sentence length as constant as possible in order to reduce experimental noise. The mean lengths of the causal clauses for the low DC/high AC and the high DC/low AC scenarios were 8.25 and 8.33 syllables respectively. The mean lengths of the complete sentences were also rather similar: 15.21 and 16.21 syllables respectively. Given the fact that the task was to judge whether the causal relation was possible, and all the experimental sentences required an affirmative answer, filler sentences requiring a negative answer were needed. The filler sentences expressed two clauses between which no causal relation was possible. The clauses themselves made perfect sense and were quite possible, so that the negative answer could only be based on the impossibility of a causal relation, as it should be. An example is: "The letters have become blurred because someone held the book upside down", and "Her nose was bleeding because she had a hair cut". There were 32 filler scenarios. The filler scenarios were incorporated in the four lists and were balanced with the experimental conditions in the same way as the experimental scenarios. Each of the four lists was presented to 8 participants. The participants were students ofTilburg University, between 18 and 27 years of age.

3.4.

Results and discussion

We are mainly interested in reading times. Since the data of participants with many errors are not reliable, the reading time data of participants whose error rate was higher than 5 % were not analysed, except in one group in which two participants had an error rate of 7.5 %. Each reading time of a particular reader for a particular item that wa~ extremely high or extremely low in comparison to the average of the participant and the average of the item was replaced by an estimate based on the average of the reader and the average of the item. In a similar way, reading times for incorrect answers were replaced, since they are not informative. The percentage of substitutions was 3.1 %. The reading times for the causal clauses are presented in Table 3.

On the processing ofcausal relations

51

Table 3. Reading times (msec) for the causal clauses.

1 2 3 content epistemic epistemic cause-effect cause-effect effect-cause low disabling conditions / high alternative causes

1567

1730

2029

4 epistemic effect cause + modal

1985

High disabling conditions / low alternative causes

1380

1489

1620

1672

1610

1824

1829

Mean

1473

The reading times for the clauses in condition 1 are a significant 137 msec shorter than for clauses in condition 2: F 1(1,28)=8.44, p. it's the cat's first Christmas, you see she doesn't know that [you get something at Christmas= -

98

Christine Gohl

7?Marie: 8Erik:

[what's going on next year you'll probably have to give her something;'

With her utterance die hat nix kriegt 'she didn't get anything' Marie gives an account for her assessment die katz isch ganz traurig- 'the cat is really sad-' . The following excerpt illustrates the interactive production of an assessment-account sequence. The participants, mother and son, are talking about the car the son has used to drive to his parents' home. This car is obviously not his own, but belongs to the company he works for. (10) AUTO (Schwab 3; 0:27:29) lMarie: < hasch du heut mit dem schone lauto Jahre> J diirfe; 2Erik: ja:: i mu.f3 ja was darschtelle oder? 'IMarie: 2Erik:

«cresc> were you allowed to drive> that beautiful car today; yeah:: I have to keep up appearances, haven't I?'

With her utterance, Marie is doing two things that call for a second action by the recipient. She is asking a question and is also making an assessment of the car the son has used to visit his parents: mit dem schone auto 'that beautiful car'. The content of this question is marked on the prosodic level as something extraordinary: first, during its production the utterance increases in volume; second, there is remarkable pitch movement on auto Jahre, with the pitch falling from a very high level. Marie's question - which constitutes the first part of an adjacency pair - calls for an answer. This is produced by Erik immediately after Marie has finished her tum (ja:: 'yeah::'). The assessment, however, requires a separate acknowledgment. This is provided when Erik gives a reason for what Marie has said (line 2). Note that Erik not only reacts to the content of Marie's utterance, but also to its prosodic design, which conveys a somewhat surprised and approving attitude towards the fact that Erik was allowed to drive this beautiful car. This attitude, however, has a slight ironic undertone, which is taken up by Erik in his utterance: both Marie's question and Erik's answer are thus produced in a teasing, slightly ironic style-judging from content and, especially in Marie's utterance, from the prosody.

Causal relations in spoken discourse

4.4.

99

Accounts after requests

Asking a question is a particular form of requesting something from the recipient, namely the provision of information (cf example (7) FINDINGS, in which the doctor asks for detailed information about the patient's medical history). Consider now example (11). The participants are discussing the appropriate time for smoking and the quality of the cigars a family member has brought along. Erik was offered one of these cigars on an earlier occasion. He now reports on the bad effects it had on him. (11) CIGARS (Schwab 3; 0:0:37) lMarie: do hocksch na un rauchsch st[att daj3 de unsre esse/ 2Erik: [ja sind die zigarre scho was, 3 also mir isch ganz ubel denoch gwase. 4Marie: er==hot==doch==gsagt die beschtn wo==s gibt5 (1.5) was 6Fritz: 7Erik: die zigarre, 'IMarie: 2Erik: 3 4Marie: 5 6Fritz: ?Erik:

then you will sit down and smoke in[stead of our eating our foodl [are these cigars okay, I felt absolutely sick after smoking one. he said the best ones you can get. (1.5.) what the cigars,'

In line 3, Erik gives a reason for asking the question ja sind die zigarre scho was? 'are these cigars okay?'. He asks if the cigars are okay, because he felt sick after smoking one of them, implying with his question that the cigars might not be okay. By offering an account, Erik is giving the recipients some insight into the motive he has for asking the question. In the speaker makes a request in the form of an imperative. The participants are talking about the food, the meat they are having for dinner in particular.

100 Christine Gohl

(12) MEAT (Schwab 8; 15:18) 1Erik: tausche mer

2Kai: gar net 3 [fwe: noi nemm du i will blo) e schtii[ck; == 4Kai: [nein (2 Silben) esse. 5Uwe: i will doch net solche brocken. 'IErik: 2Kai: 3Uwe: 4Kai: 5Uwe:

let's swap «p> yours is smaller.> not true no you take it Ijust want a pie[ce;= [no (2 syl.) eat. 1 don't want such big pieces. '

Erik asks Uwe if he wants to swap pieces of meat. Immediately after this request, he offers a reason for requesting: deins isch kleiner 'yours is smaller'. In his next tum (line 3), Uwe refuses to accept Erik's request (noi nemm du 'no you take it') and accounts for the production of this dispreferred second action in lines 3 and 5: i will blo) e schtiick; 'I just want a piece', and i will doch net solche brocken 'I don't want such big pieces' . Excerpts (11) and (12) show that requests-in these cases formulated as a question and as an imperative-are followed by accounts. Requests are face-threatening activities because they put some obligation on the recipient he or she has to do something for the speaker, either in the form of a verbal action (provide the information requested) or in the form of a non-verbal action (do the thing requested). This format makes requests conversationally sensitive actions-and these are likely to be followed by accounts.

4.5.

Accounts after complaints and reproaches

A common feature of activities like complaining and reproaching is their moral bias. "In these activities interlocutors focus on an infringement of expectations concerning situatively appropriate behaviour and in doing so, demonstrate their own orientation to moral rules and values" (Giinthner 1994: 2). Reproaches make second actions by the person the reproach is directed at relevant, i.e. they call for justifications or apologies (Giinthner 1993: 3-6). In the data analyzed here, however, reproaches are not only followed by such second actions (compare (13) below, where Fritz denies the action he has been reproached for). First speakers, too,

Causal relations in spoken discourse

101

i. e. the persons who are doing the reproaching, offer accounts for reproaching someone or complaining about something. In (13) the topic is once again the family cat. Anna is blaming Fritz for having given his cold to the cat. (13) COLD (Schwab 3; 1:49:38) JAnna: mensch jetz hasch die katz scho angschteckt mit deiner erkiiltung-== der nie:st scho de ganze d[a:g; 2 3Fritz: [des hat er heut morge scho gmacht.

'lAnna: good heavens you've given the cat your cold= 2 he's been sneezing all d[ay 3Fritz: [he did it this morning already.' From the external, visible fact der nie:st scho de ganze dag 'he's been sneezing all day', Anna draws the conclusion that Fritz, who has a cold, must have given it to the cat. She formulates this conclusion as a reproach: both the content of this utterance and the way it is expressed (e.g. the term mensch, which is very often used at the beginning or at the end of an accusatory utterance) give it its reproachful character. Right after this unit, Anna offers an explanation and a justification for the conclusion and the reproach: the cat's been sneezing all day. 16 Reproaches are face-threatening activities, i.e. activities which are likely to threaten both the face of the person the reproach is directed at and the social relationship between the parties. Offering an account after such an action makes a reproach more understandable because it reveals the reasons behind it and gives the person blamed the chance to react to the reproach in a way which is more customized than a mere refusal would be: he or she can exculpate him- or herself by referring to the reason for the reproach. This happens in line 3 in (13): Fritz makes it clear that he refuses to take responsibility for the cat's cold because the cat has been sneezing all morning. He implies that it must have caught the cold elsewhere. It will not come as a surprise that accounts after reproaches emerge not only after a speaker has blamed some other participant for something, but also after complaints about a third party. Consider the following transcript. The content is the same as in CAT I: three family members are talking about the meal they have just had and are considering the 'poor cat', who seems to be sad because it did not get anything.

102 Christine Gohl

(14) CAT II (Schwab 3; 0:07:45) lMarie: i mochtja nix (sage aber) die katz isch ganz traurig2 die hat nix kriegt. 3Fritz: , () 4

. 5Erik: des isch doch dere katz ihr erschtes weihnachte 6 die weifJ gar net daft mer do [was kriegt== 7?Marie: [was los isch; 8Erik: etz niichsch jahr () da muj3 me re vielleicht was gebe; 'lMarie: 2 3Fritz: 4 5Erik: 6 7?Marie: 8Erik:

I don't want to (say) anything (but) the cat is really sadshe didn't get anything. «f> just now I gave her something>, (.) «p> she didn't even look at it>. it's the cat's first christmas you see she doesn't know that [you get something at christmas= [what's going on; next year you'll probably have to give her something;'

With his utterance des hat er net emol aguckt 'she didn't even look at it', Fritz complains about the cat's behavior with respect to his action: he has given the cat some food, but the cat did not appreciate it. Erik and Marie account for this behavior with the utterances that follow. Thus, once again, the causal relation is constructed interactively. Fritz produces an utterance which is treated by a second speaker, Erik, as a consequence: the cat does not appreciate the food because it is her first Christmas. Therefore, you cannot expect her to react in a particular way. The second part of Erik's utterance (die weifJ gar net dajJ mer do was kriegt- 'she doesn't know that you get something at christmas-') provides another potential reason for the cat's behavior. At the same time, another speaker (probably Marie) comes in and completes another explanatory unit (die weifJ gar net was los isch; 'she doesn't know what's going on;') for what Fritz has said earlier (des hat er net emol aguckt 'she didn't even look at it' ). Both explanatory units, Erik's and Marie's, contain reasons for the cat's behavior, as well as consequences with respect to the first part of Erik's utterance (des isch doch dere katz ihr erschtes weihnachte- 'it's the cat's first christmas-'): hence, because it is the cat's first Christmas, she does not know that you receive something on this occasion and she does not know what is going on. The interactive nature of this production of accounts shows that both speakers, as recipients, consider the action of

Causal relations in spoken discourse

103

blaming the cat for not appreciating the food as an action that calls for accounting.

4.6.

Activity types as constraints on the interpretation of asyndetic causa/constructions

In my data, asyndetically linked accounts follow dispreferred second actions and various other types of action which are organized as adjacency pairs in that they make a second-verbal or non-verbal-action sequentially necessary: first assessments, for example, call for second assessments (Pomerantz 1984), requests call for responses in the form of an acceptance or refusal, reproaches for denial or admission (Heritage 1984: 269; Levinson 1983: 308,336). The causal units in the latter cases account for the production of the first actions. However, not all first parts of actions that are organized as adjacency pairs require accounts. Rather, accounting occurs, in particular, after activities which are sensitive to the social relationship between the parties. Requests and reproaches, are, like dispreferred seconds, face-threatening actions-requests, because they impose some obligation on others, reproaches, because they imply criticism of others. Although assessments are not face-threatening in the same way, they are still sensitive to possible disagreement-especially as 'strong assessments'-in that they express a speaker's highly subjective evaluative stance towards a particular fact or event. Accounting after first assessments makes a subjective position more understandable and thus more likely to lead to a preferred second assessment. The analysis of the data shows that participants are sensitive to the production of accounts in these particular environments. Therefore, I would argue that these activity types constrain the interpretations that arise when a causal unit is not explicitly marked as such. An utterance is much more likely to be interpreted as an account, i. e. as an utterance explaining and/or justifying a previous conversational move, if the preceding utterance, by virtue of its sequential and social implications, calls for accounting. Concerning accounts after dispreferred seconds, Heritage (1988: 136) argues that they help to avoid threatening the social relationship between the participants. Accounting for the inability to produce an answer, for example, leaves the responsibility for the dispreferred second with the speaker. Not accounting for a dispreferred second or denying the right of the first speaker to carry out the first action would constitute a threatening act, endangering the participants' relationship. The same holds for the

104 Christine Gohl

production of accounts after reproaches: reproaches are, by their very nature, disaffiliative acts threatening the recipient's face. Giving a reason for such a disaffiliative move makes the act of reproaching more understandable because it reveals the motives behind the reproach. With respect to requests, which, of course, are much less facethreatening than reproaches, it can be argued that a request always puts some obligation on the recipient(s): he or she is asked to do something for the speaker, for example, provide him/her with the information requested. Again, providing a reason for requesting makes the act more understandable because it gives the recipient some insight into the speaker's motives. In sum, the interpretation of an asyndetic construction depends on the features described in sections 2 and 3 and on the sequential environment of the construction in question, i. e. on the activity it can be construed as relating to.

5.

Concluding remarks

This study shows that speakers may give reasons without usmg a conventionalized causal connective element. Different sequential environments were investigated in which speakers produce such asyndetically linked causal relations in order to give accounts. These environments are constituted by different actions which, for various reasons, call for accounting. One group of these activitiesdispreferred seconds, requests, and reproaches-is particularly sensitive to the social relation between participants, either because these activities are face-threatening or because they put some obligation on the recipient. The second group, including first (strong) assessments, is also socially risky. Accounts following strong assessments make the production of a preferred second more likely because they support what is often a highly subjective attitude on the part of the speaker and thus make it easier for recipients and possible next speakers to understand the speaker's stance and react to it in a preferred format. Ford has shown in her studies (1994 and this volume) that accounts or explanations and justifications often emerge in contexts involving some form of contrast, which mayor may not be marked, for example, by contrastive markers or negation particles. The analysis of the asyndetic causal constructions in my data, however, shows that only about half of the accounts occur in contexts involving contrasts. This means that accounts have a much broader context of occurrence. In fact, much

Causal relations in spoken discourse

105

suggests that the production of accounts should be seen as motivated by making reference to the kinds of actions which call for them. This is not to deny that there may be a correlation between linguistic structure and the kind of action that is carried out by an utterance-dispreferred seconds, for example, often contain negation markers-but the correlation is by no means obligatory. Consequently, analysis must go beyond the linguistic realization of an utterance and ask what a speaker is doing with a particular utterance. Related to the issue of structure and action is the question of linguistic practices. Which ones do speakers have at their disposal for the production of reasons? What circumstances influence the choice of linguistic means, especially the choice between explicit conjunctional linkage and asyndetic linkage, when speakers produce accounts for previous actions? First tentative analyses show that one of many possible factors is whether speakers are giving reasons for their own actions or for actions produced by other speakers: asyndetic linkage seems to be preferred to conjunctional linkage in cases of interactively produced constructions. Moreover, particular environments apparently allow only asyndetic linkage. For example, when an account refers to a non-verbal action, there is no previous utterance or implication to which a conjunctional clause could be linked (cf example (6) SCISSORS). These questions, which still need investigation, touch upon the general issue of the linkage between (linguistic) practices and actions, discussed, for example, in ScheglofI (1996, 1997). Further studies along the lines suggested here could bring about a fuller understanding of the relation between various linguistic practices and the action of 'giving reasons' in spoken interaction.

Appendix: Transcription conventions [] overlap [] (.) micro-pause (-), (--), (---) short, middle or long pauses of approx. 0.25-0.75 seconds, up to ca. 1 second (2.0) estimated pause of more than 1 second slurring within units and=uh lengthening, according to its duration hesitation signals, so-called "filled pauses" uh,ah, etc. intonation phrase-final: high rise ? intonation phrase-final: mid-rise intonation phrase-final: level pitch

106 Christine Gohl

/

«

(

»

)

(such)

~ » j,

«t» «p» «all» «cresc»

intonation phrase-final: mid-fall intonation phrase-final: low fall break-off interpretive comments over a stretch of speech unintelligible passage uncertain transcription onllssion of text pitch step up pitch step down forte, loud piano, soft allegro, fast crescendo, becoming louder

Notes * 1.

2.

3.

4.

5.

6.

I am grateful to Dagmar Barth and Susanne Gtinthner as well as to the editors of this volume for their valuable comments on earlier versions. For the connection between the term 'account' and the more generic phenomenon of interactional accountability discussed in ethnomethodology and conversation analysis, see Heritage (1988: 128, 138-141). I view speaking as a form of social action (Heritage 1984: 135) and therefore linguistic utterances as a means for carrying out particular actions. From this premise it follows that coherence relations, e.g. causal relations, that hold between utterances also hold between actions, i.e. an utterance referring to some other utterance does so because of the action(s) carried out by that particular utterance. This applies to utterances to the extent that they consist of one or more turn-constructional units. For the notion of turnconstructional unit (TCD), see Sacks, Schegloffand Jefferson (1974). A first analysis of explicitly linked causal structures compared with asyndetic causal structures shows that explicit linkage allows for more conversational space inbetween accountable action and account. Exceptions are, moreover, deviant cases (Heritage 1988: 133-135), in which an account is produced with a delay, usually when a first speaker does not offer the account immediately but only after a second speaker has elicited its production (see excerpt (5) DECISION in section 4.1.). The translations are meant to convey more the content of the utterances than the stylistic subtleties of the Swabian dialect. Chafe (1988: 1) defines the prototypical intonation unit as "single coherent intonation contour characterized by one or more intonation peaks and a cadence that is recognizable as either clause-final or sentence-final." For this connective function see also Franck (1980: 253) and Lindner (1991: 174). Besides emphasizing their connective function, Ickier (1994: 377) also stresses the interactive role of modal particles: they "anchor particular utterances in larger contexts, which are usually characterized by their

Causal relations in spoken discourse

7. 8.

9. 10. 11. 12.

13.

14. 15.

16.

107

dialogic nature." For the interactive function of the particles doch, eben, and ja in spoken discourse see also Liitten (1979). Occasional remarks on the possible causal meaning of modal particles can be found for doch (Feyrer 1998: 187-192), auch (Paul 1966), halt (Abraham 1983, Hinrichs 1983) eben andja (Abraham 1983). This investigation is based on the analysis of approximately 7 hours of spoken German comprising about 4 1/2 hours of informal conversational data, i.e. family conversations, 1 1/2 hours of interactions in a medical setting, and 3/4 of an hour of radio phone-ins. Most of the data was recorded in the Swabian area of Southern Germany. I wish to thank Ute LacherLaukenikat and project C4 of the SFB 471 for making their data available to me. For the interpretation of asyndetic conditional structures, i.e. paratactic conditionals, see Thumm (1999). For the notion of 'dispreferred second' and a more detailed discussion of accounts after dispreferred second actions, see section 4.2. In fact, in excerpt (3), the utterance'.h .h Ooh I've no idea:.' also constitutes an account: M accounts for her failure to provide J with the information requested (Heritage 1984: 250). Compare also excerpt (6) in section 4.2. For more details, including the marked-unmarked distinction, and a survey of the most important issues concerning preference organization, see Levinson (1983: 332-345). On the use of causal markers in English and Japanese conversations in dispreferred environments, see Ford and Mori (1994). For another example illustrating the production of a dispreferred secondaccount-sequence after a suggestion, see excerpt (1) SMOKING. While in excerpt (6) SCISSORS the declining component is not verbalized, in excerpt (1) SMOKING the speaker formulates the declining of another speaker's suggestion in a very explicit and highly marked way. 'Pock-pock' is presumably the name for a kind of a fowl. For a discussion of assessments and their sequential implications in spoken German interaction, see Auer and Uhmann (1982). Some of their transcripts show that second speakers may explicitly ask for the reason of an assessment given by a first speaker if the latter does not account for the assessment him/herself (Auer and Uhmann 1982: 8-9). The formulation der nie:st scho de ganze da:g ('he's been sneezing all day') falls into the category of Extreme Case Formulations (Pomerantz 1986). Such formulations are used by speakers when they justify complaining about something or somebody to "legitimize a complaint and portray the complainable situation as worthy of the complaint" (Pomerantz 1986: 228).

References Abraham, Werner 1983 Starke Satzverkniipfer-Sinnsog und Partikelenquete. In: Harald Weydt (ed.), Partikeln und lnteraktion, 25-39. Tiibingen: Niemeyer.

108 Christine Gohl Atkinson, J. Maxwell and Paul Drew 1979 Order in Court. London: Macmillan Auer, Peter and Susanne Uhmann 1982 Aspekte der konversationellen Organisation von Bewertungen. Deutsche Sprache 10: 1-32. Bergmann, Jorg R. 1991 Konversationsanalyse. In: U. Flick, E. von Kardoff, H. Keupp, L. von Rosenstiel and S. Wolff (eds.), Handbuch Qualitative Sozialforschung, 213-218. Mfinchen: Psychologie Verlags Union. Blakemore, Diane 1987 Semantic Constraints on Relevance. Oxford: Blackwell. Couper-Kuhlen, Elizabeth 1996 Intonation and clause combining in discourse: The case of because. Pragmatics 6/3: 389-426. Chafe, Wallace 1988 Linking intonation units in spoken English. In: John Haiman and Sandra A. Thompson (eds.), Clause Combining in Grammar and Discourse, 1-27. AmsterdamIPhiladelphia: Benjamins. Feyrer, Cornelia 1998 Modalitiit im Kontrast: Ein Beitrag zur ubersetzungsorientierten Modalpartikelforschung anhand des Deutschen und des Franzosischen. Frankfurt am Main: Peter Lang. Ford, Cecilia E. 1994 Dialogic aspects of talk and writing: because on the interactive edited continuum. Text 14/4: 531-554. Ford, Cecilia E. and Junko Mori 1994 Causal markers in Japanese and English conversations: A crosslinguistic study of interactional grammar. Pragmatics 4/1: 31-61. Franck, Dorothea 1980 Grammatik und Konversation. Monographien Linguistik und Kommunikationswissenschaft 46. Konigstein, Ts.: Scriptor. Gfinthner, Susanne 1993 "Kannst du auch fiber ANDERE Leute LAsTERN." Vorwiirfe als Formen moralischer Kommunikation. Arbeitspapier Nr. 9. MORAL-Projekt. Universitat Konstanz. 1994 Moral voices: Prosodic features of reproaches in "why"-formats. Arbeitspapier Nr 11. MORAL-Projekt. Universitat Konstanz. 1996 From subordination to coordination? Verb-second position in German causal and concessive constructions. Pragmatics 6/3: 323370. Heritage, John 1984 Garfinkel and Ethnomethodology. Cambridge: Polity Press. Hartmann, Dietrich 1986 Semantik von Modalpartikeln im Deutschen. Deutsche Sprache 14: 140-155.

Causal relations in spoken discourse

109

Hinrichs, U. 1983 Konnen Abtonungspartikel metakommunikativ funktionieren? In: Harald Weydt (ed.), Partikeln und lnteraktion, 274-290. Tubingen: Niemeyer. IckIer, Theodor 1994 Zur Bedeutung der sogenannten 'Modalpartikeln'. Sprachwissenscho:ft 19: 374-404. Levinson, Stephen 1983 Pragmatics. Cambridge: Cambridge University Press. Lindner, Katrin 1991 'Wir sindja doch alte Bekannte' The use of Germanja und doch as modal particles. In: Werner Abraham (ed.), Discourse Particles, 163-201. Amsterdam: Benjamins. Lutten, Jutta 1979 Die Rolle der Partikeln doch, eben, und ja als KonsensusKonstitutiva in gesprochener Sprache. In: Harald Weydt (ed.), Die Partikeln der deutschen Sprache. 30-38. Berlin: de Gruyter. Pasch, Renate 1997 WElL mit Hauptsatz-Kuckucksei im DENN-Nest. Deutsche Sprache 3: 252-271. Paul, Hermann [1966] Deutsches Worterbuch, 6.A. Tubingen: Niemeyer. Pomerantz, Anita 1984 Agreeing and disagreeing with assessments: some features of preferredldispreferred turn shapes. In: 1. Maxwell Atkinson and John Heritage (eds.), Structures of Social Action, 57-101. Cambridge: Cambridge University Press. Pomerantz, Anita 1986 Extreme Case Formulations: A way of legitimizing claims. Human Studies 9: 219-229. Sacks, Harvey, Schegloff, Emanuel A., and Jefferson, Gail 1974 A simplest systematics for the organization of turn-taking for conversation. Language 50/4: 696-735 Scheutz, Hannes 1998 Weil in spoken German. Presentation given at the conference on 'Cognitive and Discourse Perspectives on Cause, Condition, Concession and Contrast'. Oktober 1998. UniversiUit Konstanz. Schegloff, Emanuel A. 1996 Confirming allusions. American Journal of Sociology 102/1: 161216. 1997 Practices and actions: Boundary cases of other-initiated repair. Discourse Processes 23: 499-549. Schleppegrell, Mary J. 1991 Paratactic BECAUSE. Journal ofPragmatics 16: 323-337.

110 Christine Gohl

Thumm, Markus 1999 How to do some things without 'if. Eine empirische Untersuchung zu parataktischen Konditionalkonstruktionen im gesprochenen Englisch. MA-thesis, Department of Linguistics, University of Konstanz. Uhmann, Susanne 1998 Verbstellungsvariation in Weil-Satzen: Lexikalische DifIerenzierung mit grammatischen Folgen. Zeitschrift .fur Sprachwissenschaft 17/1: 92-139.

Constructions with if, since, and because: Causality, epistemic stance, and clause order* Barbara Dancygier and Eve Sweetser

English causal and conditional conjunctions show significant overlap both in functions and in the ranges of grammatical constructions in which they occur. These phenomena are related: one reason why different conjunctions can express similar meanings is that the lexical semantics of one conjunction may explicitly involve meaning which is not present in the other conjunction's semantics but can be expressed by accompanying grammatical constructions. Using Mental Spaces theory (Fauconnier 1985/1994, 1997), we analyze the semantics of English if, since, and because, examining the interaction between their lexical semantics and the semantics of their grammatical surroundings. Among the factors involved are predictive function and causal relations (Dancygier 1993, 1998; Dancygier and Sweetser 1996, 1997), epistemic stance (Fillmore 1990a,b), and the relation of the mental space to the speech setting (Sweetser 1990, Dancygier and Sweetser 1996, Dancygier 1998).

1. Introduction In this paper we will analyze the uses of causal and conditional constructions using the conjunctions if, since, and because. These conjunctions specialize in distinct areas of meaning, but their areas of use overlap significantly: a comprehensive analysis, therefore, should explain the overlap as well as the distinctions between them. It has been noted, for instance, that since could substitute for if in (1) without significant change in interpretation, and that a causal relation underlies the contingency described in (2):

(1)

If (as you say) he is in town now, maybe we should phone him?

(2)

If I take my pills regularly,

the symptoms go away.

In (1), imagine that the hearer has said earlier in the conversation that the relevant person is in town now. In the speaker's reference to this previous statement, she could use since instead of if, with quite similar

112 Barbara Dancygier and Eve Sweetser

effect. (2) can naturally be interpreted to mean that taking pills regularly makes the symptoms disappear-a causal interpretation related to readings of because and since. There is thus a significant degree of overlap among the uses of the conjunctions involved. But the potential for such interpretations will often depend on factors other than the conjunction itself; notice that when is a plausible substitute for if in (2) but not in (1), suggesting that the lexical and constructional contexts are important in determinining possible construals of the relations between conjoined clauses. In positing functional overlap between these conjunctions, we are by no means suggesting that they are synonymous--or even that they have specific senses which are synonymous. On the contrary, we will try to specify the source of the similarities as well as to clearly delineate the differences. Both similarities and differences are best accounted for not by focusing only on the conjunctions themselves, but by describing the range of constructions they participate in. The main goal of our analysis is thus to distinguish aspects of constructional meaning which are involved in various uses of each of the conjunctions, and to show how particular formal parts of constructions combine with each other to contribute to varying interpretations of the whole. 1 We will bring up three relevant parameters in trying to compare and contrast the four kinds of constructions. First, we will consider the role of causal models in interpretation. Causality is involved in the constructions in a variety of ways and on various levels of interpretation. Constructional aspects of meaning, and the cognitive domains in which these constructions are interpreted, can be as important as the choice of conjunction, in bringing causality into the interpretation. The conjunctions will also be compared in terms of the type of mental space set-up they engage in; this is often also specifically signalled by the use of verb forms marking negative epistemic stance. Finally, we will review the constructions with respect to their roles in backgrounding or foregrounding elements of interpretation in different clause order configurations. Our analysis of meaning will be framed in terms of Mental Spaces theol)' (Fauconnier 1985, 1997), an extremely general formal theory which provides mechanisms for talking about the ways in which we connect cognitive structures with each other. To some extent, traditional treatments of conditional semantics in terms of Possible Worlds have aimed at capturing similar insights. Somewhat in the way that an ifclause can be said to set up a Possible World within which a then-clause

Constructions with if, since and because

113

is construed, we would argue that an ~clause sets up a Mental Space which is the background for the construal of the then-clause. For example, in If his computer gets repaired, then he'll finish the paper by Friday, the ~clause sets up a space wherein the computer gets repaired, and within that space, the speaker predicts that the paper will get finished by Friday. However, mental spaces are a more general mechanism than possible worlds, referring not only to very partial cognitive "world" or "situation" constructions as well as to more complete ones, but also to a variety of non-world-like structures which can be connected and mapped onto other cognitive structures. For example, the two mental spaces consisting of our mental list of restaurant customers and our mental list of their orders are mapped onto each other to allow utterances such as The ham sandwich wants his check now, wherein the name of the food item ordered can be used to refer to the customer who placed that order (Fauconnier 1985[1994]). World-like structures are not the only consistent, complex mental models which humans develop, and which humans can interconnect with each other. We have exemplified the use of if as a mental space builder. Numerous linguistic devices can play such a role: in 1984, in the picture, possibly, etc. All such expressions require that the content of the clause or sentence be considered in a mental space other than the base space-the reality space of the hearer and the speaker. In what follows we will show how the conjunctions if, since, and because participate in the construction of mental spaces and how the kind of space being set up affects the way the conjunctions are used.

2. Mental spaces, causality, and levels of interpretation We will start this section by outlining types of mental spaces which are particularly relevant to our interpretation of the use of conjunctions. We will then show how these types of mental space set-ups correspond to levels of interpretation of constructions and how causal relations are established at each of these levels.

114 Barbara Dancygier and Eve Sweetser

2.1.

Mental spaces and cognitive domains

Causal and conditional constructions, and particularly if-conditionals, are known to vary widely in function. It would be economical and possibly even elegant to be able to attribute some of this functional diversity to a few specific parameters of interpretation. In fact, mental spaces theory opens up the possibility for such an elegant treatment, by allowing us to talk about different kinds or classes of mental spaces. For example, in a more or less "standard" conditional such as (3),the speaker sets up a space of mental content (cf Sweetser 1990,1996; Dancygier and Sweetser 1996)-a space which is about a possible state of affairs in the world, namely the computer getting repaired. (3)

Ifhis computer gets repaired, he 'llfinish the paper by Friday.

Within this content-domain space (further filled out by the speaker's and hearer's general knowledge about the individuals referred to), the speaker predicts an added aspect of the content of this mental spacethe paper getting finished by Friday. As argued in Dancygier (1993, 1998), prediction in the content domain represents the central use of conditional constructions. "now (not Friday yet)"

Future Space = P (neutral stance) "his computer gets repaired"

Counter-Future Space = ",p "his computer doesn't get repaired"

Figure 1. If his computer gets repaired, he'll finish the paper by Friday.

Constructions with if, since and because

115

As seen in Figure 1, a predictive conditional such as (4) automatically also sets up an alternative space, where P does not hold (no computer repair) and Q also does not hold (no finished paper by Friday). Although this is not a logical property of conditionals, it is a necessary property of the predictive use of conditionals, as we have argued elsewhere (Dancygier and Sweetser 1996, 1997); (3) would not be relevant as a prediction if the paper were just as likely to be finished regardless of the computer repair. By contrast, in an example like (4), the speaker does not predict something about a good Thanksgiving on the basis of information about seeing the addressee before Thursday.

(4)

If! don't see you before Thursday, have a good Thanksgiving!2

Rather, the speaker sets up a discourse context, a speech-act space wherein Thursday has arrived without her seeing the hearer, and then utters a speech act which is to be taken as effective within that space. This can't quite literally be true, since of course the well-wishing is uttered in the current context and can't exactly be retracted; but the conditional certainly modifies the context of interpretation from the previously accepted one. In particular, the hearer cannot now reply with but isn't it a bit early for that? As we might expect from the nonpredictive nature of (4), no alternative space is set up; that is, although the speaker knows she may see the addressee before Thursday, she is not providing for that separate contingency. Her good wishes are offered regardless.

P: S doesn't see H between Monday and Thursday Figure 2. If I don't see you before Thursday, have a good Thanksgiving.

116 Barbara Dancygier and Eve Sweetser

A conditional can thus set up, imagine, and negotiate possibilities in either the world of linguistically described content, or in the world of current speech-act context and performance, where the events and participants are the Speaker, the Hearer, and the speech act with its interpretive context. Speech act conditionals have been noticed as a special and interesting case for some time (Vander Auwera 1986, Sweetser 1990). Mental spaces theory provides a simple mechanism for the description and analysis of the difference between them and content conditionals: conditional constructions may build speech-act spaces as well as content spaces. What other sorts of spaces might speakers construct with conditionals? Sweetser (1990) has argued for a distinction between content conditionals and epistemic conditionals, which follow the speaker's reasoning process in a conditional manner, as in (5):

(5)

If he finished the paper by Friday,

his computer must have gotten

repaired. In the described world of content, the paper's being fInished is not a condition for the computer's being repaired, rather the reverse. But in the epistemic world, where knowledge of effects limits possible conclusions about causes, knowing when the paper was finished is a condition which may be sufficient to warrant an inference about computer repairs. In (5), an epistemic space is set up, wherein the speaker believes or knows that the paper was finished, and it is in that space that she draws the conclusion about the repairs. (Possible paraphrase: 'If I know (or we know) that the paper was finished, then I (we) must conclude that the computer was repaired. ') A further point here is that the speaker of (5) is very possibly not doing initial setup of a space wherein the paper is known to be finished. We might say that she is evoking, rather than actively setting up, that space, if she is borrowing structure set up by an interlocutor who has just informed her that the paper was fInished by Friday. Dancygier (1993, 1998) has set out a further variety of conditional uses, as in (6), wherein metalinguistic negotiation is the subject of the conditional relationship:

(6)

Chris wants you to meet her fiance, him.

if that's

the right word for

Constructions with if, since and because

117

the speaker perhaps thinks that Chris is using fiance as a euphemism for 'live-in partner.' Crucially, the speaker is not here construed as negotiating either the facts of the case (the content), or conclusions to be drawn from them in the epistemic domain; nor is she conditionally performing the speech act of stating that Chris wants the hearer to meet a particular individual. She is apparently metalinguistically commenting on, or presenting as merely conditional, her use of the label fiance to refer to that individual. In this case, we need a somewhat more complex mental space structure, which we have referred to (Dancygier and Sweetser 1996,1997) as a metalinguistic space. A metalinguistic space (generically diagrammed in Figure 3 below) consists of a speechcontext space, within which there is a space of content being discussed and a space of linguistic forms available for referring to that content. Metalinguistic conditionality concerns mappings between the content space and the form space within a particular context. Here, the speaker constructs a particular form-content mapping (uses fiance to refer to Chris' partner), and then sets that in the context of a particular metalinguistic space by calling the mapping into question explicitly (if that's the right word). By implication, such a conditional usage at least suggests that other metalinguistic spaces could be imagined, with different form-content mappings; however, in (6) there is no explicit attempt to evoke other specific mappings (e.g., partner).

Figure 3. Metalinguistic space

118 Barbara Dancygier and Eve Sweetser

Every speech act, of course, involves interlocutors and a context and the interlocutors' ideas of each other's understandings of the content and of the forms used to express it. So all of this mental space structure is present and accessible in any speech situation. However, not all of this structure is necessarily actively at stake in a given conditional use. (6) demands that we bring all of these elements into our diagram overtly, because in a metalinguistic conditional the content-label mappings themselves are what are being conditionally represented and compared. In (4), we needed to overtly represent speech-act context spaces involving S and H to analyze a speech-act conditional, but the contentlabel mappings were not part of the conditional relationship. Rather, they were taken for granted as background to it. A diagram of (5) would need to represent possible epistemic spaces of S, but would take the S/H context and label spaces for granted as background. Our claim, then, is that in all these cases, a conditional construction involves setting up a mental space (in the case of if conditionals, this is the job of the ~clause), and requesting construal of something (in if conditionals, the then-clause or main clause) within that space. Much of the diversity of interpretation can be attributed to the fact that the spaces themselves can be quite diverse sorts of objects, related to the linguistic form in a variety of ways. Crucial to this analysis is the idea that domains related to the current speech interaction are generally privileged with respect to mental space construction (cf Sweetser 1990): they are automatically "accessible" for interpretation, even without explicit mention, in ways that other spaces are not. 3 This fact has wide-ranging effects on interpretation, far beyond conditional constructions. As noted in Sweetser (1990), broad classes of linguistic forms show the same kind of possibility for multiple interpretations depending on the level or domain accessed. Parallel to conditionals are causal and adversative conjunctions (as well as many coordinate conjunction usages, which are beyond the scope of the present paper): (7)

.Joe turned down the stereo because Sam was studying. (Causal relation is between state of affairs described in P and event described in Q, i.e. between contents of the clauses.)

(8)

Sam is (must be) studying, because Joe turned down the stereo. (Causal relation is between speaker's knowledge about content ofP and speaker's conclusion about content ofQ.)

Constructions with if, since and because

119

(9)

Could you turn down the stereo, because I'm trying to study. (Causal relation is between the contextual state expressed in P and the speech-act performance of the request in Q.)

(10)

OK, Chris introduced me to her partner, since we're being politically correct. (Causal relation is between contextual situation P and a particular form-content mapping used in Q.)

(11) Since you're a linguist, what's the Russian wordfor "blender"?

(Speech-act level: the addressee's professional status causes or enables the current speech act of questioning. ) Examples such as these can be multiplied with ease in the domain of conjunction. Some conjunctions seem to prefer certain domains: since in English has preferentially a temporal sense in the content domain4 (Chris has been studying since he came home), but has readily accessible causaVenablement senses in the other three domains (a speech-act domain example is [11]). Ranging farther afield, Hom (1985) has laid out in detail the differences between content and metalinguistic uses of negation. And the well-known contrast between deontic and epistemic modal uses is interpretable in terms of the content/epistemic contrast here invoked. Thus Joe must be home by ten, because I say so imposes a modality of compulsion in the content domain of Joe's getting home. But Joe must be home by ten, because I always see his light go on then involves instead a metaphoric 'compulsion' of the speaker's reasoning processes: 'I am forced to conclude that he gets home by ten, because 1 know that 1 always see his light go on then' (cf Sweetser 1982, 1990, Talmy 1988). What all these examples share is an implicit contribution of the speech setting to the interpretation: some part of the linguistic form (for example, because) need not be interpreted directly with respect to the content expressed by surrounding forms, but can be interpreted as meaning something about the speaker's mental processes or the speech interaction, even though those mental processes or that speech interaction have not been explicitly mentioned. So conditionals are not alone in allowing speakers to implicitly access speech-act, epistemic and metalinguistic structure as well as content, in building an interpretation. As we shall see, however, each of the conjunctions participates in mental space set-up in a somewhat different way.

120 Barbara Dancygier and Eve Sweetser

2.2.

Conjunctions versus constructions: sources of causality in the content domain

As we claimed above, causal relations in content mental spaces are understood in terms of standard causal models, including an active causation model wherein one event or state of affairs actively brings about another. Content-level causality also includes enablement (cf Sweetser [1990]). Causal because occurs in examples like He is tired because he's been up all night, where staying up late is presented as the cause of ensuing tiredness, and also in sentences such as He will come and spend Christmas with us because I'm paying for his flight, where a free ticket enables the person to spend Christmas with his family, rather than causing him to do so. Such content-level causal models constitute a major part of the actual interpretation of constructions containing because and if. 2.2.1. Two ways of depicting causality: because and if Causal conjunctions like because introduce causality or enablement into the interpretation as part of their lexical semantics, although other constructional features and the context help the hearer locate the causal link in the appropriate cognitive domain. (As we will try to show in section 4, intonation patterns and clause order in constructions with because signal content and non-content interpretation of causality in a fairly regular way.) In the cases of non-causal conjunctions, causal interpretations may arise from other aspects of the context and surrounding grammatical constructions: She insulted me and I got angry is easily interpreted as meaning that the insult caused the anger, on the basis of iconic interpretation of clause order, plus real-world knowledge about likely results. Although because lexically expresses causality, the direction of the causal chain is marked by the conjunction's placement in a larger construction: the because-clause presents the cause, the main clause presents the result. Consequently, even though the conjunction seems to be solely responsible for the causal meaning, the way in which causality participates in the interpretation of the sentence is also dependent on the subordination construction. It may initially seem trivial to focus on the obvious fact that the because-clause is marked as representing the cause, as opposed to the caused event, in the causal relation. But as Lightbown

Constructions with if, since and because

121

and Spada (1993) observe, children learning their first language may in fact hypothesize that principles of iconicity override meanings introduced at the lexical level and produce sentences like You took the towels away because I can't dry my hands. In such cases, the strategy used is to present events in their sequential iconic order (the towels disappeared from the bathroom first, and consequently the child had nothing to dl)' her hands with later on). The child needs to learn not only the causal meaning of because, but also the way to use it constructionally to mark causal connections correctly. 5 In many content-domain causal conjunction examples, only the causal relation itself is asserted, while the two events or situations between which the causal relation holds are presupposed background. In the classic I'll do it because I want to, and not because you told me to, the speaker does not assert either that she will do as she is told, or that she wants to do so, or that she was told to do so: but she is necessarily asserting that her actions are caused by her wishes and not by the addressee's commands. .[fclauses generally set up a new mental space distinct from the Base space of the speaker's assumed construal of "reality", but because-clauses do not set up such a space. Main clauses of content-domain because constructions decribe situations which may already be known to hold in the interlocutors' current base space. 6 The because-clause then offers a causal explanation of how that situation came about, also not necessarily on the basis of brand-new information. The causal relationship is at the core of the message. Since is similar to because in being semantically causal, and in that both are often said to presuppose the truth of their complements. These differences interact with other parameters, such as information structure and typical clause order. Since will be discussed in more detail in sections 3 and 4, where we focus on epistemic stance and information structure. The conjunction if presents an interesting case. Unlike because, it is not lexically causal. In the content domain, however, if-sentences typically express causal or enablement relations among events. (12), for example, would typically convey the message that the conditioner will cause the hair to get softer. (12)

Ifyou use this conditioner, your hairwillfeel much softer.

Similarly, in example (3) above, it-is likely to be assumed that the repair will enable the owner to start using the computer again.

122 Barbara Dancygier and Eve Sweetser

Dancygier (1993, 1998) has argued that causal meanings enter the interpretation of content conditionals via conditionals' primary function, prediction. 7 Prediction in conditionals is a type of reasoning which consists in setting up a hypothetical (typically future) mental space and attempting to predict its consequences based on knowledge of typical cause-effect chains and general world-knowledge. In (12) above, the speaker sets up a space in which the hearer applies the recommended conditioner, and predicts in that space that the hearer's hair will become softer as a result. Such causal interpretations tend to arise in conditionals which set up content mental spaces, where causation-based predictions can be made on the basis of assumptions of sequential correlations among events or states of affairs and general knowledge. But the interpretation of speech-act or epistemic conditionals also involves setting up causal relations: the speaker of Since you're a linguist, what do you think of Chomsky? constructs relationships of causality and enablement as part of the background to the act of asking a question. The most notable formal feature distinguishing English predictive conditionals is the combination of verb forms: for future-reference conditionals, a present-tense verb form is used in the If-clause, and the modal will in the main clause (cf Fillmore 1990a,b). This verb form pattern characterizes the whole sentence (not the two clauses in isolation from each other) as representing a certain kind of reasoning. Predictiveness (and, consequently, causality) is part of the interpretation of certain If-constructions as wholes, rather than of the conjunction if or any other lexical element of the sentence. Let us also note that such constructions do not assert any of the clauses. The If-clause cannot express an assertion because the conjunction sets up a hypothetical mental space, different from the base space where assertions could be made. The main clause describes the predicted result in that same hypothetical space, so it is not asserted either. What a predictive conditional asserts is the causal dependency and correlation between the events or states of affairs described by its clauses. To sum up, content-domain conditionals are interpreted causally as a result of predictive uses which are explicitly marked by formal markers such as verb forms. This is one of the basic differences between sentences with because, and those with if. In the former, the conjunction itself introduces causality; in the latter, constructional meaning is the source of the causal interpretation. Though the constructions with because and if are different in a number of respects (verb forms, clause order, mental space set-up, to mention just a few), they can both convey

Constructions with if, since and because

123

information about the causal relations between the content of the subordinate clause and that of the main clause. These differences and similarities are related to other aspects of the constructions, such as their epistemic stance and clause order, which we will consider in the sections to follow.

2.3.

Causality in other domains

As seen above, clauses marked by causal and conditional conjunctions are often used to set up or evoke epistemic, speech act, or metalinguistic mental spaces, in which relations are established among beliefs or speech acts, not events or states of affairs. Does causality play any role in such spaces? Causal models are certainly present in the interpretation of noncontent domain constructions, but we will have to distinguish more than one level of causal links to show how they contribute to the overall interpretation. In the content spaces, causal links can be established between events or states of affairs described. In the other domains, similar links can be established between beliefs or between beliefs and speech acts. In example (8) above (Sam is (must be) studying, because Joe turned down the stereo), the causal relation is between the speaker's knowledge about the content of P and the speaker's conclusion about the content of Q. Similarly, in (11) (Since you're a linguist, what's the Russian word for "blender"?, the belief expressed in the since clause causes or enables the act of asking about Russian. We can interpret metalinguistic uses similarly. For instance, in example (6) above (Chris wants you to meet her fiance, if that's the right word for him) the speaker uses the word fiance while also pointing out that the (possibly mistaken) belief in fiance being the right word is all that enables her to do so. It is known that causal/enablement relations in non-content domains often go in a direction opposite to the causal links in the content domain (Sweetser 1990, Dancygier 1993, 1998). This is especially common in the epistemic domain, where the speaker is often reasoning from effect to probable cause. In (8), for example, we assume that Sam's need to study normally causes Joe to tum his music down, so the fact of the music being turned down enables one to infer that Sam is studying. In fact, it is possible for epistemic sentences to be ambiguous, depending on the underlying assumption about a content causal chain. For

124 Barbara Dancygier and Eve Sweetser

example, a sentence like If Tim went to the embassy, he has his visa documents can have two epistemic interpretations. If the speaker believes that Tim was going to obtain his visa documents at the embassy, she may use the information about Tim's visit there to infer that he now has his documents. If, on the other hand, the speaker believes that acquiring the documents enabled Tim to go to the embassy, she concludes that he had them prior to the visit. In either case, knowledge of cause-effect links in the content domain is used in a reasoning process carried out in the epistemic domain. Causal models can thus participate in the construction of meaning on various levels. This is due not only to our understanding of causality, which allows us to locate it in various types of mental spaces, but to our understanding of the relations among those spaces, and, more generally, to our understanding of speech interaction. This understanding is dependent on what we may see as a nested complex of mental spaces. At the center of this complex is some representation of the content of what is said: events or states of affairs which are referred to by the speaker. The content space need not make any reference to the speaker's reasoning processes or to the speech exchange itself (although these may also be addressed as explicit COlltent: I am asking you why you need an extension on your paper). However, reasoning processes relevant to the speech interaction almost inevitably make reference to the content space: the speaker may, for example, be reasoning specifically about the structure of the content space. In He must be home; I see his coat, the speaker's use of must refers to her understanding that her reasoning processes are under compulsion, not that the described subject's actions are compelled; but the reasoning processes about the act of returning are necessarily based on the speaker's understanding of the content world, where both the return and the coat's presence reside. The structure of the speech exchange will necessarily involve both the idea that some content is expressed, and the understanding that the speaker has reasoning processes related to that content, plus the added understanding that there are a speaker and a hearer interacting in the current speech setting, with all the social interactional issues attendant upon conversational exchange. As we have seen, each of these levels of mental space structure may involve causal models appropriate to the level. But the nesting effect explains how different causal models can enter the interpretation at the same time. This includes the possibility of exploiting a cause-effect chain wherein P causes Q, as the necessary

Constructions with if, since and because

125

background for an inferential chain wherein a belief that Q leads to a conclusion that P.

3. Epistemic stance As noted above, predictive conditionals with future reference typically display a characteristic verb form pattern in English, with a present tense verb in the protasis and the predictive modal verb will in the apodosis. In fact, the same pattern is found in constructions with temporal conjunctions like when: If/when we find the paper, we'll xerox it right away. If and when have similar functions in this example, in that both mark backgrounds to main-clause predictions. The central difference between if and when is that when commits the speaker to the reality of the space described in the when clause, even if that reality will only take place in the future. If, on the other hand, makes no such commitment. This contrast exemplifies the broader phenomenon of epistemic stance (Fillmore 1990a,b), a parameter which is by no means restricted to temporal and conditional clauses. When identifies the speaker's beliefs with P as a description of the real state of affairs (expresses positive epistemic stance); if, on the other hand, expresses the speaker's lack of full positive stance with respect to the content. The non-positive stance of if need not commit the speaker to a negative or skeptical stance, but does indicate that she thereby distances herself from full commitment to the contents of the ~clause. Other aspects of a conditional construction may go further, and explicitly mark the speaker's leaning towards non-belief in the reality of the described situation. Sentences such as If we found the paper, we would xerox it right away, If we had found the paper, then we would have xeroxed it right away, and Even if we found/had found the paper, we wouldn't xerox/wouldn't have xeroxed it right away express negative epistemic stance towards the content of the conditional clause-and hence also towards the content of the main clause, since that is situated in the mental space defined by the conditional clause. The expression of negative epistemic stance is formally marked by the use of verb forms which in earlier work we have dubbed distanced, which pick up the Present + will predictive pattern and backshift it, that is, contribute added layers of past morphology. 8 We have seen examples of positive epistemic stance (with when) as well as neutral and negative epistemic stance (with ij). We noted that

126 Barbara Dancygier and Eve Sweetser

these constructions use a special predictive pattern of verb forms, in both non-distanced and distanced format, and that the distanced forms correlate with negative epistemic stance. Since and because, our two remaining conjunctions, express positive epistemic stance~ the clauses in their scope are seen as representing "factual" or presupposed information. This aspect of their use seems to result from the mental space set-up they engage in; in contrast with conditionals, they do not build new hypothetical spaces, but instead evoke aspects of whatever space is currently the base space. They also do not require any special pattern of verb forms. As one might expect, they do not allow the use of distanced verb forms, which are appropriate only in the cases of hypothetical mental spaces marked with negative epistemic stance. However, not all conditionals are predictive. We have shown above that causality-asserting predictive reasoning (and, consequently, the use of predictive and distanced verb forms) is necessarily characteristic only of conditionals setting up their mental spaces in the content domain. It has been noted (cf Dancygier 1993, 1998; Dancygier and Sweetser 1996, Sweetser 1996) that the use of if in other domains need not involve any restrictions on the use of verb forms, and that such conditionals need not necessarily bring up another alternative space9 . Recall our earlier discussion of speech-act uses of since such as Since you're a linguist, what do you think of Chomsky? A parallel speech-act conditional could serve a similar function: imagine a context where the interlocutor has stated that he is a linguist, and the speaker says If you're a linguist, what do you think of Chomsky? Here the speaker does not intend to compare the space where the addressee is a linguist with other possible spaces, nor to predict the speech act, but only to situate the actually performed speech act in a well-defined space. In such nonpredictive cases, ~clauses may rely on context in ways which are reminiscent of the more positive epistemic stance of since-clauses. Are we now calling into question our previous argument that ifhas a non-positive (neutral) epistemic stance, while since has an overtly positive stance? One might initially feel that (13) expresses a positive stance regardless of whether since or if is used.

(13) If/Since he's (so) hungry (as you say he is), he'll want a second helping.

Constructions with if, since and because

127

Sentences like (13) certainly suggest that the truth of the proposition ('he's hungry') has been asserted in earlier discourse. Rather than setting up a novel space, they evoke a space already contextually accessible. However, referring to a space available in previous discourse does not necessarily mean that the speaker always aligns herself with its content to the same degree. As argued by Dancygier (1998), ~f indicates non-commitment to the clause it marks: but non-commitment covers a wide range of possible attitudes, from strong disbelief to nearcommitment. Even in an example such as (13), where ~fand since seem interchangeable, since sounds more completely positively aligned than if, although in the presence of other strong indications of positive attitude, the degree of non-commitment conveyed by if may be negligible. So there are important differences between if and since. If maintains a neutral stance, while since takes a positive one. Also, since, unlike ~f, has explicitly temporal senses. However, in appropriate contexts, there is predictable functional overlap between the two conjunctionspredictable, in the sense that it falls out naturally from a mental space analysis of the constructions and contexts in question. But in defining the possible contexts of overlap, we need a clear understanding of the factors which make some particular assertion or claim accessible to, or assertable by, some particular participant in the speech interaction. What licenses a speaker to treat a particular content as already accessible to the hearer, or as not being subject to questioning by the hearer? We shall make use of the standard constructs of speech act theory to define contexts where if and since are naturally isofunctional. As will become evident, the relevant factors have a general application to the analysis of the relationship between space-builders and informational structure: space-building background information has to be brought up in a way appropriate to its status in the speech context. When fills a genuinely temporal space-building function, as does since when used to mean 'from some past time to the present.' Nontemporal since, however, builds up causal relations; it states a causal relationship between the space which it evokes (rather than constructs) and the main clause. An ~clause may, as in (13), merely re-enact the building of an already accessible space--or it may build a truly novel space, as in many predictive constructions. A causal since-clause, however, always evokes a space which is presupposed by the speaker, as accessible already to the audience.

128 Barbara Dancygier and Eve Sweetser

Discourse accessibility of a mental space can arise in a variety of ways. One of these ways is the speaker's right to access her interlocutor's prior conversational record, quoting or summarizing or making inferences from it. (14)

Since he's so inefficient, maybe you should make a formal complaint. (The speaker accesses her interlocutor's prior claims about inefficiency, as background to her suggestion.)

Since, with its positive epistemic stance, marks (14) as explicitly accepting the claim about the inefficiency-although of course this acceptance may be insincere or ironic. We could expect the speaker to have used if here, if she were not fully convinced or simply wanted to avoid taking sides. Another source of evocation is the immediate physical context of the exchange. In (15), since is appropriate in two situations: one where the speaker is expressing her own knowledge about the rain, and one in which the hearer has reported the rain to the speaker who has not observed it firsthand: (15)

Since it's raining, we can't go outside after all.

As (15) shows, contextually evident information should normally be marked with since rather than if, unless for some reason doubt is being thrown on its validity. For example, a speaker who is physically examining the hearer's clothing to decide whether it is warm enough or weatherproof enough could say: (16)

Since you're wearing those shoes, we'd better not go to the beach.

But for a speaker considering the possibility of the hearer's changing into sneakers before departure, ifwould be a valid choice. Typically, then, the context and the speaker's evaluation of the situation (and in particular, her epistemic stance towards it) dictate the choice between if and since. Certain categories of information are by their nature presupposed to be known to the speaker, or to the hearer, while others are not. In particular, speaker-based preconditions for a speech act are presumed to be accessible to the speaker, while hearerbased ones are not. Hearer-based preconditions, on the other hand, are

Constructions with if, since and because

129

presumed to be uniquely accessible to the hearer; the speaker cannot treat them as presupposed unless the hearer has already put them on the conversational record. Predictably, the non-presuppositional nature of if contrasts with the presuppositional semantics of since in marking clauses expressing speech-act preconditions. A speaker may hypothesize, but not presuppose, the hearer's willingness to answer a question or comply with a request or tolerate a potentially inappropriate remark:

(17) If/#Since it's not rude to ask, what made you get interested in linguistics? (since would be acceptable only if the hearer had already made it plain that the question was acceptable.) (18) If/#Since you don't mind, could you hold this stack ofbooks for a moment? (since is acceptable only if the hearer has indicated he doesn't mind.) (# is here used to indicate that the sentence so marked is grammatical, but would be unacceptable without special context; one would have to imagine that the addressees of (17)-(18) had just said things like, No, it's not rude to ask about my personal intellectual development, or qr course I don't mind holding somethingfor you.) On the other hand, speaker-based preconditions demand a positive epistemic stance; it is normal for a speaker to presuppose her authority on these subjects. She does not need to hypothesize, normally, about whether or not she is speaking technically:

(19)

Since/#IfI'm speaking technically, I'll call it a misdemeanor and not a crime. (The if form would be acceptable as a general condition, meaning 'Whenever I'm speaking technically, ... ' It cannot, unlike the since form, refer to the current speech act.)

And if the speaker is making a request on the preconditional basis of her own desires, she is presumed to know (and not hypothesize) her stated reasons for those wishes:

130 Barbara Dancygier and Eve Sweetser

(20)

Since/#If I'm stranded without a car, could you drop me at the BART station?

If for some reason the speaker were not the most knowledgeable person about her own access to a car, then if would seem more acceptable in (20). (Imagine a situation where H, a service station employee, has just told her that her husband picked up the car an hour ago, while she was expecting to pick it up and drive home in it.) Speaker and hearer are each presumed to be authorities on their own states of mind and body, and not on each other's. Hence the speaker should normally use since when evoking her own mental and physical states, and ifwhen referring to the hearer's:

(21) If/#Since you're really tired, maybe we should stop work now. (since is acceptable if the hearer has himself indicated his tiredness. ) (22)

Since/#IfI'm totally exhausted, maybe we should stop work now.

By this analysis, since should only be employed when the hearer must concede the speaker's right to take a positive epistemic stance to the content of the space-evoking clause. There remain a variety of reasons to use if rather than since. First, of course, the material may really be one towards which the speaker genuinely does not take a positive stance (truly neutral or negative-stance hypothesizing), or about which the speaker must defer to the hearer (as in hypothesizing about the hearer's tiredness). But inheritance from the conversational record is especially complex with regard to epistemic stance. On the one hand, if the hearer has just said that it is raining, and there is no reason to doubt his word, the simplest scenario may be to adopt it as given and mark it with since. On the other hand, if (cf Akatsuka 1986; Schwenter 1997, 1998) marks a non-positive stance, which may mark genuinely reserved judgement (If she was at· the party, I didn't see her), or may just acknowledge that this claim is not the speaker's own. Quotative if can thus be more polite than since, or less so, depending on whether the issue is acceptance of the contents or taking credit for the claim. How does since interact with the various parameters of interest to us? Obviously, being positive in epistemic stance like when, it does not cooccur with verb forms marking negative stance. Like if, it allows interpretation of the relationship between the two clauses as existing in

Constructions with if, since and because

131

different domains. There are content, epistemic, and speech-act uses of since; and it is easy to find metalinguistic ones as well (cf Sweetser [1990; chap. 4]): (23)

Since my plane was delayed, I'm phoningfrom London. (Content the plane's delay-a situation treated as already familiar to the hearer-is what necessitates the phone call.)

(24)

Since he's phoning from London, his plane must have been delayed. (Epistemic: the presupposed knowledge of the phoning causes the conclusion about the plane's delay.)

(25)

Since you're so smart, when was George Washington born? (Speech act the on-the-floor claim that the hearer is smart is what provokes the query.)

(26)

Since you're so picky, I'll just put another "trout" on the grill. (Metalinguistic: The addressee has just corrected the speaker's use of the term salmon to refer to the wrong species of fish. )

Sweetser comments that since may be preferentially used in the noncontent domains, precisely because of its presuppositional status; there is a relationship between presupposed or given background-clause material and non-content usage (cf. Dancygier 1998), as Nikiforidou (1990) has also pointed out for Greek conditional forms. It is in the noncontent domains that it is often most useful to contextualize a main clause, to make clear on what grounds or against what background assumptions the speaker is performing the present speech act, drawing the present conclusion, or using the present linguistic label. Unlike when, which at the content level can refer to cotemporality at any time relative to the speaker's present, since demands present or past (realis) tenses in its content-level uses. It is not used to make contentlevel predictions, and does not participate in the predictive use of tense; backshifting does not occur in since-clauses, and in general predictive uses of will in since-constructions can be interpreted as expressing prediction at some other level than that of content. This may follow from the fact that a conditional's business is to set up a (possibly entirely unreal) space, as background to some further inferences, while

132 Barbara Dancygier and Eve Sweetser

that of a causal conjunction is to link situations, claims, and conclusions causally to a given background.

4. Clause order, space set-up, and causal/conditional relations Given the flexibility of clause order in conditional and causal constructions, how do different orderings of the main and subordinate clause interact with space building? This is a very general issue indeed: not only if=-conditionals, but most of the broader class of conditionally used constructions, involve conventional use of the order of the expressions of antecedent and consequent. It seems intuitively natural that the space-builder clause should precede the clause which elaborates the space; and this is of course consistent with the observed tendency for if=-clauses to precede their main clauses. 10 However, the other order is certainly also possible, although (as we shall see) it is not possible for all conditionals-the clause containing the apparently space-building (or more neutrally, space-delimiting) material can follow the main clause. Doesn't this violate some basic understanding of how space-structuring proceeds? The answer is that perhaps it might, if the cognitive processing of mental spaces were done in a vacuum; but in fact, context may already have done a great deal of a speaker's building for her, in which case an explicit space-building clause may only be needed to confirm or clarify an already-constructed mental space. Looking first at if=-conditionals, we may note that all three of the following organizations ofP and Q are possible:

If

(27)

ifP, Q

(28)

Q, ifP I'll work at my office, down.

(29)

Q ifP

the home computer breaks down, I'll work at my office.

if

the home computer breaks

I'll work at my office if the home computer breaks down.

(27), diagrammed in figure 4, follows our intuitively basic strategy of first building a space defined by P, then adding the assertion of Q to the structure of that space. Note that the relevant space could still be in

Constructions with if, since and because

133

some sense pre-built by the discourse; (27) could respond to What 'II you do ifyour home computer breaks down?

Set up conditional space defined by P, then assert Q within that space. Space setup first, then contents. There may be a background assumption that P is already a cognitively accessible possibility, but Q is not accessible in this way. (Implicitly set up """'P, ~ space as well.)

Figure 4. If the home computer breaks down, I'll work at my office.

(28) instead asserts Q, without first specifying in what space Q applies. Here it seems almost necessary that some background space be already established, to make sure that the hearer will be adding Q to the relevant space. The speaker can make sure that has been achieved by adding a disambiguating if P of her own, following Q, but this would be an odd way to introduce a genuinely new space. Note that in both (27) and (28) we have standard predictive conditional structure, with the implicit setup of a """'P, ......,Q space alongside the P, Q space; the difference lies not in the relations expressed but in the order of their construction and its relation to the informational context of the discourse. The contrast between (28) and (29) exemplifies a more general contrast between conjuncts juxtaposed with and without comma intonation. (29) would be an answer to Would you ever work at your office? or When will you work at your office?, rather than to What will

134 Barbara Dancygier and Eve Sweetser

if the home computer breaks down? Its major assertion is the identity of the conditional relation under which the working in the office would happen; the main clause content is already "on the floor." In mental space terms, (29) begins by expressing a prediction Q, but in an intonation pattern which makes it clear that Q is not being independently asserted, and is therefore unlikely to belong to the base space. The focal stress (marking new information), and the final intonation fall (marking the end of an assertion), happen only once we are identifying the space P within which Q is to be placed. Q is "old information," while its space-setting is not. So (27)-(29) correspond to different sequences of instructions for building what are rather similar final mental space constructs. (27) is the "canonical" sequence of setting up a space and then making a prediction within it. (28) might be said to involve at least a potential pre-built Pspace, to which the utterance adds Q, subsequently confirming with an ~clause that yes, P was the intended space for elaboration by Q. (29) could be seen as involving a presupposed Q, and asserting the alternative space structure which links that Q with P rather than with ,-..;P. These differing sequences of instructions are useful in different contexts, just the way physical instructions to construct the same object might usefully differ depending on which components of the object were already pre-assembled or ready to the addressee's hand. The three space set-up options presented in (27)-(29) differ in foregrounding or backgrounding causal relations. (27), being a typical predictive conditional, asserts the causal relation (rather than any of the clauses) in the way discussed above. (28) is different in that it initially asserts Q, and thus pushes the nature of the relation between P and Q into the background. In fact, as Dancygier (1998) claims, this clause order in conditionals is perhaps most common among speech act and metalinguistic conditionals, which often perform a speech act in the current space, and use the ~clause to negotiate the background structure of that space. In (29), the stress on identifying the space in which the "old" prediction is valid gives additional focus to the causal relation on which the prediction is based. Dancygier (1998) argues that this clause order and intonation pattern is available only to predictive, causalityasserting conditionals, while all the other types of space-settings are excluded. (29) thus represents a clause pattern which constructionally requires causality to be asserted. 11 Similar clause order options are available for because-constructions, but the preferences are distinctly different. While ~clauses typically

you do

Constructions with if, since and because

135

precede their main clauses, clauses with because typically follow theirs. This regularity (observed by Ford 1993) is influenced by the fact that because-constructions do not set up new spaces, but establish causal relations in the base space. It is thus common for these constructions to first mention the fact requiring explanation, and then establish the cause of the state of affairs in question. Consequently, the main contrast among because-sentences is between two intonation contours available in the Q because P pattern. Sweetser (1990, 82ff) discusses parallel examples involving causal conjunction, as examples of Chafe's (1984) generalization about "bound" (commaless) and "unbound" (comma) intonation: comma intonation demands an interpretation involving assertion of both the main clause and the causal connection between the clauses, while commaless intonation is readily given an interpretation which presupposes the assertion (in this case, prediction) embodied in the main clause, and newly asserts only the causal connection between the clauses. Thus (30) is likely to be understood as asserting that the reason for Anna's love is Victor's resemblance to her first love, but presupposing that Anna loves Victor; (31), on the other hand, asserts that Anna loves Victor, as well as the reason why. (30)

Anna loves Victor because he reminds her o..f herfirst love.

(31)

Anna loves Victor, because he reminds her ofherfirst love.

Sweetser comments that the bound interpretation is unlikely to occur in conjunction with a speech-act or epistemic reading of the causal relationship between the clauses. Thus, (31) but not (30) is readily susceptible to non-content causal readings-for example, to conveying the message that the inference about Anna's love for Victor is caused by knowledge about a resemblance to her earlier sweetheart, rather than the content-level message that the love is caused by the resemblance. Sweetser's explanation for this contrast is that it would be exceptional to presuppose the main clause of a speech-act or epistemic conditional, since this would involve presupposing the very speech act being performed, or the new conclusion being drawn. The mental-space structures of (30) and (31) above are similar to the conditional examples (28) and (29). The main difference between the causals and conditionals is the preference for PQ clause order in conditionals and QP clause order in causals, which appears to be a

136 Barbara Dancygier and Eve Sweetser

consequence of the differing space-building functions of the two conjunctions. Space-building background has a natural reason to precede the content of a newly-built space, while a base-space state of affairs in need of explanation may well reasonably be described first, before giving the explanation for why or how it came to be that way. In contrast with because and when, since seems awkward in the commaless Q conjunction P construction: #1'1/ marry you since you're so sweet, or #He invited me since I live next door are strange, though it seems normal to say Since I live next door, he invited me or I'll marry you, since you're so sweet. Since does not seem to allow foregrounding or focus on the assertion of the causal relation, perhaps does not even assert it 12 . In this, since contrasts with a wide range of temporal conjunctions (before and after behave like when) as well as with because and if. Compare the following examples of conjunctions in constructions which employ stress or syntactic focalizing structures to mark the focus of contrast:

(32) I love you because you're sweet, not because you're rich. (33)

We'll eat when Sue arrives, not now.

(34)

They had coffee after lunch, not before dinner.

(35)

#He invited me since I live next door, not since I'm his boss.

(36) It's because you're sweet that I love you, not because you're rich. (37) It's when Sue arrives that we'll eat, not now (/not when Joe arrives). (38) It's after lunch that they had coffee, not before dinner.

(39)

#It's since I live next door that he invited me, not since I'm his boss.

Various causal and adversative conjunctions fall on one side or the other of this contrast: though and while behave like since, and phrases like in spite of the fact that or despite the fact that behave more like because. Interestingly, since in its temporal sense does allow assertion and focus of the temporal relation:

Constructions with if, since and because

137

(40) It's since the pollen came out that I've been sniffling, not since I moved to the new apartment. (41) I've been sniffling since the pollen came out, not since I moved to the new apartment. Of the two causal conjunctions we have looked at only one, because, readily allows focus on the causal relationship itself Interestingly, another conjunction which can foreground the causal link in this way is if, in its predictive use. A look at clause order and intonation options thus reveals further similarities and distinctions between the constructions under scrutiny. Both if and because constructions allow the use of sentence-final adverbial clause placement and commaless intonation to unambiguously assert the causal relation between the contents of the clauses, while the "comma" intonation usually renders the sentences ambiguous between domains. On the other hand, since does not seem to allow foregrounding or focus on the assertion of the causal relation. These facts are added evidence of the varying degrees to which asserted, presupposed, and construed causality are integral parts of building these constructional meanmgs.

5. Conclusion The functions of causal and conditional constructions can be compared along a variety of parameters. We considered both the contribution of conjunctions as lexical items, and the constructions in which these conjunctions appear. Particular conjunctions participate in mental space configurations in differing ways: some conjunctions set up their own spaces, some evoke spaces set up in preceding discourse, some merely elaborate the base space of the ongoing interaction. Further, important differences arise with respect to the cognitive domains in which the spaces are set up, but significant aspects of the overall interpretation result from the spaces being nested one within each other. Such parameters of interpretation correlate significantly with formal features of constructions, such as the choice of verb forms, clause order, or intonation. Some of these formal features may be claimed to invite certain types of mental space set-up or contextual use, regardless of the choice of conjunction. One good example of such a feature is clause

138 Barbara Dancygier and Eve Sweetser

order and accompanying intonation, since we observe recurring patterns of interpretation in constructions which use the same clause order and intonation patterns. It is often difficult to isolate the study of a small group of linguistic forms. Temporal and adversative conjunctions, which were not originally meant to appear in our data, had to be mentioned along with if or since; otherwise, important facts about the behavior of our primarily chosen conjunctions would have to be missed. As we have also claimed elsewhere (Sweetser 1990; Dancygier 1993, 1998) coordinate conjunctions and even conjunctionless and paratactic constructions 13 also share some constructional characteristics and aspects of interpretation with if, because, and since. What is needed is an analysis which uses parameters of constructional meaning (verb forms, clause order, intonation, use of mental space builders) to outline the range of constructions which participate in the construal of related meanings (causality, sequentiality, conditionality), and explores the similarities and differences between the constructions with respect to these parameters. We hope to have shown that such an analysis can reveal generalizations and aspects of constructional meaning which accounts of individual conjunctions can describe only partially.

Notes * 1.

2.

3.

4.

5.

For Ekkehard Konig, on his sixtieth birthday. It is impossible to cite all the works which have helped us to develop an understanding of the semantics of grammatical constructions. Among salient ones are Fillmore, Kay and O'Connor (1988), Fillmore and Kay (1999), Langacker (1987, 1991a, 1991b), and Goldberg (1995). Sweetser (1999) deals generally with the relation between lexical meaning and constructional composition. We are grateful to the late Suzanne Fleischman for this example as well as for characteristically insightful comments on it. By "automatically accessible," we mean that speakers do not need any special shared experience, knowledge or context besides the fact of the speech exchange, in order to have implicit access to these spaces. Though see section 3 for a full discussion of uses of since in different domains. Traugott (1982) gives an illuminating discussion of the differences between earlier stages of English, wherein both clauses in a causal relation were marked, and the modern English situation.

Constructions with if, since and because

6. 7. 8.

9.

10.

11. 12.

13.

139

Or, of course, in whatever space is currently being added to; the crucial fact is that no new space need be set up when because is used. For discussion of the relationship of then to predictive function, see Dancygier and Sweetser (1997), Schiffrin (1992), and Iatridou (1991, 1994). Conditionals with distanced form have been discussed extensively as "counterfactual" or "unreal". The discussion of this issue exceeds the limits of the present paper. The reader is referred to Sweetser 1990, Dancygier 1993, 1998, Dancygier and Sweetser (1996, 1997). For reasons why distanced conditionals should not be treated as counterfactual, see Comrie (1986); for discussion of the crosslinguistic connection between past tense forms and distanced forms, see James (1982) and Fleischman (1989). Cf. Dancygier and Sweetser (1996, 1997), for discussion of predictive and non-predictive epistemic and metalinguistic conditionals, and for analysis of reasons why content conditionals are necessarily predictive and speech-act conditionals necessarily non-predictive (see also Sweetser (1990), Dancygier (1998). Cf Haiman 1980, 1986; also Ford 1993 on because vs. if. All the patterns just discussed are also found among even if constructions, with the same priorities and restrictions involved. As Elizabeth Couper-Kuhlen has pointed out to us, this incompatibility with non-comma intonation brings us back to Sweetser's (1990) original concern about the preference of since for non-content uses, and raises doubts as to whether examples like (23) should really be treated alongside other content examples. For discussion of parallels with paratactic conjunction, see Lakoff (1971), Grice (1978), and Sweetser (1990).

References Akatsuka, Noriko 1986 Conditionals are context-bound. In: Traugott et al. (eds.), 333-52. Chafe, Wallace L. 1984 How people use adverbial clauses. BLS 10: 437-449. Comrie, Bernard 1986 Conditionals: a typology. In: Traugott et al. (eds.), 77-99. Dancygier, Barbara 1998 Conditionals and prediction: time, knowledge, and causation in conditional constructions. Cambridge: Cambridge University Press. 1992 Two metatextual operators: negation and conditionality in English and Polish. BLS 18: 61-75. 1993 Interpreting conditionals: time, knowledge and causation. Journal of Pragmatics 19: 403-434. Dancygier, Barbara and Eve Sweetser 1996 Conditionals, distancing and alternative spaces. In: Adele Goldberg (ed.). Conceptual structure, discourse and language, 83-98. Stanford CA: CSLI Publications. 1997 Then in conditionals. Cognitive Linguistics 8/2: 1-28.

140 Barbara Dancygier and Eve Sweetser Davis, W. 1983

Weak and strong conditionals. Pacific Philosophical Quarterly 64: 57-71. Fauconnier, Gilles 1985 Mental spaces. Cambridge MA: MIT Press [Reprinted 1994, Cambridge: Cambridge University Press]. 1996 Analogical counterfactuals. In: Fauconnier and Sweetser (eds.), 5790. 1997 Cognitive mappings for language and thought. Cambridge: Cambridge University Press. Fauconnier, Gilles and Eve Sweetser (eds.) 1996 Spaces, worlds and grammars. Chicago: University of Chicago Press. Fillmore, Charles J. 1990a The contribution of linguistics to language understanding. In: Aura Bocaz (ed.). Proceedings of the first symposium on cognition, language, and culture, 109-128. Universidad de Chile. 1990b Epistemic stance and grammatical form in English conditional sentences. CLS 26: 137-162. Fillmore, Charles J., Paul Kay, and Mary Catherine O'Connor 1988 Regularity and idiomaticity in grammatical constructions: the case of let alone. Language 63/3: 501-538. Fillmore, Charles and Paul Kay 1999 Grammatical constructions and linguistic generalizations: the what's X doing Y? construction. Language 75/1: 1-33. Fleischman, Suzanne 1989 Temporal distance: a basic linguistic metaphor. Studies in Language 13/1: 1-50. Ford, Cecilia E.

1993

Grammar in interaction: Adverbial clauses in American English conversations. Cambridge: Cambridge University Press. Goldberg, Adele 1995 Constructions. A Construction Grammar approach to argument structure. Chicago: University of Chicago Press. Grice, H. P. 1978 Further notes on logic and conversation. In: Peter Cole (ed.), Syntax and Semantics, Vol. 9: Pragmatics, 113-127. NY: Academic Press. Haiman, John 1980 The iconicity of grammar: isomorphism and motivation. Language 56: 515-40. 1986 Constraints on form and meaning of the protasis. In: Traugott et al. (eds.), 215-227. Hom, Larry 1984 Toward a new taxonomy for pragmatic inference: Q-based and Rbased implicature. In: Deborah SchifIrin (ed.), Meaning, form and use in context: linguistic applications (GURT 1984), 11-42. Washington, D.C.: Georgetown University Press.

Constructions with if, since and because

141

Metalinguistic negation and pragmatic ambiguity. Language 61: 121-74. Iatridou, Sabine 1991 If then, then what? NELS 22: 211-225. 1994 On the contribution of conditional then. Natural Language Semantics 2: 171-199. James, Deborah 1982 Past tense and the hypothetical: a crosslinguistic study. Studies in Language 6: 375-403. Konig, Ekkehard 1986 Conditionals, concessive conditionals, and concessives: Areas of contrast, overlap and neutralization. In: Traugott et ale (eds.), 22946. Lakoff, Robin 1971 Its, and's, and but's about conjunction. In: Charles 1. Fillmore and D. Terence Langendoen (eds.), Studies in Linguistic Semantics, 114149. New York: Holt, Rinehart and Winston. Langacker, Ronald W. 1987 Foundations ofcognitive grammar, vol. 1: Theoretical prerequisites. Stanford CA: Stanford University Press. 1991a Foundations ofcognitive grammar, vol. 2: Descriptive applications. Stanford CA: Stanford University Press. 1991b Concept, image, symbol: the cognitive basis of grammar. Berlin: Mouton de Gruyter. Lightbown, P.M. and N. Spada 1993 How languages are learned. Oxford: Oxford University Press. Nikiforidou, Kiki 1990 Conditional and concessive clauses in Modern Greek: A syntactic and semantic description. Ph.D. dissertation, University of California at Berkeley. Schiffrin, Deborah 1992 Anaphoric then: aspectual, textual, and epistemic meaning. Linguistics 30: 753-792. Schwenter, Scott 1997 The pragmatics of conditional marking: Implicature, scalarity, and exclusivity. Ph.D. dissertation, Stanford University. 1998 From hypothetical to factual and beyond: refutational si- clauses in Spanish conversation. In: Jean-Pierre Koenig (ed.), Discourse and cognition: Bridging the gap, 423-436. Stanford CA: CSLI Publications. Sweetser, Eve 1982 Root and epistemic modals: causality in two worlds. BLS 8: 484-507. 1990 From etymology to pragmatics. Cambridge: Cambridge University Press. 1996 Conditionals and mental spaces. In: Fauconnier and Sweetser (eds.), 318-333. 1985

142 Barbara Dancygier and Eve Sweetser

1999

Compositionality and blending: semantic composition in a cognitively realistic framework. In: Theo Janssen and Gisela Redeker (eds.), Cognitive Linguistics: Foundations, scope and methodology, 129-162. Berlin: Mouton de Gmyter. Talmy, Leonard 1988 Force dynamics in language and cognition. Cognitive Science 2: 49100. Traugott, Elizabeth Closs 1982 From propositional to textual and expressive meanings. In: Winfred Lehmann and Yakov Malkiel (eds.), Perspectives on Historical Linguistics, 245-271. Amsterdam: Benjamins. Traugott, Elizabeth Closs, Alice ter Meulen, Judy Snitzer Reilly and Charles A. Ferguson (eds) 1986 On conditionals. Cambridge: Cambridge University Press. Van der Auwera, Johan 1986 Conditionals and speech acts. In: Traugott et al. (eds.), 197-214.

On affirmative and negative complex conditional connectives* Estrella Montolio

This paper describes the characteristic features of clauses introduced by complex conditional connectives (CCCs). Specifically, I analyze the grammatical (§2), semantic (§3, §5, and §6) and pragmatic (§7) properties of these clauses which distinguish them from ij=-conditionals. CCCs are organized into two subsets that stand in complementary distribution: (i) affirmative CCCs and (ii) negative CCCs. These subsets show essentially the same grammatical behavior but differ in terms of their semantics (§ 1 and §4) and their pragmatics (§7). The data are drawn principally from Spanish, but the results of the analysis are, in general, equally applicable to English and French.

1. Preliminaries Though it is true that the literature on conditionals is quite extensive, conditional structures other than !fclauses have traditionally attracted little attention from linguists. This is most definitely the case with what might be called complex conditional connectives (Dostie 1987; Leard 1987; Montolio 1991b, 1999; Visconti 1994), which have received anything but extensive treatment. Perhaps this has to do with the general impression that complex conditional connectives (henceforth, CCCs) are not only equivalent to one another but are also completely parallel to !fclauses. 1 The goal of this paper is to refute such an idea and demonstrate how mistaken it actually is. Contrary to the view of equivalence, I shall show that CCCs cannot appear in the majority of the contexts in which it is possible to use !fclauses. As we shall see, this is because CCCs express a much more specific and restricted relationship between the protasis and the apodosis than does if. I concentrate mostly on Spanish data, but I believe that, for the most part, my observations are equally valid for both Romance and Germanic Languages. 2 The connectives that we shall include in the category of CCCs correspond to two complementary subsets (cf Montolio 1990, 1991b,

144 Estrella Montolio

1999). The first group, which we shall call the affirmative group (or ACCCs), is made up of constructions such as:

(1)

a. a condici6n de que (= French: acondition que) 'provided that', 'on the condition that', 'so long as' b. con tal de que (= French: pourvu que) 'provided that', 'so long as' c. siempre que 3 (= French: pourvu que) 'provided that', 'so long as'

Expressions such as the following are included in the second, negative group (or NCCCs). (2)

a. a menos que

'unless' b. a no ser que

'unless' salvo que 'except if d. excepto que 'except if

c.

amoins que) (= French: amoins que)

(= French:

(= French: saufsi)

(= French: saufsi)

Spanish conditional connectives excepto si ('except if) and salvo si ('except if) are semantically analogous to the other connectives included in this group (2), but, as if-compounds, they can combine with either the subjunctive or the indicative. Given the fact that the rest of the complex connectives to be analyzed here all trigger the subjunctive mood, 1 have preferred not to pursue their description any further here. Because these two subsets stand in complementary distribution, any member from the ACCC group can appear in the same contexts as all the others of this group but in none of the contexts required by NCCCs. The complementary distribution of these subsets is due to a difference in semantic content, which explains the pragmatic inadequacy of these connectives when an ACCC is replaced by an NCCC (or vice versa) within the same context. This can be seen in examples (3b) and (4b) below. (I use the #sign to indicate that the utterance under consideration is grammatically correct but pragmatically inadequate).

Complex conditional connectives

145

(3)

a. Ire con vosotros al cine a condici6n de que/con tal de que/siempre que me acompaneis luego a casa en coche. 'I'll go with you guys to the movies provided that/so long as you take me home afterwards by car. ' b. # Ire con vosotros al cine a no ser que/a menos que me acompaneis luego a casa. '# I'll go with you guys to the movies unless you take me home afterwards by car. '

(4)

a. No ire con vosotros al cine a no ser que/a menos que me acompaneis luego a casa en coche. 'I won't go with you guys to the movies unless you take me home afterwards by car. ' b. # No ire con vosotros al cine a condici6n de que/con tal de que/siempre que me acompaneis luego a casa en coche. '# I won't go with you guys to the movies provided that/so long as you take me home afterwards by car. '

The utterances in question become acceptable again if their polarity is changed (see example (4a». Hence, we can say that the complementary distribution of these two subsets (ACCCs and NCCCs) stems from their complementary polarities. Notice that it is not a matter of whether utterances with NCCCs should necessarily be negative (as shown by a simple example such as Manana ire a hacer jogging a no ser que llueva 'Tomorrow I'll go jogging unless it rains'). Rather, the point is that its potential commutation by an ACCC always entails a change in polarity (Manana ire a hacer jogging a condicion de que no llueva 'Tomorrow I'll go jogging provided that it doesn't rain'). As we shall see below, the reason for the strangeness of the (b) sentences is not, strictly speaking, a matter of ungrammaticality, i.e., a problem in the syntax of the constructions. Instead, it probably has more to do with a pragmatic inadequacy between the informational content of the two clauses and the processing instructions conveyed by the connectives in question. Despite these distributional constraints, it is easily shown that si ('if) and its corresponding negative counterpart si no ('if ... not') can appear in the ACCC and NCCC contexts, respectively.

146 Estrella Montolio

(5)

a. Ire con vosotros al cine si me acompafiais luego a casa en coche. 'I'll go with you to the movies if you take me home afterwards by car.' b. No ire con vosotros al cine si no me acompafiais luego a casa en coche. 'I won't go with you to the movies if you won't take me home afterwards by car.'

The distribution of the data leads me to believe that si 'if is the prototypical and superordinate member of the category of conditional structures. The connective si is semantically unmarked as compared to CCCs (as well as other possible conditional constructions), which express a more specific and restrictive type of condition. The import of this observation is that CCCs are less abstract and more specific in meaning than if and its negative counterpart if ... not. That is to say, if and if ... not constitute the superordinate terms of the relationship, in which CCCs are the subordinate terms.

2. The order and the value of p in CCCs Although research on clause order in conditional structures has not specifically examined CCC constructions, the data they provide seem to show that clauses introduced by CCCs, unlike if-clauses, usually occupy the second position in conditional constructions (for English, see Traugott, 1997: 160 and Dancygier, 1998: 173; for French, cfr. Dostie, 1987: 177-178; see Montolio 1991b: 134-135, for Spanish data). Therefore, in a high percentage of the cases, CCCs invert the normal or unmarked order of conditional expressions, which of course tends prototypically to be [if p, q]. CCC-ciause order, then, has to be related to the schema [q ifp].4 Ramsey (1985: 405) observes that while preposed protases tend to have discourse functions, postposed protases "look more like a parenthetical comment for the main clause". The corpus-based studies by Ford and Thompson (Ford and Thompson 1986, Ford 1993, 1997) have corroborated this observation, namely that when the protasis is preposed it either conveys information that is in some sense already known or creates the framework for the following discourse. However,

Complex conditional connectives

147

when !fclauses are postposed, they function more like afterthoughts, evaluations, justifications. Prosodic features also play a role. When the postposed clause is preceded by a pause [q, ifp], on occasion at least, the !fclause seems to convey the sense of less probability. The expectation that the contents in p will not be fulfilled is often taken advantage of in advertising slogans: (6)

a. Si puede mirela. 0 cambie, si puede. (Catchphrase for the Spanish TV series Serie Rosa 'Pink Series') (EI Peri6dico de Catalunya, March 21, 1991). 'If you can, watch it. Or change [the channel], if you can.' b. Resistete, si puedes. (Advertising slogan for the Renault 19) (EI Pais, September 15, 1992). 'Resist [the temptation], if you can.'

In each case the effect of postposition in conjunction with the preceding pause clearly conveys an inferential interpretation that the fulfillment of the modal puede(s) 'can' is improbable. In fact, in the television commercial corresponding to the same Renault advertising campaign as that in (6b) even a priest succumbs to the temptation. The interpretation that 'p will not be easily fulfilled' is used as sort of a challenge, a strategy to incite the consumer's interest. I would like to suggest that the CCC-schema conveys an analogous sense of improbability or exceptionality. The characteristic semantics of postposed protases headed by CCCs represents an a-posteriori specification that rectifies 5 the content of the main clause, by stipulating the conditions under which q will or will not be carried out. In these constructions the hearer is instructed that p entails a certain type of correction of what is affirmed in q. As pointed out by Dancygier (1985, 1998) for unless constructions, we can propose that the relation between q and p in CCC structures is not one of cause (as is prototypically the case of conditionals), but rather one of circumstance. That is, p constitutes a specification of the circumstances under which q can be maintained. This can be seen in the following examples: (7)

a. A mi me parece que las especulaciones son muy poco

pertinentes, a no ser que seamos capaces de imaginarnos a nosotros mismos convertidos en un anciano de setenta y siete afios [ ..}. (El Europeo, October 1988, pp. 59-60).

148 Estrella Montolio

'It seems to me that the speculations are rather inappropriate, unless we are able to imagine ourselves turned into a 77-yearold elderly person [... ]. b. La respuesta a tan dificil pregunta es sf, es posible broncearse, siempre que la exposicion solar se realice de manera inteligente. (Marie-Claire, February, 1993). 'The answer to such a difficult question is yes, it is possible to get a tan, provided that exposure to the sun is done in an intelligent manner.' The specification implied by p is so strong that, orally, it is often introduced by a long pause or, graphically, ~y a comma or even a period. In many cases, CCC-ciauses constitute phonologically and graphically independent utterances. The schema [ q. CCC p] is quite frequent in Spanish. It can be observed in the following example:

(8)

Sf, sf, vamos. Estoy convencido de que en mi vida futura la filosofia sera mi hobby. Ahora, no mi profesion, desde luego. A no ser que cambie mucho, 0 me vea en una situacion... economica muy ahogada y me tenga que dedicar a dar clases. (Esgueva and Cantarero (1981: 24). 'Yes, yes, well.... I'm convinced that in my life to come philosophy will be my hobby. But not my job, that's for sure. Unless my life changes a lot, or I find myself in a very tight financial. .. situation and I have to earn a living teaching. ' As opposed to prototypical If-clauses, which are preposed, the order

[q, CCC p] is not iconic with the sequence of events in the 'real world'.

Instead, the order appears to be iconic with the mental processing of the instructions to be given to the hearer. That is, CCC-ciauses tend not be iconic with event structure, but rather with argumentational patterns, in that their cognitive processing seems to mirror the sequence used in assertion-restriction or the specification thereof

3. The biconditional value of CCCs The hypothesis put forth here is the following: whereas the construction if p designates a very broad and generalized condition, roughly paraphrasable as 'in the event that p,' CCCs express highly specific

Complex conditional connectives

149

conditions. As noted above, CCCs establish a much more precise and restricted relationship between p and q than if does. It is true that, as it has been seen, !fclauses can convey a similar "restricted" relation betwen p and q in certain grammatical circumstances (for instance, in cases in which the subordinate clause is postposed and preceded by a pause). The difference lies in the fact that if only suggests such a strong contingency in the appropiate context, whereas CCCs make the "restricted" relation explicit, which is consistent with the greater transparency of CCCs. The relationship between p and q in a CCC-structure is so strong that, as we shall see, it cannot be canceled. In fact, the exceptionality of the condition expressed by CCCs is such that it cannot be later neutralized without leading to a semantically and pragmatically unacceptable utterance (cf. Dostie 1987: 190). Ducrot (1973: 267) and Fauconnier (1984: 147) have both pointed out that there exist mechanisms that nullify the biconditional reading of !fclauses, which is the normal interpretation in ordinary language (known as conditional perfection). Fauconnier proposes that de toute far;on ('anyway') can be employed to nullify the 'invited' inference of a conditional reading (ifp, q , if not p, not q) (Geis and Zwicky 1971). The neutralizing character of de toute far;on is confirmed by "l'apparition simultanee possible de mais ou cependant" (Fauconnier 1984: 147). (9)

a. Si tu viens,} 'irai, mais si tu ne viens pas,} 'irai de toute fa~on. b. Si til vas, ire, pero si no vas, yo ire de todos modos. 6 c. Ifyou go, I'll go, but if you don't, I'll go anyway.

However, as can be seen in (10), such expressions cannot be used with CCC-ciauses. (10) a. *J'irai a condition que tu viennes, mais si tu ne viennes pas, } 'irai de toute fa~on. b. *Ire a condici6n de que vayas, pero si til no vas, ire de todos modos. c. *1'11 go provided that you go, but if you don't, I'll go anyway. These examples corroborate the hypothesis that the condition established by CCC-ciauses is in no way defeasible. This behavior is similar to that of the connective if and only if. A condicion de que

150 Estrella Montolio

'provided that' and a menos que/a no ser que 'unless' convey unique or exceptional circumstances that will allow (or block) the event expressed in the main clause to occur. This explains why such CCC-constructions cannot be iterated in coordinate structures: (11) a. *,Sa/dre contigo a cenar a condicion de que no hab/emos de trabajo y a condici6n de que me invites. 'I'll go out to dinner with you so long as we don't speak about work and so long as you pick up the tab. ' b. *Me ire de esta empresa a no ser que me asciendan y a no ser que me aumenten e/ sue/do. 'I'll leave this company unless they promote me and unless they give me a raise. ' Notice that both constructions are perfectly viable with si: (12) a. Sa/dre contigo a cenar si no hablamos de trabajo y si me invitas. 'I'll go out to dinner with you if we don't speak about work and if you pick up the tab. ' b. Me ire de esta empresa si no me ascienden y si no me aumentan e/ sue/do. 'I'll leave this company if they don't promote me and if they don't give me a raise. ' The reason for the unacceptability of the iterated CCCs lies in the fact that it is simply not possible to present the only circumstance under which something will or will not happen and then add another exceptional condition. In short, CCCs disallow the presence of other exceptional conditions.

4. ACCCs and NCCCs Both groups (ACCC and NCCC) express exceptional conditions. The meaning of a condici6n de que 'provided that' and other ACCCs in general is restrictive: they establish the only condition under which q can be realized. That is, ACCCs have approximately the same value as [q only ifp]. ACCCs are therefore formulations of a favorable exception for something to happen. NCCCs, on the other hand, indicate the only

Complex conditional connectives

151

situation by which q will not be realized. 7 That is, NCCCs convey the instruction [q only if not pl. Hence, NCCCs convey something to the effect of "in any situation other than p." In other words, NCCCs express the exceptional circumstance by which the proposition expressed by the main clause will not be fulfilled. Both types of CCCs express tff relations. Both types of CCCs present a modal and argumentative value, since they instruct the hearer whether a given circumstance is favorable or disfavorable to the proposition in the main clause. The perspective that the speaker adopts in the face of the possibility of q 's realization is therefore quite different: while ACCCs signal that q can only be realized if p, NCCCs express that q will not be realized only in the case of p. This is what determines the affirmative character of the first group and the negative character of the second. ACCCs convey the necessary basis for something to occur; NCCCs express the circumstance under which something will not be realized. Dancygier (1985: 68) proposes quite insightfully that what unless really denies is q, not p. Her proposal seems to be clearer than the one advanced by Geis (1973): except if p, q. To observe a case in which the main clause displays negative polarity (for instance, example (4)), we can extrapolate from Dancygier's observation as follows in (13b): (13) a. q (b. not q

unlessp unlessp

q; notq;

[(not q) ifp] [ (q) ifp])

(1985: 65)

The negative character of NCCCs is the reason why a no ser que 'unless' and like connectives appear to be generally the equivalent of si no 'if ... not' -clauses. In fact, however, there are quite a few differences between the· two types of constructions (cf Geis 1973; Dancygier 1985; von Fintel 1991; Dancygier 1998: 169-171). Consider, for instance, the odd combination of NCCCs such as unless with negative-polarity clauses which are perfectly acceptable with if (Cf *Unless .John cares a bit for Mary, he shouldn't marry her vs. If John doesn't care a bit for Mary, he shouldn't marry her (Geis 1973:283)). In my opinion, regardless of whether they are equivalent to if or if ... not clauses, the fundamental value of NCCCs is to specify in p the circumstances under which q will not be fulfilled. The point is that the circumstances present in p are construed as improbable or contrary to what one might expect. The contrast between the two following examples adds credibility to the hypothesis advanced here (cf Smith 1983: 21 for similar examples):

152 Estrella Montolio

(14) a. Los cocodrilos no podrian sobrevivir si no tuvieran dientes afiladisimos. 'Crocodiles couldn't survive if they didn't have extremely sharp teeth. ' b. # Los cocodrilos no podrian sobrevivir a menos que tuvieran dientes afiladisimos. 'Crocodiles couldn't survive unless they had extremely sharp teeth. ' (14a) sounds perfectly normal but (14b) is surprising from a pragmatic point of view. In (14a), the negation of the counterfactual structure presupposes that crocodiles actually do have extremely sharp teeth. On the other hand, (14b) is decidedly odd because the clause introduced by the NCCC presupposes that the condition in p is exceptional. That is, the condition is contrary to what would normally be expected. Since NCCCs convey an exceptional obstacle for q, in an eternal-truth context such as (14b) they are obviously inappropriate: the clause suggests that crocodiles do not have extremely sharp teeth, and this conflicts whith our encyclopedic knowledge. This nuance of less expectation conveyed by unless-constructions explains why they are often inappropriate for counterfactuals. As is well known, counterfactual clauses are presupposed to be false: (15) Yo Ie habria dado ese medicamento a la nina, si no 10 hubiera desaconsejado el medico explicitamente. 'I would have given the medicine to the girl, if the doctor hadn't advised against it explicitly. ' The utterance in (15) conveys a factual interpretation in which the content of p is negated. The doctor is understood as indeed having advised against giving the girl the medicine. Clauses headed by a menos que 'unless' and a no ser que 'unless' convey information which is necessarily understood as hypothetical, precisely because they carry a sense of high improbability-and improbability is necessarily hypothetical. Thus, a version of (15) with a menos que or a no ser que indicates that the hypotheticality (i.e., nonfactuality) of p is real, since in reality we do not know if the doctor did or did not authorize the medicine:

Complex conditional connectives

153

(16) Yo Ie habria dado ese medicamento a la nina, a menos que/a no ser que 10 hubiera desaconsejado el medico explicitamente. 'I would have given the medicine to the girl, unless the doctor had advised against it explicitly. '

Precisely because they can only receive a hypothetical interpretation and are primarily future-oriented (see Traugott 1997: 158 for the case of unless), a menos que/a no ser que-clauses are incompatible with counterfactuals that presuppose the creation of an 'impossible world.' This is also shown by the behavior of the following examples. (17) a. Si Colon no hubiera descubierto America, los indios habrian sido mas felices. 'If Columbus hadn't discovered America, the Indians would have been happier. ' b. *A no ser que Colon hubiera descubierto America, los indios habrian sido mas felices. *' Unless Columbus had discovered America, the Indians would have been happier. ' As we have seen, a no ser que presents q as well as the possibility of not q. Nevertheless, counterfactuals only allow for one interpretation: q, if the clause is in negative form, or not q, if there are no overt markers of negation. Hence, (17b) is odd because we know that Columbus did actually discover America.

5. CCCs and contextually given conditions As we have just seen, the greater specificity of CCCs-in contrast to the neutral, unmarked character of si 'if-makes them inappropriate with counterfactual structures. This same reason, I contend, stands behind their inability to appear in constructions in which the content of p is contextually asserted. Take for example sentence (18): (18)

A: Ya he acabado el trabajo. 'I've just finished the job.'

154 Estrella Montolio

a. B: Pues si ya has aeabado e/ trabajo, sa/gamos a eenar por ahi. 'Well, if you've just finished the job, let's go out for dinner. ' b. 'B: *Pues a condici6n de que ya hayas aeabado e/ trabajo, sa/gamos a eenar por ahi. *'Provided that/so long as you've finished the job, let's go out to dinner. '

In other cases, the assertion hidden in the protasis is less explicit, but it is information that is present in the context of the speech event (the term context should be understood in the broad sense of Sperber and Wilson 19962) . The thematic or given nature of such rrotases is demonstrated by the fact that the protasis cannot be postposed: (19) a. Si hay una pasion franeesa que no //eva camino de desapareeer, esa es /a pasion por /a po/itiea. [E/ Pais, June 21, 1992] 'If there is one French passion that is not on its way to disappearing, it is their passion for politics. ' b. *Es /a pasion por /a po/itiea, si hay una pasion franeesa que no //eva camino de desapareeer. 'It's is a passion for politics, if there is one French passion that is not on its way to disappearing. ' c. *A condici6n de que haya una pasion franeesa que no //eva camino de desapareeer, esa es /a pasion por /a po/itiea. 'Provided that there is one French passion that is not on its way to disappearing, it is their passion for politics. ' (20) a. Si no te /0 ha dicho es porque te quiere. 'Ifhe hasn't told you it's because he loves you.' b. *Es porque te quiere, si no te /0 ha dicho. 'It's because he loves you, ifhe hasn't told you.' c. *A no ser que te /0 haya dicho es porque te quiere. *' Unless he's told you, it's because he loves you.' Elsewhere (Montolio 1999) I have called such conditional structures identificatives (19) and exp/ieatives (20), respectively. Unlike prototypical (i.e., hypothetical) conditionals, the If-clauses here are far from left-open (in the sense of Lehman 1974: 234; that is, the

Complex conditional connectives

155

proposition in the antecedent of a conditional sentence is not asserted). On the contrary, these uses of si-clauses seem to be what I propose to call left-given. This is shown by the fact that these clauses cannot be postposed, as is demonstrated by the (b) examples, nor can they be used with CCC-structures, as in the (c) examples. In Spanish, as well as in French and English, the apodoses of ifconstructions can be-indeed frequently are-yes/no-questions «21a)) as well as wh-questions «22a) and (23a)). (21) a. Sifuese obispo, ~condenaria el aborto? 'If you were a bishop, would you condemn abortion?' b. *A condici6n de quefuese obispo, ~condenaria el aborto? *'Provided that you were a bishop, would you condemn abortion?' (22) a. Si a Gonzalez Ie gusta tanto participar en la campana de un referendum, ,",por que no organiza uno en su pais? dice Philippe Seguin. [El Pais, June 21, 1992] 'If Gonzalez likes participating in the campaign for a referendum so much, why doesn't he organize one in his country? says Philippe Seguin. ' b. *A condici6n de que a Gonzalez Ie guste tanto participar en la campana de un referendum, ,",por que no organiza uno en su pais? dice Philippe Seguin. *' Provided that Gonzalez likes participating in the campaign for a referendum so much, why doesn't he organize one in his country? says Philippe Seguin. ' (23) a. Si no estas preparada para tener un hijo, ,",por que demonios has querido quedarte embarazada. 'If you're not ready to have a child, why the hell did you want to get pregnant?' b. *A menos que estes preparada para tener un hijo, ,",por que demonios has querido quedarte embarazada. *' Unless you're ready to have a child, why the hell did you want to get pregnant?' In view of examples such as these, it seems feasible to state that ACCCs are difficult to combine with yes/no-questions, as is the case with example (21b). Such a combination gives rise to utterances that are

156 Estrella Montolio

clearly very strange. Furthermore, CCCs are, in general, difficult to interpret with wh-interrogatives (e.g., (22b) and (23b». Actually, this difference in behavior with regard to the two different basic types of interrogatives does not come as much of a surprise. This is so because in wh-interrogatives the tt:protasis functions as a framework for a latent assertion (e.g.,(22a) and (23», whose superficial hypothetical form lays the groundwork for and justifies the question that follows it. That is, the use of a hypothetical form for contextually asserted information constitutes a mechanism that attempts to safeguard the relevance of the subsequent question (cf Haegeman 1984). CCCs are also incompatible with so-called indicative counterfactuals, since they also presuppose some kind of previous assertion: (24) a. Mira, si esa chica es guapa, yo soy Claudia Schiffer. 'Look, if that girl is pretty, I'm Claudia Schiffer.' b. *A condici6n de que esa chica sea guapa, yo soy Claudia Schiffer. *'Provided that that girl is pretty, I'm Claudia Schiffer.' (25) a. Si tit no eres rico, yo entonces soy un mendigo. 'Ifyou're not rich, then I'm a beggar. ' b. *A no ser que tit seas rico, yo soy un mendigo. *' Unless you're rich, I'm a beggar.' All of these kinds of conditionals fall under the heading of nonpredictive conditionals (according to Sweetser 1990; Dancygier 1993, 1999). Note that, to a greater or lesser extent, they all constitute "echoes" from some previous discourse. Furthermore, they can be characterized by their lack of strict iconicity. The sequence of clauses is not iconic with the sequence of events, though in fact there does seem to be an iconic correspondence between the linear order of the clauses and the sequence of steps taken in argumentation: Given X allows me to assert ... (cft. Sweetser 1990: 128). That is, speakers use p to show the hearer how they arrive at the assumptions in q. As we can see, CCCstructures cannot appear in any context of epistemic conditionals, where the tt:clause introduces some type of assumption present in the context of the speech situation (Sweetser 1990: 128). That is, CCC-ciauses cannot appear in any context whose content is given, assumed, or clearly presupposed. The information that CCC-ciauses convey is

x:

Complex conditional connectives

157

therefore necessarily rhematic. This is consistent with post-main-clause order, because focus tends cross-linguistically to occur in sentence-final position (Comrie 1986: 86). The insertion of a CCC, especially NCCC, seems slightly more acceptable--especially when they appear preposed-with speech-act and metalinguistic conditionals than with any of the former types. Nevertheless, such uses do not, by any means, constitute prototypical contexts for CCCs. Let us consider Austin's celebrated example: (26) a. There are biscuits on the sideboard ifyou want them. b. #There are biscuits on the sideboard provided that/so long as you want them. Clearly there is something peculiar about (26b). Consider the examples in (27): (27) a. En espanol americano a eso Ie llaman 'durazno', SI no confundo la palabra. 'In Latin-American Spanish that's what they call a 'durazno' (peach), if I'm not confusing the word.' b. ?En espanol americana a eso Ie Ilaman 'durazno', a no ser que confunda la palabra. 9 'In Latin-American Spanish that's what they call a 'durazno' (peach), unless I'm confusing the word.' The ability of NNNC clauses to be inserted in these nonpredictive, nonprototypical schemas can be explained by the parenthetical nature that both !fclauses and CCC-ciauses acquire in these types of conditionals: the independence of the propositional content of CCCclauses with regard to the main clause also corresponds to ·the nonembedded character of the subordinate clauses in speech-act and metalinguistic conditionals.

6. CCCs and the discourse deictic then Another feature that characterizes CCC-constructions as noncentral members of the family of conditional structures is their inability to combine with the phoric then: 10

158 Estrella Montolio

(28) a. Si hace sol, entonces iremos a la playa. '/fit's sunny, then we'll go to the beach.' b. *A condici6n de que/con tal de que/ siempre que haga sol, entonces iremos a la playa. 'Provided that/so long as it's sunny, then we'll go to the beach.' (29) a. Si no llueve, entonces iremos a la playa. '/fit doesn't rain, then we'll go to the beach.' *A no ser que/ a menos que/ salvo que/ excepto que llueva, entonces iremos a la playa. *' Unless/except ifit rains, then we'll go to the beach.' The presence of entonces ('then') in conditional schemas is a matter that remains to be thoroughly studied in Spanish. Nonetheless, as pointed out by Haiman (1978: 576), in contradistinction to the traditionally studied schema if p, then q, the phoric then does not constitute by any means an obligatory element in ordinary-language conditionals. In fact, certain types, such as speech-act conditionals or metalinguistic conditionals, reject it. (30) a. ?? Si no me equivoco, entonces ese es tu primo. ??'/fI' m not mistaken, then that guy is your cousin. ' b. ?? Si se Ie puede llamar asi, entonces es una estructura condicional. ??' /f you can call it that, then it's a conditional structure. ' Of course, speech-act and metalinguistic conditionals, by their very nature, do not express a grammatical relationship where p entails q. By contrast, entonces 'then' explicitly expresses that q follows from p. Therefore, the semantics of these constructions clashes with that of then and thus produces an unacceptable utterance. In those constructions that do accept its presence, the insertion of then is neither pragmatically, nor semantically identical to its omission: it is one thing to imply that q follows from p, as in prototypical conditionals using the schema [if p, q], and it is another to mark this explicitly, as in [ifp, then q]. In any event, it should be noted that those if-constructions that do allow the presence of then do not allow it when the normal order of the clauses is altered, which happens to be the schema normally used by CCCs:

Complex conditional connectives

159

(31) *Entonces ire a la playa, si llueve. 'Then I'll go to the beach, ifit rains.' Nevertheless, the examples in (28) and (29) show that CCCs are incompatible with entonces 'then' even when they appear in a preposed position. Such a fact invites the question of just what the function of then in a conditional is. There seems to be agreement that one of its main functions is to act as an anaphoric emphatic correlate to the protasis, i.e., something like a resumptive pronoun (Dancygier 1998; Dancygier and Sweetser 1997; latridou 1992; Montolio 1991b, 1999). One might say that the use of then constitutes an attempt on the part of the speaker to insure that the interlocutor retains the hypothetical framework on which the following clause or utterance is based. This is clearly the case of conditionals in which various protases are linked to one apodosis: (32) Todos creemos tener el mismo derecho: si uno se puede hacer millonario en dos dias, si no es el resultado de una vida de esfuerzo, si se consigue con solo jugar a la Bolsa, con disponer de informacion privilegiada, entonces todo el mundo cree que tiene derecho a ganar mas de 10 que gana. [EI Pais, May 17, 1998] 'We all believe that we have the same right: if one can become a millionaire in two days, if it's not the result of a life of hard work, if it can be obtained by playing the stock market, by having privileged information, then everybody believes that they have the right to earn more than they do.'

As opposed to prototypical ~clauses, CCC-ciauses do not function as a hypothetical framework for the assertion of q. Rather, as we have seen, they restrict or specify what is expressed in q. They are not, however, the basis for asserting the content of the main clause. These facts account for why CCCs cannot collocate with then, since the cohesive function of this element is to serve as an elliptical framework for the apodosis. However, when the CCC-structure is included in a larger framework, entonces 'then' is indeed acceptable. (33) A: Papa, iftle Ilevaras al cine? 'Dad, will you take me to the movies?'

160 Estrella Montolio

B:

J.~i,

a condici6n de que acabes los deberes. Entonces te llevare al cine. 'Provided that/so long as you finish your homework. Then I'll take you to the movies. '

As can be seen, in (33) the insertion of entonces 'then' sounds perfectly normal because its presence is not linked solely to the information in the CCC-ciause. If this were the case, it would have given rise to an unacceptable utterance: (34)

*A condici6n de que acabes los deberes, entonces te llevare al cine. *'Provided that/so long as you fmish your homework, then I'll take you to the movies. '

Instead, in (33) B's answer, entonces ('then') is linked to a larger discourse context, to the whole previous sentence ("yes, I will take you to the movies, provided that you fmish your homework. Then --'if this is the case' --, I'll take you").

7. The pragmatics of CCCs In various papers, Fillenbaum has analyzed the relationship that exists between a series of conditional constructions and certain speech acts (see, specially, Fillenbaum 1986). He points out that the semantic content of unless makes it appropriate to appear in deterrents, while at the same time making it unnatural-sounding in inducements. This semantico-pragmatic relationship between unless-constructions and the illocutionary value of threats!! and prohibitions has been documented diachronically for English (Traugott 1997) as well as for Spanish. Bartol Hernandez (1992: 89) briefly comments that, historically, in the majority of the cases, they are prohibitions (or negative orders, expressed with can, should, or the subjunctive) whose compliance carries an exception, a restriction, presented as a necessary condition. (35) is an example given by this author:

(35) E si por aventura el Rey fuere de tan gran piedat que 10 quiere dexar vivir, non 10 pueda facer a menos que no Ie saque los ojos. (Fuero Real, 350)

Complex conditional connectives

161

'And if by chance the King were so kindhearted as to want to let him live, he cannot do so unless his eyes are tom out. (Regional Code, 350)' Fillenbaum has shown that in a high percentage of cases provided that appears in promises or inducements (e.g., Fillenbaum 1976: 148). The fact is that it is very difficult to invert these roles; that is to say, the formulation of a promise with unless or that of a threat with provided that is highly infrequent (being generally inadequate). This is true even if we change the polarity of the clauses, in accordance with the complementary distribution of the CCCs that we have analyzed. As can be seen, (36b) is no longer a deterrent (as (36a» but a promise not to do something undesirable. In contrast, (36d) seems a threat more than an inducement (in contradistinction to (36c»: (36) a. Te dejare todo el verano sin vacaciones, a no ser que mejores estas notas. 'I'll take away your vacation, unless you improve these grades. ' b. # No te dejare todo el verano sin vacaciones, a condici6n de que mejores estas notas. #'1 won't take away your vacation, provided that you improve these grades. ' c. Iremos de viaje a Disneylandia, a condici6n de que mejores estas notas. 'We'll go to Disneyland, provided that you improve these grades.' d. # No iremos de viaje a Disneylandia, a no ser que mejores estas notas. #'We won't go to Disneyland, unless you improve these grades.' The favoring character of positive stimuli presented by provided that is corroborated by the fact that it can paraphrase-with the right modifications in the clauses-any paratactic conditional construction with the conjunction and, but only in the event that it expresses an inducement stimulus, such as: (37) a. Dame un besito y te dare un caramelo. 'Give me a little kiss and I'll give you a piece of candy.'

162 Estrella Montolio

b. Te dare un earamelo, a condici6n de que me des un besito. 'I'll give you a piece of candy, provided that you give me a kiss. ' However, when these coordinate structures are the equivalent of a threat the paraphrase is impossible. Compare the following sentences: (38) a. Toea a mi hermano y te parto la eara. 'Touch my brother and I'll bash your face in. ' b. #Te parto la eara, a condici6n de que toques a mi hermano. #'1'11 bash in your face, provided that you touch my brother. ' The illocutionary specialization of these connectives is derived from their semantics. The lessening of expectation or high improbability conveyed by a menos que ('unless') conveniently explains its inadequacy as a stimulus: it does not seem pragmatically profitable to try to act upon one's interlocutor under the pretext that something probably WILL NOT happen. This same reason explains the adequacy of unless for prohibitions and threats. The coercive power of a threat lies in the fact that the negative event for the hearer is presented as a general assertion, which only has a distant possibility of being revised: X, unless Y. The hearer must therefore either comply with the proposition in the unless-clause or suffer the consequences. 12 To exemplify this, we can recreate a cliche-ridden dialogue from the film-noir genre between a gangster type and "the other guy's girl." In the improbable event that the gangster were to use a CCC-construction instead of the more predictable paratactic structure, he could choose between formulating his threat with provided that or unless. Nevertheless, one of them would seem much more persuasive than the other: (39) a. No matare a tu novio, a condici6n de que te fugues eonmigo. 'I won't kill your boyfriend, provided that you run away with me.' b. Matare a tu novio, a no ser que te fugues eonmigo. 'I'll kill your boyfriend, unless you run away with me.' The greater force of (39b) lies precisely in that the threatening action (i.e., kill the boyfriend) is expressed in an affirmative factual manner.

Complex conditional connectives

163

This seems to be more effective than posing the matter in the converse terms, as in (39a), where the threatening action is negated and the restriction on this negated threat appears later. The same pragmatic explanation can be applied to resolve the greater or lesser inadequacy of utterances (36b) and (36d). As can be seen, exchanging connectives causes the speech acts to be exchanged too. Thus, (3 6b) seems to be a much weaker version of a threat than that formulated in (36a), though it also could be interpreted as an unconvincing stimulus. Insofar as the clear stimulating promise in (36c) is concerned, it is converted directly into a threat as soon as its connective is changed (36d). Besides the illocutionary value generally equated with CCCstructures, we can also pose the question of what their pragmatic behavior is as compared to that of if-clauses. Also worth considering is the question of whether there exist differences between the two subsets of CCCs from such a point of view. An if-clause can appear in any context that a CCC can, and in fact, speakers prefer to use the unmarked connector if even when the insertion of a CCC is possible. The very subtle differences that these constructions convey with regard to if (and if ... not) are so faint that most speakers probably do not consider them to be very operative. Furthermore, in the Romance languages, CCCclauses trigger the subjunctive mood, which according to the literature on the matter is in the decline. This seems to be why CCC-ciauses are more characteristic of formal, written registers than of spontaneous, oral ones. CCC-structures are, in effect, noticeably less frequent than if-clauses in oral and spontaneous registers. In this sense, we can see the correspondence between ontogenesis and linguistic universals, because such constructions are acquired much later than if-structures. Although most authorities on the matter of the acquisition of conditionals only deal with if-clauses, Wing and Kofsky (1981) point out that just as common sense would lead us to believe, CCC-structures are among the latest to be acquired. Given their formal complexity and the specificity of the condition they introduce, CCC-structures surface in child language at a later stage than if-clauses and other types of synthetic conditional structures, such as paratactic structures (McCabe et al. 1983, Wing and Kofsky 1981). In the case of NCCCs, this late appearance seems to be even more justified. NCCCs imply a greater cognitiveprocessing complexity in that to the uncertainty of the truth value they add a negative relationship between the two clauses.

164 Estrella Montolio

My claim is that no matter what the speech act may be, ACCCs and NCCCs are pragmatically very different from one another. When speakers choose to use an unless-clause, they are often trying to avoid committing themselves completely: q is said first, and afterwards, as a kind of afterthought, the only situation in which their utterance will not be true is signalled (P). This claim is consistent with the historical corpus-based analysis of Traugott (1997), who concludes that in general unless-clauses express 'possible events that will block promises, threats, expectations, and even probable recurrence. '

(40) Creo que [el condicionamiento social y economico] impiden que en la practica se pueda decir: yo me caso cuando quiera. A no ser que la ... no se, tambien la situacion economica familiar del novio y de la novia, su plan, claro (Esgueva and Cantarero 1981: 34). 'I think that [social and economic conditioning factors] keep one from saying in practice: I'm going to get married whenever I want. Unless the ... I don't know, also the boy's and the girl's family financial situation, their plans, of course. ' Thus, in most cases, the use of unless can be interpreted as aface-saving repair (cf Goffman 1967). In such cases the speaker tries to take precautions against an interpretation of his utterance in terms of overall truth, which could jeopardize the speakers "face" in the future. In this sense, the use of unless is much like metadiscursive if=-clauses of the type if I'm not mistaken, if that's the right word, if I remember correctly, etc., and has to do with the strategy of guarding oneself from possible criticisms from one's interlocutor. On the other hand, ACCCs are more problematic from an interactive perspective: once established, provided that cannot be mitigated. Such a specific meaning explains-at least in part-its low frequency: speakers prefer to propose a condition with an if=-clause, rather than impose a condition with provided that. In fact, according to the data I have worked with, speakers often use an ACCC-ciause when they are talking about a third party. In effect, ACCCs can be too imposing, and therefore are used typically when there exists a certain asymmetry ("power") in the social status of the interlocutors. Thus, a tenured professor can easily indicate to his graduate student:

Complex conditional connectives

165

(41) De acuerdo, Ie dirigire la tesis, a condici6n de que trabaje sobre el subjuntivo. 'Agreed, I'll supervise your thesis, provided that you work on the subjunctive. ' On the other hand, this utterance would not be very likely the other way around; that is, it would be precarious if the graduate student declared something like the following to his superior: (42) Muy bien, dare sus clases a condici6n de que usted me ayude a preparar la bibliografia. 'Okay, I'll teach your classes provided that you help me with the bibliography. ' Such an utterance, without a doubt, would seem less risky formulated as an ~clause: (43) Muy bien, dare sus clases si usted me ayuda a preparar la bibliografia. 'Okay, I'll teach your classes if you help me with the bibliography. ' The rule of linguistic politeness which indicates that indirectness is more courteous than direct expressions also explains why the ACCCstructures can seem not very polite: with these structures, there is no interpretation other than the imposition of a necessary condition. Such constructions are learned; furthermore, they can also seem impolite if they are used in direct address.

10.

Summary

To summarize, I shall try to list the principal traits of CCC-structures and place them in relation to other structures belonging to the family of conditional structures: (i)

Non-predictive conditionals are non-prototypical. a. CCCs cannot appear in non-predictive contexts (i.e., where, in most cases, [ifp] equals [givenp]); b. CCCs are related to predictive and future-oriented contexts.

166 Estrella Montolio

(ii)

Hypotheticals and counterfactuals are non-prototypical (Dancygier 1993: 428). a. CCCs rarely appear in counterfactual contexts; b. CCCs are basically related to central, predictive conditionals (i. e., prototypical conditionals).

(iii)

In prototypical, predictive conditionals, CCCs are non-central because: a. whereas, in prototypical If-conditionals, clauses are related through causal relationships, in CCC structures, clauses are related through circumstantial relationships; b. whereas the prototypical order of central If-conditionals if p, q makes them iconic with the sequence of events, the prototypical order of CCC-conditionals q, CCC P makes them iconic with the order involved in processing assertionrestrictions; c. whereas the function of If-clauses is, prototypically, to establish a state of affairs to be taken as the basis for considering the consequent, the function of CCC-ciauses is to restrict or specify the contents of the apodosis (hence, their incompatibility with then).

Notes *

1.

2.

I would like to thank Maria Rosa Vila and Joseph Hilferty for the many hours that they spent with me discussing many of the points expanded upon here, as well as the editors of this volume, Bernd Kortmann and Elizabeth CouperKuhlen, for their exhaustive observations on this chapter. I would also like to thank lbon Sarasola for providing me with the Basque example in endnote 4. I would further like to mention the help received from Joseph Hilferty in making an intelligible English version. It goes without saying, however, that all errors are of my own responsibility. The lack of quality grammatical descriptions regarding CCCs is reflected in the fact that most Spanish textbooks and reference grammars for foreign learners give a rather jumbled treatment of these constructions. Furthermore, they tend to present them as if they were mutually commutable with one another as well as with ij=-clauses in general. This has the unfortunate side effect, as I have seen in my own classes, of prompting foreign students to produce unacceptable utterances. It may be possible that the conclusions reached here are also largely applicable to languages such as Italian and German. Given my unfamiliarity with these languages, however, I leave it to others to confirm or disprove this

Complex conditional connectives

3.

4.

5.

6.

possibility. It should also be noted, however, that in Spanish--as in other Romance languages--clauses containing CCCs are generally constructed with the subjunctive mood. The inflectional contrast between the present and the imperfect subjunctives allows speakers of the Spanish language to differentiate the values that conditional constructions with if express through the indicative/subjunctive opposition. The Spanish ACCC siempre que (literally: 'always that') maintains a temporal meaning in many of the utterances in which it appears. An interesting question is which formal elements must be involved so as to obtain a conditional reading; in this sense, it seems that the presence or absence of a pause between the relevant clauses is decisive. Conditional readings tend to be marked by a pause, whereas a lack of a pause seems to signal a temporal interpretation. Compare the following examples: (i) a. Nos traera un regalo siempre que vue/va de viaje. (temporal) 'She'll bring us a present every time she returns from a trip.' b. Nos traera un regalo, siempre que vuelva de viaje. (conditional). 'She'll bring us a present, provided that she returns from a trip.' The relation between CCC-ciauses and postposed ~clauses appears to be corroborated by the fact that languages such as Basque, which do not possess specific unitary connectives to convey such conditional nuances, often express them by postposing the ~clause (as in the Basque ba): (i) Hondartzarajoango ara eguzkia egiten Beach ABL go FUT AUX(lPL/PRES/ITR) sun-ART badu bederen make-IMP COND-AUX(CONJ/TR/3SG) at least 'We'll go to the beach if it's at least sunny out.' Something similar is suggested by Leard (1987: 164) upon isolating a particular group of conditional expressions which he calls restrictives. Dostie (1987: 191-192) also associates this rectifying mechanism with CCCs, though she only considers those structures that I have included in the ACCC group. Of course, such utterances would sound more natural if an optative concessive conditional were used: (i) Vayas 0 no vayas, go-PRES-SUBJUNCT-you or NEG go-PRES-SUBJUNCT-you yo ire. I

7.

8.

167

gO-FUT

'I'm going whether you go or not.' NCCCs are, in a manner of speaking, related to concessive-conditional connectives. This applies, specifically, to incluso si ('even if), since concessive-conditionals express the idea of 'contrary to expectation'. On the other hand, a no ser que ('unless') and incluso si ('even if) have, in a certain sense, complementary meanings: a no ser que conveys the only circumstance that will be an impediment for the fulfillment of q; in contrast, incluso si entails that p will never be an obstacle for q to be realized. In Spanish there is another type of conditional structure labeled a contrastive conditional, which is characterized by the thematic nature of the protasis. It does not appear to have a clear counterpart in English:

168 Estrella Montolio

(i)

Si Chatwin eligi6 vlaJar caminando, Paul Theroux redescubri6 las glorias y zozobras del ferrocarril (El Europeo, May 1993). *lfChatwin chose to travel by foot, Paul Theroux rediscovered the glory and angst of the train. However, just as in the case of other conditionals whose protasis expresses a contextually given condition, this conditional cannot appear with a CCC, as is demonstrated by the grammatical behavior of the sentences in (ii). (ii) *A condici6n de que Chatwin eligiera viajar caminando, Paul Theroux redescubrio las glorias y zozobras del ferro carri I. *Provided that Chatwin chose to travel by foot, Paul Theroux rediscovered the glory and angst of the train. For more information on Spanish contrastive conditionals, see Montolio (1999). 9. In fact, speech-act conditionals are a habitual context for a natural unlessclause anteposition.Thus, for instance, the example at hand would seem more plausible with a preposed CCC: A no ser que confunda la palabra, en espanol americana a eso Ie llaman udurazno" 'Unless I'm confusing the word, LatinAmerican Spanish they call this [i.e., a peach] "durazno'" 10. In Spanish there exist two particles that correspond to this value of then: entonces (which, like then, coincidentally also has a temporal meaning) and pues (which historically meant 'after'). In fact, the connective pues 'thus'seems actually to be much more frequent than entonces 'then' in spoken Spanish conditionals and seems to correspond to a marker of sequentiality (Dancygier, 1998: 178 et passim), that is, a particle that introduces conclusions. This matter still needs to be examined further, but a first approximation can be found in Montolio (1991a, 1999). 11. It is worthwhile pointing out that in a conversational context, especially where direct reference is being made to the interlocutor, conditional threats in Spanish frequently take the form of a special construction: como + VPRESSUBJUNCT. This construction seems to be quite similar to the Japanese -tewa (Akatsuka, 1997) and the disjunctive paratactic conditional, as in: Cena 0 no veras la television ('Eat your dinner or you can't watch TV'). For more on such threatening conditionals in Spanish, see Montolio (1999). 12. To this effect, Brown and Levinson distinguish concrete types of acts that threaten negative face and that are formulated in the following way: Threats, warnings, dares (S[peaker] indicates that he (or someone or something) will instigate sanctions against H[earer] unless he does A) (Brown and Levinson 1987: 65-66; the italics are mine).

References Akatsuka, Noriko 1997 Negative conditionality, subjectification, and conditional reasoning. In: Angeliki Athanasiadou and Rene Dirven (eds.), On Conditionals Again, 323-354. Amsterdam: Benjamins.

Complex conditional connectives

169

Austin, John L. 1956 1ft and cans. Proceedings ofthe British Academy: 109-132. Bartol Hernandez, Jose Antonio 1992 Algunas conjunciones y locuciones condicionales en la Edad Media. In: Jose Antonio Bartol Hernandez, Juan Felipe Garcia and Javier de Santiago (eds.), Estudios jilol6gicos en homenaje a Bustos Tovar, vol. 1,85-110. Salamanca: Ediciones Universidad de Salamanca. Brown, Penelope and Stephen C. Levinson 1987 Politeness. Cambridge: Cambridge University Press. Comrie, Bernard 1986 Conditionals: A typology. In: Elizabeth Traugott, Alice Ter Meulen, Judy Snitzer and Charles Ferguson (eds.), On conditionals, 77-99. Cambridge: Cambridge University Press. Contreras, Lidia 1963 Las oraciones condicionales. Boletin de Filologia de la Universidad de Chile XII: 273-290. Dancygier, Barbara 1985 If, unless, and their Polish equivalents. Papers in Contrastive Linguistics XX: 65-72. 1990 Conditionals: Sequence of events and sequence of clauses. In: Jaeck Fisiak (ed.), Further insights into contrastive analysis, 357-373. Amsterdam: Benjamins. 1993 Interpreting conditionals: Time, knowledge, and causation. Journal ofPragmatics 19: 403-434. 1998 Conditionals and prediction. Time, knowledge and causation in conditionals structures. Cambridge: Cambridge University Press. Dancygier, Barbara and Eve Sweetser 1997 Then in conditional constructions. Cognitive Linguistics 8(2): 1-28. Dostie, Gaetane Etude semantique de quatre connecteurs conditionnels: a condition 1987 que, pourvu que, en autant que et d'abord que. Le jran9ais moderne LV/3-4: 174-203. Ducrot, Oswald 1973 La preuve et Ie dire. Paris: Marne. Esgueva, Manuel and Margarita Cantarero (eds.) 1981 El habla culta de la ciudad de Madrid. Madrid: Consejo Superior de Investigaciones Cientificas. Fauconnier, Gilles 1984 Espaces mentaux. Paris: Minuit. Fillenbaum, Samuel 1986 The use of conditionals in inducements and deterrents. In: Elizabeth Traugott, Alice Ter Meulen, Judy Snitzer and Charles Ferguson (eds.), On conditionals, 179-195. Cambridge: Cambridge University Press. von Fintel, Kai 1991 Exceptive conditionals: the meaning of unless. North Eastern Linguistics Society 22: 135-148.

170 Estrella Montolio

Ford, Cecilia E. and Sandra A. Thompson 1986 Conditionals in discourse: A text-based study from English. In: Elizabeth Traugott, Alice Ter Meulen, Judy Snitzer and Charles Ferguson (eds.), On conditionals, 353- 372. Cambridge: Cambridge University Press. Ford, Cecilia E 1993 Grammar in interaction. Adverbial clauses in American English conversations. Cambridge: Cambridge University Press. Geis, Michael and Arnold M. Zwicky 1971 On invited inferences. Linguistic Inquiry 2: 61-66. Geis, Michael 1973 If and unless. In: Braj B. Kachru, Robert B. Lees, Yakov Malkiel, Angelina Pientrangeli and Sol Saporta (eds.), Issues in Linguistics. Papers in honor of Henry and Renee Kahane, 231-253. Urbana: University of Illinois Press. Goffman, Erving 1967 Interaction ritual: Essays on face to face behavior. New York: Garden City. Haegeman, Liliane 1984 Pragmatic conditionals in English. Folia Linguistica XVIII/3-4: 485502. Iatridou, Sabine 1991 If Then, then what? North Eastern Linguistics Society 22: 211-225. Leard, Jean-Marcel 1987 La syntaxe et la classification des conditionnelles et des concessives. Le Fran~ais Moderne LV (3/4): 158-173. Lehmann, Christian 1974 Prinzipien flir "Universal 14". In: Hansjakob Seiler (ed.), Linguistic Workshop 11, 69-97. Munich: Wilhelm Fink Verlag. McCabe, Anne E., Susan Evely, Rona Abmmovitch, Carl M. Corter and Debra 1. Pepier 1983 Conditional statements in young children's spontaneous speech. Joumal ofChild Language 10: 253- 258. Montolio, Estrella 1990 Expresi6n de la condicionalidad en espano!. Ph.D. dissertation. Barcelona: Universitat de Barcelona. 1991a Asi, pues entonces, 10 mejor sera que pienses bien 10 de casarte. Acerca de los procondicionantes en espana!. Foro Hispanico 2: 4353. 1991b On conditional structures with complex connectors. In: Johan van der Auwera (ed.), Eurotyp working papers: Adverbialia, volume 1: 33-52. 1999 Capitulo 57: Las construcciones condicionales. In: Ignacio Bosque and Violeta Demonte (eds.), Gramatica descriptiva de la lengua espanola, Volume III: 3643-3737. Madrid: Espasa Calpe.

Complex conditional connectives

171

Ramsey, Violeta 1987 The functional distribution of preposed and postposed if- and whenclauses in written discourse. In: Russell S. Tomlin (ed.), Coherence and Grounding in Discourse, 383-408. Amsterdam: Benjamins. Smith, Neil V. 1983 On interpreting conditionals. Australian Journal ofLinguistics 3: 124. Sperber, Dan and Deidre Wilson 19962 Relevance. Communication and cognition. Oxford: Blackwell. Sweetser, Eve E. 1990 From etymology to pragmatics: Metaphorical and cultural aspects ofsemantic structure. Cambridge: Cambridge University Press. Traugott, Elizabeth Closs 1997 Unless and but conditionals: a historical perspective. In: Rene Dirven and Angeliki Athanasiadou (eds.), On conditionals again, 145-167. Cambridge. Cambridge University Press. Visconti, Jacqueline 1994 Entre hypotheses et conditions. Sur la semantique des connecteurs conditionnels complexes. Cahiers de linguistique fran~aise 15: 77108. Wing, C.S. and Ellin Kofsky 1981 Children's comprehension of pragmatic concepts expressed in because, although, if, and unless. Journal ofChild Language 8: 347365.

Pre- and post-positioning of wenn-clauses in spoken and written German * Peter Auer

This paper is concerned with the distinction between pre- and postpositioned (initial and final) wenn-clauses in German, and with the distinction between written and spoken language. A simple cross-tabulation of the two features [spoken/written] and [pre-/postpositioned] shows that initial wennclauses are preferred in spoken German, but final wenn-clauses are preferred in written German. An attempt will be made to explain these findings.

1.

Introduction

The fmdings and discussions in this paper are corpus-based. They are partly quantitative, partly qualitative. With respect to both dimensions, the claim is that a full understanding of the syntax of (particularly) spoken language eludes the possibilities of a purely introspective methodology. Of course, no (quantitative or qualitative) corpus-based investigation can do without a strong reliance on the analyst's knowledge ('intuition') about the language being researched; in fact, fmding valid generalizations always involves Gedankenexperimente, playing with structural changes in and recontextualizations of the 'examples' found to be used by the informants. On the other hand, not even the empirical starting point of the present investigation (i.e., the (differing) preferences of spoken and written language for post- and prepositioning) is available to a purely introspective approach, since it is of a quantitative kind. In order to reach an explanation of these findings, this quantitative analysis has to be complemented by an in-depth analysis of individual cases of usage. Such an analysis will pay attention (a) to the in-time emergence of syntactic patterns, including the details of their delivery such as hesitations, reformulations, break-offs, etc., and (b) to the interactional aspects of this emergence, including hearer feedback (or lack of it) and sequential placement. In this respect, spoken language research can profit in important ways from conversation analysis.

174 Peter Auer

2.

German wenn-clauses and English tactic and semantic overview

~clauses:

a brief syn-

This, of course, is not the place for a full contrastive analysis of the two constructions. In order to facilitate the reader's access to the examples to be discussed below, and in order to link up the present investigation with previous ones on tf-clauses in spoken (and written) Engiish, a short sketch of some important similarities and differences may, however, be in order here.

2.1. Syntax

The syntax of English suggests a (misleading) parallel between pre- and postpositioned adverbial clauses (including conditionals) since they can usually be exchanged without structural changes in either the main or the dependent clause. German syntax, on the other hand, treats the two positions quite differently. While post-positioned adverbial clauses always occupy the so-called post-field (Nachfeld), which is not obligatory, and are thus tagged on to an already complete syntactic pattern, pre-positioned subordinated clauses may be (and in written, normative language usually are) more tightly integrated into the syntactic structure of the following main clause: they occupy the so-called front field (Vorfeld) of the sentence, i.e., the uniquely available and obligatory position before the finite verb. Moving adverbial clauses from one to the other position therefore involves structural changes in the main clause: (1) a. wenn sie==n JOB haben wollen, () mUssen sie==n bisschen da aufn PUNKT kommen. (original utterance) 1 'if you want to have ajob, you need to get down to the point' b. sie miissen n bisschen auf==n PUNKT kommen wenn sie==n jOb haben wollen. (fabricated) 'you need to get down to the point if you want to have a job' Positioning the wenn-clause in the post-field (lb) instead of the frontfield (la) implies that another constituent will fill this position (in the present case, it is the subject pronoun Sie). The dominant syntactic pattern in which pre-positioned adverbial clauses occur in written German may therefore be called "integrative", while the English treatment is "non-integrative" (cf Konig and van der Auwera 1988: 103-109 for this

Positioning ofwenn-clauses in German

175

terminology and some further remarks). In spoken German, however, the fully integrated placement of the pre-positioned adverbial clause in the front-field is only one possibility. Alternatively, pre-positioned wenn-clauses may be followed by a resumptive particle (a localtemporal adverbial such as dann or one of its regional equivalents, e.g. na, denn, no, etc.; cf (2b»;2 or they may even be used in a nonintegrative way, rather like in English (cf (2c»: (2) a. wenn sie==n JOB haben wollen, () mUssen sie==n bisschen da aufn PUNKT kommen. (original utterance) 'if you want to have ajob, you need to get down to the point' b. wenn sie==n JOB haben wollen, () dann mOssen sie==n bisschen da aufn PUNKT kommen (fabricated) (same meaning) c. wenn sie==n WIRKlich n JOB hiitten haben wollen; () sie hiitten dann SCHON n==bisschen aufn PUNKT kommen miissen (fabricated) 'if you had really wanted a job, (then) you would have needed to get down to the point' For (2c), special conditions of use hold, and it may therefore be considered to be "marked" (cf Konig and van der Auwera 1988, Kopcke and Panther 1985, Giinthner 1999, and below).

2.2. Semantics

The semantics of German wenn-clauses3 is not strictly equivalent to English conditional tf-clauses either. The cognate of English if, German ob (> Germanic *eba), has lost its original (OHG/MHG) function of introducing conditional clauses, although remnants of this usage may still be found in (etymologically) composite concessive conjunctions such as Modem German ob+wohl and ob+gleich and in concessive conditionals of the type ob X oder nicht...('whether X or not'). Filling the gap, the temporal conjunction wenn (or rather, its predecessors, MHG swenne/swanne), a cognate of English when, has taken over most of its functions. As a consequence, the semantics of wenn-introduced clauses oscillates between a temporal and conditional reading in the indicative mood. (For this reason, German wenn-clauses, other than English tfclauses, cannot be called conditional clauses. 4 ) Other conjunctions are

176 Peter Auer

available for a nonambiguous encoding of a conditional or a temporal relationship. Disambiguation of wenn is possible on the basis of contextual information (i.e., the semantics of the remainder of the clause and/or its conversational context) in some, but certainly not all cases. Even in the following examples (which are among the clearest in my data) the paraphrases are not entirely beyond dispute. However, they do represent the prevalent readings ofwenn-clauses: (3) (temporal: consecutive) dann MELD ich mich morgen bei ihnen? (-) wenn «=SOBALD, SOWIE)) ich bei AUror angerufen hab, 'so I'll be in touch with you tomorrow as soon as I have given "Auror" a ring' (4) (temporal: simultaneous) TRIFFST du den () wenn «=WAHREND, sOLANGE)) du in PEking bist? 'will you see him while you are in Peking?' (5) (temporal: iterative) ich sprEch UNdeutlich, (1) LISpel auch n=b' etwas, (-) und (-) ich sprEch dann after zu SCHNELL. (-) wenn «=JEDESMAL WENN)) ich beGEIStert bin, oder (-) eh im element bin; (-) dann sprech ich zu SCHNELL, 'I speak inarticulately, I also lisp a little, and then I often talk too fast. whenever I am enthusiastic about something, or ehm get carried away; (-) then I talk too fast,' (6) (conditional: hypothetical) er will sie jetzt wieder HElraten, und die haben so ne FRIST ehm in den islamischen liindern dass innerhalb von nem halben JAHR oder so, muss die frau dann wieder zum MANN zuriick wenn «=FDRDEN FALL DASS, FALLS)) er sie DOCHwiederwil1 h. 'he now wants to marry her again and they have kind of a deadline ehm in the Islamic countries that within half a year or so the wife has to return to her husband in case he wants her back again'

Positioning ofwenn-clauses in German

177

(7) (conditional: factual) «radio phone-in, psychotherapeutic consulting; the caller has complained about having no-one to tum to with his marriage problems; the therapist recapitulates and formulates her advice» denn LETZTlich .h wenn ((=DA)) sie in ihrer verwAndtsch~ft niemand HAbn mit dems REden kOnnen, .h iih is=danns BESte, (-) sie wiirden zu am Eheberater gehn? 'for in the end, since you have nobody among your relatives who you could talk to, ehm it's best then to tum to a marriage counselor' Clearly, the temporal readings of wenn are not covered by English if, but by when instead. Some verbal and prosodic features of wenn-clauses may facilitate or even enforce one or the other reading: (a) the temporal, non-iterative reading is not available in sentences referring to past events; here, the temporal conjunction als takes over (while English allows when); (b) focussing adverbials such as stressed nur ('only') in the main clause strongly suggest a conditional reading of the (following) wenn-clause; (c) the particle schon (no English equivalent) in the wenn-clause suggests a factual-conditional reading; (d) immer wenn ('always when') instead of a simple wenn as a conjunction enforces a habitual-temporal/ contingent reading; (e) selbst wenn (' even if) and wenn ... iiberhaupt ('if ... at all') instead of a simple wenn enforce a conditional reading; (f) subjunctive (Konjunktiv II) in the wenn-clause enforces a hypotheticalconditional reading; (g) stressed wenn-conjunctions suggest a conditional instead of a temporal reading. Two special uses of wenn-clauses need to be mentioned. The first is the expression of concessivity through the combination of wenn and auch (wenn + auch or auch + wenn), roughly similar to Engl. even if/even though: 5 (8) a. auch wenn sie KEInen job haben wollen, () miissen sie=n bisschen da aufn PUNKT kommen. (fabricated) 'even if you don't want ajob, you need to get down to the point' b. wenn sie auch keinen JOB haben wollen, () sie miissen n bisschen da aufn PUNKT kommen. (fabricated) 'even though you don't want a job, you need to get down to the point' c. obWOIa sie KEInen job haben wollen, () miissen sie=n bisschen da aufn PUNKT kommen. (fabricated)

178 Peter Auer

'although you don't want to have ajob, you need to get down to the point' The auch wenn-construction (8a) differs from obwohl-concessives (Eng!. although, (8c» in that the truth of the proposition it expresses can but need not be taken for granted ("neutral epistemic stance"; cf Fillmore 1990, Couper-Kuhlen, 1999): whereas the proposition 'you don't want a job' is not asserted in (8a)/auch wenn, it is in (8c )/obwohl. Auchwenn-clauses therefore differ from if=-conditionals and resemble true (obwohl-) concessives in that the presupposed generic statement is negative (for the above example: 'someone who does not want a job does not have to get down to the point'). At the same time, they differ from true concessives and are similar to true conditionals in that the truth of the antecendent may but need not be asserted. Note that, differently from auch wenn pre-positioned wenn auch-clauses (8b) co-occur with nonintegrative word order in the consequent, and always receive a factual interpretation. Finally, it should be noted that German wenn-clauses occur sometimes as obligatory constituents of the verb. 6 (English often uses nonfinite forms such as participle or infinitive clauses for this purpose, although if=-clauses are also possible.)

(9) das EINfachste, da ham sie RECHT, das war fiir uns, wenn sie mal==n MOnat (-) im teleFONmarketing ARbeiten wiirden. 'the simplest solution for us, and here you are right, would be if you could work in our direct marketing sector for a month' Syntactically speaking, wenn-clauses of this kind can be replaced by dass-(complement) clauses (das Einfachste ware, dass sie mal im Telefonmarketing arbeiten). Semantically speaking, various differences result from the choice between dass- and wenn-complements; most of them pertain to the presumed status of the information in the complement clause (cf Eisenberg 31994: 365f for some further discussion).

3.

German wenn-clauses from a quantitative perspective

The observations in this section are based on a collection of 500 wenn7 clauses taken from a corpus of spontaneous, direct conversations. All instances of wenn were considered for analysis, apart from obvious syn-

Positioning ofwenn-clauses in German

179

tactic break-offs in the wenn-clause, some non-reconstructable utterances, and the comparative uses of als/wie wenn (see note 5). In Figure 1, the total of n=500 tokens is broken down according to the position of the adverbial clause relative to the main clause: pre-positioning, postpositioning, parenthetical positioning within the clause, 8 independent use of the wenn-clause as a turn-constructional unit of its own, and a residual category of ambiguous cases (e.g. apo-koinu constructions, see below). There can be no doubt that the front position is preferred in spoken German. 60,00 50,00 40,00 30,00 20,00 10,00 0,00

IJ percentage

prepositioned

parenthetical

postpositioned

single

others

56,00

4,00

32,60

5,40

2,00

Figure 1. Position of German wenn-clauses relative to the main clause~ n=500.

The results agree with Ford and Thompson's fmdings on If-clauses in English conversations, according to which initials outnumber finals by a ratio of 4:1 (n=316, initial=81%, fmal=19%; Ford and Thompson 1986: 362), with Ford's findings based on a smaller collection (n=52, 50% of which where preposed, 35% postpositioned, and 15% single; cf Ford 1993: 24), and with more general claims about a universal preference for pre-positioning of antecedents in conditional constructions (Greenberg 1963). Note, however, that the preference found in the English data for pre-positioning of conditional clauses does not extend to temporal (e.g., when-) clauses; rather, Ford (1993: 24) found these to follow their main clauses by a ratio of 2:1. Given the ambiguity of German wennclauses (in the indicative mood) between a conditional and a temporal reading, it may be asked if the preference for initial placement of wennclauses holds for both. To answer this question, those instances ofwennclauses have been singled out (n=203) which have either a clear tempo-

180 Peter Auer

ral or a clear conditional reading (based on the substitution tests and criteria discussed in section 2.3).9 Of the n=203 disambiguated wenn-clauses, 24% have temporal, the remainder conditional meaning. Exactly half of the wenn-clauses with temporal meaning are prepositioned and postpositioned respectively. (Of the 76% unambiguously conditional examples, about two thirds are pre-positioned (55% of the total), one third is postpositioned (21 %).) There is, then, a clear difference between conditional and temporal uses: only for the former does the preference for pre-positioning hold. Since the majority of German wenn-clauses are semantically ambiguous between a temporal and a conditional reading, this fmding also suggests that, taken as a whole, they behave syntactically like (English) conditional rather than temporal clauses. Figure 2 shows the percentage of integrative, resumptive and nonintegrative constructions among the pre-positioned wenn-clauses in the sample (n=280).

60,00 50,00 40,00 30,00 20,00 10,00 0,00 integrative

resumptive

nonintegrative

others

25,72

47,83

18,48

9,42

Figure 2. Integration (%) of pre-positioned wenn-clauses into the subsequent main clause (n=280)

The relatively large residual category ("others") covers wenn-clauses plus subsequent main clauses within larger hypotactical constructions (see below example (26)-(28)). Again, the results are very clear: resumptive constructions are preferred to fully integrated and totally nonintegrated constructions. The canonical, integrative construction of standard written German only plays a secondary role in spoken German.

Positioning ofwenn-clauses in German

181

Some comments on non-integrative wenn-clauses in German are necessary at this point. Pre-positioned wenn-clauses occurring in the pre-front field of a sentence are basically of two types (cf Auer 1996). We find instances which cannot be positioned in the front field (i.e., integrated into the main clause); in other words, the only available pattern for them is non-integrative syntax. This is sometimes for syntactic reasons; in particular, yes/no-questions and imperatives, which are verbinitial syntagms in German, do not have a front field, and in wquestions, the w-question word is usually said to occupy the frontfield. 10 In these contexts, adverbial clauses either need to be postpositioned (despite the general preference for pre-positioning), or to be non-integrative. Of the 45 questions/imperatives in the sample, 16 have pre-positioned wenn-clauses, i.e., non-integrative word order (cf (10», while 29 have post-positioning; this means that the normal preference is reversed in this syntactic environment. (10) ich mein ich muss ihnen (-) ganz SCHNELL und GANZ: vehement sagen wenns IRgendwie gEht (-) fahrns InN 'I mean I have to tell you without hesitating and very vehemently: if you can make it at all, go there!'

However, there are also semantic reasons why certain wenn-clauses have to occur in the pre-front instead of the front field. This is the case for "speech-act related" 11 wenn-clauses which do not conjoin two propositions on the content level; often, they are used in order to mitigate subsequent face-threatening acts (such as, in the following example, an interruption). The apodosis is asserted independently of the protasis, and this semantic independence corresponds with obligatory syntactic non-integration: (11) (Gob interview» wenn ich (-) grad WElter ausfiihren darf; (0.5) Sie wissenja in de: in der A Utoinduschdrie .h herrschen SEHR grofJe k' konkurRENZ, markt 'if I may continue elaborating on that; (0.5) you know that in the car industry there is a lot of competition «etc.»' In such cases, the marked position of the wenn-clause in the pre-front

field helps to contextualize a marked (non-referential) semantic interpretation.

182 Peter Auer

But there are also contexts in which non-integrative syntax is frequent although not obligatory. For instance, there is a tendency for nonintegrative clause-combining to occur in concessive wenn nichtconstructions: (12) wenn auch die theoRIE; (-) eh (-) so IRgendwo mal gehOrt wurde im KOPF? (-) eh das UMsetzen das ist ja das entSCHEldende, 'even though the theory (-) ehm (-) may have been heard somewhere in one's head (-) the decisive thing is putting it into practice' Another frequent function of non-integrated wenn-clauses is topicalization; in this case, the wenn-clause is typically followed by an anaphoric pronoun back-referencing the proposition expressed in the wenn-clause as a whole, or an element contained in it. In the following example, the wenn-clause introduces a new discourse referent or topic; it is in many ways equivalent to other topicalization constructions (such as a cleft construction: was Ihre Fragen angeht, die konnen Sie jetzt stellen), with the additional implication that the speaker is not certain about the relevance of the new discourse referent for the co-participant.

(13) also wenn sie FRAgen ham zwischendufch, eh DIE konnen Sie ruhig STELlen? 'well if you have any questions in between, ehm you can ask THEM of course. ' A similar topicalization (not of a single referent, but of a whole proposition) is involved in the following example:

(14) un wenn ich mein Eltem anrufh wUrde, ==DS wiirde AUCH nix bringn. 'and if I called my parents, that wouldn't be any use either. ' Here, the wenn-clause could even be entirely replaced by an infinitival construction (meine Eltern anzurujen), since potentiality is already expressed by the conditional verb form wiirde... bringen and redundantly coded by wenn. Finally, non-integrated wenn-clauses often express emphasis and lend an emotional meaning to the utterance: 12

Positioning ofwenn-clauses in German

183

(15) wenn WIRKlich==n ganzen tag das telefon klingelt, und acht STUN' (-) man IS hinterher' < man WEISS was man> getan hat. () geb ich ehrlich ZU 'if the phone really rings all day, and eight hOD' (-) afterwards you are' you know what you have done. i have to admit that. ' In (15), the speaker describes her working-day in a call-centre and wants to underline that dealing with callers is a tiring job~ one of the strategies used to convey this meaning is the non-integration of the protasis into the apodosis.

4.

Some reasons for pre- and post-positioning

What are the advantages of pre-positioning wenn-clauses? This question seems less difficult to answer than the opposite one of why a certain number of these clauses - roughly a third in our data - are postpositioned. We will deal with each question in turn.

4.1. The advantages ofpre-positioning To start with, it should be noted that the preference for pre-positioned wenn-clauses is not just a quantitative fmding but is reflected in speakers' changes in the design of an emerging syntactic pattern 'in midstream'. Particularly striking are cases such as Ex. (5), in which a postpositioned wenn-clause is retrospectively turned into a pre-positioned one via what might be called an apo-koinu construction: The K01VOV here, of course, is wenn ich beGEIStert bin, oder (-) eh im element bin. It seems that the speaker, having completed the threepart list of his verbal handicaps, wants to qualify the last item retrospectively. He could have done this by simply adding the wenn-clause in the post-field but recycles this last component instead, with the wenn-clause inserted before it. The wenn-clause here is both final and initial. instances in which a clause is broken off and a wenn-clause is inserted before it is re-started (as in (16)) are also evidence for the interactional relevance of pre- vs. postpositioning.

(16) ichfahr (-) wenn (-) wenns iiberHAUPT geht dennfahr ich NA:CH(er) erscht in Urlaub,

184 Peter Auer

'I'll have if if it works out at all then I'll only have my holidays afterwards' So why this additional effort? There seems to be some kind of cognitive 'naturalness' in the way in which conditionals create the ground - or, in more recent but equally metaphorical parlance, set up a "mental space" (Fauconnier 1985) - in which some hypothetical or factual proposition is located. 13 For cognitive reasons, it is the grounding which (iconically) precedes the focal proposition, and not the other way round. Ford, for instance, suggests that "the prevalence of initially placed if-clauses may reflect the general tendency to signal «... » that the interpretation of the coming clause will be, in some general way, limited by the contents of the if-clause" (1993: 15). Further evidence for the 'naturalness' of this position can be derived from the affinity of conditional clauses and topic-introducing devices (topics precede comments), for which some evidence has been given in the preceding section (see Haiman 1978, Ford and Thompson 1986 for an in-depth treatment of this line of argumentation), and from the affinity of conditional and causal clauses (where causes iconically precede their effects). The advantages of this discourse function seem to outweigh the cognitive costs linked to the deployment of a syntactic pattern which projects considerably into time. It may not have been sufficiently taken into account in previous research on clause positioning, however, that this projection in time has an interactional side as well: 14 speakers who open up far-reaching syntactic gestalts claim the tum for at least the time which is necessary to bring them to a well-formed conclusion. In other words, producing a wennclause gives the speaker the right and obligation to go on talking; it functions as a tum-holding device until the formulation of the consequent is completed. There are numerous cases in the data in which highly complex turns emerge in this way, since the speaker uses the space between a gestalt-opening wenn-clause and a terminating main clause for detailing the "mental space" opened up by the first component. Two elaborate examples (as they seem to be typical for institutional talk) are (17) and (18): (17) (Gob interview; applicant B is talking about his previous employment in a West German consultant company which, however, withdrew from East Germany, despite the fact that it had highly experienced consultants»

Positioning ofwenn-c/auses in German

B:

I: B:

185

zum beispiel einen herren, () KELler? (-) eh der () is: () FONFundzwanzig jahre unterNEHmensberater? () der hat=n STAM!vfklientel in uh es ah KAnada? mhm, und DER war naturlich, (-) ein FACHmann. (-) aber er KOMMT,

()

I: B:

I: B:

I: B:

I: B:

in die neuen BUNdesliinder? () «acc>er war ja nu> (-) eh hatte es ja gar nicht mehr NDtig gehabt; =da () so [VIEL] (-) zu REIsen, [mhm,] aber (-) er IS in die neuen BUNDdesliinder gekommen, (-) um auch etwas zu be WEgen. (-) aber wenn er dann nur auf der STRASse () sitzt, (-) und DANN (-) den () kliENten () mit nach schweRIN nehmen muss um=n FDRderantrag zu stellen; (-) dann wieder zur BANK, (-) und die BANK sagt () wir brauchen erst=ne ZUstimmung von dem FDRderinstitut, =

mhm, vor[her () ko]nnen wir nicht die geSAMTfinanzierung, [(h) ]

wie mit KDpenick. ja und () fund er da]nn NUR aufder STRAsse ist; [(h)] (-) dann SAGT er das LOHNT sichfiir mich nicht. (-) dann bleib ich LIEber () in nordrhein westFAlen.

'B:

for instance a Mr Keller ehm who has been a consultant for 25 years he has his regular clients in the U.S. and Canada

I:

mhm,

B:

and he was a specialist of course. (-) but he is coming to the New States (.) he certainly had (-) ehm he had no need to do that any more; to travel so [much there

I: B:

[mhm, but (-) he did come to the New States, (-) in order to get something moving. (-) but if he is on the road all the time (-) and then (-) he has to take his client with him to Schwerin in order to hand in the proposal for the subsidies; (-) and then back to the bank, (-) and the bank says (.) first we need the subsidizing body's consent

186 Peter Auer

I: B: I: B:

mhm, before [that we cannot (do) the total financing [(h) like with Kopenick 15 and (.) and he is just on the road then he says this isn't worth it for me. then I'd rather stay in North Rhine-Westphalia «a West German state»'

This passage is embedded into a larger report the applicant gives of his participation in a West German consultant agency in the New States, which however closed down its East German office, making him redundant. The interviewer does not seem to know the company and questions its importance on the market. The applicant counters by stating that although small, the company had very professional consultants. At the same time, he has to deal with the interviewer's innuendo that the company withdrew from the East German market because it was not working successfully. In this context, the case of "Mr. Keller" is mentioned, an experienced consultant who was disappointed by the kafkaesque way in which state and bank authorities made it hard for new enterprises to get subsidies, and returned to the Old States. After he has been portrayed as a successful consultant who came to East Germany mainly for idealistic reasons, "Mr. Keller's" dissatisfaction with the situation is described in a complex tum construction which starts out with a wenn-clause (wenn er nun aufder Straj3e sitzt... ). In the given context, the interpretation is not hypothetical but refers to a (factual) state of affairs ('since he was always on the road... '), which is established as the ground from which some conclusion can be drawn. Before this conclusion is reached, however, the speaker elaborates at considerable length on the unfortunate situation in which "Mr. Keller" and his clients found themselves; in four clauses each introduced by (und) dann, the various fruitless journeys between the financing bank and the state authorities in Schwerin are described. Towards the end of this elaboration (securely produced by the speaker within the realm of his own tum, since a syntactic projection - that of the when-clause - still remains to be taken care of), the interviewer produces some recipiency tokens which, although not claiming the tum (cf their reduced loudness, indicating non-competitiveness), nevertheless acknowledge the speaker's point: two laughter particles and one comment (wie mit Kopenick) display understanding. Only after this feedback does the

Positioning ofwenn-clauses in German

187

speaker close the syntactic gestalt with two resumptive dann-clauses. Their content is higWy predictable, given the fact that it has been mentioned before that the company closed down its East German branch. It seems, then, that what the speaker wanted to convey by this complex turn is not so much this consequent but rather the details of the situation which led to it. The relevant information of this complex construction is what is produced BETWEEN the initial wenn-clause and the final dannclauses. The speaker employs the projecting force of the first in order to claim conversational space for himself: and makes use of this space as long as he needs it to 'convince' the recipient of his point (as evidenced by the recipient's responses). The 'orderly' conclusion of the turn is produced as soon as this purpose is reached. The following extract similarly shows how pre-positioned wennclauses can be used to claim conversational space: (18) (bulimia therapy) M: aso ich hab ma mit einer zuSA!vfMgewohnt, == ==und .h die hab ich EH nich so leidn kOnn un sie mich AUCH nich, und dann hab ich IMmer so.h (0.5) und (-) DIE: is schon wesentlich DICker als ich; == ==und dann hab ich ECHT immer gedacht (0.5) ich hab so alles des (-) AUF se projeziert und wenn se viel geGESsn hat, == ==die hat sich .h SAHne n ganzn becher SAHne mit Apfelschnittchen drin gegessn. == ==und das warfiir mich ECHT der ABscheu. == .h des is ja wohl (1.0) des is FURCHTbar (1.0) wie KAMmer denn sowas ESsn un auch noch mit gUtm ge WISsn. 'M:

you see I once lived with a girl and I couldn't really stand her and neither could she me and then I always and she really was a lot bigger than I was and believe me I always thought I projected everything on her and when/if she ate a lot, she put cream a whole cup of cream she ate with slices of apple in it

188 Peter Auer

and to me that was really disgusting. then I thought .h isn't that (1.0) that is really appalling how can you eat anything like that and without even feeling guilty. ' Once more, a speaker is involved in telling a story which in this case is supposed to show how she projected her own feelings of guilt for eating too much onto her flatmate. And once more, a wenn-clause is the first component of a syntactically cohesive turn construction which spans six intonation units. The speaker does not go into gestalt closure (apodosis) after the wenn-clause, but rather parenthetically includes information detailing the claim that the roommate 'ate a lot', and how she herself reacted to that emotionally. Only then does a (dan)n-clause follow which ties back to the initial part of the turn, where a story concerning 'projection' (ich hab so alles des aufse projeziert) was announced. There is only one legitimate way for a recipient to share (or rather, intrude into) the conversational space which a wenn-projection creates for the current speaker: by becoming a co-speaker herself, i.e, by collaboratively producing the gestalt-closing apodosis matching the already produced protasis (cf. Lerner 1991). Both inserted material between protasis and apodosis and collaborative constructions pivoting around this transition suggest that there is some interactional work going on, and that, at least in a substantial subgroup of examples, the construction is not planned and executed as one whole, but rather develops in (at least) two steps.

4.2. Why post-positioning at all?

If pre-positioned wenn-clauses are both cognitively more 'natural' and interactionally more advantageous than post-positioned ones, why do the latter occur at all? Two reasons have already been mentioned in section 3: wenn-clauses may be used for expressing the temporal circumstances of an event, and since temporal adverbial clauses do not follow the preference for pre-positioning, wenn-c1auses of this semantic type need not do so either. I Secondly, it was shown that the absence of a front-field in questions and other verb-initial syntagms makes their postpositioning more likely. 17 There are, however, other important reasons. First of all, it may be asked if there are any further syntactic environments in which post-positioning is preferred or even necessary.

Positioning ofwenn-clauses in German

189

There is indeed another construction in which the front-field is not available: that in which the wenn-clause plus subsequent clause are themselves embedded into a larger construction. The various types of embedding show different patterns with respect to the possibility of prepositioning. As in Ford and Thompson's English data (1986: 359), final positioning is preferred "when a conditional clause occurs within a nominalization, an infinitive, or a relative clause". Take, for instance, the following case of a relative clause: (19) (therapy session) TM: s==ESsn isch wie? ein Teddybiir. == TW: ==ja, TM: den' den sie: (-) .h mit sich RUMtragn. (2.0) damit SIE <

nich allEin sein miissn. > un dem==mer (-) sich RANzieht, ja? (-) (-) wenns HARTwird; (3.0) an dem=mer sich FESCHThiilt, (2.0) wem==mer EINsam isch, (1.0) nd der ii: berall MIT muss. 'TM: eating is like a teddy bear. TW: yeah TM: who who you carry around with you.

so that you don't have to be alone. and whom one holds close, right? (-) (-) when life becomes hard; whom one clings to, (2.0) when one is lonely, and who has to come along all the time. ' Both wenn-clauses in this extract are part of a relative clause introduced by an oblique relative pronoun, i.e., their matrix clause is itself subordinated, and therefore has verb-final syntax (cf the placement of the finite verbs ranzieht and festhiilt). Here, the wenn-clause cannot be placed in front of the relative clause (*und wenn's hart wird den man sich ranzieht); pre-positioning would require a superordinate main clause instead of the relative clause (und wenn's hart wird, zieht man sich den ran).18 The same applies to dependent clauses introduced by wie 'as', obwohl 'although', weil 'because', etc. which likewise do not allow initial wenn-clauses. However, subordination by the most frequent complementizer dass ('that') shows a different pattern. Here, we frequently encounter initial placement of the pre-positioned wenn-clause be..fore the complementizer dass:

190 Peter Auer

(20) MEIN interesse is natiirlich WENN ich da: .h schon als POSTdoc auf==m ZEITvertrag bin; dass ich wahrend de dieser ZEIT dann; (-) auch==n paar ergebnisse MITnehme 'my interest of course is if I am there as a post-doc on a temporary contract that I can take at least some results with me during this time' (21) und DESwegen war es natiirlich; (-) fir uns WUNschenswert; ()

dass sie so friih wie MOGlich natiirlich anfangen. 'and therefore of course it would be desirable for us if we could come to an agreement that you start as soon as possible' The additional stress on wenn in these examples may give us a clue to the origin of this construction; arguably, it underlines the semantic link between antecendent and consequent. Fronting the wenn-clause to a position before the dass-complementizer may be another way of focussing on the semantic link established by wenn. 19 Note in passing that the fronting of the wenn-clause renders its scope ambiguous both in (20) and (21): it mayor may not include the initial phrases mein Interesse ist natiirlichl ...ware es natiirlichfir uns wiinschenswert (i.e.: 'of course, if I am only there as a post-doc on a temporary contract, then my interest is to take at least some results with me' and 'if we could come to an agreement it would of course be desirable for us that you start as soon as possible' respectively).20 In addition to these syntactic constraints, there are semantic-syntactic reasons for post-positioning wenn-clauses. In particular, wenn-clauses in complement function are usually post-positioned (cf (9) above). As a rule, the main clause contains an evaluative two-place predicate, with the wenn-clause expressing the proposition which is evaluated.21 The opposite serialization is not unacceptable; nevertheless, it is very rare. The dominant pattern obviously parallels that of dass-introduced complement clauses which can, but rarely do, precede the main clause as well. Complements make up ca. 25% of all the post-positioned wennclauses in the spoken materials investigated. Finally, and most importantly, post-positioning of wenn-clauses is linked to the pragmatic status of the proposition they express, and to the interactional possibilities this position opens up both for the speaker and

Positioning ofwenn-clauses in German

191

the hearer. As outlined in section 1, final subordinated clauses in German are added onto an already complete syntactic structure. They are therefore a straightforward means for expanding a syntactic gestalt, and thereby the tum-at-talk. This is particularly obvious in cases where syntactically complete syntagms preceding the wenn-clause are marked as terminal by intonation, e.g. by a pitch fall to the speaker's base line (full stop in the transcription); the wenn-clause then appears as an afterthought, or epexegesis (cf Auer 1991):

(22) «hypothetical talk about a situation in which two people are in conflict over where to put the cup for the coffee; B is asked to mediate» ich wiird [einfach] die () die tasse kaffee NEHmen, B: II: [

h:m,] B: und eh (-) WEGste//en. (-) .hja? () eh==s da stunden/ang streiteREIen gibt, wiird ich sagen, a/[so: ] jetzt is:II: [hm,} (0.5) II: da hiitt==ich iirger [mit IHnen. B: [sch/uss aus ENde? II: [wenn SIE mir dann auch noch den KAFfee (wegschlieften.)] B: [he he he he he he 'B: II: B:

II:

II: B: II: B:

I would [simply take the (.) the cup of coffee

[mhm and ehm (-) put it away. (-) you see? before they start quarreling for hours, I would say right [now it is [hm,] (0.5) then I would have trouble [with you. [over and out [if you (shut away) my coffee [he he he he he he'

At a point where B has already suggested to simply 'taking away' the disputed coffee cup, but is in the middle of a syntactic construction elaborating on this proposal (eh's da stundenlang Streiterei gibt wiirde ich also sagen: Schluss, aus, Ende) II intervenes during an intra-tum

192 Peter Auer

hesitation pause to refute this solution: 'if you did that, there would be trouble between the two of us' (i.e. between the mediator, B, and one of the two people quarreling, i.e. himself). The utterance is linked to B's proposal by the initial anaphoric da; it is semantically and syntactically complete, and being marked by a final fall, it certainly is a candidate for a complete tum. However, B does not pick up this refutation, but continues with the production of the unfinished syntagm in another piece of simultaneous talk. Sequential structure and temporal development are now out of phase: a response has been produced to an utterance which is still in need of being completed, and is only completed after the response. In this context, II's following wenn-clause, syntactically expanding an already complete turn/syntagm, can be seen as a skillful way of re-aligning sequentiality and timing: it re-instantiates II's refutation of B' s proposal without repeating it, by retrospectively transforming a simple construction into a hypotactical one with a post-positioned adverbial clause. Semantically, this expansion adds nothing new: it just restates what B herself has said before. The possibility of such an expansion is not only available to the speaker but also to the recipient, of course, who may become a cospeaker and co-producer of the emerging syntactic pattern by adding a wenn-clause himself/herself: (23)

L: S: L: S: L:

S:

.h dann: eh () wir' der Hundwird auchjetz zunehmend ruhiger; mHM (-) des GUT so; (-) JAja des==also wird langsam (a)==richtiger HUND; aHA hm, (-) wenn==er (nicht mehr) abhaut, (-) hat (name) des ANgebot jetz fiir den zaun? ==

'L: S: L: S:

then ehm beco the dog is becoming more and more calm now; this is how it should be; yeah slowly he's tuming into a real dog. I see

L:

hm, if he doesn't escape (any more),

S:

did NN get the offer for the fence in the meantime?'

Positioning ofwenn-clauses in German

193

So it is not only the transition between a wenn-clause and its subsequent main clause which is sensitive to turn-taking, but also the inverse transition between a (main) clause and its subsequent wenn-clause. But obviously, there is an important difference: while in the first case an open syntactic projection is in play, in the second case the first speaker has already come to an orderly completion of the sentence/turn. Post-positioned wenn-clauses thus offer the possibility not only of expanding a turn, but also of expanding a sentence by adding a postfield constituent. At least example (22)22 also points to an important pragmatic feature of such expansions: its low information value. Indeed, this applies to a very large number of post-positioned wenn-clauses. Often it is the whole previous text which functions to build up the 'mental space' that is necessary to come to the conclusion expressed in the main clause, while the post-positioned wenn-clause only summarizes this preceding text, sometimes slightly changing the focus. In (24), the introductory adverbial insofern explicitly establishes this resultative link between pre-text and conclusion, while the post-positioned wenn-clause just repeats what is known from the previous conversation anyway (the wenn-clause is factual here): (24)

«after a long discussion of the applicant's career aspirations in the bank, and an equally long description of the branch of the bank in Stralsund and its sophisticated private customer service, which seems to match these wishes» (( ..)) das HAM wir alles in stralsUnd, also inSOfern, (-) eh ware das==ne ideAle (-) STELle, (-) wenn sie (-) praktiZIEren wollen im verTRlEBSbereich. im KUNDdennahen bereich. 'we've got all that in Stralsund, so in that regard, this would be an ideal position if you want to be a trainee in the sales department. in client-oriented business.'

As in other, similar cases of low-relevance wenn-clauses in final position, the front-field is used here for a connecting (anaphorical) adverbial, which is preferentially placed in sentence-initial position, where its indexical meaning is most easily processed. Since only one constituent may be placed in the front-field, this position is not available for the wenn-clause any longer. Since post-positioned wenn-clauses are often of low pragmatic relevance, upgrading their informational value requires special means; a standard technique for doing so is the use of focussing particles such as

194 Peter Auer

(stressed) auch, nur or dann, or a combination of these. In this case, it is the focussing particle which projects syntactically: it requires a constituent to follow which is in its scope. Therefore, wenn-clauses such as the following cannot be treated in the same way as post-field wenn-clauses in general: they do not expand an already complete syntactic pattern but rather close a gestalt projected by the particle. (25) «about wearing glasses)) ich zieh==se nur DENN ouf wEnn==i==se wIrklich (-) Effektiv brOuch 'I only put them on then when I really and positively need them' Ex negativo, the necessity of using such focussing particles in order to upgrade the following wenn-clause to rhematic status is evidence for the (sub)thematic status which wenn-clauses usually have in the post-field.

5.

Pre- and post-positioning of wenn-clauses in written German

In the last sections, it has been shown that wenn-clauses are preferentially pre-positioned with respect to their main clauses in spoken German, and that this serialisation has a number of cognitive and interactional advantages. It has also been shown that the more marked structure, i.e. post-positioned wenn-clauses, which does occur in about a third of all instances, has its own specific contexts of usage. These are partly due to (a) syntactic constraints on pre-positioning in superordinate clauses without a pre-field or in which two complementizers occur in adjacent position; partly to (b) tum-taking (afterthought position); partly to (c) semantic-syntactic reasons (wenn-clauses in complement function are postpositioned); and partly to (d) pragmatic reasons (postpositioned dependent clauses are thematic or subthematic, unless focussing particles indicate the contrary). In written German, wenn-clauses are generally less frequent than in spoken discourse, a finding which contradicts the frequent claim that spoken language avoids syntactically complex constructions: the frequencies of wenn-clauses per 100 words in the corpus of spoken language used above is 0.54, but in a corpus of written language, taken from the newspapers DIE ZEIT (politics section) and Frankfurter Allgemeine Zeitung (culture section), it is 0.33. 23 To put it differently,

Positioning ofwenn-clauses in German

195

every 186th word is wenn in our spoken corpus on an average, but only every 300th word in our written texts. Once more, this fmding is in line with comparative work on written and spoken English (Ford and Thompson 1986, 354: 0.72 vs. 0.46; similarly: Beaman 1984 and Biber 1986), but also with previous work on German (Leska 1965: 450). There are of course also qualitative differences between the wennclauses used in the two corpora; in particular, certain rather idiomatic patterns (constructions) seem to be more or less exclusively used either in spoken or written language. For instance, the reduced wenn-(dann)constructions of spoken German (cf section 2.2. above) do not occur in the newspaper corpus, while, on the other hand, the topicalizing causative construction wenn p dann (deswegen), weil q ('if p, then that is because of q'), as in (26), seems to be used exclusively in writing. (26) (DIE ZEIT Nr. 8 17.02.1995) Wenn wir das Leugnen von Auschwitz, anders als das Leugnen der kopernikanischen Wende, unter Strafe stellen, dann deswegen, weil es uns nicht nur hypothetisch angst macht. 'If we punish the denial of Auschwitz, and not the denial of the Copernican revolution, then that is because it does not make us feel afraid only hypothetically. ' Also, and contradicting received wisdom according to which written language is more logical and more explicit, we fmd instances of wenn as a conjunction in the newspaper texts with semantics which are exceedingly vague, as for instance in (27): (27) (F.A.Z 19.06.1993, S. 27 / Nr. 139) Aber wenn zum stets und instiindig angestrebten "Weltniveau" der DDR die Stellvertreterschaft des iiberragenden, fortschrittlichen, darin sagar selbst "biirgerlichen" Kulturerbes gehorte, dann iiufJert sich nun das AufschliefJen zur Weltoffenheit seltsam kleinmiitig. 'But if the representation of an outstanding and progressive cultural heritage, one which includes even the "bourgeois", belonged to the ever and urgently sought after "international standard" of the GDR, then growing into cosmopolitan openmindedness expresses itself rather timidly nowadays. '

196 Peter Auer

Here, the wenn-dann construction seems to vaguely express something between adversativity and concessivity. However, these differences only affect a relatively small number of examples and are not directly linked to the positioning of the wennclause. The important question for the present discussion is rather whether the preference for pre-positioning of wenn-clauses is also to be found in written German (as it is in written English, cf Ford and Thompson 198624). Fig. (3) shows that this is not the case: 60,00 50,00 40,00 30,00 20,00 10,00 0,00

a percentage

Figure 3.

51,60

1,60

1,00

Percentage of post- and pre-positioned wenn-clauses in written German (n=626)

In the written materials, post-positioned wenn-clauses are almost one and a half times more frequent than pre-positioned ones, while the opposite ratio is found in the spoken material. 25 As would be expected, the number of integrative pre-positioned wenn-clauses is higher than in conversational language (65.56%), and both the number of nonintegrative (9.28%) and resumptive (26.16%)26 structures is markedly reduced. The preference for post-positioning is only slightly less pronounced in unambiguously conditional wenn-clauses than in unambiguously temporal ones (65 % vs. 71%) (n=301). What could be the reason for this reversal of preferences for postand pre-positioning in written German compared to spoken language? Three factors seem to be primarily responsible for it. First, the number of (almost exclusively final) wenn-clauses with a preceding focus particle in the main clause is about eight times as high in the written as in the spoken material (40 vs. 5 occurrences). An example is:

Positioning ofwenn-clauses in German

197

(28) (DIE ZEIT Nr. 7 10.02.1995) Danach durfen Frauen ungewollte Schwangerschaften in den ersten drei Monaten nur beenden, wenn sie sich vorher haben beraten lassen: in einer Beratungsstelle und vom abtreibenden Arzt. 'Accordingly, women may only terminate an unwanted pregnancy during the first three months if they have undergone counselling: in an advice centre and also by the physician who does the abortion. ' Secondly, although embedded wenn-clause plus main clause constructions are not more frequent in the written than in the spoken material, all 52 wenn-clauses of this type are post-positioned, while a majority of them (34 of 54, all of which are embedded into dass-constructions) are pre-positioned in the spoken materials. In other words, fronting of wenn-clauses before the embedded matrix clause such as in (20) or (21) does not occur in the newspaper texts. Finally, one of the important reasons outlined above for frequent prepositioning in interactional language use is simply not applicable to writing: this is the need for the speaker to claim conversational space for the production of a larger tum, through projecting syntactically beyond the current clause. Instead, another factor becomes relevant: in writing, the wenn-clause may become so complex that processing it would become difficult even in reading if it was pre-positioned with respect to its main clause; cf for instance: (29) (DIE ZEIT Nr. 03 13.01.1995) Und man denkt an Talleyrands Feststellung: "Hochverrat ist eine Frage des Datums", wenn man sich daran erinnert, daft Hans Modrow in einer Phase als Reformer und Hoffnungstriiger galt, aber in der niichsten fiir schuldig erachtet wurde, weil er mitverantwortlich warfiir das DDR-System. 'And one thinks of Talleyrand's statement: "High treason is a question of the date" when one recalls that Hans Modrow was regarded as a reformer and as a source of hope in one phase, but that he was found guilty in another, because he shared responsibility in the GDR political system. '

198 Peter Auer

Neither of these reasons for post-positioning in written German can explain the difference between the English and the German results, of course. Why should these same reasons not lead to a preference for post-positioning in written English as well? If one was looking for a structural explanation, one would probably try to find an answer based on the most prominent difference between English if- and German wenn-clauses, i.e. the semantic ambiguity of the latter. For instance, it might be argued that since wenn can often be interpreted either conditionally or temporally, German newspaper journalists try to disambiguate their sentences by using other, strictly conditional conjunctions instead, such as falls, im FaIle dass, fiir den Fall dass, sofern or soweit. However, this hypothesis receives little empirical support in my data: not only are these conjunctions very rare in the newspaper texts (a total of 31 tokens!), they also fail to show a positional distribution different from that of wenn (6 initial vs. 17 fmal tokens, with 8 parentheticals).27 Another possibility to express conditionality in German which is not available in present-day English (apart from peripheral cases) is inversion (as in: kommst du zu spat, bestraft dich das Leben := wenn du zu spat kommst, dann bestraft dich das Leben 'if you are late, you will be punished by life'). This possibility is almost never used in spoken German because of its bookish and high-register connotations but its occurrence cannot be excluded in rather conservative newspapers such as DIE ZEIT and F.A.Z.; and since the distribution of pre- and post-positioned conditional clauses with inversion is unknown we cannot exclude a balancing effect, for instance due to a preference for pre-positioning in this case. Since this syntactic pattern can only be quantified in syntactically labelled corpora, there is no possibility to test this hypothesis in a straightforward way in our materials. However, preliminary analyses of some texts suggests that inversion hardly occurs in newspapers. Alternatively, one might look for a non-structural explanation which would locate the reason for diverging English and German patterns on the textual level, possibly in the stylistic preferences of English and German text composition. The matter clearly awaits further investigation.

6.

Conclusion

In this paper, I have looked at the placement of German wenn-clauses in spoken and written texts. Various explanations for the general quantitative results - i. e., that spoken German prefers pre-positioning, written

Positioning ofwenn-clauses in German

199

German post-positioning - have been presented and discussed on the basis of individual conversational contexts in which wenn-clauses occur. The general conclusion of this study is that the supposed parallel between 'left' and 'right' in syntax (suggested by parlances such as 'left extraposition' vs. 'right extraposition', or 'left-adjoined' vs. 'rightadjoined') is fundamentally mistaken when applied to spoken syntax; in speaking, to be sure, there is no 'left' and 'right', but only 'earlier' and 'later'. At least for an approach to syntax which takes the in-time ('online') emergence of (particularly) oral language units seriously, what is dealt with first and what is taken care of later cannot be seen as a decision between two logical equivalents (as between 'right' and 'left'). Rather, it involves one of the most basic and far-reaching decisions a speaker can make, with all kinds of cognitive, interactional and structural repercussions. 28 Appendix: Regularisation ofTranscriptions ofWord-Count(example) original transcription:

regularised transcription:

M .hh ich will UMgehend den (-) nachsten (-) ANrufer wieder einen Horer begrii.6en guten Abend? A guten Abend, B guten Abend? (0.5) A .hhh ja; ALso hh das proBLEM ah ALler steht ah auch ah so fib vor MIR jetzt irgendwo-wie wie und wo ANfangen am besten? B rnhm, A = ahrn: (--) es is eine etwas AUSsergewohnliche Proble ProbleMAtik, die vielleicht nicht nicht so ganz ah: : HAUfig in ihrer sendung erscheint, .hh obgleich des eigentlich ah::: ne total MENSCHliche angelegenheit isich lebe: seit =um gleich mal auf FAKten mal zu KOMM: ich lebe seit hh fib circa FONFzehn JAHRN mit einern MANM zusArnrnen? [.hh ] B [ja] A und fib auch ah recht GU:T eigentlich =natiirlich mit den iiblichen ALLtagsschwierigkeiten die iiberall existieren

Ich will umgehend den nachsten Anrufer, wieder einen Horer, begrii.6en: Guten Abend. Guten Abend. Guten Abend. Ja, also das Problem aller steht auch so vor mir jetzt irgendwo. Wie und wo anfangen am besten? Es ist eine etwas auBergewohnliche Problematik, die vielleicht nicht so ganz haufig in ihrer Sendung erscheint, obgleich das eigentlich eine total menschliche Angelegenheit ist. Ich lebe seit, urn gleich mal auf Fakten zu kornrnen, ich lebe seit circa fiinfzehn Jahren mit einern Mann zusarnrnen.

Ja? Und auch recht gut, eigentlich. Natiirlich mit den iiblichen Alltagsschwierigkeiten, die iiberall existie-

200 Peter Auer original transcription:

B A

regularised transcription:

auch in HEterobeziehungen, ja, u:nd fih das=(dies) ist nicht unsere probleMAtik eigentlich, wir kommen mit diesem mit unserer verANlagung sehr gut zuRECHT? also=[das]= ist alles o.k. soweit,

B

[ja:]

ren, auch in Heterobeziehungen. la. Und das ist nicht unsere Problematik, eigentlich. Wir kommen mit unserer Veranlagung sehr gut zurecht, also das ist alles o.k. soweit. Ja.

Notes

* 1.

2. 3.

4.

5. 6.

7.

The corpus research on which this paper is based was supported in many and substantial ways by Benjamin Stoltenburg. Thanks to Jessica Wallace for correcting the worst blunders in my English, and to Susanne Giinthner as well as to the editors of this volume for their comments on a previous version. Transcription of the spoken extracts follows GAT-conventions (cf. Selting et al. 1998); capital letters indicate stress positions. English translations are simplified, particularly with respect to prosody and hesitation phenomena. In case of conflict, less idiomatic versions have been chosen in order to give a better impression of German syntactic structure. The wenn-clause itself should be seen as adjoined to the resumptive particle, i.e., as a co-constituent of the front-field; cf. Eisenberg (31994: 364f). For a detailed discussion, see Metschkowa-Atanassowa 1983 and Zifonun et al. 1997: 2280-2293. A certain kind of ambiguity between a temporal and a conditional reading can also be observed in the German question word wann (usually asking for temporal information, 'at which time') which, when followed by a verb in the subjunctive mood, often takes on a conditional meaning (wann wiirdest du kommen = 'under which conditions would you come' or 'at which time would you come'). Thanks to Bernd Kortmann for drawing my attention to this parallel. For an analysis of these "concessive conditionals", cf. Konig 1985. Among these wenn-clauses in the role of obligatory constituents, we may also count comparisons using wie wenn and als wenn, as in: du kOmmsch dir vielleicht vor wie wenn dir deine wErte verLORN gangn sin. ('maybe it seems to you as ifyour values had been lost '). This usage of wenn will not be taken into account in the following discussion, nor has it been included in the quantitative analysis. A note in passing: some grammarians believe that wenn-clauses in complement function are obligatorily marked by a resumptive es (e.g. Eisenberg 31994: 365); this is not supported by my data, however. Some 400/0 of the corpus are job interviews, mainly collected among north and east German speakers, some 30% are therapeutic conversations, both in faceto-face and in radio phone-in contexts, and the remaining 30% represent pri-

Positioning ofwenn-clauses in German

8. 9.

10.

11.

12. 13. 14. 15. 16. 17.

19.

20.

201

vate everyday conversations, partly on the telephone. In the latter two types of data, southern German speakers prevail. Embeddings of wenn-clauses into complex hypotactic constructions were not counted as parenthetical. The count excludes, in addition to all polyvalent cases, all factual conditionals (Le. those expressing a positive epistemic stance), which are always nontemporal, and all concessives, as well as wenn-clauses used as complements, but includes counterfactual conditionals and reduced wenn-dann routines. *Warum [wenn Du Kirschen magst] pfliickst Du Dir keine vom Baum? 'Why [if you like cherries] don't you pick any from the tree?' therefore has to be understood as parenthetical. Resumption by dann is also excluded here, Le., non-integration is the only option. The term is used in a broader sense here than in Sweetser 1990. Details on this construction may be found in Gfinthner 1999. Note that Sweetser's "epistemic conditionals", although not "content conditionals", do not allow pre-front field placement in German (* Wenn er sich jeden Tag voillaufen lasst, sie hat ihn verlassen. 'If he gets drunk everyday, she has left him. '). Cf. Konig and van der Auwera 1988: 128 ("assertive emphasis on a consequent of a concessive allows non-integration"), Kopcke and Panther 1989: 700 ("high degree of ego involvement") and Gfinthner 1999 for details. Cf., among others, Ford and Thompson 1986: 370; Ford 1993; Dancygier and Sweetser 1996. But see Ford 1993: 56. Speaker I is most likely alluding to Zuckmayer's play (and a famous German movie) Der Hauptmann von Kopenick, in which the Prussian state and army authorities are caricatured. This of course, leaves the question open why temporal adverbial clauses should behave differently from conditional ones-a question which requires an investigation of its own. For a similar remark on English, cf. Ford and Thompson 1986: 369. In some cases, however-though not in (26) with its oblique relative pronoun-the wenn-clause can follow the relative pronoun (das Essen ist wie ein Teddybtir, der, wenn es hart ist, immer bei Ihnen ist, und der, wenn man einsam ist, zum Festhalten da ist). But here we are dealing with parenthetical placement in the middle field of the sentence; this is exceedingly rare in spoken German. Of course, wenn is not always stressed in fronted wenn-clauses. Cf. the following example: ich (.) hab (-) FUNF jahre lang an der schule franzOsisch geHABTh, mir jEhlts eigentlich an (.) PRAxis, .h aber: (-) ich bin aberZEUGT davon, =wenn ich: eh eh OFters mal die geLEgenheit hatte zum beispiel in FRANKreich, eh mich aufzuhaltn, .hh dass des: (-) eh SICherlich Ausbauftihig is. 'I had French at school for five years, actually I'm lacking practice, but I'm convinced if I on occasion had the chance to spend some time for instance in France, that I could work on it. ' The tendency to place the wenn-clause early in dependent constructions is also evidenced by the fact that parenthetical placement immediately after dass is

202 Peter Auer

21. 22.

23.

24.

25. 26.

27. 28.

frequent (cf. Note 18). Often, a second, resumptive dass is added at the beginning of the consequent: kAnnst du ihm vielleicht (-) A USrichten dass ich ANgerufn habe? und dass wenn er mit dem trelitz geSPROChen hat uber meine priifungk, dass er sich dann irgenwie=mal=GANZ kurz bei mir mElden soli? 'could you perhaps tell him that I called? and that if he has talked to Trelitz about my exams, that he should give me a quick ring some time?' The same was found in English conversations by Ford and Thompson 1986: 368. In other-speaker produced post-positioned wenn-clauses this does not always hold, since second speakers may choose this way of intimately linking their speech to a preceding syntactic pattern but nonetheless produce unexpected and even contradictory information under this 'disguise'. More exactly, the corpus included the F.A.Z. Feuilleton-Glossen from Jan 8, 1993 (Ausgabe Nr. 6) to Dec 31, 1993 (Nr. 304) and DIE ZEIT politics section of Dec 30, 1994 (No.1) to Feb 17, 1995 (Nr. 8). Transcriptions of conversational speech were regularized in order to make a comparative computerbased word-count possible; see the appendix for an example. The written corpus used by Ford and Thompson consisted of philosophical essays, a professional text for automobile mechanics, and a personal narrative account (1986: 355). The preference for pre-positioning held for all these sources. A separate count for the two newspapers shows that the quantitative results are identical. The resumptive particle is dann in 46 cases and so in 16 cases. Dann and so are not freely interchangeable; in particular, so can be used with concessives (introduced by auch/selbst wenn), while dann cannot. Cf.: Auch wenn das Abgeordnetenhaus erst im Herbst, voraussichtlich am 22. Oktober, gewtihlt wird, so hat mit der Urabstimmung schon der Wahlkampf begonnen. ('Even though parliament will not be elected until autumn, probably on October 22, SO the electorial campaign has already begun with the strike ballot. ') Of the 15 tokens in the spoken material, the three positions were about equally distributed. For a similar argument, cf. Thompson 1985.

References Auer, Peter 1991 1996

Das Ende deutscher Satze-Rechtsexpansionen im deutschen Einfachsatz. Zeitschriftfur germanistische Linguistik 1991: 139-157. The pre-front field in spoken German and its relevance as a grammaticalization position. Pragmatics (Special issue, eds. Johannes Wagner and Cecilia Ford), Vol. 6, No.3: 295-322.

Positioning ofwenn-clauses in German

203

Biber, Douglas 1986 Spoken and written textual dimensions in English: Resolving the contradictory findings. Language 62: 384-416. Couper-Kuhlen, Elizabeth 1999 Varieties of conditionals and their emergence in discourse. In: Aditi Lahiri, Alexander Patschosvky and Christoph Schwarze (eds.), Issues in interdisciplinary research on the lexicon = Working Paper No. 99, University of Konstanz, Fachgruppe Sprachwissenschaft, 89130. Dancygier, Barbara and Eve Sweetser 1996 Conditionals, distancing, and alternative spaces. In: Ann E. Goldberg (ed.), Conceptual structure - Discourse and language, 83-98. Stanford: CSLI Publications Eisenberg, Peter 1994 GrundrifJ der deutschen Grammatik. Stuttgart: Metzler. Fauconnier, Gilles 1985 Mental spaces. Cambridge: Cambridge University Press. Fillmore, Charles J. 1990 Episternic stance and grammatical form in English conditional sentences. In: Chicago Linguistic Society 26, University of Chicago, 137-162. Ford, Cecilia E. 1993 Grammar in interaction. Adverbial clauses in American English conversations. Cambridge: Cambridge University Press. 1997 Speaking conditionally: Some contexts for if-clauses in conversation. In: Angeliki Athanasiadou (ed.), On conditionals again, 387-413. Amsterdam: Benjamins, Ford, Cecilia E. and Sandra A.Thompson 1986 Conditions in discourse: a text-based study from English. In: Elizabeth Traugott et al. (eds.), On conditionals, 353-372. Cambridge: Cambridge University Press. Greenberg, Joseph H. 1963 Some universals of grammar with particular reference to the order of meaningful elements. In: Joseph H. Greenberg (ed.), Universals of language, 73 -113. Cambridge, Mass. : MIT-Press. Gtinthner, Susanne 1999 Wenn-Satze im Vor-Vorfeld: ihre Formen und Funktionen in der gesprochenen Sprache. Deutsche Sprache 27(3): 209-235. Haiman, John 1978 Conditionals are topics. Language 54: 564-89. Konig, Ekkehard 1985 Where do concessives come from? On the development of concessive connectives. In: Jacek Fisiak (ed.), Historical semantics, 263282. Berlin: de Gruyter.

204 Peter Auer Konig, Ekkehard and Johan van der Auwera 1988 Clause integration in German and Dutch conditionals, concessive conditionals, and concessives. In: John Haiman and Sandra A. Thompson (eds.), Clause combining in grammar and discourse, 101133. Amsterdam: Benjamins. Kopcke, Klaus-Michael and Klaus-Uwe Panther 1989 On correlations between word order and pragmatic function of conditional sentences in German. Journal ofPragmatics 13(5): 685-711. Lerner, Gene 1991 On the syntax of sentences-in-progress. Language in Society 20(3): 441-458. Leska, Christel 1965 Vergleichende Untersuchungen zur Syntax gesprochener und geschriebener deutscher Gegenwartssprache. Beitrage zur Geschichte der Deutschen Sprache und Literatur 87: 427-464. Metschkowa-Atanassowa, Sdrawa 1983 Temporale and konditionale "wenn "-Satze. Dusseldorf: Schwann. Selting, Margret, Peter Auer, Birgit Barden, Jorg Bergmann, Elizabeth CouperKuhlen, Susanne Gunthner, Christoph Meier, Uta Quasthoff, Peter Schlobinski, Susanne Uhmann 1998 Gesprachsanalytisches Transkriptionssystem. Linguistische Berichte 34(173): 91-122. Sweetser, Eve E. 1990 From etymology to pragmatics. Metaphorical and cultural aspects of semantic structure. Cambridge: Cambridge University Press. Thompson, Sandra 1985 Grammar and written discourse: initial vs. final purpose clauses in English. Text 5: 55-84. Zifonun, Gisela, Ludger Hoffmann and Bruno Strecker 1997 Grammatik der deutschen Sprache. Vol. III. Berlin: de Gruyter.

Counterfactual reasoning and desirability* Noriko McCawley Akatsuka and Susan Strauss

This paper calls into question the popular view of counterfactual thinking under the influence of formal logic, which considers the utterance of a counterfactual sentence as an instance of complex and intricate reasoning skills. Using naturally occurring data from English, Japanese, and Korean, we will demonstrate that there exists a pattern of prototypical counterfactual reasoning which appears to be natural and spontaneous to every human being. Speakers/writers express a particular stance of desirability versus undesirability toward a particular event, based on their subjective evaluation of reality. Counterfactual conditionals are invoked as a necessary step in this line of thinking.

1. Introduction l This chapter represents a continuation of our inquiry into the most typical usage of counterfactual conditionals in everyday life across language and culture (Akatsuka 1997, 1999). We use the word "typical" here in the sense of "natural and spontaneous," and thus underscore our claim that human beings appeal to counterfactual reasoning in dealing with the many aspects of everyday life that we find ourselves facing. In the tradition of mathematics, philosophy and formal linguistics, the conditional sentence has been regarded as the epitome of Man's rational capacity, the height of Man's ability to reason logically. In the same vein, it has long been argued that underlying the counterfactual conditional is a similar, and perhaps even more logically complex type of reasoning, engendering such analytical sentences as in (1), from Fauconnier (1985: 118). (1) A:

If Napoleon had been the son ofAlexander,

he would have won the battle of Waterloo. B: But he would have died long before that. A: Well, suppose he lived a very long life, without ever ageing, or that Alexander was resurrected in Corsica in the eighteenth century.

206 Noriko McCawley Akatsuka and Susan Strauss

Obviously, this string of utterances for Fauconnier represents a typical instance of counterfactual conditionality. The utterance is clearly an invented dialogue in which both participants realize the gross counterfactuality (and absurdity) of the proposition 'if p'. And while the type of reasoning displayed here could indeed be considered "complex," the view of counterfactual conditionality that we adopt actually has little to do with this type of exercise in logic, truth values, and absurdity. Based on the discourse data that we collected in American English, Korean, and Japanese just following the 1994 Los Angeles earthquake (hereafter, the earthquake data), we will show how and to what degree a qualitative, cross-linguistic analysis can illuminate the role of the speaker's evaluative stance of desirability. More specifically, we will demonstrate the importance of the dichotomy of desirable/undesirable in understanding the inherent ·relationship between counterfactual thinking and everyday human existence. The outline of the paper is as follows. In section 2, we will introduce Fauconnier's (1985) "mental spaces" hypothesis as one current view on counterfactuals. We will show first that while Fauconnier considers his framework as "cognitive," it is actually inherently non-cognitive. We will systematically demonstrate why this is so and why the framework fails to account for usages of counterfactuals in everyday situations, by introducing and then explicating a number of his invented examples in the light of Akatsuka's desirability framework. In section 3, we will examine spontaneous instances of counterfactual conditionals in the earthquake data in the three languages (English, Korean, and Japanese) and will contrast these with Fauconnier's constructed examples. We will also present a discussion of the theoretical implications of this study. In section 4, we will demonstrate that the present study sheds critical light on the treatment of counterfactuals in Ford and Thompson (1986) and Labov (1972). Section 5 is the conclusion.

2.

Fauconnier (1985) and counterfactuality

2.1.

Fauconnier (1985) and the tradition offormallogic

A close and careful examination of Fauconnier's (1985) analysis of counterfactuality reveals that his approach is actually inseparable from the tradition of formal logic. That is, the major concern of previous scholars in philosophy and logic has centered on the logical problem of truth conditions. Within that discipline, the primary aim was to specify under

Counterfactual reasoning and desirability

207

what strict conditions a particular counterfactual statement can be viewed as a case of 'true' or 'valid' reasoning (cf Jackson 1991; Fauconnier 1985; McCawley 1981). Example (2) excerpted from Fauconnier (1985: 109) encapsulates his approach to the phenomenon of counterfactuality. As is clearly stated, rather than grappling with issues of logic and truth conditions, the focus of analysis shifts to the process by which we establish "counterfactual spaces"-an issue which Fauconnier labels a "cognitive semantic question."

(2) Linguistically, we do not directly tackle the logical problem of truth conditions for counterfactuals, but rather the cognitive semantic question of how counterfactual spaces are set up and structured...Counterfactuality is a case of forced incompatibility between spaces (emphasis added); a space Ml is incompatible with another space M2 if some relation explicitly specified in Ml is not satisfied for the corresponding elements in M2 (1985:109).

Thus, he solves the truth conditional problem of logic by simply proposing a hypothetical world. What is crucial here is that he characterizes counterfactuality as a case of "forced incompatibility between spaces." He does not, however, specify any other conditions relating to the spaces M1 and M2, so that it is irrelevant which of the two 2 is the real world, R, and which is a hypothetical world, H. Further, in spite of his labeling the problem as a "cognitive semantic" one, we actually find no evidence of a true theoretical attempt to illuminate the notion of counterfactuality from a cognitive semantic point of view. We discuss this in detail in the next section.

2.2.

"Space-builders"

In order to understand Fauconnier's view of counterfactuality, it is perhaps best to first examine his theoretical construct "space-builders." According to this framework, "space-builders" include various linguistic elements such as conditionals (if ), verbs like wish, and negatives such as not and prevent. The particular semantic properties of these elements, according to Fauconnier, are supposed to be capable of setting up a counterfactual mental space which functions in such a way that 'some relation,' which does not hold in the parent space, may be satisfied in the newly created counterfactual space. In this light, consider the utterances listed in (3a)-(3d), which appear in the original as examples (3), (4), (5) and (6), respectively (1985: 109-110).

208 Noriko McCawley Akatsuka and Susan Strauss

(3) a. If Lucky had won, I would be rich. I would have moved to Tahiti. b. I wish Lucky had won. I would be rich. c. Fortunately, the fire did not cross the highway. My house would have been destroyed. d. Luckily, the fire was prevented from crossing the highway. My house would have been destroyed. According to this framework, it is simply the linguistic elements not in italics which are assumed to create the so-called counterfactual spacethat is, the conditional element if in (3 a), the verb wish in (3b), the negation not in (3c), and the verb prevent in (3d). However, there is one crucial observation missing here: what these examples have in common is the fact that they are all subjective statements. Utterances such as (3a) through (3d) can by no means be construed as objective descriptions, whether invented by the researcher or culled from spontaneous discourse data. In other words, the speaker of (3a) and (3b) sounds unhappy or disappointed because what s/he was hoping to happen did not actually happen. Conversely, the speaker of (3c) and (3d) sounds happy and relieved because his/her home was not destroyed by a particular fire which did not cross a particular highway. In this light, Fauconnier's framework of mental spaces takes no account whatsoever of an individual speaker's mental attitude, and thus pays no special attention to the function of the adverbs fortunately and luckily, which figure overtly in his own constructed examples (3 c) and (3d). The function of these adverbials also accounts for analytic gaps in examples (3a) and (3b). We will explicate this function in the next section from the perspective of the concept of desirability.

2.3.

Desirability and conditionals

We will begin this discussion by examining Fauconnier's analysis of (3c) and (3d). According to Fauconnier, the lexical items not and prevent represent the strongest type of lexical space-builder by virtue of the fact that they deal in the domain of negation, and hence counterfactuality, since semantically they are capable of reversing and/or negating some expression of reality (i.e., factuality). Accordingly, lexical items such as these are much stronger than the if and wish types, since the latter do not always build counterfactual space. Fauconnier states:

Counterfactual reasoning and desirability

209

(4) Clearly, counterfactuality may be lexically imposed, as it is by strong negatives such as not and prevent in (5) [i.e., our (3c)] and (6) [i.e., our (3d)] (1985: Ill). What is interesting here is the focused concentration on negation markers as lexical agents for the construction of this so-called counterfactual space. However, close examination of these sample sentences reveals the force of other semantic factors beyond these negators which playa significant role, if not a more significant role, in the expression of counterfactuality. To illustrate, let us begin by omitting the attitudinal adverbs fortunately and luckily from Fauconnier's examples in (3c) and (3d). The sentences so altered appear as (5a) and (5b) below:

(5) a. ??Thefire did not cross the highway. My house would have been destroyed. b. ?? The fire was prevented from crossing the highway. My house would have been destroyed. While the change consists simply of a seemingly minor lexical deletion, the occurrence of the counterfactual in the second sentence is rendered unnatural in both examples. The negation markers nonetheless are still intact-a fact which now casts serious doubt on the claim that counterfactuality is 'lexically imposed' by such elements as negatives. We propose instead that the phenomenon of counterfactuality is inherently related to the critical role of the speaker's stance of desirability. And while we do not deny the importance of negation in this phenomenon, we do not recognize negation as the single factor involved. That is, negation comes into play in conjunction with, and only in conjunction with, an expression of the speaker's stance of desirable vs. undesirable. This becomes much clearer when we compare Fauconnier's examples with their Japanese counterparts. The English expression, 'Luckily/Fortunately + p (proposition) happened' corresponds most closely to the Japanese sentence pattern, p-te yokatta. What this grammar pattern expresses is precisely the speaker's attitude, desirable, towards the past event/state of affairs described by the proposition, p. Observe (6):

210 Noriko McCawley Akatsuka and Susan Strauss

(6)

p-te, [desirable}

yokatta.

was lucky '(I, it, etc.) was lucky that p.'

Now, let's compare Fauconnier's original example from (3c), repeated below as (7), with the Japanese translation of the same utterance in (8).

(7)

Fortunately, the fire didn't cross the highway. My house would have been destroyed.

(8)

Kaji ga

dooro

no mukoo ni moeutsura-nakute, yokatta. fortunate Fire SM. highway over burn-cross-not (Moshi) moeutsutte-itara, uchi ga yakete itadaroo. IRREALIS burn-cross-if my house SM had burnt down-MDL 'I was lucky that the fire did not cross the highway. house would have been destroyed. '

If it had,

my

The essential difference between the two examples is that the conditional antecedent appears in the Japanese example, (8). That is, unlike English, Japanese syntax does not normally tolerate monoclausal conditionals. And once again, we see the overt representation of the speaker's attitude, desirable, towards the past event/state of affairs described by the proposition, p. In this case, p is synonymous with 'the fire did not cross the highway' (=fact). To illustrate in English, we repeat examples (Sa) and (5b) below. (5)

a. ??The fire did not cross the highway. My house would have

been destroyed b. ??Thefire was preventedfrom crossing the highway. My house would have been destroyed. Note that when we supply an overt conditional antecedent in both examples, the awkwardness disappears. Observe (Sa') and (5b'): (5)

a' The fire did not cross the highway. If it had, my house would

have been destroyed. (5)

b' The fire was prevented from crossing the highway. If it hadn't been, my house would have been destroyed.

Counterfactual reasoning and desirability

211

Now, (Sa') and (Sb') both contain the explicit conditional antecedent 'if not P,' and it is here that we can begin to understand what the speaker is actually doing by invoking counterfactuality: she is presenting the reason why she considers it desirable that p, i.e., 'the fire did not cross the highway' (=fact), happened. The counterfactual situation, i.e., 'not p' or 'the fire's crossing the highway,' would have led to the loss of her house. We informally represent this line of reasoning, in (9): (9)

It was desirable that p (=fact) happened. If 'not p' (=counter to fact) had happened, it would have led to undesirable consequences (='not q' ).

Contrary to Fauconnier's claim, then, it is not the lexical items not and prevent that build the so-called counterfactual space. Rather, we maintain that it is the speaker who invokes counterfactual thinking as an integral step in the line of reasoning, as delineated in (9). Thus, the use of negation figures strongly in the expression of p vs. 'if not P,' as well as q vs. 'not q,' or in other words, in the speaker's expression of stance of desirable vs. undesirable, and not in the construction itself of the socalled 'counterfactual space. ' Let us now tum to Fauconnier's analysis of (3a) and (3b), repeated below as (lOa) and (lOb) for convenience. (lOa) IfLucky had won, I would be rich. I would have moved to Tahiti. (lOb) I wish Lucky had won. I would be rich. According to Fauconnier's framework, it is the "space-builders" if and wish that are responsible for the appearance of the counterfactuality in these examples. Now, compare the same examples, with the first sentence of each altered to express the fact p that Lucky (a racehorse) did not win a particular competition, and the second sentence altered to include an overt conditional antecedent expressing 'if not p,' as noted in (lOa)' and (lOb)': (lOa)' Unfortunately, Lucky did not win. If he had won, I would be rich. IfI were rich, I would have moved to Tahiti. (lOb)' Unfortunately, Lucky did not win. I wish he had won. won, I would be rich.

If he had

212 Noriko McCawley Akatsuka and Susan Strauss

We claim that an adequate theory of counterfactuality must explain on a principled basis that the workings of counterfactual reasoning in Fauconnier's examples (lOa) and (lOb) are exactly the same as those in (lOa)' and (lOb)'. According to our analysis, the line of reasoning involved here will be illustrated informally as in (11). (11) It was undesirable that p (=fact) happened. If 'not p' (=counter-tofact) had happened, it would have led to desirable consequences ('not q"). It will be obvious that (9) and (11) represent the identical pattern of reasoning, with the sole difference being the speaker's desirability assessment taking opposite values. The line of reasoning delineated in (9) and (11) will be referred to henceforth in this study as "prototypical counterfactual reasoning." We have demonstrated that Fauconnier's framework is an i1mpersonal one, not unlike those proposed by scholars from the tradition of formal logic and mathematics; the view is necessarily limited to an incomplete observation of the distribution of counterfactuals. Crucially, the framework fails to account for all four examples (i.e., (3a)-(3d» as squarely representative of prototypical counterfactual reasoning. Unwittingly, too, even example (1), based on an invented dialogue positing absurd conditions ('if Napoleon had been Alexander's son'), expresses a strong stance of desirability ('he would have won the Battle of Waterloo'). And what Frenchman would find this anything but desirable? We now tum our discussion away from constructed examples and examine authentic discourse excerpted from the earthquake data in English, Japanese, and Korean, where "prototypical" counterfactuals emerge spontaneously.

3. A qualitative analysis of the earthquake data3 On January 17, 1994, at approximately 4:30 a.m., a strong earthquake hit Los Angeles, significantly startling its residents. January 17 happened to be a Monday, and that Monday happened also to be a national holiday (Martin Luther King's birthday). Because of this, many people had planned not to go to work that morning. Commuter traffic was exceptionally light, and many people were still asleep when the

Counterfactual reasoning and desirability

213

earthquake hit. Because of these and other factors, the casualties were 4 rather light, given the magnitude of the quake . Approximately three weeks following the incident, researchers from UCLA collected 'earthquake stories' from dyads of native speakers of English, Japanese, and Korean. Subjects were predominantly UCLA graduate students, though some undergraduates also participated. In all cases, we attempted to ensure that the dyad participants did not know each other previously, though here too some of the Japanese data reveals that a handful of subjects did know their interlocutor to some extent. For the most part, however, subject pairs were intended to be strangers. The only instructions given to the subjects were to tell each other about the experiences they had had during the earthquake, and that the data recording session would last for approximately 20-25 minutes. Other than these criteria, no explicit instructions were given. The interactions were both audio and video taped and were transcribed using the conventions of Conversation Analysis (Atkinson and Heritage 1984). What is remarkable is that counterfactual conditionals emerged in all three languages, and in a number of cases we find significant overlap in terms of the types of issues captured with counterfactuals (e. g., 'if it had happened during rush hour,' 'if it had happened in a place other than LA,' 'if it had happened before the recent earthquake retrofitting of some of the campus buildings,' etc.). None of the conversational dyads was aware of what the other dyads had said in other recording sessions. Further, because the study was designed solely to elicit spontaneous talk in the respective languages, no mention of counterfactuals was ever made at the inception. It was not until we transcribed and analyzed the data that we discovered this parallelism. In the discussion that follows, we limit ourselves to one representative sample from each language. In all three examples, the subjects systematically express their reactions to the relatively light damage and small number of injuries and deaths, in the light of the strength of the earthquake itself Interestingly, the abstract line of reasoning of each case is quite similar to what we have noticed in Fauconnier's examples (3c) and (3d), even though Fauconnier's examples are monologic and constructed, while our data derive from spontaneous, natural dialogue. Instead of the initial occurrence of such attitudinal adverbs as fortunately and luckily as we witnessed in the Fauconnier examples, we observe here the speakers exchanging and building on the expression "we were lucky", in all three languages. We will call this activity 'the opening.' Just following this sequence emerges the counterfactual utterance, 'if not p' had happened,

214 Noriko McCawley Akatsuka and Susan Strauss

'not q,' would have happened in all three languages. Henceforth, we will adopt the formula 'if not p, then not q' to represent counterfactual conditionality. Let us now examine each excerpt, starting with English.

3.1.

English: "Actually it's good that this happened now"

The speakers, T and H, were just talking about the fact that friends living in the dorms said that the building seemed to sway at the time of the earthquake; T then adds that some of the UCLA dormitories had recently undergone seismic retrofitting: "actually, it's good that this happened now because..." In other words, "we were lucky," hence the opening sequence. Observe, too, that after this opening T immediately appeals to counterfactuals. What is of particular importance here is T's initial clause "I jus' kept thinking," which precedes the counterfactual construction, underscoring just how naturally and just how spontaneously we appeal to counterfactual reasoning.

(12) English: «participants: T=Female, H=Male)) (a)

Opening

T1

Actually it's good that this happened now because (then) these dorms were just respir- retrofitted () fer- () y'know () earthquakes () two to three years ago.

(b)

'if not p, not q'

==> T2: ==> ==> ==>

==>

I jus' kept thinking if this earthquake had been in the middle ofthe afternoon on a day that wasn't a holiday how ba: :d this would have been, like the San Francisco quake. 1- That was at five 0' clock in the ==> afternoon on a Friday "hh or something like that It ==> was just. y'know I think it-I think there would have ==> been so many more lives lost and this would have been a rea: : I tragedy.

Counterfactual reasoning and desirability

215

In the stretch of talk in section (b) 'if not p, not q' we witness the gradual emergence of a complex set of thoughts, all of which express some counter-to-fact notion with respect to some aspect of the reality of this earthquake. That is, first T proposes a time and a day different from the actual time and actual day of the incident p and then indicates that the consequences q would have been much worse (or undesirable), a line of reasoning we encountered earlier in example (9), repeated below as (13):

(13) It was desirable that p (=fact) happened. If 'not p' (=counter-tofact) had happened, it would have led to undesirable consequences (='not q'). Crucially, the speaker attitude underlying this utterance surfaces in the linguistic manifestation of how T expresses the degree of potential undesirability of the consequences ('not q'), had the situation played out under the, counterfactual conditions she just established (i.e., 'different time, different day'): her utterance is structured with an exclamatory degree of intensity marker how, coupled with an ostensible sound stretch in the negative adjective ba:d (I jus' kept thinking how ba::d this would've been). But she doesri't stop there. We actually witness her imagination emerge linguistically, as she now upgrades the potential counterfactual consequences by not only changing the time of the occurrence, but now the place, i.e., San Francisco, where in 1989 a comparable earthquake occurred during the height of rush hour traffic. And once again, the pattern of reasoning is identical to that in (13). The significance of this upgrade is twofold: first, it establishes a recognizable frame of reference for this new level of negative consequences by recalling an actual incident where those very same conditions held (i.e., different time of day, different day of the week, and now a different place); it is a well-known fact that many more people suffered and died in the San Francisco quake. Second, it underscores all the more the crucial relationship between desirability stance and counterfactual thinking by again linguistically matching the upgraded scenario with an equally upgraded utterance: there would have been so many more lives lost and this would have been a rea:: I tragedy, clearly more semantically intense than the initial 'how bad this would've been, and notice too the audible sound stretch on the intensifier in a rea:: I tragedy.

216 Noriko McCawley Akatsuka and Susan Strauss

3.2.

Co-construction in Korean and Japanese

We have just observed an example from the English database where a single speaker imagines a scenario different in specific respects from the actual one, in order to emphasize her current stance of desirability toward the actual incident. In the Korean and Japanese examples that follow, we note a similarity in this type of complex thinking; only here, the imagination belongs to both parties as they collaboratively participate in picking out specific aspects of the actual incident which could have been different, and in jointly constructing the potential undesirable circumstances.

3.2.1. Korean co-construction First we present the Korean extracts to illustrate the co-construction of counterfactual reasoning. Example (14) below is framed much in the same way as its English counterpart in (12). That is, it begins with an assessment in (a) concerning the fortunate fact that the earthquake happened very early in the morning, followed by the utterance of a counterfactual scenario in (b) where the participants underscore just how lucky they were. There, we will observe that the counterfactual conditional 'if not p, not q' is collaboratively constructed by the two speakers. H opens this sequence with the Korean expression 'be lucky' tahayng ita, and also notably ends the sentence with the same expression. He continues his turn with mannyak-ey, 'if foreshadowing that some conditional utterance will most likely follow. And T provides an agreement token, kurehcho 'that's right. ' (14) Korean: «participants: H=Male, T=Male)) (a)

Opening

==> HI: tahayng-i-n kes-un cihca saypyek-ey cicin lucky-COP-ATTR thing-TM dawn at earthquake tn-ssunikka tahayng-i-cyo occur-CONN lucky-COP-TAG 'The lucky thing is that the earthquake happened at dawn, that was lucky, huh. '

Counterfactual reasoning and desirability

H2:

==> TI:

217

mannyak-ey.. if (lit. in a one in ten thousand chance) 'if kureh-cho, ku.. be like-TAG FLLR 'That's true.'

Directly following this assessment sequence, T then builds upon H's foreshadowed conditional, by introducing a counterfactual element concerning the time that the earthquake occurred: 'traffic' manha-ssu-I ttaykena, 'if it had been at a time when there was a lot of traffic,' in other words 'if not p.' H produces the consequent clause 'not q' koyngcanghi manhun salam-tul-i ama, 'my, a lot of people would probably have died,' which itself is an emotionally charged statement, underscored all the more by the exclamatory marker ama '(oh) my. ' (b) ==> T2:

'if not p, not q'

ce 'traffic' manha-ssu-I -ttaykena xxx ... freeway' FLLR traffic be a lot-PST-ATTR time freeway 'ifit had been at a time when there was a lot of traffic or...the freeway."

==> H3:

koyngcanghi manhun salam-tul-i ama rather many-ATTR people-PL EXCL cwuke-ss- keyss-cho die-PST-MDL-MDL 'my, a lot of people would probably have died.'

The Korean example represents a very neat case where one speaker produces the conditional antecedent, 'if not p' clause and the second speaker the consequent, 'not q.' And the interactional dynamic which engenders such a collaborative construction is an interesting one. Even more revealing, however, is the Japanese example, where we find the most elaborate instance of collaboratively produced counterfactual reasoning of the three excerpts under investigation.

218 Noriko McCawley Akatsuka and Susan Strauss

3.2.2. Japanese co-construction In the stretch of talk that follows, we observe an opening interactional pattern much like that in example (14), where one participant, in this case, H, expresses the assessment that it was "lucky" the earthquake happened so early in the morning. This is responded to in the very next line by A with an expression indicating a stance of agreement, "I thought about that, too." What is interesting here is that in both the Korean and the Japanese examples, the participant who utters the assessment does so in such a way that co-participation is invited from the outset through the use of interactional particles. In the case of the Korean utterance, the speaker ends the assessment with -ci, functioning at times like a TAG-like particle. Similarly, in the Japanese example below, we find the use of nee, which among other functions, also serves as a TAG-like marker. And, in each case, their interlocutors do provide both an appropriate and relevant response, i. e., utterances providing agreement.

(15) Japanese: «participants: H=Male, A=Female)) (a)

Opening

==> HI : demo nee asa de yokatta desu yo nee but PRT morning in lucky-PST COP PRT PRT gozen yo)i han de? a.m. four 0' clock halfpast at 'But, see, we were lucky it happened in the morning. at four thirty, right?' ==> AI:

atashi mo omotta-n desu yo. 1 too think-PST-PLN-NOML COP PRT 'I thought about that, too.'

In the lengthy collaboratively built sequence that follows, H expands his own initial utterance and begins to weave a "complex" mesh of counterfactual thought with his interlocutor, first by altering the time of the earthquake, from the early morning hours (4:00 a.m.) to the current time, i.e., the approximate time of recording (4:00 p.m.), and then by altering the place in conjunction with the changed time ('if it had

Counterfactual reasoning and desirability

219

happened in the library at a time when students were in it'). And finally, the relevance of an altered time emerges in the talk, referring to the freeway overpass which collapsed as a result of the quake, beginning at line A3, and resumed again from H7 through the end of the excerpt. (b)

'ifnotp,notq'

==> H2:

moshi nee kore ga nee imagoro= ifPRT this 8M PRT now around 'If (it had happened) around now [i.e., about 4 p.m.],'

==> A2:

=kore ga nee, yuugata no: this 8M PRT evening PRT '(if it had happened) in the evening,'

H3:

==> A3:

H4:

ima toka ne now or something PRT 'like now or something. ' rasshuji no go-ji toka ne ha ha ha ha rush hour GEN five o'clock or something PRT 'at rush hour like five 0' clock ha ha ha ha ha' soo SOD SOD yuugata no right right right evening GEN rasshuji no goji toka rush hour GEN five 0' clock or something 'Right, right, right at rush hour, at like five o'clock in the evening. '

A4:

==> H5:

ha haha haha ha 'ha haha haha ha' anD toshokan nanka de mada gakusei nanka FLLR library or something LOC still students something ir A9:

Closing

nn atashi mo omotta-n su asa de yokatta-n daroo naa tte I too think-PST-PRT moming-INST good-PST-PRT-MDL-PRT

Counterfactual reasoning and desirability

221

'So I thought too that we were lucky it happened in the morning.'

HIO: nnn 'yeah' AIO: minna neteru toki d~tta kara everyone sleep-te PROG time COP-PST-PLN because 'Everybody was sleeping at that time, so' HII: honto hukoo chuu no saiwai desu yo. really disaster in GEN blessing COP PRT 'Right, it was the only blessing within this disaster. ' All: nn= 'yeah.'

HI2: =nn nn 'yeah.' In the first line of this sequence, H2, foreshadows his conditional thought with the word moshi 'if,' and then sets the counterfactual hour to be 'around now' imagoro. In response, A builds on H's utterance by repeating it almost verbatim, with the exception of one change: rather than recycling imagoro, a temporal deictic term whose meaning can only be recovered from the situated context, she substitutes the time reference with yuugata, thus slightly upgrading the level of specificity of the counterfactual time in question. And she substitutes and upgrades the time reference yet once more in A3, by specifying even more detail in this counterfactual switch: rashuji no go-ji toka '(if it were) at rush hour, like 5:00 p.m. '-a piece of information which will become supremely relevant later in the talk. From H2 through A4, the speakers have created a counterfactual scenario in which the earthquake could have occurred at a time other than at the actual time that it did. At H5, they carry this one step further by now zooming in on a specific location which could have suffered serious effects, had that quake indeed happened at the altered hour. What is crucial here, as well, is the fact that this specific location is one which both purportedly know very well-the library, ('well [suppose this happened] like at the library, if it had happened at a time when students were still there'). Note too, that H is the one who proffers this new

222 Noriko McCawley Akatsuka and Susan Strauss

location (H5), and in mid-utterance, as he struggles through the verb iru ('to be-at a location'), A actually begins her tum, by supplying the verb iru as well as the temporal specification of 3 or 4 0' clock. And their collaborative imagining continues, as they jointly create a scene, complete with vivid visuals, supplied by H (moo atama kara saa hon ga ochitekite 'books would have fallen on their heads'), and onomatopoetic expressions (aa zaa 'crash!'), supplied by A. Observe that this very description represents the first instance of a consequent clause in this entire sequence, and that this consequent clause, just like its antecedent, is collaboratively produced by both participants, ending with the assessment of sugokatta desu ne 'it would have been terrible,' at A6. At H7, H continues to build on the counterfactual scenario to include all freeways and highways, thereby creating a much more intense and dangerous situation than the actual one, zenbunee sundanshichattaramoo 'if all freeways and highways had been cut (like the one that actually was cut'). A in tum concurs with the first half of the undesirable consequent kanbotsushichatte: 'they would have collapsed, and', which H completes by providing the actual upshot of this entire conversational sequence and line of reasoning, shisha nanka juubai gurai n natta deshoo nee, 'the death toll would have been ten times higher, right?' Crucially, as soon as the above conclusion is jointly reached, the speakers go back to where they left the conversation in the opening. That is, they again congratulate themselves by saying asa de yokatta 'we were lucky it was in the morning. '

3.3.

Summary

To sum up, in the excerpts from all three languages, we find the consistent parallel whereby speakers invoke counterfactual thinking to express the pattern of reasoning originally introduced as (9) and (13), and repeated below as (16): (16) It was desirable that p (=fact) happened. If 'not p' (=counter-tofact) had happened, it would have led to undesirable consequences (='not q'). Every instance of counterfactual conditional is invoked to highlight the speaker's appreciative attitude towards the actual incident. That is, in each language, we find that speakers have created a domain of

Counterfactual reasoning and desirability

223

imagination in which they alter one or more specific aspects of a given reality (either on their own, as in the English example, or in collaboration with their interlocutor, as in the Korean and Japanese examples). In all three cases, we have witnessed how speakers shifted coordinates of time, coordinates of place, or both, and then viewed the situations from a counterfactual perspective which enabled them all to appreciate just how lucky they were that the damages caused by such a large earthquake were so mmor.

3.4.

Discussion

Why are there such significant overlaps in the contents of counterfactual conditionals uttered by the American, Korean, and Japanese speakers when they recall the occurrence of the earthquake from the point of view of the common experience of the community? In the discussion that follows, we suggest that this seemingly simple question is the key which will shed light on the cognitively dynamic aspects of counterfactual reasomng.

3.4.1. The predictability of 'not p' In response to the question above, we propose that the counterfactual utterances by the American, Korean, and Japanese speakers overlap in the way they do for the reason that the contents of 'not p' are highly predictable. The speakers in all three languages, in producing counterfactual utterances, reflect their shared common sense about the earthquake. In other words, irrespective of language and culture, residents of metropolitan cities like Los Angeles share the view that if an earthquake were to occur early in the morning while most residents are still asleep, casualties would be relatively small, even if freeways and large public buildings were to suffer heavy structural damage, precisely because few people would have been present in those areas. Thinking this way, we realize that the processes of imagination we have observed in the cross-linguistic data are not really "complex" mental activities. This is the reason why the two Japanese speakers can enjoy the lengthy collaborative construction of counterfactual reasoning, as illustrated by the laughter tokens ha ha ha ha ha (lines A3 and A4) and their enthusiastic endorsement of the other party's comments soo SOD SOD 'right right right' (H4). Note that they are not co-reasoning to create new

224 Noriko McCawley Akatsuka and Susan Strauss

ideas. There is absolutely no exchange of new ideas going on here. Instead, in these extended stretches of talk the speakers are jointly repeating what they have been repeating to themselves and to others and what they have heard repeated on so many occasions since the earthquake first hit five weeks earlier. And by repeating what they know and what basically everybody else already knows, they are congratulating themselves by conveying the message, "We are really lucky. It could've been so much worse."

3.4.2. The speaker's belief: "if 'not p' actually happens, then terrible

consequences will result" In sharp contrast to Fauconnier's 'absurd' conditions in example (1) ('if Napoleon had been the son of Alexander'), all instances in the earthquake data of 'not p' of "if 'not p' had happened" actually represent a variety of situations which could have indeed happened and could have resulted in much greater casualties. Why is this the case? We suggest that when the speaker utters "if 'not p' had happened," this utterance reflects the speaker's belief that if 'not p' does indeed happen, then terrible consequences will result. In other words, each instance of counterfactual reasoning in the earthquake data is endorsed by the same speaker's belief that "if 'not p' actually happens, then 'not q' (=terrible consequences) will result." What we observe here is the co-existence of conditional judgment and counterfactual judgment in the consciousness of the speaker-which explains why the earthquake data contains so many instances of expressions of great relief 3.4.3. The 'causal link' between p and q5

In order to explain the co-existence of conditional reasoning and counterfactual reasoning in the consciousness of the speaker that is observable in the earthquake data, we will need a dynamic theory of conditionals which postulates that the connection between p and q is an integral part of the grammatical meaning of natural language conditionals. Logicians have long noted that normally there is some kind of connection between p and q. It was partly due to the difficulty in pinpointing the exact nature of this connection in the framework of formal logic that they generally concluded, regardless of their personal stand on the analysis of ~f, that this connection should be treated as a problem of pragmatics

Counterfactual reasoning and desirability

225

rather than grammar (e.g., Quine 1950, Grice 1975, Stalnaker 1975). The position of these logicians has been more or less inherited by many linguists (e.g., Geis and Zwicky 1972, Haiman 1978, Karttunen and peters 1979, McCawley 1981). It is important to note that Fauconnier (1985: 113-115) explicitly rejects the idea that there is an inherent connection between p and q of both indicative conditionals and counterfactual conditionals. Fauconnier even takes the position that the counterfactuality of q in counterfactual conditionals is implicature and therefore cancelable. Basically, then, Fauconnier (1985) treats natural language conditionals very much like mathematical conditionals. Note that in the case of mathematical conditionals, there is no connection between p and q of 'p ::> q'; they are independent propositions. Further, there is no sequential relationship between the two, either. However, McCawley (1981) observed that in all English conditionals, p is temporally and/or causally or epistemologically prior to q. We now come to the conclusion that it is far from being an accident that Fauconnier's mental space theory has failed to account for the prototypical counterfactual reasoning phenomenon in our discussion in Section 2.

4.

More on aspects ofprototypical counterfactual reasoning

4.1.

Ford and Thompson (1986): 'contrast'

It will be illuminating here to examine Ford and Thompson's (1986) analysis of conditionals in the light of the present discussion. Ford and Thompson (1986) is a pioneering study of discourse functions of the ifclause as a link between the preceding context and the following discourse. Based on their quantitative analyses, they have identified four basic discourse functions shared by their written and spoken data, one of which is to present a contrast with the preceding context. This view is shown by the formula in (17) below. What is relevant to our study is that (18) is provided as a clear case illustrating that such a contrast can be indicated by means of a counterfactual. (18) is an excerpt from a personal narrative about the behavior of a chimpanzee by the name of Nim Chimpsky, trained to use American Sign Language. (17) and (18) appear in the original as examples (6) and (8), respectively (1986: 359). (17) X. (But) if not X, then Y.

226 Noriko McCawley Akatsuka and Susan Strauss

(18) Nim's aggression increased mainly because of the necessity of introducing more and more teachers into his life... 1f it had been possible for him to have grown up with a small and stable group of caretakers, he would have experienced fewer separations from his trusted caretakers and had far fewer opportunities to test his dominance through aggression. (Terrace 1979: 145).

According to our analysis, what underlies (18) is the same line of reasoning as delineated in (19) below. This reasoning is identical to that expressed earlier in (16), with the exception that the speaker's desirability stance has taken the opposite value. (19) was previously introduced as (11), and together with (16) these expressions represent instances of 'prototypical counterfactual reasoning.' (19) It was undesirable that p (=fact) happened. If 'not p' (=counter-tofact) had happened, it would have led to desirable consequences ('not q"). The writer of (18) has a warm and sympathetic understanding of the poor chimp's aggressive behaviors resulting from his unfortunate upbringing. This has been made unmistakably clear by his appeal to counterfactual reasoning. We agree with Ford and Thompson that 'contrast' is indeed a useful notion in understanding the discourse function of the counterfactual conditional; however, we propose that what is contrasted in (18), is not 'if not X' and its preceding context X, but rather the undesirable reality 'caused by X' and the desirable situation Y which could have taken place.

4.1.1. Clancy, Akatsuka, Strauss (1997) Strong empirical support of our re-analysis of Ford and Thompson's formula comes from our recent cross-linguistic research project (Clancy, Akatsuka, and Strauss 1997). Based on a total of 84 hours of spontaneous discourse data, we examined how American, Japanese, and Korean parents use conditionals when talking to children less than three years of age. Since the subjects were so young, these interactions were particularly rich in such adult-produced speech acts as commands, prohibitions, and granting permission. We discovered that overwhelmingly the most frequent type of conditional uttered by the caregivers was what we called "predictive" conditionals, which predict

Counterfactual reasoning and desirability

227

that if one does an undesirable thing, then undesirable consequences will result (American English 63%, Japanese 74%, Korean 80%). Observe the following example where an American mother is talking to her two-year-old child who wants Santa Claus to bring her two dolls for Christmas:

(20) Mother: You have to be a good girl. If you are not a good girl, Santa Claus won't bring you anything. Throughout the data, the caregivers often gave warnings, criticisms, and advice to their children, constituting a particular type of persuasive speech act. Typically, this pattern of persuasion takes the following form: An initial utterance expressing "do X" or "don't do X" is followed by the provision of a reason, couched in a conditional statement. In other words, the message takes the general shape of "if children don't listen to what adults say, undesirable consequences will result." What is being contrasted here are desirable consequences versus undesirable consequences. We can informally represent this line of reasoning involved here as (21): (21) It is desirable that p will happen. If 'not p' happens, it will lead to undesirable consequences. There is an obvious parallelism between prototypical counterfactual reasoning and the pattern of reasoning delineated in (21). From this, it will be clear to the reader that the speaker's stance of desirability plays a critical role in understanding the many usages of conditional utterances in our daily lives, not just the role it plays in prototypical counterfactual reasoning alone.

4.2.

The discourse domain: The case ofa book review

Because the prototypical counterfactual serves to underscore the speaker's stance of desirability or undesirability toward a particular consequence, the discourse domain in which this kind of reasoning manifests itself can be quite broad. In order to illustrate this point, let us look at a one-page book review which appeared in the "Dividing Line" column of a recent issue of Time magazine (October 4, 1999: 90). The book is entitled Gone Boy and was written by a father whose 15 year old son was killed by a fellow student in a college campus shooting incident.

228 Noriko McCawley Akatsuka and Susan Strauss

The reviewer, Jack E. White, is a long-time friend of the author~ the review is directly related to White's stance on gun control. The column heading reads "Elegy for a Gone Boy," in large, bolded black letters, accompanied by the subtitle, "A father's searches for meaning in his son's murder," which appears in much smaller font. It is clear that the function of the heading and its subtitle is not simply to provide an efficiently abridged abstract of the contents of the book-it is an unmistakable indicator of the reviewer's mental attitude toward an undesirable reality. The review begins as a personal narrative, which appears in (22) below: (22) GREGORY BIGSON AND I MET 36 YEARS AGO AS FRESHMEN AT Swathmore College. Greg was 15-the same age his elder son Galen had reached in 1992 when he was slaughtered in an act of senseless violence. Galen was in his second year at Simon's Rock College in Great Barrington, Mass., when a fellow student named Wayne Lo went berserk and shot up the campus with a cheap imported rifle, killing Galen and a teacher and wounding four others.

White devotes a sizable amount of space to the body of the article, which provides an account of how Galen's grieving father spent seven long years in the investigation into the cause of his son's murder. He goes on to report that the result is "a tapestry of shared pain and guilt in which everyone, including the killer is in some sense a victim." The review ends with the concluding remarks reproduced in (23) below. Notice that this utterance is a textbook case of prototypical counterfactual reasoning, "if 'not p' had happened, it would have led to desirable consequences." (23) None of this would have happened if Wayne Lo, at the age of 18, had not been able to walk into a gun store, flash his driver's license and $129 and walk out with a deadly weapon. Or if he had not been able to have 200 bullets sent to him at Simon's Rock College by a mail-order arms company. To my friend, Greg, there is a straightforward conclusion to be drawn from the mystery of Galen's death. "We've just got too many guns in this country. We've got to get rid of them". Anyone who reads Gone Boy will find it hard to disagree.

By appealing to counterfactual reasoning, the reviewer is providing the reader with detailed and specific new information about just how easily Wayne Lo was able to buy a gun, have the ammunition sent to him at school, and actually carry out this fatal shooting. The bottom line here is that there are indeed too many guns in America.

Counterfactual reasoning and desirability

229

Thus, counterfactuals do not just describe unrealized possibilities; they can and do eloquently report what actually happens in the real world.

4.3.

Labov (1972): 'comparators'

This section will serve as a preview of our next project, which involves a re-examination of Labov's (1972) seminal study of narratives in the light of our prototypical counterfactual reasoning hypothesis. Specifically, we are interested in Labov's treatment of counterfactuals in relation to his key notions of "evaluation" and "comparators". Simply stated, Labov's "evaluation" is a syntactic unit with which the narrator shows why a particular event is a reportable experience; "comparators" are grammatical devices that the narrator utilizes to evaluate the narrative. According to the original framework, neither notion is related to the speaker's evaluative stance of desirability. Consider the following excerpt (1972: 387) (emphasis added), where Labov cites one of the most dramatic danger-of-death stories to illustrate the notion of what constitutes a comparator: (24) One of the most dramatic danger-of-death stories was told by a retired postman on the Lower East Side: his brother had stabbed him in the head with a knife. He concludes: And the doctor just says, "Just about this much more", he says,

"and you'd've been dead". Comparators then include negatives, futures, modals (emphasis added), quasimodals, questions, imperatives, or-clauses, superlatives, and comparatives, more or less in increasing order of syntactic complexity.

As stated here, Labov (1972) treats counterfactual utterances as evaluating clauses because they utilize "modals". We are impressed by Labov's perceptive observations. For example, he continues the above discussion by saying that "A comparator moves away from the line of narrative events to consider unrealized possibilities and compare them with the events that did occur." Unfortunately, however, the question of "what precisely is it that the speaker is evaluating with those devices called comparators?" remains unaddressed. Further, as we have just pointed out in the previous section, counterfactuals do not just consider unrealized possibilities. Skillful speakers/writers such as the narrator of this story know that they can effectively report what actually happened by appealing to counterfactual reasoning, and in so doing, they convey the message "I was really lucky."

230 Noriko McCawley Akatsuka and Susan Strauss

5. Concluding remarks Throughout this study, we have attempted to demonstrate that there exists a pattern of "prototypical counterfactual reasoning," which appears to be natural and spontaneous for every human being. We represent this pattern informally as in (25) and (26). (25) It is desirable that p happened. If 'not p' had happened, it would have led to undesirable consequences (='not q'). (26) It is undesirable that p happened. If 'not p' had happened, it would have led to desirable consequences (='not q'). These two represent the identical pattern of logic. We have argued that prototypical counterfactual reasoning in the earthquake data calls for a cognitively dynamic theory of natural language conditionals which can account for the fact that there is a necessary, simultaneous co-existence of conditional and counterfactual conditional reasoning in the consciousness of the speaker. That is, each occurrence of "if 'not p' had happened" in the cross-linguistic data entails the same speaker's belief that if 'not p' actually happens, then 'not q' (=terrible consequences) will result. This analysis explains why the earthquake data is permeated by expressions of such a great sense of relief In this line of thinking, we realize that the seemingly "complex" imagining processes that we observed in the cross-linguistic data are not really "complex" mental activities because 'not p' is highly predictable; it is a reflection of the shared common sense of the residents of Los Angeles about the earthquake and its damages, irrespective of language and culture. We have further argued that prototypical counterfactual reasoning calls for a theory of natural language which postulates that the antecedent and consequent of both indicative and counterfactual conditionals are not only sequentially ordered but they are 'causally' related to each other. What emerges from the study of prototypical counterfactual reasoning, then, is the understanding of the fundamental differences between the grammar of the natural language conditional and that of the mathematical conditional. Recall that in the case of the mathematical conditional, the antecedent and consequent are independent propositions. Moreover, there is no ordering relationship between them. Our study suggests strongly that the popular view of counterfactual thinking as "complex" is a byproduct of theoretical frameworks which consider that the natural

Counter/actual reasoning and desirability

231

language conditional has a similar structure to that of the mathematical conditional. In conclusion, we hope to have demonstrated instead that counterfactual reasoning as it occurs typically in everyday situations is inextricably related to an expression of the speaker's stance toward the desirability or undesirability of a particular outcome. Abbreviations ATTR COMPLTV COND CONN COP EXCL FLLR GEN GER :MDL NML ONM PLN PRT PST SM TAG TM

ATTRIBUTIVE COMPLETIVE CONDITIONAL CONNECTIVE COPULA EXCLAMATION FILLER GENITIVE GERUNDIVE MODAL NOMINALIZER ONOMATOPOETIC PLAIN FORM PARTICLE PAST SUBJECT MARKER TAG MARKER TOPIC MARKER

Notes

*

1.

We would like to dedicate this study to the memory of James D. McCawley. If Jim had not written McCawley (1981) we would never have learned the joy of discovering the logic of natural languages. The authors have contributed equally to the production of this chapter and are both equally responsible for its contents. The authors are grateful to the two volume editors, Bernd Kortmann and Elizabeth Couper-Kuhlen, and to Sandy Thompson for their insightful comments and constructive criticisms, which greatly enhanced the quality of this paper.

232 Noriko McCawley Akatsuka and Susan Strauss 2. 3.

4.

5.

See Akatsuka (1985) for a full discussion of the epistemic scale where this relationship is discussed in cognitive terms. The Japanese data were collected and transcribed by Shoichi Iwasaki and his research assistants, Eri Yoshida and Yumiko Kawanishi. Data collection and transcription for the Japanese data appearing here were partially supported by the University of California's Pacific Rim Study Grant (1992-94). The English and Korean used in this study data were collected and transcribed independently from the Iwasaki data. The magnitude of the Northridge earthquake was 6.6. In all, 51 people lost their lives and over 5,500 people were injured. For the sake of comparison, the earthquake in Kobe, Japan, which occurred exactly one year later at approximately the same time of the morning, measured 7.2 on the Richter scale~ the number of fatalities in the Kobe earthquake exceeded 5,000. See Akatsuka (1986) for an earlier discussion of this point based on introspective data

References Akatsuka, Noriko 1985 Conditionals and epistemic scale. Language 61: 625-639. 1986 Conditionals are discourse-bound. In: Elizabeth Closs Traugott, Alice ter Meulen, Judy Snitzer Reilly, Charles A. Ferguson (eds.), On conditionals, 333-352. Cambridge: Cambridge University Press. 1991 Dracula conditional and discourse. In: C. Georgopoulos and R. Ishihara (eds.), Interdisciplinary approaches to language: Essays in honor of S.-Y. Kuroda, 25-37. Dordrecht/BostonILondon: Kluwer Academic Publishers. 1997 On the co-construction of counterfactual reasoning. Journal of Pragmatics 28: 781-794. 1999 Towards a theory of desirability in conditional reasoning. In: Akio Kamio and Ken-ichi Takami (eds.), Function and Structure, 195213. AmsterdamlPhiladelphia: Benjamins. Atkinson, J. Maxwell and John Heritage (eds.) 1984 Structures of social action: Studies in conversation analysis. Cambridge: Cambridge University Press. Clancy, Patricia M., Noriko Akatsuka and Susan Strauss 1997 Deontic modality and conditionality in adult-child discourse: a cross-linguistic study. In: A. Kamio (ed.), Directions in functional linguistics, 19-57. Amsterdam: Benjarnins. Fauconnier, Giles 1985 Mental spaces. Cambridge, MA: MIT Press.

Counterfactual reasoning and desirability

233

Ford, Cecilia E. and Sandra A. Thompson 1986 Conditionals in discourse: A text-based study from English. In: Elizabeth Closs Traugott, Alice ter Meulen, Judy Snitzer Reilly, Charles A. Ferguson (eds.), On conditionals, 353-372. Cambridge: Cambridge University Press Geis, Michael L. and Arnold M. Zwicky 1971 On invited inferences. Linguistic Inquiry 2: 561-566. Grice, Paul H. 1975 Logic and conversation. In: Peter Cole and Jerrold L. Morgan (eds.), Syntax and semantics 3: Speech acts, 41-58. New York: Academic Press. Haiman, John 1978 Conditionals are topics. Language 54: 564-589. Jackson, Frank (ed.) 1991 Conditionals: Oxford readings in philosophy. New York: Oxford University Press. Karttunen, Lauri and Stanley Peters 1979 Conversational implicatures. In: C.-K. Oh and D.A. Dinneen (eds.), Syntax and semantics 11: Presupposition, 1-56. New York: Academic Press. Labov, William 1972 The transformation of experience in narrative syntax. In: William Labov (ed.), Language in The Inner City: studies in the Black English vernacular, 354-396. Philadelphia: University of Pennsylvania Press. McCawley, James D. 1981 Everything that linguists have always wanted to know about logic **(but were ashamed to ask). Chicago: University of Chicago Press. Quine, Willard Van Orman 1950 Methods oflogic. New York: Henry Holt and Co. Sacks, Harvey, Emanuel Schegloff and Gail Jefferson 1974 A simplest systematics for the organization of turn-taking for conversation. Language 50: 696-735. Stalnaker, Robert 1968 A theory of conditionals. Reprinted in: William L. Harper, Robert Stalnaker and Glenn Pearce (eds.) (1981), 1ft: conditionals, belief, decision, chance, and time, 41-56. Dordrecht/Boston: Reidel/ Kluwer. Terrace, Herbert 1979 Nim: A chimpanzee who learned sign language. New York: Washington Square Press. White, Jack E. 1999 "Elegy for a Gone Boy" Book review in Time magazine October 4, 1999.

234 Noriko McCawley Akatsuka and Susan Strauss

TEXTS: 1994

Los Angeles earthquake data: English data: Susan Strauss, 1994 Japanese data: Shoichi Iwasaki, 1994 Korean data: Susan Strauss and Yumiko Kawanishi

Adversative connectors on distinct levels of discourse: A re-examination of Eve Sweetser's three-level approach * Ewald Lang

While basically acknowledging the discemability of three levels on which a sentence is interpreted (content, epistemic and speech act level), the paper tackles Sweetser's claim that the correct interpretation of interclausal connections depends not on form, but on a pragmatically motivated choice. I argue for a division of labour between grammar and pragmatics proper by (i) showing that there are more level-specific structural differences at issue; (ii) proving the notion "epistemic" to be feasible only if statements, assumptions, and inferences are kept distinct; (iii) providing an account for the unavailability of content-level interpretation for adversative conjunctions.

1. Introduction In Chapter 4 of her exciting book From Etymology to Pragmatics (1990), Eve Sweetser offers a fresh view of the notorious diversity of interpretations we observe with causal and adversative connectors. She rejects the view that connectors like because, although or and and but are lexically polysemous in the classical sense and suggests instead that the extremely varied uses of these connectors should be explained in terms of pragmatic ambiguity. Drawing on Sweetser (1990:76-78) the main tenets of her approach can be summarized as in (I)-(ill): (I)

One meaning, several uses

In polysemy, a morpheme has several related semantic values; in pragmatic ambiguity , a single semantics is pragmatically applied in different ways according to pragmatic context. (1990:76)

236 Ewald Lang

(II)

Three levels of interclausal connection ("conjunction"l) ... conjunction may be interpreted as applying in one of (at least) three domains [where] the choice of a "correct" interpretation depends not on form, but on a pragmatically motivated choice between viewing the conjoined clauses as representing content units, logical entities, or speech acts. (1990:78 ) [emphasis mine -EL]

The domains of interpretation are illustrated by clear-cut examples such as (la-c) below, all of which have in common that the second clause asserts a proposition in the form of a statement. (1)

a. John came back because he loved her.

b. John loved her, because he came back. c. What are you doing tonight,

because there's a good movie on.

[content or fact level] [epistemic level] [speech act level]

The levels are distinguished depending on what the clauses are taken to represent and hence on what sort of entities the causal relation is taken to obtain. Thus, in (la) it is two content units between which a factual causal relation is asserted to obtain. More exactly: the content units are two propositions whose couching in declarative sentences enables them to render assertable statements. The linking pattern for a causal relation on the content level thus amounts to STATEI\ffiNT because STATEI\ffiNT. In (1 b), the first clause is not asserted as a statement but only as an assumption which is inferred from what is stated as fact in the second clause. The causal connector marks the second clause as sufficient evidence justifying the assumption rendered by the first clause. The linking pattern on the epistemic level is ASSUMPTION, because STATEMENT (EVIDENCE). The hypothetical status of what is asserted by the first conjunct can be made explicit by embedding it under modals cf (1 b-i, ii) or by rephrasing it as an inference - cf (1 b-iii): (1)

b-i. John must have loved her, because he came back. b-ii. John undoubtedly loved her, because he came back. b-iii. From the fact that he came back, I conclude that John loved her.

Adversative connectors on distinct levels ofdiscourse

237

In (Ic), the first clause serves the purpose of a speech act of asking whose performance is being justified by the statement rendered by the second clause. The linking pattern thus is SPEECH ACT, because STATE:MENT (JUSTIFICATION). The because-clause does not form part of the question but is linked instead to an understood assertion which, if explicitly inserted, may read like this:

(1)

C-l.

What are you doing tonight? I am asking that because there's a good movie on.

In the course of presenting a variety of cases that are less clear-cut than the ones in (la-c) above, Sweetser does of course mention disambiguating features that may (but need not) be present in the grammatical structure of the conjunction. For example, she hints at the role of intonation: commaless intonation is linked with the content reading in (Ia) but unavailable to the other readings in (lb) or (lc); the latter require a so-called "comma intonation" such that the clause-final intonation drop marks the first clause as an independent assertion. Sweetser also discusses the role of iconic ordering of clauses in different domains (see her Ch. 4.2; more on this in § 3 below). The overall tendency of her approach, however, is based on the following claim :

(ill) Pervasiveness of pragmatic ambiguity Given sufficient context, we can almost always force either a content-conjunction reading or an epistemic-conjunction reading on any pair of clauses conjoined by because; it is just harder to find reasonable contexts for some readings than for others. [in addition to because] therefore, since, so, although, and despite < ... > all show such multiple usage. (1990:77-8)

2. Objections and amendments While 1 fundamentally agree with claim (I) One meaning, several uses (I consider this to be the only promising way of dealing with connectors!), and while 1 am basically sympathetic to the notion of "pragmatic ambiguity", 1 see some serious difficulties in accepting the claims in (II) and (ill).

238 Ewald Lang

The notion of "pragmatic ambiguity" is in need of refmement and elaboration. To achieve this aim Sweetser's approach will be relativized by (N) a. challenging the ambiguity claim advocated in (IT) and (ill); b. inspecting more closely the features based on which the supposed levels are distinguished; c. checking the consistency and exhaustiveness of the three-level distinction.

Pragmatics is not considered to be the sort of stop-gap device which Sweetser purports it to be; rather, what Sweetser calls "a pragmatically motivated choice of interpretation" (see (IT) above) will be shown to draw heavily on structural information available from the conjoined clauses as well as the connectors occurring in the conjunctions at issue. Specifically, I will defend (and provide evidence for) the following counter-claims, each of which has its bearing on the three points of criticism in (N). For ease of reference, these claims will be prefixed by "C-". The rest of this section will concentrate on (C-l), the others will be discussed in § 3. (C-l)

There are more level-specific structural differences in the way in which clauses are conjoined than Sweetser's claims (I) - (ill) would admit.

(C-2)

The lexical items, i.e. the connectors that are used to connect clauses on the three levels of interpretation, display more levelspecific distributional restrictions than the sample of cases discussed by Sweetser would suggest.

(C-3)

The three-level distinction proposed by Sweetser must be supplemented by another level-provisionally called "textual progression" or "discourse perspective"-which is not located within the Sweetserian hierarchy of levels but cuts across them.

(C-l) emphasizes the role of structural information in determining the interpretation of a given conjunction. Its import can easily be demonstrated by taking up Sweetser's introductory set of examples. A closer look at (la-c) immediately reveals that there is a whole range of structural dimensions with respect to which the clauses that are linked together by connectors differ. Spelling out (C-l) in detail means checking the clauses of what Sweetser calls a "conjunction" for a possible display

Adversative connectors on distinct levels ofdiscourse

239

of some crucial structural features, e.g. those listed in (CI-I) to (CI-5) below. (CI-I) Are the clauses of the same or of different clause type? (in morphosyntactic terms of being declarative, interrogative, or imperative sentences) It is obvious that a content-level interpretation in the sense of Sweetser's paragon example (la) is confined to clauses which share declarative mood and thus may render STATEMENTS. Taking this as a necessary condition for content-level interpretation implies that any conjunction whose clauses differ in type is excluded from being interpreted on that level and is thus confined to being interpreted on epistemic and/or speech act level. 2 Sharing declarative mood is a necessary, though not a sufficient, condition for a conjunction of clauses to be interpretable on the content level. Consider again (lb) and a version of (1 c) in which the SPEECH ACT of ASKING is not grammatically overt but concealed in a declarative: (1)

b. John loved her, because he came back. c'. I am asking you what you are doing tonight, because there's a good movie on.

This is the point where further criterial features enter the picture. One of them, of course, is comma intonation. Due to its interaction with other features the indicative role of comma intonation is more far-reaching than Sweetser assumes. Below the level of sharing declarative mood there are important indicators of clause-internal differences which interact with comma intonation. Thus, in isolation, the first clause of (1 b) may render a STATEMENT as does the first clause in (la). Both behave differently, though, if we add an adjunct like presumably or a parenthetical like I guess, which mark the hypotheticality or assumption status of the assertion being made: (1)

a'. Presumably John came back because he loved her. John, I guess, came back because he loved her. b'. Presumably John loved her, because he came back. John, I guess, loved her, because he came back.

While in (la') hypotheticality markers like presumably take scope over both clauses, their scope in (1 b') is confined to the first clause. Thus the

240 Ewald Lang

function of comma intonation to mark the first clause as an independent assertion is confirmed by its effect of marking scope boundaries. Similar observations can be made for negation and other operators. Despite looking alike on the surface, the first clauses in (la, Ia') and (Ib, Ib') differ in scope determination as well as in intonation, which undoubtedly both form part of core grammar, and hence provide distinct cues for the interpretation to be assigned to the conjunction as a whole. Thus despite the fact that the first clauses in (1a) and (1 b) are alike in sharing declarative mood, they also differ. The first clause in (Ia) asserts a proposition in the form of a factual statement thus allowing for a content-level interpretation of the conjunction. The first clause in (1 b), however, asserts a proposition in the form of an ASSUMPTION as evidenced by comma intonation and characteristic scope properties. The assumption status of the first clause is crucial to epistemic-Ievel interpretation of the conjunction. The next dimension in support of level-specific structure formation as claimed by (C-l) is mentioned by Sweetser in discussing "iconic ordering" in coordinate structures. However, the point at issue can be extended to causal connectors and exploited in a more systematic way. (C 1-2)

Restrictions on the order of clauses with subordinators

(Ia) on the content level allows both orders-John came back because he loved her and Because he loved her John came back, though there is a difference that distinguishes backgrounded vs. focussed material on the level of discourse-related information structure. In any case, preposing the subordinate clause does not prevent a content-level reading as shown by (2a). Coherence is maintained independently of the order of clauses. However, as shown in (2b-c'), Sweetser's examples (Ib) and (Ic) as well as their German equivalents do not allow of preposed because/weil-clauses if they are to keep their respective readings on the epistemic or speech act level (marked * and ??, respectively): (2)

a. Weil er sie liebte, kam John zuriick. 'Because he loved her John came back. ' b. Weil er zuriickkam, liebte John sie. 'Because he came back, John loved her. '

[content]

*[epistemic]

Adversative connectors on distinct levels ofdiscourse

241

b'. Weil er zuriickkam, LIEBTE John sie vermutlich / denke ich. [epistemic] 'Because he came back, John presumably / I guess LOVed her. ' c. Because there's a good movie on, what are you doing tonight? ??[speech act] c'. Because there's a good movie on, I'd like to ask you: what are you doing tonight? [speech act] Thus, unlike (lb), a conjunction like (2b) cannot be interpreted as justifying the conclusion "he loved her" from the fact that "he came back". What is lacking here is the sort of coherence that is rendered by the original ordering. To ensure coherence in such cases requires additional lexical or prosodic focus indicators, as is illustrated in (2b'). Similarly, while clause type difference and intonation break may serve as compensatory cues, (2c) is still incoherent. An amended version of it should read like (2c'). Sweetser recognizes the role of conjunct ordering in discussing conjunctions with coordinators like and, or and but by stating that pragmatically "the primary, independent conjunct precedes the secondary, dependent conjunct" (1990:98). Yet, she simply ignores the ordering effects in conjunctions with subordinators. The facts in (2b, 2c) above show that (Cl-2) has to be taken as a condition on level-specific structure formation, which in tum has serious repercussions on Sweetser's ambiguity claim in (II)-(ill) above. First, it is a grammatical category feature of subordinators like because, since, etc. to allow for clause preposing while keeping their interpretation constant-cf (2a). If this built-in property fails, as with (2b) for example, this has to be acknowledged as a disambiguating grammatical fact, too. It is on this basis that pragmatic interpretation starts to work, contrary to the ambiguity claim in (II). Second, unlike subordinators, coordinators like and, or, but (though notfor or German denn) are grammatically unrestricted as to the order of conjuncts and hence do not provide category-based grammatical information that is taken up by pragmatic interpretation. Among other things, it is the sequential ordering of the conjuncts that is submitted to pragmatic interpretation (along the lines discussed by Sweetser). Third, in view of this, we should expect coordinate conjunctions to be less fertile, hence more ambiguous or vague, in providing grammatically determined information on which, following Sweetser, the "pragmatic choice of the' correct' interpretation" may draw.

242 Ewald Lang

In what follows it will be shown that this assumption is by no means justified. Due to space restrictions the further pursuit of (C-I) to (C-3) will be confined to coordinate conjunctions whose clauses are linked by adversative connectors. Moreover, we will employ German data3 and present arguments which draw on results from current work being carried out in the connector project at the Institut fur deutsche Sprache Mannheim (for details see HdK [=Pasch et al. (forthcoming)]).

3.

Adversative connectors

3.1.

How ambiguous are aberlbut-conjunctions?

From the set of about 30 lexical items that may be subsumed under the label "adversative connector" in German we will select only those needed to illustrate (C-I) to (C-3). Let us first turn to coordination involving the connector aber, which (like but in English) is commonly considered as the standard adversative connector. Compared with connectors such as trotzdem ('nevertheless'), hingegen ('whereas'), or dennoch ('nonetheless', 'all the same'), aber is much less restricted regarding distribution and hence to a much higher degree prone to "pragmatic ambiguity". Nevertheless, the supposed ambiguity of aber-conjunctions is drastically reduced by cues which emerge from structural features of the coordinate clauses in the domain of sentence structure. In addition to CI-I and CI-2, the relevant features include: (CI-3)

Matching or independently varying syntactic clause structure

(CI-4)

Parallel or independently varying information structure

(C 1-5)

(Un)availability of clause-internal deletions like "gapping", "right node raising", etc. (Lang 1991, see also th. 4 below).

Let us now look at some data that are representative of the range of cases to be taken into account. (3)

a. Dein Vortrag war sehr lang, aber das soli keine Kritik sein.

'Your talk was rather lengthy, but this is not a criticism. '

Adversative connectors on distinct levels ofdiscourse

243

b. Wir sind mit dem Manuskript in Verzug, aber welcher Autor halt schon Termine ein? 'We are late with the manuscript, but what author keeps the dead-lines?' c. Willst du heute etwa ins Kino-aber ich komme nicht mit. 'Do you want to go to the movies today-but I'm not coming along' All examples in (3) are easily recognizable as cases to be interpreted on the speech act level. Just recall the relevant indicators mentioned in (C II), namely "clause type difference" - cf (3b, 3c), "intonational break" cf (3a), and what was said about "independent clause-internal structures" in CI-3 to CI-4 (i.e. lack of parallelism as to constituency and topic-focus articulation). Just as in (lc) above, it is the conjunct-internal differences which prevent the examples in (3) from being interpretable on the content level and qualify them for speech act level interpretation. Generally speaking, the reason is this: aber/but marks what is rendered by the second clause as being in contrast with inferences that might be drawn from the speech act that is performed by uttering the first clause. In short, what counts are the conjunct-internal differences in terms of C1-1 to C14. If such differences are missing, more effort is needed to determine the respective levels of interpretation. Consider the following examples, which Sweetser tends to locate at the epistemic level. (4)

a. Hans hat Unmengen von Zigarren gehortet,

aber er raucht gar nicht. 'Hans has hoarded an obscene number of cigars, but he doesn't even smoke' b. John keeps six boxes ofpancake mix on hand, but he never eats pancakes. (cf Sweetser (1990: 100» Sweetser comments: "But presents two conjuncts which clash with each other in some way ... At the epistemic level, the available premises may clash with an apparently necessary conclusion" (1990: 100). The first part of the quote is unproblematic; but what exactly is "epistemic" about the entities connected by aber/but? (4a, b) instantiate the well-known type of "denial of expectation" interpretation of but-conjunctions (Lakoff 1971:133). In both cases, the clauses are mutually independent, hence semantically compatible STATEMENTS. Given this, (4a, b) qualify as coordinations of statements at the content level with the additional requirement that the interpretation involves inferring an ASSUMPTION

244 Ewald Lang

which is in contrast with the two statements being presented side by side. This analysis reduces the "epistemic" aspect of (4a, b) to the inferences to be drawn and thereby does justice to the structurally determined leveldistinction. The same point was made in Foolen (1992:83).

3.2.

Problems with interpreting but/aber at the content level

The awkwardness of Sweetser's use of the notion "epistemic" is furthermore revealed by the stand she takes on cases of what since Lakoff (1971:133) is commonly known as "semantic opposition" but. Sweetser (1990:103) claims: "The but in (56) [= (5a) below] does indeed indicate contrast: an epistemic contrast between two semantically opposed propositions" (italics mine - EL). From this we may conclude that (5b) lacks such an epistemic contrast, other things being equal. Again, (5b) with and is a clear-cut case of content-level conjunction. (5)

a. John is rich but Bill is poor. [= Sweetser's (56), 1990:103 ] b. John is rich and Bill is poor. c. John is rich, but Bill is rich, too. d. *John is rich but Bill is rich.

What we are left with is the problem of where to locate (5a): it does indeed indicate a different contrast from (5b). I doubt, however, whether this contrast can be accounted for in terms of the "epistemic level" without rendering this notion completely void. I would contend that the alleged "epistemic" difference between (5a) and (5b) is that the interpretation of (5a) requires some additional inferencing which the interpretation of (5b) does not. Thus, (5a) and (5b), like (5c) when compared with (5d), clearly differ as to the conditions they impose on the context in order to get interpreted properly. According to an observation first made by Bellert (1972), the interpretation of but/aber-conjunctions involves at least two semantic contrasts to be read off or inferred from the conjuncts. (5d) contains the overt contrast John-Bill but fails to provide cues to reconstruct the second contrast needed, i. e. an INFERRED ASSUMPTION which is in opposition to the STATEMENT Bill is rich. This pragmatic insufficiency has its grammatical counterpart: (5d) can be considered deviant because it does not allow a contrastive stress pattern - just try and you·will see' Moreover, contrary to most work done in the field, we claim that the "semantic opposition" (more exactly, the occurrence of antonymous

Adversative connectors on distinct levels ofdiscourse

245

predicates in logically compatible STATEMENTS) in (Sa) is not a crucial feature of this type of conjunction (as shown by (5c) which lacks opposition); and that this opposition is by no means the carrier of the contrast indicated by aber/but. Instead, the contrasting entity has to be inferred from outside the coordination. However, the opposition in (5a) as well as its absence in (5c) serve as structural clues to reconstructing it. We will return to this in 3.3 below. Now, if (5a) cannot be located on the epistemic level, why not consider it a content-level conjunction - very much like (5b)? Sweetser (1990:103) rejects this idea by stating: "I have not been able to unearth any indubitable content-conjunction examples with but" and by offering "as a plausible explanation for the use of but in only two domains" [epistemic level and speech act level - EL] the following consideration: "what does it mean to say that A and B "clash" or "contrast" in the real world? How can discordance or contrast exist outside of the speaker's mental concept of harmony or non-contrast?" Sure enough, the contrast induced by aber/but is not something to be looked for in the real world but something to be established by the interlocutors in assessing what, by way of coordinate conjoining, is presented as co-existing in the world. But this does not prevent aber/butconjunctions from being interpretable at the content level, that is, as conjoining two STATEl\1ENTS which lay the ground for additional inferencing. Sweetser's problem with interpreting but at the content level is caused by confusing two notions which, though not unrelated, have to be kept distinct: inference (whether a proposition is asserted or inferred) and epistemic status (whether a proposition forms the content of a factual STATEMENT or an ASSUMPTION). The alternative approach suggested here tries to avoid this confusion. It draws on the fact that the interpretation of an adversative conjunction involves relating the conjoined clauses to an ASSUMPTION which-if not available in explicit terms (see (V) below)-has to be inferred from outside the conjunction. Based on these considerations it is posited that (V)

(a) adversative (and probably also concessive) connectors inherently contain pointers to previous information available from the context, and due to this (b) adversative conjunctions necessarily involve some backtracking that may well go beyond the domain of sentence structure and operates on (what may be called) the level of "textual progression" or "discourse perspective".

246 Ewald Lang

Following (C-3) it is this level which provides the appropriate means to interpret adversative conjunctions, and therefore we consider it a necessaIy supplement to Sweetser's approach. Note that the level of "discourse perspective" must not be understood as an additional level within, or on top of, the Sweetserian hierarchy, but rather as embracing the three levels discussed so far as grammatically determined patterns of discourse formation. Recall that C1-1 to Cl-5 apply to sentence structure and thus clearly contribute to discourse formation and comprehension. 3.3.

Adversative conjunctions and the division of labour between grammar and pragmatics

This section will provide an outline of the way in which sentence and discourse level interact in determining the interpretation of adversative conjunctions. To begin with, note that the semantic contribution of aber/but in combining two clauses into a conjunction is twofold. In a nutshell, it may be conceived of as: (VI) (a) conjoining semantically compatible and non-inclusive propositions that-depending on the number of structural parallels between the conjuncts-may be bundled up or enumerated as instances of a "common integrator" (Lang 1984, 1991, in print; see also fn. 4 below); (b) indicating that the assertion rendered by the second clause is in contrast to an ASSUMPTION that either may be read off, or must be inferred from, previous information (Braul3e 1998). (VIa) states what aber/but share with und/and; therefore (VIa) is the basis for (Sa) and (Sb) having identical truth conditions. This is a feature of aber/but that probably nobody will deny. (VIb) is what separates aber/but from undland, and hence (VIb) is what causes (Sa) and (Sb) to have distinct conditions of use as regards semantic coherence. As such, (VIb) is perhaps not as uncontroversial as (VIa), so it may be worth commenting on. What has been stated in (VIa + b) should not be understood as amounting to "but = and + pragmatics". This would defmitely miss the target. The conditions of use of aber/but and other adversative connectors are not simply a matter of "pragmatically motivated interpretation" (as Sweetser suggests); rather they are rooted in the lexical semantics as well

Adversative connectors on distinct levels ofdiscourse

247

as in the syntactic category features of these connectors. On the other hand, I do not deny the role which contextually induced pragmatic interpretation inevitably has to play in specifying the "correct" reading which a conjunction of clauses eventually gets assigned. I differ from Sweetser, however, by claiming that in order to account for (what she calls) "pragmatic ambiguity", the division of labour between grammar and pragmatics has to be redefined. Regarding aber/but as characterized in (VIb), this division of labour may be determined as follows: That part of (VIb) which reads "indicating that the assertion rendered by the second clause is in contrast to an ASSUMPTION" has to be spelled out primarily in terms of grammatical structure formation (including C1-1 to CI-5), while the other part, which reads "an ASSUMPTION that either may be read off, or must be inferred from, previous information", is where pragmatics proper enters the picture. It seems reasonable to consider the interpretation of an adversative conjunction as involving a search device which is determined by a specific source-target relation. Triggered by the lexical indication that a contrast must be established, the device takes the assertion rendered by the second clause as the source from which to look for an appropriate ASSUMPTION that meets the condition of contrast, i.e. the target. Though the search device is to some extent guided by grammatical structure, as claimed by (Cl-l) to (CI-5), it is inevitably accompanied and supplemented by pragmatically based reasoning. So the part to be played by pragmatics proper is to provide supplementary means to find out what the target, i.e. the INFERRED ASSUMPTION, is. Note that by dividing up (VIb) between grammar and pragmatics as suggested afore, it is not an accident that the notion "ASSUMPTION" figures in both, and rightly so. The INFERRED ASSUMPTION forms the bridge between grammar and pragmatics in interpreting adversative conjunctions. Almost all studies of adversative connectors have attempted to account for it in one way or another (for a survey see Rudolph 1996). What is new is the way to tackle it as proposed in (VIb). Source and target are both propositional but differ as to their epistemic status. That the target (i.e. the proposition which the second clause is put into contrast with) has to be an ASSUMPTION, not a factual STATEMENT, can be seen if, for example, (4a) is extended by a clause that renders the target in explicit terms, cf the bold-faced additions in (6):

248 Ewald Lang

(6)

a. Hans hat Unmengen von Zigarren gehortet,

aber er raucht gar nicht. 'Hans has hoarded an obscene number of cigars, but he doesn't (even) smoke' b. Hans hat Unmengen von Zigarren gehortet, *folglich raucht er, aber er raucht nicht. ' ... hence he smokes, but he doesn't smoke' c. Hans hat Unmengen von Zigarren gehortet, also konnte man vermuten, daB er raucht, aber er raucht nicht. , ... so one could assume that he smokes, but he doesn't smoke'

[= 4a]

If the supposed target of contrast is expressed in the form of a FACTUAL STATEMENT, the result is a contradiction-cf (6b), but if it appears with overt markers of ASSUMPTION status as in (6c), the inter-

pretation of the entire construction is consistent. This is what follows from the grammar part of (VI), which requires the target of contrast to be an ASSUMPTION. Let us now turn to the pragmatic aspects of (VI). If the clauses conjoined by aber/but both represent STATEMENTS (as is the case with the majority of adversative conjunctions), the ASSUMPTION forming the target of contrast has to be inferred. Here is where pragmatics comes into play by providing the means to infer the target assumption from previous information. Reconsidering the conjunctions in (4) and (5) above will then amount to the following analysis. In (4a/6a), the inference is obviously drawn by resorting to world knowledge about smokers such that the STATEMENT rendered by the first clause (Hans has hoarded an obscene number ofcigars) is taken as a premise which by abductive reasoning makes it plausible to assume "Hans smokes" and thus to yield a suitable target assumption. By the way, the fact that this inference is not based on deductive but on abductive reasoning provides additional evidence of the ASSUMPTION status of the target. In (Sa-d), repeated as (7a-d) below, the inferential steps leading to the target are less clearly set out. (This is what caused the trouble for Sweetser's approach, cf § 3.2 above.) How come? The clauses conjoined by aber/but in (5/7) are clear-cut STATEMENTS, so the target of contrast has to be inferred from outside the conjunction. Moreover, the clauses conjoined are absolutely parallel as to their internal structure. So they cannot but be taken as instances of a

Adversative connectors on distinct levels ofdiscourse

249

"common integrator".4 The common integrator, which has to be identified within the conjunction (see fu. 4 below) or else has to be inferred from outside the conjunction, then serves as the location frame within which the contrast is to be placed. Thus, in comparison to (4a16a), the conjunctions in (5/7) require additional cues from the discourse level in order to be assigned an appropriate interpretation. The simplest way to illustrate these context conditions consists in showing what sort of questions the respective conjunctions in (7) might serve to answer. Suitable questions (added in 11...11 ) thus reveal the distinct effects that are caused by the structural differences of the conjoined clauses. (7)

a. John is rich but Bill is poor. liAs for John and Bill, are they both rich? II b. John is rich and Bill is poor. II As for John and Bill, what about their income? II c. John is rich, but Bill is rich, too. II As for John and Bill, they aren't both rich, are they? II d. *John is rich but Bill is rich. II??? II

The structural differences are clearly brought out by the fact that none of (7a-d) is replaceable with any of the others as an appropriate answer to the respective question-context given in II. ..11. Assessing (517) once more, we may say: in isolation and in written form these conjunctions are not really ambiguous but are underspecified regarding the cues they need to get interpreted. Finally, we will look at those grammatical features of adversative connectors that form the basis of the division of labour between grammar and pragmatics just outlined. The most relevant point to note is this: indicating a contrast in the way suggested in (VIb) is a semantically builtin feature of aber/but which correlates with likewise built-in syntactic properties of these connectors. In contrast to and, or, etc., adversative coordinators like aber/but display the following features: First, aberlbut allows for binary connection only. This is the basis for marking the assertion following aberlbut as the source from which to look for an appropriate CONTRASTING ASSUMPTION, that is, the target. Second, aber Ibut is intrinsically asymmetric in marking what has been called a "change of perspective" (Spooren 1989): for example, (4a) or (7a) with the reversed order of conjuncts would not render the same

250 Ewald Lang

information on the level of discourse perspective, because source and target are reversed. Third, unlike und/and, which allows for scope ambiguities, aber/but is a marker of scope boundaries: negation or other scope-taking operators cannot take scope over aber/but, cf:

(8)

Hans ist nicht dumm undfaul 'Hans is not stupid and lazy'

(9)

Hans ist nicht dumm, aberfaul [scope of negation unambiguous] 'Hans is not stupid but he is lazy'S

[ambiguous as to scope of negation]

With these features in mind let us now examine a couple of aber/but conjunctions which display subtle but telling differences in the information structure of the conjuncts. To begin with, note that in written form (10) below is underspecified as to its information structure and hence to the way it is linked with previous discourse. However, when (10) occurs as a continuation of (11) or (12), which (along the lines of (5/7) above) form distinct discourse contexts which (10) has to match, we realize that (10) assumes correspondingly distinct patterns of information structure shown in (11) and (12), respectively. (In the following examples, SMALL CAPS indicate main stress, I and \ contrastive stress, labeled square brackets topics and foci, respectively.)

(10) Hans ist krank, aber seine Frau geht arbeiten. 'Hans is off sick but his wife is at work' (11) II Wie geht's den beiden? II 'How are the two doing?' a.

Tl[/Hans] F[ist KRANK], aber T2[seine Frau\] F[geht ARBEITEN] 'Hans is off sick but his wife is at work'

b.

T[[Hans] und [seine Frau]] sind F[[IKRANK] bzw. [auf ARBEIT\]] 'Hans and his wife are off sick and at work, respectively'

(12) II Was regt dich so auf?11 'What's upsetting you?' a.

F[[Hans] ist KRANK], aber F[[seine Frau] geht ARBEITEN] 'Hans is off sick but his wife is at work'

Adversative connectors on distinct levels ofdiscourse

b.

F[[Hans] ist KRANK], dennoch F[[seine Frau] geht ARBEITEN] 'Hans is off sick, nevertheless, his wife is at work'

c.

F[Obwohl [Hans] KRANK ist], F[geht [seine Frau] ARBEITEN] 'Although Hans is off sick, his wife is at work'

251

Let us take a closer look at the way in which information structure at sentence level and at discourse level interact. The contexts adduced in (11) and (12) each form a discourse topic whose internal structure determines the way in which it is elaborated by the sentences following it. The question in (11) marks the NP den beiden 'the two' as a (continued) topic and specifies the VP as question focus, thus determining focussed VPs as answers. In view of this, (I1a) shows a topic-focus articulation which elaborates that of the discourse topic in a strictly parallel way: the topic den beiden is resumed by the topical subjects Hans and seine Frau, respectively; the VP question focus is instantiated by the VPs ist krank and geht arbeiten, respectively. This mode of parallel elaboration is evidenced also by two symmetry features: (i) the clauses in (lIa) could be reordered without changing their links to the discourse topic; (ii) (IIa) can be regrouped into a respectively construction like (lIb). Hence, (11a,b) instantiate what might be called "strictly parallel elaboration of a discourse topic". In (12), where was ('What') refers to a whole proposition, the question is such that it requires all-focus sentences as answers. Given this, (10) is assigned the information structure in (I2a), which determines the conjuncts as all-focus sentences at the discourse level. The informational status of Hans and seine Frau is open: syntactically they are sentence topics, but it depends on further discourse factors whether or not they resume a discourse topic as they do in (I1a,b). Now, (12a) is a predetermined elaboration of the discourse topic Was regt dich so auf? ('What's upsetting you?'), but there is no need for the elaboration to spread over a conjunction of clauses-Hans ist krank ('Hans is off sick') would suffice as an answer to (12) alone. In case the elaboration involves a conjunction of clauses, these are not bound to the sort of parallelism we observe with (11a,b). The relation between the all-focus clauses in (12a) is not dependent on the discourse topic. Hence, (I2a) is what might be called a "non-symmetric elaboration of a discourse topic. " This is evidenced by (12b,c), which answer the question by couching the same asserted propositions in concessive constructions. Concessive constructions6 , no matter whether coordinate like (I2b) or subordinate

252 Ewald Lang

like (12c), are inherently asymmetric regarding the relation between the clauses they link. To summarize: due to their distinct information structure, (12a-c) are not replaceable with (1Ia,b) while keeping the respective discourse contexts constant, and vice versa. Furthermore, the examples in (11) and (12) reveal an important difference in the way interclausal aber/but relate to the discourse. In the non-symmetric cases, the target from which the ASSUMPTION that contrasts with the STATE1vffiNT in the second conjunct ('His wife is at work') has to be derived is supplied by the first conjunct. So, in (12a-c) it is the assertion of John's being off sick that triggers the relevant inference: sick people need care, and spouses make good candidates for this job. In (11a,b), which form symmetric continuations of the given discourse topic, the target that licenses the use of aber/but is not as easy to reconstruct. Structurally, the question Wie geht's den beiden? ('How are the two doing?') does not provide an appropriate target per se; all it does is to license a list of and-conjoined items, cf (11 b). This is where pragmatics enters the picture. In interpreting (1Ia), we face the following situation: if searching the wider context for a suitable target fails, we are left with no alternative but to take the first conjunct Hans ist krank ('Hans is off sick') and generate from it the implicature that the news about his wife will be of similar quality. Given that, as claimed in (VIb), aber/but has the built-in feature of triggering an inference, the notion of a but of "semantic opposition" (introduced by Lakoff (1971» seems to be misconceived: semantic opposition per se never suffices to license the use of aber/but.

4. Concluding remarks Sweetser's three-level approach has proved to be valuable as a heuristic means in sorting the various interpretations that conjunctions can be assigned. However, in order to serve as a means of explanation, it needs to be modified in some respects. Regarding the level of grammatically determined sentence structure, Sweetser's level-distinction can be more precisely reconstructed in terms of criteria like Cl-l to CI-5, among which Cl-l (sameness or difference of clause type) is the most crucial one. Given this, we might reconstruct what should be kept of Sweetser's three-level distinction of conjunction interpretation in the following way:

Adversative connectors on distinct levels ofdiscourse

253

(VII) a. Content-level interpretation is reserved for conjunctions of declarative clauses that render STATEMENTS (i.e. entities to be related to the world via truth-conditions); see (la, 3a, 4a,5b,6a). b. Epistemic-Ievel interpretation is confined to conjunctions of declarative clauses, of which at least one is marked as rendering an ASSUMPTION; see (lb',lb-i, 1b-ii, 1b-iii, 2b'). c. Speech act level-interpretation is available for conjunctions of non-declaratives (with reservations about or-conjunctions like (2» as well as for conjunctions of clauses differing as to type but with one clause being marked as rendering a SPEECH ACT; see (lc, 1c', 1c-i, 2c, 2c', 3a-c). In addition, Sweetser's approach has to be modified in a way that would enable it to account for contextual requirements imposed on the interpretation of adversative conjunctions. The level that is assumed in (C-3) to do this is the all-embracing level of "discourse perspective". This is the level where (a) the information structure of sentences or clauses interacts with that of the discourse containing them-as exemplified by the cases in (7) and (10)-(12); (b) where the peculiar syntax and semantics of adversative connectors can be accounted for in a promising way; (c) where the division of labour between grammar and pragmatics can be defined - cf (VI a+b). Pragmatics comes in when grammar does not provide sufficient cues to get an adversative conjunction properly interpreted.

Notes *

1.

2.

This paper is dedicated to Ekkehard Konig on the occasion of his 60 th birthday. I wish to thank the editors for useful comments on an earlier draft of this paper and Chris Wilder for improving my English. Note that Sweetser uses the term "conjunction" for the syntactic operation of clause-linking without distinguishing coordination and subordination. In order to avoid confusion with the word class "conjunctions" I will call the latter "connectors" (subdivided into "coordinators" and "subordinators") and the constituents they link "conjuncts"~ otherwise I will follow Sweetser's terminology. It may be noted in passing that sameness of clause type with non-declaratives raises some problems concerning the conjoining of clauses on the speech act

254 Ewald Lang

3.

4.

level. I will take for granted that pragmatically (i) is asymmetric, while (ii) is symmetric regarding the alternatives presented: Give me liberty or give me death! [= Sweetser's (46), p. 98] (i) (ii) Give me a salami sandwich or a hotdog! What needs to be clarified, however, is this: is it justified to consider (ii) as an or-{;onjunction of two speech acts? Speech acts normally resist being presented as alternatives. Another open question is whether we are to take (ii) as a conjunction of two speech acts or rather as a single speech act of request which comprises an alternative of how the request might be complied with? Again, these questions show that a closer examination of the internal structure of the conjuncts seems to be unavoidable. The advantage of illustrating the issue at hand with German data consists in the fact that aber is an unambiguous adversative connector. It does not share the range of additional uses and readings that are covered by but, as in German these are rendered by separate lexical items like sondern, auf3er etc. Analysing but-constructions in comparison with aber-constructions thus will help us to pick out only those having adversative readings. Without going into details (see Lang 1984, 1991, in print), let me just point out that the derivation of the Common Integrator (CI) of a given conjunction (a) is sensitive to surface structure parallels shared by the conjoined clauses, and (b) can best be illustrated by means of a (multiple) question to which the given conjunction would serve as an appropriate answer. So while (i)-(iv) share the propositional formula "someone transfers something to someone" as constituent part of their respective Cis, the specific make-up of the latter rests on the information structure of the clauses, specifically on the (non-)coreference of the items instantiating the argument places. (Non-)coreference is indicated by (non-) coindexing. Coreference requires destressing: if e.g. the beast is coindexed with the dog the former serves as an anaphoric epithet for the latter. (i) Grandpa will bequeath the dogk to his grandchildreni and Grandma (will)(bequeath) the beastv*k to us j/*i CI: WHO will bequeath WHAT TO WHOM? (ii) Grandpa will bequeath the dogk to his grandchildreni and/but Grandma will sell the beastk/*l to us i/*j CI: WHO will transfer the dogk to USi IN WHAT WAY? (iii) Grandpa and Grandma will bequeath the dogkand the beastv*k to their grandchildreni and to us j/*i, respectively CI: WHO will bequeath WHAT TO WHOM? (iv) Grandpa will bequeath, and Grandma will sell, the dog to us CI: WHO will transfer the dogkto USi IN WHAT WAY? Note that the internal structure of the conjuncts in (i) allows for regrouping the parallel constituents into a respectively construction like (iii) which leads to the same CI as (i) but is, for instance, unavailable for (ii). On the other hand, (ii) can also be rendered by a right node raising construction like (iv) which leads to the same CI as (ii) but is unavailable for (i). So (iii) and (iv) may serve as overt syntactic diagnostics for information structure and CI

Adversative connectors on distinct levels ofdiscourse

5.

6.

255

differences that seem to be concealed when conjunctions like (i) or (ii) are presented in written form only. (8) and (9) point to an important difference between German aber and English but regarding scope determination under negation. Since German adversative coordinators for contrast (aber) vs. correction (sondern) are lexically distinct, their scopal behaviour is built-in and need not be indicated by other means-ef. (i, ii). English but-constructions, however, have to render the distinction of contrast vs. correction under negation by distinct conditions on deleting non-focussed material in the second clause. Thus, butconstructions like (iii) expressing contrast disallow, whereas butconstructions like (iv) expressing correction require, the deletion of repeated material in the second conjunct (for details see Lang (1991). (i) Hans ist nicht dumm, aber (er ist) faul [contrast] (ii) Hans ist nicht dumm, sondem (er ist) faul [correction] (iii) Hans is not stupid but *(he is) lazy [contrast] (iv) Hans is not stupid but (*he is) lazy [correction] The same point can be made to explain the difference in the following cases: (v) John ist reich, aber Bill (ist) arm. (vii) John is rich but Bill *(is) poor. (vi) John ist reich, aber Bill nicht lauch. (viii) John is rich but Bill *(is) not Itoo. Despite the differences shown, aber and adversative but both mark scope boundaries in preventing propositional operators from taking scope over them. The relationship between aber-conjunctions and concessive constructions cannot be tackled here, but see the contributions of Barth and of Crevels (in this volume). Another point that deserves mentioning is the fact that German aber (unlike English but) is not confined to interclausal positions, but can also float in the second clause (this is why in HdK [=Pasch et a1. (forthcoming)] aber has been classified as a "connective particle"). At first glance, this distributional flexibility seems to support the view that aber is most prone to "pragmatic ambiguity". It can be shown, however, that being allowed to float within the second clause is what enables aber to syntactically mark the information structure relation of the second clause with respect to the discourse topic as a non-symmetric one, similar to (12a-c) above: (12) d. F[[Hans] ist KRANK], F[[seine Frau] geht aber ARBEITEN] 'Hans is off sick, his wife is, nevertheless, at work' The prosodic details of contrastive vs. concessive clause linking are being investigated in a current DFG project of mine which forms the source of the data discussed here.

256 Ewald Lang R~ferences

Bellert, Irena 1972 On certain properties of the English Connectives and and but. In: Senta Plotz (ed.), Transformationelle Analyse, 327-356. Frankfurt am Main: Athenaum. BrauBe, Ursula 1998 Was ist Adversativitat? aber oder una? Deutsche Sprache 26:138159. Foolen, Ad 1992 Review of Sweetser 1990. Lingua 88:76-86. HdK [= Pasch, Renate, Ursula Brausse, and Eva Breindl] (to appear) Handbuch der deutschen Konnektoren. 2 vols. Ms. Mannheim: Institut flir deutsche Sprache. Kortmann, Bernd 1997 Adverbial subordination. A typology and history of adverbial subordinators based on European languages. (EALT 18.) Berlin/New York: Mouton de Gruyter. Lakoff, Robin 1971 If's, and's and but's about conjunction. In: Charles J. Fillmore and D. T. Langendoen (eds.), Studies in linguistic semantics, 114-149. New York: Holt, Rinehart & Winston. Lang, Ewald 1984 The semantics of coordination. (SLCS 9.) ArnsterdamlPhiladelphia: Benjarnins. 1991 Koordinierende Konjunktionen. In: Semantik/Semantics, Art. 26, 597-623. (HSK 6.) BerlinlNew York: Mouton de Gruyter. (in print) Die Wortart "Konjunktion". In: Lexicology/Lexikologie, Art. 88, (HSK 17.) Berlin/New York: Mouton de Gruyter. Raible, Wolfgang 1992 Junktion. Eine Dimension der Sprache und ihre Realisierungsformen zwischen Aggregation und Integration. Heidelberg: Winter. Rudolph, Elisabeth 1996 Contrast: adversative and concessive relations and their expressions in English, German, Spanish, Portuguese on sentence and text level. (Research in text theory 23.) BerlinlNew York: Mouton de Gruyter. Spooren, Willibrordus 1989 Some Aspects of the form and interpretation ofglobal constrastive coherence relations. PhD Dissertation, University of Nijmegen. Sweetser, EveE. 1990 From etymology to pragmatics. Metaphorical and cultural aspects ofsemantic structure. Cambridge: Cambridge University Press.

Viewpoints and polysemy: Linking adversative and causal meanings of discourse markers Scott A. Schwenter

This article examines the use of a connective discourse marker in Spanish (si), which can be used to convey both adversative and causal meaning, depending on the discourse context. The semantic/pragmatic link between the two uses of the connective, which are considered distinct, but overlapping, polysemies, is mediated by viewpoint. The refutational meaning of the adversative polysemy is paralleled by the high-scalar meaning of the causal polysemy. This parallelism results from the transfer of dialogical features from the adversative use, typically found in dialogues, to the causal use, typically found in monologues. The study provides a general framework for analyzing the role of viewpoint in cases of polysemy.

1. Introduction The representation of more than one viewpoint in a single lexeme or construction can be illustrated by a multitude of forms in English or any other language. A prototypical example of how multiple viewpoints or "voices" get encoded into linguistic structure can be seen clearly by looking at one of the "e" relations forming the theme of this book: concession. In an English concessive sentence like the following, (1)

Even though it rained, they played the baseball game.

what is strictly encompassed by the speaker's point of view is clear: the rain did not affect whether the baseball game was played or not. But beyond this, another point of view is evoked which cannot necessarily be ascribed to any particular individual, namely the expectation (based on knowledge of "how the world works") that the rain would have been sufficient cause for the baseball game not to be played. Though not explicitly stated in (1), this expectation, and the point of view from which it arises, is taken into account by the speaker, and grammatically indexed by the concessive ('even though') construction. In this paper, I intend to show that taking multiple viewpoints into account is a useful practice for establishing connections between distinct

258 Scott A. Schwenter

polysemies of individuallexemes. I will focus in particular on adversative (contrastive) and causal polysemies of the Spanish discourse marker si, which derives from the conditional marker si 'if (and is unrelated to to the sf of affirmation meaning 'yes'; indeed, as will be shown below, its meaning is often quite close to 'no'). Although I will be interested primarily in the synchronic connections exhibited by these polysemies, such an analysis is not necessarily restricted to synchrony. In fact, this kind of analysis has been advanced for some time now with regard to the rise and structure of conditional constructions, which are claimed to present in condensed form (i.e. in a complex sentence) a discursive sequence consisting of a (yes-no) question (speaker A), an implied affirmative answer (speaker B), and a stated consequence following from B's answer (speaker A) (Haiman 1978; Jespersen 1940). Although the title of this paper makes explicit the connections to two of the four "C" relations that make up the theme of this volume-contrast (of which I take adversativity to be a subtype; cf Schwenter 1999a: Ch. 4 and below) and causality-the full story of what I am about to present actually brings another of the "C" relations into the picture, namely conditionality. I will focus most of my attention on discourse marker (DM) uses of Spanish si, which, in its most canonical use, is a conditional marker meaning 'if, marking "non-assertion" of the protasis, as in the constructed example (2a). With other functions, however, it is found in assertions with declarative sentence form as an adversative DM (2b), akin to certain uses of 'but', or as a "justificatory" DM or connective (2c-d), akin to certain uses of 'because' (the latter two often, but not always, with exclamatory intonation): (2a) Si traen el pastel, me 10 como. 'If they bring the cake, I'll eat it.' (2b) A: jComete un poco de pastel! 'Eat some cake!' ~ B: jSi ya he comido! 'But I've already eaten!' (2c) Date prisa. jSi se va el autobus! 'Hurry up. (Because) the bus is leaving!' (2d) Juan no viene a la fiesta. Si su madre no 10 deja. 'Juan won't come to the party. (Because) his mother won't let him.'

Viewpoints and polysemy

259

In the constructed examples (2b-d) si does not contribute to the propositional content of the utterance-indeed, it could be removed without affecting this content in the least-but rather signals how the utterance in which it appears relates to the preceding utterance (cf Fraser 1988). In this function, then, si is like many other DMs, which primarily express procedural as opposed to conceptual meaning (Blakemore 1996) as well as a "metatextual" comment on the text being constructed (Traugott 1999). In (neo-)Gricean terms, si in each of these examples conveys a conventional implicature which is non-truth-conditional but also uncancelable; the specific semantic content of this conventional implicature, as well as its pragmatic effects, are what I intend to flesh out in the discussion below. Throughout the paper, I will take a neo-Gricean perspective on implicature, and in particular on scalar implicature (cf Hom 1989, among others). The remainder of this paper is structured as follows. In section 2 I provide an overview of the analytical framework used throughout. Section 3 analyzes uses of si as an adversative DM. Section 4 analyzes uses of si as a causal DM in two different contexts, exhortative and epistemic, and summarizes the polysemies of DM si. Section 5 offers concluding remarks, including ways of extending such an analysis to some English forms. The data to be analyzed in the paper come principally from a 40hour corpus of colloquial conversational Spanish collected by the author in Alicante, Spain, and indicated by [ALC]. The sources of other naturally-occurring examples will be given in the text, and constructed examples will also be used when necessary.

2. Analytical framework Let me begin by stating explicitly my position regarding the concepts of contrast and adversativity. Contrast is not a strictly linguistic notion, but rather a general cognitive ability linked to structures of expectation. I define contrast as "[t]he perception of difference, at some level and by any means possible, between two entities comparable on some other dimension" (Schwenter 1999a: 126). Adversativity, on the other hand, is a strictly linguistic notion. It involves "[c]ontrast between different points of view as these are constructed in language use (Schwenter 1999a: 127, emphasis in original). Contrast can be manifested indirectly in language, for example in a sentence like John is short and he's a good basketball player. Here, the conjunction does not explicitly signal contrast, but hearers familiar with the link between height and basketball-playing ability

260 Scott A. Schwenter

will perceive it nevertheless. Adversativity differs in that it typically provides explicit marking of clashes between different viewpoints, as in John is short but he's a good basketball player, where the conjunction but forces an interpretation in which the speaker asserts (supposed) incompatibility between shortness and basketball-playing ability. Now, although there are a number of ways to conceptualize the interplay between viewpoints in language use and linguistic structure (for some examples, see Ducrot 1984, 1996), in this paper I will utilize a rather basic schema for analysis. Following Roulet (1984), I distinguish between monologues (one physical speaker) and dialogues (two physical speakers), as well as between monological (one "viewpoint") and dialogical (two "viewpoints") discourses. These two sets of distinctions can be merged to yield four different combinations of speaker/viewpoint structure, as displayed in Table 1: Table 1.

Structure of discourse by speakers and "viewpoints"

No. of Speakers One (Monologue) Two (Dialogue)

No. of "Viewpoints" Two (Dialo~ical) (Monolo~ical ) Monologue/ Monologue/ Monological Dialo£ical Dialogue/ Dialogue/ Dialogical Monological One

The distinction between monological and dialogical viewpoints in this paper is closely tied to the argumentative function of utterances in their discourse context, and in particular to whether the viewpoints are construed as argumentatively parallel, i.e. oriented towards the same conclusion, or as argumentatively opposing, i.e. oriented towards different conclusions. To illustrate what is meant by this, consider again the English concessive sentence in (1) above. Here it is possible to isolate both the stated viewpoint that the game was played despite the rain, and the underlying expectation that rain would cause the game to be cancelled. Thus, in this case, a single speaker uttering (1) is simultaneously advancing two viewpoints which are in some sense incompatible. Therefore, in accordance with the schema in Table 1, a concessive sentence is a monologue, yet expresses a dialogical viewpoint. As the discussion below will make clear, the discreteness of the four combinations of speakers and viewpoints illustrated in Table 1 is somewhat illusory. Instead, a gradient perspective on viewpoints appears necessary, since the vestiges of indisputably dialogical interactions remain

Viewpoints and polysemy

261

palpable in meanings/functions that otherwise appear monological in viewpoint. It is possible to determine, for instance, different degrees of "addressee orientation" or, from another perspective, the degree to which the meaning/function of a lexeme is based strictly in the perspective of the speaker. This is not meant to imply, however, that the schema provided in Table 1 is misguided. Indeed, it would have been much more difficult to recognize the gradient nature of viewpoints without this starting point for the analysis.

3. Si as an adversative DM The adversative polysemy of si is typically found in dialogues, and is always dialogical-it signals a confrontation of "incompatible" viewpoints. The utterance introduced by si presents a "retort" (Montolio 1999) to another speaker/viewpoint, and has a "refutational" (Schwenter 1998) pragmatic value, indicating the exclusive relevance of the speaker viewpoint. As a retort, the speaker's viewpoint is presented in the context of another, competing viewpoint (often the immediately preceding utterance of an interlocutor), and via inference excludes the competing viewpoint from further consideration in the discourse. To illustrate this process in action, consider the exchange in (3), taken from the Valencia corpus of colloquial conversational Spanish (Briz [ed.] 1995). Here, a couple is engaged in a rather heated argument, and appear to be on the verge of "breaking up" (§ = "latching" between two speakers; / = .5 second pause): (3) A: § mira yo te quiero// y cre- y creo que /0 SABES/// pero NO/no puedo DEMOSTRARTELOJ 0 sea no no puedo dedicarte todo /0 que tit necesitas 'Look, I love youl/ and I thi- and I think that you KNOW it/II but NO/ I can't DEMONSTRATE IT TO YOU, I mean I can't give you everything that you need. ' ~ B: pero si yo no te pido que me /0 demuestres 'but SI I'm not asking you to demonstrate it to me.' [Briz (ed.): 1995: 83] B's response specifically picks up on A's claim no puedo demostrarte/o and contravenes an assumption inferrable from A's utterance: (A thinks that) B wants A to demonstrate his love for B. In other words, what B is refuting here is not any part of A's utterance per se, but rather his (prag-

262 Scott A. Schwenter

matic) presupposition about what B wants him to do (viz. demonstrate his love for B). The adversative nature of these si-marked utterances is often reinforced by the adversative DM pero, as in (3). However, such reinforcement is not necessary for the adversative meaning to be conveyed, and si is often used on its own to signal refutation of a competing viewpoint. 1 In (4) si is used by M to introduce a refutation of (part of) the propositional content derivable from A's utterance: (4)

[A and M want advice for their next vacation] S~ .ha viajado por toda Europa. 'S has traveled all over Europe. ' M: iSi no ha salido de Espana! 'SI she hasn't been out of Spain!'

A:

~

(ALC) Here, A's position that S has traveled extensively is directly denied by M via negation. Since A's utterance also conversationally implicates that S would be a good person to ask for advice, M's retort functions, via implicature, to refute the validity of this implicit line of reasoning. Semantic contradiction as in (4) is not a necessary feature of simarked utterances. The refutation involved may be based in considerations of relevance (in the Gricean sense), as in the next example, where the rhetorical argumentation introduced by si is presented as the only one relevant to the discourse situation. (5)

~

[R is watching A drink orange juice] R: A, no te bebas el zumo tan rapido, que te va a doler la barriga. 'A, don't drink the juice so quickly, because your stomach is going to hurt. ' A: Si tengo sed. 'SI I'm thirsty.' R: Bueno, pues 10 que te de la gana. 'Fine, then (do) whatever you want.' (ALC 7b.OI15.338)

The argumentative sequence of the first two turns in (5) can be summarized as in (5'), where the parenthesized segments are conclusions left implicit, but easily inferrable from the exchange.

Viewpoints and polysemy

263

~ dolor de barriga (:. no beber rapido) 'drink quickly' ~ 'stomach ache' ( :. 'don't drink quickly') A: tener sed ~ (heber rapido) 'be thirsty' ~ ('drink quickly')

(5') R: beber rapido

It is important to notice that the utterer (=A) of the si-marked turn does not dispute the "truth" of the proposition ("your stomach is going to hurt"), but she does dispute its relevance and the relevance of the associated argumentative sequence within the discourse situation. Her response to R signals the exclusive relevance of the "be thirsty" argument and, as a consequence, justifies the conclusion "drink quickly". Refutation with si may also be directed towards conversational implicatures that go beyond the literal meaning of the proposition, like the one generated by A's utterance in (6): (6)

~

[B (age 6) looks strangely at a glass of a purple beverage] B: i,Que es eso? 'What is that?' A: Es el que te gusta a ti. 'It's the one (flavor) that you like.' M: Si a ella Ie gustan todos los sabores. 'SI she likes all of the flavors. ' (ALC)

The cleft construction in A's utterance leads to an upper-bounded scalar implicature from "it's the flavor that you like" to "it's the only flavor that you like" (cf Atlas and Levinson 1981). M's si-marked utterance contravenes the upper bound implicated by A's utterance. Finally, si can also head an utterance employed to object to more strictly metalinguistic aspects of utterances, such as the pronunciation of a word:

(7)

A: Manana tengo que dar clase, sobre la GENEOLOGiA.

~

'Tomorrow I have to teach a class, about geneology.' R: Si es GENEALOGiA. 'SI it's genealogy.' (ALC)

R's objection in (7) is to the appropriateness of A's pronunciation of the word genealogia 'geneology', not to any propositional content derivable

264 Scott A. Schwenter

from A's utterance. The correction of A's pronunciation by R implicates that A's rendering of the word was inappropriate. As the preceding examples illustrate, si-marked utterances can introduce a wide range of opposition types between two (or more) viewpoints. This should not be surprising, since refutation as a discourse-pragmatic strategy is not limited to denying and correcting the "truth" of prior propositions. Like negation, it is also free to target "oppositions which are created in the pragmatic aspects of utterance interpretation" (Ladusaw 1996: 323; cf Hom 1989), e.g. rhetorical argumentation, conversational implicatures, and judgments of metalinguistic appropriateness.

3.1.

"Hybrid" uses

The bulk of the examples of si presented in 3.1 above had an overwhelmingly adversative flavor, since in each case two speakers were in "opposition" to each other. There are examples of other interactions which on the surface appear to express a similar kind of adversativity, but on further inspection reveal themselves to be subtly-yet crucially-different from most of those seen above. The one exception among the preceding examples is (6), which is a "metapragmatic" (Schwenter 1999a) example objecting to a scalar implicature, but not to the propositional content or to the argumentative orientation of the prior utterance. Examples like (6) are therefore distinct because their interpretation is not so obviously construable in strictly adversative terms, i. e. strictly as a case of opposition between (two) competing viewpoints. Another example of this type is (8):

(8)

M:

~

B:

Creo que te han tintado MAL el pelo. 'I think they dyed your hair POORLY. ' iSi me 10 han tintado FATAL! 'SI they dyed it HORRIBLY!' (ALe)

In (8), M makes a hedged assertion about the dye job done on B's hair, and characterizes it as done mal 'poorly'. B' s response disputes this characterization of the dye job; but instead of characterizing the dye job positively, B strengthens the negative orientation already made explicit by the adverb in M's assertion, stating that the dye job was donefatal 'horribly, lit. fatally,.2 Such an example expresses some degree of adversativity, since there is a clear correction of one speaker's evaluation by another speaker's evaluation. At the same time, because the two speakers actually

Viewpoints and polysemy

265

AGREE on the overall implicated point-the dye job was not a good one-the rhetorical purpose of each utterance in the exchange is parallel. Here there is no ambiguity between a dialogical, adversative interpretation and a monological interpretation with parallel arguments. Rather, both components of meaning combine to produce the overall interpretation of the example. The adversativity expressed by a particular si-marked utterance can be diminished even more when a "correction" of (some aspect of) an interlocutor's utterance cannot be identified:

(9)

~

[Looking at a bicycle for a 4 year-old boy] S: Mira, se puede bajar el asiento tambien. 'Look, you can lower the seat also.' A: Si incluso se puede bajar mas. ,SI you can even lower it more. '

(ALe) In the situational context of this example, S and A are inspecting the bicycle, and S' s utterance is accompanied by the physical act of lowering the bicycle seat a certain amount. A's utterance is in response to the (nonlocutionary) implication that the seat cannot be lowered beyond the point to which S has lowered it. This is where the adversativity of the example lies. Nevertheless, the contributions of both speakers are argumentatively oriented in the same direction, each constituting positive assessments of the bicycle in question, and implicating a conclusion like "we should buy this bicycle" (which, in fact, they did). The co-oriented nature of the utterances is made clear by the appearance of incluso 'even' in A's tum, which marks the expressed proposition as a parallel, but stronger, argument for the aforementioned conclusion (which A took to be implicated by what S said and did) (Kay 1990; Konig 1991). The two preceding examples, along with (6) above, illustrate that, even in dialogues, the degree of opposition marked by si is not fixed, but sensitive to contextual factors, lexical relationships, etc. The "hybrid" nature of each example is due primarily to the parallel argumentative orientation of the asserted propositions: in (8) both the mal proposition and the fatal proposition are directed towards a negative evaluation of B's hair, while in (9) the propositions derivable from both utterances are being used to evaluate the bicycle positively. The greater adversativity expressed in (8) as opposed to (9) results from the explicit correction of a weaker predicate with a stronger one.

266 Scott A. Schwenter

4.

Si as a "causal" DM

4.1.

"Exhortative" contexts

DM uses of si where it functions to signal "connection" between two utterances that form part of a complex discursive act in monologue, such as (2c) and (2d) above, can be divided into two classes. The first is comprised of a paratactic sequence of two utterances, normally separated by period intonation. Specifically, this class consists of an utterance expressing a command or exhortation which is "motivated" by a following utterance prefaced by si. The second class consists of a similar two-utterance sequence linking a speaker's subjective assessment with an argument serving to justify that assessment. The distinction between the two classes corresponds roughly to the now well-known division between "speechact" and "epistemic" causality (Sweetser 1990). Hence from a pragmatic perspective they are distinct and, following Sweetser, I consider them to be pragmatically ambiguous. Both differ from adversative DM si in that they cannot be prefaced by pero 'but'. However, they are also similar in th~t both are typically expressed in paratactic structures. This is the norm for "speech-act" and "epistemic" causals, which assert the content of their clauses. Both types are moreover opposed to "content" ("realworld" or "direct") causals, which presuppose the content of their clauses and tend to display features of subordination (Couper-Kuhlen 1996, inter alia). In this section I examine the first class of uses, where "exhortation" is the central pragmatic act being performed. To the extent that any uses of si like those in (2c) and (2d) above have been considered in the literature, the explanation for them has been assimilated to dialogue-based, adversative uses like those in §3 above. Consider for instance the following naturally-occurring example, from a recent article analyzing the functions of "independent" si: (10) [To the disgusted husband whose favorite soccer team has just lost three goals to none] No te preocupes, jhombrel, si eljUtbol no es importante. 'Don't worry about it, man! SI football [=soccer] isn't important.' (Porroche 1998: 236) In her analysis of this example, Porroche points out that the target of the husband-directed utterance is not an actual utterance produced by the husband, but rather "a belief that ... forms part of the mental universe of the speaker, but which, at times, can also be attributed to the interlocutor"

Viewpoints and polysemy

267

(Porroche 1998: 236~ my translation). The second part of this explanation is, of course, implicitly recognizing the multiple viewpoints discernible in these examples: the wife believes that her husband thinks soccer is (very) important, and attempts to provide a correction to his viewpoint. However, at the same time, the segment introduced by si can also be understood in a causal sense, since it presents a piece of motivation for the negative imperative No te preocupes 'Don't worry', and, moreover, goes on to provide a reason not to worry. The causality involved in examples like (10) above or (12) and (13) below thus concerns the relationship between a command/exhortation and motivating material which provides "just cause" for what the command requests. In this "exhortative" use, there exists a clear overlap between si and discourse connective uses of que (lit. 'that') which perform a similar function (cf Garcia 1996: 49-52)-indeed, que could have been used in place of si in (10) above. Consider the constructed (near) minimal pair in (11 ):3 (IIa) Come, que no te queda casi nada. 'Eat, QUE there's hardly anything left.' (11 b) Come. Si no te queda casi nada. 'Eat. SI there's hardly anything left.' Either of these utterances could be uttered by a parent to a child who is still sitting at the table with food in front of her, long after the rest of the family has finished eating. Both markers preface an asserted proposition that functions pragmatically as an argument in favor of carrying out the action demanded by the imperative come 'eat', and both could be suppressed, i. e. the causal relationship between the imperative come and the declarative sentence that follows is inferrable without any intervening connective element. Pragmatically, the difference between the two utterances is that the second one with si is more "emphatic"~ it would be interpreted as a stronger (more persuasive) attempt on the parent's part to get the child to finish the food than the first utterance with que, which presents the "motivation for eating" in a more neutral way, pointing out to the child the fact that she has little food remaining, and possibly softening the command to eat (cf Pons 1998). Put somewhat differently, the second utterance seems to assume a priori stronger opposition from the child towards eating-it strives to take into account the child's perspective and presents a corrective measure to that perspective. In short, it displays dialogical features. Thus, though the propositional content of the two exam-

268 Scott A. Schwenter

pIes in (11) is equivalent, the version with si offers stronger motivation for eating than the version with que. Unlike most of the dialogue-based examples shown above, example (11 b) would most likely not be interpreted as a reply to an explicit utterance by an interlocutor. As a result, further reinforcement of si by the adversative DMpero, a possibility in all of the adversative examples in §3, would be odd in (11 b), unless the child had previously stated or implied that there was an overabundance of food left on her plate (i. e. as justification for not eating more). The (im)possibility of prefacing si-marked utterances with pero thus constitutes a handy diagnostic for distinguishing adversative uses (usually dialogues) from those with a more causal sense, like the monologue example in (11 b). A naturally-occurring example where eating was precisely at issue can be seen in (12), where A (the mother) produces two utterances which relate back to the command to eat in the first utterance: (12) [Mother to her 18-month-old son, who is sitting at the dinner table] A: jVenga hombre! jCome! 'Come on man! Eat!' A: te quedan dos bocaditos. 'there are two bites left. ' ~ A: Si ya no te queda casi nada. 'SI there's hardly anything left. ' [ACA,3/21/99] Here, both of A's final two utterances can be seen as trying to persuade the child to eat, or more accurately, to finish eating. Both of these utterances can thus be considered "pragmatically subordinate" to the initial tum where the command is uttered; the si-marked utterance being the "stronger" of the two. Indeed, it is interesting to note that if the propositional contents of A's latter two utterances are switched, the sequence will become pragmatically anomalous, as shown in (12'): (12') A: iVenga hombre! iCome! A: Ya no te queda casi nada. ~ A: ?? Si te quedan dos bocaditos. The reason for this is that the sequence introduced by si must be interpretable not only as motivation for carrying out the action demanded by the imperative, but also as the "strongest" piece of motivation among

Viewpoints and polysemy

269

those available. Here, the negative sentence explicitly minimizes the quantity of food left, especially through the use of casi nada 'almost nothing', thereby providing the strongest argument among those put forward in favor of eating. By contrast, the sentence te quedan dos bocaditos 'there are two bites left' presents the quantity in positive terms, leaving strictly to conversational inference the pragmatic connection between it and the imperative come. What this type of example illustrates, then, is that si in this use encodes a component of scalar meaning, in effect imposing a ranking on propositions that are being used as motivation for the action requested. The dialogical nature of the preceding examples, though not linguistically explicit, does appear to be easily inferrable from the discourse context. In each case, one can assume a reluctance or "opposition" on the part of the addressee towards carrying out an action (e.g. eating) or, as in (10), an emotional opposition to "not worrying" about the losing soccer team. Moreover, the si-marked utterances are plainly addressee-directed, since they attempt to motivate or persuade the addressee to carry out the action specified in the command. But one might ask whether it is possible to generalize this analysis to all "exhortative" examples. An example which, at first glance, might not seem analyzable in this way is (13): (13) B: jcono! aqui /e da e/ so/// jmecagiien /a put-!§ 'Damn! the sun is shining here// son of a bi-!' D: § i,ahora te enteras?/ ((estam-)) una mierda ahi 'you just figured that out? «we-)) a piece of shit there' ~ C: mira! a/ Toni/ si /e sobra un pedazo dee- dee- de hierba por ahi/ co/gando 'Look at Toni/51 he's got a piece of plant there/ hanging' D: [dos dos] 'two two' A: [i,a quien?J 'who?' C: en e/ bocadi//o//parece que //evas unn§ 'in the sandwich// it looks like you've got a' A: § jcono! es- son ajos tiernos 'damn! it's- they're green garlic' [Briz (ed.) 1995: 68] Here there is no expressed opposition to the command 'look at Toni' uttered by C. However, this utterance represents an obvious and abrupt

270 Scott A. Schwenter

change in topic, as is evidenced by the exchange between B and D at the beginning of the example. C's goal is to capture the other participants' attention and redirect it towards Toni. It appears, then, that C's si-marked utterance-presenting an argument in favor of looking at Toni-can also be interpreted as attempting to overcome opposition, specifically as a way of persuading the others to heed his command by refocusing their attention on Toni. Given the discussion above of si and que (which could replace si in [13]), C's choice of si to provide linkage between the command and the assertive motivation that follows is not arbitrary: the scalar strength of si lends the propositional content a sense of "noteworthiness" that que would not necessarily convey, and thereby makes a stronger request for the addressees' attention. Because the illocutionary force of the two utterances linked by si in the above examples differs (a command and an assertion related to that command), it is tempting to assimilate them fully to the "speech-act" examples of Sweetser (1990). However there exists an important difference. Sweetser's speech-act examples consist of sentences in which it is the act itself that is motivated or "enabled" by the causal clause, as in What are you doing tonight? Because there's a good movie at the theater. In this sentence, the causal clause provides the motivation for performing the speech act of asking the question-in other words, it provides a "context of relevance" for the question. The "exhortative" examples with si differ: they do not provide motivation for the speech act itself, but rather for the action specified in the command (e.g. looking, eating). Instead of being based in considerations of speech-act relevance, their purpose is fundamentally persuasive in nature.

4.2.

Epistemic contexts

Like the exhortative examples seen in 4.1, in this section we will once again consider discursive sequences consisting of two utterances with the structure [VI + si V2]. Once again, there is a paratactic relationship between the two utterances, and in this relationship the si-marked utterance V2 remains "pragmatically subordinate" to VI. Specifically, VI represents an inferential conclusion that the speaker presents as "warranted" or "justified" by the content of an assertion in V2 (cf Schiffrin 1987: 202). Because they refer to speaker-internal cognitive positions, these examples are labeled "epistemic"; indeed, they appear to be good examples of causality in the epistemic domain, to use Sweetser's (1990) terminology.

Viewpoints and polysemy

271

Two representative examples of si used as a DM in epistemic contexts can be seen in (14) and (15): (14) [talking about a friend who has cellulitis] Q: Se que todas las mujeres la tenemos. 'I know that all women have it. ' Q: Pero ella, es que es una cosa por de mas. 'But her, it's a thing that is too much. ' ~ Q: Si encima Ie han dicho que caminara un poco para reducirla. 'SI they even told her to walk a little to reduce it. ' A: Pero no. 'But she doesn't. ' [ALC] (15) [A has been telling C how she likes the actor Mel Gibson] A: Y a L tambien Ie gusta mucho. ,And L likes him a lot too. ' ~ A: Si incluso Ie grabe yo una pelicula de el. 'SI I even recorded one of his movies for her.' A: i,C6mo se Ilamaba? 'What was it called?' C: i,EI hombre sin rostro? 'The Man without a Face?' A: Si, esa es. 'Yes, that's it.' [ALC] In (14) Q evaluates her friend's cellulitis negatively as "a thing that is too much". This evaluation is then justified by the si-marked utterance that follows. Presumably the underlying reasoning involved is this: given that doctors have told the friend to walk to reduce her cellulitis, it is reasonable to conclude that the cellulitis is "too much". The force of this argument gains strength in the context of Q's initial assertion that "all women have it", but despite that fact not all women require the intervention of doctors. In (15) A's statement that her friend L likes Mel Gibson a lot is justified by A's following assertion that she recorded one of Gibson's movies for L. Both examples show si in combination with scalar additive particles (encima and incluso) which, in these contexts, can both be translated as 'even'. The pragmatic force of such particles is to mark a proposition as a strong-if not the strongest-argument for a contextual

272 Scott A. Schwenter

conclusion (Kay 1990; Konig 1991; Schwenter 1999c), and to imply that other, weaker, arguments for that same conclusion also exist. The combination of si + scalar particle thus signals very strong speaker commitment towards an inferential conclusion, which, as in (14), is often a controversial and/or critical assessment. Notice that, unlike the exhortative examples seen above, epistemic examples like (14) and (15) do not contain explicitly addressee-directed utterances. To the extent that an opposing viewpoint can be identified, or even inferred, it is construed in opposition to the conclusion, not to any aspect of the justificatory utterance introduced by si. As a result, it is in the speaker's best conversational interest to support the conclusion to the greatest degree possible. It is interesting to note that most of the examples of si introducing an assertion that justifies (some aspect of) speaker viewpoint, as in (14) and (15) above, are functionally quite similar to the epistemic use of the causal connective porque 'because'. Indeed, porque could replace si in (14) and (15) without affecting the propositional content in the least, with the proviso that there would typically be a shorter intonation break (comma intonation) between the connected segments. The most salient semantic/pragmatic difference between the two markers is that speakers intuit a tighter "fit" between the content of the two segments with si so that, for instance, (14) would be more likely used in a situation in which the speaker felt that there already was or possibly would be opposition to her opinion ("her cellulitis is too much") stated in the first segment. The use of si instead of porque thus tends to reflect emphatic epistemic commitment to the truth of the conclusion. While it is possible that porque could be used in such a context, the relevant point regarding the epistemic porque construction is that it has a wider domain of application than, and is pragmatically unmarked with respect to (cf Lambrecht 1994: 16-17), the epistemic si construction. Further evidence for this particular analysis comes from the two connectives' distinct co-occurrence possibilities with epistemic expressions like me parece 'It seems to me' or creo 'I think.' These epistemic expressions function to attenuate speaker commitment to the truth of the propositional content of an assertion. While such expressions seem completely natural with porque, they are by contrast pragmatically bizarre with si, as the following examples illustrate (one might imagine these uttered by a speaker A in a context in which another interlocutor B assumes that Juan is coming to their party that same night):

Viewpoints and polysemy

273

(16a) (Me parece que/Creo que) Juan esta en(ermo, porque 10 he visto en el medico. '(It seems to me thatlI think that) Juan is sick, because I saw him at the doctor's office. ' (16b) (*Me parece que/*Creo que) Juan esta enfermo. Si 10 he visto en el medico. '(It seems to me thatII think that) Juan is sick. SI I saw him at the doctor's office. ' Likewise, the well-known epistemic use of Spanish synthetic future morphology to convey probability or inferred certainty (like English That'll be John calling upon hearing the phone ring) is wholly compatible withporque, but not with si: (17a) Juan estara enfermo, porque 10 he visto en el medico. 'Juan must be (lit. will be) sick, because I saw him at the doctor's office. ' (17b) *Juan estara enfermo. Si 10 he visto en el medico. 'Juan must be (lit. will be) sick. SI I saw him at the doctor's office. ' The principal explanation behind these patterns of pragmatic acceptability appears to reside in the strong degree of epistemic commitment that si expresses, not only to the asserted segment that it introduces (whose content is usually not under dispute), but also, and principally, to the justificatory relationship between that segment and another segment that expresses a potentially more controversial speaker position. In this use, si is "modally harmonic" (cf Bybee, Perkins, and Pagliuca 1994),5 i. e. in epistemic concord, with non-attenuated forms like the indicative mood, but is clearly not harmonic with markers of lessened epistemic commitment like those seen in (16) and (17). By contrast, porque carries fewer restrictions with respect to strength of speaker epistemic commitment it can appear felicitously in both epistemically attenuated and non6 attenuated environments. As a causal DM in epistemic contexts, then, si expresses justification of an emphatic sort, as shown by its scalar contrast with porque. The cooccurrence of si with scalar additive particles as in (14) and (15) above is therefore not surprising, as these too are informationally emphatic forms.

274 Scott A. Schwenter

Note however that it would be misleading to ascribe an additive meaning to si, as is made clear by (18): (18) R: "Que vas a hacer con la cortina? 'What are you going to do with the curtain?' A: Pues arreglarla. 'Well fix it.' A: ;,No ves que esta rota? 'Don't you see that it's ripped?' [A takes the curtain and shows the ripped part to R] A: Sera muyfacil. 'It'll be very easy. ' ~ A: Si total, nada mas es cuestion de coserle esta parte de aqui. 'SI after all, it's only a question of sewing up this part here. ' [ALe] Here A's inferential conclusion that fixing the curtain "will be very easy" is warranted by the si-marked assertion that follows. The DM total, which is most plausibly translated in (18) as 'after all', indicates that the following proposition represents the only argument relevant to the conclusion. And nada mas 'only, lit. nothing more' further minimizes the description of the effort involved in fixing the curtain. A's argument for how easy fixing the curtain will be, then, constitutes the only argument for coming to this conclusion, i.e. si appears in a non-additive context. All of these markers thus work in tandem to justify the conclusion, which "follows" from the assertion in A's fmal tum. The exhortative examples in the preceding section included a strategy of lexical minimization (e.g. through negation) in a si-marked utterance as motivation for carrying out the action specified in an imperative-mood command. But as the preceding example shows, such minimization is not limited to exhortative examples-epistemic examples display it also. Another epistemic example that uses minimization in support of a previously stated conclusion is (19). Here, Y is attempting to justify her opinion, stated at the outset, that the man she's talking about doesn't pay attention to anyone: (19) [Y describing a man that she knows] Y: Es que el, es una persona que no Ie hace caso a nadie. 'it's that him, he's a person who doesn't pay attention to anyone.'

Viewpoints and polysemy ~

275

Y: Si una vez fuimos a su casa y se qued6 ahi sentado sin decir nada ... ini se movi6! 'SI once we went to his house and he just sat there without saying anything ... he didn't even move"

[ALe] The negative words in Y's second utterance mark an absolute endpoint: nada 'nothing' and ni 'not even'. They are used in service ofY's conclusion that the man doesn't pay attention to anyone, and run counter to expectations that Y holds regarding another person's behavior when in the presence of visitors in that person's home. In addition, the clause ini se movi6! 'he didn't even move" licenses an inference about other things the man refuses to do when in the company of others, further sanctioning the conclusion "he doesn't pay attention to anyone". The connection between all the examples in this section is that in each case it is strongly implicated that (i) the si-marked justification offered for the conclusion is the strongest available, and (ii) the conclusion simply follows, as a matter of course, from this justification.

4.3.

Summary

Implicit in the organization of this paper has been the position that the adversative DM si and the causal DM si are distinct polysemies of the 7 same lexeme. This is supported on discourse-structural grounds: the adversative DM presents, within a dialogue, a dialogical "retort" directed at an interlocutor's utterance, while the causal DM introduces an assertion which provides motivation/justification for a preceding utterance by the same speaker. By contrast, the exhortative and epistemic uses of the causal DM are not distinct polysemies but rather offer a case of pragmatic ambiguity (Hom 1989; Sweetser 1990): what si is expressing in each use is the same semantic content, the interpretation of which differs due to the pragmatic domain of application. The two polysemies also differ with respect to collocational properties: perhaps most importantly, only the adversative DM can be prefaced, perhaps in every case, by the adversative DM pero 'but', which reinforces the contrast expressed by si (Montolio 1999). Utterance-initial pero is not found with either sub-type 8 of causal DM si. Despite these distinctions, however, it should also be obvious from the foregoing discussion that the boundaries between each use of si as a DM are not clear-cut, and that the strong dialogical nature of the adversative

276 Scott A. Schwenter

examples endures, albeit to a lesser extent, in the causal examples, where it is now partially interpretable as an aspect of scalar meaning. Indeed, one can arrange the uses presented above in relative fashion with respect to the degree to which they are dialogical or monological. This is shown in the table below: Table 2.

Uses ofDM si, by viewpoint Dialogical

Adversative Exhortative Epistemic

Monological

+++ ++

+

+

++

This display helps illustrate what has been stressed above: exhortative and epistemic uses of si do not lack dialogical features completely. The importance of these features is somewhat diminished in these uses, but paradigmatic contrasts with other forms like porque still indicate a dialogical residue: a more salient feature now is high scalar commitment to a conclusion, as the examples in 4.2 illustrate. The same cannot be said for adversative uses of si, which always appear to be exclusively dialogical (even in monologue examples, see Montolio 1999). The greater dialogical nature of exhortative uses of DM si, as compared with the epistemic uses in Table 2, results from the addressee-directed characteristics of the former-most importantly, the assumption of strong opposition on the part of the addressee-which are exhibited to a lesser extent by the epistemic uses.

5. Conclusion An obvious implication of the analysis presented in this paper is that there is an underlying affinity between (some types of) adversativity and (some types of) causality. This affinity seems commonsensical, especially if one considers a pragmatic function/strategy like "justification" to be one that speakers employ when they sense that their positions may potentially be in dispute. But it also shows that certain pragmatic relationships, for instance, refutation, which hold between (strictly) dialogical viewpoints, may have as their (relatively more) monological counterparts other kinds of pragmatic relationship, like justification. Thus, while it is not the case that all types of adversative expression have a counterpart in the domain

Viewpoints and polysemy

277

of causality, the meaning conveyed by Spanish si in the first conceptual domain does have an analogue in the second. The analysis also shows that the distinction between dialogical and monological viewpoints is not absolute, but rather gradient. The meaning associated with adversative DM si, especially its function as a marker of often vehement opposition between viewpoints, is carried over into the strong scalar meaning of causal uses of si, which, I have argued, conserve aspects of the dialogical features characteristic of adversative uses. A strong type of adversativity, namely refutation, is mirrored by the highscalar argumentative strength expressed by causal si, which, indeed, strives to present "irrefutable" justification for speaker-based epistemic conclusions. Put more generally, then, dialogical "footprints" can be found in lexemes and constructions which otherwise appear similar to their more strictly monological counterparts. The gradience revealed here between types of viewpoint parallels the gradience among the conceptual domains which are under investigation in this volume (cf Konig 1986; Kortmann 1997). Though still under-utilized as an analytical device, the framework employed in this paper can and has been extended to the study of other lexemes. For instance, Traugott (1997: 6) argues th~t the historically distinct uses of English after all are linked by "dialogic use, whether in dialog or monolog". In addition, adversative and additive uses of English infact, as in (20) and (21), respectively, are also amenable to this type of analysis: (20) They think he's crazy, (but) in fact he's quite sane. (21) That movie was bad, in fact it was terrible.

These uses, which arise at different times diachronically (Schwenter and Traugott 2000), can be distinguished synchronically by the dialogical viewpoint in the adversative example and the monological viewpoint in the additive example (Schwenter 1999a). Many other differences between the two types of in fact can also be related to this difference in viewpoint: the greater bleaching of the nounfact in (21), the sense of "reformulation" in the same example, and the epistemic scale in (20) versus the "rhetorical" or "argumentative" scale in (21). Finally, even lexemes whose different uses do not appear to constitute separate polysemies convey nuances of meaning which are sensitive to the number of viewpoints involved. This is the case, for instance, of epistemic sentence adverbs like certainly in dialogues like these:

278 Scott A. Schwenter

(22a) A: John isn't smart. B: He certainly is! (22b) A: John sure is smart. B: He certainly is! In both examples the main semantic contribution made by certainly is to signal and reinforce speaker B' s full epistemic commitment to the proposition "John is smart". Because of this semantic parallelism, it does not seem plausible to consider the two uses to be different polysemies. Nonetheless, as a result of the distinct argumentative orientation of B' s utterance with respect to A's in each case-anti-oriented in (22a) vs. cooriented in (22b)-it is feasible to identify certainly as a dialogical pragmatic marker of disagreement in (22a) but as a monological marker of agreement in (22b).

Notes 1. 2.

3. 4.

5. 6.

See Schwenter (1999b) and Montolio (1999) for discussion of the pero si combination. The scalar relationship between the two adverbs in (8) is easily demonstrated. Consider the results when they are inserted into a syntactic frame of the type "X but not Y", where X represents the "weaker" adverb, y the "stronger" one: (ia) Te 10 han tintado mal, pero no fatal. 'They dyed it badly, but not horribly.' (ib) *Te 10 han tintado fatal, pero no mal. 'They dyed it horribly, but not badly.' The infelicity of (ib) results from the fact that any speaker of Spanish would order these adverbs as along a semantic ("Horn") scale, such thatfatal semantically entails mal, but not vice-versa. The intonational break between the imperative and the utterance segment justifying it tends to be longer when si introduces the latter. This is reflected by the comma intonation in (11a) as opposed to the period in (11b). Note too that the translation equivalent of because in Spanish, porque, could not replace si in the examples seen above. In fact, upon translating examples (10) through (13) into English I am unable to replace si with any English connective form. My use of "modally harmonic" with respect to discourse markers extends the term beyond that of the scholars cited, who limit it to cases involving a modal verb and other devices like adverbs. The differences between this connective use of si and epistemic porque are very similar to those separating discourse marker uses of English after all

Viewpoints and polysemy

7.

8.

279

and epistemic because (Traugott 1997). Nevertheless, after all and si in these uses are not always possible translations of each other. In Schwenter (1999b), I have pointed out the extreme similarity between si and the Japanese connective datte, which can be translated as either 'but' or 'because', depending on its context of use (Mori 1994), and which "by itself indicates disagreement" (Ford and Mori 1994: 56). Unfortunately, these authors do not tackle the issue of whether the separate uses of datte constitute distinct polysemies of a single lexeme. Another potential difference not mentioned in this paper is the intonation pattern of each use. According to Montolio (1999), utterances introduced by adversative DM si are characterized by a rise-fall pattern with strong stress on the final tonic syllable. In my data from Alicante, this pattern is found mainly with adversative examples in which si is reinforced by pero, and therefore could be interpreted as an iconic reflection of the speaker's strengthened adversative stance. But whatever the case may be for the adversative DM, it appears that causal OM si does not occur with this unique intonation pattern, though at times increased loudness (as in an exclamatory utterance) can be found.

References Atlas, Jay David, and Stephen C. Levinson 1981 It-clefts, informativeness, and logical form: radical pragmatics (revised standard version). In: Peter Cole (ed.), Radical pragmatics, 161. New York: Academic Press. Blakemore, Diane 1996 Are apposition markers discourse markers? Journal of Linguistics 32: 325-47. Briz, Antonio (ed.) 1995 La conversacion coloquial (Materiales para su estudio). (Cuadernos de Filologia 16.) Valencia: Universitat de Valencia. Bybee, Joan, Revere Perkins, and William Pagliuca 1994 The evolution ofgrammar. Chicago: University of Chicago Press. Couper-Kuhlen, Elizabeth 1996 Intonation and clause combining in discourse: the case of because. Pragmatics 6: 389-426. Ducrot, Oswald 1984 Le dire et Ie dit. Paris: Minuit. 1996 Slovenian lectures/Conferences Slovenes: Argumentative semantics/Semantique argumentative. Ljubljana: ISH. Ford, Cecilia, and Junko Mori 1994 Causal markers in Japanese and English conversations: A crosslinguistic study of interactional grammar. Pragmatics 4: 31-61. Fraser, Bruce 1988 Types of English discourse markers. Acta Linguistica Hungarica 38: 19-33.

280 Scott A. Schwenter Garcia, Serafina 1996 Las expresiones causales y finales. Madrid: ArcolLibros. Haiman, John 1978 Conditionals are topics. Language 54: 564-89. Horn, Laurence R. 1989 A natural history ofnegation. Chicago: University of Chicago Press. Jespersen, Otto 1940 A modern English grammar on historical principles, Volume 5: Syntax. London: George Allen and Unwin. Kay, Paul 1990 Even. Linguistics and Philosophy 13: 59-111. Konig, Ekkehard 1986 Conditionals, concessive conditionals and concessives: Areas of contrast, overlap and neutralization. In: Elizabeth Closs Traugott, Alice ter Meulen, Judy Snitzer Reilly, and Charles A. Ferguson (eds.), On conditionals, 229-46. Cambridge: Cambridge University Press. 1991 The meaning offocus particles: A comparative perspective. London: Routledge. Kortmann, Bernd 1997 Adverbial subordination: A typology and history of adverbial subordinators based on European languages. Berlin: Mouton de Gruyter. Ladusaw, William A. 1996 Negation and polarity items. In: Shalom Lappin (ed.), The handbook ofcontemporary semantic theory, 321-41. Oxford: Blackwell. Lambrecht, Knud 1994 Information structure and sentence form. Cambridge: Cambridge University Press. Montolio, Estrella 1999 iSi nunca he dicho que estuviera enamorada de ell Sobre construcciones independientes introducidas por si con valor replicativo. Oralia 2: 37-69. Mori, Junko 1994 Functions of the connective datte in Japanese conversation. In: Noriko Akatsuka (ed.), Japanese/Korean Linguistics, Vol. 4, 147-63. Stanford: CSLI Publications. Pons Borderia, Salvador 1998 Conexi6n y conectores: estudio de su relaci6n en el registro informal de la lengua. (Cuadernos de Filologia 27.) Valencia: Universitat de Valencia. Porroche Ballesteros, Margarita 1998 Sobre algunos usos de que, si yes que como marcadores discursivos. In: Ma Antonia Martin Zorraquino and Estrella Montolio Duran (eds.), Los marcadores del discurso: Teoria y analisis, 229-42. Madrid: ArcolLibros.

Viewpoints and polysemy

281

Roulet, Eddy 1984 Speech acts, discourse structure, and pragmatic connectives. Journal ofPragmatics 8: 31-47. Schiffrin, Deborah 1987 Discourse markers. Cambridge: Cambridge University Press. Schwenter, Scott A. 1998 From hypothetical to factual and beyond: refutational si-clauses in Spanish conversation. In: Jean-Pierre Koenig (ed.), Discourse and cognition: Bridging the gap, 423-35. Stanford: CSLI Publications. The pragmatics of conditional marking: Implicature, scalarity, and 1999a exclusivity. (Outstanding Dissertations in Linguistics.) New York: Garland. 1999b Discourse markers and the PAISN distinction. Ms., The Ohio State University. 1999c Two types of scalar particles: evidence from Spanish. Advances in Hispanic linguistics. In: Javier Gutierrez-Rexach and Fernando Martinez-Gil (eds.), 546-61. Somerville, MA: Cascadilla Press. Schwenter, Scott A. and Elizabeth Closs Traugott 2000 Invoking scalarity: the development of in fact. Journal ofHistorical Pragmatics 1: 7-25. Sweetser, Eve E. 1990 From etymology to pragmatics. Cambridge: Cambridge University Press. Traugott, Elizabeth Closs 1997 The discourse connective after all: A historical pragmatic account. Paper presented at the 16th International Congress of Linguists, Paris, July. 1999 The rhetoric of counter-expectation in semantic change: A study in subjectification. In: Andreas Blank and Peter Koch (eds.), Historical semantics and cognition, 177-196. Berlin: Mouton de Gruyter.

The treatment of contrasts in interaction * Cecilia E. Ford

Relying on data from natural conversations and emphasizing the on-going analyses afforded by speakers as they respond to one another, this study explores the functions of contrasts in interaction. Considering contrast as broadly including neutral contrast, concession, and antithesis, and recognizing that contrasts can be manifested as disagreements across interlocutors, a recurrent rhetorical pattern is found wherein contrasts are followed by explanations or solutions. Furthermore, cases of contrast deployed without explanation or solution are found to be followed by talk in which interlocutors treat the contrast as requiring further elaboration. Unelaborated contrasts are also found to be used in a variety of ways related to activity and speaker role. The methodology is an adaptation of conversation analysis, with the talk of interactants themselves-specifically talk following contrasts-used as a primary resource for interpreting the functions of the rhetorical patterns under scrutiny.

1. Introduction Documenting the recurrent resources and practices of language use is a primary goal for researchers in both discourse-functional linguistics (for recent reviews see Cumming and Ono 1997; Tomlin et al. 1997; Hopper 1998, among others) and conversation analysis (see Atkinson and Heritage 1984; ten Have and Psathas 1990; Whalen 1992, among others). Understanding linguistic structures and how such structures are combined in use, is fundamental to the study of language development, language change, and language as social action. As part of a functional research program focusing on language in use, the present study examines the realization of a rhetorical relation which is recurrently deployed in spoken interaction: the relation of contrast. I explore how speakersparticipants in interaction-treat contrasts through their talk directly following the textual relation, with attention to how that subsequent talk both builds on and interprets the functions of the contrasts themselves. My attention will thus be on the functions of contrasts as they are elaborated on and responded to by participants in face-to-face interaction. This method of inquiry is an adaptation of the conversation analytic

284 Cecilia E. Ford

practice of relying on the talk of interactants as a primary source for the analysis of functions of talk-in-interaction (see Petrakyla 1997). For the purposes of this study, a contrast consists of two text spans which present an opposition. This type of textual relation subsumes the definitions of contrast, antithesis, and concession in Rhetorical Structure Theory (RST) (Mann and Thompson 1988). Included in my collection are cases where two parts of a text, generally (but not exclusively) clauses, are presented as entities, actions, or propositions which could be: (a) comprehended as the same in many respects (b) comprehended as differing in a few respects (c) compared with respect to one or more of these differences (Mann and Thompson 1988)1 A feature of RST which makes it especially useful for the broad study of rhetorical relations is that it allows for relations to be either explicitly marked or inferable based on the content of text spans in question; thus no explicit linguistic marking is required for the interpretation of two spans as being in an oppositional relation. Furthermore, RST places no fixed limit on the size of text spans to be related, though in practice the lowest unit has often been the clause. Example (1) offers an initial sense of the type of pattern I will be examining. The quote is from a magazine interview, and the speaker is a student who organized a protest at her high school. She is describing the effort it took: (1) (From Ford 1994)

R: [NO: I won't get that for a couple of weeks yet.= A: =Oh, (

A:

=>R:

.)

[W'l

['Cause i t takes a long time.

A's extension, "at least?", displays the pursuit of agreement from R. At the first arrow, R instead produces a negative and disaffiliative response, which is followed, at the second arrow, by a because-prefaced account for this disagreement. In a subsequent study (Ford 1994), following up on the observation that because introduced accounts for disaffiliative responses, I looked at the use of because in different types of talk and writing. I found that if disaffiliation was understood as a form of contrast, then the pattern of

294 Cecilia E. Ford

because-prefaced units introducing accounts could be generalized across text types. While my findings in that study supported the observation that because regularly functioned to introduce accounts for contrasts, I did not look for a broader pattern beyond units connected with that specific causal conjunction. In other words, the 1994 study involved a collection of cases based on linguistic form, the connector because: in a data base of instances of because, I found that a pattern in their immediately preceding textual environments regularly involved contrasts. 6 In the present study, however, I fmd evidence of a more general pattern whereby the contrasts and explanations come in sequenced combinations regardless of the use of any particular formal connector. This pattern of explanation following contrast was common in the more freely interactional portions of the science class. In (8), just before the class officially begins, student J has asked for change for his fifty dollar bill. His intent is to buy some candy from S. S now offers him change in the form of a check she has received from someone else in payment for the candy she is selling (a second-party check): (8)

CONT~T=>

EXPLANATION=>

s: I have a check. S: Eight fifty. J: Nah. I won't take- I don't take secondparty checks. S: eh huh huh huh huh J: I don't got no way of telechecking 'em,

J's refusal (contrast) is followed by an explanation in which he cites his inability to verify whether the check is good ("telechecking" is a process used by stores to get clearance for cashing checks). The tum format includes normative features of disagreeing responses: the declining "Nah" and the account portion (what 1 am calling an explanation), "I don't take second party checks... 1 don't got no way of telechecking 'em." Note also that there are nested contrasts and explanations here, as in (5), above. J's "Nah" is followed by the explanation that he does not take second party checks, and that negation is further followed by the explanation that he has no way of electronically verifying the validity of the checks. As a final example of contrast plus explanation in my data, (9), below, is an instance which supports the claim that explanations after con-

The treatment ofcontrasts in interaction

295

trasts are relevant not only from a discourse analyst's perspective, but also, and in a significant manner, to the interactants themselves. In (9) T presents an assertion which contrasts with a prior assumption (line 8), but she does not immediately provide an explanation that might resolve the asserted incompatibility. Note that both A and R treat the absence of an explanation as problematic and deserving of further attention. A and R have been remarking on the fact that there are many pictures of people and sheep on the walls of T and P's dining room, where they are seated. At lines 4-7, R begins to refer to the people in the pictures as relatives, "grandma" and "grandpa": (9) 1 R: I see there's a lo:t of pictures of sheep. 2 A: Oh yeah. there's more in the other [r(h)oo:m 3 t(h)oo. 4 R: [There 5 there's sheep there, Right by grandma. 6 A: ha ha ha 7 R: Gramma, (.) an' (.) an' (.) and grandpa,and the 8 the sheep and [baby sheep. [We don't know these people. 9 =>T: 10

(. )

11 12 13

T: t>or< these people. R: y~ dotn't?

14

T:

No.

15 16

A:

Do you know the one [just below that one though? [just

17 18 19 20 21 22

(

.) (

?:

.)

(0.5)

A: T: T: A:

the one in the mid[dle? [No.

The little cryin' girl? No? 23=>R: [You just like the >pictures.< 24=>T: [We just likesilver: frames and 25 pictures, 26 R: And what about the sheep.=Do you know the 27 sheep?

296 Cecilia E. Ford

28 29 30 31

A: R: T:

32

R:

Eh(h) [heh heh heh [Ub heh heh heh These have been in our family [for gener(h) at(h) ions [Ha ha ha ha ha

At line 9, T produces a tum which contrasts with R's representation (47) of the people as relatives. Instead of agreeing with R's statements, T asserts that she and her partner P "don't know these people." It is after this contrasting assertion that T fails to provide an explanation. What happens next supports the existence of an expectation, on the parts of these participants, for an explanation to be offered as a resolution to the incompatibility expressed through the contrast. They treat the contrast here as a remarkable announcement needing elaboration. At 12, with marked pitch height, R expresses surprise, with "You do l'n' t?", a form of repetition which regularly functions as a prompt for further elaboration on the part of the prior speaker. But at 14, T, rather than providing an explanation now that her contribution has been treated as noteworthy, simply reasserts her negation. From 16 through 22, A tries to salvage the assumption of T and P's familiarity with the subjects of the pictures: she draws attention to persons in other pictures, asking whether T and P are at least familiar with them. However, T continues to respond with unelaborated, unexplained denial (line 20). It is not until lines 23 and 24 that the sought after explanations are offered, and at that point they are produced by R and T, in overlap. R offers her guess at a possible account, "You just like the pictures," and simultaneously, T gives her explanation, "We just like- silver: frames and pictures." Only after these explanations does the talk move beyond the problematic contrast. What follows is a joking question from R about the sheep; and T joins in the joking by pretending that the sheep have been in her family "for generations." In the clear orientations of the interactants in (9) to something problematic and unresolved after T's contrasting assertion, we find further evidence for a shared expectation that explanation should follow at least certain types of contrasts. In addition to the cases in my data, I have found other instances of CONTRAST + EXPLANATION in published studies of conversation. Notably, cases reported for English in the volume Analysing Causal Conversation (Eggins and Slade 1997) and in a study of tumconstructional units in German by Margret Selting (1998). While space

The treatment ofcontrasts in interaction

297

limitations preclude the inclusion of those cases here, and while the authors do not draw specific attention to a CONTRAST + EXPLANATION rhetorical cluster in their data, I take its occurrence in their examples as suggestive of a more general pattern, one that is likely not exclusive to my data nor to English conversations. In sum, in the present data, we find a rhetorical pattern involving contrasts, in a variety of functions, regularly followed by explanations. As shown here, contrasts may be used pedagogically to highlight puzzles or problems; they may be the vehicles for observing remarkable states of affairs; and, through interactional sequences, they may involve disagreements between speakers. Significantly, explanation is a treated by interactants as normatively relevant subsequent to contrasts in a variety of functions. Whether produced by single speakers or across speakers, and whether prefaced by because or not, speakers in these data regularly treat explanations as relevant after contrasts. In addition to the pattern of contrasts followed by explanations, another set of cases from my data involve contrasts followed by solutions. Examining these cases provides further perspective on sequential contexts in which elaboration is added after a contrast. These cases also support the observation that contrasts are used to present incompatibilities or problems which participants treat as needing resolution.

6. Contrasts + Solutions A further pattern was evident in the data whereby contrasts were followed either by solutions or by displays that solutions, if not actually stated, were understood as relevant or expected. In Rhetorical Structure Theory, a relation of SOLUTIONHOOD requires that one span present a problem and that the "situation presented in [the other text span be] a (partial) solution to the problem" (Mann, Matthiessen, and Thompson, 1992:72). While Mann et al. do not point to the rhetorical relation of CONTRAST as a pattern used in expressing the problem in the solutionhood schema, they do note that the "scope of a problem includes... conditions that carry negative values, either expressly or culturally, including...frustrations" (1992:72). This resonates with Longacre's (1983) notion of frustration in relations between predications. Frustration in a text relation involves an expectancy which is countered. Contrasts appear to be common textual expressions of situations that go against ex-

298 Cecilia E. Ford

pectations. The expression of solutions after contrasts evidences participants' interpretation of contrasts in just this way. Solutions also have commonalities with explanations in such rhetorical contexts. Both types of additions treat contrasts as problematic and needing further talk in the service of resolution. In the case of explanations, resolution is in the form of support or justification. With solutions, the contrast is resolved through the expression· of a potential remedy. In (10), R offers a solution to the problem A has just expressed through a contrast. A and P have been Christmas shopping at a store that carries bird feeders. P is now on the phone in the another room, and A is telling R about their shopping dilemmas: (10)

"she"

=P

A: She's trying to decide if sh[e wants to SPEND= R: [She has a lot of sheep. A: =thirty five dollars on her dad. for Christmas. R: Oh.

A: And she really doesn't.

=>

(0.4)

A: But- (.) the gift that she thought of was thirty five dollars, and she can't find i t anyplace else. R: Mm-mrmn.

A: so, => R: Has she tried Chickadee Depot.

There are actually two contrasts in this example: (1) spending $35 or not, (2) not wanting to spend $35 (for which finding the gift more cheaply would be a solution) but being unable to find the item anywhere else for less. Note the pause after the first contrast, an indication that R is treating A's tum as not yet complete. Given the pattern I am observing, this would be a relevant place for an expression offering resolution, either an explanation or a solution. A normative pattern whereby contrasts are elaborated in such ways could explain R's lack of uptake following this first contrast and during the pause at the first arrow; she may be awaiting an explanation, for example.

The treatment ofcontrasts in interaction

299

The second arrow points to R's response to the second contrast. This contrast is between P's not wanting to spend $35 but not being able to find the birdfeeder elsewhere for less. By asking whether P has tried "Chickadee Depot," a local bird store, R is proposing a potential solution. In suggesting this solution (in the form of a question), R displays her interpretation of the contrast as expressing a problem, for which a search for solution is relevant. In the science class, not surprisingly, there was a great deal of talk focusing on problems and solutions. In the following case, the teacher treats a solution as relevant in the context of a contrast, but he suggests that the students solve the problem on their own. This strategy seems particularly appropriate to one of the pedagogical goals in this discourse context: to teach the students to formulate and solve physics problems. (11)

T: If I stand thi-Oand this is the mirror th- the screeno here, (0.3) I'm at a «pointing to wall in front of ~

CONTRASTS

him»

certain height, right? Look in. If I go back farther, am I smaller? I'm still the same height, right? (

.)

J: Yeah,= ORIENTATION TO SOLUTION

T: =Try (.) If you care to. (.) try this uh (.) problem, ( .) but doing i t wi th uh scale drawing.

The teacher's contrasts here are formulated as questions, and although he does not provide a solution, he does explicitly name what he has presented a "problem," and he suggests that the students may want to solve it. So in addition to being treated as presenting incompatible states of affairs needing explanation, as described in section 5, contrasts can also be treated as presenting problems, after which solutions are relevant. Both solutions and explanations are elaborations which contextualize preceding talk as unfinished, problematic, and needing resolution. The present data provide us, then, with a basis for positing two kinds of rhetorical clusters and for using these clusters as a basis for interpreting the work that contrasts do in talk-in-interaction.

300 Cecilia E. Ford

7. Contrasts without explanations or solutions In a major portion of the cases I have analyzed, speakers either produce or treat as problematically absent the patterns described in sections 5 and 6; that is, contrasts are either followed by explanations or solutions, or speakers treat the lack of such elaboration as notable, through delays or through further talk in pursuit of resolution. However, the collection also includes instances of contrasts after which explanations and solutions are not produced nor are they explicitly treated as relevant. In a small subset of these instances, the contrasts appear to function perfectly well alone, that is, they are in no way treated as unresolved in their contexts. The functions of contrasts which are treated in this way is worthy of further investigation. In the majority of these non-explanation, nonsolution cases, however, there are good sequential and functional reasons for a lack of orientation to resolution. One type of contrast which was left unelaborated involves connectives used as markers of discourse organization, without any literal contrast. Contrastive marking works as a format for text cohesion without a concomitant expression of a problem or other content in need of resolution. Thus, while most of the contrasts in the collection involved some contrastive marking and some real incompatibility at the content level, several cases involve conjunctions serving text organizational functions but not encoding any content level incompatibility. In such cases, contrastive marking structures chunks of discourse, rather than focusing on contrasts in meaning. In (12), the science teacher is involved in the activity of discussing itemized tasks with which the students have already been working. He refers briefly to one item and then moves to another, using contrastive markers as a way of shifting the addressees' attention, without drawing any problematic incompatibility into focus. In introducing the next item in his list, he uses contrastive expressions and marked stress: "Now this one, (.) but this one." In addition to the use of but, we can see a use of now which Schiffrin (1987:234-40) describes as expressing "a speaker's progression through the discourse time of a comparison." (12)

T's first and second uses of this refer to different tasks on which the students are working T: After today this one will be easy. Now this one, (.) but this one >I don't want to take an hour on it. although I was as I said before a member of the junge . 11 UnIon, > the mao bible with me, by chance. only to show in school'

With this complex sentence, Drs asserts two facts (he was a member of the Junge Union and he carried the Mao Bible with him) against the background of the assumption that these two do not normally go together: if you are a member of a conservative party, you do not carry around a Mao Bible. The satellite status of the inserted parenthetical obwohl-clause

444 Susanne Gunthner

is supported by an increase in tempo and thus by the contextualization of a low degree of relevance. 12 As the previous examples have shown, obwohl-constructions with subordinate verb order have the freedom of different types of ordering relative to the main clause: "q, obwohl p", "obwohl p, q", and even inserted obwohl-clauses. Furthermore, concessive obwohl-constructions show additional features of subordinate clauses: they can be within the scope of a question: (4) WOLLT sie's mitnehm, obwohl es noch uberarbeitungsbedurftig isch? 'did she want to take it along although it still needs to be revised?'; and they can be transformed into a prepositional phrase with trotz 'in spite of: Trotz der Tatsache, daft es noch iiberarbeitungsbediirftig ist, wollt sie's mitnehm 'in spite of the fact that it still needs to be revised, she wanted to take it along'.

3. Corrective uses of obwohl In this section we will consider a use of obwohl-constructions, frequently found in informal spoken German, which differs formally and functionally from the concessive use of obwohl discussed in the previous section.

3.1.

The initiation ofself-correction with obwoW

Speakers in colloquial German frequently use obwohl to initiate selfcorrection. 13 In (2) we observed a speaker using obwohl with mainclause syntax to introduce a correction of her former statement: (2) MALDNE ISLANDS 22Mara: ich diit SO GERN mit. (echt) 23 (0.5) 24 obwohl ZEHN stundenflug () desfiind ich doch net so gut. 25Lisa: hm () des macht mirjetzt gar nix [aus.] [i-ich] 26Mara: 27Mara: find- () ich halt (so) lange Fluge echt kaum aus. 28Mara: (finds) gruselig.

'22Mara: I would love to come along. (really) 23 (0.5) 24 although a ten hour flight (.) I wouldn't like that.

From concessive connector to discourse marker

25Lisa: 26Mara: 27Mara: 28Mara:

445

hm (.) 1 don't mind [it.] [1- I] find- (.) 1 can hardly stand (such) long flights. (find it) horrible.'

In contrast to complex sentences with concessive obwohl-clauses,where speakers assert that p and q are both valid but normally do not go together-the speaker (Mara) in (2) first asserts "q" (ich dat SO GERN mit. (echt)); then, after a short pause, she utters obwohl "p" (obwohl ZEHN stunden flug [J des land ich doch net so gut.), thereby correcting the validity of "q".l This type of obwohl-clause contradicts one of the central semantic characteristics of concessives: the entailment of "p" and "q". From the perspective of Rhetorical Structure Theory, too, this use of obwohl can no longer be subsumed under the definition of concessive relations. Contrary to a concessive relation, in which "the speaker acknowledges the apparently contrary information, but then advances the nucleus anyway, showing that s/he does not regard the two as genuinely incompatible" (MannfThompson 1992: 39), the speaker treats the two utterances as incompatible in corrective obwohl-constructions. By uttering the obwohl-clause, the information in the "nucleus" is rectified. The meaning of this corrective use of obwohl is somewhat like "as against what 1 just said", and thus corresponds to what Halliday/Hasan (1976: 252ff) call the "corrective type" among "adversative relations". 15 Syntactic non-integration of the obwohl-clause (in 2) is supported by prosody: both utterances (ich dat SO GERN mit. (echt) and obwohl ZEHN stunden flug () des fand ich doch net so gut.) have their own intonation contours and are separated by a pause. Furthermore, as the example shows, syntactic non-integration provides the option of adding "main clause phenomena" (Green 1976), e.g. left-dislocation phenomena: obwohl ZEHN stunden flug () des fand ich doch net so gut' although a ten hour flight (.) 1 wouldn't like that'. In the following example (6), Mia first suggests (lines 17 and 18) that Bea should ring her doorbell, then she will come down and join Bea for their planned bike tour. But then she changes her mind (lines 21 and 22), using obwohl to introduce a correction of her original proposal:

(6) 17Mia: 18 19Bea:

BIKE TOUR klingel einfach; ich [komm] dann runter. [okay.]

446 Susanne Gunthner 20

21Mia: 22

'17Mia: 18 19Bea:

20 21Mia: 22

(0.5) obwohl () ich komm doch besser bei dir vorbei. des isch glaub =ich gschickter.

just ring the bell; I'll [come] down. [okay.] (0.5) although (.) I'd better come and get you at your place. that's more convenient l=think. '

Also in this case, the two constructions-the obwohl-clause and the preceding one-are produced in two steps. First, Mia suggests: klingel einfach; ich komm dann runter. After a moment of reflection-signalled by the pause-she revises her former proposal and produces a correction: obwohl () ich komm doch besser bei dir vorbei. Again, as in (2), syntactic non-integration is supported by prosody: the two clauses (lines 18 and 21) carry their own intonation contours, and the pause between the two utterances indicates their discontinuity. Corrective obwohl-utterances no longer represent a satellite or a minor act; on the contrary, they achieve the status of a nucleus or major act, rectifying the preceding statement. 16 The syntactic organization (main clause syntax) as well as the prosodic design (prosodic non-integration) iconically reflect the elevation of its status to a major act. Consequently, certain syntactic operations typical of subordinate clauses are no longer possible in the case of corrective uses of obwohl (i. e. they would change the meaning of the construction): *"Obwohl ich doch besser bei dir vorbei komm, komm ich dann runter?"; *"Trotz meines Vorbeikommens komm ich runter". Iconicity is also at work in the linear ordering of corrective obwohl-utterances: the linear organization "q, obwohl p" corresponds to the fact that the obwohl-construction is added after the speaker realizes that the preceding utterance needs to be rectified. Syntactic and prosodic non-integration go hand in hand with the pragmatic independence of the obwohl-utterance and the preceding one: both utterances have their own illocutionary forces. 17 As the following two examples (7) and (8) show, the obwohlconstruction and the preceding utterance can even represent two different sentence types. In (7) Eva has prepared green tea and offers some to Hans:

From concessive connector to discourse marker

447

(7) GREEN TEA 44Eva: willsch mal proBIERe? 45Hans: hm. ich MAG kein griinen tee. 46 (0.5) 47Hans: obwohl GEB mir doch mal ne (-) h ' HALBE tasse vol/. '44Eva: 45Hans: 46 47Hans:

you wanna try some? hm. I don't like green tea. (0.5) although pour me half a cup.'

Hans first rejects the offer and states his dislike for green tea (line 45). After a short pause, he changes his mind and asks for a half a cup (line 47). Here, the obwohl-clause corrects the validity of the preceding speech act (Hans' refusal). Not only do both clauses have their own prosodic contours, they also have their own illocutionary forces. In the next example (8), the obwohl-utterance corrects a preceding question:

66Kati: 67Lars:

PARTY kommt der eigentlich au? (-) [0 obwohl) () s=isch mir eigentlich ega:/. 0 [[glaub schon))

'64Kati: 65 66Kati: 67Lars:

is he coming too? (-) [0 although] (.) 1 don't care really.o [[I assume so)]'

(8) 64Kati: 65

Examples such as (7) and (8) clearly reveal that the obwohl-utterances are no longer within the scope of the illocutionary force of their main clauses (Lehmann 1991: 526). The correction introduced by obwohl may involve various levels of the preceding statement: whereas in the case of (2), (6) and (7) the correction was related to an explicitly stated utterance, the following example (9) shows that obwohl can also be used to initiate the correction of a possible implication. Ulla is talking about a neighbor who has suddenly fallen seriously ill, and comes to the conclusion that one should consider oneself lucky to have enjoyed good health so far.

448 Susanne Gunthner

(9) 36Ulla:

43Rolf: 44Ulla: 45Rolf:

ILLNESS do: kann man bis=jetzt=no=eigentlich (-) Tal. Tal. Tal () no ganz FROH sei. gell? (0.5) OBWOHL man weij3 jo gar net was in oim SCHLUMMERT (1.5) vielleicht sen mir au scho bald mol DO. GWESE. des weij3 mer halt [nie:.] [hajo] do steckt mer halt net drinne.

'36Ulla: 37 38 39 40 41 43Rolf: 44Vlla: 45Rolf:

up to now one can actually (-) knock on wood (.) be glad. right (0.5) although you never know what's slumbering inside of you (1.5) perhaps our time in this world will soon be up one never knows [for sure.] [of course] it's not up to us. '

37

38 39 40

41

In lines 36-37 VIla states that up to now one can be glad, inferring that she has been healthy so far. However, after a short pause, the obwohlclause introduces a sudden change in her perspective: obwohl man weij3 jo gar net was in oim SCHLUMMERT 'although you never know what's slumbering inside of you' and thus corrects the suggested inference. The correction does not affect the factual content of the preceding statement, but a possible conclusion which may be drawn from that utterance. Thus, corrective obwohl-constructions may also operate on the level of possible conclusion, i.e., what Sweetser (1990) calls the "epistemic domain". Syntactic non-integration (do: kann man bis=jetzt=no==eigentlich (-) Tal. TOl. TOl () no ganz FROH sei. gell? (0.5) OBWOHL man weij3jo gar net was in oim SCHLUMMERT.) is again echoed by prosodic discontinuity: the two clauses have their own intonation contours and are separated by a pause of 0.5 seconds. The obwohl is marked by a local increase in volume and a rise in pitch. Corrections using obwohl can also relate to a term used in a preceding utterance. 18 In (10) Elli talks about a relative who plays various instruments:

From concessive connector to discourse marker

449

(10) VIOLA 46Elli: der spielt do- doch au (soviel=[i=weifi)] CELLO. 47Hans: [de karl] 48Hans: alles MDGLICHE [(spielt=der).} 49Elli: [obwohl} (-) des isch glaub=i 50E/li: a BRATSCHE (was er noch spielt.) '46Elli: 47Hans: 48Hans: 49Elli: 50Elli:

he ev- even plays (as=[far=as=i=know)] CELLO. [karl] all sorts of things [(he=plays).] [although](-)I=think it is it's a VIOLA (that he also plays).'

Again, a concessive interpretation is not possible. The obwohl-construction does not presuppose: *if I think it's a viola, then normally he does not playa cello. Instead, Elli corrects herself In the examples presented in this section so far, the scope of the connecting function of obwohl is extended from the clause level (interclausal linking) to the linking of two (relatively) independent utterances. As the following example (11) illustrates, the scope .of obwohl can even exceed the level of linking two utterances to that of linking an utterance with a larger discursive sequence. 19 Anja and Lisa are talking about John, a mutual acquaintance: (11) 56Anja: 57Lisa: 58Anja: 59Lisa: 60Anja: 61Anja: 66 67 68Lisa: 69Anja: 70 71Lisa: 72Anja: 73 74

NEW YORK jaja (so is er halt) ein LANG[WEILER}. [ja ((hi)) ja}. null pep. echt. ja des stimmt. obWOHL du:, () als wir (gero und ich) () damals in new york waren, da da HAT der sich ENGA[GIERT}, [hm} uns die stadtis =ja=seine=heimat[stadt,}= [ja] gezeigt, und (uns) da () rumgefiihrt, war to- () total voll engagement und [WIEF],

450 Susanne Gunthner

75Lisa: 75Anja:

[echt?] ich hab den (hi) echt kaum wieder(hi)erkannt.

'56Anja: 57Lisa: 58Anja: 59Lisa: 60Anja: 61Anja: 66 67 68Lisa: 69Anja: 70 71Lisa: 72Anja: 73 74 75Lisa: 75Anja:

yes yes (that's the way he is) a [BOREl [yeah «hee» yeahl no pep really yeah that's right although you know:, (.) when we (gero and I) (.) were in new york that time, he he really was very [active], [hm] showing us the cityit=s=his=home[town,]= [yes] and he showed (us) around there, was to- (.) totally full of enthusiasm and [involved], [really?] I hardly (hee) recognized (hee) him. '

After Anja has agreed to Lisa's assessment that John is a "LANGWEILER" (bore; line 57) and has "null pep. echt" (no pep really; line 59), she introduces a correction sequence (61) initiated by obwohl. Here, obwohl does not introduce a corrective utterance but a short narrative sequence. This narrative reconstruction refutes the preceding evaluation of John in the sense of "as against what we just said" (Halliday/Hasan 1976: 254). Obwohl assumes the role of a framing device which not only marks the discourse-pragmatic function of the following clause but of the larger narrative sequence as well. Thus, the scope of corrective obwohl can go far beyond the clause level, contrary to adverbial subordinators. Their scope is rather "local" (i.e., it does not exceed their immediate environment).

3.2.

The initiation ofdisagreement with obwohl

Obwohl in spoken interaction is not only used by a speaker to correct his or her own utterance. Second speakers frequently use obwohl to displ~ their disagreement with their co-participant's preceding assessments. 0

From concessive connector to discourse marker

451

Consider (12): Lina and Ira are discussing Ira's plans to defend her doctoral thesis once she will have started her teacher training program: (12) 54Lina: 55Ira: 56

'54Lina: 55Ira: 56

TEACHER TRAINING des referendariat ischja auch ganz schon stressig. ach obwohl (-) d 'eva die sagt immer, in baden-wiirttemberg ist das nichts im vergleich zu bayern. teacher training is quite demanding. ah although (-) eva she always says, the situation in baden-wiirttemberg is nothing compared to that in bavaria. '

After Lina's statement about teacher training, Ira produces a disagreement introduced by the interjection ach and an obwohl: ach obwohl (-) d' eva die sagt immer, in baden-wiirttemberg ist das nichts im vergleich zu bayem 'ah although (-) eva she always says the situation in badenwiirttemberg is nothing compared to that in bavaria'. Thus, obwohl-in combination with ach - is used here as a "disagreement preface" (Pomerantz 1984: 72). Ira then quotes Eva as an authority in order to modify Lina's previous assessment. Due to its syntactic non-integration, the obwohl-clause functions like a non-integrative obwohl-clause uttered by a single speaker: it corrects the validity of the preceding turn. Obwohl is also used in the following example to introduce an upcoming disagreement. Klaus and Hans are discussing the quality of non-alcoholic beer: (13) lKlaus: 2 3

4Hans: 5 , 1Klaus: 2 3 4Hans:

SUMl\1ER HEAT das is echt s 'BESTE BIER. (-) ich mein von den alkoholfreien. (-) hhm. obwohl es gibt schon BESSERE. zum Beispiel BECKS is bei weitem TRINKBARER. this· is really the best beer (-) I mean among the alcohol-free ones (-) hhm. although there are better ones for example Becks is far more drinkable'

452 Susanne Gunthner

After Klaus' assessment concerning the high quality of the beer he is drinking, a short pause occurs, which may already indicate upcoming disagreement. Then, Hans uses a combination of hhm and obwohl to introduce his disagreeing second assessment: obwohl es gibt schon BESSERE. 'although there are better ones' . As we have seen, corrective obwohl is not restricted to self-correction but can also be used as a dissent marker by a second speaker to frame an upcoming disagreement. The obwohl-construction refutes the preceding utterance and provides a partially or totally contradictory second assessment. It reads something like: "As against what you just said...". What is it that makes obwohl such a suitable candidate for a disagreement token? A plausible explanation seems to be that obwohl does not articulate an explicit, "on record" dissent (such as 'no' or 'that is not true'), but-due to its potential concessive character (i.e., its rhetorical indication of a concession)-a weakened and more face-saving form of indicating an upcoming disagreement. 21 In all the examples presented in sections 3.1. and 3.2. obwohl was shown to no longer function as a concessive subordinator, but as a marker indicating an upcoming correction or disagreement. The degree of the correction can vary from a partial correction of the preceding utterance (11) to a total correction which cancels the validity of the preceding statement or speech act (6). It can correct the factual content of the preceding utterance (2), as well as a (performed) communicative act (6), (7) and (8), a possible conclusion or inference which the recipients may draw from the ~receding statement (9), or a term used in the preceding utterance (10).2 Besides self-correction it can also be used to introduce a disagreeing second assessment and thus function as a dissent marker (12) and (13). The various levels on which these corrective uses of obwohl differ from concessive obwohl-clauses are summarized below:

Discourse function:

Concessive obwohl (q, obwohl p)

Corrective obwohl (q, obwohl p)

in spite of seemingly contrary information "q" is presented as valid ("q" is the major act)

change in perspective: the validity of "q" is corrected by "p". ("p" is the major act) dialogical use: disagreement

From concessive connector to discourse marker

453

Formal characteristics: validity of "q"

"q" remains valid

the validity of"q" is cancel1ed~

dialogical use: the validity of "q" is questioned syntactic properties

syntactic integration (final positioning of the finite verb; subordinate clause operations are possible)

syntactic non-integration (second positioning of the finite verb; main clause phenomena may appear)

initial and final

final only

prosodic features

prosodic integration and non-integration are possible

only prosodic non-integration23

pragmatic features

the obwohl-clause is within the scope of the illocutionary force of the main clause

the obwohl-clause and the preceding one have their own illocutionary forces

scope of connection

connection of two clauses

connection of two utterances or larger chunks of discourse

position of the obwohl syntagma

4. Restrictive obwohl as a link between concessive and corrective obwohl So far, the analysis of the use of obwohl in colloquial German suggests that there are two variants of obwohl-constructions which represent strictly separate entities. This is not the case, however. The semantics of the concessive subordinator obwohl constrains the functional variation of the corrective obwohl in various ways, and the two uses of obwohl still share major features. (i) Both variants of obwohl function as cohesive elements (Halliday/Hasan 1976). But whereas the concessive obwohl is used to join two clauses (or smaller units), the corrective obwohl connects utterances or even larger discursive chunks and represents a cohesive element on the discourse level. (ii) Both variants of obwohl depend on a sort of dissonance or incompatibility. However, whereas in the case of

454 Susanne Gunthner

concessive obwohl the speaker asserts "two facts against the background of an assumption that situations of these types are incompatible and do not normally go together" (Konig 1994; and Konig in this volume), in the case of corrective obwoh/ the speaker asserts a dissonance or incompatibility between her/his upcoming utterance and her/his preceding one. The element of dissonance is transferred to the discourse level. Thus, the semantics of concessive obwoh/ is still partly inherent in the corrective use of obwohl: it also expresses a dissonance or incompatibility. It loses the conceding element, however. Thus, there is a loss in the semantic complexity of the source item. These two uses of obwoh/ (as a concessive subordinator and a corrective marker) can be seen as two ends of a scale, which has intermediate gradations. One such intermediate option is the restrictive use of obwohl. This use does not share the typical characteristics of a concessive relation (it does not presuppose "if p, normally"" q"), nor does it cancel the validity of the preceding utterance (as the corrective obwohl does). Instead, it restricts the preceding statement or a conclusion which can be drawn from the preceding utterance. 24 Restrictive uses of obwohl show verb-final as well as verb-second position. The following transcript shows a restrictive obwoh/-construction with final ordering of the finite verb. Elli and Fritz are talking about Greta and her husband Hans, who "adores" her: (14) 34Elli: 35Fritz: 36E//i:

FRIEDA'S BIRTHDAY der der hot doch vie:: Iflir die [gmacht.] [hajo] obwoh/==s scho au grenze flir==en gebe hot.

'34Elli: 35Fritz: 36Elli:

he he's done a lot for [her.] [right] although he has reached his limits at times. '

The obwohl-clause does not express a general dissonance or a factual conflict between the two statements: der der hot doch vie/ flir die [gmacht.] 'he he's done a lot for her.' and obwoh/==s scho au grenze {fiir==en] gebe hot. 'although he has reached his limits at times'. Instead it prevents a possible conclusion that could be drawn from the first statement: 25 that Hans has done just everything for her, without any limits; i.e., a possible inference is restricted. The traditional definition of a concessive relation "if p, then normally"" q" cannot be applied here: *"if someone has his limits, then normally he will not do a lot for someone

From concessive connector to discourse marker

455

else". However, as with concessive relations, both "p" and "q" are entailed and treated as valid; i.e., "p" (obwohl==s scho au grenze {fiir==en] gebe hot., 'although he has reached his limits at times' as well as "q" (der der hot doch viel fiir die [gmacht.], 'he he's done a lot for her.' are logically true. In our example (14), syntactic integration (via verb-final positioning in the obwohl-construction) marks a close connection between the two clauses. The prosodic realization, however, indicates that the two clauses are produced as distinct units: the utterance in line 34 carries a fmal intonation contour which marks completeness. Syntactic non-integration (via verb-second positioning in the obwohl-construction) would be possible without any change in meaning: der der hot doch vie:: 1fiir die gmacht. obwohl s==hot scho au grenze fiir==en gebe. And indeed, in colloquial German, speakers also use main clause syntax with this kind of restrictive obwohl-construction. In (15), Karl is talking about the composer Penderecki and his musical oeuvre. In line 4 he produces an obwohl-construction in order to restrict the validity of his preceding statement: (15) lKarl: 2

3 4 5

, 1Karl: 2 3

4

5

PENDERECKI nein==nein () der is (sicher) nicht hi SCHRILL. (-) was sicherlich der fall ist,

seine Musik ist zeitgenossisch aber nicht dogmatisch. no=no (.) he (surely) isn't hee jarring. (-) which surely is the case, «all> although I am not familiar with his complete works,> his music is contemporary but not dogmatic. '

With the obwohl-utterance Karl restricts the validity of his former utterance was sicherlich der fall ist 'which surely is the case': he admits that he is not familiar with Penderecki's complete works. However, the obwohl-construction here again does not yield the traditional presupposition for concessive relations; i.e., "if p', then normally -, q'" cannot be applied: *"ifI am not familiar with his complete works, then normally it is surely not the case". The restriction of a preceding statement or a possible conclusion from a preceding statement has the character of an afterthought. Consequently,

456 Susanne Gunthner

it is not surprising that the linear order is such that the obwohl-clause is generally post-posed: having stated the "matrix", the speaker sees the necessity of restricting the validity of this statement or of a possible conclusion which may be drawn from it. One should add, however, that the distinctions between the different uses of obwohl are not clear-cut. Sometimes it remains unclear whether an obwohl-utterance totally cancels or only restricts a preceding utterance. Thus, we arrive at a continuum ranging from the concessive subordinator obwohl via the restrictive use to obwohl as a correction marker. Table 1.

Synchronic variants of obwohl in colloquial German

Concessive subordinator Restrictive obwohl q obwohl p q, obwohl p p and q are valid q or a conclusion from q which can be inferred is restricted verb final position prosodic integration and non-inte~ration

initial and final positioning of the obwohl-clause

verb final or verb second position prosodic nonintegration final positioning of the obwohl-clause

Corrective obwohl q, obwohl p correction of q (or of a conclusion from q)~ dialogical use: the validity of q is questioned verb second position prosodic non-integration final positioning of the obwohl-clause

Due to its possible main clause syntax (and resulting main clause phenomena), its prosodic independence and its loss of freedom in the linear order, obwohl in its restrictive use can no longer be classified as a subordinate conjunction. 26 Furthermore, due to the fact that restrictive obwohl-constructions no longer presuppose "If p, then normally --, q", they no longer represent concessive relations in the traditional sense. The corrective use of obwohl is even further away from its original use as a concessive subordinator: it does not show features of subordination nor does it possess the central semantic characteristic of concessives; i.e. the entailment of "p" and "q". Finally, its scope is extended to the level of connecting utterances and units of discourse. Given that corrective obwohl has lost syntactic, semantic, pragmatic and discourse-functional properties which are characteristic for concessive subordinators,27 the question arises: has the concessive

From concessive connector to discourse marker

457

subordinator obwohl been reanalyzed as a discourse marker in colloquial German?

5. Corrective obwohl as a discourse marker? In his chapter on discourse deixis, Levinson (1983: 87£) refers to those words which are used to indicate "the relationship between an utterance and the prior discourse": ...there are many words and phrases in English, and no doubt most languages, that indicate the relationship between an utterance and the prior discourse. Examples are utterance-initial usages of but, therefore, in conclusion, to the contrary, still, however, anyway, well, besides, actually, all in all, so, after all, and so on. It is generally conceded that such words have at least a component of meaning that resists truth-conditional treatment. .. What they seem to do is indicate, often in very complex ways, just how the utterance that contains them is a response to, or a continuation of, some portion of the prior discourse.

Levinson's description comes close to what Fraser (1996: 186) defmes as "discourse markers": ... an expression which signals the relationship of the basic message to the foregoing discourse. [...] They (Le. "discourse markers"; S.G.) provide instructions to the addressee on how the utterance to which the discourse marker is attached is to be interpreted.

Although definitions vary a great deal, there are a number of characteristics which most studies on discourse markers identify:28 (i) discourse markers are predominantly a feature of oral discourse; (ii) they are drawn primarily from the syntactic classes of conjunctions, adverbials, and prepositional phrases; (iii) because of their frequency in oral language, discourse markers are stylistically stigmatized and negatively evaluated (especially in written or formal discourse); (iv) they are usually "short items"; (v) they appear in utterance initial position; (vi) they occur either outside the syntactic structure or loosely attached to it and hence have no clear grammatical function; (vii) they are optional; (viii) they are marginal grammatical categories;

458 Susanne Ganthner

(ix)

they. are multifunctional, operating simultaneously on the local (semantic, syntactic etc.) and global (pragmatic) level.

Based on these characteristics, corrective obwohl appears to be a straightforward instance of a discourse marker: it is a feature of oral discourse and is stylistically stigmatized as "ungrammatical", it is a "short item", which is only loosely attached to the syntactic structure and hence has no clear grammatical functions, it is optional, it mainly functions as a pragmatic and discourse grammatical category and is multifunctional at the same time. One could even argue that it occurs initially, as the corrective utterance is independent from the preceding utterance, and obwohl cannot be moved from its utterance-initial position to a position within or at the end of the utterance. 29 Since corrective obwohl indicates a relationship of dissonance between an utterance (or a longer discursive chunk) and the prior discourse, it can be assigned to the sub-group of "corrective markers" (Holker 1990; GulichIKotschi 1995) or "contrastive discourse markers" (Fraser 1998). However, contrary to Fraser's (1990) claim that the pragmatic meaning or function of discourse markers is clearly separate from the semantics of their "homophonous forms" (1990: 389), I maintain that the meaning and function of the discourse marker obwohl is still related to its traditional meaning/function as a concessive subjunctor: corrective obwohl still carries the meaning potential of dissonance and incompatibility, even though it has lost its conceding element. As for the different variants of obwohl in colloquial German, we can now suggest the following cline, which from a synchronic perspective can be thought of as a continuum, on which the forms available are systematically arranged (Hopperffraugott 1993: 6£): subordinate conjunction> (coordinate conjunction» discourse marker. 30 This cline is to be seen as a gradual process with overlaps: the different variants co-exist and the traditional concessive meaning of obwohl constrains the meaning/function of the later variants. At the same time, this cline-which reveals features of semantic and syntactic reclassification and an increase in pragmatic function-contradicts traditional assumptions of grammaticalization, e.g. the unilinear view of 31 clause combining from loose paratactic to tight hypotactic constructions.

From concessive connector to discourse marker

459

6. Uses of concessive conjunctions as corrective markers in other languages The different variants of obwohl, ranging from a concessive subordinator to a discourse marker, raise the question of whether there are similar reinterpretations of concessive subordinators as corrective discourse markers in other languages. Is there a more general tendency in spoken interaction, namely that of a development from subordinate conjunction (via coordinate conjunction) to discourse marker? Can although, for example, also be used as a discourse marker indicating a correction in colloquial English? And do we fmd similar developments in nonEuropean languages? In discussing the development of although-constructions from less to more paratactic, Hopperffraugott (1993: 184) refer to a similar case in English. They treat the paratactic use of although, which is apparently frequently used by college students, as a "counterexample" to the unidirectionality in clause combining and claim that this use of although (in the sense of however) is a "hypercorrection, resulting from literacy and learned punctuation, and therefore a temporary innovation". In his analysis of "contrastive discourse markers", Fraser (1998) also refers to although in the use of however. "Contrastive discourse markers'·' are used to signal a "contrastive relationship between the S2 they introduce and a foregoing S2, although in some cases they signal more than simple contrast"; i.e., they "signal that the speaker intends the explicit message conveyed by S2 to contrast with an explicit or indirect message conveyed by SI" (Fraser 1998: 303ft). According to Fraser (1998: 313), although in the sense of however "can have as its target the direct SI message (... ) and an indirect message conveyed by S 1". As Barth's study (in this volume) shows, speakers of colloquial English (not only college students) make use of although in order to introduce a partial correction of previous utterances. However, in contrast to the German corrective marker obwohl, the correction introduced by although does not reach the extent of a total annulation of the validity of the preceding statement. In discussing the different uses of the Spanish concessive conjunction aunque, Crevels (1998) adds a "textual level of linking" to Sweetser's three domains (content, epistemic and speech act), and states that in this textual use, aunque takes over the function of a "rectifying concessive" (Crevels 1998: 140). It signals an "unexpected tum in the discourse context, a possible paraphrase being 'if I conclude all of a sudden that (... ), then normally I would not previously have said that (... )" (Crevels

460 Susanne Gunthner

1998: 134). The examples Crevels provides show that in its textual use, Spanish aunque comes close to the restrictive obwohl: "the concessive clause is not meant as a plain comment on the felicity of the preceding assertion, but rather modifies the preceding text unit by excluding certain interpretation options which have become available to the hearer in the course of it. (... ) Rectifying concessive clauses are even more loosely linked to a main clause: They can only follow the main clause and typically exhibit main clause order in languages like German and Dutch" (Crevels 1998: 141). Onodera's (1995) analysis of the development of the connectors demo and dakedo in Japanese illustrates a similar process from concessive conjunctions to discourse markers. Demo and dakedo develop from clause-final concessive subordinators, indicating referential contrast to clause-initial discourse markers used to signal contrastive actions. Parallel to this semantic, syntactic and pragmatic change "the scope of the connecting function is enlarged" from the sentence to the discourse level, and the pragmatic functions are expanded (Onodera 1995: 422ff). These studies show that the use of a traditionally concessive subordinator as a contrastive discourse marker is not restricted to German, but that various languages show similar developments. 32 Still, there are differences concerning the extent of the contrast: in English, paratactic although may be used to restrict a preceding utterance, but "q" is still treated as valid. In German and Japanese, however, those contrastive markers originating from concessive connectors (obwohl, demo and dakedo) can be used to cancel a preceding utterance or communicative action.

7. Conclusion In this paper I have discussed obwohl-constructions in contemporary spoken German. Speakers not only use obwohl as a subordinator expressing concessive relations, they also use obwohl-constructions to restrict and cancel the validity of prior discourse. Thus, speakers reanalyze this traditional subordinator as a discourse marker indicating an upcoming correction or disagreement. These shifts in discursive function involve shifts in the grammatical features of obwohl-constructions. However, the different uses of obwohl are not arbitrary or disparate; they are related to a certain potential inherent in obwohl, namely the display of an incompatibility or dissonance. This dissonance can be between the two

From concessive connector to discourse marker

461

statements expressed, between a statement and a possible conclusion, or between two actions. Furthermore, the ongoing development of obwohl from a concessive subordinator to a corrective marker, and thus from an element of sentence grammar to an element of discourse grammar, is in accordance with tendencies attested for various aspects of the grammar of spoken discourse; i. e. functional shifts of linguistic elements from the sentential to the textual and discursive level. 33

Appendix: Transcription conventions (The transcription conventions follow GAT "Gesprachsanalytisches Transkriptionssystem", Selting et al. 1998): [ja das] finde ich [du ab]

conversational overlap

( -)

short pauses of less than 0.5 sec. pauses of 0.5 sec. and longer unintelligible uncertain transcription latching intonation phrase-final: rising intonation phrase-final: slightly rising intonation phrase-final: falling intonation phrase-final: slightly falling global high pitch global low pitch lengthening soft voice loud voice laugh particles within the utterance laughter nonlexical phenomena (e.g. coughing)

(0.5)

( ) (gestern) ?

, word word> a: °no o NEIN mo«hi»mentan hahaha «hustet»

Notes

*

1. 2.

This study is closely connected to the project "Adverbial clause relations in spoken English and German" in the Sonderforschungsbereich 471 at the University of Konstanz. I would like to thank Peter Auer, Elizabeth CouperKuhlen, Dagmar Barth, Christine Gohl, Verena Haser, Bernd Kortmann and Manfred Krug for their comments on an earlier version of this paper. Cf. also Giinthner (1996~ 1999). Cf. Eisenberg (1993).

462 Susanne Gunthner

3. 4. 5. 6. 7 8. 9.

Cf. Kortmann (1998) on characteristics of adverbial subordinators, and cf. Hengeveld (1998) on adverbial clauses in various European languages. In these two examples a word-by-word gloss will be given in addition to the English translations. Cf. see also Metrich (1980), Gaumann (1983), Kiiper (1991), Giinthner (1996; 1999). Due to its main clause order, left dislocation is possible in this obwohlsentence. Cf. also Gtinthner (1999). In general, definitions of concessive relations operate with terms such as "dissonance", "incompatibility", "contrast", "unexpectedness", "surprise", etc. Cf. Di Meola (1997) and Crevels (1998). I will use "p" and "q" to refer to concessive relations: "p" refers to the obwohl-clause, "q" to the matrix clause (i.e. "obwohl p, q" and "q, obwohl p").

10. Initial obwohl-clauses are rarely used in my data. Among the 28 uses of concessive obwohl in my data there are only two cases of obwohl in initial position. . 11. The "Junge Union" (young union) is the youth organization of the conservative party (the Christian Democrats) in Germany. 12. Cf. Gumperz (1982) on the use of prosody to contextualize meaning in interaction; cf. Uhmann (1992) on speech rate and the contextualization of relevance. 13. For the conversational organization of repair and correction sequences cf. ScheglofIlJeffersonlSacks (1977) and Jefferson (1987). 14. Syntactic integration would suggest a different meaning (i.e. a concessive reading): ich dat so gern mit, obwohl ich zehn Stunden Flug nicht so gut fand. If we include the modal particle doch, which is used in the original version, the concessive meaning is less obvious. 15. As corrective obwohl-constructions lack the acknowledgment that an assertion holds, despite its potential incompatibility, they no longer represent concessives but mark contrast relations (cf. Couper-KuhlenlThompson, this volume). 16. Cf. Grote/Lenke/Stede (1997: 92) for "major" (nucleus) and "minor acts" (satellite) in concessions. 17. Cf. Givan (1990: 826) for iconicity and syntactic, semantic and pragmatic independence. 18. Cf. also Gaumann (1983). 19. Cf. Giinthner (1999). 20. Cf. Pomerantz (1984) for the interactive organization of disagreement. Cf. Giinthner (1996; 1999) for obwohl as a disagreement marker. 21. Cf. Brown/Levinson (1978) for face-saving strategies. 22. Thus, as with the concessive use of obwohl, the corrective use of obwohl operates on different levels of discourse: (i) it can link propositions ("content domain"); (ii) it can link aspects of knowledge; i.e. a statement and a possible conclusion ("epistemic domain"); and (iii) it can link speech acts ("speech act domain"); cf. Sweetser (1990) and Crevels (1998).

From concessive connector to discourse marker

463

23. Corrective obwohl is sometimes used in those written genres which reproduce dialogical, colloquial language (e.g. reconstructions of interviews, advertisement dialogues, private letters, e-mail communication, chat groups, etc.). In these cases, the discontinuity is marked by a comma or period before the obwohl-clause. Furthermore, the obwohl is sometimes separated from the following correction by a colon or dash. The following text segment is taken from an interview in a weekly newspaper with a German model. The interviewer asks her whether the prejudice that models have no brains hurts her: Q: So ein Urteil tut weh? A: Ach, es start mich nicht, obwohl: Es argert mich schon ein biJ3chen. 'Q: Does such a judgement hurt? A: Ah, it doesn't bother me, although: it actually does make me a bit angry.' (WOCHENPOST, 6 April 1995). 24. Lerch (1929: 334) refers to the special case of "restrictions" among concessive relations, and has treated them as an intermediate stage between adversativity and concessivity. On restrictive uses of concessive conjunctions also cf. Rudolph (1996: 411), who states: "The state of affairs affirmed in the main clause is restricted in its validity by the state of affairs affirmed in the concessive clause. The procedure of restricting serves to make clearer what was meant and offers the opportunity of avoiding misunderstandings." Cf. also Konig (1994: 681) on "rectifying" concessive clauses. 25. This use of obwohl is related to Sweetser's (1990: 79) "epistemic domain". Furthermore, it comes close to what Halliday/Hasan (1976: 252f.) describe as the "internal use of adversative relations". The meaning is something like "as against what the current state of the communication process would lead us to expect, the fact of the matter is..." (Halliday/Hasan 1976: 254). 26. The fact that restrictive obwohl-constructions also show main clause syntax and thus reveal typical characteristics of coordinate constructions contradicts Rudolph's (1996: 415) assumption that "the restrictive type is of exclusively semantic characteristics", and can thus be considered "as a subordinate clause". However, as Rudolph's data is limited to written material, it is not surprising that she does not find restrictive uses of obwohl with verb-second position. Cf. also Konig (1994: 681) on "main clause order" in (what he calls) "rectifying concessive clauses". 27. Cf. Kortmann (1998: 458-460) on characteristics of adverbial subordinators. 28. Cf. Schiffrin (1987: 34ff.)~ Brinton (1996: 29ff.), Fraser (1998: 301), Jucker/Ziv (1998: 1), Giinthner (1999). 29. It occupies the topological position of the German Vor-Vorfeld (pre-frontfield)~ cf. Auer (1996). 30. Cf. Kortmann (1998) on characteristics of subordinators and coordinators. Cf. also Traugott (1995a) and Tabor/Traugott (1998) on the grammaticalization of discourse markers. 31. Cf. Giinthner (1999) and Gohl/Giinthner (1999) who discuss present uses of obwohl and weil as discourse markers and thus "the exploitation of old forms for novel functions" within the theory of grammaticalization.

464 Susanne Giinthner

32. Cf. also Konig's (1994: 681) discussion of rectifying concessive clauses. He mentions the use of French encore que for this subtype of concessive constructions. Even in Latin the concessive connectors quamquam and etsi (although) were sometimes used to restrict or correct the preceding sentences. Cf. Kuhner/Stegmann's (1962) Latin grammar. 33. Cf. Traugott (1995a~ b), Auer (1996), Gtinthner (1999) and Gohl/Gtinthner (1999).

References Auer, Peter 1996

The pre-front field position in spoken German and its relevance as a grammaticalization position. Pragmatics 6 (3): 295-322. Brinton, Laurel J. 1996 Pragmatic Markers in English. BerlinlNew York: de Gruyter. Brown, Penelope and Stephen C. Levinson 1978 Universals in language usage: Politeness phenomena. In: Esther Goody (ed.), Questions and Politeness, 56-311. Cambridge: Cambridge University Press. Crevels, Mily 1998 Concession in Spanish. In: Hannay, Mike and A. Machelt Bolkestein (eds.), Functional Grammar and Verbal Interaction, 129-148. Amsterdam: Benjamins. Di Meola, Claudio 1997 Der Ausdruck der Konzessivittit in der Deutschen Gegenwartssprache. Tubingen: Niemeyer. Eisenberg, Peter 1993 Der Kausalsatz ist Dicht zu retten. Praxis Deutsch 118: 10-11. Eroms, Hans-Werner 1980 Funktionskonstanz und Systemstabilisierung bei den begrundenden Konjunktionen im Deutschen. Sprachwissenschaft 5: 73 -115. Fraser, Bruce 1990 An approach to discourse markers. Journal of Pragmatics 14: 383395. 1996 Pragmatic markers. Pragmatics 6 (2): 167-190. 1998 Contrastive· discourse markers in English. In: Andreas Jucker and Yael Ziv (eds.), Discourse Markers. Descriptions and Theory, 301326. AmsterdamIPhiladelphia: Benjamins. Gaumann, Ulrike 1983 'Weil die machen jetzt bald zu '. Angabe- und Junktivsatz in der Deutschen Gegenwartssprache. Goppingen: Kummerle. Giv6n, Talmy 1990 Syntax. A Functional-Typological Introduction. Amsterdam: Benjamins.

From concessive connector to discourse marker

465

Gohl, Christine and Susanne Giinthner 1999 Grammatikalisierung von weil als Diskursmarker in der Gesprochenen Sprache. Zeitschrift fur Sprachwissenschaft 18: 3975. Green, Georgia 1976 Main clause phenomena in subordinate clauses. Language 52: 382397. Grote, Brigitte, Nils Lenke and Manfred Stede 1997 Ma(r)king concessions in English and German. Discourse Processes 24: 87-117. Giilich, Elisabeth and Thomas Kotschi 1995 Discourse production in oral communication. A study based on French. In: Uta Quasthoff (ed.), Aspects of Oral Communication, 30-66. Berlin: de Gruyter. Gumperz, John J. 1982 Discourse strategies. Cambridge: Cambridge University Press. Giinthner, Susanne 1993 ' ...weil-man kann es ja wissenschaftlich untersuchen'Diskurspragmatische Aspekte der Wortstellung in WEIL-Satzen. Linguistische Berichte 143: 37-59. 1996 From subordination to coordination? Verb-second position in German causal and concessive constructions. Pragmatics 6 (3): 323370. 1999 Entwickelt sich der Konzessivkonnektor obwohl zum Diskursmarker? Grammatikalisierungstendenzen im Gesprochenen Deutsch. Linguistische Berichte 180: 409-446. Halliday, Michael A. K. and Ruqaiya Hasan 1976 Cohesion in English. London: Longman. Hengeveld, Kees 1998 Adverbial clauses in the languages of Europe. In: lohan van der Auwera (ed.), Adverbial Constructions in the Languages ofEurope, 335-420. Berlin/New York: Mouton de Gruyter. Holker, Klaus 1990 Franzosisch: Partikelforschung. Lexikon der Romanischen Linguistik V (1): 77-88. Hopper, Paul and Elizabeth C. Traugott 1993 Grammaticalization. Cambridge: Cambridge University Press. Jefferson, Gail 1987 On exposed and embedded correction in conversation. In: Graham Button and 1. R. E. Lee (eds.), Talk and Social Organization, 86100. Clevedon: Multilingual. Jucker, Andreas and Yael Ziv 1998 Discourse markers: Introduction. In: Andreas Jucker and Yael Ziv (eds.), Discourse Markers. Descriptions and Theory, 1-12. AmsterdamIPhiladelphia: Benjamins.

466 Susanne Gilnthner

Keller, Rudi 1993 Das epistemische WElL. Bedeutungswandel einer Konjunktion. H. J. Heringer and G. Stotzel (eds.), Sprachgeschichte und Sprachkritik, 219-247. BerlinINew York: de Gruyter. Konig, Ekkehard 1991 Konzessive Konjunktionen. In: Arnim von Stechow and Dieter Wunderlich (eds.), Semantik. Ein internationales Handbuch der Zeitgenossischen Forschung, 631-639. Berlin: de Gruyter. 1994 Concessive Clauses. In: R. E. Asher (ed.), The Encyclopedia of Language and Linguistics, 679-681. Oxford: Pergamon. Konig, Ekkehard and Jan Van der Auwera 1988 Clause integration in German and Dutch conditionals, concessive conditionals and concessives. In: John Haiman and Sandra A. Thompson (eds.), Clause Combining in Grammar and Discourse, 101-133. Amsterdam: Benjamins. Kortmann, Bernd 1998 Adverbial subordinators in the languages of Europe. In: Johan van der Auwera (ed.), Adverbial Constructions in the Languages of Europe, 457- 562. BerlinlNew York: Mouton de Gruyter. Kuhner, Raphael and Stegmann Carl 1962 Ausfuhrliche Grammatik der Lateinischen Sprache. Darmstadt: Wissenschaftliche Buchgesellschaft. Kuper, Christoph 1991 Geht die Nebensatzstellung im Deutschen verloren? Zur pragmatischen Funktion der Wortstellung in Haupt- und Nebensatzen. Deutsche Sprache 19: 133-158. Lehmann, Christian 1991 Grammaticalization and related changes in contemporary German. In: Elizabeth C. Traugott and Bernd Heine (eds.), Approaches to Grammaticalization, 493-535. Amsterdam: Benjamins. Lerch, Eugen 1929 Historische Franzosische Syntax. Leipzig: O. R. Reisland. Levinson, Stephen C. 1983 Pragmatics. Cambridge: Cambridge University Press. Mann, William C. and Sandra A. Thompson 1992 Relational discourse structure: A comparison of approaches to structuring text by 'contrast'. In: S. Hwang and W. R. Merrifield (eds.), Language in Context: Essaysfor Robert E. Longacre, 19-46. Arlington: University of Texas Press. Metrich, Rene 1980 Zur Syntax und Semantik von 'obwohl' und 'wenn auch'. Paris: Centre Universitaire du Grand Palais. Onodera, Noriko O. 1995 Diachronic analysis of Japanese discourse markers. In: Andreas Jucker (ed.), Historical Pragmatics. Pragmatic Developments in the History ofEnglish, 393-437. AmsterdamlPhiladelphia: Benjamins.

From concessive connector to discourse marker

467

Pasch, Renate 1997 Weil mit Hauptsatz-Kuckucksei im Denn-Nest. Deutsche Sprache 3: 252-271. Pomerantz, Anita 1984 Agreeing and disagreeing with assessments: some features of preferred/dispreferred tum shapes. In: John M. Atkinson and John Heritage (eds.), Structures ofSocial Action: Studies in Conversation Analysis, 57-101. Cambridge: Cambridge University Press. Quirk, Randolph, Sidney Greenbaum, Geoffrey Leech and Jan Svartvik 1985 A Comprehensive Grammar of the English Language. London: Longman. Rudolph, Elisabeth 1996 Contrast. Adversative and Concessive Relations and their Expressions in English, German, Spanish, Portugese on Sentence and Text Level. Berlin/New York: de Gruyter. Schegloff, Emanuel A., Gail Jefferson and Harvey Sacks 1977 The preference for self-correction in the organization of repair in conversation. Language 53 (2): 361-382. Schegloff, Emanuel A., Elinor Ochs and Sandra Thompson (eds.) 1996 Interaction and Grammar. Cambridge: Cambridge University Press. Schiffrin, Deborah 1987 Discourse Markers. Cambridge: Cambridge University Press. Schlobinski, Peter 1992 Funktionale Grammatik und Sprachbeschreibung. Eine Untersuchung zum Gesprochenen Deutsch sowie zum Chinesischen. Opladen: Westdeutscher Verlag. Schlobinski, Peter (ed.) 1997 Syntax des Gesprochenen Deutsch. Opladen: Westdeutscher Verlag. Selting, Margret, Peter Auer, Birgit Barden, Elizabeth Couper-Kuhlen, Susanne Giinthner, Uta Quasthoff and Peter Schlobinski 1998 Gesprachsanalytisches Transkriptionssystem (GAT). Linguistische Berichte 173: 91-122. Sweetser, Eve E. 1990 From Etymology to Pragmatics. Metaphorical and Cultural Aspects ofSemantic Structure. Cambridge: Cambridge University Press. Tabor, Whitney and Elizabeth C. Traugott 1998 Structural scope expansion and grammaticalization. In: Anna Giacalone-Ramat and Paul Hopper (eds.), The Limits of Grammaticalization, 229-272. Amsterdam: Benjamins. Traugott, Elizabeth C. 1995a The Role of the Development of Discourse Markers in a Theory of Grammaticalization. Paper presented at the 12th Inter~ational Conference on Historical Linguistics. University of Manchester. 1995b Subjectification in grammaticalization. In: Dieter Stein and Susan Wright (eds.), Language, Subjectivity and Subjectivisation, 31- 54. Cambridge: Cambridge University Press.

468 Susanne Gunthner

Uhmann, Susanne 1992 Contextualizing relevance: On some forms and functions of speech rate changes in everyday conversation. In: Peter Auer and Aldo di Luzio (eds.), The Contextualization of Language, 297-336. Amsterdam: Benjamins. 1998 Verbstellungsvarianten in weil-Satzen: Lexikalische Differenzierung mit grammatischen Folgen. Zeitschrift fur Sprachwissenschaft 17.1: 92-139. Wegener, Heide 1993 weil-das hat schon seinen Grund. Zur Verbstellung in Kausalsatzen mit WElL im gegenwartigen Deutsch. Deutsche Sprache 4: 289-305.

Index account 87,89-105,292-294,296 adverbial clause post-positioned 174, 197, 181, 428f. pre-positioned 174, 179, 428f. adversative connector 118, 136, 138,235,242,249,253 adversativity 2, 5, 196, 258f., 264f., 276f. argumentation/argumentative 61, 148, 151, 156, 260, 262-265, 277f., 356f., 369f., 373, 383, 404,411,424,429-431 assessment 78, 92f., 97f., 103f., 212, 216-218, 222, 265f., 272, 284, 301, 304f., 384, 386, 389, 403, 405, 450-452 assumption 4, 40, 42, 52, 123, 156, 236, 239-241, 243-246, 248, 252f., 261,276, 295f., 313, 341, 348f., 353, 365- 367, 393, 443, 454 asyndetic construction 6, 27, 83-87, 92, 103-105, 321f., 417

background 14, 16, 20, 113, 118, 121f, 125, 127f., 131, 133f, 136, 240, 291, 313, 348f., 353, 365f., 370, 372f., 376, 393,424, 443,454

causal connective 3, 57-60, 63, 66, 77,89,104,118,136,272,306, 342, 346f., 352, 375 causal constraint 35, 38, 41, 43-45, 47,49,51,53-55 causal relation 2, 9, 12, 14, 25-27, 35-38, 40, 42f., 47f., 50, 52-55, 57f., 64-66, 77f., 83-85, 87-90, 102, 104, 111, 113, 120-123,

134-138, 236, 347f., 354, 359, 363, 370f. conceptual order 35, 38f., 4043,46-49,51-55,58 linear order 35, 38, 40-42, 5355,156,456 causality 1-3, 5f., 9-14, 18, 21-27, 29f., 35- 42, 57-59, 63f., 77, 88, 112f., 120, 122-124, 126, 134, 137f., 141, 258, 266f., 270, 276f., 341-343, 348, 350f., 354, 361f., 364f., 367, 369-373, 375f.,378 hidden causality 348f. causals 4, 37f, 138,266, 342~see also causal connective content 35-38, 370f. epistemic 38, 370f. causals, concessives, equivalence of 344, 350f., 354,362 causals, concessives, paraphrase relations 344, 351, 361 causation (concept of) 12, 22f., 27, 120 clause marking 284, 320f., 325, 391 clause order 111f., 120-122, 132, 134f., 137f., 146, 156-158, 240, 443f., 446, 460~ see also causal relation~ linear order~ adverbial clause, pre-positioned, postpositioned~ conditional clause, pre- vs. post-posed clause-combining strategy 321f. co-construction 188, 192, 216, 218222 cognition 2, 24, 35f., 57, 65, 368, 378 coherence relation 11, 24, 26-28, 56, 60f, 63, 65, 84, 87, 256 common integrator 246, 248 complaint 92, 100f., 303f., 306 conceding!concession/concessivity 1-6, 26, 177, 196, 257, 283f.,

470 Index 288, 290, 341, 343, 348-353, 355, 361-369, 371-374, 376, 378, 381-387, 389f., 393-400, 403-405, 411-414, 417-420, 422, 425, 430f., 439, 452, 454, 458 Cardinal Concessive Schema 356f., 384-392, 394-398, 402-405,442-446,427-431 concessives 5, 178, 251f., 313f., 318-320, 336, 342f., 357, 362f., 365,370,373,445,456 conditional clause pre- vs. postposed 146, 148, 154f., 157, 159, 173, 179f., 181, 183-194, 196, 322 conditional perfection 149 conditionals 5, 114-116, 118f., 122, 125f., 132, 134f., 139, 143, 147, 154, 156-159, 163, 165-168, 174f., 178, 184, 205-208, 210, 213,223-226,230, 342f., 369 complex conditional connective 143 content 116, 139 counterfactual 2f., 12f., 139, 152f., 156, 166, 205-209, 211-218, 221-231, 342 epistemic 116, 122, 135, 156 speech-act 118, 126, 139, 157f., 168 conjunction 1-3, 6, 46, 48, 52, 59, 83f., 88f., 111-113, 118-123, 125-127, 132, 135-139, 147, 161, 175, 177, 195, 198, 209, 218, 236-253, 259f., 289, 294, 300, 313, 315f., 318, 341, 346, 352, 354, 363f., 374, 391, 4Q2, 456- 460 connective If., 5f., 37, 57-63, 68, 7Of. , 75, 77f., 143-146, 149, 151, 162f., 167f., 257f., 267, 272, 300, 317, 341, 344, 346f., 352f., 355, 359, 363, 370, 375, 383,418 content domain/level/relation 36-38, 40, 43, 51f., 54, 59f., 63, 114,

119-121, 123, 126, 131, 181, 236, 239f., 243-245, 300, 313, 318, 320, 323, 326, 329, 349, 363,370 contrast 2, 136, 225f., 244f., 247249, 258f., 283-285, 288-294, 296-301, 303-306, 364, 382, 459f. conversation 84, 89, 92, 96, 111, 173, 193, 222, 283, 286, 291293, 296, 349, 381, 384, 388, 393, 401-403, 405f., 413, 422, 439-441 conversation analysis 84, 89, 92, 173,213,283, 292f. corpus analysis 5, 70, 77f., 146, 164,173, 178f., 194f., 259, 417, 420 correction 147, 264-267, 302, 305, 404, 422, 439, 441, 444-448, 450,452,456, 459f. corrective marker 454, 458f., 461 desirability 205f., 208f., 212, 215f., 226f.,229,231 disagreement 5, 87, 93, 103, 278, 283, 286, 288f., 292f., 297, 355, 382, 403f., 405, 419, 422, 439, 441, 450-453, 460 discourse function 146, 184, 225f., 411f., 420f., 427, 431f., 439, 441,452 discourse grammar 461 discourse marker 5f., 8, 257f., 275, 439, 441, 457-460 division of labour between grammar and pragmatics 235, 246f., 249, 253 duality 21, 66, 350f., 355, 362, 376 Dutch 3, 6, 46, 52, 57-61, 63f., 66, 70f., 77f., 317, 319, 347, 375, 460 dyadic 381f., 403, 414-416, 427

Index

enablement 119-123 entity type 316 epistemic domain/level/relation 3, 35-40, 42f., 46f., 51, 53f., 59f., 62f., 65-68, 70, 74-76, 78, 116, 118, 126f., 131, 236, 240f., 243245, 270-274, 313, 315, 317f., 320, 323, 326, 329, 331, 363f., 370,448 epistemic stance IIIf., 121, 123, 125f., 128-130, 178, 370-373, 377 exceptional condition 147, 150, 152 explanation 26f., 283, 285, 289-294, 296-306 causal explanation 21 focus/focussing 134, 136f., 177, 193f., 243, 250f., 314, 344f., 254 French 60, 143, 155

471

213, 226, 239, 260, 264, 283286, 293, 299, 305f., 343f., 350, 356f., 368, 383, 385f., 390, 399, 403, 406, 432, 439, 441, 450, 459 interclausal connection 235f. interclausallinking 449 intonation 85-87, 91, 105f., 120, 133-135, 137-139, 188, 191, 237, 239-241, 258, 266, 272, 291, 344f.,361, 390, 396,402f., 405f., 432, 445f., 448,455, 461 Japanese 6, 168,205f., 209f.,212f., 216-218, 223, 226f., 323, 327, 329f.,460 justification 2, 5, 26f., 90, 100, 103f., 147, 237, 258, 268, 272276,292 Korean 6, 205f., 212f., 216-218, 223,226

German 2, 4-6, 8, 60, 73, 87, 97, 166, 173-175, 178-181, 184, 186f., 191, 195, 198, 240-242, 296, 319,336, 346f., 353, 421f., 439-441,444,453,455-461 spoken German 83f., 173, 175, 179f., 194f., 198,444,460 writtenGerman 173f., 180, 194, 196, 198f.

levels of interpretation 3, 36-38, 5963, 112f., 238, 243, 252f., 313, 315,332,370,373,459 lexical contiguity 328, 330 mental space 3f., 65, 111-114, 117128, 132, 134, 137f., 184, 193, 206-208,225,361,365-378

hypotaxis/hypotactic 4, 6, 180, 192, 420,431,458 iconicity/iconic 41f., 120f., 140, 148, 156, 166, 184, 237, 240, 319,446 imagination 215f., 223 informativeness (scale of) 26, 317 interaction 5, 18, 24, 49, 51, 83f., 87f., 94, 105, 111, 118f., 124, 127, 137, 173, 183-188, 199,

narrative 27, 29, 225, 228f., 429432,450 negation/negative 104f., 119, 144146, 151-153, 208f., 211, 240, 249f., 262, 264, 274, 287-290, 294, 296, 301, 314, 341, 344f., 347, 350f., 353-355, 361f., 374f.,377f.

472 Index

parataxis/paratactic 4, 6, 83, 138, 161-163, 266, 270, 320f., 325, 327, 345, 431, 458f. point of view 35, 64, 88, 152, 163, 207, 223, 257, 366, 369, 383, 416,419,430 politeness 88, 165, 418f., 431 polyphony 4, 368f. polysemy 235, 257-259, 261, 275, 277f.,363 pragmatic ambiguity 235, 237f., 242, 246f.,275 prediction 38, 41f., 48f., 51f., 68-70, 114f., 122, 125, 131, 134f., 165f.,350 predictive function 111, 139, 226f. presupposition 262, 313f., 336, 342, 346, 352-354, 365f., 442, 455 projectability 398 prosody 84-87, 98, 147, 177, 289, 381, 398, 401, 445f.

semantic domain 78, 315 semantic partitioning 381, 399 semantic primitives 22 Spanish 2, 4-6, 143f., 146-148, 155, 157f., 160, 166-168, 257-259, 261,273,277, 315f.,459f. speech act domain/level/relation 3, 59f., 119, 126, 131, 157f., 181, 235f.,237,239f.,243,245,253, 270, 313, 317f., 320f., 323f., 326f., 329, 331f. speech setting 111, 119, 124 spoken language 5f., 173, 194, 196, 199, 286, 288, 411f., 417f., 420, 430 statement 235-237, 240, 243, 252f. subject of consciousness/SOC 3, 57, 64-70, 73-78 subjectivity 3, 57, 59, 63f., 67f., 7478 subordinator 2, 5f., 240f., 336, 392, 440f., 450, 452-453, 456f., 459461

quantifier 347, 350 reasoning process 36, 38, 116, 119, 124, 364f., 372 refutation 192, 262-264, 276f. Relation (maxim of) 20f. Relevance (principle of) 20-22 representation 35f., 38 request 89, 92, 96, 99f., 103f., 267, 418 restriction 58, 61, 126, 139, 148f., 159f., 163, 166, 238, 240f., 273, 336, 422, 428f., 455 restrictive use 421f., 454, 456 rhetorical pattern 283, 297 Rhetorical Structure Theory/RST 284, 288f., 297, 306, 412, 442, 445 scope 190, 239f., 250, 319, 344f., 254, 272f. semantic change 346

topic-focus articulation 243, 251 topos 366-368, 373 viewpoint 4, 64, 257, 260-262, 264, 267, 272, 276f., 368f., 375, 378 volitionality/volitional 63-68, 70-74 word order 1, 178, 181, 319, 322, 439-441

Authors. addresses Noriko McCawley Akatsuka Dept. of East Asian Languages & Cultures University of California at Los Angeles 276-B Royce Hall Los Angeles 90095 USA [email protected]

Barbara Dancygier English Language and Culture Program Simon Fraser University 580-515 West Hastings Street Vancouver, B.C. V6B 5K3 Canada [email protected]

Peter Auer Deutsches Seminar I Albert-Ludwigs-Universitat Freiburg Postfach D-79085 Freiburg Germany [email protected]

Femke de Blijzer Faculty of Arts Tilburg University P.O. Box 90153 NL-5000 LE Tilburg The Netherlands [email protected]

Dagmar Barth Germanistisches Institut Martin-Luther-Universitat HalleWittenberg Luisenstr. 2 0-06099 Halle/Sa. Germany [email protected]

Cecilia E. Ford Dept. of English University of Wisconsin-Madison 600 N. Park Street Madison, WI 53706 USA [email protected]

Elizabeth Couper-Kuhlen Sprachwissenschaft (Anglistik) Universitat Konstanz Postfach 5560 0-78434 Konstanz Germany [email protected] Mily Crevels Dept. of Comparative Linguistics Leiden University PO Box 9515 NL-2300 RA LEIDEN The Netherlands [email protected]

Christine Gohl Sprachwissenschaft Universitat Konstanz Postfach 5560 0-78434 Konstanz Germany Christine. [email protected] Susanne Giinthner Sprachwissenschaft Universitat Konstanz Postfach 5560 0-78434 Konstanz Germany Susanne. [email protected]

474 Authors' addresses Ekkehard Konig Institut flir Englische Philologie Freie UniversiUit Berlin Gosslerstr. 2-4 0-14195 Berlin Germany [email protected]

Leo Noordman Faculty of Arts Tilburg University P.O. Box 90153 NL-5000 LE Tilburg The Netherlands [email protected]

Bernd Kortmann Englisches Seminar I Albert-Ludwigs-Universitat Freiburg Rempartstr. 15 0-79098 Freiburg Germany [email protected]

Henk Pander Maat Dept. of Dutch/Centre for Language and Communication Utrecht University Trans 10 NL-3512 JK Utrecht The Netherlands [email protected]. nl

Ewald Lang Zentrum flir Allgemeine Sprachwissenschaft Jagerstr. 10/11 0-10117 Berlin Germany [email protected] Paul Georg Meyer Institut flir Anglistik III RWTHAachen Karmanstr. 17-19 0-52062 Aachen Germany [email protected]

Estrella Montolio Facultat de Filologia Dept. de Filologia Hispanica (Lengua Espanola) Universitat de Barcelona Gran Via de les Corts Catalanes, 585 08007 Barcelona Spain [email protected]

Ted Sanders Dept. of Dutch/Centre for Language and Communication Utrecht University Trans 10 NL-3512 JK Utrecht The Netherlands Ted. [email protected] Scott Schwenter Dept. of Spanish and Portuguese The Ohio State University 266 Cunz Hall 1841 Millikin Road Columbus, OR 43210-1229 USA [email protected]

Peter Siemund Institut flir Englische Philologie Freie Universitat Berlin Gosslerstrasse 2-4 0-14195 Berlin Germany [email protected]

Authors' addresses Susan Gayle Strauss Dept. of Speech Communication The Pennsylvania State University 234 Sparks Building State College, PA 16802 USA [email protected] Eve Sweetser Dept. of Linguistics University of California at Berkeley 1203 Dwinelle Hall Berkeley, CA 94720-2650 USA [email protected]

Sandra A. Thompson Dept. of Linguistics University of California Santa Barbara, CA 93106 USA [email protected] Arie Verhagen Dutch Linguistics Leiden University P.O. Box 9515 2300 RA Leiden The Netherlands Arie. [email protected]

475

Topics in English Linguistics Edited by Bernd Kortmann and Elizabeth Closs Traugott Mouton de Gruyter . Berlin· New York 1 Niels Davidsen-Nielsen, Tense and Mood in English. A Comparison with Danish. 1990. 2 Historical English Syntax. Edited by Dieter Kastovsky. 1991. 3 English Computer Corpora. Selected Papers and Research Guide. Edited by Stig Johansson and Anna-Brita Stenstrom. 1991. 4 Donka Minkova, The History ofFinal Vowels in English. The Sound ofMuting. 1991. 5 Lia Korrel, Duration in English. A Basic Choice, Illustrated in Comparison with Dutch. 1991. 6 Andreas H. Jucker, Social Stylistics. Syntactic Variation in British Newspapers. 1992. 7 Ken-ichi Takami, Preposition Stranding. From Syntactic to Functional Analyses. 1992. 8 Bas Aarts, Small Clauses in English. The Nonverbal Types. 1992. 9 New Directions in English Language Corpora. Methodology, Results, Software Developments. Edited by Gerhard Leitner. 1992. 10 History of Englishes. New Methods and Interpretations in Historical Linguistics. Edited by Matti Rissanen, Ossi Ihalainen, Terttu Nevalainen and Irma Taavitsainen. 1992. 11 Early English in the Computer Age. Explorations through the Helsinki Corpus. Edited by Matti Rissanen, Merja Kyto and Minna Palander-Collin. 1993. 12 Towards a Standard English: 1600-1800. Edited by Dieter Stein and Ingrid Tieken-Boon van Ostade. 1993. 13 Studies in Early Modern English. Edited by Dieter Kastovsky. 1994. 14 Ronald Geluykens, The Pragmatics of Discourse Anaphora in English. Evidencefrom Conversational Repair. 1994. 15 Traute Ewers, The Origin ofAmerican Black English. Be-Forms in the HOODOO Texts. 1996. 16 lIse Depraetere, The Tense System in English Relative Clauses. A CorpusBased Analysis. 1996. 17 Michiko Ogura, Verbs in Medieval English. Differences in Verb Choice in Verse and Prose. 1996. 18 Spanish Loanwords in the English Language. A Tendency towards Hegemony Reversal. Edited by Felix Rodriguez Gonzales. 1996. 19 Laurel J. Brinton, Pragmatic Markers in English. Grammaticalization and Discourse Functions. 1996.

20 Christiane Dalton-Puffer, The French Influence on Middle English Morphology. A Corpus-Based Study on Derivation. 1996. 21 lohan Elsness, The Perfect and the Preterite in Contemporary and Earlier English. 1997. 22 Carl Bache and Niels Davidsen-Nielsen, Mastering English. An Advanced Grammar for Non-native and Native Speakers. 1997. 23 English in Transition. Corpus-based Studies in Linguistic Variation and Genre Styles. Edited by Matti Rissanen, Merja Kyto and Kirsi Heikkonen. 1997. 24 Grammaticalization at Work. Studies of Long-term Developments in English. Edited by Matti Rissanen, Merja Kyto and Kirsi Heikkonen. 1997. 25 Axel HUbler, The Expressivity of Grammar. Grammatical Devices Expressing Emotion across Time. 1998. 26 Negation in the History ofEnglish. Edited by Ingrid Tieken-Boon van Ostade, Gunnel Tottie and Wim van der Wurff. 1998. 27 Martina Hacker, Adverbial Clauses in Scots: A Semantic-Syntactic Study. 1998. 28 Ingo Plag, Morphological Productivity. Structural Constraints in English Derivation. 1999. 29 Gustav Muthmann, Reverse English Dictionary. Based on Phonological and Morphological Principles. 1999. 30 Metaphor and Metonymy at the Crossroads. A Cognitive Perspective. Edited by Antonio Barcelona. 2000. 31 Generative Theory and Corpus Studies. A Dialogue from 10 ICEHL. Edited by Ricardo Bermudez-Otero, David Denison, Richard M. Hogg and C. B. McCully. 2000. 32 Manfred G. Krug, Emerging English Modals. A Corpus-Based Study ofGrammaticalization. 2000. 33 Cause - Condition - Concession - Contrast. Cognitive and Discourse Perspectives. Edited by Elizabeth Couper-Kuhlen and Bernd Kortmann. 2000.