CONSTRUCTIONS IN COGNITIVE LINGUISTICS
AMSTERDAM STUDIES IN THE THEORY AND HISTORY OF LINGUISTIC SCIENCE General Editor E. F. KONRAD KOERNER (University of Ottawa)
Series IV - CURRENT ISSUES IN LINGUISTIC THEORY
Advisory Editorial Board Raimo Anttila (Los Angeles); Lyle Campbell (Christchurch, N.Z.) Sheila Embleton (Toronto); John E. Joseph (Edinburgh) Manfred Krifka (Austin, Tex.); Hans-Heinrich Lieb (Berlin) E. Wyn Roberts (Vancouver, B.C.); Hans-Jürgen Sasse (Köln)
Volume 178
Ad Fooien and Frederike van der Leek (eds.) Constructions in Cognitive Linguistics Selected papers from the Fiflh International Cognitive Linguistics Conference, Amsterdam, 1997
CONSTRUCTIONS IN COGNITIVE LINGUISTICS SELECTED PAPERS FROM THE FIFTH INTERNATIONAL COGNITIVE LINGUISTICS CONFERENCE Amsterdam, 1997
Edited by AD FOOLEN University of Nijmegen FREDFRIKE VAN DER LEEK University of Amsterdam
JOHN BENJAMINS PUBLISHING COMPANY AMSTERDAM/PHILADELPHIA
The paper used in this publication meets the minimum requirements of American National Standard for Information Sciences — Permanence of Paper for Printed Library Materials, ANSI Z39.48-1984.
Library of Congress Cataloging-in-Publication Data International Cognitive Linguistics Conference (1997 : He Amsterdam) Constructions in cognitive linguistics : selected papers from the Fifth International Cognitive Linguistics Conference, Amster dam, 1997 / edited by Ad Fooien, Frederike van der Leek. p. cm. -- (Amsterdam studies in the theory and history of linguistic science. Series IV, Current issues in linguistic theory, ISSN 0304-0763 ; v. 178) Includes bibliographical references and index. 1. Cognitive grammar-Congresses. I. Fooien, Ad. II. Leek, Frederike van der. III. Title. IV. Series. P165.I58 2000 415-4IC21 99-089845 ISBN 90 272 3684 4 (Eur.) / 1 55619 955 4 (US) (Hb; alk. paper) CIP © 2000 - John Benjamins B.V. No part of this book may be reproduced in any form, by print, photoprint, microfilm, or any other means, without written permission from the publisher. John Benjamins Publishing Co. • P.O.Box 75577 • 1070 AN Amsterdam • The Netherlands John Benjamins North America • P.O.Box 27519 • Philadelphia PA 19118-0519 • USA
CONTENTS
Acknowledgements
vii
Editors' Foreword
ix
Pragmatic Conditionals Angeliki Athanasiadou & René Dirven
1
How Polish Structures Space: Prepositions, Direction Nouns, Case, and Metaphor Barbara Dancygier
27
Case Meaning and Sequence of Attention: Source Landmarks as Accusative and Dative Objects of the Verb Robert B. Dewell
47
Fijian Children's Possessive Categories and Constructions Patrick Griffiths
67
Facing up to the Meaning of 'face up to': A Cognitive Semantico-Pragmatic Analysis of an English Verb-Particle Construction Beate Hampe
81
Gerundive Nominalization: From Type Specification to Grounded Instance Liesbet Heyvaert A Cognitive Approach to Errors in Case Marking in Japanese Agrammatism: The Priority of the Goal -ni over the Source -kara Hiroko & Ikuyo Fujita
103
123
VI
CONTENTS
Verbal Aspect and Construal Agota Kochañska
141
How I got myself arrested: Underspecificity in Grammatical Blends as a Source for Constructional Ambiguity Nili Mandelblit & Gilles Fauconnier
167
Konjunktiv II and Epistemic Modals in German: A Division of Labour Tanja Mortelmans
191
Subjectivity and Conditionality: The Marking of Speaker Involvement in Modern Greek Kiki Nikiforidou & Demetra Katis
111
English Imperatives and Passives Hidemitsu Takahashi
239
Lexical Causatives in Thai Kingkarn Thepkanjana
259
Cognitive Models in Transitive Construal in the Japanese Adversative Passive Eijiro Tsuboi
283
Caused-Motion and the 'Bottom-Up' Role of Grammar Frederike van der Leek
301
Addresses
333
Index
335
Acknowledgements
This volume was realized with the help of a great many people. First of all we would like to thank all authors for their contributions and cooperation. Thanks are also due to the following people who acted as anonymous reviewers: Michel Achard, Willem Botha, Melissa Bowerman, Claudia Brugman, Eve Clark, Herbert Clark, Kenneth Cook, Barbara Dancygier, Bob Dewell, Matthew Dryer, Patrick Duffley, Peter Harder, Joe Hilferty, Bob Kirsner, Ron Langacker, David Lee, Ricardo Maldonado, Yo Matsumoto, Haruko Minegishi Cook, Laura Michaelis, Kiki Nikiforidou, Jan Nuyts, Jan-Ola Östman, Maria Polinsky, Klaus-Uwe Panther, Jo Rubba, Masayoshi Shibatani, Simon Slings, Michael Smith, Leon Stassen, Eve Sweetser, John Taylor, Friedrich Ungerer, Johan van der Auwera, Marjolijn Verspoor and Margaret Winters. We thank Anke de Looper of John Benjamins Publishing Company for her professional advice and her patience and, last but not least, Rob van den Berg for his moral support and his expertise in preparing the final versions of the papers. Nijmegen/Amsterdam November 1999 Ad Fooien & Frederike van der Leek
Editors' Foreword
This volume contains a selection from the proceedings of the 5th International Cognitive Linguistics Conference, which was held at the Free University in Amsterdam, from July 14-19, 1997. The volume is a companion of two others that also contain selected papers of the same conference: Metaphor in Cognitive Linguistics, edited by Raymond W. Gibbs Jr. and Gerard J. Steen (CILT 175) and Discourse Studies in Cognitive Linguistics, edited by Karen van Hoek, Andrej A. Kibrik and Leo Noordman (CILT 176). The title of the present volume may be assumed to speak, to a certain extent, for itself. As Goldberg (1995:1) observes, "[t]he notion construction has a time-honored place in linguistics", but was temporarily renounced in the Chomsky an Government and Binding framework. In Cognitive Linguistics, it is given pride of place again, though the notion is, on the whole, subject to different interpretations as far as scope and organizational level are concerned. Langacker (1987:409) sees grammatical constructions as the form-meaning "integration of two or more component structures to form a composite expression". The Construction Grammar framework (cf. Fillmore & Kay to appear) employs the much more rigorous notion that a particular form-meaning combination is a construction only if it has some (form and/or meaning) property that "is not strictly predictable from [its] component parts, or from other previously established constructions" (Goldberg 1995:4). Beyond the consensus, then, that constructions constitute recognizable form-meaning patterns, opinions differ. This is also evident from the papers making up the present volume. They seem to vary significantly in their vision on what constructions are. The research areas they are concerned with, also cover a wide range of different topics. For this reason we have decided to make no attempt to organize the papers thematically; we simply present them in alphabetical order. In order to give the reader some preliminary idea of what this volume has to offer, we will, however, first outline its contents from various angles. Except for Griffiths' paper, which deals with child language, all the papers are directly concerned with particular constructions as used by mature speakers. A variety of languages is covered; in alphabetical order: English, Fijian, French, German, (Modern) Greek, (Modern) Hebrew, Japanese, Polish and Thai. Two of the papers, by Griffiths and Ihara & Fujita, are experimentally based and approach the constructions dealt with from psychological angles, i.e.
X
EDITOR'S FOREWORD
that of acquisition and agrammatism respectively, a widening of the field that is very welcome. Thematically, we can recognize, very roughly, the following constructional fields of interest: (verbal) Aspect, (Kochanska), Case (Dancygier; Dewell; Griffiths; Ihara & Fujita), Causation (Mandelblit & Fauconnier; Thepkanjana; Van der Leek), Conditionals (Athanasiadou & Dirven; Nikiforidou & Katis), Gerundive Nomináis (Heyvaert), Modality (Mortelmans), Particles (Hampe), and Passives (Takahashi; Tsuboi). Various authors furthermore both argue and illustrate the importance of corpus-based data (Athanasiadou & Dirven; Hampe; Mortelmans; Nikiforidou & Katis). As for what we might call 'cognitive mechanisms', we find the notion Construal to be pervasive throughout all the papers, while the following types of mecha nisms play a central role in more specific analyses: Blending and Underspecificity (Mandelblit & Fauconnier), Compositionality (and beyond) (Dancygier; Kochanska; Van der Leek), Grounding (Heyvaert; Mortelmans); Cognitive Models/Frames (Takahashi; Tsuboi; Van der Leek); Lexical Alternation (Thepkanjana; Van der Leek), Subjectivity (Mortelmans; Nikiforidou & Katis), and, lastly, Sequence of Attention (Dewell). Obviously, the above classification is a rough cross-specification, not intended as exhaus tive; too many specifically cognitive issues crop up in individual papers to be all mentioned here. Assuming that the above has nevertheless provided the reader with enough of a first impression of what to expect from this volume, we will now turn to a brief characterization of each individual paper. Athanasiadou & Dirven discuss English pragmatic, non-prototypical conditionals, a class which, in their conception, also comprises epistemic/ logical conditionals. Their paper, which is corpus-based, argues that the four subtypes they distinguish, 'identifying' and 'inferencing' (both logical) and 'discourse' and 'metacommunicative' (both conversational) all share certain basic pragmatic characteristics that are extended in different ways and the paper aims to show that these differences can account for the variety of forms each type can manifest. Dancygier analyzes the conceptualization of space in Polish, arguing that the way this language construes space is dependent on three subsystems, i.e. direction nouns, prepositions and case. Her overall conclusion is that the spatial expressions in question are constructed compositionally, with each of the three structural subsystems making a consistent meaning contribution of its own. Dewell discusses the schematic meanings of accusative and dative cases, arguing that these do not simply reflect different roles in the action chain, but,
EDITORS' FOREWORD
XI
instead, impose a certain way of construing an event. In particular, he points out a central difference between the above two cases that concerns the notion 'sequence of attention'. The accusative, he argues, typically makes one change focus from the subject to the direct object referent, while atypical variants (with the accusative NP itself evoking a pathway) effect a construal that makes one stay attentive to both at the same time. The dative, on the other hand, has the effect that the NP referent never gains central attention, but is construed as staying, in a sense, outside the direct action chain. In his comparison of English and German, Dewell shows that the structure of the languages involved (absence of a difference between dative and accusative case in English) is a factor that plays its own role in the type of construals that the two languages make available for an event. Griffiths analyzes children's acquisition data on Fijian possessive constructions. Fijian differentiates between alienable and inalienable possession and has, besides an unmarked possessive marker, special markers for possession of 'food' and 'drink'. The data suggest that acquisition of this type of markers is cognitively driven in that, firstly, possessive meanings get lin guistically expressed before the relevant constructions are mastered, secondly, conceptually simpler categories, e.g. the singular marker and the default possessive marker, are learnt prior to the 'food' and 'drink' ones, and, thirdly, the latter type of markers are applied first to prototypical food and drink categories, and only later to more abstract members of the category. Hampe shows that the existence in English of the phrasal verb 'to face up to' side by side with the simple verb 'to face' is not to be seen as 'wordy speech', but as a multiply motivated construction, showing that the particle combination conveys that the challenge induced by some problem/obstacle is actually met, whereas the simple verb merely expresses that the problem/ obstacle does not go, so to speak, unnoticed. Conceptually speaking, the meaning of the phrasal verb is 'motivated' by the meaning of its parts, though motivation, as Hampe is careful to point out, is not the same as strict compositionality. From a pragmatic perspective, the phrasal verb is motivated by expressivity, and a corpus analysis method is used to argue this latter point. Heyvaert presents a proposal for refining Langacker's analysis of English gerundive nomináis. She argues that the -ing marker in gerundive nomináis indicates, differently from the same marker in action nomináis, that the complex predication made up of the verb and its object (which she analyzes, pace Langacker (1991), as a clausal type rather than a clausal instance) has the status of nonfinite clausal head of the gerundive construction, a construction which requires a (periphrastic) subject for its instantiation. This analysis enables her to distinguish intrinsically between action nomináis, gerundive
Xll
EDITORS' FOREWORD
nomináis and that-clmse nomináis, and thereby account for the semantic differences these three types of nomináis exhibit, as well as for the way they differ in their external behaviour. Ihara & Fujita discover, from an experimental study with three Japanese subjects suffering from agrammatism, that these patients wrongly use the source marker -kara instead of the goal marker -ni in a significant number of cases, and not the other way around. This behaviour, they argue, is in line with certain linguistic phenomena relating to source and goal markers: source markers are marked in comparison to goal markers and source constructions can, diachronically, develop into goal constructions. Both the linguistic phenomena and the agrammatic data can be explained by the cognitive action chain model, in which the goal is in line with the energy flow that originates in an actor, whereas a source conceptualizes a point that is opposite to the direction of the action chain. Kochañska, dealing with verbal aspect in Polish, restricts herself to accounting for the acceptability and possible interpretations of coordinated VP constructions involving perfective/imperfective VPs, seen from a temporally extended or a temporally momentary Viewing Frame. On the basis of her findings, she argues, with Langacker (1991, inter alia), that the notions of temporal boundedness/unboundedness and internal heterogeneity/homogeneity should be kept distinct; both, she argues, pace Verkuyl (1993), are relevant for verbal aspect. Her general conclusions are that Polish perfective and imperfective VPs are both polysemous, that the meaning of coordinated VPs is constructional rather than purely compositional, and that notions like boundedness are a matter of construal rather than constituting 'objective' facets of the world: an event is, for instance, bounded if we (can) think of it as such, not because it is bounded regardless of how one looks at the situation. An important consequence of this latter viewpoint is that only a truly subjectivist analysis of aspectual phenomena can, according to Kochañska, deal with the facts in a revealing manner. Mandelblit & Fauconnier argue that a generic conceptual causative schema underlies not only grammatical causative constructions, but passive, middle and reflexive constructions as well. They analyze, to concentrate on the first type, causative syntactic constructions in English, Modern Hebrew (hif'il) and French. In their view, a sentence is the result of blending a conceived event with a schematically meaningful syntactic construction. Because not every element of the conceived event is projected on the construction, possible ambiguities may arise, for example between middle and passive constructions in Modern Hebrew. Because, furthermore, mappings from the conceived event onto the meaningful syntax may differ in systematic ways, the blending
EDITORS' FOREWORD
Xlll
mechanism can also account for the difference in mapping responsible for the meanings of Rachel sneezed the napkin off the table on the one hand (here the construction's subject referent is understood to do the sneezing), and of She trotted the horse into the stable on the other (here the trotting is understood as done by the object referent). Language being to a large extent under specified, grammatical blending can thus account for various kinds of constructional ambiguity. Mortelmans shows, on the basis of original data, that the Konjunktiv II mood in German on its own signals negative epistemic stance (irrealis), that epistemic modals in the Konjunktiv II form, e.g. könnte (cf. Indicative kann), still signal positive epistemic stance (potentiality), be it in a weakened form. However, in conditionals with Konjunktiv II functioning as grounding predi cation in the protasis, the effect on e.g. könnte is that it now can only express negative epistemic stance (irrealis). Interestingly, the form dürfte has grammaticalized in a different direction, expressing a highly subjective epistemic evaluation; it is therefore on the whole considered unacceptable if Konjunktiv II has a dominant grounding function. Her conclusion is, therefore, that mood and modals exhibit what she aptly calls 'division of labour'. Nikiforidou & Katis show, on the basis of authentic material, that the traditional claim that the conditional markers ama, ean and na are free variants of the basic conditional marker an is wrong, and that the three markers under discussion code extra semantic dimensions, in that each involves, in its own way, a "subjective construal ... of the situation depicted in the protasis" (section 5; 'subjective' in Langacker's (1987) sense of 'egocentric construal'). The authors show that the three markers display semantic extensions as well, extensions which seem to point more in the direction of Traugott's (1995) notion of subjectivity by marking highly personal facets like empathy or involvement, thus providing motivated links from the basic egocentric semantics of the markers in question. As Takahashi points out, imperatives do not, normally speaking, occur in the passive voice in English, cf. *Be helped by Jill; however, there are felicitous examples of passive imperatives as well, cf. Be flattered by what he says, it'll make his day. Takahashi shows that it is conceptual incompatibility that leads to the typical unacceptability of passive imperatives. The (second person) subject of an imperative is, given a prototypical Imperative Event Model, an agent, whereas the subject of a passive, given a prototypical Passive Model, is a patient. Only when used in non-prototypical ways, can the two constructions become compatible. It is this cognitive viewpoint that enables Takahashi, moreover, to explain why in Japanese passive imperatives are hardly (if at all) acceptable: unlike English, the meaning of the Japanese
XlV
EDITORS' FOREWORD
imperative is restricted to the prototypical Imperative Event Model, hence passive and imperative constructions are bound to clash in Japanese. Thepkanjana, analysing the semantics of lexical causative verbs in Thai, makes a major distinction between alternating and non-alternating lexical causatives, and a further subdivision into (a) different types of causatives, (b) agent-oriented causatives and (c) patient-oriented causatives. On the basis of these distinctions, and with the help of Talmy's notion of 'windowing of attention', she neatly accounts for the kind of constructions the verbs in question can, or cannot, occur in, while also making it clear that the notion 'accidental lexical gap' is a misnomer as far as suppletive lexical causatives in Thai are concerned: their distribution is by no means random. Tsuboi deals with the adversative passive construction in Japanese, a construction which involves the quite fascinating phenomenon of valence increase: particular verb types, which are intransitive when used actively, can passivize provided an extra argument, marked by ablative -ni, is added to the passive construction. Earlier analyses essentially account for this valence increase in random terms: the construction requires an extra-thematic NP. Tsuboi, on the other hand, argues that the adversative passive is motivated by a particular Japanese cultural cognitive model that involves a person's 'sense of responsibility'. This models a person A responsible for something bad happening to a person B, provided A could have prevented this from happen ing; A is thus seen as having affected adversatively (regardless whether this was intentional or not). The meaning of the passive adversative can only be understood against this ICM, with the --marked actor mapping onto person A, provided, moreover, that the passive subject referent (the focus of attention in the passive) is actually annoyed at the actor in question. This model of responsibility ascription shows itself not only in the passive construction, but also in some active contexts that also involve valence increase. The adversative passive is, thus, not an isolated phenomenon in Japanese, as previous analyses suggested. Van der Leek, lastly, argues against the analysis of the English causedmotion construction as proposed by Goldberg (1995), who claims, essentially, that this is a construction in the Construction Grammar sense (cf. Fillmore & Kay to appear), in that its syntax prototypically pairs up with a caused-motion sense, and with extended senses in nonprototypical uses of the construction; thanks to this claim, she can argue that verbs have basically only one meaning, and that a verb is licensed in the construction provided its basic meaning is compatible with one of the construction's conventional senses. Van der Leek argues instead that, where mature speakers are concerned, the pure syntax of the construction makes no semantic contribution of its own, and that the
EDITORS' FOREWORD
XV
caused-motion pattern (or any of its extensions) gets extracted by the language user as a superordinate conceptual category, a category that, indeed, licenses particular usages without, however, lending them their meaning. She further argues that pure syntactic configurations have no semantic value prior to conceptualization, even though children begin by assuming that there is a direct mapping between their prelinguistic conceptualizations of caused-motion and the meaning of the syntactic configuration in question. References Fillmore, & P. Kay. To appear. Construction Grammar. Stanford: Center for the Study of Language and Information. Goldberg, A. 1995. A Construction Grammar Approach to Argument Structure. Chicago: The University of Chicago Press. Langacker, R. 1987. Foundations of Cognitive Grammar. Vol. I. Theoretical Perspectives. Stanford: Stanford University Press. 1991. Foundations of Cognitive Grammar. Vol. 2. Descriptive Application. Stanford: Stanford University Press. Traugott, E. 1995. "Subjectification in Grammaticalization". Subjectivity and Subjectivization in Language eds. D. Stein & S. Wright, 31-54. Cambridge: Cambridge University Press. Verkuyl, H. 1993. A theory of aspectuality. Cambridge: Cambridge University Press.
Pragmatic Conditionals Angeliki Athanasiadou University of Thessaloniki, Greece
René Dirven University of Duisburg, G
1.
Introduction: Types of conditionals * In previous papers on conditionals, we have distinguished between three major types of conditionals. The first type is the traditionally perceived 'hypo home, in which the consequent fully depends on the realisation of the antecedent. A second type has largely remained unobserved in English grammars and in the recent literature,1 though it has a very high statistical occurrence in the corpora excerpted.2 It is marked by the absence of modal forms in the consequent, but describes regularly co-occurring sequences of events as in If there is a drought, the eggs remain dormant. Such a sequence of events is somehow also causally related, but not construed as such, i.e. the two events are not construed as simply triggering each other but just as co-occurring. Therefore we have labelled it as 'course-of-events conditional'. The we are having dinner at six, which Fauconnier (1994:121) analyses in terms of a hidden conditional of the type If you are hungry, join us at six. But the real complexity of this pragmatic type of conditional is much greater and we will devote this whole paper to this subtype of a pragmatic conditional amongst various other members of this hybrid category.
Before beginning the analysis of pragmatic conditionals proper we want to point out that both course-of-events conditionals and pragmatic conditionals, contain their own subclass of 'inferencing' conditionals. This raises the question of the fundamental difference between the conceptual domains of course-of-eventness and of pragmaticity. This problem can be best approached by looking into some distinctions that can be made between these two types of conditionals: inferencing course-of-events conditionals as in (1) contrast in various respects with inferencing pragmatic conditionals as in (2): (1) He looked at his watch; if the soldier was coming, it was nearly time. [COB:243] (2) If she's divorced, (then) she's been married. [Sweetser 1990:116]
2
ANGELIKI ATHANASIADOU & RENE DIRVEN
The main difference between the two inferencing processes is a question of the nature of the evidence: in (1) the inferencing process is about a really occurring situation, whereas in (2) the inferencing is about the explicitation of a given concept; the mental process in (2) need not involve any actual situation, but may be a purely logical operation in the mind.
In the course-of-events conditional of (1), the course-of-events character fol soldier's coming is always so punctual that the observer automatically knows what time it is and what to do next; he just checks the time of the situation by looking at his watch. The inferencing process is embedded in an actually occurring situation, so that the inference is based on an obser consequently linked with a second situation which tends to co-occur with the first. In the pragmatic conditional of (2), there is no observation of situations, but only a logical operation of spelling out entailments of a given statement or concept; the notion of 'being divorced' implies a previous state of 'having been married before'. Here the consecutive aspect of the two events is adequately expressed by the originally temporal, i.e. consecutive use of the adverb then. Historically, this temporal adverb has been grammaticalized as a logical operator, associating temporal succession with logical inference. In (2) the temporal meaning is no longer present, but only the logical 'inference' meaning is given. The two inferencing processes are thus of a different nature: in an observational inference' we make simultaneous ad hoc inferences, applicable to the situational context; it is the logic of simultaneous events. But in a purely 'logical' inference, we make generic inferences, based on the meanings of consecutive events found in our conceptual world; it is the logic of concepts. But why can the logical inferencing process, as illustrated in (2), be called a 'pragmatic' conditional? This question only arises if one takes the term pragmatic in a narrow sense. Very often the term pragmatic is used only in relation to speech acts, but its original meaning, as specified by Charles Morris (1946), is much wider. Morris made the distinction between syntactics, semantics, and pragmatics, such that each level involves different relations: syntactics specifies the relations between the signs themselves, semantics specifies the relations between signs and the world, and sense that we want to use the term 'pragmatic' for our present purposes. When speaking of 'pragmatic conditionals', we thus suggest that there is always a special emphasis on the presence of the user of the signs. This is
pragm
PRAGMATIC CONDITIONALS
most emphatically felt in 'interpersonal' speech acts,3 such as requests or offers, but it can also be present in speech acts of the ideational type such as inferencing processes of the logical type. Indeed, here the logical
inferenc
In an interpersonal speech-act situation the pragmatic aspect is explicitly or implicitly. The speaker, when using a pragmatic conditional in a discourse situation, performs a speech act, and the if-clause explicates one of the preparatory conditions, i.e. the speaker suggests to the hearer that he believes that the state of affairs in the consequent may become true. Consider sentence (3) below taken from Athanasiadou and Dirven (1996): (3) If anyone wants me, I'm downstairs.
This issue has led to a great diversity of opinions concerning the categorisation of types of conditionals. On the one hand, we have those such as Sweetser (1990:123), who maintain that logical conditionals as in (2) and conversational conditionals as in (3) are strictly separate categories;4 on the other hand, we have those such as Comrie (1986:81), who group logical and conversational conditionals together. By establishing the superordinate category of pragmatic conditionals we adopt the latter position. Our various degrees for the three main types of conditionals. The causal dependency conditionals; causal dependency decreases considerably, but remains imp a purely logical, i.e. non-causal, relationship in pragmatic conditionals of the inferencing type as found in (2), and to a merely conversational point of relevance in pragmatic conditionals of the discourse type found in (3). But the dependency relation is never totally absent.
The categories and subcategories we propose for discussion in this paper are summarised in Table 1. This table reflects a taxonomic hierarchy, consisting of the 'logical' and 'conversational' conditionals (where the term 'basic level' is taken in a wider sense than usual), and the four subordinate categories or subcategories.
superordi
4
ANGELIKI ATHANASIADOU & RENÉ DIRVEN
Table 1. Categories and subcategories of pragmatic conditionals (PCs)
-
The paper, then, deals with each of these categories in their turn: General characteristics of pragmatic conditionals, Identifying conditionals, Inferencing conditionals, Discourse conditionals, Metacommunicative conditionals.
2.
General characteristics of pragmatic conditionals (PCs) As Table 1 shows, pragmatic conditionals comprise four main types: two of them involve logical processes, i.e. the 'identifying' PC as illustrated in (4) below, and the 'inferencing' PC as depicted in sentence (5). The two other types involve conversational contexts, and comprise the 'discourse' PC as in (6) and the 'metacommunicative' PC as in (7). (4) If there's one human species that ought to be put out to pasture, it's Presidents and Prime Ministers. [COB: 136] (5) If the super-organism created by a colony of termites can be compared to an antelope, then the disciplined aggressive columns of the army ants must be reckoned to be the insect equivalent of a beast of prey. [COB:268] (6) What about the parents demonstrating, if there are no friends? [COB:84] (7) I've come to offer my congratulations, if that's the right word. [LDC:42] Before characterising the internal structure of each of these four categories, we must first address more general questions which are relevant
PRAGMATIC CONDITIONALS
5
to the present discussion. First of all, these four types of conditionals belong to two basic-level categories, namely logical conditionals (examples (4) and (5)) and conversational conditionals (examples (6) and (7)). Whereas the logical conditionals involve analytic reasoning processes, and as a result the antecedent can only be preposed to the consequent, conversational conditionals involve speech acts in actual discourse (6) or aspects of the discourse such as metalinguistic references to the linguistic choices made by the speakers (7) and the antecedent tends to be postposed. Secondly, the separate category of a logical inferencing conditional is not an additional complication or multiplication of categories, since we need the category of 'logical conditionals' anyhow because of the existence of a logical identifying conditional. We can therefore make a distinction between two types of inferencing conditionals, viz. inferencing course-of-events conditionals and inferencing pragmatic conditionals. The latter is much closer to the 'identifying' pragmatic conditional than to the inferencing course-of-events conditional, because both pragmatic logical conditionals express merely reasoning operations, but do not involve any actual observed reality. Thirdly, the diversity of subcategories within the two basic-level categories of logical and conversational conditionals and the conceptual difference between these two basic-level categories themselves calls for the introduction of a unifying superordinate category, which groups these course-of-events and hypothetical conditionals. The term 'pragmatic of the four subcategories: as will be shown in the next sections, they are all, each in its own way, more strongly either 'speaker-oriented' or this respect, pragmatic conditionals contrast very strongly with the primordially ideational functions associated with the use of course-of-events and hypothetical conditionals.
Having established these internal commonalities and distinctions in the area of pragmatic conditionals, we will now try to specify the conceptual link between the degree of independency of the antecedent and the consequent, on the one hand, and the functions of the four types of How great is the degree of independency between antecedent and consequent in logical pragmatic conditionals? How much smaller is it in comparison with hypothetical conditionals and course-of-events conditionals? As shown in Athanasiadou and Dirven (1997), hypothetical conditionals are
subcate conditio
'hearer
pragm
6
ANGELIKI ATHANASIADOU & RENÉ DIRVEN
the central or prototypical member of the class of three types since they show the highest level of dependency between antecedent and consequent. In course-of-events conditionals the dependency is still present but less co-occur but they are not explicitly linked in any causally or otherwise determined way. Let us examine somewhat more closely the degree of dependency implied in sentence (1) and contrast it to the logical pragmatic conditional in example (5), both of which are repeated here for the sake of convenience:
pr
(1) He looked at his watch; if the soldier was coming, it was nearly time. (5) If termites are the equivalent of antelopes, army ants are the equivalent of beasts of prey. The dependency between the two events in (1), viz. the soldier coming and it nearly being time is based on recurring experience, that is to say, on observation of regularly, consecutively occurring events. The dependency between the two events in an inferencing course-of-events conditional like (1) is greater because the construal of this dependency is based on repeatedly observed facts. In an inferencing pragmatic conditional like (5), the dependency between antecedent and consequent is much weaker because the link is based upon a construal of analytic definitions: termites are only comparable to non-carnivorous antelopes from a very special, namely nutritional point of view, and on the basis of that specific parallel we infer a similar comparability between army ants and carnivorous beasts of prey. It is even the case that in (5) the antecedent and the consequent are theoretical constructions in the mind which in themselves have no validity and thus cannot be dependent on one another. In its simplest form, such parallel between a=b and c=d created by the logical inferencing ifconstruction that makes some slight form of dependency plausible. In the scale of dependency of antecedent and consequent — as already shown in Athanasiadou and Dirven (1996, 1997) for hypothetical and course-of-events conditionals — we can see the following global picture of decreasing dependency for all the types of conditionals:
construc
PRAGMATIC CONDITIONALS
7
Table 2. Scale of (in) dependency between antecedent and consequent in conditionals
3.
Identifying conditionals The above scale of (in)dependency of conditionals in Table 2 implies that identifying pragmatic conditionals show a relatively greater dependency degree than inferencing pragmatic conditionals. This has not been (4) If there's one human species that ought to be put out to pasture, it's Presidents and Prime Ministers. Such an identifying conditional consists of two parts: an antecedent which functions as an identifying description and a consequent which is the identified one or the identifier. The matching of the entity to be identified and the identifier depends to a large extent on the detailed description in the antecedent. Although there is no causal relationship between the antecedent and the consequent, the two parts are strongly interrelated: the description in the antecedent is made to fit the identifier in the consequent. But in actual fact, things are more complicated. Although identifying pragmatic conditionals seem to function in order to identify the unknown or
demon
8
ANGELIKI ATHANASIADOU & RENÉ DIRVEN
as yet unidentified member of a category, this may be a very misleading interpretation of their communicative function. What the speaker is doing by means of the identifying construction is not only just to accomplish an act of revealing an entity's identity, but rather to create expectations, which is done by a very long description of the entity to be identified: (8) Now, if any part of the Bible is assuredly the very Word of God speaking through His servant, it is John's Gospel. [BRO:91] Thus, the identifying pragmatic conditional assumes a strong rhetorical function emphasising the important features of the category to be identified and raising the expectation for the identity of the variable of this category precisely by postponing the naming of it. Preposing of the consequent would therefore not make any sense in identifying pragmatic conditionals. Our thesis that the main function of the identifying pragmatic conditional is not necessarily that of revealing someone's identity can also be shown by the fact that the construction can be used with a negation thus rather excluding possible candidates than carrying through an identification process: (9) Luckily both women knew my position and if anyone suffered in their opinion it was not I. [BRO: 104] The rhetorical use of the identifying pragmatic conditional can even be strengthened by the use of adverbs like ever in the antecedent and the reduction of the consequent to a very short formula, as in the following examples: (10) If ever a rifle met the needs of the whitetail hunter, this is it. [BRO: 189] (11) Brains and beauty, high position in both the social and intellectual worlds, athlete, fabled lover — if ever the world was any man's oyster it was his. [BRO: 191] As with most pragmatic phenomena in language, here too we are left again with the rather predictable conclusion that there is no one to one relationship between the syntactic form of this construction and its various pragmatic uses. The syntactic form typically serves an identifying function, but the pragmatic use overrides this syntactic function, as is usually the case
PRAGMATIC CONDITIONALS
9
in the pragmatic use of language. This conclusion contains another argument for categorising identifying pragmatic conditionals as a member of the category of pragmatic conditionals and not as a separate category. Whereas the examples (8) to (11) contain cases of subject sentence of the antecedent may be selected for the identification process. The entity to be identified may be the topic of the sentence (12), the location (13), or the reason why something is the case (14): (12) If there was one thing Julie couldn't be doing with, it was anonymity. [COB:20] (13) If there's trouble in a home, it isn't always in the bedroom - it's quite often in the budget. [COB: 19] (14) If, indeed, we are inclined to deny one particular thing, this is because of the failure to recognize some other aspect of it. [LOB:3, adapted] Sentence (12) illustrates a colloquial and popular means of using identifying conditionals: it is purely emotional. Sentence (13), does not seem to provide an identification of what the main marriage trouble might be, but rather an identification of where the trouble might have to be situated, whereby the locatives in the bedroom and in the budget are metonymically used for the sexual aspect of marriage and for its financial management, respectively. In sentence (14), the identification process is related to a subclause denoting reason. In the identification formula this is because, the pronoun this indeed refers to the whole of the antecedent. So the pragmatic type of identifying conditional in (14) follows the pattern 'if a, then because of b'. From a cognitive point of view, however, examples (12-14) represent a borderline case of identifying conditionals in the sense that they form a transition from the process of subject identification to a process of emphasising a particular constituent in the sentence. Still in all the examples discussed in this section, the dependency between the antecedent and the consequent is manifestly, though weakly present. 4.
Inferencing conditionals Identifying conditionals are used for other communicative purposes than mere identification. Similarly, inferencing pragmatic conditionals are not only used to make inferences, but also and perhaps even more
identifi
10
ANGELIKI ATHANASIADOU & RENE DIRVEN
importantly, to emphasise the inferencing force of the utterance, which again foregrounds the pragmatic character of the construction. One of the many items emphasising the 'inferencing' force of an inferencing logical then still preserves a strong deictic link with some previous state of affairs. This presupposes an interdependency between antecedent and consequent. If the interdependency is not strong enough, then is ruled out. Thus, it cannot be used in discourse conditionals such as (3). Hence we cannot say *Iƒ anyone wants me, then I'm downstairs. In this respect it is worth noting that there is a very strong contrast between inferencing pragmatic conditionals and inferencing course-of-events conditionals. In inferencing pragmatic conditionals the use of then may sometimes be absolutely indispensable as in the following example:
(15) If Arthur Williams was involved in the fraud or the murder, then he too had another identity. [BRO: 114] This use of the logical operator then can be further clarified by its compatibility with the epistemic auxiliary must, i.e. 'then he too must have had another identity'. Without then it is much more difficult to interpret this inferencing pragmatic conditional since we would not be inclined to stress the pronoun form he and make it refer back to Arthur Williams. In inferencing course-of-events conditionals, on the other hand, such as the soldier example in (1), the inference is drawn without any further focus on the inferencing process as such.5 That may also be the reason why inferencing conditionals with then and without then do not always have the same implications. Let us, for instance, repeat the soldier example without and with then for the sake of easy comparison: (16) a. He looked at his watch; if the soldier was coming, it was nearly time. b. He looked at his watch; if the soldier was coming, then it was nearly time. In the course-of-events context of (16a) the conclusion follows from observation: the two events in the antecedent and the consequent are contiguous, in the sense that the one stands metonymically for the other. That's why both are factual: the soldier was coming and this was the sign that it was nearly time. The subject checks the time on his watch to confirm his factual inference. So both the antecedent and the consequent are based
condit
PRAGMATIC CONDITIONALS
11
on observational deduction. Here the observational conclusion seems to exclude the use of the logical operator then, which requires a fully different conceptual scene. The speaker of (16b) would not even need to see the soldier coming. Moreover, he could also use the form with must, i.e. then it must nearly be time. Consequently, in (16b) the soldier's coming is much less certain because it is no longer based on an observation of a course of events, but on a logical, truth-finding deduction. Moreover, (16b) now looks like an event only occurring once, whereas (16a) clearly implies a recurring event, which is precisely the basis for the inference. The use of then in (16b) changes the nature of an inferencing course-of-events conditional into that of a logical conditional. The possibility of using then is merely the tip of the iceberg in that it is but one of a whole range of epistemic, inferencing expressions, all of which stress the epistemic or truth-finding nature of the inferencing process and as such stress the pragmatic relationship between 'the sign' and һ user'. These epistemic expressions in inferencing conditionals may range from epistemic modal auxiliaries like may, must, should, have to, and paraphrasing constructions for modality such as it is possible, surely, it seems likely, etc. to explicit lexical expressions denoting the truth-finding process, such as (we) conclude, it follows that, is it any wonder, it stands to reason, you must admit, or negative conclusions like it does not follow, it is not necessary, it is strange, I see no reason, etc. Even a certain rhetorical flavour may be present occasionally, especially through the use of rhetorical questions introduced by why shouldn't, why not, how much less, etc. Of all these possibilities the preferred devices are modal auxiliaries or modal paraphrases: (17) On the other hand, if there have been no signs of active infection for some time, the murmur may be due to old scars left over from a previous attack. [COB: 106] (18) If this is so, why is it not also possible that they achieved their original dominance by stealth and cunning and bloodshed.... and not, as they claim, by Heaven's will? [COB:336] These modal auxiliaries are, more often than not, combined with other inference-focusing devices such as I think it should, should+progressive, there may be some reason, then...should, etc.
12
ANGELIKI ATHANASIADOU & RENÉ DIRVEN
(19) If arranging for the girls must take time into account, I think a day or two should be enough to finish our business. [BRO: 112] (20) If, as Reid says, nearly all his poetry was produced when he was not taking opium, there may be some reason to doubt that he was under its influence in the period from 1896 to 1900. [BRO: 119] The inference-focusing devices can be even more elaborate and contain a certain degree of redundancy: (21) Nor is it necessary to hold that if a man wants to get food, he must be tempted to steal. [LOB:4] (22) It stands to reason that if a horse is too backward to race during his first season in training, he is most unlikely to be sufficiently mature to beat the best of his generation in the late May or early June. [LOB:81] The phrases in (21-22) are characteristic of the jargon of logic. In more colloquial language the inference-focusing devices tend to be less elaborate, as the following expressions from the Leuven Drama Corpus show. The expressions found here are: I'd say, is it any wonder, I see no reason, it follows, etc. (23) If I saw a man eating grass I'd say he was hungry. [LDC:31] (24) If he fails to see me in a room, is it any wonder he has difficulty finding the enemy in the whole of Flanders. [LDC:70] (25) If the sun never sets on your empires, it follows that it never rises either. [LDC:109] It turns out that not only the logical adverb then, but several other colloquial devices such as so, or after all, or whole phrases like you must admit, etc. are devices emphasising the act of inferencing. (26) So there is a God after all, if it is reason. [LDC:93] (27) If you persist in maintaining the almighty Jove aimed a thunderbolt at me, then, you must admit he had the unusual experience of missing. [LDC:102]
PRAGMATIC CONDITIONALS
13
But most clearly, the rhetorical force of an inference-focusing construction is found in an interrogative conditional· (28) If birds can fly, why shouldn't man? [LDC:65] (29) If he is a man, how much less of a man am I? [LDC:72] The analysis of the many examples in this section was necessary to pave the way to one important conclusion. The linguistic realisation of the process of inferencing includes more than mere inferencing, that is, drawing all sorts of logical consequences. It also includes an awareness of the as in example (2) If she 's divorced, (then) she 's been married, is rather the exception, not the rule, since it only suggests a minimum of focusing. The rule is rather that the speaker uses numerous inference-focusing devices to refer to the ongoing process of inferencing. In this respect, the speaker is signalling very explicitly that he is engaging upon making the inference and thus these inference-focusing devices are a pragmatic strategy to signal the pragmatic force of the utterance. Of course, such pragmatic self-monitoring is not to be confused with performative acts, but these devices reveal that speakers' interests are far beyond making purely analytic inferences. The many inference-focusing devices also invite us to use the term 'pragmatic conditionals' in a much wider sense than that of conditional speech-acts, as found in the literature, (e.g. Sweetser 1990:118). Whereas traditionally the concept 'pragmatic' was limited in its application to performing speech acts, presuppositions, hedging, etc., we now propose to use it again in the wider, original — as intended by Morris — sense of the relationship between the sign and its user, so that it does not only cover the things we do, but also the things we think and say, while performing speech acts or any other communicative acts. Moreover it comprises various sorts of hints at the thought processes behind our speech acts, at the special appeals to the hearer or at comments on the speaker's choice of certain wordings. 5.
Discourse conditionals Discourse conditionals link the consequent to some or other, usually hearer-oriented, pragmatic factor in the conversation. The split between the antecedent and the consequent is extremely great. In English, this split cannot show up in different word orders in antecedent and consequent, as in German and Dutch, but only in the intonation pattern. Thus, discourse con-
14
ANGELIKI ATHANASIADOU & RENÉ DIRVEN
(i) (ii) (iii) (iv) (v) (vi) (vii)
no possibility of using then no hypothetical forms no change of tenses (see Sweetser 1990; and see (31) below) intonational 'caesura' no explicit use of performatives very great ambiguity of possible speech-act forces (see (32) below) preferred order: antecedent before consequent Table 3. Characteristics of prototypical discourse conditionals
ditionals show a very important 'caesura' between the antecedent and the consequent. Intonationally, this 'caesura' is shown by the fact that one cannot read or pronounce a discourse conditional (30a) in the same way as a hypothetical conditional as in (30b) below: (30) a. If you're hungry, there's food in the fridge. b. If you ask him kindly, John will help you. If one reads the hypothetical conditional in (30b) with a 'caesura' or an intonational break, an odd effect is achieved, but in (30a) it is quite normal. Conversely, (30b) must be read as one intonational unit but if one applies this intonational pattern in the discourse conditional in (30a) the result is very odd. The discourse conditional is much more a construction sui generis, i.e. a highly idiosyncratic category with very specific characteristics of its own, summarised in Table 3, which can also be seen as a summarising definition of the prototypical discourse conditional. (31) ?If you were hungry, there was food in the fridge. (32) If anyone wants me, I'm downstairs. The question mark in front of the sentence in (31) means that this sentence cannot be given the interpretation of a discourse conditional. But in itself the sentence is not impossible; on the contrary, it is a very clear instance of a course-of-events conditional which is to be interpreted as 'Whenever you were hungry, there was always food in the fridge'. In (32) the ambiguity of the possible conversational implicature of the speech act is very great. Possible candidates are 'please tell X', 'don't tell X', 'I don't want to be disturbed', 'they can find me', etc.
PRAGMATIC CONDITIONALS
15
5.1. The great variety of discourse conditionals Most discourse conditionals exhibit the following central or prototypical features of their category: they have a very clear 'caesura', they use present tenses in both antecedent and consequent, they are of the affirmative type and they have the sequence antecedent-consequent. However, structural possibilities with discourse conditionals, which differ in some interesting respects from the prototypical ones discussed before. (i) The first discourse conditional ever pointed out is found in Austin's paper "Ifs and cans" (1961:210-12) and its structure has the reversed order of a prototypical discourse conditional, i.e. consequent before antecedent: (33) There are biscuits on the sideboard if you want them. Not only is there a reversed order of consequent and antecedent in this example, but there is also absence of a strong 'caesura'. Here in the marked order of consequent and antecedent, this 'caesura' can almost be reduced to zero. This may be due to the fact that with a preposed consequent, as in (33), there is no communicative need for an intonational break, since the //-clause cannot be misunderstood now. But note that here, too, at the end of the antecedent, there is a slightly rising intonation contour, typical of questions or offers. What is even more striking is that instead of a postposed antecedent, one can use an intonation question or any other type of question as in (34): (34) There are biscuits on the sideboard. You want some? This possibility of using two independent sentences instead of a discourse conditional is a very strong indication for the very low degree of dependency between the antecedent and consequent in this type of pragmatic conditional. (ii) A second structural variant found with the discourse conditional is the interrogative form, as in the following example: (35) If it is found that in recent years the habit of discussing with the chair has increased, may I ask you not to hesitate to say so? [LOB: 131]
in co
16
ANGELIKI ATHANASIADOU & RENE DIRVEN
The hearer-oriented character of the discourse conditional triggers the use not only of you forms, but also of invitations, requests, etc. (iii) Another pragmatic form frequently and commonly used to express a discourse conditional is a why not-question in the main clause, as in (36): (36) a. If he attaches little importance to personal liberty, why not make this known to the world? [BRO:243] b. If anti-Semitism was on trial in Jerusalem, why was it not identified? [BRO:78] The negated why-question with infinitive in (36a) has an urging force and comes close to a challenge to do something about some negative discoveries. But in (36b) the same negated wh-question is not a challenge to do something about a situation but, since it is in the past tense, it is rather ambiguous between a neutral information question and a reproach on moral grounds. Here the position of the negation adverb not is very significant: it can remain unstressed and then it is an information question, but when it is stressed, it becomes a reproach. But if it is used as a contracted form with the auxiliary (why wasn't it identified?), the contracted form rather suggests the reproach meaning, i.e. 'it should have been identified', as the preferred interpretation. (iv) An equally common structural realisation of the discourse conditional is an imperative form in the consequent, which is not only postposed as in (37a), but probably more often preposed, as in (37b,c). (37) a. If it is hallucinations you are after, get drunk. [LDC:96] b. Keep them, if you feel so lost. [LDC:52] c. Prove it if you can. Such an imperative form in discourse conditionals may express many different speech-act forces, such as letting someone down or cursing (get lost, go to hell), typically found in postposed position (37a), and acts of reassuring (37b), challenging (37c), etc. typically found in preposed positions. (v) Other structural possibilities are offered by the combinations of different tenses in the antecedent and the consequent. In addition to the prototypical pattern 'present + present' in antecedent and consequent there
PRAGMATIC CONDITIONALS
17
are also the unexpected combinations of 'past + present', 'past + past', 'past + past perfect', etc. Most of these tense combinations are motivated by the different times referred to in the antecedent and in the consequent: (38) Castle thought: that's a bad slip if the telephone call this morning was from the office. [COB:275] Here, the different use of tenses (present in the consequent + past in the antecedent) is motivated by the different moments when the evaluation occurs (consequent) and when the event evaluated took place (antecedent). Since speech acts are grounded in the present time of the speech act itself, acts of offering, inviting, challenging, evaluating tend to occur in the present tense, and they do not normally occur with past tense forms hungry, there was food in the fridge. 6
Metacommunicative conditionals
The function of the if-clause in a discourse conditional is force expressed in the antecedent is relevant. In this respect a differs markedly from discourse conditionals:
he metacommunicativ
(39) a. ...and if I might say so at this stage, that is one of the sentences it may impose on a young male in prison. [Haegeman 1984:487] b. If I may say so, that's a crazy idea. [Sweetser 1990:118] Here the link between the antecedent and the consequent is a preparatory link: the if-clause has the function of softening the possible unpleasantness of the statement of the consequent. Besides softening there may be many other metacommunicative functions of pragmatic if-clauses, a very frequent case being that of relativising one's presuppositions in the making of the speech act. Therefore we can define a metacommunicative conditional as the pointing out of some aspects of the global communicative act which need special attention. Consequently, these ¿/-clauses can be seen as comments on various aspects of the communicative act and are therefore metacommunicative in nature.
18
ANGELIKI ATHANASIADOU & RENÉ DIRVEN
6.1. Various uses of metacommunicative conditionals Since there are so many different types of communicative acts, of which speech acts are just one, though very important, subcategory, it may be advisable to make further distinctions between a number of metacommunicati the whole speech act is concerned we can speak of the metapragmatic use of the conditional. If separate aspects are involved, we can distinguish between a metalinguistic or a restrictive use of the conditional; the //-clause relates either to the 'wording' of the speech act, i.e. the locution used by the speaker or to the presuppositions made by the speaker. In the former case we have a metalinguistic use — as this term is traditionally understood in linguistics — and in the latter we usually have a 'restrictive' use. In this section we therefore address the following three functions of metacommunicative
(i) The metapragmatic use of metacommunicative conditionals, (ii) The metalinguistic use of metacommunicative conditionals, (iii) The restrictive use of metacommunicative conditionals. (i) The 'metapragmatic * use of metacommunicative conditionals Alongside the stereotypical formula if I may say so in (39b), we find many other metapragmatic references to the appropriateness or distance may be too great for the speaker in a lower hierarchical position to perform a blunt speech act of evaluation, and therefore he must 'preface' his evaluation by means of a softening metapragmatic cajoler, as in (40):
inappropri
(40) Good gracious me, sir, if I may make so bold - it's a bit shocking, isn't it? [LDC:86] This comment in the if-clause is, in fact, just as in the two examples in (39), a parenthetical construction, which can as easily be left out as not, and which therefore further emphasises the fact that the metacommunicative function is not an essential part of the utterance. This does not mean of course that parenthetical constructions are always optional, e.g. the following construction if you'll excuse us could not be omitted because this would produce an awkward, if not impossible, utterance:
m
PRAGMATIC CONDITIONALS
19
(41) a. Splendid! Now ladies, if you'll excuse us, we have a lot to do. [LDC:54] b. ** Splendid! Now ladies, we have a lot to do. Here the if-clause in fact expresses the speech act of asking for the permission to leave and the consequent gives the reason for this. The metacommunicative conditional may also be used (42a), or to challenge the hearer to use the correct procedure to find
metapragmatica info
(42) a. I've always been interested in people, sir, if that's what you mean. [LDC:107] b. It's all on the back pages of the paper, just before the sport, if you know where to look. [LDC:16] (ii) The 'metalinguistic ' use of metacommunicative conditionals Whereas in the above cases the metacommunicative conditional relates to the speech act as a whole, in the following cases it has a more limited function, and relates only to particular aspects of the speech act, such as to its formal aspects, or more importantly, its functional aspects. Speakers of a language are aware of performing a large number of metalinguistic acts. But, in fact, very few examples occur in the selected corpora: (43) a. I've come to offer my congratulations, if thaťs the right word. [LDC:42] b. He was beginning to feel sure that she had recognised him and wanted to talk but could not summon up the courage - if courage was the word. [LOB:180] As these examples illustrate, the metalinguistic reference always concerns a single word or expression, and therefore the metalinguistic reference is made by means of an anaphoric pronoun such as that (43a), or by the full word or expression courage, as in (43b). This formal device also shows the enormous difference between the metapragmatic use (42) of the metacommunicative conditional, which can have any possible structural form, and its metalinguistic use which can only have a nominal form such as that in (43), because it must refer to some element in the consequent.
20
ANGELIKI ATHANASIADOU & RENÉ DIRVEN
(iii) The 'restrictive' use of metacommunicative conditionals When uttering speech acts the existence of (some of) the referents of the noun phrases used, or of the events described are presupposed. Metacommunicative conditionals are often used to draw attention to these presuppositions and are typically used either to strengthen the belief of the existence of the referents (assertive use of some or one) or, on the contrary, not to commit oneself to the existence of the referent (non-assertive any) and to restrict the validity of such a belief. These opposite functions are expressed by the parenthetical formula if there is one for the assertive attitude (44), and if there is any for the non-assertive attitude (45): (44) a. This is a time — if there ever was one — for parents to show their thoughtfulness and generosity towards each other. [COB: 117] b. That, I smarted, is a royal rebuff if ever there was one. [BRO: 192] (45) The shooting-season opens Saturday and the birds'11 be scattered all over the place after that — if there's any left. [COB: 165] In (44a) the main clause implies that now the occasion is present, so the speaker takes an affirmative stand to the existence of such a single referent. This attitude completely changes if one is replaced by any. Then the assertive attitude becomes non-assertive, and the speaker no longer commits himself to the existence of such a referent. In (45) the use of any is the only possibility because the non-assertive suggestion is that there might not be a single bird left. In addition to these two possibilities of the metacommunicative conditional, its non-assertive use seems to be the only possibility in the elliptic formulaic expression if any, as in (46): (46) a. Secondly, we must pool and ration our supplies of food, if any. [LOB:64] b. The change, if any, in foreign policy will consist rather of a freshness of approach. [LOB:98] The elliptical use of if any in these two examples relates to the non-affirmed existence of the referents of the noun phrases supplies of food in (46a), and change in (46b). Structurally, the if any-phrase does not function as an antecedent to the consequent, but it is simply attached to the
PRAGMATIC CONDITIONALS
21
noun phrase it modifies, and therefore functions as a small afterthought related to the presupposition made by the use of the definite determiners in the noun phrases. When the presupposition is related to a locative expression such as here, the elliptic phrase uses the adverbial form anywhere as in (47): (47) For here if anywhere in contemporary literature is a major effort to counterbalance Existentialism. [BRO: 108] The use of if anywhere in (47) is ambiguous: it can either restrict the belief in the existence of a given place in contemporary literature referred to by here, or it can also function in a rhetorical way and stress the fact that it is precisely here that Existentialism will be counterbalanced. This rhetorical function has become the only use of the elliptical expression if anything, but now it is not the existence of the referents or the processes that is focussed upon; on the contrary, the phrase if anything no longer has a reference-modifying function, but rather a sentential function leading to some strengthening effect as in (48): (48) a. If anything, he had merely become a little more reserved, and much more polite. [LOB: 127] b. The effect of make-believe was, if anything, heightened by the arrival in the room of the German uniforms. [LOB: 126] Thus the phrase if anything has become an idiomatic expression relating to the sentence as a whole and meaning something like 'the least one can say is that...'. This idiomatic use may also be the explanation why the phrase if anything can be used at the beginning of a sentence, as in (48a) or with an almost adverbial function in mid-position between an auxiliary and a verb as in (48b). The elliptical use of the metacommunicative conditional occurs in the context of adjectives as in (49): (49) a. And I suppose now that you've finally grown up, if a little late, you'd go on producing kittens every six months or so. [BRO:20] b. As usual, delegates were entertained to dinner by the Royal Yachting Association and a very pleasant affair it was, if a little cramped in its temporary quarters. [LOB:87]
22
ANGELIKI ATHANASIADOU & RENÉ DIRVEN
(49) c. Hobhouse was Byron's intimate, if a little stuffy and unimaginative friend from Cambridge. [LOB:88] Here the elliptical construction is in fact a double elliptical construction: it is (i) syntactically elliptical as in the former cases since there is no subject or verb, but only an adjective and some adjuncts; and (ii) the conjunction if itself is an elliptical form of the concessive conjunction even if. Therefore, the elliptical construction does not relate to a utterance: the positive characteristic in the consequent, for instance you've finally grown up in (49a), is somewhat restricted by contrasting it with the concession in the antecedent that this achievement happened a little late. So examples (49a, b, c) illustrate a merely concessive construction rather than a metacommunicative conditional and rather fall in a different, though related category of conceptual constructions.6
presuppo
7.
Conclusions The above analyses have shown that although pragmatic conditionals cover a wide variety of specialized conditional uses, they also have very typical common characteristics as well as specific idiosyncratic ones.
(i) Pragmatic conditionals all have one thing in common. Their an of a communicative act contained in the consequent. In identifying conditiona rhetorical impact on the hearer in his communicative act. In inferencing conditionals the speaker uses all possible types of inference-focusing devices in order to emphasise the speaker's process of inferencing. In discourse conditionals, speakers express the many different conditions under which their speech act forces are relevant for the hearer. And finally, in metacommunicati highlighting some or other important aspects of their communicative acts. In spite of these common aspects of pragmatic conditionals, which consist in the emphasis put on either the speaker's or the hearer's role in the process of communication, their specific functions or uses stand out clearly, too.
(ii) A second common and distinguishing element between the four types of pragmatic conditionals is the relatively great to almost absolute amount of independency of antecedent and consequent. This is a common element in so far as it distinguishes pragmatic conditionals from courseof-events conditionals and hypothetical conditionals, but it also differentiates
PRAGMATIC CONDITIONALS
23
between the four subcategories of pragmatic conditionals, i.e. the distinction between identifying and inferencing conditionals on the one hand, and discourse conditionals and metacommunicative conditionals, on the other. This distinction is extremely great, since in the latter two (conversational) conditionals the if-clause can even be omitted. In its turn, the distinction between discourse conditionals and metacommunicative ones is again very great because the former have no formal link with the consequent whereas the latter always have a referential link. Of all the four pragmatic deviation from the prototypical member of the category of conditionals, viz. the hypothetical conditional, since it conveys the highest degree of be it intonational, syntactic, referential or pragmatic ones.7 (iii) Although the two conversational conditionals, i.e. discourse conditionals and metacommunicative conditionals, have some elements in common, the differences between the two have proved to be even greater than originally thought. The present analysis has shown that the biggest difference between the two is precisely what the term 'metacommunicative' suggests: in each metacommunicative conditional there is always a comment on the consequent and the communicative act contained in it. This pragmatic link is syntactically expressed by means of a reference device such as so, in If I may ask so, any in phrases such as if any, if anything, if anywhere, substitution elements such as one in If there is one or anaphoric elements such as that in If that's what you mean. Discourse conditionals contrast very strongly with metacommunicative conditionals in this respect: they do not contain any referential link, but only a relevance link, to the consequent. In fact, as will be remembered, they are characterised by an intonational and syntactic 'caesura'. The 'caesura' which reflects a major conceptual point of view. Anything that can be left out in the communicative transfer is both linguistically and conceptually less prominent. (iv) The more general conclusion to be drawn from the analysis of the above structural patterns is that authentic discourse yields many more structural possibilities of any type of conditional, or, more generally, of any type of conceptual structure, than could ever be identified by mere introspection. While analysing these many examples, it was not primarily our purpose to give an exhaustive list of all the types and features of pragmatic conditionals, but in the first place to show that any conceptual category has many more structural realisations than an approach solely based on introspection can guarantee. Probably introspection will or tends to
conditi
indep
distinc
24
ANGELIKI ATHANASIADOU & RENÉ DIRVEN
produce the prototypical cases only, and corpus-based examples will provide the whole range of structural possibilities. Both approaches complement one another, which is a conclusion everyone could probably agree with. Thus one could say that Austin, an introspectivist if ever there was one, put the pragmatic conditional firmly on the agenda. But at the same time, it is only thanks to corpus linguistics that the many diverse subtypes could be brought out. The ultimate purpose of this paper is to take this conclusion one notch higher: neither of the two approaches can do without the other. Without an extensive corpus study, descriptive linguistics would never have managed to single out a class of 'course-of-events conditionals' and similarly, we would never have arrived at the insight of the existence of a superordinate category of 'pragmatic conditionals', comprising a fairly heterogeneous number of subcategories. In fact, we might still be concentrating only on the most prototypical category, i.e. that of 'hypothetical conditionals'. Notes * We want to thank the two anonymous reviewers for their highly insightful remarks and queries. 1 It is, for instance, not discussed in Comrie (1986), Haegeman (1984), Sweetser (1990), Sweetser (1996), and Dancygier and Sweetser (1996). This latter paper discusses both (our) hypothetical and pragmatic conditionals, but not (our) course-of-events conditionals. This category is vaguely present in Dancygier and Sweetser (1997:123) under the name 'generic conditional' and exemplified by non-corpus-based examples such as If Mary bakes a cake, then she gives a party. 2 The four corpora used here are abbreviated as follows: BRO for Brown Corpus, COB for Cobuild, LDC for Leuven Drama Corpus, and LOB for Lancaster-Oslo-Bergen Corpus. 3 This term is one of a set of three terms, designating Halliday's functions of language use, i.e. the ideational, interpersonal, and textual functions. It is not implied that pragmatic conditionals solely serve the interpersonal function, but that they do so prototypically. The inverse applies to hypothetical and course-of-events conditionals, which mainly serve the ideational function, but allow many other indirect speech acts as well. Sweetser (1990) and Dancygier and Sweetser (1996:88ff.) even refer to hypothetical conditionals as 'content' conditionals, which is an even stronger ideational characterisation. 4 Although it is not explicitly stated, it looks as if Dancygier and Sweetser (1996, 1997) still keep this position. 5 The function of then in conditionals is extensively discussed in Dancygier and Sweetser (1997) and is seen as a deictic reference to a mental space built up by an if-clause, which must therefore precede the consequent. However, this analysis can at best only say when the consequent can take then, not what its presence actually adds to the meaning of the whole sentence. Starting from the iconic principle 'that more linguistic material also implies more meaning', we suggest in the text that the additional function of then is to highlight and to focus on the inferencing process. Thus in (16a) there is an 'inductive leap' from an observation to a conclusion, but in (16b) with the use of then there is a deliberate
PRAGMATIC CONDITIONALS
25
inference, which may also be marked by many other syntactic devices, as shown in the further discussion. 6 This view is elaborated in Dirven (1997). 7 This conclusion seems to be somewhat in conflict with the scale of dependency in Table 2 where the metacommunicative conditionals are claimed to have the lowest degree of dependency. However, there is not a real contradiction, since in Table 2 we look at the relationship between antecedent and consequent. Metacommunicative conditionals do not relate to the complete relationship but only to one element in the antecedent. But the link with this selected element is much stronger than the link between the antecedent and the consequent in a discourse conditional.
References Athanasiadou Angeliki and René Dirven. 1996. "Typology of if-clauses". Cognitive Linguistics in the Redwoods ed. Casad Eugene, 609-654. Berlin: Mouton de Gruyter. . 1997. "Conditionality, hypotheticality, counterfactuality". On Conditionals Again eds. Athanasiadou Angeliki & René Dirven, 61-96. Amsterdam & Philadelphia: J. Benjamins. Austin, J.L. 1961. 'Ifs and cans". Philosophical papers eds. J.O. Urmson & G.J. Warnock, 153-180 (3rd edition, 1979). Oxford: Oxford University Press. Comrie, Bernard. 1986. "Conditionals: A Typology". On Conditionals eds. E.C. Traugott et al., 77-99. Cambridge: Cambridge University Press. Dancygier, Barbara & Eve Sweetser. 1996. "Conditionals, Distancing, and Alternative Spaces". Conceptual Structure, Discourse and Language ed. Adele E. Goldberg, 83-98. Stanford, Calif.: CSLI Publications. . 1997. "Then in Conditional Constructions". Cognitive Linguistics 8:2. 109-136. Dirven, René. 1997. "The conceptual space between conditionals and so-called concessives". Grammar and Text in Synchrony and Diachrony. In Honour of Gottfried Graustein eds. Mechtild Rheinhardt & Wolfgang Thiele, 77-99. Frankfurt: Vervuert / Madrid: Iberoamericana. Fauconnier, Gilles. 1994. Mental Space. Aspects of meaning construction in natural language. Cambridge: Cambridge University Press. Haegeman, Liliane, 1984. "Pragmatic Conditionals in English". Folia Linguistica 18.485-502. Morris, Charles. 1946. Signs, Language and Behavior. New York: Prentice-Hall. Sweetser, Eve. 1990. From Etymology to Pragmatics. Cambridge: Cambridge University Press.
26
ANGELIKI ATHANASIADOU & RENÉ DIRVEN
. 1996. "Mental Spaces and the Grammar of Conditional Constructions". Spaces, Worlds, and Grammars eds. Gilles Fauconnier & Eve Sweetser, 318-333. Chicago & London: The University of Chicago Press. Traugott, Elizabeth , Alice ter Meulen, Judy Snitzer Reilly & Charles A. Ferguson, eds. 1986. On Conditionals. Cambridge: Cambridge University Press.
How Polish Structures Space Prepositions, Direction Nouns, Case, and Metaphor Barbara Dancygier Simon Fraser University, Burnaby, Canada
1.
Introduction Numerous studies have now shown that spatial construais in different languages can rely on the cooperation among various subsystems of the language (cf. Talmy 1983, Ameka 1995, Sinha and Kuteva 1994). Even in languages in which prepositions are primary carriers of spatial information, their interaction with other subsystems is an important aspect of the resulting spatial interpretation (cf. Smith 1987, 1993, 1995, Serra-Borneto 1997). In this paper I will present some data from Polish to show how spatial construais result from an interaction of three major subsystems: direction nouns, prepositions, and the case form which is marked on the noun, following the choice of a given preposition. In Polish, as in a number of languages, prepositions make the most significant contribution to building spatial construais and, not surprisingly, they also display a variety of meanings and uses. At the same time, the main orientation axes ('up/down' and 'front/back') are not expressed by prepositions, but by nouns (which I will call 'direction nouns'). The direction nouns with prepositions, so that the two subsystems often interact closely. The third subsystem which contributes to constructing spatial similar to Smith's 'two-way' prepositions, in that the nouns following them can be marked with two different case endings, and that each choice of case results in a different spatial interpretation of the whole expression (cf. Smith 1987, 1993, 1995, Serra-Borneto 1997). Cases themselves can thus be claimed to carry their own meaning, possibly the kind of meaning relevant to spatial interpretations (as regards meaningfulness of case, it has been proposed independently of prepositional contrasts by Smith 1987, Langacker 1991, Janda 1993). On the other hand, at least one case in Polish, the instrumental, can express spatial information independently of prepositions, as well as in interaction with them. It can also be marked on the direction nouns. It seems, then, that, although the three subsystems are mostly seen
in
info
28
BARBARA DANCYGIER
in interaction (preposition and case, preposition and a direction noun and case, a direction noun and case), direction nouns and case endings can make their meaningful contributions to spatial interpretation independently of the remaining subsystems. In Polish, as well as in other Slavic languages, spatial information can also be found in verbal prefixes. The prefixes are often similar to function is that of providing aspectual information. As was shown in Janda (1986) on the example of some Russian prefixes, the numerous meanings of such prefixes can be related to their spatial source, but their representation of spatial information is not quite straightforward and regular. If they are used to represent spatial information, they do not seem to interact with the remaining subsystems; in fact, the same information is often repeated in a preposition quite like the prefix itself. The analysis of verbal prefixes thus seems to fall beyond the scope of the present paper.
prep
In what follows I will also not attempt to tackle individual prepositions of Polish in any way; rather, I will try to show how broadly defined groups of prepositions interact with the other systems in building spatial construais. In other words, I will argue that spatial construais are motivated compositionally, and that the role of prepositions here is significant, but not unique. The first section will show the ways in which direction nouns on the role of case marking options, primarily in prepositional phrases, but, in the case of the instrumental, also independently of prepositions. Finally, I will consider some metaphorical mappings motivated by the spatial construais reviewed in the main body of the paper. 2.
Direction nouns The spatial nominals that I refer to as 'direction nouns' are used to describe regions in space or regions of landmarks and usually come in pairs representing the contrasting directions along major orientation axes. For example, nouns such as góra and dół ("up" and "down"), or przód and tyl ("front" and "back") are used, with prepositions and appropriate case of objects with respect to the observer, as in (2), or parts of objects, as in (3):
co
HOW POLISH STRUCTURES SPACE
29
(1) Poszliśmy w górę / w dół We-walked in up-ACC / in down-ACC "We walked up/down" (2) Miasto zostało w tyle City stayed in back-LOC "I left the city behind me" (3) Kino jest na dole Cinema is on down-LOC "The cinema is on the ground level/downstairs" The two orientation axes exemplified above seem to be the most salient in Polish, since the direction nouns marking the contrast are used in the greatest number of 'preposition + noun' combinations, and all of the case-marking options are also used. Another orientational nominal commonly used is the noun środek ("middle" and "the inside"), which used with a number of prepositions, as in w środku (in middle-LOC, "in the middle" or "inside") or do środka (into middle-GEN, "into the middle"). Also the noun bok ("side") has a number of similar uses, even though it is not clear whether środek and bok make a contrasting pair in the same way as góra and dół.
rep
As regards other spatial nominals, however, they are used with much less regularity and their use with prepositions is much more conventiona in use than the nominals mentioned above, and its natural counter-part denoting the exterior part of a landmark, the noun zewnętrze, is perceived as archaic and not used at all. Interestingly enough, the exterior region is commonly described by an adverbial expression composed of a preposition and an adverb: na zewnątrz (on exterior-ADV).
The left/right orientation axis is also represented by highly conventionalized ex so the only expressions available rely an adjectives and adverbs, as in po lewej stronie (on/at left-ADJ-LOC-SING-FEM side-N-LOC-SING-FEM, "on the left") or w lewo (in left-ADV, "to the left"). The pairs of direction nouns which represent the up/down and front/back orientation are also conventionalized to some degree and express essentially adverbial meanings, as can be seen in examples (1-3). Neverthe-
BARBARA DANCYGIER
less, the contribution of all elements of the phrases (that is, of prepositions, case markings and particular nouns) matches the use of the same elements in spatial expressions which are less conventionalized or not built in the same way in expressions with all nouns, not just direction nouns. Direction nouns have a special role only in the sense that certain basic spatial concepts (e.g., up/down or front/back orientation) are not systematically expressed by other lexical elements. Contrasting pairs of direction nouns divide space into complementary regions. The nature of such regions in space does not have to be specified any further, which seems to be the case in (1). All the spatial information provided by the sentence can be summarized by saying that the participants of the walk entered a spatial region above or below their position prior to walking. Also, the boundary between the complementary regions is defined with respect to the position of the moving subject, so that whatever counts as 'up', involves the subject moving or looking upwards. A very similar construal involving front/back orientation seems to be offered by (2). In the case of (3), where the regions are defined within a three dimensional, bounded object (a building) the object may impose a bounded, possibly three-dimensional interpretation on the regions within it. Part of the problem with coming up with a more specific interpretation of (3) seems to be the use of the preposition na ("on"). As it was observed by going far beyond the central uses of similar prepositions in other languages. For example, on is most commonly associated with two-dimensional regions and with concepts such as 'support' or 'attachment'. In Polish, however, na has been found to invoke the concept of a three-dimensional region (at least in the sense in which the boundaries are boundaries of volume, as in the phrase na rynku ("in the market"). Also, in absence of a preposition like at (as described by Cuyckens 1984) in Polish, na seems to be developing into a preposition signifying "general direct location" (Cienki 1989). characteristics of the spatial region represented by the direction noun are specified with respect to the preposition used and the actual object described. This seems to confirm the claim proposed above that spatial construais arise on the basis of information offered by all the relevant subsystems, rather than by any of the systems alone.
At the same time, some of the direction nouns have been extended to denote features of landscape which are primary examples of the spatial concepts in question. For example, a noun like góra ("up") is also (or
convent
Sysak-Bo
Interpr
HOW POLISH STRUCTURES SPACE
31
perhaps first of all) used in the sense of "a mountain", while dół ("down") also means "a ditch", "a hole in the ground". These nouns, however, are used with prepositions and case according to the general rules, that is, in the same way as nouns not related to direction nouns. 3.
Prepositions and case Case plays a significant role in spatial construais in Polish, first of all because the choice of case marking may distinguish different uses of prepositions. More specifically, a given preposition may be used in two different spatial construais and the choice of the actual interpretation will depend on the choice of case marking on the noun following the preposition. Smith (1987, 1993) describes a similar phenomenon in German and refers to such prepositions as 'two-way prepositions'. Smith's interpretation of case after two-way prepositions attributes different semantics to the cases under analysis, and offers an account of spatial construais which shows how the two subsystems of German complement each other in the construction of spatial information. In what follows (section 3.2), I will try to show how two-way prepositions and their accompanying cases create spatial information in Polish. The meaning of case outside prepositional phrases has recently More specifically, the semantics of case in Slavic languages came under a close scrutiny in a number of studies (to mention only Dąbrowska 1994 and 1997, Janda 1993, Rudzka-Ostyn 1992, Brecht and Levine 1986). It is interesting to note that some uses of case outside prepositional phrases also give rise to spatially relevant construais. One such instance is the use of the instrumental, which will be discussed in section 3.1 below. The instrumental, as well as other Slavic cases, has been shown to have a variety of meanings. However, the scope of this paper will allow me to discuss only those uses which contribute to the construction of spatial information. 3.1
The spatial use of the instrumental The Polish instrumental seems to be the most interesting of the cases which have meanings related to spatial construais. The reason is that the instrumental can be used to represent spatial information without the help of a preposition, but it is also found as one of the choices following a two-way preposition (to be discussed in section 3.2) and that the two uses seem to be related. The concept most commonly associated with the instrumental is that of a 'path'. For example, when we use direction nouns in the instrumental
in
32
BARBARA DANCYGIER
case, as in (4), the meaning is that of a path allowing motion within or through the region designated by the noun. (4) Marysia poszła do jeziora górą, a ja dołem Mary went to lake up-INSTR, while I down-INSTR "Mary took the upper path to the lake, while I took the lower
path" Sentence (4) presents a situation where the area between the point where the two people separate and the lake which they are both trying to reach is divided into an upper and a lower part (perhaps this means going over the hill or along the foot of the hill). It is not in fact implied that there are specific paths or trails to choose between, rather, the options differ in altitude only (which may involve other concerns, such as the time spent hiking or the strenuousness of the walk, but this is not a part of the spatial construal here). The role of the instrumental case does not seem to be to 'draw' a path across the region, but rather to 'redefine' the whole region as a path. The use of the instrumental in the way described above seems to be motivated by its basic meaning of 'instrument' or 'manner'. In both cases the instrumental describes the 'way', or 'method' of accomplishing a goal or getting to a destination. In fact, when asking a question to which the instrumental-as-path is expected to be an answer, we would use the question word którędy?, which should be translated into English as "which way?". Interestingly, a similar motivation has been postulated by Janda (1993) to account for related uses of the Russian instrumental. This 'region is path' meaning of the instrumental remains unchanged if the case marking appears on other nouns denoting regions. The contrast between the uses involving the instrumental and those involving prepositions and other cases confirms the interpretation of the instrumental as an forest-INSTR, "We were/started walking through the forest") the speaker communicates that the people referred to in the subject were walking and that there was forest where they walked, but not that they entered the forest and were crossing it. The latter interpretation, indicating a specific, ("across"). Interestingly enough, region nouns marked with the instrumental are often used in the plural, as in Poszliśmy lasami (We-walked-PERF forests-INSTR, "We were/started walking through forests"). The intro-
goal-ori
HOW POLISH STRUCTURES SPACE
33
duction of the plural has the effect of broadening the region, and is usually interpreted to mean that the walk took a long time, and/or that the walkers were in and out of the forest as they walked. This addition, however, does not affect the overall 'region-as-path' construal. The instrumental 'path' is not necesarily more specified even if the noun denotes a region which is canonically used with a certain orientation. For instance, a noun like droga ("road") represents a region of a specific shape which is typically used to get from one place to another. To put it simply, roads are perfect examples of paths, and their users move either along them or across them. This is reflected in the fact that sentences such as Żołnierze maszerowali drogą (Soldiers were-marching road-INSTR, "The soldiers were marching along the road") are conventionally interpreted to mean that the soldiers were moving along the road. However, the instrumental itself does not imply the direction of the movement, as we can see in the sentence Wąż pełzł drogą (Snake was-crawling road-INSTR, "The snake was crawling on the road"), which does not carry any information about the actual path of the snake's motion. All we know is that the snake was moving within the region delineated as 'the road'. The instrumental is sometimes (though admittedly rarely) used to indicate motion with respect to concepts which are not easily construed as regions. One such example is the noun powietrze ("air"), which is in fact sometimes conceptualized as a region which can be entered (consider The bird flew up into the air). If one says Zapach dymu płynął powietrzem (Smell smoke-GEN floated air-INSTR, "The smell of smoke floated through the air"), there is no indication of motion along a specific path; still, as in the remaining cases, the region itself defines the path travelled. It is interesting to compare the uses of the Polish instrumental with those of the instrumental in Russian, as described in Janda (1993). Janda discusses the Russian instrumental of setting (which includes space and time), such that the case ending appears on the nominal denoting a setting in which a nominative acts on an accusative. In the most typical examples of the instrumental of space, however, the action is described by a verb of motion. In Polish, at least in the examples I have considered, motion (expressed most naturally by the verb) is an indispensable part of the construal. The instrumental in Polish could not be used to mark the setting for an action not involving motion. In fact, it is most naturally used in the cases where the nominative (the subject) is moving, and it is difficult to find examples where the accusative (the object) is set in motion and the instrumental specifies the path. Perhaps a sentence such as Generał posłał nasz oddział górami (General-NOM sent our unit-ACC mountains-INSTR,
34
BARBARA DANCYGIER
"The general sent our unit via the mountains") could serve as an example, but in fact the accusative (the soldiers in our unit) is involved in self-propelled motion, rather than set in motion by the nominative (the general). It seems, then, that the concept of setting is not specific enough to adequately describe the use of the Polish instrumental, which involves motion via a region. At the same time, it should be clear from the discussion above that the concept of 'region-as-path' is not rich in semantic detail either. In fact, the instrumental would be the likely form to be chosen in portraying motion only if certain more specific elements of the spatial construal are missing. For example, if crossing the boundaries of the region becomes a part of the construal, the speaker will probably choose the preposition przez ("through, across"), while the use of the preposition po ("on the surface of, along, all over") will be more likely if the region involved in the motion can also be distinguished in terms of a topmost surface, or if the motion does not cross the boundaries of the region. That is, prepositions przez and po will not, in all likelihood, be used to indicate motion through the air (no surface, no boundaries to be crossed), while the instrumental will be less likely if we are describing a mouse running away from the cat and using a table-top as its route of escape (clear boundaries, elevated surface). On the other hand, if the region in question is the forest, all three uses are possible. The phrase przez las (through forest-ACC, "through the forest") can indicate either crossing the forest (as a whole region) or walking through the trees, being surrounded by the forest as you walk. In the latter case, the construal does not necessarily include the moment of crossing the forest boundary. The use with the preposition po, as in po lesie (all over forest-LOC, "all over the forest"), indicates motion within the region. Finally, the instrumental use, as in lasem (forest-INSTR, "via the forest"), will indicate being in the forest at least part of the way, and the context may enrich the interpretation in the direction of either przez or po construais. This, however, is possible thanks to the rich experiential basis of our use of the concept 'forest'. In view of the above remarks, it is not surprising that the more of the instrumental. The pair of direction nouns przód/tyl ("front/back") is used in Polish to mark the orientation of objects (e.g., przód/tył samochodu, "the front/back of the car"), but not of people; different words are also used to denote body parts. As a result, our front/back orientation in the surrounding space is conceptualized via the front and back regions adjacent
HOW POLISH STRUCTURES SPACE
35
to the human body. However, the actual expressions motivated by the conceptualization are not always interpreted symmetrically. For example, the expression iść przodem (go front-INSTR) describes going in front of someone in the same direction (through the 'region' in front of them, which is moving along with the forward motion). On the other hand, a with your back (not your face) facing the direction of the motion (like a crayfish); in other words, this is walking through the 'region' behind you. Thus, both uses are based on the conceptualization of regions surrounding the human body, but in the former case the 'walker' walks in 'the front region of the person behind him', while in the latter he walks in his own 'back region'. Both uses are based on the instrumental's meaning of 'region-as-path', but also highly conventionalized.
mi
3.2
Case after two-way prepositions In the literature on Polish it has often been assumed that the choice of case marking for the noun following a preposition is largely arbitrary. It seems, however, that the use of case with particular prepositions is quite regular and compositionally significant. In the combinations considered cases such as the locative, the accusative, the genitive, and the instrumental are used with groups of prepositions to contribute similar aspects of meaning to the construal. Let us first consider the examples with direction nouns: (5) Ptak wzlecial w górę Bird flew in up-ACC "The bird flew up" (6) Patrzyłam na szybującego w górze ptaka I-was-looking at flying in up-LOC bird-ACC "I was looking at the bird flying above" The preposition in (5) and (6) is w ("in"), the most common either with the locative (as in (6), or with the accusative (as in (5)). The contrast between the uses of w with the two cases has often been described in terms of opposition between 'static' meanings, expressed by the locative, and 'dynamic' ones, expressed by the accusative. However, as (6) suggests, the contrast does not really consist in the presence or absence of motion. I
representa
36
BARBARA DANCYGIER
want to argue that the accusative is used when the trajector has to cross the boundary of the region to enter it, while the locative is appropriate when the trajector is already contained within the landmark region. This is clearly the case for (5) and (6): in the former the bird starts flying and gains hight, thus entering the 'upper region', in the latter the bird that the speaker is watching is flying quite high already. Let us also note that in the case of (5), as in the cases discussed above, the region does not have any natural boundaries, but is delineated with respect to the moving agent's position. That is, the region above the bird's position before flying is the 'upper region' which the bird enters upon taking off. In (6), on the other hand, the bird continues its motion without changing the altitude in any significant way and is already 'up' from the point of the observer or its previous position. It can be argued, then, that the contrast between the accusative and the locative is not between 'motion' and 'lack of motion', but between 'entering the region' and 'location or motion within the region'. The nature of the contrast is the same in the case of another Type A preposition — na ("on"). By saying Weszłam na górę (I-went on up-ACC, "I went upstairs") the speaker describes the agent's motion from the lower to the upper part of the building, while a sentence such as Mieszkam na górze (I-live on up-LOC, "I live upstairs") locates the subject in the upper part without indicating motion from the contrasting lower region. 'Preposition + direction noun' combinations are used in a similar way to mark the orientation of other landmark nouns. For example, one can mark the 'upstream' and the 'downstream' part of the river by using appropriate prepositions and direction nouns, and by choosing case according to the rules outlined above. Thus, a sentence such as Poszliśmy dalej w górę/dół rzeki (We-went farther in up-ACC/down-ACC river-GEN. "We went farther up/down the river") talks about motion (into and in the upper/lower part of the river), while a sentence like Miasto leży w górze/dole rzeki (Town lies in up-LOC/down-LOC river-GEN, "The town is up/down the river") locates the town in relation to the parts of the river defined with respect to the position of the speaker. Type A prepositions are used in essentialy the same way with most nouns denoting regions. For example, nouns like las ("forest") are freely used with either case pattern (so that we can say both Byliśmy w lesie (We-were in forest-LOC, "We were in the forest") and Weszliśmy w las (We-went in forest-ACC, "We went into the forest")). Perhaps less construais.
typical
HOW POLISH STRUCTURES SPACE
37
The contrast between the uses of w requiring the locative and those where the accusative is appropriate is best represented in those cases in which the region in question does not have very clear boundaries. However, if the region can be construed as having strict boundaries, and especially if it invites the concept of containment, only the locative variant is acceptable with w. Consequently, we can talk about someone being w pokoju (in room-LOC, "in the room"), but the form ??w pokój (in room-ACC) is inappropriate. The form w + ACC with nouns denoting bounded regions and containers is considered archaic now, and can be found only in the uses where traditional form has been preserved, as in proverbs or old texts. It is understood, however, similarly to the modern use with unbounded regions — to denote entering the region. When w + ACC form is inappropriate, motion into the region will be described with the preposition do ("to") and the genitive case marking. The main difference between w + ACC on the one hand and do + GEN on the other is that the former describes crossing the boundary into a region without specifying whether any distance (path) had to be covered prior to entering, while the latter adds a path/goal component to the 'entering the region' part. That is, it enriches the conceptualization with the image of covering a certain path to get to the goal (the goal is then located within the region considered). Do receives a genitive case marking and belongs to a different group of prepositions, which I will refer to as Type B. These are prepositions such as do ("to") and od ("from"), which are understood as inherently dynamic, which typically involve path/goal and source/path construais, and which accept only the genitive case marking for the nouns they precede. (7) Podrzuciłam piłkę do góry I-threw ball to up-GEN "I threw the ball up" In (7), which uses a direction noun góra, the goal of throwing the ball is to send it, along a predictable trajectory, into the 'upper region'. This use differs from the w + ACC cases discussed above in that the trajector follows the path leading upwards, rather than immediately entering the region perceived as higher than the initial position. The contrast between the dynamic construais offered by w and do is the first case we have seen so far where the concepts such as boundedness, three-dimensionality, or containment play an important role in the choice of
38
BARBARA DANCYGIER
the preposition. It is interesting to note, then, that both w + ACC and do + GEN can appear in phrases with direction nouns, even though direction nouns do not seem to impose any a priori characterization of the regions they represent. In fact, as we could see above, the only boundary that seems to be relevant in the standard case is the reference point established with respect to the moving subject (or the speaker) which separates the 'up' from 'down' or 'front' from 'back'. If there are other boundaries, they are imposed by other elements of the construal, rather than by the direction nouns themselves. As a result, the image of a path involved in the construais motivated by do is often much less clearly present in the interpretations of phrases with direction nouns than in the phrases using other, more specific landmark nouns. For example, the interpretation of Spojrzałam w górę (I-looked in up-ACC, "I looked up") is very close to the understanding of Spojrzałam do góry (I-looked to up-GEN, "I looked up"). This is because in such a case 'entering the region' and 'a path leading into the region' mean practically the same thing. The contrast between w + ACC and do + GEN is clearer in the cases of nouns representing regions with more specific characteristics. The sentence Poszliśmy w las (We went in forest-ACC, "We entered the forest") suggests that the people referred to walked until they found themselves in the forest, and continued their walk within its boundaries. The central aspect of the interpretation is that they are now in the forest, walking in it, but not across it, and that getting to the forest was not the primary goal of the trip. For comparison, Poszliśmy do lasu (We went to forest-GEN, "We went to the forest") presents the forest as the goal of the walk, reached along a certain path. The contrast is also seen in the inappropriateness of using do with some nouns which accept w. For example, we have a phrase strzelać w powietrze (shoot in air-ACC, "fire a shot into the air"), which seems to be acceptable thanks to the fact that such shooting has no aim, that the path the bullet travels does not matter, and that we are surrounded by air (so the bullet does not have to 'reach the air' along any path. The same factors seem to account for the fact that a phrase such as ??strzelać do powietrza (shoot to air-GEN, "shoot at air") requires a rather unlikely construal whereby the air is the target. In other cases, the contrast gives rise to interesting conventional collocations. For example, iść do miasta (go to town-GEN) is interpreted as taking a special trip downtown from a suburb or a less central area of the city, while iść w miasto (go in town-ACC) is
HOW POLISH STRUCTURES SPACE
39
used to describe wandering around the city, exploring it. The latter use does not imply the stage of reaching the area (no path), and focuses on 'being in', rather than on 'getting to'. As I observed above, direction nouns do not carry any built-in information concerning boundedness or dimensionality. Consequently, they are not used with the third set of basic prepositions, Type C, which are mainly used to talk about three-dimensional landmarks; they are prepositions like przed/za ("in front of/behind") and nad/pod ("above/under"). Type prepositions also show the static/dynamic contrast through case: the former is marked by the instrumental, the latter again by the accusative. (8) Schowałam się za drzewo / I-hid myself behind tree-ACC / Byłam schowana za drzewem I-was hidden behind tree-INSTR "I hid myself behind a tree / I was hidden behind a tree" As in the cases analysed above, the case marking is making a consistent contribution to the meaning of the spatial expression overall. In the accusative variant of (8), the case indicates again that the landmark noun is a goal to reach in order to be able to hide, the preposition locates the hiding place with respect to the landmark noun. The conceptualization behind the use of the instrumental, for comparison, is that the hearer's vision or thought has to travel an imagined path around, above, or case of 'fictive motion', as described by Talmy 1996). The instrumental thus seems to be consistently used to introduce the concept of a path, but, contrary to the genitive, it does not involve the concept of a goal. It was pointed out to me that the use of the instrumental in the way exemplified in (8) may also be interpreted differently. Namely, it could perhaps be argued that the tree is in fact an instrument the subject uses to hide. However, this interpretation does not seem plausible. First of all, not all uses of the instrumental with za and other such prepositions require the presence of an agent in the construal and many are not compatible with any use of instruments. For example, za + INSTR would be appropriate with most verbs which are not motion verbs (which arises out of the primary static/dynamic contrast between the accusative and the instrumental with Type prepositions). We could say, among others, that somebody was, slept, sat, ate, sang a song, or painted a picture behind a tree and in each
40
BARBARA DANCYGIER
case the form za + INSTR would be used. This is the reason why the second sentence in (8) is in the passive voice, for the active voice variant implies motion to the hiding place. Also, if the instrumental were to mean 'instrument' here, why would the first sentence use the accusative to do the same? Secondly, it might be interesting to note that verbs involving 'fictive motion' belong with the dynamic construal, not the static one. That is, in order to say I looked behind the tree in Polish one would have to use za + ACC, not za + INSTR. The difference is that in the accusative case it is the subject's vision that has to reach behind the tree, while in the instrumental case, it is the hearer's. It seems, then, that the interpretation of the phrase is quite specific in this case. Finally, there are examples of the uses with za where the meaning of the instrumental seems more directly related to the 'region-as-path' interpretation I described in section 3.1. In the sentence Mieszkam za lasem (I-live behind forest-SING-INSTR, 'I live over the forest") the forest is easily conceptualized as a region (possibly as a three-dimensional region) and as a path to be travelled (mentally) to establish where the subject lives. Furthermore, the similarity is even clearer in the use of the instrumental with plural nouns. When one describes something as being za lasami (behind forests-INSTR, "over many forests, far away"), the use of the instrumental is almost identical to the one we saw in the phrases like iść lasami (walk forests-INSTR, "walk through many forests"), where the noun lasami indicates that the walking covered a long distance and part of the route led through the forest, but the length of the walk is unspecified and the 'regions' passed on the way are not clearly conceptualized at all. There is practically no difference in the way lasami is understood on its own or with za, while the difference in the understanding of the path (physical motion versus fictive motion) arises on the basis of the absence or presence of the preposition. To sum up, the choice of case after prepositions of all three types makes a consistent meaning contribution to the spatial expression overall. Thus, Type A prepositions always require the locative in the so-called static uses to denote the landmark region; they always require the accusative in the so-called dynamic uses - this seems to indicate that the trajector enters the landmark region. For Type B, the case required is the genitive, and the description of motion through space involves path/goal or source/path structure. Finally, Type prepositions use either the accusative (in essentially the same sense as in Type A phrases), or the instrumental (which marks a mental path leading to the location). Consequently, the instrumental
HOW POLISH STRUCTURES SPACE
41
can be seen as consistently marking the concept of a path - whether in prepositional phrases or in other spatial construais. 4.
Metaphorical extensions The construais outlined above to describe the primary uses of basic prepositions are further extended into non-spatial domains. Not surprisingly, the most common domain of metaphorical extension is time, but specific construais of spatial regions are also involved in describing states or activities. In the present section I will review several mappings, to show how spatial interpretation motivates the non-spatial one. First, it is interesting to note that the spatial regions usually marked by the direction nouns, which organize our basic spatial orientation, can also be interpreted in a metaphorical way. (9) Ty pójdziesz górą a ja doliną You go-FUT mountain-INSTR while I valley-INSTR "You'll go up the mountain, I'll go down the valley" Sentence (9) is a line from a song where a boy and a girl separate and decide to go through life their different ways (the interpretation is similar to the well know line from "Loch Lommond": I'll take the high road, you'll take the low road). The regions associated with the words 'mountain' and 'valley', used in the 'path' instrumental, suggest taking two different It is almost trivial to observe that spatial expressions are often metaphorically extended to describe time and it is by now a well recognized fact that prepositions are used to express temporal relations (for some observations concerning prepositions in Polish see Kochańska 1996). It is thus not surprising that the use of case which marks a certain spatial construal can also be extended into the domain of time. In the sentence Nie lubię spacerować nocą (Not I-like to-walk night-INSTR, "I don't like to walk at night") the instrumental reinterprets the 'time-region' (the night) in the same way as in the case of space. It is in fact common for nouns denoting major units of time (such as an hour, a week, a month, a year) to be used in the plural and with the instrumental, marking (as in the spatial cases) to indicate the speaker's annoyance with the length of time spent working towards the goal (as in the case of spatial uses of the instrumental achieving a goal is not part of the construal). For example, one can describe working on a project for a long time by saying Pracowałem nad tym
dir
42
BARBARA DANCYGIER
godzinami / tygodniami / miesiącami / latami (I-worked over this hours-INSTR / weeks-INSTR / months-INSTR / years-INSTR, "I spent hours / weeks / months / years working on this"). Interestingly enough, the word czas ("time") itself can also be used with the instrumental. The singular form has the meaning similar to 'occasionally, by chance', while the plural is closer to 'from time to time'. Both uses seem to refer to 'time' as 'any point or section of the time axis' and conceptualize units of time as regions in space. Many of the spatial construais described in section 2 motivate temporal uses where units of time are conceptualized similarly to spatial regions. That is, we can talk about seeing a doctor once a month using the expression raz w miesiącu (once in month-LOC), thereby locating the event anywhere in the region represented by the word miesiąc. The pattern w + ACC is also used to mark entering temporal units, as in wkroczyć w nowy miesiąc (enter in new month-ACC, "start a new month"). Perhaps the most interesting use, though, is that of za + ACC, as in wyjeżdżamy za miesiąc (we-leave behind month-ACC, "we leave in a month")· The use of the accusative and the preposition suggest motion through time until we (mentally) reach a point beyond the suggested 'temporal region' (a month). The contrasts among w + LOC, w + ACC, and do + GEN give rise to interesting construais in the domain of states and activities. For example, we śnie (in sleep-LOC) describes a person in the state ('region') of being asleep (via the STATES ARE LOCATIONS metaphor). Zapaść w sen (fall in sleep-ACC, "fall asleep"), for comparison, represents an inception of the state via the 'entering the region' construal represented by the w + ACC pattern. Finally, ułożyć się do snu (lie oneself to sleep-GEN, "lie down to be able to fall asleep") uses the path/goal structure of the spatial construal to represent taking steps to reach/enter the state ('region'). Similar extensions can be found in descriptions of activities. An activity noun like taniec ("dancing") can represent a landmark region in a number of different prepositional phrases. W tańcu (in dancing-LOC, "while dancing") describes just the 'goings-on' within the landmark region/activity, as in Uśmiechała się w tańcu (She-was-smiling herself in dancing-LOC, "She was smiling while dancing"). W taniec (in dancing-ACC, "into dancing") also involves the stage of beginning the activity, just as it involves entering the region in the spatial use (as in Poszli w taniec [They-went in dancing-ACC, "They started dancing"]). Finally, do tańca (into dancing-GEN) maps the path/goal structure onto taking steps to enter
HOW POLISH STRUCTURES SPACE
43
an activity, as in Zagrali do tańca (They-played-PERF to dancing-GEN, "They started playing so that people could dance"). Finally, it is interesting to note that nouns which are ambiguous between spatial and other meanings use preposition + case combinations in the way which matches each of the meanings independently. For example, a noun like droga ("road") is commonly used with the instrumental (drogą [road-INSTR, "along the road"] or with prepositions like na ("on"), przy ("by"), or phrases like po drugiej stronie ("on the other side/across"). Such construais are clearly related to the salient aspects of the spatial configuration of a road: being a path, having a surface, being bounded on two sides. However, when the meaning of the word is extended into the concept of 'travel', it is now conceptualized broadly as a region, and expressions such as w drogę (in road-ACC), w drodze (in road-LOC), or do drogi (to road-GEN) are used to describe aspects of travel. Thus w drogę (in road-ACC) indicates starting on one's way (entering the travel 'region'), w drodze (in road-LOC) represents being in the process of travelling (in the travel 'region'), while do drogi (to road-GEN), as in przygotować się do drogi ("get ready for travelling"), indicates the steps (the path) one needs to take to begin travelling (enter the travel 'region'). 5.
Conclusion I hope to have shown, then, that there are three subsystems responsible for spatial construais in Polish: direction nouns, prepositions, and case. Each of the subsystems makes a consistent meaning contribution to the spatial construal represented by the expression. Elements of the subsystems can be combined in different (though well defined) ways, but each of the possible configurations receives its interpretation on the basis of the contribution each of the systems makes. References Ameka, Felix K. 1995. "The Linguistic Construction of Space in Ewe." Cognitive Linguistics 6:2/3.139-182. Brecht, Richard D. & James S. Levine, eds. 1986. Case in Slavic. Columbus, Ohio: Slavica Publishers. Cienki, Alan J. 1989. Spatial Cognition and the Semantics of Prepositions in English, Polish, and Russian. München: Verlag Otto Sagner. Cuyckens, Hubert. 1984. "At: a Typically English Preposition". Papers and Studies in Contrastive Linguistics 19.49-64.
44
BARBARA DANCYGIER
Dąbrowska, Ewa. 1994. "Radial Categories in Grammar: the Polish Instrumental Case". Linguistica Silesiana 15.83-94. . 1997. Cognitive Semantics and the Polish Dative. Berlin: Mouton de Gruyter. Janda, Laura. 1986. A Semantic Analysis of the Russian Verbal Prefixes za-, pere-, do-, do- and ot-. München: Verlag Otto Sagner. . 1993. A Geography of Case Semantics: the Czech Dative and the Russian Instrumental. Berlin: Mouton de Gruyter. Kochańska, Agata. 1996. "Temporal Meanings of Spatial Prepositions in Polish: the Case of przez and w". In: The Construal of Space in Language and Thought, eds. Martin Pütz & René Dirven, Berlin: Mouton de Gruyter, 491-508. Langacker, Ronald W. 1991. Foundations of Cognitive Grammar. Vol II: Descriptive Applications. Stanford: Stanford University Press. Serra-Borneto, Carlo. 1997. "Two-way Prepositions in German: Image and Constraints". In: Lexical and Syntactical Constructions and the Construction of Meaning, eds. Marjolijn Verspoor, Kee Dong Lee & Eve Sweetser, Amsterdam: John Benjamins, 187-204. Sinha, Chris & Tania Kuteva. 1994. "Distributed Spatial Semantics". Paper presented at LAUD International Symposium on Language and Space, Duisburg, March 1994. Smith, Michael B. 1987. The Semantics of Dative and Accusative in German. Ph.D. dissertation. San Diego: UCSD. . 1993. "Cases as Conceptual Categories: Evidence from German". In: Conceptualization and Mental Processing in Language, eds. R.A.Geiger & . Rudzka-Ostyn, Berlin: Mouton de Gruyter, 531-565. . 1995. "Semantic Motivation vs. Arbitrariness in Grammar: Toward a More General Account of the DAT/ACC Contrast With German Two-way Prepositions". Insights in Germanic Linguistics: Methodology and Transition, 293-323. Berlin: Mouton de Gruyter. Sysak-Borońska, Maria. 1975. "Some Remarks on the Spatio-Relative System in English and Polish". Papers and Studies in Contrastive Linguistics 3.185-208. . 1980. The Spatial System in English and Polish: Prepositions of Direct Location. Ph.D. dissertation. Katowice, Poland: University of Silesia. Talmy, Leonard. 1983. "How Language Structures Space". Spatial Orientation: Theory, Research, and Application, eds. Herbert L. Pick & Linda P. Acredolo, New York: Plenum Press: 225-282.
HOW POLISH STRUCTURES SPACE
45
— . 1996. "Fictive Motion in Language and 'Ception'". Language and Space, eds. Paul Bloom, Mary Peterson, Lynn Nadel & Merrill Garrett, Cambridge, Mass.: MIT Press, 211-275.
Case Meaning and Sequence of Attention Source Landmarks as Accusative and Dative Objects of the Verb Robert B. Dewell Loyola University, New Orleans, U.S.A.
1.
Introduction This paper will examine a set of source-path constructions in German and English with an eye toward isolating the semantic contribution of the grammatical cases. It will pay particular attention to some semantically unusual constructions in which the flow of energy and motion in the event being described does not coincide completely with the flow of our focal attention as we construe the event. On the basis of these constructions we will be able to isolate meanings for the cases at the level of prominence relations, a level which is more abstract than their obvious prototypical instantiations in terms of semantic roles (agent, patient, experiencer, and so on). The key to the analysis will be a more explicitly dynamic conception of prominence relations than those previously proposed. Case meaning and shifting prominence relations Langacker (1991) proposes schematic meanings for the clausal subject and direct object, and thus for the nominative and accusative cases that signal them, in terms of relative prominence: the nominative signals the 'primary clausal figure', while the accusative signals the 'secondary clausal figure'. These abstract grammatical relations are 'a matter of construal rather than conceptual content' (1991:312), and they underlie more specific prototypical instantiations as agents and patients respectively. On the other hand Langacker (1991:324-9) analyzes dative indirect objects without a schematic meaning based on prominence relations. They are characterized in terms of the semantic role of an 'active experiencer in the target domain', with the prototypical experiencer role frequently blending with that of recipient (or possessor). When Langacker speaks of 'primary' and 'secondary' figures, he is referring to a static, hierarchical rank ordering of relative prominence — i.e., the clausal subject is the principal figure in the clause and the direct object ranks next in importance. If I understand him correctly, both the subject and the object are in the focal 'spotlight' throughout the construal (1991:301), but the light constantly shines more brightly on the subject. 1.1
48
ROBERT DEWELL
The terms 'primary' and 'secondary' potentially suggest another meaning, though — namely that the subject and direct object are the 'first' and 'second' clausal figures in the construal. It is not fully clear what role the notion of temporal sequence plays in Langacker's analysis of prominence relations, but he apparently considers it to be a matter of conceptual content. In other words, the entity profiled by the subject typically precedes the entity profiled by the object in the temporal flow of the action chain, but temporal sequence does not play an essential role in the figure/ground construal itself. Even when Langacker (1991:293) characterizes the nominative subject as the 'starting point' for a 'natural path' from primary figure to secondary figure to ground, he apparently has in mind a static relationship that does not change during the construal. In this paper I want to suggest that the spotlight of focal attention can shift during the course of construing an event, independently of the action chain, and that we need to keep a temporal dimension in mind when we speak of prominence relations. We may begin a construal focused centrally on one entity as the primary figure while another entity has secondary prominence, but then transfer our focal attention during the course of the construal so that the original secondary figure takes over primary prominence. Shifting from the spotlight metaphor, we can think of construing an event as watching a motion picture in which we control the camera (and thus the size and relative centrality and focus of the entities in view). Now it is possible to isolate single images in freeze-frame, 'synchronically' so to speak, so that each will have its own prominence relations. If we then compare frames we may note that 'diachronic' changes occur in relative prominence as well as in the content of the event being described. To make explicit this distinction between static synchronic prominence relations and a diachronic sequence of focal attention, I will avoid the word 'primary' and speak of participants that attract either 'central' or 'secondary' focus in a frame or a series of frames, paying particular attention to a comparison of the 'initial' and 'final' images of the construal. Thus we can entertain the hypothesis that the nominative subject designates an entity with 'initial central' focus, and the accusative direct object an entity that acquires 'final central' focus.1 Moreover, once we have added a temporal dimension to prominence relations a plausible schematic meaning emerges for dative indirect objects as well (i.e., as a grammatical relation underlying specific semantic roles such as active experiencer, recipient or possessor). Specifically, I will suggest in the final section of the paper that the indirect object is a participant with constant 'secondary' focus outside the sequential flow of central focal attention (i.e., it is the dative that is defined
CASE MEANING AND SEQUENCE OF ATTENTION
49
mainly in terms of secondary prominence independent of the temporal dimension of construal). 2.
Trajector-centered source paths To set the stage for the constructions that require us to separate the flow of focal attention from the flow of energy in the action chain being described, we will begin with some normal source-path descriptions in which the two sequences are in harmony. Imagine the events described in (1) and (2) as if you were watching a film of schematically structured images — a film with a beginning, an end, and a variable number of images in between, ordered according to the temporal and force-dynamic flow of an action chain. (1) Bill Bill (2) Bill Bill
came out of the house. ist aus dem Haus gekommen. came out. ist herausgekommen.
The trajector (TR) Bill is in central focus at the beginning of the construal, and he remains so as he moves steadily downstream in space and time. As we construe the event, we simply follow Bill as he moves in a straightforward trajectory away from his original location. The concluding image (the resulting state) of the event as described has Bill as its central focus, just as the opening image did. As for the source-landmark (LM) house, it can be specified obliquely in a PP as in (1) or gapped altogether in a construction with a source-path particle such as off, out, or away (or German ab-, aus-, or weg-), as in (2). Either way, it is relegated to peripheral status relative to Bill. We focus primarily on Bill throughout the construal, and as he moves away from the house it fades progressively into the periphery, left behind where Bill used to be. Once Bill has reached his new location, that is where our focus is and the house is 'history'. 2.1
Source paths with an accusative subTR The situation is more complicated in transitive constructions such as (3)-(4) since the initial focal participant is the agent Helmut, who enters the construal before the letter does and, in these particular examples, continues to participate in the event all the way to the end of the construal (i.e., he never releases the letter). So the letter never receives the kind of exclusive focus that the intransitive source-path TR did in (1)-(2). Still, it is fair to say that the
50
ROBERT DEWELL
letter does become the most central entity in the construal as soon as it begins to move — it becomes a figure in its own right, the TR of a source path. (As the final image of the construal you might focus almost exclusively on the letter with Helmut extremely marginalized, the way you would if the agent were left upstream in a sentence like 'Helmut threw the letter out the window into a puddle'; but you could not focus almost exclusively on Helmut with the letter marginalized at the end.) (3) Helmut took the letter out of the drawer. Helmut nahm den Brief aus der Schublade. (4) Helmut took the letter out. Helmut nahm den Brief heraus. As far as the source path itself is concerned, the constructions in (3)-(4) are not essentially different from (1)-(2), except that the source-path TR (the letter) is now a subTR (Langacker 1987:268-9) and the source-path LM (the drawer) is now a subLM. We follow the letter's path just as we followed Bill in the intransitive constructions. The resulting state is a new location for the letter. As for the source LM (the drawer), it can be specified in exactly the same patterns as in (1)-(2): it can appear obliquely in a source PP as in (3), or it can be gapped in the presence of a source-path particle as in (4). The drawer is only peripherally relevant to the resulting state; it is left somewhere upstream as part of the 'history' of the construal. In all the sentences of (1)-(4), the conceptual flow of our focal attention is in harmony with the objective flow of energy and motion being described. Our attention focuses steadily on a nominative TR until it transfers energy to a subTR, in which case we focus steadily on that subTR until the end of the event. Everything about the construal moves, merrily merrily, down the stream. 3.
LM-centered source paths Things change in sentences such as (5)-(9), though, in ways that highlight the distinctive semantic contribution of grammatical cases to the construal of an event. (5) George peeled the apple. George schälte den Apfel. (6) Ronald deiced the windshield. Ronald enteiste die Windschutzscheibe.
CASE MEANING AND SEQUENCE OF ATTENTION
51
(7) Willy wiped off the table. Willy wischte den Tisch ab. (8) Ike cleared the table (of dishes). Ike räumte den Tisch ab. (9) Kurt Georg unpacked the suitcase. Kurt Georg packte den Koffer aus. The denominal constructions in (5) and (6), for example, describe events which are essentially like those in (3)-(4): an agent (George) removes a subTR (the peel) from a subLM (the apple). The same can be said for the combination of a 'wipe' verb with a source-path particle and an accusative source LM in (7),2 'clear' verbs with an accusative source LM as in (8),3 and other verbs such as the one in (9). In fact, (7)-(9) all have accusative-subTR alternations with the same verb, as in (10)-(12). (10) Willy wiped the crumbs off (of the table). Willy wischte die Krümel vom Tisch/ (vom Tisch) ab. (11) Ike cleared the dishes (from the table). Ike räumte die Teller ab. (12) Kurt Georg unpacked the clothes (from the suitcase). Kurt Georg packte die Kleider (aus dem Koffer) aus. Despite the similarity of events, however, (5)-(9) are interpreted quite differently from the constructions with an accusative source-path TR. 'Removing the peel from an apple' describes doing something to the peel, and the resulting state focuses primarily on the peel's new location; 'peeling an apple' describes doing something to the apple, and the resulting state is a state of the apple, which holds center stage at the conclusion of the event as construed. Similar comments can be made comparing (7)-(9) with (10)-(12). In effect (3)-(4) and (10)-(12) on the one hand, and (5)-(9) on the other, represent two opposed strategies for imagining a source path: a TR-centered strategy in which we follow the moving TR sequentially through space and time from the source location to somewhere else; and a LM-centered strategy in which we hold a bounded spatial setting constantly at the center of our attention and watch the TR leave that frame gradually without conceptually leaving it ourselves. If you think about it, though, the LM-centered strategy of (5)-(9) is a strained one. To interpret (9), for example, we are aware of a path in which
52
ROBERT DEWELL
the clothes move (gradually) away from the suitcase; but at the point when they begin to draw our attention away from the suitcase, we have to resist the tendency to follow the moving objects. We loop back instead to hold the suitcase ultimately at the center of our attention — a construal process I'll refer to as a separation loop. As for the clothes, they move into the periphery and become marginalized; they are out of the picture at the end, much as the subLM was out of the picture at the end of the TR-centered paths. Now it is the clothes which have become interpreted history at the end; but they are downstream history, which is an odd concept. In sum, when we have finished understanding sentences like (5)-(9), we are focused at a place upstream from the final location of the only moving entity in the source-path image. Our construal is in harmony with the flow of energy from Kurt Georg to the (contents of the) suitcase, but it does not continue on from there to follow the clothes' trajectory away from the suitcase. At the end of the construal we are 'going against the flow' of space and time, resisting the momentum of objective motion, and ultimately focusing on an upstream resulting state. While the source-path TR is floating downstream, we are conceptually swimming back upstream to remain with the source-path LM. 3.1
The 'end at the accusative' principle I want to suggest that we construe the events described in sentences like (5)-(9) in the odd way that we do because the accusative case directs us to do so. Its core schematic meaning is roughly: At some point in the construal of the event, shift the central focus from the nominative subject to the accusative direct object, and conclude the construal that way. In the case of constructions involving separation loops, this means we are not allowed to leave the accusative object behind as interpreted history even if it is a source LM. In normal transitive path descriptions such as those in (3)-(4), the role of the grammatical cases is not particularly obvious since we would naturally construe the event that way anyway. The sequence of focal attention converges with the intrinsic action-chain dynamics, so we don't need grammatical case to tell us to conclude the construal with our attention focused on the direct object (especially given English word order). In constructions requiring a separation loop, however, something in the sentence has to direct us to the marked LM-centered reading. 3.2
Relative salience and boundedness Sentences like (5)-(9) are subtle examples, however, because the source-path TR is marginalized. It takes unusual mental effort to become aware that these sentences describe a source path at all, because we tend to think only
CASE MEANING AND SEQUENCE OF ATTENTION
53
of a flow of energy from the agent to the source-path LM. For example, we tend to imagine that someone who unpacks a suitcase is manipulating the suitcase directly rather than the clothes, and that sentence (9) simply describes an action chain from the agent to the suitcase and nothing more. In other words we construe the accusative object to be the tail of the action chain for the complete event being described, even when it is not. Generally speaking, the constructions in (5)-(9) only occur when the subtrajectory can be marginalized in this way by concentrating primarily on a flow of energy from the agent to the direct object. The source-path TRs — the apple peel of (5), the ice in (6), the dirt, crumbs, liquid or whatever was wiped off in (7), the cleared dishes of (8), the unpacked clothes etc. of (9) — do not have bounds of their own, because that would make them distinct from the accusative LM. They are construed as distributed TRs that 'fill' the whole source-path-LM space at the beginning of the construal.4 When they depart, they do so gradually in a diffuse path, ultimately disappearing vaguely into the margins of the scene as we hold our focus steadily on the LM space they are vacating. If the thing being removed were a clearly separate entity from the source LM, the accusative-subTR constructions in (3)-(4) and (10)-(12) would be called for. By the same token, the accusative LM is the only entity with its own independent shape involved in the source path as described. To be in central focus at the end of the construal, it needs to be a setting within whose bounds the source path is construed to take place. In other words, it is the accusative counterpart to a setting-subject that hosts an internal source-path process (cf. 'The snake shed its skin', 'The bathtub drained slowly'). It follows, then, that source-path descriptions with a salient bounded TR and a salient separately bounded accusative LM are hard to find, since they would force us to shift our attention back upstream away from a focal TR moving downstream. The action chain would be in obvious opposition to the sequence of focal attention. We will see in section 4 that occasional examples do occur, though, illustrating the semantic contribution of the accusative case in a more unmistakable way. 4. Nominative source-path TRs with accusative source LMs 4.1 Nominative distributed TRs: vacating an area Clear and empty (and German raümen) can also be used on rare occasions with a distributed subject that is the source-path TR, as opposed to an agent removing a gapped source-path subTR.
54
ROBERT DEWELL
(13) ?The audience emptied the auditorium. ?Die Zuschauer räumten den Saal. The constructions are problematic, though, because the subject is more commonly read as a normal causal agent as in (8) (cf. 'The audience emptied the auditorium of debris'). Evacuate and vacate, which also lexicalize a resulting state of vacancy (complete departure by a distributed TR), are somewhat more normal with a source-path-TR subject. (14) The troops evacuated the area. (15) The tenants vacated the building. There is still ambiguity, though, since evacuate and vacate are also used with a bounded subject construed as a causal agent removing a gapped source-path TR (e.g. 'The general evacuated the area')· The subject is often metonymically included in a distributed set together with something such as possessions ('Gerald vacated the premises/ the apartment').5 To the extent that (14)-(15) do in fact have nominative source-path TRs, they provide excellent examples of separation loops with salient TRs (albeit construed as distributed and without clear separate bounds), and with direct-object source LMs that can scarcely be characterized as patients. All of the motion being described is away from the accusative direct object, yet we still conclude our construal of the event focused on the accusative and its resulting state.
4.2 Nominative bounded TRs and accusative participants: abandoning Sentences such as (16), with verbs like abandon and desert, elicit a separation loop even though the nominative source-path TR is a bounded individual. Jimmy is syntactically specified like the TR of the TR-centered constructions in (1)-(2); he has his own clear bounds and shape independent of the LM. That makes it difficult to interpret (16) with a simple LM-centered reading. Wc in effect focus on an abstract space with the children at its center and imagine a resulting state that they are left 'in' after the TR departs. (Compare 'Jimmy left his children alone'.) (16) Jimmy abandoned his children. Jimmy verließ seine Kinder.
CASE MEANING AND SEQUENCE OF ATTENTION
55
Sentences like (16) are particularly strong evidence for the 'end at the accusative' principle. We are forced to turn our attention back from a highly salient and moving subject and focus squarely on the upstream source LM at the end. Such sentences also strain the notion of 'patient', since the flow of energy is all away from the children (even as our focal attention moves toward them). 4.3 Atypical accusatives: leaving a place Sentences such as (16) are relatively uncommon, however, no doubt because the action chain and the sequence of focal attention are so strikingly at odds. A more entrenched construction that is superficially similar is illustrated in (17)-(18), namely the straightforward 'leaving' type with the verbs leave and verlassen and their synonyms (including depart and exit). Although 'abandoning' variants like 'Jimmy left his wife' are still possible, most 'leaving' constructions take LMs that directly describe the place itself rather than a participant that metonymically defines an associated space. These constructions cause difficulties for any approach to the meaning of the accusative. (17) Harry left the house by the back door/ at 8:00. Harry verließ das Haus durch den Hinterausgang/ um 8 Uhr. (18) Lyndon left school. Lyndon verließ die Schule. The significant difference between (16) and (17)-(18) is readily apparent in the behavior of the English sentences. (17)-(18) can take a pure place expression like here or there in the direct-object slot: contrast 'Harry left here an hour ago' with *'Jimmy abandoned here' (or *'George peeled here' or *'Willy wiped off here'). Their LMs are construed to be substantially unaffected by the event: the house itself is not changed because Harry left it, whereas Jimmy's children are definitely affected by being abandoned (compare 'an abandoned child' with the absurdity of an adjectival passive participle like *'a left school'). They also do not normally take passive: contrast *?'The house was left an hour ago/ by the rear door' with 'Rosalyn was left by her husband' or 'The children were abandoned by their parents'. Moreover, these constructions have intransitive counterparts ('Harry departed', 'Harry left from home'), which was not the case with abandoning (*'Jimmy abandoned', *'Jimmy left from his children').
56
ROBERT DEWELL
Simply put, the 'direct objects' of (17)-(18) are not very good direct objects. We do not really end our construal of the event focused centrally on the accusative source LM. (For the accusative to become the focal center it has to become a subTR participant, as in augmented resultative constructions like 'leave the house in shambles'.)6 On the other hand, (17)-(18) do not really violate the 'end at the accusative' principle either. We may not wind up looking squarely at the house as a whole bounded entity, but we cannot leave the house behind as interpreted history, either — i.e., we do not follow Harry's further path downstream (cf. 'Harry left from the house'). The event as described is a punctual interaction frozen at the moment of separation. It ends, without a clear resulting-state image, at the moment when our attention would otherwise be drawn away from the source LM. These constructions are semantically too complex to be treated adequately here, but I believe they are actually a special punctual case of a general type with an accusative pathway-LM. Compare 'walk the plank', 'run the course', 'drive the whole way', as well as 'swim two miles' — constructions with an intransitive verb of motion augmented to take an accusative that defines (delimits, 'contains') its whole path, so that the construction describes the interaction between a path and a pathway.7 There is a kind of fusion between the nominative TR and an accusative place LM that co-extends with its trajectory, reflecting the general lack of a clear distinction between a 'path' as the sequential motion of a TR and a 'path' as the route the TR follows (or the pre-existing linear-shaped surface the TR is in contact with). The nominative subject and the accusative object share the central focus for the whole duration of the TR's path (and the accusative functions as an 'incremental theme' in the sense of Dowty 1991:567-571) — or more precisely, there is an ongoing interactive transfer of both energy and focal attention from the subject to the direct object (similar to more normal transitive constructions such as 'rub the surface'). Leave (like enter) is a marked variant of this type in two ways. To begin with, it incorporates the path type into the meaning of the verb, adopting the pattern of a 'verb-framed' language rather than the usual English 'satellite-framed' oblique specification (Talmy 1992). Moreover, its built-in path type specifies punctual separation at the moment the TR clears the bounds of the LM. In normal TR-centered intransitive constructions, that moment of separation is followed by a continued path away from the LM, but when leave occurs with an accusative LM the interpretation is effectively frozen at the moment of separation, since we cannot leave the accusative object behind. In effect then, sentences like (17)-(18) are coerced into the typical transitive pattern as a special variant.
CASE MEANING AND SEQUENCE OF ATTENTION
57
4.3.1 Accusative route-path LMs A similar kind of punctual separation reading occurs with sentences such as (19)-(20), in which the accusative object is a route-path LM (characterizing the path in terms of an intermediate point) rather than a source LM. (19) Jack overshot the intersection. Jack hat die Kreuzung überfahren. (20) Jack jumped the ditch. Jack hat den Graben übersprungen. Although we are fully aware that the bounded TR has moved past the LM, we still conclude our construal holding the accusative upstream LM in focus — a pronounced separation loop. The LM is an unaffected place, and these sentences are not prototypically 'transitive' (*?'The intersection was overshot'). Nevertheless, the accusative directs us not to leave the LM behind as interpreted history, and we tend to wind up with a punctual reading focused on the moment of separation from it. The direct object becomes a sort of participant that interacts with the TR, either because the TR aimed at it or because it is an obstacle (adversary) to be 'overcome'. Compare Dewell (1996). 5. Interactive separation, shared focus, and the dative 5.1 Escaping Up to this point, English and German have looked like twins. The two languages part company in interesting ways, though, when it comes to sentences like (21). (21) Ludwig escaped his captors/ pursuers. Ludwig ist seinen Verfolgern entkommen. On the surface, the English sentence looks like exactly the same syntactic pattern as the 'abandoning' sentence in (16). It is also like 'leaving a place' in that it relates to intransitive constructions ('Ludwig escaped from his captors'). As its German counterpart with a dative LM (and a sein-perfect) suggests, though, escaping is different from other kinds of leaving in ways that are obscured by the English syntactic pattern. In order to focus the discussion as clearly as possible, consider the minimal pair represented in (22), again construing the event in slow motion while noticing the prominence of the participants.
58
ROBERT DEWELL
(22) a. Ludwig abandoned his family. b. Ludwig escaped his family. When Ludwig abandons his family, we end up focused mainly on them and their situation after he has departed, much as we ultimately focused on the apple without its peel or the windshield without ice. It doesn't really matter exactly where Ludwig is now, as long as he is not with the family, because we have transferred our attention from him to his poor accusative victims in their new state. If Ludwig escapes his family, on the other hand, they do not stand still and they are anything but passive victims; they are construed to be active, moving participants that pursue him, or at least try to reach out to him and pull him back. Moreover — and this is what really separates escaping from abandoning — at the end of the escaping event we are still focused principally on Ludwig! We might even imagine him breathing a sigh of relief, free at last from his family's clutches. The family is still prominent in the picture, to be sure, but we have not transferred our central focus from Ludwig to them. In the final image of the construal the relative prominence of Ludwig and the family is unchanged since the beginning image. The principal focus is steadily on the nominative TR, and the source LM retains secondary focus as an additional participant. Put another way, (22b) is not prototypically transitive. It does not exactly violate the 'end at the accusative' principle by leaving the direct object behind as interpreted history, but it does not leave us focused centrally on it either. It is like the accusative-LM types discussed under (17)-(20), in that an atypical variant is being coerced into the transitive pattern; but now the direct object is a participant rather than a place. This semantic analysis is supported by the observation that there is no passive version of escaping ('His family was abandoned' is acceptable, *'His family was escaped' is not). Similarly, it is acceptable to say 'What Ludwig did to his family was abandon them', but not *'What Ludwig did to his family was escape them'. It is also worth mentioning that escape is restricted to an animate LM when it appears with an apparent direct object, while intransitive escape is not: Ludwig can escape from his captors or from jail, but he can't *'escape jail'. Moreover, if we augment (22a) to 'Ludwig abandoned his family to a life of solitude', the additional phrase clearly has the family and not Ludwig as its TR; (22b) does not readily admit such an expansion because we cannot leave the family behind as history at the end of the construal, but if it did we would certainly interpret *?' Ludwig escaped his family to a life of solitude' with Ludwig rather than his family as the TR of the additional phrase.
CASE MEANING AND SEQUENCE OF ATTENTION
59
The result of (22a) is an 'abandoned family', not an *'abandoned father'; the result of escaping, on the other hand, is not an *'escaped captor' but an 'escaped captive'. From the point of view of the speaker, the situation is roughly as follows. When we want to describe an escape so that it is not simply TR-centered like the intransitive expressions, we have no obvious alternative in English other than the transitive pattern of (22b). But that means we have to press the intended meaning into a syntactic mold that it does not completely fit. The 'accusative' is still there at the end of the interpretation, but it retains secondary status relative to the nominative TR. Interestingly, the closest we can come to capturing the relevant semantic distinctions explicitly in English is the creative use of a 'dative' object in a double-object construction such as (23).8 (23) Ludwig gave his pursuers the slip. 5.2 Shared-focus separation in German: 'ablative' ent-verbs with dative LMs While English is basically limited to a choice between two opposing syntactic constructions (an apparently transitive direct object or no non-oblique object at all), German offers a middle road via the dative construction in the German version of (21), further illustrated in (24)-(25). This construction is particularly associated with what Barbara Stiebels (1996) calls the 'ablative' ent-verbs, so that ent- again emerges as a form uniquely associated with a separation loop. The status of the nominative subject as the central focal participant throughout the path construal is reflected in the use of sein as the perfect auxiliary. (24) Dick ist seinen Wächtern entflohen. Dick is his guards (DAT) away-fled "Dick escaped (from) his guards." (25) Die Kassette ist seinen Händen entglitten. The cassette is his hands (DAT) away-slipped "The cassette slipped out of his hands." 5.3
How to 'steal' and 'rob' at the same time in German The additional German option is also available in transitive constructions such as (26)-(27), with an accusative subTR and a dative source subLM. Again the constructions are particularly associated with ablative ent-, although they also occur with plain base verbs such as nehmen and stehlen. The resulting
60
ROBERT DEWELL
state of (27) is primarily a new home for the letter, but it also includes an affected former owner as a secondary participant. (26) Sie hat dem Alten ihre Hand entzogen. She has the old man (DAT) her hand (ACC) away-pulled "She withdrew her hand from the old man." (27) Konrad hat ihr den Brief gestohlen. Konrad has her (DAT) the letter (ACC) stolen "Konrad stole the letter from her." The construction with a dative source LM still involves a kind of separation loop, but it does not require transferring our attention to the source LM and ignoring the source-path TR at the end. We only need to allow room in the final picture for an upstream dative participant that lags behind, still in the TR's wake. To describe similar events, English offers only the usual basic alternatives: either a TR-centered accusative-subTR pattern like (28) (Levin's 'steal' verbs), or a LM-centered accusative-subLM pattern like (29) that gaps the vanishing subTR (Levin's 'cheat' verbs). (28) Konrad stole the letter from her. (29) a. Konrad robbed her (of everything she had/ ??of the letter). b. Konrad relieved her of her duties. 5.4 A schematic meaning for the dative Smith (1993) offers a good schematic analysis of the German dative that can easily be adapted to the approach being proposed in this paper. He contrasts the 'asymmetry' characteristic of the clausal accusative with the 'bilateral involvement' of a clausal dative, which means that dative objects are simultaneously object-like (affected by the flow of energy in the action chain) and subject-like ('conceived to exert control over their immediate sphere of influence'). They are active participants that do something, something related to but separate from whatever the other TRs in the clause are doing. They may take possession ('ihm das Buch geben'), lose possession or control ('ihm das Buch nehmen', 'ihm entfliehen'), follow ('ihm vorausgehen/ vorangehen'), lead or be an ongoing target ('ihm folgen/ nachgehen', 'sich ihm nähern'), give orders ('ihm gehorchen'), carry on an activity assisted by the verb activity
CASE MEANING AND SEQUENCE OF ATTENTION
61
('ihm helfen'), or simply undergo an internal reaction to a state of affairs they are aware of ('ihm kalt sein'). The core insights of Smith's analysis can be readily reformulated in terms of the sequence of focal attention. That is, the subject-like nature of dative participants reflects their presence as secondary participants at the beginning of the construal and their object-like nature reflects their continued presence as secondary participants at the conclusion of the construal. Dative objects lack the strong asymmetry of the nominative-accusative relationship because their secondary status is constant throughout the construal. They share the stage with the nominative and the accusative (if there is one), but they never compete with them for central focus. There are two main advantages to thinking of the dative's schematic meaning in terms of relative prominence with a temporal dimension. One is beyond the scope of this paper and w i l l be mentioned only in passing, namely that the analysis extends naturally to the objects of two-way prepositions as well (compare Smith 1993). That is, 'in dem Haus' with a dative object means that the house is constantly present as a locational L M throughout the construal, while 'in das Haus' is marked by the accusative case to mean the location is not a profiled part of the scene at the beginning of the construal but it becomes so at the conclusion. The other advantage is that the schematic meaning becomes clearly distinct from the conceptual content of the action chain, so that we are less apt to confuse the level of focal attention with that of prototypical instantiations. We especially need to avoid presuming that the dative's activity is particularly tied to a downstream position within the action chain, a presumption implicit in Langacker's 'target domain' and clearly evident in Smith's earlier work as well. For example, Smith (1985:395) speaks of dative participants whose 'further action' is a subsequent causal result of the main clausal action, occurring 'after being set in motion by the TR'. 9 It is admittedly tempting to think of dative participants primarily in terms of a downstream position in the action chain as long as we are thinking of prototypical transfers to a recipient, or possibly even of helping or feeling cold; but clearly the activity of the dative object is not downstream from the subject's action in examples with folgen or gehorchen, or with entfliehen, entkommen and entgleiten. The semantic fact is that the dative participant is there in the construal all along, not just near the end — or more precisely, it stands outside the main flow of central attention. Even in a prototypical sentence like 'Dick gave Gerald the keys', the dative recipient is a presence during the whole event, interacting with Dick and motivating his act. We begin with a scene that focuses centrally on Dick but also includes Gerald, and then we shift our principal focal attention from Dick to the keys, while Gerald
62
ROBERT DEWELL
retains his secondary focus. Contrast 'Dick gave the keys to Gerald', which really does introduce Gerald as a subsequent downstream subLM that appears only in the final stage of the construal. Dative objects are in a sense the real 'secondary figures', active but semifocal participants that are always 'also involved', subordinated either to the nominative or to the accusative. Their status is relatively constant, as reflected in the fact that dative remains dative even in passive constructions in German. While we may shift our primary focus rather abruptly from nominative to accusative, the dative is always already there. In English, the German dative objects are consistently translated with coerced direct objects and verbs such as escape (and precede, follow, approach, obey, help, etc., which are less problematic because they do not require a separation loop), or else with double-object constructions and verbs such as give and buy. Otherwise, English has no alternative to oblique phrases with to, from or for, sacrificing the object's participant status. 6.
Summary Speaking very broadly, we might summarize the lessons gained from examining separation loops this way: The accusative case typically directs us to transfer our central focus from the nominative subject to the direct object in the course of construing the event, so that we conclude our construal looking mainly at the accusative. Accusative pathway LMs are atypical variants which fuse the trajectory of the nominative with an accusative place and direct us to look at both together (albeit with ongoing conceptual 'motion' directed from subject to object). English also has 'dative' objects — participants that are still there at the end of the construal, but which never gain central focus — that are coerced into the transitive pattern. We are never allowed to leave an accusative object behind as interpreted history. Dative objects of the verb are included in an event construal as secondary participants outside the main flow of central attention. They have special prominence as participants that are still present and 'doing something' in the final image, somehow interacting with the nominative subject and the accusative object (if there is one) — but they never compete for central focus.10 Notes 1 Langacker (1991:382) contrasts two opposing 'natural paths': the one from primary figure to secondary figure to ground that serves as the basis for the nominative-accusative system (with the nominative as the more basic 'starting point'), and the one defined in terms of conceptual autonomy/dependence that serves as the basis for the ergative-absolutive system
CASE MEANING AND SEQUENCE OF ATTENTION
63
(with the absolutive case as the more basic starting point representing the conceptually autonomous core thematic participant). In effect, I am recasting this description slightly by identifying the core theme of A / D layering with the concluding central focus (the TR of the resulting state). The nominative is thus construed as the beginning central focus (which may or may not be replaced as the centrally focal participant by a marked accusative before the interpretation has concluded), while the absolutive is understood to be the concluding central focus (which may or may not be preceded by a marked focal ergative). I do not mean to deny that the nominative subject will be more salient than the accusative direct object in overall semantic effect. The subject has privileged status as the starting point. Moreover, it is prototypically associated with highly salient human agents that are the responsible, autonomous heads of an action chain. As far as their most schematic meaning is concerned, though, the nominative and the accusative both occupy central focal attention at some point in the construal, in contrast with the dative and with oblique LMs. 2 See Levin (1993:125-128). The 'wipe' verbs are a large class that includes wipe, sweep, brush, rake, wash, lick, suck, squeeze, and many others. The source-path subtrajectory is explicitly introduced by the deceptively simple device of augmenting the verb with the particles off or out (ab- or aus-), even though the subTR itself is completely gapped. Sentences like (7) are essentially like those with prefixes such as de- or ent- — compare 'wiping off a table' and 'dusting a table'. The only difference is that the 'wipe' verb refers specifically to the manner of the agentive activity and the subTR is not even designated generically in the root of the base verb. 3 The so-called 'clear' verbs, according to Levin (1993:124-125), are clear, empty, drain and clean. This class is difficult to define satisfactorily, and German has no real counterpart to it, but for our purposes what matters is that a sentence such as (8) lexicalizes the achievement of a state which results when a distributed subTR is removed from a subLM. That means the constructions always involve a separation loop, even without any explicit augmentation with a particle such as out or off, or an of-pp that mentions the subTR. They are semantically very similar to resultative 'wipe'-verb constructions (compare 'He cleared the table' and 'He swept the table clear'), except that the 'clear' verb does not specify the agentive manner. Since they lexicalize a resulting state, clear, drain, and empty can occur in an intransitive inchoative variant with the source-path L M as a nominative subject when there is no definite agent in the image ('The sky slowly cleared (of clouds)'). Compare Langacker's (1991: 345-355) discussion of 'setting-subjects'. 4 Ent-verbs like enthaupten ('behead') and entkorken ('uncork') do exist, with a root subTR known pragmatically to be an individual with its own shape. Although these verbs are frequently chosen as typical examples, they are actually quite atypical. They are specialized instances in which the pattern for removing a distributed subTR is used to describe removing an individual subTR, thus implying a vaguer reverberating functional effect on the whole L M space — the whole bottle is put in a new state when its cork is removed. Similar comments can be made about the English 'pit' and 'debone' verbs (Levin 1993: 130-131). The overwhelming majority of these verbs incorporate a distributed TR, and an expression such as 'debone the fish' clearly describes removing all of the bones and not just a single bone. Potential counterexamples with incorporated subTRs that are known to be a single defined entity not spread over or through the L M are essential core parts whose removal has effects that reverberate through the whole L M and change its overall state (core, pit).
64
ROBERT DEWELL
5 Unlike most 'clear' verbs, vacate and evacuate do not have nominative-subLM variants (*'The area vacated/ evacuated (of troops)'). On the other hand, clear is not very good with subjects like this either (??'The area cleared (of troops)'. Given that 'The general evacuated his troops from the area' is acceptable, evacuate does alternate an accusative subTR with an accusative source L M (albeit without an of-PP): 'The general evacuated/cleared (the troops from) the area'. Levin (1993) does not mention vacate, and she lists evacuate only as a 'banish' verb. As for German raümen, it behaves like the English verbs treated in this section (e.g. 'ein Hotelzimmer räumen', 'Polizei raümte die Straße von Passanten', 'Bücher vom Tisch räumen'). 6 'Harry left the apartment in shambles' is a good example of an accusative that directs a shift of central focus without necessarily implying an action-chain transfer of energy to a 'patient'. The sentence does not necessarily mean that Harry put the apartment in that state — he could have found it that way. 7 The type is more common in extended variants with abstract pathways such as 'sing a song', and is related to more typically transitive extensions in which the accusative object is also altered as a result of the path that it defines — e.g. 'mow the lawn', 'eat the soup'. There are also temporal extensions like 'all day' or 'den ganzen Tag', which are clearly accusative in German. 8 The atypical accusative of escaping is similar to the atypical nominative of stative experiencer-subject verbs (Levin's (1993:191-192) 'admire' verbs, e.g. like, love, admire, enjoy, esteem, detest, fear, despise, hate), which often correspond to dative-experiencer verbs in a language like Russian. Contrast experiencer-object verbs (Levin's (1993:189-191) 'amuse' verbs, e.g. please, amuse, bore, astonish, surprise, thrill, frighten, scare, terrify), which are basically causal. Experiencer-object constructions have a true 'accusative' (transitive) shift of central focus to a subTR that contains a separate reflexive-TR process; experiencer-subject constructions in English have a 'dative'-style subject that yields primary focus to the direct object but remains in the picture to the end, with a stative yet interactive relation drawn between the subject and the object. The same kind of 'dative' relation is found in possessive constructions (e.g. with have). 9 The tendency to assume vaguely that the activity of dative objects is located downstream from the activity of other participants in the action chain is still subtly evident in Smith (1993), for example when he says of a sentence with folgen (1993:560) that the energy transmitted by the nominative agent 'stimulates some type of reaction' in the dative object. Of course I agree that the interaction between the nominative and dative participants continues throughout the construal, but the follower's motion is certainly not located upstream from that of the dative participant being followed. Incidentally, I do not mean to invalidate semantic analyses in terms of semantic roles within prototypical instantiations, only to guard against confusing that level of analysis with more schematic meanings involving dynamic relative prominence. I agree, for example, with analyses such as Janda's (1993) that consider communicative transfers with the pattern 'She told him a story' to be metaphorically extended from an established prototype construction like 'She gave him a book'. 10 A further advantage of emphasizing the sequence of central focus, aside from opening the way to a schematic account of the dative, is that it encourages us to isolate highly general dynamic construal processes from the 'content' of the scene being described. For example, I suspect that the general role of interpretative 'history' in construing an event is related to
CASE MEANING AND SEQUENCE OF ATTENTION
65
'fictive' constructions (Talmy 1996). They can be understood as constructions in which stative marking directs us to conclude by 'forgetting' the path history and considering only the resulting state to be objectively real. Compare the sequential scanning operations that underlie the generation of any schematic gestalt, but which are normally sublexical as we concentrate only on the result.
References Dewell, Robert B. 1996. "The separability of German über-: A cognitive approach". The Construal of Space in Language and Thought ( = Cognitive linguistics research 8) eds. Martin Pütz & René Dirven, 109-133. Berlin: Mouton de Gruyter. Dowty, David. 1991. "Thematic proto-roles and argument selection". Language 67.547-619. Janda, Laura. 1993. "The shape of the indirect object in central and eastern Europe". Slavic and East European Journal 37.533-563. Langacker, Ronald W. 1991. Foundations of Cognitive Grammar II: Descriptive Application. Stanford: University Press. . 1987. Foundations of Cognitive Grammar I: Theoretical Prerequisites. tanford: University Press. Levin, Beth. 1993. English Verb Classes and Alternations: A Preliminary Investigation ( = Linguistic Inquiry Monographs 26). Chicago: University Press. Smith, Michael B. 1993. "Cases as conceptual categories: Evidence from German". in Conceptualizations and Mental Processing in Language eds. Richard A. Geiger & Brygida Rudzka-Ostyn, 531-565. Berlin: Mouton de Gruyter. . 1985. "Event chains, grammatical relations, and the semantics of case in German". Chicago Linguistic Society 21.388-407. Stiebels, Barbara. 1996. Lexikalische Argumente und Adjunkte: Zum semantischen Beitrag von verbalen Präfixen und Partikeln. ( = studia grammatica 39). Berlin: Akademie Verlag. Talmy, Leonard. 1996. "Fictive motion in language and 'ception'". Language and Space eds. Paul Bloom, Mary A . Peterson, Lynn Nadel, & M e r r i l l F. Garrett, 211-276. Cambridge, Mass.: M I T Press (Bradford Book). . 1992. Path to realization: A typology of event integration. Technical report 92-01 of the Center for Cognitive Science, State University of New York, Buffalo.
Fijian Children's Possessive Categories and Constructions Patrick Griffiths University of the South Pacific,
1.
Fiji
Introduction * The pronoun system of Fijian is elaborate. Possessive proforms constitute a substantial subsystem of this. This report concentrates on Fijian children's learning of the morphology of possessive constructions of their language, mainly in a set of proforms, but also the use of the same morphemes as possessive suffixes on nouns. A table of possessive suffixes w i l l be presented later. It may be useful, however, to make a summary statement now regarding the salient features of the possessive system of Fijian. In the first person non-singular forms, there is an 'inclusive' versus 'exclusive' distinction, depending upon whether or not the addressee is in cluded: "belonging to us and you" as opposed to "belonging to us, but not to you". There is a four-way system of number distinctions: 'singular', 'dual', 'paucal' (denoting groups of three or more, up to about a dozen or a score) and 'plural' (when there are more entities than the number denoted by a paucal form). Two types of possessive relationship are recognised: 'inalienable' and 'alienable'. Inalienable possession is signalled by putting the possessive suffix on to a noun denoting the possession, e.g. the noun ulu means "head" and the suffix -mu is a second person singular possessive marker; so "your head" is na ulumu, not something you can readily give away. Na is an article. For alienable possession, on the other hand, special possessive stems are used as bases for suffixation. By contrast with one's head, a bag (kato in Fijian) is a possession that can easily change ownership, and "your bag" is na kato, not *na katomu. I n na noma kato the suffix -mu is attached to the possessive stem no. No is the 'default' possessive stem. There are two other stems used in constructing alienable possessive forms. They are ke for 'food', possessions that are for eating, are being eaten or have been eaten by the possessor, and me for 'drinks', possessions that have been, are being or w i l l be imbibed by the possessor. Thus there is a considerable amount of detail for Fijian children to gain control over in learning this part of their language. See Geraghty (1994) for
68
PATRICK GRIFFITHS
a concise, non-technical account of the relevant constructions and meanings, and Schütz (1985) for a detailed treatment. Fijian is a VOS language, except that when the subject is a pronoun it goes in front of the verb. 1.1
Overview I n a cross-sectional study of 2- to 5-year-old Fijian children: (a) there are indications of meanings being acquired before mastery of the constructions which conventionally express those meanings; (b) there are strong indications of systematicity in acquisition; and (c), in connection with the 'food' versus 'drink' alienable possessive distinction, prototypicality effects are observable. (a) Possession appears first to be expressed by just naming the possessor, or using a non-possessive pronoun to refer to the possessor, or by simply putting together two nouns in the order Possession+Possessor, not a well formed construction in adult Fijian. The earliest well-formed possessive constructions follow the alienable pattern, in which a possessive suffix is attached to one of the possessive stems. The inalienable pattern noun+suffix is acquired rather later. (b) The distinctions that show evidence of systematic acquisition among possessives are number, for which the starting point is 'singular', and the alienable categories where the 'default' stem no is the first to be used. (c) Two observations testify to prototypicality as a foundation for the cognitive organisation underpinning possession in Fijian. One is that there are restrictions on the meanings associated with the ke stem: the 24 children in the sample gave no evidence of an abstract use that adults have for ke-based possessives. Secondly, there are some errors whereby certain non-liquid foods, which nonetheless belong in the 'drink' category — a category for which they could be regarded as non-prototypical — are referred to by children using constructions with the 'food' or 'default' possessive stems. Arguments in support of these conclusions are set out in section 2 of the paper. 1.2
The data The child language data on which this report is based come from a project carried out jointly with my colleague Asela Bavadra. The subjects were 24 children whose first language is Fijian. Each child was recorded once only. The investigation was therefore cross-sectional, rather than longitudinal. The children were spread across the age range 1 year 10 months to 4 years 10 months. I n the 24 transcripts there are 346 occurrences of possessive forms. Discussion w i l l principally be concerned with these, though I begin with some
FIJIAN CHILDREN'S POSSESSIVE CATEGORIES AND CONSTRUCTIONS
69
instances in which children conveyed possessive meanings without producing conventional possessive expressions. The methodology of the study is described in greater detail in Griffiths & Bavadra (forthcoming), a paper which also reports more general findings about the first language acquisition of Fijian pronouns. A leap has to be made in inferring developmental sequences from cross-sectional data. The researcher has to assume that the younger children w i l l eventually come to have language systems very similar to the older ones, and that the older children were previously linguistically like the younger ones, i.e. that all the children are on essentially the same trajectory. Cross-sectional sampling is appropriate for exploratory research in language development, however. And, until there is evidence to the contrary, it is a justifiable initial assumption that all the children are following the same developmental path. I have not found any other studies of Fijian first language acquisition; so the work reported on here is indeed exploratory. A start has been made on language acquisition research in a related language, Samoan (Ochs, 1985, for instance). In the presentation of examples, I replace personal names with plausible pseudonyms, chosen because they have equivalents in English. 2. 2.1
Patterns in the data Ill-formed expressions of possession There are instances where the intention to express a relationship of possession seems, clearly enough, to be there, but the child does not explicitly mark possession. I first quote cases in which the child avoids the morpho logical choices that have to be made for the production of well-formed Fijian possessive constructions. Morphological contrasts side-stepped. A child can evade the complexities of Fijian possessive marking by just naming a possessor. The result is generally understandable but ill-formed. Examples (1) and (2) show this starting point for talking about possession. These examples, as well as others below, also illustrate how adult possession questions are worded. I n example (1) the child is asked whom a serving of rice standing nearby is for. He answers with his younger brother's name. (1) O cei na kena na raisi? O ART who ART 'food'-3rdSgPOSS ART rice? ART Who? "Whose is the rice? Who?" [Adult]
cei?
70
PATRICK GRIFFITHS
(1) O Samu. A R T Sam "Sam." [Child 9 (3;9.15)] 1 I n example (2), the same child, who was being recorded in his family's home, is asked whose house it is. He supplies his own name as an answer. (2) O cei nona vale qo:? A R T who 'default'-3rdSgPOSS house this "Whose house is this?" [Adult] Siti. A R T Steve "Steve." [Child 9 (3;9.15)] Example (3) shows another pattern. Here the possessor's name is put into construction with a label for the possessed object. This too is ungrammatical, but, in our data, it appears to be the most basic way of producing some sort of construction for conveying the possessive relationship. I n this respect Fijian children are similar to children acquiring English as a first language, where Mummy sock can mean "Mummy's sock" (Bloom 1970), though it is found in the present corpus at a later age than the second year, which is when English speaking children characteristically produce Noun+Noun possessives. This may be a pointer to the complexity of Fijian possession as a language learning task. I n example (3) the child is asked who a bag in the room belongs to. She starts to respond with the preposition used in the question, but then just juxtaposes the owner's name with the word denoting the possession. (3) Na kato nei cei? A R T bag of who "Whose bag?" [Adult]
Nei... kato Pita. Of bag Peter "Of... Peter bag" [Child 10 (3;9.6)]
Non-possessive pronouns used as possessives. Examples (1-3) showed children expressing possession without recourse to morphology. There are 18 tokens in the data, from 7 children, where non-possessive pronouns are pressed into service as possessives. I n almost all of these the utterance is not well formed, but they can be rated as more complex than (1-3) because virtually any use of a pronoun in Fijian generally requires some morphological distinctions to be observed. Examples (4) and (5) are illustrative. I n (4), the two research assis-
FIJIAN CHILDREN'S POSSESSIVE CATEGORIES AND CONSTRUCTIONS
71
tants, K T and L V , ask Child 1 about the ownership of a toy car on the floor. He answers with a non-possessive pronoun. His mother then shows, in a recast, how a preposition could turn his answer into a well-formed one, i f he insists on not using possessive pronouns. (4) Nei cei na motoka:? of who A R T car "Whose is the car?" [KT]
M? hm? " H m ? " [Child 1 (2;8.15)]
Na noqu. Nei cei? A R T 'default'-lstSgPOSS of who
Koya. 3rdSg
"Mine.
" H i m / H e r " [Child 1]
Nei of
Whose?" [ L V ]
koya. 3rdSg
" O f him/her." [Child 1's mother] I n (5) the child answers a question about the ownership of the house by claiming that it belongs to her and her four-year-old cousin, Kata. (5) Nei cei of
na
vale
qo:?
who A R T house this
"Whose is this house?" [Adult] Kata, keirau mama. Kate 1stExclDu both. "Kate, both us-two." [Child 11 (2;11.27)] Use of non-possessive pronouns requires the child to select grammatical number: examples (4) and (5) showed a singular and a dual. And person contrasts have to be handled as well: 3rd person in (4) and 1st person (exclusive of addressee) in (5). However, by answering like this, the children are avoiding further choices that have to be made for appropriate use of Fijian possessive proforms. 2.2
Possession expressed with possessive proforms I n example (6), Child 11 is asked about a radio standing in the room. She first answers with a non-possessive pronoun — a grammatically ill-formed answer — and then, after doubt is expressed over the truth of her answer, reaffirms it, but this time with the appropriate possessive pronoun.
72
PATRICK GRIFFITHS
(6) cei nona re:tio: ART who 'default'-3rdSgPOSS radio this "Whose radio is this?" [Adult]
qo:?
Au. 1stSg " M e . " [Child 11 (2; 11.27)] Seti, seni ! disapproval-interjection not 'default'-2ndSgPOSS "Tut! It's not yours." [Adult] Noqu! 'default'-lstSgPOSS " M i n e ! " [Child 11] Example 6 shows three of the possessive suffixes of Fijian, looking at what the adults said as well as the child's utterances: -na 3rdSgPOSS, -mu 2ndSgPOSS and -qu 1stSgPOSS. I n each case they are attached to a stem no, the 'default' alienable possessive stem. This is a suitable point to provide more details about Fijian possessive constructions. 2.3
The Fijian system of possessive suffixes Table 1 catalogues the possessive suffixes of Fijian. They attach to possessive stems and to nouns denoting possessed objects. The stem no has been exemplified several times above. No is used for possessions that are not possessed for eating or drinking. I n Fijian, the word for "chicken" is toa. I am a vegetarian. I f I had a chicken it would be a pet, not possessed for eating, and I could talk of it as noqu toa " m y chicken". I f I was not a vegetarian and possessed a chicken with a view to eating it, then that would be kequ toa "my chicken". There is another possessive stem me for things that the possessor drinks, e.g. mequjusi " m y fruit juice that I drank, am drinking or w i l l drink". In the 1st person exclusive row of Table 1 there is some allomorphy which should be mentioned. The forms of the suffixes as shown in the table are for attachment to nouns, when one is constructing inalienable possessives. When they go on to the possessive stems no, ke and me the sequence -ike- is omitted. Furthermore, when these non-singular 1st person exclusive possessive suffixes are applied to no the stem then surfaces as an allomorph ne. Example (7) illustrates this. Child 15, commenting on a tap from which water no longer flows, because the settlement's supply had been cut off that day, correctly uses neitou, instead of *noikeitou.
FIJIAN CHILDREN'S POSSESSIVE CATEGORIES AND CONSTRUCTIONS
73
Number singular
dual (exactly 2)
paucal (3-~20)
plural ( > paucal)
1st Excl
-qu my
-ikeirau us-two's, not yours
-ikeitou our group's, not yours
-ikeimami us-many's, not yours
1st Incl
—
-daru me-and-yours
-datou me&you& others in a group's
-da me&you& many others'
2nd
-mu your, Singular
-mudrau you-two's
-mudou your group's
-muni you-many's
3rd
-na her/his
-drau them-two's
-dratou their group's
-dra them-many's
Person
Table 1. Fijian possessive suffixes
(7) Sa: maca na neitou ASP empty ART 'default'-lstExclPauPOSS pipe "Our water's run dry." [Child 15 (3;4.25)]
palpo.
2.4 Systematic aspects of the acquisition of the Fijian possessive system Number. Every one of the 24 children used at least one singular possessive suffix, and 10 of them manifested only singular possessives. Eighty-four percent of correct possessive uses were singular. Nine children had not only singular but also paucal possessives. Although paucal possessive forms were greatly outnumbered by singulars in our data, paucals occurred twice as often as dual and plural possessives added together. Only one child produced possessives in all four number categories. The remaining four children had Sg&PIu, S g & D u & P l u , Sg&Du&Pau, Sg&Pau&Pl. From all this I infer that number distinctions are fairly systematically acquired in the possessive system, in the following order (where ' > ' indicates that the item to its left is acquired earlier than the item to its right): singular > paucal > {dual, plural}
74
PATRICK GRIFFITHS
Stem categories. Among the possessive stems it is the 'default' (neither for eating nor drinking) stem no that appears to be acquired first. It is over whelmingly the most frequent in our child data (82% of tokens). A l l 24 of the children use it and nine of them were not recorded with any other possessive stem. The 'food' stem ke appeared in the transcripts of nine children, with a total of 30 tokens. Eight children used me the 'drink' stem (16 tokens). The acquisition order that I infer from this is: no 'default' > ke 'food' > me 'drink' The dearth of errors. Errors are not common in our corpus. Only 8 of the children were detected in making possessive errors and the total number of tokens involved was only 24, by comparison with 340 correct uses. One set consisting of 18 of the 24 possessive errors has already been illustrated. They were the ones where a non-possessive pronoun was used in place of a possessive form. The many correctly used possessive proforms included quite a number of instances where the complicated allomorphy mentioned earlier (e.g. neitou < *noikeitou) was successfully negotiated. A possible explanation for this impressive level of accuracy is that children start with possessive constructions that they have heard from more proficient speakers and that they tend to be cautious about generalising beyond what has been attested in their experience, a beginning reminiscent of Tomasello's (1992) account of his daughter treating English verbs as 'islands'. Some of the children having noqu Na:2 "my M u m " as their only possessive phrase tends to support this idea. However, they soon do generalise. For instance, Child 3 (2; 11.15) gave us 20 possessive proforms, all first person singular. Eighteen were noqu " m y " but it was in construction with not just Na: " M u m " , but also with words meaning " D a d " , "Granddad", " b a l l " and "car". His other two possessives were correct uses of the 'drink' stem in the phrase mequ loli "my l o l l y " , which he said twice while sucking a lolly. (Things that are sucked, like lollies, count as drinkable.) Which possessive stem to use in an answer can often be picked up from the question. I n example 1, for instance, Child 9 could have discovered or confirmed from kena in the question that raisi "rice" for his little brother to eat is in the ke category, as a basis for constructing a correct possessive answer, such as na kena o Samu "Sam's". Example (8) shows one of the frequent instances we have in the data of a possessive stem matching across a question and answer pair; so this quite likely is a way in which the children learn which stem to use. I n this example, it can be seen that the questions provide three cues that ke is the appropriate stem.
FIJIAN CHILDREN'S POSSESSIVE CATEGORIES AND CONSTRUCTIONS
(8) O
cei
na
kena
sos
ya:?
75
Ya:?
ART who ART 'food'-3rdSgPOSS sauce over-there over-there "Whose sauce is over there? Over there?" O cei na kena? Ya:? Kei Art who ART 'food'-3rdSgPOSS over-there ' food'-of who? "Whose (food) is it? Over there? Whose (food)?"
cei?
[Adult] Kequ. 'food'-1stSgPOSS "Mine (food)." [Child 5 (2;4.5)] However, possessives were also used by the children to assert or demand possession, and to construct noun phrases for use in sentences that were not answers, or were not answers to questions about ownership. I n such cases the child does not get discourse cues regarding which stem to use. Furthermore we have a few instances in which children went against the model offered in a question, as in example (9), below. 2.5
Prototype effects The small number of errors in which wrong possessive stems are used is interesting. They generally involved children failing to include in the me 'drink' class items that belong there. Child 6 (2;0.10) wrongly used kequ (constructed on the ke 'food' stem) instead of mequ for a lolly. I n Fijian anything that is sucked belongs in the 'drink' category, and lollies are commonly sucked. Child 10 (3;9.6) used keitou ('food'-1stExclPauPOSS) and neitou ('default'-lstExclPauPOSS) for sugar, which also counts as a drinkable (because it too melts in the mouth into a liquid and has as, probably, its most common use in F i j i that of being an additive in tea, which again means that it goes down the throat as a liquid) and should therefore have been spoken about with meitou ('drink'-lstExclPauPOSS). Child 24 (3;9.12), speaking of citrus fruit (moli) that he said he had picked, used both keitou ('food'lstExclPauPOSS) and neitou ('default'-lstExclPauPOSS), instead of the required meitou ('drink'-lstExclPauPOSS), citrus being juicy fruits. The learning needed would appear to be the semantic subtleties of the Fijian 'food'/'drink' boundary. A reasonable surmise is that prototypical substances in the 'drink' category are liquids in our ordinary experience, and citrus fruits, lollies and sugar are non-prototypical in only becoming liquid as they are taken in.
76
PATRICK GRIFFITHS
A further pointer to an influence of prototypicality is that there are no instances in our data of an abstract use of the possessive stem ke. As well as being for 'food' possession, ke is used to mark deverbal nominals in which the possessor is not the agent of what the verb denotes, e.g. a punch that you receive, a picture of you or a story about you. See the account of 'PASSIVE POSSESSION' in Geraghty (1983, 248f). This suggests that the children had induced a simpler category for ke, taking it as covering just ' f o o d ' , with the abstract use perhaps being learned beyond the 4; 10 upper age boundary of our sample. 2.6
Possessive suffixation on nouns The suffixes shown in Table 1 can be added directly to nouns. When they
are, ' I N A L I E N A B I L I T Y ' is signalled.
There are 16 occurrences in the corpus of possessive forms constructed by direct suffixation to nouns. Examples (9) and (10) w i l l serve for illustration. I n example (9) the child is asked what kin relationship another little girl bears to her. (9) Qo: this
nornu cava? Nomu cava qo:? 'default'-2ndSgPOSS what 'default'-2ndSgPOSS what this
"This (one) is your what?
Your what is this (one)?" [Adult]
Taciqu. younger-sibling-1stSgPOSS "My younger sister." [Child 10 (3;9.6)]
(10) cei drau dau moce vota na ART who 2ndDu usually sleep together ART night "Who do you usually sleep with at night?" [Adult]
bogi?
Noqu... na taciqu. 'default'-lstSgPOSS ART younger-sibling-1stSgPOSS "My... my younger sister." [Child 23 (4;2.21)]
The false start in example (10) is interesting. It suggests that she was going to put her younger sister into the no possessive category, which is what she, entirely appropriately, did with the words Na:, : and Bu.bu: ( " M u m " , "Dad" and "Grandmother", respectively) in the same transcript, but that would have been an error and she avoids it. Another child did not self-correct a simi lar error, as seen in the final utterance of example (11), where Child 16 is
FIJIAN CHILDREN'S POSSESSIVE CATEGORIES AND CONSTRUCTIONS
77
asked about relationships in a picture that shows a mother, a father, a boy called Qalo and Qalo's younger brother. The questioner points out the individuals in the picture as she asks about them. (11) Nona
cava o
Qalo qo:?
'default'-3rdSgPOSS what ART Qalo this "Qalo's what is this (one)?" [Adult] Nona Na:. E? Va:cava qo:? 'defauIt'-3rdSgPOSS Mum Eh? how this "His M u m . " [Child 16 (4;6)] "Eh? How about this (one)?" [Adulti Nona Ta:. Qo:? 'default'-3rdSgPOSS Dad this "His D a d . " [Child 16] Nona gone. 'default'-3rdSgPOSS child "His child." [hid16]
"This (one)?" [Adult]
I n saying Nona gone "his c h i l d " , it seems that Child 16 was wanting to indicate that the smallest boy in the picture was the child of the father who, in the picture, was helping him ride a tricycle. This is an ill-formed response. She should have said Luvena (luve "offspring" + na 3rdSgPOSS), w i t h the possessive suffix attached to a noun. The number of different nouns occurring in our data with direct possessive suffixation is small: ulu "hair", gusu "mouth" and taci "younger sibling". This suggests rote learning. However, there is an indication of systematic acquisition too. Every instance we recorded of possessive suffix ation to a noun is singular, i.e. the suffix is a singular one. Given the restric tion to singular number, I infer that direct suffixation to nouns is acquired after proforms constructed with me, which also appeared 16 times in the data and also came in the records of 8 different children, but were found with singular, paucal and plural suffixes. This makes it possible to offer the following order of acquisition for Fijian possessive constructions (where ' > ' indicates that the morphological construction to the left is acquired before the construction to the right): n+suFFix 'default' > k e + s u F F I X 'food' > me+suFFIX 'drink' NOUN+SUFFIX
>
78
PATRICK GRIFFITHS
The first three in the sequence are predominanlty alienable possessives and they are proforms. The last type in the sequence is inalienable and non pronominal. 3.
Conclusion Fijian expression of possession is complicated. Table 1 shows 15 possessive suffixes, varying according to person and number. Each suffix can appear on any of three possessive stems {no 'default', ke ' f o o d ' , me 'drink') to signal 'alienable possession', or on common nouns to mark 'inalienable' possession. The cross-sectional dataset that has been reported on here, from 24 Fijian children, points to a systematic developmental sequence in which possession is first conveyed without morphological marking (by just naming the possessor or via an ill-formed construction Possession + Name of Possessor); later by using a pronoun of appropriate person and number (but ill-formed because it is not possessive). Next comes possessive suffixation, starting with the default stem and singular number, progressing eventually to suffixation on common nouns. Four aspects of this sequence suggest that cognition may be a driving influence: (i) Possessive meanings are expressed before the necessary constructions are controlled, (ii) Singular is intuitively simpler conceptually than the other numbers, (iii) The default possessive category (using the no stem) likewise appears conceptually simpler than the other two ('food' and 'drink'), (iv) It might be that 'alienable' forms are learnt before 'inalienables' because possession is easier to understand when it can be seen to be transferable. That the children in this study were learning the possessive categories as semantic ones is hinted at by prototypicality effects: the transcripts yielded no instances of an abstract use that is available to adults for the ke 'food' stem; and, in a few cases, non-liquid items, such as sugar, failed to be put into the me 'drink' category, where they belong in Fijian. Children's language errors are informative. They are signs that generalisations are being made. However, the low rate of error in the present study (24/364 = 0.003) is itself an observation in need of explanation. Because this investigation was cross-sectional and relatively small in scale, the conclusions have to be regarded as tentative. Data on the frequency of Fijian children's exposure to different forms, patterns and categories must be gathered and brought into consideration.
FIJIAN CHILDREN'S POSSESSIVE CATEGORIES AND CONSTRUCTIONS
79
Notes * The project was funded by the Research Committee of the University of the South Pacific (award 6133-1201-70766-15). I am grateful to my co-researcher, Asela Bavadra, for her contributions to all aspects of the investigation. Of especial value in connection with the present paper was her checking of the translations in the examples used here. Karalaini Tubuna and Louise Vakamocea have my gratitude for making the recordings and putting considerable time and effort into transcribing the tapes, a task on which we worked as a team. Mrs Vatira Cereilagi provided invaluable liaison with the community. I offer her my thanks. Two anonymous referees made useful suggestions for improving the paper, not all of which I am able to adopt without undertaking further research. Finally, but very importantly, I express my appreciation to the 24 children who were recorded in Nadonumai settlement and to their families. The research could not have been done without their cooperation. 1 Children's ages are given in brackets in the form years;months.days. The following abbreviations are used in the glosses: 1st '1st person', 2nd '2nd person', 3rd '3rd person', ART 'Article', ASP 'aspect marker', Du 'dual', Excl 'excluding 2nd person', Pau 'paucal', Pl 'plural', POSS 'possessive', Sg 'singular'. Long vowels are marked with a colon. 2 It is a fact about Fijian that not all inalienable possession is explicitly marked as such. The terms for quite a number of family members simply go into possessive phrases constructed with the 'default' stem no. Noqu Na: "my M u m " is one such. This does not mean that alienability is not an important contrast in the system. It is, because when a possessive suffix is put directly on to a noun we can be sure that we are dealing with an inalienable, e.g. Tamaqu "my father", which has the possessive suffix -qu attached to the noun Tama "father". The inference does not go the other way round, however: when the suffix appears on the possessive stem no 'default', the possession is often alienable, but in some cases is not. References Bloom, Lois. 1970. Language development: form grammars.
and function
in
emerging
languages.
(=
Oceanic
Cambridge, Mass.: M I T Press.
Geraghty, Paul. 1983. The history
of the Fijian
Linguistics Special Publication, 19). Honolulu: University of Hawaii Press. . 1994. Fijian phrasebook. Hawthorn, Victoria: Lonely Planet Publications. Griffiths, Patrick & Asela Bavadra. Forthcoming. "Systematic development in preschool Fijian children's production of pronouns". Ochs, Elinor.
1985. "Variation and error:
a sociolinguistic
language acquisition i n Samoa". The crosslinguistic acquisition.
approach to
study of language
Vol.I: The data, ed. Dan I. Slobin, 783-838. Hillsdale, N.J.:
Erlbaum. Schütz, A . J . 1985. The Fijian
language.
Honolulu: University of Hawaii
Press. Tomasello, Michael. 1992. First verbs: a case study of early grammatical velopment. Cambridge: Cambridge University Press.
de
Facing up to the Meaning of 'face up to' A Cognitive Semantico-Pragmatic Analysis of an English Verb-Particle Construction Beate Hampe Friedrich Schiller University, Jena, Germany
81 "No longer does man face a problem, he faces up to it." ' (TIMES, 8 September, 1958) 1.
Goals of this analysis The above quote from The Times implies that the use of the phrasal verb is just an example of 'wordy' speech and that the adverbial particle and preposition can be regarded as superfluous. Such opinions can also be found in the literature on phrasal verbs (cf. Potter 1965:289). The purpose of this paper is to pin down exactly in which respects, both semantically and pragmatically, the phrasal verb to face up to differs from the simple verb to face in the relevant, non-literal sense. Even though it is true that the simple verb is often used in similar contexts denoting a seemingly similar state of affairs (in fact, in the corpora checked, the simple verb is used in the overwhelming majority) it will be shown that the particle is not an arbitrary, meaningless or redundant addition to the verb, but that its presence is meaningful in the sense of 'motivated', determining both semantic and pragmatic properties of the construction. Parallels to other verb-particle constructions with up (to/with) will be drawn in the course of the semantic analysis. While the description of the phrasal verb's meaning is based on the general approach offered by cognitive semantics, the pragmatic properties of the construction are determined by means of frequency counts of both verbs in different types of text and discourse as well as by an analysis of the characteristics of the immediate linguistic environments of hundreds of actual occurrences of both verbs in a corpus of spoken and written British English. 2.
The phrasal verb in lexicography Lexicography has only recently started to deal with the existence of such seemingly redundant phrasal verbs as to face up to. The phrasal verb is not recognized at all in many dictionaries of the German-English and monolingual variety before 1950.1
82
BEATE HAMPE
Early attempts to characterize the phrasal verb have mainly stated its synonymy with some senses of the simple verb to face. None of the major monolingual dictionaries of 20th century British English which recognize the existence of the phrasal verb allows the use of the phrasal verb in any of the following senses of the simple verb: 'to have/turn the face to or in a certain direction', 'to be situated opposite of (of inanimate things)', 'to look in the face of/stand confronting','to meet face to face (weak sense)', 'to present itself to (of a problem etc.)', 'to cover with a layer of a different material'/'to put a facing on' or 'to coat tea with a colouring substance'.2 The only senses in which both the simple and the phrasal verb are said to be possible are the following: 'meet confidently/accept and deal with/ confront/oppose', 'show no fear at time of trial/danger/difficulty' and 'recog nize existence of (facts etc)'. 3 The Universal Dictionary of the English Language, edited by H.C. W y l d in 1932, for example, defines the phrasal verb by relating it to the simple verb in the following way: (i)
to face:
to be ready to undergo examination, trial etc. to dare, have courage, to come face to face with to confront, to oppose, also "face up to", to recognize existence of, and be prepared to meet, also "face up to ".
But even in those cases where (more recent) monolingual dictionaries try to provide the phrasal verb with a definition of its own, these definitions do not actually set apart the meaning of the phrasal verb from that of the simple one in one of the three relevant senses (c.f. e.g. (ii) Collins Cobuild: English Language Dictionary, edited by J. Sinclair in 1987, (iii): Oxford English Reference Dictionary, edited by J. Pearsall in 1995): (ii)
to face:
15. I f you ~ sth. difficult or unpleasant, it is going to affect you and you have to deal with it. 16. I f you ~ the truth, a fact or a problem, you accept that it is true or really exists and respond to it in a suitable way although you would prefer to ignore it.
to face up to: I f you ~ a difficult situation you accept it and deal with
FACING UP TO THE MEANING OF 'FACE UP TO' (iii) to face:
83
meet resolutely or defiantly, confront, not shrink from, present itself to (sb.)
to face up to: accept bravely, confront, stand up to I have so far come across one dictionary only (Lexicon of Contemporary English, edited by T. McArthur in 1981) which approached the meaning of the phrasal verb differently, defining to face up to as an emphatic version of the simple verb: (iv)
to face:
to go, turn or look towards or standing looking at (Fig.) The men faced great dangers fearlessly.
to face up to: (emph.) to face, accept (sth. unpleasant) confront = face or face up to From these observations the following questions arise: 1. What exactly is the semantic difference between the simple and the phrasal verb? 2. Why is the adverbial particle compatible with some senses of the simple verb and not with others? In which way can this situation be accounted for by an appropriate notion of compositionality that seeks to uncover the motivations underlying the construction while viewing all elements of the verb-particle construction as meaningful? 3. Can a semantic analysis explain the increased emphaticness of the phrasal verb? Can the semantic properties of the verb-particle construction be seen as motivating its pragmatic characteristics? 3.
The phrasal verb as a marked lexical item What seems to be clear from the cursory look at the dictionaries is that the phrasal verb to face up to can be said to be roughly synonymous with some of the non-literal senses of the polysemous simple verb to face (1a-c), while it cannot combine with the various literal (both spatial and specialized) senses (2a-d): (1) a. During the last months we have faced (up to) a lot of difficulties. b. They have learned to face (up to) the facts of parenthood. c. has managed to face (up to) his illness.
(aper
84
BEATE HAMPE
(2) a. Grandmother was still facing (*up to) the wall, not talking to anybody. b. She turned round and faced (*up to) her visitors. c. Our living room windows face (*up to) the main street. d. The dress-maker has faced (*up to) the woolen coat with silk. But even concerning the abstract senses of to face ('to confront, meet'), the phrasal verb cannot always replace the simple verb (3). (3) a. We are facing (up to) great problems, b. Great problems are facing (*up to) us. The fact that replacing the phrasal verb by the simple one always yields an acceptable sentence but not vice versa, suggests that the meaning of the phrasal verb is more specific than that of the simple one. This is not at all unexpected, assuming, that is, that the particle up is a meaningful element of the construction. It also corresponds to the fact that in my corpus of 3.5 mill. words containing samples of both spoken and written British English, the simple verb in its relevant abstract senses was used much more frequently than the phrasal verb, with the particle occurring in only 6.01 % of all the relevant non-literal contexts.4 As a first approximation, we claim that the verb-particle construction is more marked than that containing the simple verb, for the following reasons: 1. The verb particle construction involves additional linguistic material and is thus more 'costly'. 2. The verb particle construction is significantly less frequent than the simple verb, even if only the relevant senses of the simple verb are counted (cf. also section 5.1). 3. The simple verb can occur in all the contexts that the verb particle construction occurs in, but not vice versa; the verb-particle construction is therefore more restricted in its use. 4. Moreover, the phrasal verb is additionally marked stylistically as an informal lexical item. At the same time we are aware of the fact that this perception of 'markedness' itself is only the observation of a surface phenomenon and needs further explanation. The following cognitive semantic analysis will show in which way the meaning of the phrasal verb can be said to be more specific and what sort of conceptualizations are 'behind' the markedness perceived.
FACING UP TO THE MEANING OF 'FACE UP TO'
85
4. A cognitive semantic analysis of 'to face' and 'to face up to' 4.1 Foundations A semantic analysis of the kind we are employing here rests upon the recognition of the fact that linguistic meaning can only be analysed with reference to the conceptualizations that give rise to it. The following analysis of the meaning of the verb-particle construction to face up (to) is therefore dependent on the basic assumptions of cognitive semantics as they are outlined in Langacker's Cognitive Grammar under the cover term 'construal'/'conven tional imagery' (cf. e.g. Langacker 1988/1993), the theory of metaphor and image-schemas (cf. e.g. Lakoff & Johnson 1980, Lakoff 1987, Johnson 1987, Lakoff 1990, Lakoff 1993), the theory of 'mental spaces' (cf. Fauconnier 1994) and the notion of 'conceptual blending' (cf. Turner & Fauconnier 1995). It is also compatible with central notions of construction grammar (cf. e.g. Goldberg 1995, Fillmore 1988). Any analysis of the meaning of the phrasal-prepositional verb to face up to needs to take into consideration the central literal meanings of its highly polysemous components, though none of these literal meanings features in the construction itself. However, all the central literal meanings serve as source domains for the construction of the non-literal meaning of the construction as a whole, the target being the abstract realm of our being confronted and dealing with problems and difficulties. The following analysis does not claim that the meaning of the phrasal verb is 'compositional' in the sense of being completely predictable. Rather, it wants to point out that a small range of conceptual metaphors which are conventionalized in English is responsible for a series of extensions of the meaning of its components and thus 'motivates' the meaning of the con struction as a whole. In other words, the components point in a heavily abbre viated manner at the complex (though conventionalized) conceptual blend 'behind' the construction as a whole. 4.2
The meaning of 'to face' The literal meaning of the verb to face is derived from the noun denoting the part of the head we call face. In its central literal meaning the verb denotes a spatial configuration in which a living agent is situated in front of/opposite an (animate or inanimate) object (the 'landmark') so that s/he can (visually) perceive it (4a-c): (4) a. Grandma was still facing the wall, not talking to anybody. b. She turned around and found herself facing a lion. When they came round the corner they were facing a huge rock.
86
BEATE HAMPE
By means of a shift of the 'deictic centre', a 'body projection' of faces onto objects (in other words the creation of fronts) and possibly a metaphorical construal that ascribes agenthood to external objects, it is also possible to re verse the perception of trajector and landmark in a scene as the one described by example (4c) and construe it in the following way: (4) d. Around the corner a huge rock was facing us, obstructing our way. Via a projection from animate to inanimate involving the creation of fronts just mentioned, the verb to face conventionally also refers to the (perceived) fronts of objects, e.g. buildings, which are situated opposite other objects. (4) e. All our sitting room windows face the main street. To sum this up in cognitive grammar terms, in each of these transitive constructions the subject-NP instantiates a trajector whose 'active zone' (the face or front) is situated opposite a landmark which is instantiated by the direct object NP. It is, however, the central literal sense (4a-b) which is the source for further extensions of meaning. Under the EVENT STRUCTURE metaphor, location version (cf. Lakoff 1993:219-229/Lakoff, Espenson & Schwartz 1991), activity is conceptualized as motion, and purposes are locations to be reached. Thus, purposeful activity can be understood as movement along a path leading to that goal. Objects obstructing that path are impediments to motion, preventing further progress. Via the ontological metaphor ABSTRACT THINGS ARE CONCRETE OBJECTS, difficulties and problems can be understood as such concrete obstacles. It comes as no surprise that humans can face abstract things the same way as they do concrete objects once the vast and complex conceptual background that is provided by the EVENT-STRUCTURE metaphor is taken for granted, i.e. recognized as conventional in English (5a): (5) a. She has really faced a lot of serious difficulties lately. As with the literal use of the verb, the perception of what is landmark and what is trajector can also be reversed (5)b: (5) b. These are the difficulties that have faced her recently.
FACING UP TO THE MEANING OF 'FACE UP TO'
87
Sometimes, both the literal reading and the reading under the EVENT metaphor are simultaneously possible, e.g. when the landmark object is a problem, e.g. potentially dangerous, as in (5c): STRUCTURE
(5)
She turned round and found herself facing a lion.
Situations such as the one described by example (5c) provide a kind of 'cognitive bridge' between the literal and the non-literal sense, since both senses are simultaneously active and thus provide an actual correlation in experience. They have recently been described as 'primary scenes', the results of which are 'primary metaphors' (cf. Grady 1997a,b). Since they are directly grounded in experience, they serve as the foundations for more complex conceptual structures elaborating upon and integrating these 'basic' ones (cf. ibid.). Interestingly, the corpus analysis5 of all instances of to face up to and to face in the non-literal sense revealed that the object NPs instantiating the landmark mainly stemmed from the semantic fields of difficulties and problems (or were explicitly described as difficult, unpleasant, challenging etc.), indicating that the EVENT-STRUCTURE metaphor indeed provides the basis for the extension of the meaning of to face (see Table 1). In the target domain of event structure, the verb to face also has emergent structure of its own due to the nature of the target domain, i.e. it also has structure which does not stem from the metaphorical mapping providing a set of fixed correspondences between the source and the target domain. In the case of the event structure reading of the construction to face + NP, being confronted with an entity that is a potential problem or danger may (urgently) require a certain course of action. This is neither a central feature of the central literal meanings of to face denoting perceived constellations in space, nor of its abstract readings: whether such an action is taken or not, is something that the construction itself does not reveal. In other words, there is nothing in the EVENT-STRUCTURE reading of the linguistic construction itself that determines whether an animate subject is merely 'confronted by' or else 'actively confronting and dealing with' the entity instantiating the landmark. Hence, the construction in the non-literal sense is ambiguous concerning the semantic role of its grammatical subject — it could either be construed as AFFECTED or as AGENT.
88
BEATE HAMPE
semantic field of object-NP instantiating the landmark
to face (total: 297 examples)
to face up to (total: : 19 examples)
117
3
task, challenge, decision
40
5
hardship, danger, risk
18
1
future, change
17
3
reality, things as they are
15
6
criticism, failure
12
0
various others
78
1
problem, crisis
Table 1: Most frequent semantic fields of direct-object NPs
The following sentences show that with the simple verb both construals are actually possible; the selection of the proper reading is determined by the context: (6) a. We doubt that he is capable of meeting the challenges he is facing now. (= 'is confronted with') b. Don't worry, she has faced problems like this for years and will manage. (= 'has confronted and dealt with') Further linguistic means, such as modal verbs and adverbial adjuncts, are often necessary to avoid ambiguity and limit possible interpretations to one of the readings: (7) a. We have known her to face reality with as brave a heart as she could. b. God, help us to know that when we face life boldly, we won't be lost. Labour should face the challenge. As can be seen from the examples below, this is where the phrasal verb comes in. The answer as to why it is possible that to face up to a
FACING UP TO THE MEANING OF 'FACE UP TO'
89
challenge entails meeting it, is also the description of the semantic contribution of the particle to the meaning of this construction: (8) a. He suddenly faced that huge challenge, but was unable to meet it. b. *He suddenly faced up to that huge challenge, but was unable to meet it. 4.3
The contributions of the particle and preposition to the meaning of to face up (to) ' The particle up in its central literal sense denotes verticality in oriented physical space, a notion that has arisen out of our bodily experience (of gravity), i.e. the relative position of head and feet to each other when we stand, sit or move. Thus, an experiential correlation exists between verticality and being awake, active and conscious, giving rise to such well-known conceptual metaphors as ACTIVE IS UP, WELL-BEING IS UP, RATIONAL IS UP, CONTROL IS UP (cf. Lakoff & Johnson 1980/Lakoff, Espenson & Schwartz 1991). That the notion of verticality and its metaphorical extensions also underlie the use of the particle up in verb-particle constructions was first described by Susan Lindner (cf. 1983), although without the help of the concepts that metaphor theory has provided since. Lindner (Lindner 1983:132) discovered that " U P paths have as either point of departure or as goal a region which we may call the region of interactive focus — the realm of shared experience, existence, action, function, conscious interaction and awareness." Under the conceptual metaphor ACTIVE IS UP, the up-path, therefore, can denote (among other things) "the transition from a resting and motionless state to an active, operational, prototypically moving state" (ibid. :129). This explains why there is a sense of action, and possibly even control, in the phrasal verb that the simple verb lacks. There are, however, two more metaphoric links that motivate the use of the particle and preposition in this construction. They are both provided by the EVENT STRUCTURE metaphor and its entailments. Firstly, the non-literal use of up in to face up to (sth.) can be regarded as a metaphorical extension of the literal meaning of up in such constructions as the following: (9) a. climb up to the top, go up to the third floor
90
BEATE HAMPE
In these literal examples, the prepositional phrase with to denotes the goal of a movement in space, while the adverbial up specifies that the movement is along the vertical axis. The topology of this is retained in the non-literal meaning of to face up to sth under the EVENT STRUCTURE metaphor. Supplying the metaphorical frame that enables us to conceptualize difficulties etc. as physical obstacles obstructing the way to some goal, the EVENT STRUCTURE metaphor entails that the height of these obstacles should correspond directly to the degree of difficulty encountered. The higher an obstacle is, the more difficult it is to overcome. We find independent linguistic evidence for the fact that this is conventionalized in English in examples such as (9b): (9) b. high requirements/standards, the next examination loomed in the distance Related phrasal verbs such as to stand up to (a challenge), own up to (a mistake) or live up to (expectations) also support such an analysis. We may say, thus, that facing up to a problem involves a metaphorical 'upwards gaze' whose goal is the metaphorical object obstructing the metaphorical path of action.6 But there is a lot more to it. Lindner (cf. 1983:142) also observed that up can refer to reaching a goal, either in space or in some abstract domain. This, in turn, she connected to the observation that up can also describe "a path from far to near" (cf. ibid. : 143). Recasting her two observations in terms of metaphor theory, we can state that the use of up is due to the conceptual metaphor CLOSE IS UP, which captures the experiential fact that things which come closer to an observer rise in his/her visual field, as in (10a): (10) a. She came up to me. A transfer of the deictic centre makes it possible that any entity approaching another entity being the deictic centre can be seen as coming up to the latter. Up, therefore, can denote "increasing proximity" or the movement of a trajector to a landmark that is the deictic centre. The use of the preposition to indicates the deictic centre of the movement in both (10a) and (10b). (10) b. The car drove up to the entrance door.
FACING UP TO THE MEANING OF 'FACE UP TO'
91
A n example for a literal use of the notion of 'increased proximity' i n a verb-particle construction with up to is provided by to meet up, which codes 'coming together', in contrast with the simple verb to meet that merely denotes 'be together' (cf. Lindner 1983:144). We claim that the meaning o f the verb-particle construction to face up to likewise contains an element o f motion, namely o f the trajector (the grammatical subject) toward the landmark (the prepositional object), which may be a part of the conceptual base of the verb to face, but which is not an especially salient or 'profiled' dimension of its meaning. However, since the verb-particle construction to face up to is only used non-literally, no literal 'movement' can be involved in facing up to a problem. Under the metaphors K N O W I N G IS SEEING/PERCEIVING (cf. Sweetser 1991) and A C T I O N Is M O T I O N , the latter being a part of the EVENT-STRUCTURE
METAPHOR complex, actively dealing with a problem can be conceptualized as moving towards an obstacle which obstructs further progress on a path. I t is important to note that i n the metaphorical extensions the inference patterns are retained (cf. Lakoff 1990/1993): approaching an object entails seeing it clearly and i n more detail, approaching a problem entails realizing its nature and getting to know it i n more detail. Furthermore, up also provides the phrasal verb with a sense of completion: the metaphor CLOSE IS U P has been grammaticalized under the EVENT-STRUCTURE metaphor yielding C O M P L E T I O N IS UP (cf. Lakoff, Espenson
& Schwartz 1991). Therefore, any activity approaching completion can be understood as motion towards a goal which is the deictic centre and can be furnished with the aspect marker up. 4.4
Summary We have tried to demonstrate why facing up to a problem entails meeting it. The complex conceptual blend 'behind' the phrasal verb is not 'contained' by its components or by the way they are put together. The complex f o r m , i n other words, underspecifies this rich conceptual structure. The components are mere hints that provide some 'guidance' to the target structure (which is now conventionalized and need not be actively created). This guidance to a specific target conceptualization is not coded i n the simple verb, but left unspecified and needs to be supplied by contextual clues. Hence, the simple verb to face can also code a mere (static) constellation i n space where no movement o f the trajector towards the landmark is involved. I n such static situations what is perceived as landmark or trajector may mainly be a matter of perspective or of thematic focus:
92
BEATE HAMPE
(11) a. The hospital faces the old church. b. The old church faces the hospital. As we have mentioned before, the same is true for the non-literal use: (12) a. We are facing a huge problem. b. A huge problem is facing us. Obviously, this cannot be done with the phrasal verb since it is simply nonsensical to construe people as obstacles that need to be overcome by the problems moving towards them: (13) a. We are facing up to a huge problem. b. *A huge problem is facing up to us. We claim that the use of the particle up in the verb-particle construction is motivated by conceptual metaphors in multiple ways. This simultaneous motivation by more than one metaphor is a very common property of particles in verb-particle constructions. Firstly, the EVENT-STRUCTURE metaphor (location branch) and its entailments accounts for the non-literal use of to face — allowing us to conceptualize event structure in terms of physical experience while retaining the basic topology of the literal use. It also accounts for the grammaticalized use of the particle up as an aspekt/aktionsart marker. Secondly, the conceptual metaphors CLOSE IS UP, ACTIVE IS UP, CONTROL
IS UP — allow us to use the image schema up, literally coding verticality, to denote both increasing proximity of the trajector (grammatical subject) to the landmark (prepositional object) and the transition of the trajector to a state of activity and, possibly, control. Thirdly, though the target interpretation of the construction ('to actively confront an entity that poses a difficulty, danger etc.') cannot be 'predicted' by means of a simple calculus operating on the literal meanings of its elements, the construction is 'compositional' in the sense that the presence of all its elements is 'motivated' and that all elements contribute to this target interpretation: the verb by supplying the source domain for the activity, the particle by 'profiling' an image-schematic dimension of the verbal base which is left unspecified when the simple verb alone is used. Thus, it can indeed be claimed that the meaning of the verb-particle construction is more specific than that of the simple verb — the profiling of the relevant image-schema restricts possible metaphorical construals of the verb
FACING UP TO THE MEANING OF 'FACE UP TO'
face
up
to [NP]
source domains
being situtated in front of/opposite an entity
vertical movement
goal in space
conceptual metaphors
- event structure metaphor, loc. - problems are obstacles
-
- purposeful action is motion towards a goal (event structure met.)
target domains
confront sth.
activity, control, completion
target interpretation
close is up active is up control is up completion is up (event structure met.)
93
goal of activity
'to actively confront an entity that poses a problem, difficulty etc. for one's further actions'
Figure 1 : Metaphors motivating the components of the verb-particle construction 'to face up to'
content by effectively providing the relevant clues to the construal. Thus, the profiling of the image-schema up 'guides' the construal of the activity in theabstract target domain. In this way, possible ambiguities that may arise from construing the simple verb alone in a non-literal way can be avoided. Figure 1 is a representation of the complex metaphors involved in the verb particle construction to face up to. 5.
Pragmatic characteristics of the verb-particle construction 'to face up (toy) 5.1 Frequencies of the phrasal verb in different types of text and discourse In general, as was to be expected, the verb-particle construction to face up to is much more infrequent than the simple verb. We carried out a frequency count of the occurrences of both the simple and the phrasal verb using the British and American parts of the Collins Online corpus (totalling about 50 mill words).7 While there were 4,578 occurrences of the simple verb in BE ( = 127.17 per million words) and 3,723 occurrences in AE ( = 413.67
94
BEATE HAMPE
per million words), only 110 instances of the phrasal-prepositional verb were counted in BE ( = 3.05 per million words), and 51 instances in A E ( = 5.67 per million words). We put the occurrences of the simple and phrasal verb in relation to each other and found that 2.33% of all forms of to face appeared in the company of the particles up (to) in BE, and 1.35% in A E . We furthermore wanted to check whether the percentage of the phrasal verb increased significantly when it was related to the occurrences of to face in one of the three relevant non-literal senses only. Obviously, this could not be done with a corpus of the size of the Collins Online, so we used a collection of four small British corpora comprising a total of 3.5 million words (cf. note 4) and sorted all occurrences of to face manually, ending up with 297 examples (84.86 per million words) as compared to 19 instances of to face up to (5.43 per million words). We could say, in other words, that the particle and preposition up (to) accompanied the verb to face in only 6.01 % of all instances in the sense 'to confront/actively meet/recognize existence of and deal w i t h ' . The smaller corpus also had the advantage that minute details about the text sources could easily be obtained. Table 2 displays the percentage of the verb-particle construction with respect to the text types in the corpus. Concerning the distribution of to face up to in the corpus, the following details are interesting to note: (i) I n the text type literature, two of the three phrasal verbs found occurred in the genre humour, the third in a creative writing text. (ii) I n semi-official spoken discourse, one of the two phrasal verbs appeared in a speech in the House of Commons, the other in an interview. (iii) In business texts, one phrasal verb occurred in a reader's letter, all others in headlines. (iv) In the news texts, two of the three phrasal verbs detected appeared in direct speech, i.e. were no part of the commentary itself, the third example was found in a headline. (v) The percentage of the phrasal verb in headlines that contained either to face in the relevant meaning or to face up (to) was 15,38%, i.e. as high as in private spoken discourse. (vi) No phrasal verbs occurred in scientific and religious discourse. The general tendency in the corpus analysed seems to be that the phrasal verb predominantly occurs in spoken (vs written) and informal (vs formal) types of discourse. This is a confirmed property of all phrasal verbs. A t the same time, informal speech situations and text types are also those where evaluative dimensions of meaning may be of increased relevance. This is also consistent with the high percentage of phrasal verbs in headlines. Indeed, the use of informal lexical items has been linked to indirect evaluative speaker strategies (cf. Powell 1992), and the phrasal verb is certainly (stylistically) marked as an informal lexical item. But this is not the
95
FACING UP TO THE MEANING OF 'FACE UP TO'
instances of to face vs to face up to
percentage of to face up to
297 vs 19
6.01 %
spoken discourse: private conversation between intimates and equals
11 vs 2
15.38%
press: arts, journals, general articles
27 vs 3
10.00%
literature: biography, creative writing
27 vs 3
10.00%
spoken discourse: semi-official and public (speeches, interviews)
20 vs 2
9.09%
press: reviews
14 vs 1
6.66%
written: business
62 vs 4
6.06%
press: sports
17 vs 1
5.55%
100 vs 3
2.91%
7 vs 0
0.00%
type of text or discourse
average occurrence of the phrasal verb (all types of text/discourse)
press: editorial, reportage, news written: scientific, learned/religion: sermons, books
Table 2: Types of discourse with the highest amounts of the phrasal verb
only respect in which the preference of the phrasal verb over the simple one may be significant in a pragmatic perspective. We interpret the fact that a speaker uses a lexical item that is marked in two ways — both as more informal and as more specific than the simple verb — as 'indexical' of an extralinguistic dimension of the speech situation: the speaker's emotional state and/or attitude towards the state of affairs described (cf. Hübler 1987/1998). Seen from a slightly different angle, the fact that the speaker takes the effort to make explicit the particular construal intended by inserting the particle up (and thus avoids possible ambiguities) can be interpreted in the same way. I n this respect it must be noted that instead of choosing further lexical means that merely 'describe' the intended non-literal meaning of to face, such as the adverbial adjunct in face boldly, the speaker chooses the kinaesthetic/ schematic meaning of the particle up, which seems to be a lot closer to non verbal expressions such as vivid iconic gestures (cf. Lindner 1983:231, quotation), the latter being often a by-product of vivid speaker emotions.
96
BEATE HAMPE
Moreover, we may say that this 'emphasis' or 'expressivity' — though certainly present in all phrasal verbs — is especially foregrounded in phrasal verbs with an increased conceptual overlap between their components, i.e. in such phrasal verbs where the particle 'profiles' an image-schema that is already part of the conceptual base of the verb. A very similar observation was made almost 40 years ago: "Such particle verbs or phrasal verbs [i.e. hurry up, lend out, gather together, link together, meet together, open up, rise up, settle up, swallow down etc.] are felt to be more dynamic than their older synonyms, which have grown static and dull with old age and much use. For these particle verbs are composed of some simple and common verb combined with a preposition that expresses the idea of motion (up, down, off, over, away etc.) in both literal and figurative ... signification, whereby they translate into ... 'kinaesthetic' ... images ... these 'kinaesthetic metaphors' arouse the imagined sensations of muscular effort, and thereby become highly suggestive and expressive. Being the expression of muscular effort and motion, these verbs may easily be accompanied by a strong emotional feeling ... it is especially the adverb up that is filled with 'motor suggestion' . . . " [Charleston 1960:131] Thus we hypothesized that the immediate contexts8 of the phrasal verbs should be more expressive than those of the simple verbs, i.e. contain a higher amount of semantically, grammatically or stylistically marked items. 9 To this end we analysed the immediate contexts of all occurrences of both to face in the relevant sense (297 examples, 22 in headlines) and to face up to (19 examples, 4 in headlines) in the small collection of British corpora totalling 3.5 m i l l words. Since 19 instances of the phrasal verb seemed insufficient for our purposes, we additionally analysed the contexts of all occurrences of the verbparticle construction in the British part of the Collins Online (110 examples, 4 in headlines).10 We are aware of the fact that our analysis is merely suggestive in that it compiles evidence for the plausibility of our approach, rather than actually proves it. We think the latter option is not available when talking about speaker emotions and attitudes which are not directly expressed, ie which are not part of the signification itself, but occur on a user-level. Concerning the results, our expectations were largely confirmed. Though there was a large amount of variation according to text types, the average over all text types showed that the environments of the phrasal verb contained a
97
FACING UP TO THE MEANING OF 'FACE UP TO'
contextual feature
number of contextual fea tures counted (average occurrence of contextual feature per context) to face up to, 121 examples
number of contextual fea tures counted (average occurrence of contextual feature per context) to face, 215 examples
stylistic connotation: colloquial/slang
84
(0.69)
85
(0.31)
other VPCs
53
(0.44)
73
(0.26)
stylistic connotation: formal/literary/biblical
10
(0.08)
74
(0.27)
evaluative connotations
85
(0.70)
99
(0.36)
intensifying adjuncts/modifiers
98
(0.81)
190
(0.69)
metaphor/metonymy/ idioms
111
(0.92)
183
(0.66)
irony
14
(0.12)
21
(0.08)
marked word order (clefts, pseudo-clefts, fronting)
34
(0.28)
59
(0.21)
repetition/parallelism etc.
46
(0.38)
57
(0.21)
expressive use of doperiphrase/get-passive/ expanded form/perfect form
10
(0.08)
11
(0.04)
other rhetorical figures
38
(0.31)
34
(0.12)
non-finite and verbless clauses
174
(1.44)
271
(0.98)
Table 3: Marked linguistic features in the immediate textual environments of phrasal vs. simple verbs
higher amount of nearly all the features we looked for (cf. Table 3). The only exception to this finding regards the use of lexical items marked as formal or literary; here the result indicates that it is the simple rather than the phrasal verb that tends to occur in more formal environments.
98
BEATE HAMPE
We concluded that both the phrasal verb's frequency in/distribution over the text types in the corpus and the fact that the immediate linguistic environ ments of the phrasal verbs contained on average more marked items than those of the simple verb supported our claim that the use of the phrasal verb is more expressive and can index an evaluative/emotive dimension at the speaker level. The fact that informal lexical items in general have been observed to point at indirect evaluative strategies of speakers corroborates this view. 6.
Summary Firstly, the analysis at the level of the construction has made it clear that the particle is in no way void of meaning or superfluous. By providing the construction with the topology of the image schema up, it guides the construal of the non-literal meaning of to face, thus restricting possible interpretations to a particular one, namely the active and more dynamic reading. This fact and the possible closeness of particles to iconic gestures, may be the reasons why the phrasal verb appears to be an emphatic variant of the simple verb used in the same sense. Secondly, the analysis at the level of actual usage has suggested that the phrasal verb may very well index some emotional or attitudinal dimension at the speaker level — considering that the frequency of the phrasal verb corre sponds with the degree of informality of a type of text or discourse and, hence, also with the importance of emotive and evaluative functions of language within that type of discourse. We regarded the fact that the immediate lin guistic environments of the phrasal verbs contain an increased amount of expressive linguistic elements as supportive of such a view. Notes 1 This is the case, for example, with the Encyclopaedic Dictionary, edited by Prof. Muret. German-English/English-German (1900), The Concise Oxford Dictionary of Current English by Fowler & Fowler (1911-1951), the English-German/German-English Dictionary by Wildhagen (1932) and by Wildhagen & Heraucourt (1956). 2 This is true, among others, for The Oxford English Dictionary (1989), The Universal Dictionary of the English Language (1932), The Concise Oxford Dictionary (1976-1990), The Oxford Advanced Learner's Dictionary (1974-1989), The Collins Cobuild English Language Dictionary (1987), The Longman Dictionary of Current English (1987), and The Collins Concise Dictionary of the English Language (1988). 3 In a variety of other verb-particle constructions with 'superfluous' particles, the same pattern has been discovered: the simple verb is polysemous, and the corresponding phrasal verb is said to be synonymous with a very restricted number of senses only: cover up, rise up, tighten up, seek out, narrow down, cool down, sketch out, meet up (with) etc. Interestingly, these senses tend to be metaphorical, abstract or otherwise specialized.
FACING UP TO THE MEANING OF 'FACE UP TO
99
4 The corpora I had access to in 1997 included the London-Lund corpus of spoken English: 0.5 mill, the LOB Corpus of British English: 1 mill, the Microconcord Texts A, The Independent: 1 mill and parts of the British National Corpus, section K: 1 mill words. Only in 1998 could I augment some of my findings by additionally using the Collins Online corpus (parts of the Bank of English): about 50 mill words. 5 We used the 3.5 mill word corpus of British English described in note 4. 6 That such a construal additionally depends on the independently existing metaphor KNOWING Is PERCEIVING (cf Sweetser 1991) is discussed below. 7 The Collins Online (part of the Bank of English) contained about 36 million words of written and spoken British English and about 9 million words of written and spoken American English at the time we used it. 8 We chose the surrounding 50 words. 9 We looked for stylistically marked lexical items in the immediate environment, such as items marked as informal/slang vs formal/literary/biblical etc. Concerning the former, we additionally counted the occurrences of other verb-particle constructions. We also searched for lexical items with strong emotive/evaluative connotations as well as for modifiers and adjuncts with an intensifying/emphatic/hyperbolic or evaluative function. Concerning syntax, marked word orders, as those produced by fronting (topicalization), clefts and pseudoclefts, were included for their emphatic function, as were get-passives, doperiphrases and the expanded and perfect form, the latter two only i f not used with an aspectual, but an expressive function (cf. Hübler 1998). Furthermore, we searched for other syntactic features, such as non-finite and verbless clauses, which might signal an increased level of informality. Also included were rhetorical devises, such as metaphor/metonymy, repetitions, syntactic paralellisms etc. The list of features is, of course, in no way exhaustive (marked phonological forms were not included, for example). 10 We excluded from the analysis any occurrences of either of the two verbs in headlines since these lack a sufficient amount of context. References Charleston, Britta M . 1960. Studies on the emotional and affective means of expression in modern English. Bern: Francke Verlag. Fauconnier, Gilles. 1985/1994 [2nd edition]. Mental Spaces. Cambridge, Mass.: MIT Press. Fillmore, Charles J. 1988. "The Mechanisms of Construction Grammar". Proceedings Society.
of the 14th Annual
Meeting
of the Berkeley
Linguistic
General Session and Parasession on Grammaticalization
eds.
Shelley Axmaker, Annie Jaisser & Helen Singmaster, 35-55. Berkeley: Berkeley Linguistics Society. Goldberg, Adele. 1995. Constructions. A Construction Grammar Approach to Argument Structure. Chicago & London: Chicago University Press. Grady, Joseph. 1997a. "Typology of Motivation for Conceptual Metaphor". Paper at the 5th International 1997 in Amsterdam.
Cognitive Linguistics
Conference, July 7
100
BEATE HAMPE
. 1997b. Foundations of Meaning: Primary Metaphors and Primary Scenes. PhD Dissertation. Berkeley: Department of Linguistics, University of California. Hübler, Axel. 1987. "Communication and Expressivity". Functionalism in Linguistics eds. René Dirven & Villem Fried, 357-380. Amsterdam & Philadelphia: John Benjamins. . 1998. The Expressivity of Grammar. Grammatical Devices Expressing Emotion across Time. Berlin & New York: Mouton de Gruyter. Johnson, Mark. 1987. The Body in the Mind: The Bodily Basis of Meaning, Imagination and Reason. Chicago: University of Chicago Press. Lakoff, George. 1993. "The Contemporary Theory of Metaphor". Metaphor and Thought (2nd edition) ed. Andrew Ortony, 202-251. Cambridge: Cambridge University Press. . 1990. "The Invariance Hypothesis: Is Abstract Reason Based on Image Schemas?" Cognitive Linguistics 1:1.39-74. . 1987. Women, Fire and Dangerous Things. What Categories Reveal about the Mind. Chicago: Chicago University Press. & Mark Johnson. 1980. Metaphors we live by. Chicago & London: Chicago University Press. , Espenson, Jane & Alan Schwartz. 1989. "Master Metaphor List". Second Edition. MS at the Cognitive Linguistics Group, University of California at Berkeley. Langacker, Ronald W. 1988. "A View of Linguistic Semantics". Topics in Cognitive Linguistics ed. Brygida Rydzka-Ostyn, 49-90. Amsterdam & Philadelphia: John Benjamins. . 1993. "Universals of Construal". Proceedings of the 19th Annual Meeting of the Berkeley Linguistics Society, Febr. 12-15, 1993. eds. Joshua Guenter, Barbara A . Kaiser & Cheryl Z o l l , 447-463. Berkeley: Berkeley Linguistics Society. . 1995. "Viewing in cognition and grammar". Alternative Linguistics: Descriptive and Theoretical Modes ed. Philip Davis, 153-212. Amsterdam & Philadelphia: John Benjamins. Lindner, Susan J. 1983. A Lexico-Semantic Analysis of Verb-Particle Constructions with UP and OUT Trier: L . A . U . T . Series A , No 101. Powell, Mava J. 1992. "Semantic/Pragmatic Regularities in Informal Lexis". Text 12:1.19-58. Potter, Simon. 1965. "English Phrasal Verbs". Philologica Pragensia 8:23. 285-289.
FACING UP TO THE MEANING OF 'FACE UP TO'
101
Sweetser, Eve. 1991. From Etymology to Pragmatics. Metaphorical and Cultural Aspects of Semantic Structure. Cambridge: Cambridge University Press. Turner, Mark & Gilles Fauconnier. 1995. "Conceptual Integration and Formal Expression". Journal of Metaphor and Symbolic Activity 10:3.183-204.
Gerundive Nominalization From Type Specification to Grounded Instance Liesbet Heyvaert University of Leuven, Belgium 1.
Introduction Langacker (1991) claims that the semantic concepts of type specification, instantiation and grounding tend to be reflected iconically in the nominal and clausal structure. A type "specifies the basis for identifying various entities as being representatives of the same class but is not tied to any particular instance of that class" (Langacker 1991:53). An instance "is thought of as having a particular location in the domain of instantiation" (ibid.:57) and a grounding predication "presupposes that an instance has been established and gives some indication concerning its relation to the speech-act participants: instantiation per se provides no such indication" (ibid. :58). Langacker (1991:422) illustrates this claim by means of an elaborate analysis of nominalizations of the following kind: (1) Zelda's reluctant signing of the contract surprised the entire crew. (2) Zelda's reluctantly signing the contract surprised the entire crew. Traditionally, the distinction between these two constructions has been explained as a difference in degree of nominal features (see, among others, Lees 1960, Fraser 1970, Chomsky 1970, Kiparsky andKiparsky 1970, Menzel 1975, Wierzbicka 1988 and Declerck 1991). In (1), the nominalization signing behaves as a typical nominal head: it can be modified by an adjective {reluctant signing of the contract), does not tolerate an auxiliary verb {*Zelda's having signed of the contract), it can take a determiner {the signing of the contract) and it is postmodified by an ^/-prepositional phrase. This type is called 'action' nominalization (Lees 1960:64). The second nominalization, on the other hand, has preserved a clause-like internal structure: it takes an adverb rather than an adjective {reluctantly signing the contract), can have auxiliaries {Zelda 's having signed the contract), it cannot be preceded by an article or by a pronominal determiner (except for the possessive) {*the signing the contract), and the direct object follows the gerund like in a clause. Lees (1960:71) calls this 'gerundive' nominalization.
104
LIESBET HEYVAERT
Langacker attempts to offer a semantic motivation for the structures' features. He argues that structural differences exist between action, gerundive and that-nomimls (e.g. That Zelda signed the contract surprised us all), because they are derived from three different "levels of organization in the assembly of a finite clause" (1991:33), i.e. verb stem, intermediate processual structure and finite clause. These structural levels are said to correspond to a type, an ungrounded and a grounded instance respectively. The action nominal the signing of the contract is analysed as a 'type' nominalization because the nominalizing process involves the reification of a verb stem (e.g. sign) into a noun (signing), whose "semantic function (...) is limited to specifying a type" (Langacker 1991:33). If that noun is used as head of a nominal, it becomes an instance and the nominal means of instantiating and grounding apply. In the case of a gerundive nominal, however, the nominalization derives not from a verb stem (process type), but from an intermediate processual expression or ungrounded instance (e.g. sign the contract), which Langacker defines as "a structure that is like a finite clause except for the absence of an explicit subject and a predication of tense or modality" (1991:34). The result is a "complex noun" (1991:34) which functions as head noun, the schematic trajector of which is expressed by means of possessive periphrasis (Zelda's signing the contract). Finally, unlike gerundive nomináis, í/z¿zf-nominals derive from a grounded instance or a finite clause, whose internal structure is retained in the nominalization. The type/instance approach is especially worthwhile in as far as it brings out semantic parallels and differences across clausal and nominal structures: it offers the necessary descriptive tools for the analysis of the process of nominalization from clausal starting point to nominal construction. However, Langacker seems to be primarily concerned with illustrating the explanatory power of the concepts of type specification, instantiation and grounding, rather than with a thorough examination of the various kinds of nominalizations. In other words, he does not attempt to apply the conceptual framework to action and gerundive nomináis to the full, as a result of which he leaves some interesting phenomena unaccounted for. The analysis of both the internal structure and the external functioning of the nominal Zelda's signing the contract needs to be refined if one is to come at a plausible explanation for the various differences with the action nominal the signing of the contract. To start with, the status of the -ing form in gerundive nomináis should be clarified. In his discussion of action nomináis, Langacker explicitly attributes a nominalizing function to the -ing suffix, claiming that it turns a verb stem into a noun. Similarly, in gerundive nomináis, -ing is said to attach to the verb stem, only this time the verb belongs to a processual expression
GERUNDIVE NOMINALIZATION
105
(cf. supra), identified as an intermediate level of organization of a finite clause. The initial, clause-like internal organization of the processual expression is preserved in the complex noun that is formed through the addition of -ing (signing the contract). It seems to be implied in Langacker's discussion that the change from processual expression to complex noun itself is brought about by the -ing suffix (1991:32). Or, in other words, that the -ing suffix has the same, i.e. nominalizing, function as in action nomináis (Taylor's 1996 analysis of gerundive nomináis follows Langacker in this respect). Unfortunately, this is as far as Langacker takes the issue: he does not question the nature of the -ing form (signing) in the complex noun (signing the contract) any further. Nevertheless, I feel that the key to understanding gerundive nomináis lies precisely in the status of the -ing form. It is my opinion that careful consideration of the nature of the -ing suffix in gerundive nominalization is essential to elucidate the category's features and to correctly describe it in terms of the semantic concepts of type specification, instantiation and grounding. Firstly, it has to be noted that -ing in gerundive nomináis does not create a noun (like in action nomináis), but preserves the verbal character of the verb it attaches to. In other words, the -ing form in gerundive nomináis is not responsible for the reification process. Secondly, and most importantly, I claim that gerundive nomináis do not have the intermediate structural level of a finite clause as starting point (e.g. sign the contract), but rather derive from a structure that underlies a nonfinite or atemporal clausal structure (e.g. signing the contract). More specifically, I claim that the gerundive verb form which is created by the -ing suffix in gerundive nomináis should be analysed as nonfinite clausal head. In his discussion of the verb group (1991 :ch.5), Langacker argues that, parallel to the noun in the nominal, the verb group minus the grounding predication (tense or modality) can be considered as clausal head, providing a clausal type specification. But Langacker only takes into account finite clausal heads, i.e. heads that can function in a finite clause once they have been combined with a grounding predication (e.g. be playing; have opened; be killed ...). I intend to show that the grammatical system of clausal heads offers both a finite and a nonfinite option, and that this distinction is crucial for the analysis of gerundive and action nomináis. Not only does Langacker only consider finite clausal heads, he also analyzes signing the contract as an instance rather than a type, on the basis of the inclusion of the direct object. Following Davidse (in press), I claim that objects do not have an instantiating role, but rather provide a further specification of the clausal type. They can thus be compared to adjectival and nominal modifiers accompanying nominal heads.
106
LIESBET HEYVAERT
As pointed out above, Langacker claims that the -ing form in an action nominal (the signing of the contract) creates a noun type specification. I argue that in the gerundive nominal her signing the contract, -ing produces a (nonfinite) clausal type specification and does not itself reify the construction. Consequently, action and gerundive nomináis should also be discriminated in terms of the kind of nominalizing process that they result from. Two different kinds of nominalization will be distinguished, namely word nominalization (involving a verb being turned into a noun by adding -ing) and clause nomi nalization, in which a clausal structure, e.g. signing the contract, externally functions within the structure of a nominal while retaining the internal outlook of a clausal type. Exactly how a clausal structure can come to function in a nominal environment is yet another issue that needs to be raised. Langacker (1991) has discussed the similarities between a finite clause that is nominalized, on the one hand, and personal pronouns and proper names on the other hand. The question of how nonfinite clausal types behave in the nominal, however, is left unanswered. It will be shown that, even though at first sight the heads of action and gerundive nomináis seem to behave similarly, significant differences exist. Finally, the obligatory inclusion of the clausal subject in the form of a possessive will be looked at. Different kinds of instantiation and grounding are shown to be at work in action and gerundive nomináis: the noun type speci fication resulting from word nominalization (signing) is instantiated and grounded with nominal means, namely determiners. The complex process type specification that is preserved in clause nominalization, however, requires what will be classified as a mixture of nominal and clausal instantiation, namely the presence of the periphrastic subject. 2. Nonfinite clausal type specification 2.1 The semantic functions of type, instance and grounding are reflected both in the nominal and clausal structure: nouns and verbs represent types, nomináis and finite clauses profile grounded instances of these types. More specifically, Langacker suggests that in the verb group "the specification of tense and modality be analyzed as the grounding predication, with the remainder of the group (other auxiliaries and the main verb) regarded as complex clausal head analogous to a head noun" (1991:195): (the clausal heads are in italics; in (3a) the clausal head itself is inflected for tense).
GERUNDIVE NOMINALIZATION
107
(3) a. She cleaned the room. b. She may clean the room. She must have been cleaning the room. If a verb group is to function as clausal head and provide a clausal type specification, it must profile a 'process', i.e. a series of component states which are distributed continuously through time and scanned in a sequential way (ibid. :21). Even though each auxiliary element imposes its own profile on the main verb and thereby derives a higher-order type specification, it is only "the leftmost verb in the sequence" (ibid. : 196) which determines the profile of the entire verb group. Langacker argues that the function of the auxiliaries have (perfect aspect), be (progressive aspect) and be (passive voice) is pre cisely to impose a processual profile on the atemporal relation that is formed by the affixal/inflectional pair members -ed (perfect aspect), -ing (progressive aspect) and -ed (passive voice) (ibid.: 199): he claims that "the members of a given pair must co-occur if the resulting expression is to be processual and thus able to serve as clausal head" (ibid.:200). Examples of clausal heads in which auxiliary pairs function are: (4) a. perfect aspect They must have known it all the time. b. progressive aspect She will be singing in several operas this year. c. passive voice You will be notified in due course. The auxiliary pairs can also occur in combination (examples drawn from the Collins COBUILD Corpus): (5) a. perfect and progressive They have argued that the Inkatha leader has been deliberately orchestrating the recent violence. b. progressive and passive Samples from 90 horses are being tested at the Australian Animal Health Laboratory, in Geelong. c. perfect and passive A computer program which diagnoses schizophrenia has been developed by a University of Queensland research team.
108
LIESBET HEYVAERT
(5) d. perfect, progressive and passive And will the Queensland Government agree? Well, the papers have been being put into the Cabinet in-basket and taken out of the Cabinet in-basket since some time in March. 2.2 However, verb groups with a processual profile cannot only be grounded in the speech event. What Langacker does not consider in his discussion of clausal heads is that processual verb groups or clausal heads can also be atemporalized. In what follows I will first present structural evidence in favour of extending the concept of clausal head to include both finite and nonfinite heads: more specifically, nonfinite clausal heads or processual structures which have been given an atemporal profile provide a complex type specification which cannot in any way be grounded in the speech event. In a second part, the structural similarities with gerundive nominalizations will be pointed out and illustrated, and it will be concluded that the -ing suffix in gerundive nomináis must be analyzed as being part of a nonfinite clausal head, rather than having a reifying function. How a clausal type specification can come to function in a nominalization will be dealt with later on in the discussion. To start with, one of Langacker's arguments in favour of the existence of a fundamental division between tense/modality on the one hand and the remainder of the verb group on the other hand is the following: an infinitive (with to) can be formed on a main verb together with its auxiliaries but without markers of tense or modality: thus, constructions like to wash, to be washing, to have washed, to have been washed are possible, while Ho washed, Ho should wash, Ho had been washed are unacceptable. However, it seems to me that this does not only illustrate that it is the non-grounding auxiliaries (like be and have), rather than the grounding expressions of tense and modality, that form a strong unit with the main verb. These structures also show that the combination of main verb and auxiliaries which is termed 'clausal head' by Langacker cannot only be turned into a finite expression, but that it can just as well be elaborated into a to-infinitive, while using an equally wide range of auxiliaries and auxiliary combinations as with a grounded clausal head. Moreover, note also that, besides a to-infinitive, the -ing suffix can also function as an alternative for grounding the clausal head, witness the following examples: (6) washes/washed has washed has been washing
washing/ to wash having washed/ to have washed having been washing/ to have been washing
GERUNDIVE NOMINALIZATION
109
Grounding the non-finite verb group which is thus formed is impossible: *shoulding wash, *hadding been washed. Only when the clausal head is progressive in profile, is -ing impossible: e.g. be playing, but *being playing. In his discussion of the -ing suffix, Langacker attributes this to the fact that "simpler alternatives are available that do essentially the same job" (1991:232). In other words, playing is a simpler alternative for being playing and therefore makes the latter structure redundant. What do the finite and nonfinite way of elaborating clausal heads have in common and in which way do they differ? Firstly, the examples make clear that both the atemporal markers (-ing and to-infinitive) and elements of tense/modality need a processualprofile as basis: atemporal verb phrases like playing and played have to be complemented with the auxiliaries be or have, which profile a process, before they can be grounded or atemporalized. Moreover, grounding or atemporalizing a clausal head prove to be mutually exclusive choices. This means that, once a head has been grounded, it can no longer be atemporalized, neither can an atemporalized head be grounded. In short, a processual structure or clausal head is either grounded or turned into a structure which is nonfinite and cannot, in any way, be grounded: (7) a. *to was sitting on that bench [Langacker 1991:421] *shoulding have sat on that bench [ibid.] b. *should being killed *will to play in the garden Of course, if the atemporal -ing suffix is elaborated into a structure with be, a new processual structure or clausal head is formed and grounding be comes possible again: being dropped -*• be being dropped -> is being dropped. In addition, like finite clausal heads, nonfinite heads can be made to function in a clausal structure with a subject, direct object, complement, adjuncts and adverbials (examples drawn from the Collins COBUILD Corpus): (8) a. Those who ate the apples ran off wild and mad into the forest the apples having been poisoned by a woman who had loved Merlin (...). b. (...) there has been another explosion at an office of the ruling National Party, with evidence pointing to it having been the work of white extremists.
LIESBET HEYVAERT
(8) . Falling water levels at Somerset Dam have led to restrictions being placed on boating in the area. d. There his father (...) lived the life of a recluse, conscientiously tending the woods he loved. e. Last Friday, he sent a telegram to the annual meeting of Quebec's independence party, openly endorsing their separatist demands. If the atemporal markers -ing and to-infinitive do not ground the type specification provided by the clausal head, then what exactly is their semantic impact? I follow Langacker when he argues that atemporal markers primarily affect the conceptualizer's mode of scanning the component states of the process: rather than the step-by-step, serial way of scanning typical of processual structures (e.g. play, have played, be playing) and finite clauses, nonfinite clausal heads offer a holistic view on the component states. This means that the component states are viewed "as a single gestalt" (1991:21). In other words, turning a clausal head into a finite structure implies grounding the type specification which it provides, without changing the mode of scanning. In contrast, when a clausal head is atemporalized, it still profiles a clausal type, but the way in which the component states of the verb type are scanned changes from processual to holistic. Characteristic of -ing is that it moreover restricts the profile to "a series of component states that does not include the initial and final states" (ibid.:209). Because the basic semantic effect of -ing and to-infinitive is that the component states of the process are scanned holistically, it does not matter whether the component states of the process to which they are added are identical or not. In other words, the atemporal markers -ing and to-infinitive can be used with both 'imperfective' and 'perfective' verbs: imperfective verbs profile processes with identical component states (e.g. know -*■ to know /knowing), perfective verbs profile some kind of change due to the differences that exist among the component states (e.g. hit -> to hit/hitting) (Langacker 1991:21). However, when turning the atemporal -ing profile into a processual head with be, sequential scanning of the component states is reestablished. More specifically, the distinctive semantic feature of the progressive pair which is thus created is what Langacker calls its 'imperfectivizing' effect, i.e. "the component states are construed at a level of schematicity that neutralizes their differences" (ibid.:209). Therefore, only perfective verbs (with different component states) can be used in a progressive construction: imperfectivizing an already imperfective verb would not make sense (ibid.:208).
GERUNDIVE NOMINALIZATION
111
To sum up, I hope to have shown that there are strong structural indications that the atemporal markers -ing and to-infinitive can give rise to a second type of clausal head, next to grounded clausal heads. Even though both types of clausal head display similarities, there are of course also differences, especially as far as the semantic function which they fulfill is concerned: a finite clausal head grounds the clause in the speech event, whereas a nonfinite clausal head merely creates a different, i.e. atemporal or holistic, way of looking at verbal types. Now, if we consider gerundive nomináis, a striking resemblance with atemporal clausal heads can be witnessed, both structurally and semantically. More particularly, gerundive nomináis can also take a range of auxiliaries and, secondly, other constituents typical of clausal structures (such as adverbials, complements, direct objects and adjuncts) can be included: (9) a. perfect aspect One of those individuals who is unaccountably gifted as a contemplative, she knew nonetheless that her discovery of that hidden track inward, and her having been able to follow it into full, ecstatic awareness of God, was sheer grace. It has been observed that the emphasis on Christ's humanity — his having rendered himself up without resistance to the authorities and the fact that his body was broken and bleeding — allowed women in particular to identify with him at a very deep level. b. passive voice But his outliving her made him a rare specimen in this setting. Also her being awed by their vast drawing-room and the grand piano that nobody ever played. Vice President Quayle announces that his being dropped from the ticket is a closed issue, as though the decision is his to make. perfect aspect and passive voice Its volume alone would be impressive, but its originality and sheer brilliance vindicate completely her having been declared Doctor of the Church. d. perfect aspect and progressive voice Their having been camping in the nature reserve got them a fine, [personal example]
112
LIESBET HEYVAERT
What's more, the other atemporal marker, namely to-infinitive, can also function as part of a nominalization: e.g. For her to have been able to follow it into filli, ecstatic awareness of God, was sheer grace; (...) For them to have been camping in the nature reserve got them a fine (for a further description of this type of nominalization, see Lees 1960). Together with nominals of the kind That Zelda signed the contract surprised the crew (see above), in which the clausal head is grounded, these to-infinitive nominalizations complete the picture of clausal heads which can function in a nominal environment. Finally, similar to nonfinite clausal heads, gerundive nominals can draw on both perfective (see examples) and imperfective process types: e.g. John('s) knowing the answer surprised us all; Lucy('s) being late worried her father. I n short, I claim that gerundive nominals systematically draw on nonfinite clausal heads, or heads which cannot be grounded in the speech event, which present the action holistically and from "an internal perspective" (Langacker 1991:209), ignoring its initial and final states. The possibility of using such an 'ungroundable' clausal type may well motivate the construction and partly explain its semantic import (for instance with regard to its 'factive' status, see below). Langacker's 'ungrounded instance' analysis of gerundive nominals can now be refined. I n the first place, it has been shown that the structure from which gerundive nominalization starts already contains the -ing suffix, which provides an atemporal profile. Being a nonfinite clausal head, the initial structure is not only ungrounded, but, more particularly, it can never be grounded. Secondly, the nonfinite type specification constituted by the head signing is not instantiated by the direct object (cf. signing the contract), but rather described in more detail by it. I n other words, gerundive nominais derive not from an ungrounded instance, but from a complex type specification that cannot be grounded. The latter claim concerning the nature of the direct object, however, w i l l be further elaborated upon in the next section. 2.3 Nominal heads can take nominal, adjectival and prepositional modifiers subcategorizing the 'thing' designated by the head noun without instantiating it: Langacker describes the construction excellent convention site in the Midwest (1991:53) as a type specification which is "rendered progressively more specific". He argues that a noun type can be characterized with any desired degree of precision, thanks to adjectival and other modifiers that are not determining (ibid. :54). Similarly, verb stems can be 'modified' without being instantiated: Langacker considers auxiliaries as part of clausal types. Following Davidse (in press), I claim that a clausal type specification does not
113
GERUNDIVE NOMINALIZATION
only consist of the main verb and its nonfinite auxiliaries, but that also objects should be included: in other words, I disagree with Langacker when he states about the gerundive nominal Zelda's signing the contract that "The fact that the nominalized structure incorporates a fully specified direct object supports the claim that it represents an instance and not just a type" (1991:34). I view signing the contract as a whole as providing a complex process type specifi cation. Instead of fulfilling an instantiating function, the direct object is thus claimed to merely 'subcategorize' the process type expressed by the verb stem "into more specific process types" (Davidse 1997:422). Davidse (in press) gives the following arguments for a type-specifying role of the object and an instantiating function of the subject: her claims are based on the observation that a fundamental structural split exists between the subject and the finite verb (and polarity) on the one hand (called ' M o o d ' by Halliday 1994), and the rest of the clause on the other hand (which Halliday calls 'Residue'): Mood element Subject
Finite
Residue Polarity
Predicator IO DO
GROUNDING
TYPE
INSTANTIATION
SPECIFICATION
I
am
You
're
Complement
writing you a letter not
being
a pest (are you ?)
Formal evidence for this grammatical split is found in tags (which isolate the Mood element) and in elliptical clauses (Mood ellipsis: e.g. What were they doing? Holding hands/ Being chased by a bull; and Residue ellipsis: e.g. I am writing you a letter? Oh, are you? Yes, I am). Other formal properties of objects that are in line with their typespecifying role can be found in Keenan (1976, as cited in Davidse, in press: 161): he points out that objects are typically lowly referential (relatively indefinite), while subjects are usually highly referential (relatively definite). Moreover, Keenan claims that one of the most universal properties of the subject is that its reference "must be determinable by the addressee at the moment of utterance. It cannot be made to depend on the reference of other NPs which follow i t " , e.g. *He-self admires John cannot replace John admires himself (Keenan 1976:313). The reference of the object, however, can be made dependent on that of the subject, and is generally controlled by the subject.
114
LIESBET HEYVAERT
That clausal type specifications are formed by the predicator and its objects is also reflected in the general grammatico-semantic categorization of process types/ clausal heads, which is based on whether or not an object is included: intransitive (incl. passive of the transitive): no object (e.g. sleeping; being hit) transitive (incl. passive of the ditransitive): object (e.g. hitting so.;being given sth.) ditransitive: indirect and direct object (e.g. giving so.sth.) Why does Davidse prefer to call the subject rather than both subject and object instantiating? Langacker (1991) defines 'grounding' as relating an instance to the ground (i.e. the speech event and its participants). Grounding presupposes that a type specification has been tied to a particular instance, and it involves pointing out the relation between the instance and the ground, thus rendering the instance identifiable to the hearer. For the clause, this means that it is finite. The finite verb establishes 'subjective grounding' by indicating the expression's temporal and modal relation to the speech participants (the ground or reference point is maximally subjective within the scope of predication, i.e. it remains implicit, cf. Langacker 1991:93-94). Davidse (in press) claims that, in addition to being grounded by the finite element, a clause is also grounded by the person deixis which the subject expresses: the finite verb does not only indicate tense or modality, it also implies (implicit hence 'subjective') PERSON deixis (grammatical person), which is yet another way of relating the instance to the speech event. The subject can be said to make explicit this grammatical person of the finite verb, which is expressed on the finite element by person and number marking, and on the subject by the nominative case (note that no information about objects is expressed on the English finite verb). Therefore, the subject can be said to constitute an objective form of grounding, via what could be called 'objective' person deixis. Davidse moreover argues that 'finiteness' is also crucially instantiation of a process type: 'infinitives' designate types (be, break . . . ) , whereas finite verbs designate instances of types (was, broke, has broken . . . ) : "the process type specification is instantiated in the first place (...) by the Finite [verb]" (in press: 161). The subject is claimed to make explicit which instance the finite verb designates: it explicitates the person implied in the finite verb, thus instantiating the process type, and opening the way for relating the instance to the ground or speech event.
GERUNDIVE NOMINALIZATION
115
To conclude, I claim that the starting point or the structure which the reification applies to in gerundive nominais is a nonfinite clausal head, which, together with its object, provides a complex process type specification (signing the contract). So far, Langacker's analysis of gerundive nominais has thus been modified with regard to two points: -ing is claimed to be part of the initial, nonfinite clausal head structure and is therefore not itself nominalizing (as it is in action nominalizations). And secondly, the direct object is not instantiating, but rather belongs to the original type specification. 3. Different kinds of nominalization 3.1. It has been shown that action and gerundive nominais differ with regard to the structure that is nominalized: in nominalizations like the signing of the contract (action), a simple process type specification (sign) is reified. Gerundive nominais, on the other hand, derive from a complex process type specification (e.g. signing the contract). Because of these different starting points, distinct kinds of nominalization apply: action nominalization turns a simple process type into a noun type by adding -ing. Because the nomina lization is carried out at word level (from verb to noun), I suggest to call it word nominalization. Apart from -ing ('action' noun), other suffixes can be used, e.g. -er ('agentive' noun, e.g. signer). The resulting noun can be used as head of a nominal, cf. The signing of the contract took an hour; The signer of the contract had used the wrong pencil. Gerundive nominalization, on the other hand, retains the internal structure of a complex process type, but makes it function within a nominal structure, itself fulfilling the role of a main-clause participant (e.g. Zelda's signing the contract shocked us all). I n other words, the nonfinite clausal head comes to function as nominal head externally (and nominal means of instantiation and grounding the head apply), but internally it retains its status as clausal head. Because the internal clausal structure is left intact, I prefer to call this clause nominalization. 'Factive' nominais (cf. Lees 1960; e.g. That Zelda signed the contract) also fall under this type of nominalization. 3.2. I n how far can action and gerundive nominais be considered as prototypical nominal structures? A prototypical nominal reflects its semantic function directly, i.e. it incorporates both a head noun and a determiner, each fulfilling a specific semantic function: the head noun specifies a type, and the determiner grounds an instance of that type (Langacker 1991:143). Nominal structures which deviate from this canonical organization arguably only differ from prototypical nominais in that they lack (visible) structural implementation
116
LIESBET HEYVAERT
of the semantic functions. Langacker (1991:148-149) describes the nominalization of a finite clause (e.g. That Zelda signed the contract surprised the crew) as an illustration of a non-prototypical nominal: the nominalization itself makes up the nominal, and determiners cannot be added. This is attributed to the fact that the clause, functioning as head noun, designates a grounded instance: "The specified type therefore has only a single instance, with the consequence that the derived noun is inherently definite" (ibid. :149). As a result, no separate grounding predication is required: "conceptual reification alone converts a finite clause into a nominal" (ibid. :149). Finite clauses that are nominalized are thus comparable to proper names and personal pronouns: due to the fact that they designate single instances of a type and that they are inherently definite, their being grounded does not need separate structural symbolization. Gerundive nominais, containing a nonfinite clausal head, differ from finite nominalizations in that they do take a premodifying structure, namely the possessive (e.g. Zelda's / Her signing the contract). I n action nominais, the nominal head is prototypical in that it can be pre- and postmodified (e.g. the signing of the contract). As a result, one is tempted to classify action and ge rundive nominais as typical nominal constructions. However, when comparing the way in which the heads of action and gerundive nominais function within the nominal structure, some interesting differences show up, suggesting that in gerundive nominais, the internal clausal structure of the head crucially influences the way in which it can function in a nominal structure. I n what follows, I w i l l show that action heads are fully nominalized in that they take the nominal means of instantiation and grounding, whereas gerundive heads take a mixture of nominal and clausal instantiation and can therefore be said to be only partly nominalized. 4. Means of instantiation and grounding 4.1 The noun type signing (as in Zelda's [signing] of the contract) can only function as head of a nominal after it has been instantiated and grounded. Grounding implies that the type specification is tied to a specific instance and that the speaker makes this instance identifiable to the hearer. I n the case of an action nominal, the following grounding options are available: definite article 1 : e.g. The [signing] of the contract took place in this room. demonstrative pronoun: e.g. We must try to stop this massive [killing] of innocent people. possessive: e.g. Zelda's/Her [signing] of the contract shocked us all.
GERUNDIVE NOMINALIZATION
117
What strikes the eye is, firstly, that the various ways of grounding the head are all distinctly nominal, and, secondly, that it is only the addition of determiners that enables the head to take up its function in a nominal: adjectives and postmodifying structures (like of the contract) only render the noun type more specific, more complex, without instantiating or grounding it. Moreover, it has to be pointed out that the Agent of the verb forming the basis of the action nominal is not necessarily expressed (cf. the signing of the contract), and when it is expressed (periphrastically of course), it is not necessarily located in front of the head, functioning as possessive grounding element. It can take the form of an of-prepositional phrase (namely when the verb which is nominalized is intransitive): e.g. The cooing of the pigeons; The coming of the Prince of Wales (Declerck 1991:499). When, on the other hand, the verb does take a direct object (i.e. it is transitive), and this direct object is periphrastically expressed in the form of an -of prepositional phrase, the Agent is indicated by means of a by -prepositional phrase, e.g. The climbing of Mont Blanc by a solitary girl is an incredible performance (Declerck 1991:499). 4.2. When functioning as head of a nominal, the complex process type signing the contract has only one grounding option available: possessive: Zelda's/His signing the contract surprised the crew. article: *the signing the contract demonstrative: *that killing innocent people A t first sight, one is tempted to conclude that, the possessive being a nominal means of grounding, the gerundive nominal functions as a 'normal' nominal head. However, things are more complicated. Firstly, the subject of the nonfinite clausal head (which may be conflated with Agent, e.g. his having rendered himself up, or Patient, e.g. his being dropped from the ticket) must be expressed: not only is it impossible to use other nominal means of construing the nonfinite clausal grounding (*the/that signing the contract...), head immediately as nominal entails a shift in meaning. Lees (1960:72) speaks of a semantic change leading from a 'fact' to an 'action', but he does not analyze the 'action' variant of gerundive nominais any further (see also ibid. 1960:106): (10) a. *Signing the contract surprised the entire crew. b. Signing the contract was a stupid thing to do.
118
LIESBET HEYVAERT
Secondly, in informal language, the periphrastic form of the subject is often replaced by a bare noun: John writing us a letter is rather unusual Lees (1960:72) claims that an interesting feature of the gerundive nominal is that in "colloquial English" and " i f it is very long, it usually takes no nominal affixes", e.g. I don't approve of the crew's only remaining man going there. The least one can conclude from this is that the structural position filled by the possessive differs in action and gerundive nominais: in action nominais the possessive is only one of a series of nominal grounding options; i n gerundives the expression of the subject is obligatory; i n gerundives the possessive can be replaced by a bare noun, i.e. a clausal form of expressing the subject 4.3. Starting from the observation that it is always the subject of the nonfinite clausal head that functions as premodifier in gerundive nominais, I claim that the possessive form of the subject instantiates and grounds the clausal head in its (external) function as head of the nominal. A t the same time, however, the subject which is periphrastically expressed by the possessive is needed to instantiate the (internal) clausal structure signing the contract, thus allowing it to figure i n 'factive' contexts (see below). I n short, in view of the fundamental difference between direct object and subject which I discussed earlier, and considering the close relation between subject and finite as far as instantiation and grounding are concerned, I would like to propose an alternative analysis for the gerundive nominal Zelda's [signing the contract]: instead of calling [signing the contract] an instance whose subject is periphrastically expressed, I prefer to call it a type which is necessarily instantiated by the subject. As was argued before, the subject realizes instantiation of the clause together with the finite verb: only the addition of a finite verb w i l l fully instantiate the clause, i.e. indicate the location of the clausal type specification in terms of tense or modality. But the presence of a subject can be seen as a first step towards instantiation. 5.
Factive Nominais? Langacker's discussion of -ing nominalizations in terms of type/instance also attempts to account for a semantic difference which has often been observed between action and gerundive nominais: action nominais seem to "focus on the event as a physical activity" (e.g. Sam's washing of the windows was meticulous, Langacker 1991:32), whereas gerundive nominais seem to imply that the designated event has occurred, that it is a 'fact' (e.g. Sam's
GERUNDIVE NOMINALIZATION
119
washing the windows was a shock to everybody, Langacker 1991:32). Langacker argues that it is the instantiated nature of gerundive nominais which allows them to figure in factive contexts. Due to the ungrounded nature of the reified process, however, this need not always be the case (e.g. I would definitely object to your taunting the bear (should you ever decide to do it), ibid. :32). For various reasons I w i l l not elaborate on the topic of 'factivity': comparing the many different definitions (cf. Lees 1960; Kiparsky and Kiparsky 1971; Menzel 1975; Davidse 1991; Halliday 1994) of the concept alone would require another article. But, more importantly, 'factivity' is not an exclusive property of gerundive nominais: as Lees (1960) points out, 'factive nominais' 2 (e.g. That he came was obvious, ibid.:59) and 'for ... to' nominais (e.g. For him to eat vegetables is surprising, ibid. :71) also refer to 'facts'. A n account of factivity should therefore incorporate a detailed analysis of these types of nominalizations as well. Due to time and space limitations I have had to restrict myself to remarks about Langacker's type/instance analysis itself. However, my claim concerning the instantiating function of the periphrastically worded subject in gerundive nominais certainly offers interesting perspectives on the analysis of factivity: i f Langacker's hypothesis that a factive meaning is linked with instantiation is correct, and i f the subject is, next to the finite verb, an instantiator, then it should logically follow that subject-less, nonfinite nominalizations cannot have a 'factive' meaning. Interestingly enough, observations in that direction have already been made by Lees (1960), who distinguishes gerundive and 'for ... to' nominais with a factive meaning from those with an 'action' interpretation; the latter type "cannot have any expressed subject" (ibid.:72; e.g. It's great fun swimming there; It's great fun to swim there). 6.
Conclusion This article has been an attempt to refine Langacker's type/instance analysis of nominalizations. His account has the merit of offering a whole new perspective on phenomena whose structural properties had, despite extensive discussion in the literature, as yet not been motivated satisfactorily. However, I claim that Langacker does not use his conceptual framework to the f u l l : as far as gerundive nominais are concerned, I suggest an analysis which deviates from Langacker's interpretation in three ways: firstly, I want to stress the importance of describing the -ing form in detail. Gerundive nominais turn out to contain a nonfinite clausal head, which forms a complex clausal type specification with its object. Secondly, two kinds of reification processes can be distinguished, depending on the internal outlook of the nominalization.
120
LIESBET HEYVAERT
Finally, the complex clausal type specification in gerundive nominais is argued to be necessarily instantiated by the periphrastically worded subject. As already hinted at in the previous paragraph, several other types of nominais and a range of semantic peculiarities should be taken into account as well if one is to come at an understanding of the concept of nominalization and its various 'instantiations'. I hope to have shown that such a large-scale analysis of -ing nominalization will necessarily have to be 'grounded' in a detailed description of the head, of the paradigmatic variants or options that are available to the language user to fill the modifier positions, and of the syntagmatic or contextual potential of each subtype. Notes 1 Though less frequently occurring, the indefinite article, the zero determiner, and relative quantifiers like no, some, any ... can also be found in Action nominais: e.g. (...) removal of references to the monarch is an act of sedition and a breaking of the oath of office taken by politicians; (...) the Ministry of Agriculture will allow burning of herbage seeds, reeds, lavender (...); There was no breaking of ranks with Cabinet colleagues; Certainly there seemed to be no burning of bridges for these two; ... [Source: Collins COBUILD Corpus] 2 Langacker (1991) uses the term 'factive' as a synonym for 'gerundive' nominais. This does not correspond to Lees' terminology: he argues that gerundive nominalization (i.e. of the kind Zelda's signing the contract) "refers to a fact" (Lees, 1960:65), but he reserves the term 'factive' nominalization for finite nominalizations (e.g. that Zelda signed the contract). Chomsky (1970) follows Lees in his choice for the term 'gerundive'. It is only Fraser (1970) who uses the term 'factive' for 'gerundive' nominais. Considering Langacker's explanation for the factive meaning which gerundive nominais often have (i.e. their being instances), and his analysis of the fact that they can also have a non-factive meaning (because they are ungrounded instances), it seems best to follow Lees' terminology and reserve the term 'factive' only for that type of nominal which, being a grounded instance, is always factive in meaning.
References Chomsky, N. 1970. "Remarks on Nominalization". Readings in English Transformational Grammar eds. R.A. Jacobs & P.S. Rosenbaum, 184221. Waltham, Mass.: Ginn. Davidse, K. 1991. Categories of Experiential Grammar. PhD Dissertation, Dept. of Linguistics, University of Leuven. Davidse, K. 1997. "The Subject-Object versus the Agent-Patient asymmetry". Leuven Contributions in Linguistics and Philology 86.413-431. Davidse, K. in press. "The Dative as participant role versus the Indirect Object: on the need to distinguish two layers of organization". The Dative: Theoretical and Contrastive Studies. eds. W. Van Langendonck & W. Van Belle, 143-184. Amsterdam & Philadelphia: John Benjamins.
GERUNDIVE NOMINALIZATION
121
Declerck, R. 1991. A Comprehensive Descriptive Grammar of English. Tokyo: Kaitakusha Co. Fraser, B. 1970. "Some Remarks on the Action Nominalization in English". Readings in English Transformational Grammareds.R.A. Jacobs & P.S. Rosenbaum, 184-221. Waltham, Mass.: Ginn. Halliday, Μ.Α.K. 1994. An Introduction to Functional Grammar. London, Melbourne & Auckland: Edward Arnold. Kiparsky, P. and C. Kiparsky 1971. "Fact". Semantics: An Interdisciplinary Reader in Philosophy, Linguistics, and Psychology eds. D.D. Steinberg & L.A. Jakobovits, 345-369, Cambridge: Cambridge University Press. angacker,R. 1991. Foundations of Cognitive Grammar 2: Descriptive application. Stanford: Stanford University Press. Lees, R.B. 1960. The Grammar of English Nominalizations. Bloomington: Indiana University Research Center in Anthropology, Folklore, and Linguistics, Publication 12. Menzel, P. 1975. Semantics and Syntax in Complementation. The Hague, Paris: Mouton. Taylor, J. 1996. Possessives in English. An Exploration in Cognitive Grammar. Oxford: Clarendon Press. Wierzbicka, A. 1988. The Semantics of Grammar. Amsterdam & Philadelphia: John Benjamins.
A Cognitive Approach to Errors in Case Marking in Japanese Agrammatism The Priority of the Goal -ni over the Source -kara Hiroko Ihara Tokyo Zokei University Tokyo, Japan
Ikuyo Fujita International University of Health and Welfare Tochigi, Japan
1.
Introduction * Agrammatism is a syndrome that occurs in aphasia caused by brain damage. It is generally admitted that one of its characteristics is the omission of, or errors in, function words (Caplan & Futter 1986; Caramazza & Berndt 1985). In Japanese agrammatism, case particles are very often omitted (Sasanuma et al. 1990) or incorrect case particles are sometimes substituted for them (Kamio 1979) in spontaneous speech. It is also ob served that incorrect case particles are substituted for correct ones in many cases in elicited speech (Fujita 1991; Fujita &Ihr1992,1993; Ihara & Fujita 1995). The aim of this paper is to show that the priority of the dative -ni representing the goal role over the ablative -kara representing the source role in the sentences produced by Japanese agrammatic patients is in accord with the goal-orientedness observed pervasively in human language. It also aims to show that the priority of the dative -ni can be explained using the notion of action chain proposed in Langacker (1991). Section 2 gives an outline of Japanese case particles and section 3 gives a brief outline of the characteristics of agrammatism. Section 4 gives details of the experiment and shows the results. Section 5 argues that the results of the experiment indicate goal-orientedness pervasively observable in human language, and attempts an explanation using the notion of action chain proposed in Langacker (1991). The final section gives concluding remarks. 2.
Outline of case particles in Japanese Japanese is an SOV language and grammatical and semantic relations between a predicate and its participants are in principle realized as case particles placed after nouns.1 Although a sentence typically begins with the subject, word order is relatively free except for the position of verbs, which come at the end of the sentence, as the following examples show (NOM = NOMINATIVE, ACC= ACCUSATIVE):
124
HIROKO IHARA & IKUYO FUJITA
(1) a. Keiko-ga Taroo-o but-ta Keiko-NOM Taro-ACC hit-Past "Keiko hit Taro." b. Taroo-o Keiko-ga but-ta Taro-ACC Keiko-NOM hit-Past "Keiko hit Taro." c. ??But-ta Taroo-o Keiko-ga2 hit-Past Taro-ACC Keiko-NOM There are a number of case particles. In principle the nominative -ga and the accusative -o indicate that the accompanying nouns are the subject and the object of the sentence respectively.3 The other case particles represent semantic roles. As far as transfer-of-possession verbs are concerned, the dative -ni in principle represents the goal role in GIVE-type verbs and the ablative -kara represents the source role in RECEIVE-type verbs, as the following examples show (DAT=DATIVE, ABL=ABLATIVE): (2) a. Taroo-ga Keiko-ni hana-o age-ta Taro-NOM Keiko-DAT flower-ACC give-Past "Taro gave Keiko flowers." b. Keiko-ga Taroo-kara -o uketot-ta Keiko-NOM Taro-ABL money-ACC receive-Past "Keiko received money from Taro." However, as will be shown in section 5.1, the dative -ni can be substi tuted for the ablative -kara with a limited number of RECEIVE-type verbs. 3.
Characteristics of agrammatism Before giving details of the experiment, we will take a brief look at the characteristics of agrammatism. In general, aphasia is a language disor der caused by brain damage after a language has been acquired, and agram matism is one kind of aphasic syndrome. It very often overlaps with one traditionally classified type of aphasia known as Broca's, "resulting from a lesion of the anterior part of the language zone of the dominant (in most cases) left hemisphere" (Tonkonogy 1986:53).4 The characteristics of agrammatism are speaking effortfully with phonemic distortion and producing sentences with minimal syntactic structure (Schwartz et al. 1987). It is also generally agreed that one of the characteristics is the omission or
ERRORS IN CASE MARKING IN JAPANESE AGRAMMATISM
misselection of grammatical markers and function words (Caramazza & Berndt 1985; Linebarger 1990). In the case of Japanese agrammatism, it is reported that in spontaneous speech case particles are frequently omitted and are sometimes substituted (Kamio 1979; Sasanuma et al. 1990), and in elicited speech case particles are very often substituted (Fujita 1991; Fujita & Ihara 1992, 1993; Ihara & Fujita 1995). 4. Experiment and results 4.1 Experiment Purpose. The aim of the experiment was to see whether there is any difference between the rates of errors made in supplying the dative -ni and the ablative -kara by agrammatic aphasics as far as transfer-of-possession verbs are concerned, and how the errors are made. Materials. A picture-description task was used. Twenty pictures were prepared and the names of the objects and people in the pictures were written beside them. Each picture was to be described with a verb that was provided. Two GIVE-type verbs, i.e. watasu "hand" and ageru "give", and two RECEIVE-type verbs, i.e. uketoru "receive" and azukaru "be entrusted with" were used. Both types of verbs take -ga for a subject, and GIVE-type verbs take -ni for the goal role and RECEIVE-type verbs take -kara for the source role. Each verb was used five times with different pictures. Procedure. The subjects were asked to describe each picture orally in a complete sentence, using a given verb. Each subject was shown the same 20 pictures, and was given one picture and one verb at a time. The verb was given to them orally as well as visually. The pictures were arranged in such a way that the same type of verbs were not in succession. Each subject was given a preliminary test to confirm his/her understanding of the task. Subjects. Three Broca's aphasics (T.S./I../F.T.), and two normal controls served as the subjects of the study. All the subjects were native speakers of Japanese. The aphasie patients were all right-handed, and had lesions in the left hemisphere. They produced agrammatic speech and had difficulty in producing function words. They often made errors in using case particles. 4.2 Results The rates of -ni representing the goal role and -kara representing the source role correctly supplied and those incorrectly supplied are shown in Table 1. It shows that the rate of the dative -ni correctly supplied is significantly higher than that of the ablative -kara correctly supplied (χ2 = 52.325,
126
HIROKO IHARA & IKUYO FUJITA
-ni
-kara
subject
correct
error
others
correct
error
others
T.S.
10/10 (100%)
0/10 (0%)
0/10 (0%)
0/10 (0%)
10/10 (100%)
0/10 (0%)
I.K.
10/10 (100%)
0/10 (0%)
0/10 (0%)
1/10 (10%
8/10 (80%)
1/10 (10%
F.T.
10/10 (100%)
0/10 (0%)
0/10 (0%)
0/10 (0%)
10/10 (100%)
0/10 (0%)
N1
10/10 (100%)
0/10 (0%)
0/10 (0%)
10/10 (100%)
0/10 (0%)
0/10 (0%)
N2
10/10 (100%)
0/10 (0%)
0/10 (0%)
10/10 (100%)
0/10 (0%)
0/10 (0%)
Table 1. Rates of -ni and -kara correctly or incorrectly supplied
p<0.001). When the ablative -kara was not assigned correctly, in most cases the nominative -ga or dative -ni was erroneously substituted for it, as Table 2 shows. The following are some examples in which -ni and -ga were substi tuted for -kara. Incorrect case particles are non-italic and particles in paren theses are the expected forms. (3) a. *Otousan-ga okaasan-ni (-kara) okane-o uketoru father-NOM mother-DAT (-ABL) money-ACC receive "Father receives money from mother." [T.S.] b. *Kakari no hito-ga onnanohito-ni (-kara) the person in charge-NOM woman-DAT (-ABL) nimotsu-o azukaru luggage-ACC be entrusted with "The person in charge is entrusted with the luggage by the woman." [I.K.] *Okaasan-ga (-kara) onnanoko-ni (-ga) sofutokuriimu-o mother-NOM (-ABL) girl-DAT (-NOM) ice cream-ACC uketoru receive "The girl receives an ice cream from mother." [F.T.]
ERRORS IN CASE MARKING IN JAPANESE AGRAMMATISM
subject
-ni
-ga
T.S.
4/10
(40%)
6/10
(60%)
I.K.
2/8
(25%)
6/8
(75%) 0/8
F.T.
3/10
(30%)
6/10
(60%)
127
others 0/10
1/10
(0%) (0%) (10%)
Table 2. Case particles substituted for -kara
(3) d. *Kuriininguyasan-ni (-ga) onnanohito-ga (-kara) youfuku-o laundryman-DAT (-NOM) woman-NOM (-ABL) clothes-ACC azukaru be entrusted with "The laundryman is entrusted with the clothes by the woman." [F.T.] We can sum up the results as follows: (1) There is a significant differ ence between the rate of the goal -ni correctly supplied and that of the source -kara correctly supplied; (2) When -kara is not produced, in most cases either -ni or -ga is substituted for -kara incorrectly. 5. Discussion 5.1 Goal-orientedness The results of the experiment in the previous section indicate the asymmetry between the source -kara and the goal -ni, which represent the starting point and the end point of transfer of possession respectively, and are logically symmetrical in the sense that both of them represent one point of the same event. In other words, the results show a tendency for the goal -ni to have priority over the source -kara in sentences produced by agrammatic aphasics.5 However, a question may arise here: is there any possibility that the agrammatic patients do not know the word -kara? The answer is negative. It is clear that they do know the word, -kara, because two of them produced -kara correctly in this experiment (I.K.) and in another experiment (F.K.) and the other, T.S., produced it in another experiment, though incorrectly, as the following examples show (Fujita & Ihara 1992, 1993).
128
HIROKO IHARA & IKUYO FUJITA
(4) a. Shichiya-ga onnanohito-kara kamera-o pawnbroker-NOM woman-ABL camera-ACC azukaru be entrusted with "The pawnbroker buys a camera from a woman." [I.K.] b. Otokonoko-ga saru-kara ringo-o toriageru boy-NOM monkey-ABL apple-ACC take away "The boy takes away an apple from the monkey." [F.T.] c. *Onnanoko-ga ringo-kara (-o) watasu apple-ABL (-ACC) hand girl-NOM "The girl hands an apple." [T.S.] The priority of the goal -ni over the source -kara can be pervasively observed in Japanese, and similar phenomena showing goal-orientedness can be observed in other languages. Now let us look at some linguistic phenomena showing goal-orientedness. First, the goal marker is sometimes substituted for the source marker, but the reverse never occurs. In Japanese, the goal marker -ni can be substituted for the source marker -kara with a limited number of RECEIVEtype verbs such as morau "get", kariru "borrow" and narau "learn", as the following examples show. (5) a. Keiko-ga Taroo-kara/ni hana-o Keiko-NOM Taro-ABL/DAT(=source) flowers-ACC morat-ta get-Past "Keiko got flowers from Taro." b. Taroo-ga Keiko-kara/ni hon-o Taro-NOM Keiko-ABL/DAT( = source) book-ACC kari-ta borrow-Past "Taro borrowed a book from Keiko." However, not all the RECEIVE-type verbs can take -ni as a source marker, and there are some verbs with similar meanings for which only the source marker -kara is used, such as uketoru "receive" and kau "buy".
ERRORS IN CASE MARKING IN JAPANESE AGRAMMATISM
129
(6) Taroo-ga Keiko-kara/*ni hon-o uketot-ta Taro-NOM Keiko-ABL/*DAT( = source) book-ACC receive-Past "Taro received a book from Keiko." On the other hand, the source marker -kara is never substituted for the goal marker -ni. Similar phenomena are found in other languages. One is the use of the goal marker and the source marker in Korean. As is the case in Japanese, the goal marker -eke is used as a source marker with some transfer-of-possession verbs in Korean, as the examples in (7) show. (7) a. John-i -eke chek-ul cu-ess-ta John-subject I-goal book-object give-Past-Decl "John gave me a book." b. nay-ka John-eke chek-ul pat-ass-ta I-subject John-source book-object receive-Past-Decl "I received a book from John. " Another is what is called 'dative of separation'. As the following German example shows, the dative case is used to represent the source role. (8) a. Hans nimmt dem Kind das Spielzeug Hans takes the child(dative) the toy(accusative) "Hans takes the toy from the child." [Ikegami 1981:123] b. Inge hat mir Geld genommen Inge has me-Dat money taken "Inge took money from me." [Smith 1985:392] According to Ikegami (1987), the dative of separation is found with other verbs such as entziehen "take away", rauben "rob", stehlen "steal" and wegnehmen "remove", though this use seems to be restricted to verbs of taking. The dative of separation is much more restricted in English but can be found, as the example in (9) shows. (9) It takes me an hour to finish the work. [Ikegami 1981:123] Second, the goal marker and source marker are strongly contrasted in terms of markedness. The source marker is relatively more marked, as has
130
HIROKO IHARA & IKUYO FUJITA
been pointed out by a number of researchers such as Leech (1969), Bennett (1975), Ikegami (1981,1987) and Jackendoff (1983). (10) a. The cat ran under the table. b. The cat ran from under the table. Under in (10a) means either the location or the goal, while from is necessary for under to mean the source as (10b) shows. Similar phenomena can be found in Japanese. In informal Japanese the goal marker -ni as well as the subject marker -ga and the object marker -o is sometimes omitted. On the other hand, the source marker -kara can never be omitted. (11) a. Keiko(-ga) kaet-ta-no? Keiko(-NOM) go back-Past-Question "Did Keiko go back?" b. Hon(-o) kat-ta book(-ACC) buy-Past "I bought a book." Kinou doko(-ni) it-ta-no? yesterday where(-DAT) go-Past-Question "Where did you go yesterday?" d.Do* (-kara) ki-ta-no ? where*(-ABL) come-Past-Question "Where did you come from?" In addition, the relative markedness of the source is found in passivization (Ikegami 1987). (12) a. Hanako-ga Taroo-ni eigo-o oshieru Hanako-NOM Taro-DAT English-ACC teach "Hanako teaches English to Taro." b. Taroo-ga Hanako-ni eigo-o oshier-areru Taro-NOM Hanako-DAT English-ACC teach-Pass "Taro is taught English by Hanako."
ERRORS IN CASE MARKING IN JAPANESE AGRAMMATISM
131
(13) a. Taroo-ga Hanako-kara eigo-o manabu Taro-NOM Hanako-ABL English-ACC learn "Taro learns English from Hanako." b. *Hanako-ga Taroo-ni eigo-o manabar-eru Hanako-NOM Taro-DAT English-ACC learn-Pass "*Hanako is learned English by Taro." The sentences in (12a) and (13a), which contain the goal marker and the source marker respectively, refer to the same event.6 As the sentence in (12b) shows, the sentence in (12a) can be passivized with the goal role Taroo as the subject, while the sentence in (13a) cannot be passivized with the source role Hanako as the subject.7 Third, as is pointed out in Ikegami (1987), we can see a diachronic semantic shift from 'source-oriented' to 'goal-oriented'. For example, in (14a) the person is considered as a source, while in (14b) the person is considered as a goal. (14) a. ask a favor of a person b. ask a person a favor [Ikegami 1987:127] Source-oriented expressions like (14a) are on the decrease and have been replaced by goal-oriented ones like (14b) in Modern English. The following example exhibits a similar shift in meaning: (15) The wind blows south. [Ikegami 1981:130] Originally sentence (15) meant "The wind blows from the south" like a south wind meaning "a wind from the south". However, the meaning of this expression has changed; it now means "The wind blows to the south". A similar diachronic phenomenon can be found in Japanese as well. According to Nishida (1977), -kara was used as a case particle representing the source in the Heian period (794-1192) but -yori was more frequently used at that time. It was after the Muromachi period (1338-1573) that -kara was more frequently used as the source marker. On the other hand, -ni was used as the goal marker as early as the Nara period (710-794), but it was not used as the source marker until the middle of the Edo period (16031867) as far as we surveyed.8 This historical phenomenon suggests a strong possibility that the goal marker -ni has been substituted for the source marker -kara.
132
HIROKO IHARA & IKUYO FUJITA
5.2 Explanation On the basis of the linguistic phenomena presented in the previous section, it is clear that the asymmetry between the goal -ni and the source -kara observed in the sentences produced by agrammatic aphasics is not aberrant but is one of the linguistic phenomena showing goal-orientedness. Then what does the goal-orientedness result from? How can the unmarkedness of the goal role and the markedness of the source role be explained? Let us look at the linguistic phenomena from the cognitive linguistic point of view. According to Langacker's billiard-ball model (Langacker 1991), the world is viewed as space full of physical objects making contact with one another. Some objects move driven by energy from internal resources, while others put themselves in motion with energy coming from outside. When one object makes physical contact with another object as a result of motion, energy is transmitted from the moving object to the other one, which may then be put in motion. The notion of an action chain, which is useful for describing aspects of clause structure, derives from this billiardball model. An action chain diagrammed in Figure 1 is defined as follows: "A chain of interactions, such that each 'link' involves one participant trans mitting energy to a second, which is thus induced to interact energetically with the next, and so on. The initial energy source is the head of the chain, and the ultimate energy 'sink' is its tail" (Langacker 1991:543).
Looking at the linguistic phenomena in the previous section carefully, we notice that the asymmetry between the goal marker and the source marker is related to the following: the difference in the position of the most salient participant in an action chain between verbs with the goal role and those with the source role, and the question whether the conceptualization is in accord with an action chain in terms of energy flow. For example, the verb, oshieru "teach", in (12) conceptualizes an event of transfer of infor mation, taking the head of the energy flow, the starting point, as the most salient participant, while the verb, manaba "learn", in (13) conceptualizes the same kind of event in the reverse direction, taking the tail, the end point, as the most salient participant. The former is in accord with the action chain with respect to energy flow and the conceptualization is cogni-
ERRORS IN CASE MARKING IN JAPANESE AGRAMMATISM
133
tively natural. This seems to be the reason why the example in (12a) can be passivized as the example in (12b) shows: it is possible to derive an unnatural, marked pattern from a natural, unmarked one. On the other hand, the latter is, as it were, water flowing backward and the concep tualization is cognitively rather unnatural. It is difficult to derive a natural pattern from an unnatural pattern, which seems to be the reason why the example in (13a) cannot be passivized. The predominance of the goal role of transfer-of-possession verbs in the previous section can be explained in the same way.9 In the case of transfer-of-location verbs, the mover is the most salient participant as well as the head, and the goal role is the tail, located downstream. In other words, the transfer-of-location verbs with the goal role conceptualize events in accord with an action chain with respect to energy flow. The location with the source role is, on the other hand, the one that the mover leaves and is located upstream. In other words, the transfer-of-location verbs with the source role conceptualize events in a direction opposite to that of an action chain. In short, goal-orientedness boils down to the way verbs with the goal marker are conceptualized: their conceptualization is in accord with an action chain in terms of energy flow.10 Now we will take a look at the results of the experiment in section 4.2 from the cognitive linguistic point of view and try to give an explanation using the notion of action chain. The head of an action chain is intrinsically salient because it serves as the starting point with respect to energy flow. On the other hand, the subject can be defined as the most prominent participant in a selected predication according to the cognitive point of view (Croft 1991; Givón 1979, 1984; Langacker 1991; Talmy 1985). Therefore, the prototypical subject represents the starting point of an action chain.11 In the case of GIVE-type verbs, the subject represents the giver, who is the starting point of an action chain. On the other hand, in the case of RECEIVE-type verbs, the subject represents the receiver, who is the tail of an action chain. Therefore, the -ga that represents the subject of GIVE-type verbs is prototypical and the -ga that represents the subject of RECEIVEtype verbs is nonprototypical. Taking into consideration that GIVE-type verbs take the goal role and the object as participants in addition to the subject, we can assume that not only -ga but also all the case particles of GIVE-type verbs, i.e. - ga, -o, -ni, make up a prototypical case-marking set of transfer-of-possession verbs, where -ga and -ni represent the head and the tail of an action chain respectively. On the other hand, as RECEIVEtype verbs take the source marker and the object as participants in addition to the subject, it can be assumed that the case particles of RECEIVE-type
134
HIROKO IHARA & IKUYO FUJITA
verbs, i.e. -ga, -o, -kara, make up a nonprototypical case-marking set, where -ga and -kara represent the tail and the head of an action chain respectively, as is diagrammed in Figure 2.12 Therefore, the results of the experiment where -ni was supplied correctly in all cases, while -kara was supplied correctly only in one case reflects the prototypicality of -ga, -o, -ni and the nonprototypicality of -ga, -o, -kara. In other words, the prototypical case-marking set is preserved or reconstructed in agrammatism.13
Next let us turn to the fact that -ga or -ni may be substituted for -kara when -kara is not correctly produced. In the case of the substitution of -ga for -kara, -ga representing the subject is erroneously assigned to the head of an action chain for RECEIVE-type verbs, which assign the subject to the tail. These kinds of errors indicate that -ga representing the subject of GIVE-type verbs is prototypical and is retained in agrammatism and that their unstable access to case particles leads agrammatic aphasics to overuse prototypical ones. As is pointed out in Clancy (1985), the same tendency can be found in the acquisition by children of the verbs of giving and receiving used in benefactive constructions in Japanese. According to Horiguchi (1979), in her observation ageru "give" emerged earlier than morau "receive" and even after morau was acquired, it was used with a higher percentage of errors than ageru. It is reported that the observed child used verbs of giving rather than those of receiving in expressing benefactive relations. This tendency is also found in the results of experiments on children in Fisher et al. (1994), which indicate that the subcategorization of GIVE-type verbs is acquired earlier than that of RECEIVE-type verbs in English. The substitution of -ni for -kara reflects the above-mentioned prototypicality of the case particle set -ga, -o, -ni. Agrammatic aphasics have
ERRORS IN CASE MARKING IN JAPANESE AGRAMMATISM
135
unstable access to case particles and when they are not sure to which par ticipant they should assign the subject -ga, they depend on prototypicality, which is cognitively persuasive. However, as the results of the experiment show, they can once in a while produce a correct nonprototypical case marking, assigning -ga to the tail of an action chain for RECEIVE-type verbs. This occasional correct production suggests that they have latent in formation on the usage of case particles which is difficult for them to ac cess. The fact that -ni is substituted for -kara in most cases even when -ga is correctly produced shows that the case particle set, -ga, -o, -ni, is proto typical and -ni is dissociated from its semantic role, the goal role, in these cases.14 6.
Concluding remarks Based on the results of the experiment, we have seen the priority of -ni representing the goal role over -kara representing the source role in senten ces produced by Japanese agrammatic aphasics: the rate of -ni correctly supplied is significantly higher than that of -kara correctly supplied. We have also seen that the asymmetry observed in agrammatism is one of the linguistic phenomena found pervasively. Then we argued that goal-orientedness, which is seemingly attributable to the issue of semantic roles, is essentially derived from the naturalness of the way the verbs with the goal role depict a scene. We also discussed how the priority of -ni over -kara in Japanese agrammatism can be explained in terms of the notion of action chain. It has been shown that -ga, ~o, -ni of GIVE-type verbs, where -ga and ~ni represent the head and the tail of an action chain respectively, is a prototypical case-marking set, because it is in accord with the energy flow in an action chain, and that is why this case-marking set can be preserved or reconstructed and overused in agrammatism. In short, the priority of -ni over -kara in agrammatism gives a piece of linguistic evidence supporting the idea that human language use tends to follow certain preferred concep tualization patterns of events. It is left for future research to determine whether or not the asymmetry between the goal role and the source role observed in Japanese agrammatism can be observed cross-linguistically.
136
HIROKO IHARA & IKUYO FUJITA
Appendix Subject
T.S.
I.K.
F.T.
Aphasia
Broca
Broca
Broca
Sex
Male
Female
Age
38
Handedness
N1
N2
Male
Female
Male
43
27
45
41
Right
Right
Right
Post Onset
122M
37M
19M
Etiology
LCVA
LCVA
LCVA
WAIS PIQ
80
89
82
Table 3. Information on the subjects
Notes * The authors are very grateful to the two anonymous reviewers for their helpful comments on an earlier version of this paper. We are also grateful to Ad Fooien and Frederike van der Leek for their helpful advice and remarks. 1 Some researchers (Miyagawa 1989 among others) distinguish the nominative -ga and accusative -o from the others, calling the former 'Case markers' and the latter 'postpositions'. We, however, do not distinguish them here and call all of them 'case particles'. 2 Even this kind of word order can be allowed in very informal colloquial speech or emphatic expressions. 3 Some researchers (Chomsky 1981, Miyagawa 1989 among others) claim that the subject and object, which are assigned a structural case, do not provide a particular semantic role to their NPs and that they are defined in terms of their positions in syntactic structure. Other researchers (Langacker 1987, 1991, Croft 1991 among others) claim that the subject and the object are defined in terms of their meaning other than semantic roles such as the perspective point. 4 Broca's aphasia is not necessarily diagnosed as agrammatism and vice versa. 5 It might be too much to draw a definite conclusion from the results of a single experiment in which only three aphasic subjects participated. However, as can be seen in section 4.2, the results are very clear as far as the three patients are concerned, and it seems appropriate to regard the results as sufficient evidence to accept the strong tendency observable in them. 6 Here a question may arise whether the sentences in (12a) and (13a) refer to the same event in a strict sense, because the sentence in (12a) does not necessarily imply that Taro learns English as the following sentence shows: Hanako-ga (Hanako-NOM) Taroo-ni (TaroDAT) eigo-o (English-ACC) oshieta (teach-Past)-ga (but), Taroo-wa (Taro-Top) eigo-o
ERRORS IN CASE MARKING IN JAPANESE AGRAMMATISM
137
(English-ACC) manabanakatta (learn-Neg-Past), meaning "Hanako taught English to Taro, but he didn't learn it." However, assuming that oshieru "teach" and manabu "learn" cover a different range of the same event, we can say that both of them refer to the same event. 7 One of the editors of this book gave us very helpful remarks, suggesting the following: this kind of passivization could be explained with the notion of 'affectedness' based on the difference in affectedness of NP2 between NP1 V NP2 NP3 and NP1 V NP3 to NP2 in English. In Japanese, however, although NP- (DAT) is not 'affected' as the example in note 6 shows, sentences with it as the subject can be passivized as we can see in sentence (12b), in contrast to the fact that NP2 in NP1 V NP3 to NP2 is not affected and this construction cannot be passivized with NP2 as the subject. Therefore, as far as transferof-possession verbs in Japanese are concerned, affectedness does not seem to play a decisive role. 8 Unfortunately we could not find any research concerning the question when exactly -ni began to be used as the source marker. On checking a number of literary works from the Heian period through to the Edo period, we could find only two examples in which -ni seems to be the source marker, in Ugetsu Monogatari ("The Story of Rain and the Moon") written in 1776. This led us to conclude that -ni was not used frequently as the source marker until modern times. 9 Kumashiro (1993,1994) explains how -ni with the Recipient Sense is extended to -ni with the Agent-Source Sense via the Causee and the Agent Sense, using the network-based approach. Kabata & Rice (1997) account for the question how -ni with a basic sense is extended to -ni with more abstract senses in a similar way. 10 According to Smith (1985, 1993), the dative case in German has the subject-like properties of control or potency as well as the object-like properties of affectedness and it is shown that the two-sided nature of the dative can be explained in terms of the energy flow of an action chain. Although the Japanese dative case-marker -ni also has this two-sided nature, as is pointed out by a number of researchers such as Ikegami (1981) and Teramura (1982), we have not taken it into consideration here, because properties of control and affectedness do not seem to play decisive roles concerning the asymmetry between -ni representing the goal role and -kara representing the source role: both NP- used with GIVE-type verbs and NP-kara used with RECEIVE-type verbs have control and neither of them is affected. 11 Some researchers such as MacWhinney (1977) and Croft (1991) point out that the starting point of an event tends to be the unmarked subject. 12 In addition to GIVE-type and RECEIVE-type verbs, there is another kind of verb in transfer-of-possession verbs: PASS-type verbs, in which -ga represents the possessed, -kara represents the source (the giver) and -ni represents the goal (the receiver). For example, Houseki-ga (jewel-NOM) dorobou-kara (thief-ABL) shichiya-ni (pawnbroker-DAT) wataru (pass), "The jewel goes from the thief to the pawnbroker." The same aphasics served as subjects in another experiment with PASS-type verbs, but none of them assigned -ga to the possessed correctly. 13 As the head of transfer-of-possession verbs is higher in Agency than the tail and Agency is one of the parameters of Transitivity in Hopper & Thompson (1980), it may seem at first sight that the difference in height in Transitivity between GIVE-type and RECEIVE-type verbs can explain the prototypicality of GIVE-type case marking. However, the tendency for the head of an action chain to have high potency comes from the head's
138
HIROKO IHARA & IKUYO FUJITA
being the source of energy flow. In other words, high potency is attributable to the energy flow of an action chain. In addition, it is clear that high potency cannot explain all the linguistic phenomena indicating goal-orientedness mentioned in section 4.1. 14 We can predict from the results of our experiment that these patients will correctly produce -ni used with verbs which take -ni and -kara for the source role such as kariru "borrow" by using the prototypical case particle set. However, whether our prediction is correct or not will be left for future research.
References Benett, David C. 1975. Spatial and temporal uses of English prepositions: An essay in stratificational semantics. London: Longman. Caplan, David & Christine Futter. 1986. "Assignment of thematic roles to nouns in sentence comprehension by an agrammatic patient". Brain and Language 27.117-134. Caramazza, Alfonso & Rita S. Berndt. 1985. "A multicomponent deficit view of agrammatic Broca's aphasia". Agrammatism. ed. Mary-Louise Kean, 27-63. New York: Academic Press. Chomsky, Noam. 1981. Lectures on government and binding. Dordrecht: Foris. Clancy, Patricia M.. 1985. "The acquisition of Japanese". The crosslinguistic study of language acquisition, Vol. 1: The data. ed. Dan I. Slobin, 373-524. Hillsdale, N.J.: Erlbaum. Croft, William. 1991. Syntactic categories and grammatical relations: The cognitive organization of information. Chicago & London: University of Chicago Press. Fisher, Cynthia, Hall, D. Geoffrey, Rakowitz, Susan & Lila Gleitman. 1994. "When it is better to receive than to give: Syntactic and conceptual constraints on vocabulary growth". The acquisition of the lexicon, eds. Lila Gleitman & Barbara Landau, 333-375. Cambridge, Mass.: MIT Press. Fujita, Ikuyo. 1991. "Nihongo no shitsubunpou to sakubunpou no tokusei to kaifuku patan" [The characteristics and recovery patterns of Japanese agrammatism and paragrammatism]. Higher Brain Function Research 11.96-103. Fujita, Ikuyo & Hiroko Ihara. 1992. "Processing of verbs in agrammatic aphasics". Proceedings of the 30th Annual Meeting, Academy of Aphasia 47-48. . 1993. "Processing of verbs in sentence production in aphasie patients". Japan Journal of Logopedics and Phoniatrics 34.7-13.
ERRORS IN CASE MARKING IN JAPANESE AGRAMMATISM
139
Givón, Talmy. 1979. On understanding grammar. New York: Academic Press. . 1984. Syntax: A functional typological introduction, Vol.1. Amsterdam & Philadelphia: John Benjamins. Hopper, Paul & Sandra Thompson. 1980. "Transitivity in grammar and discourse". Language 56.251-299. Horiguchi, Junko. 1979. "Nenshouji no jukyuu hyougen" [Transfer-ofpossession expressions by young children]. Kotoba no Hattatsu: Brain function and language acquisition. eds. Fred C. Peng & Motoko Hori, 51-76. Hiroshima: Bunka Hyouron Publishing Co. Ihara, Hiroko & Ikuyo Fujita. 1995. "On dissymmetry between -ni and -kara in sentences produced by agrammatic aphasics". Proceedings of the 111th Annual Meeting, the Linguistic Society of Japan 53-58. Ikegami, Yoshihiko. 1981. Suru to naru no gengogaku [DO-language and BECOME-language]. Tokyo: Taishukan. . 1987. "'Source' vs. 'goal': a case of linguistic dissymmetry". Concepts of case. eds. René Dirven & Günter Radden, 122-146. Tübingen: Gunter Narr Verlag. Jackendoff, Ray. 1983. Semantics and cognition. Cambridge, Mass.: MIT Press. Kabata, Kaori & Sally Rice. 1997. "Japanese ni: The particulars of a somewhat contradictory particle". Lexical and syntactical constructions and the construction of meaning. eds. Marjolijn Verspoor, Keedong Lee & Eve Sweetser, 107-127. Amsterdam & Philadelphia: John Benjamins. Kamio, Akio. 1979. "Shitsugoshou kanja no tougo nouryoku no shougai" [Syntactic deficit of aphasics]. Shitsugoshou to sono chiryou [Aphasia and therapy]. ed. Sumiko Sasanuma, 80-138. Tokyo: Taishukan. Kumashiro, Toshiyuki. 1993. "How the goal can be the source: The semantics of the Japanese dative marker ni". LACUS 20.401-417. . 1994. "On the conceptual definitions of adpositions and case markers: A case for the conceptual basis of syntax". Chicago Linguistic Society 30.236-250. Langacker, Ronald W. 1987. Foundations of cognitive grammar, Vol. 1. Stanford, Calif.: Stanford University Press. . 1991. Foundations of cognitive grammar, Vol. 2. Stanford, Calif.: Stanford University Press. Leech, Geoffrey N. 1969. Toward a semantic description of English. London: Longman.
140
HIROKO IHARA & IKUYO FUJITA
Linebarger, Marcia C. 1990. "Neuropsychology of sentence parsing". Cognitive neuropsychology and neurolinguistics. ed. Alfonso Caramazza, 55-122. Hillsdale, N.J.: Erlbaum. MacWhinney, Brian. 1977. "Starting points". Language 53.152-68. Miyagawa, Shigeru. 1989. Syntax and semantics, Vol. 22: Structure and case marking in Japanese. New York: Academic Press. Nishida, Naotoshi. 1977. "Joshi" [Particle]. Iwanami kouza nihongo 7, bunpou II [Iwanami Japanese language series 7, grammar II]. eds. Susumu Ohno & Takeshi Shibata, 191-290. Tokyo: Iwanami. Sasanuma, Sumiko, Kamio, Akio & Masahito Kubota. 1990. "Agrammatism in Japanese: Two case studies". Agrammatic aphasia: Cross-language narrative sourcebook. eds. Lise Menn & Loraine K. Obler, 1225-1307. Amsterdam & Philadelphia: John Benjamins. Schwartz, Myrna F., Linebarger, Marcia C, Saffran, Eleanor M. & D.S. Pate. 1987. "Syntactic transparency and sentence interpretation in aphasia". Language and Cognitive Processes 2.85-113. Smith, Michael B. 1985. "Event chains, grammatical relations, and the semantics of case in German". Chicago Linguistic Society 21.388-407. . 1993. "Cases as conceptual categories: Evidence from German". Conceptualizations and mental processing in language. eds. Richard A. Geiger & Brygida Rudzka-Ostyn, 531-565. Berlin: Mouton de Gruyter. Talmy, Leonard. 1985. "Lexicalization patterns: Semantic structure in lexical forms". Language typology and syntactic description. ed. Timothy Shopen, 57-149. Cambridge: Cambridge University Press. Teramura, Hideo. 1982. Nihongo no shintakusu to imi [Japanese syntax and semantics]. Tokyo: Kuroshio. Tonkonogy, Joseph M. 1986. Vascular aphasia. Cambridge, Mass.: MIT Press.
Verbal Aspect and Construal Agata Kochańska Warsaw University, Warsaw, Poland 1.
Introduction * In the present paper I am going to analyse only a very minute area in the vast territory of the semantics of verbal aspect in Polish - namely, the behaviour of perfective and imperfective verb phrases in sentences in which a number of such phrases are coordinated in a sequence. Verbs in Polish, like verbs in other Slavonic languages, may be clas sified as either morphologically perfective or imperfective. Any detailed discussion concerning the ways of forming perfective and imperfective verb forms in Polish is beyond the scope of the present paper. Let me only note here that in all the examples discussed below imperfective forms are mor phologically simple and perfective ones are derived by prefixing imper fective stems. Typically, perfective verb phrases denote whole completed processes, while imperfective ones designate processes in progress, viewed from within (cf. Dahl 1981:82; see also Comrie 1985, ch.1 for an overview of characte rizations of the perfective and the imperfective morpheme). Imperfective verb phrases may also have repetitive or habitual meanings. For reasons of space I will limit the present discussion to sentences in the past tense in which each verb phrase refers to a single occurrence of a process. Thus, I am not going to deal with habitual or repetitive uses of verbs. Moreover, I am going to consider only sentences with verb phrases denoting processes which, rather than being momentary, occupy a certain span of time. In my analysis I will adopt a cognitive approach to meaning and grammar in general, and to verbal aspect in particular. First of all, I will adopt Langacker's idea that the grammar of a language should be charac terized as "those aspects of cognitive organization in which resides a speaker's grasp of established linguistic convention" (1987a:57) or as "a structured inventory of conventional linguistic units" (ibid.), where a unit is understood as any semantic, phonological or symbolic structure mastered by a native speaker to such an extent that it becomes a cognitive routine for him (cf. Langacker 1988a: 11).
142
AGATA KOCHAŃSKA
Such an understanding of the notion of grammar leads, among other things, to granting the status of a unit to any meaning of a given linguistic structure arising in a specific context of use, provided that this meaning is sufficiently conventionalized for a sufficient number of speakers of a lan guage (cf. Langacker 1988a: 11). Whether or not this specific meaning may be derived from some more general principles does not matter for its unit status. Also, adopting this understanding of grammar results in the claim that a characterization of an expression's meaning should not be limited to those of its aspects which are common to all uses of the expression in question. Instead, a full characterization of an expression's meaning should take into account all its conventionalized meanings, even if some of them appear only in specific contexts. Therefore, it may be expected that such a characteriza tion does not normally reduce to a single sense but rather is a network of interrelated conventionalized meanings. Among further theoretical assumptions that I would like to make here is an encyclopaedic and subjectivist view of meaning (cf. Langacker 1988b). More specifically, I would like to adopt Langacker's claim that the meaning of a linguistic expression reduces to the conceptualization this expression evokes. As such, it should be characterized not only in terms of the expres sion's objective content but also in terms of how this content is construed (cf. ibid.). Let me now turn to the more specific assumptions made in the present analysis. First of all, following Langacker, I would like to claim that every verb profiles a process, that is, a series of states distributed through concei ved time and scanned sequentially (cf. 1987a:491; 1987b:75). This means that a conceptualization of a process is a conceptualization of a series of relational configurations between (or among) the process participants. These component configurations are not thought of as holding simultaneously. Instead, they are conceptualized as being distributed through conceived time, where conceived time is understood as an object of conceptualization and not its medium. In the conceptualization of a process, the component relational configurations (or states) are scanned sequentially, which means that "[t]he various phases of an evolving situation are examined serially, in non-cumulative fashion; hence the conceptualization is dynamic, in the sense that its contents change from one instant to the next" (Langacker 1987b:72). I will furthermore assume (cf. Langacker 1991:33) that every verbal stem considered in isolation profiles a process type, that is a process not thought of as having any particular location in time, which is its domain of instantiation. On the other hand, every finite clause profiles a grounded instance of a process type. A process instance is thought of as having a
VERBAL ASPECT AND CONSTRUAL
143
particular location in time. Being grounded, it is conceptualized as located in time and reality/irreality specifically relative to the speaker, the hearer and the speech event. Grounding in a finite clause is achieved through the use of particular tense and modal predications (cf. ibid. :241). Finally, a finite clause profiles a process instance which is not only grounded but also quantified - a given sentence may profile either a whole instance of a pro cess, or some portion thereof as falling within the adopted viewing frame (cf. ibid.:421). A viewing frame, in turn, is to be understood as that part of the conceived scene which is "immediately accessible for focused observa tion" (ibid. : 441). The quantification over a process instance is of special interest for the present analysis since it is effected by aspectual predications (cf. ibid.). According to Langacker (cf. 1987b:78ff.), process types may be classified into two main groups: those that are conceptualized as being bounded in time within the viewing frame and those that are thought of as temporally unbounded.1 It should be noted that the bounded/unbounded distinction made in cognitive grammar differs from the traditional telic/atelic contrast (cf. Garey 1957:106). According to Langacker, a process type is bounded not only if it inherently involves the conception of a terminal point, whose attainment results in completing the process in question. A process type is temporally bounded simply if its component states are normally thought of as being distributed over a limited span of time and its temporal boundaries are conceptualized as falling within the adopted viewing frame (cf. ibid. :80ff.). It is important to stress here that what matters for classifying a process type as either temporally bounded or unbounded is how this process is typically conceptualized. The objective properties of a real-world situation are not important. Thus, for example, the real-world situation that may be described by means of the sentence Harry knows French has an objective beginning (the moment when Harry mastered French to a certain degree) and ending (for instance, Harry's death). However, this is not how we nor mally think of a situation of this kind. We conceptualize it as a state which is indefinitely extended in time because its endpoints do not fall within any 'human-sized' temporal viewing frame. Therefore, the process type profiled by the verb know should be classified as temporally unbounded. The second criterion used by Langacker in classifying process types is their internal homo- or heterogeneity (cf. ibid. :80ff.). A process type is internally heterogeneous if its component states are changing through time. On the other hand, a process type is internally homogeneous if all its component states are "construed as being effectively identical" (ibid.:80). It
144
AGATA KOCHAŃSKA
should be noted here once again that homogeneity in Langacker's sense has nothing to do with objective properties of real-world situations. A conceptu alization may be said to be internally homogeneous if "it is construed as homogeneous for expressive purposes, i.e. its internal variability is not in focus as a salient or explicit matter of predication" (Langacker 1987a:205). According to Langacker (cf. 1987b:80ff.) temporally unbounded pro cesses are at the same time internally homogeneous. On the other hand, temporally bounded processes are typically internally heterogeneous - a con ceptualization of a process type traditionally classified as telic encompasses its natural beginning and ending, as well as the intermediate states that constitute the transition between them. However, a temporally bounded process type does not have to be in ternally heterogeneous. As mentioned above, a process type is temporally bounded simply if its temporal expanse is thought of as being limited. Thus, internal heterogeneity is not a necessary prerequisite for temporal bounding. As a result, in Langacker's classification there exist process types which are at the same time internally homogeneous and temporally bounded (cf. 1987b:85ff.).2 Among such process types are those profiled by verbs such as, for instance, sleep, walk, swim etc. As Langacker claims, "[p]rocesses like these typically occur in bounded episodes, rather than continuing in definitely. Their episodic nature is evidently incorporated as part of the conventional value of these verbs, and is responsible for their categorization as [temporally bounded]" (ibid.86).3 I would like to make two comments here. First, it should be noted that process types which are classified by Langacker as internally homogeneous but temporally bounded are treated as temporally unbounded in many other analyses. It is often claimed that verbs such as sleep denote temporally unbounded states (cf. Klein 1994:10; Talmy 1988:179), while verbs such as walk or swim denote activities (in Vendler's sense of this term - cf. Vendler 1967), which are again considered to be temporally unbounded (or durative, as opposed to terminative - cf. Verkuyl 1993).4 It seems to me that, unlike the above-mentioned analyses, Langacker's classification of process types such as sleep or walk as temporally bounded allows for differentiating them from true states such as love or hate, thereby making it possible to explain why verbs profiling the former typically occur in progressive tenses, while those designating the latter do not (cf. Langacker 1987:85ff.). The second comment that I would like to make here has to do with the usefulness of the notion of homogeneity in aspectual considerations. The idea of homogeneity as an explanatory notion in an analysis of aspect has been severely criticized by, for example, Verkuyl (cf. 1993:196ff.). One of the points that he raises (cf. ibid. :206ff.) is that this notion is useless in
VERBAL ASPECT AND CONSTRUAL
145
differentiating sentences which are durative in his analysis - such as John pushed a cart - from sentences that he classifies as terminative - such as John ate a piece of fruit-cake. The first sentence is classified as durative because it can combine with a for + NPspan of time adverbial and cannot combine with an in + NPspan of time adverbial. The second sentence behaves in the opposite way and is thus classified as terminative. This is illustrated in (1) and (2) below. (1) a. b. (2) a. b.
John pushed a cart f or hours *John pushed a cart in an hour *John ate a piece offruit-cake for hours John ate a piece offruit-cake in an hour
Nevertheless, both sentences are said to be characterized by homogeneity since in both cases if χ V-ed is true for any interval of time, it is also true for any portion of this interval. Therefore, the notion of homogeneity is said to be irrelevant for aspectual considerations. My answer to a criticism of this kind is that, first of all, under Langacker's understanding of the notion of homogeneity, it is only the process instance profiled by John pushed the cart that is internally homoge neous. On the other hand, the process instance profiled by John ate a piece of fruit-cake is internally heterogeneous since in each component phase of this process its landmark (a piece of fruitcake) gets smaller and smaller as a result of what the trajector (John) does. Therefore, the profiled relationship between the process participants is construed as changing through time. The fact that any portion of the internal component states of this process may be referred to by the same sentence is irrelevant here. Furthermore, it seems to me that the notion of internal homo- or heterogeneity may be useful in explaining the behaviour of verb phrases combined with for + NPspan of time and in + NPspan of time adverbials. This behaviour, illustrated in (1) and (2) above, is a classic test used for differentiating temporally bounded and unbounded processes.5 However, I have argued elsewhere (cf. Kochańska 1996:500ff.) that it is not the absence of temporal bounding which is responsible for the unacceptability of combining a verb phrase with an adverbial of the in + NPspan of time type, but rather the internal homogeneity of the profiled process instance. One argument for this claim is that in Polish there exists a group of morpho logically perfective verbs designating internally homogeneous but clearly temporally bounded process types and that a combination of such a verb
146
AGATA KOCHAŃSKA
with an adverbial of the in + NPspan of time type leads to unacceptability. This is illustrated by (3) below. (3) *Piotr pospacerował w godzinę Piotr po(Perf.)-walked in hour *"piotr had a walk in an hour" Let me now return to the main theoretical assumptions of the present analysis. As mentioned above, according to Langacker every verbal stem profiles a process type. Process types may be divided into three groups, which are represented, respectively, in Figures la, b, and below. (Note that in all the figures used in the present paper those states of the respective processes which are put in profile are drawn in heavy lines.)
Figure 1. A classification of process types: (a) temporally unbounded, internally homo geneous process, (b) temporally bounded, internally heterogeneous process, (c) temporally bounded, internally homogeneous process. t = conceived time, VF = viewing frame
In (a) we have processes which are internally homogeneous and temporally unbounded, like have black hair. Figure (b) represents processes which are typically conceptualized as internally heterogeneous and temporally bounded, like repair a car or read a letter. Finally, Figure (c) depicts processes which are internally homogeneous but nevertheless temporally bounded (like sleep or read). A process type profiled by a verbal stem is quite schematic in nature, due to the maximally schematic characterization of the process participants inherent in its profile. The intermediate levels of organization in the assem bly of a finite clause profile instances of higher-order process types.6 First,
VERBAL ASPECT AND CONSTRUAL
147
when a verb stem is integrated with a subject NP, a direct object NP and/or possibly a PP, the nature of the process participants becomes specified. This results in a higher-order process type specification - for example, plurality of the subject or object NP may result in a conceptualization of a replicate process (cf. Langacker 1991:112). Also quantification effected by aspectual predications may bring about a conceptualization of a higher-order process type. For example, the semantic import of the progressive construction in English (cf. Langacker 1995:175ff.) is that the profile imposed upon the base conceptualization of a temporally bounded process encompasses only a portion of its internal states. Therefore, the endpoints of the process in question are excluded from the adopted viewing frame (for convenience, I speak of the endpoints of a process rather than of its beginning and ending). The internal states are viewed as homogeneous by virtue of being construed at a high level of abstraction simply as representative internal states of the process in question (cf. ibid. :180). In this way the progressive construction derives a higher-order temporally unbounded and internally homogeneous process type specification. After discussing the theoretical framework of the present study let me now turn to an analysis of the data. In what follows I will investigate the behaviour of perfective and imperfective verb phrases under coordination, as well as how this behaviour is influenced by the choice of a particular temporal viewing frame, which should be a decisive factor in determining whether it is a whole instance of a given process type or only a portion of it that is put in profile. 2.
An analysis of the data Since I am going to analyse the behaviour of perfective and imperfec tive verb phrases coordinated in a sequence, let me start the discussion of the data with a few remarks concerning coordination. According to Langacker, coordination in its pure form, such as coordination by means of the logical and, "reduces to the mental juxtaposition of co-equal structures, on either a simultaneous or an alternating basis" (1991:472). This means that no relationship between (or among) the coordinated entities constitutes an object of conceptualization but their mental juxtaposition is just a part of the conceptualization process. However, as Langacker observes (cf. ibid. :429 fn. 8), instances of such pure coordination are very rare in actual language use. For example, coordinating a number of verb phrases in a sentence often results in profi ling process instances which are thought of as occurring in a temporal sequence. It should be noted that this meaning of temporal sequencing arises
148
AGATA KOCHAŃSKA
at the level of the whole coordinated construction and cannot be derived from the meanings of the construction's component parts - the meaning of temporal sequencing arises even if a number of verb phrases are coordinated by a mere juxtaposition in a linear sequence, without using any conjunction. A sequence of coordinated verb phrases may also be understood as profiling a number of process instances occurring simultaneously. This meaning of simultaneous occurrence again goes beyond the purely composi tional meaning of the construction, which reduces to the mental juxtapositi on of the conceived process instances. When the meaning of a series of verb phrases coordinated in a sentence involves the notion of temporal sequencing or simultaneity (or, for that matter, any other kind of relationship between the profiled process instances, such as e.g. causation) then the evoked conceptualization does not reduce to a number of process instances which are separately profiled and merely mentally juxtaposed. In such a case the notion of a temporal rela tionship between the profiled process instances receives at least some degree of objective construal. This means that it becomes also an object of conceptualization and not only a part of the conceptualization process. After these general remarks concerning coordination I finally turn to the discussion of the data. Let me start with the sentences under (4) and (5), where the viewing frame imposed on the conceived situation is momentary. This viewing frame is specified by an adverbial phrase at the beginning of each sentence. The sentences under (4) illustrate the behaviour of sequences of perfective verb phrases under a momentary viewing frame. (4) a. *W chwili, gdy zadzwonił telefon, Piotr zjadł At moment when za(Perf.)-rang phone, Piotr z(Perf.)-ate kanapkę i przeczytał list sandwich and prze(perf.)-read letter *"When the phone rang, Piotr ate a sandwich and read a letter"
VERBAL ASPECT AND CONSTRUAL
149
(4) b. *W chwili, gdy zadzwonił telefon, Piotr At moment when za(Perf.)-rang phone, Piotr przeczytał list i zreperował samochód prze(Perf.)-read letter and z(Perf.)-repaired car *"When the phone rang, Piotr read a letter and repaired a car" It may be observed that under a momentary viewing frame a sequence of perfective verb phrases is generally unacceptable. This is because using a perfective verb phrase in a sentence results in profiling a whole instance of the process in question, together with its endpoints. This characterization of perfective verb phrases is in keeping with their traditional description, in which they are said to denote whole completed processes. Since each pro cess profiled in (4) is thought of as having a certain temporal duration, the conceptualization of even a single instance of such a process, let alone a sequence of instances, is inconsistent with the adopted momentary viewing frame. This situation is illustrated in Figure 2, where the profiled states of a process (drawn in heavy lines) go beyond the boundaries of the adopted viewing frame:
Figure 2. A perfective VP interpreted under a momentary VF. t = conceived time, VF = viewing frame
Sequences of perfective verb phrases may be acceptable under a momentary viewing frame if the profiled processes are thought of as having a punctual duration and their simultaneous occurrence is not excluded by our world knowledge. Each perfcctivc vcrb phrase in (4) may alternatively be interpreted as profiling only the transition from the last but one to the final state conceptualized against the base of a whole process instance. Under this interpretation the meaning of each VP would not be V-ed but rather finished V-ing. This situation is depicted in Figure 3.
150
AGATA KOCHAŃSKA
Figure 3. A series of perfective verb phrases interpreted under a momentary viewing frame as profiling a simultaneous completion of the base processes. t = conceived time, VF = viewing frame
Under such an interpretation sentence (4a) would become acceptable since a simultaneous occurrence of the momentary processes it profiles is possible according to our world knowledge. Let us now turn to the sentences under (5), in which the viewing frame is again momentary, but the verb phrases put in a sequence are imperfective. In examples (5a) and (b) these imperfective verb phrases refer to real-world situations which are normally thought of as telic. In (c) and (d) we have imperfective verb phrases referring to atelic situations. And finally, the imperfective verb phrases in (e) and (f) refer to states. (5) a.
W chwili, gdy zadzwonił telefon, Piotr jadł At moment when za(Perf.)-rang phone, Piotr ate(Imperf.) kanapkę i czytał list sandwich and read(Imperf.) letter "When the phone rang, Piotr was eating a sandwich and reading a letter" b. *W chwili, gdy zadzwonił telefon, Piotr At moment when za(Perf.)-rang phone, Piotr czytał list i reperował samochód read(Imperf.) letter and repaired(Imperf.) car *"When the phone rang, Piotr was reading a letter and repairing a car"
VERBAL ASPECT AND CONSTRUAL
151
(5) c.
W chwili, gdy zadzwonił telefon, Piotr leżał At moment when za(Perf.)-rang phone, Piotr lay(Imperf.) w łóżku i czytał in bed and read(Imperf.) "When the phone rang, Piotr was lying in bed and reading" d. *W chwili, gdy zadzwonił telefon, Piotr At moment when za(Perf.)-rang phone, Piotr spał i czytał slept(Imperf.) and read(Imperf.) *"When the phone rang, Piotr was sleeping and reading" e. W chwili, gdy otworzyłam oczy, iglica kościoła, eyes, steeple of-church, At moment when I-opened jak zawsze, górowała nad miastem i as always, towered(Imperf.) over town and kierowała myśli wiernych ku Bogu directed(Imperf.) thoughts of-believers to God "When I opened my eyes, the church steeple, as always, towered over the town and directed the thoughts of believers to God" f. *W chwili, gdy otworzyłam oczy, iglica kościoła, At moment when I-opened eyes, steeple of-church, jak zawsze, górowała nad miastem i as always, towered(Imperf.) over town and ginęła w morzu innych zabudowań got-lost(Imperf.) in sea of-other buildings *"When I opened my eyes, the church steeple, as always, towered over the town and was lost in the sea of other buildings"
It seems that under a momentary viewing frame a sequence of imperfective verb phrases in a sentence is acceptable only when a simultaneous occurrence of the profiled processes is possible according to our world knowledge. This is the case with telic processes such as eat a sandwich and read a tetter in (5a), with atelic ones such as lie in bed and sleep in (5c), and with true states such as tower over the town and direct the thoughts of believers to God in (5e) - hence these sentences are acceptable.7 On the
152
AGATA KOCHAŃSKA
other hand, a simultaneous occurrence of processes such as read a letter and repair a car in (5b), sleep and read in (5d), and tower over the town and be lost in the sea of other buildings in (5f) is judged impossible. Consequently, these examples are unacceptable. An imperfective verb phrase used in a sentence may be interpreted as profiling only a portion of the internal states of a process type that constitu tes its base - it is traditionally characterized as denoting an ongoing process viewed from within. Such an interpretation is in agreement with the adopted momentary viewing frame. Also a sequence of imperfective verb phrases may receive an interpretation consistent with the momentary viewing frame if the profiled portions of processes may be thought of as occurring simulta neously. This is illustrated in Figure 4.
Figure 4. A series of imperfective VPs interpreted under a momentary VF. (a) VPs profiling a portion of internal states of temporally bounded and internally heterogeneous process instances (sentence (5a)), (b) VPs profiling a portion of temporally bounded and internally homogeneous process instances (sentence (5c)), (c) VPs profiling a portion of internal states of true states, that is, of temporally unbounded process instances ((sentence (5e)). t = conceived time, VF = viewing frame
VERBAL ASPECT AND CONSTRUAL
153
Variants (a), (b) and (c) of Figure 4 represent the interpretations assigned to examples (5a), (5c) and (5e), respectively. Sentence (5a) thus profiles temporally bounded and internally heterogeneous process instances, (5c) profiles process instances which are temporally bounded but internally homogeneous, and (5e) designates true states (temporally unbounded process instances). In all those cases only a portion of the internal states coextensive with the viewing frame is profiled. These states are construed as effectively homogeneous, no matter whether the base process is conceptualized as internally homo- or heterogeneous. On the other hand, when a sentence with a sequence of imperfective verb phrases denotes processes which may only be thought of as occurring in a temporal sequence, such a sentence is unacceptable. This is because the conception of temporal sequencing of processes requires the conception of their endpoints and these are excluded from the viewing frame when the viewing frame is momentary and only selected internal states of the base processes are profiled. Let us now turn to examples in which the viewing frame adopted in a sentence encompasses an extended period of time. This situation is illustrated in the examples under (6) to (9). Again, the viewing frame is specified by an adverbial phrase at the beginning of each sentence. In (6) we have examples of sentences in which it is perfective verb phrases that are put in a sequence. These verb phrases denote telic processes. (6) a. Wczoraj między 8 a 11 rano Piotr Yesterday between 8 and 11 in-morning Piotr zjadł kanapkę i przeczytał list z(Perf.)-ate sandwich and prze(Perf.)-read letter "Yesterday between 8 and 11 in the morning Piotr ate a sandwich and read a letter" b. Wczoraj między 8 a 11 rano Piotr Yesterday between 8 and 11 in-morning Piotr przeczytał list i zreperował samochód prze(Perf.)-read letter and z(Perf.)-repaired car "Yesterday between 8 and 11 in the morning Piotr read a letter and repaired a car"
154
AGATA KOCHAŃSKA
When the viewing frame adopted in a sentence encompasses an extended period of time, a sequence of perfective verb phrases is accep table, no matter whether the profiled processes may occur simultaneously or not. When our world knowledge excludes the possibility of a simultaneous occurrence, the sequence of verb phrases receives a sequential interpreta tion. Since we usually can't read a letter and repair a car at the same time, this is the only interpretation for sentence (6b). On the other hand, example (6a) may be understood as profiling process instances which are thought of as occurring either in a sequence or simultaneously, since both a sequential and a simultaneous occurrence are, in principle, possible. This seems to be the case because a conceptualization of a whole process instance or a sequence of such instances is perfectly consistent with the extended viewing frame adopted in (6). A sequential interpretation of each sentence is possible, since, as illustrated in Figure 5 below, the endpoints of the profiled process instances fall within this kind of viewing frame. It may be observed that the sequential interpretation is in fact not only possible here but also seems to be favoured. This may be attributed to its iconic character. Iconicity, in turn, seems to be a natural tendency in languages since it makes a coding of experience more isomorphic to the experience itself and "[a]11 other things being equal, a coded experience is easier to store, retrieve and communicate if the code itself is maximally isomorphic to the experience" (Givón 1985:189).
Figure 5. A series of perfective VPs under an extended VF — sequential interpretation. t = conceived time, VF = viewing frame
Let us now turn to sentences with imperfective verb forms in (7).
VERBAL ASPECT AND CONSTRUAL
155
(7) a.
Wczoraj między 8 a 11 rano Piotr Yesterday between 8 and 11 in-morning Piotr jadł kanapkę i czytał list ate(Imperf.) sandwich and read(Imperf.) letter "Yesterday between 8 and 11 in the morning Piotr ate from a sandwich and read from a letter" b. Wczoraj między 8 a 11 rano Piotr Yesterday between 8 and 11 in-morning Piotr czytał list i reperował samochód read(Imperf.) letter and repaired(Imperf.) car "Yesterday between 8 and 11 in the morning Piotr read from a letter and worked at repairing a car" c. Wczoraj między 8 a 11 rano Piotr Yesterday between 8 and 11 in-morning Piotr leżał w łóżku i czytał lay(Imperi.) in bed and read(Imperf.) "Yesterday between 8 and 11 in the morning Piotr lay in bed and read"
d. Wczoraj między 8 a 11 rano Piotr Yesterday between 8 and 11 in-morning Piotr spał i czytał slept(Imperf.) and read(Imperf.) "Yesterday between 8 and 11 in the morning Piotr slept and read" Strangely enough, the interpretations assigned to these sequences are very similar to those given to the sequences of perfective verb phrases in (6). Again, both the simultaneous and the sequential interpretations are possible, unless the former are excluded by our world knowledge. When the possibility of a simultaneous occurrence of the two profiled processes is excluded on the basis of our world knowledge, the sequential interpretation is chosen. Hence all the sequences of imperfective verb phrases in (7) are acceptable. This is true both for sequences of verb phrases referring to telic real-world situations (as in examples (7a) and (b)) and for those which denote atelic situations, as in examples (7c) and (d).
156
AGATA KOCHAŃSKA
Of course, there are certain differences between the use of the perfec tive verb forms in (6) and the imperfective ones in (7). For example, the sentences in (6) profile whole, completed processes, while the use of imperfective forms in (7a) and (b) suggests that the final point inherent in the prototypical conception of eating a sandwich, reading a letter or repair ing a car may not have been achieved, despite the fact that these processes had been terminated. Therefore, when the perfective verb phrases are used in (6), whole instances of internally heterogeneous process types are profiled. The use of their imperfective equivalents in (7a) and (b) results in profiling instances of process types without the terminal point beyond which the processes in question cannot continue. What is important, however, is that under the extended temporal viewing frame both the sequences of perfective verb phrases in (6) and those of the imperfective ones in (7) may receive a sequential interpretation. My suggestion is that we experience all processes profiled by the sentences under (7) as occurring in what Langacker calls "bounded episo des" and that temporal boundedness resulting from their episodic nature is a part of the conventional value of the verb phrases in (7), just as it belongs to the meaning of the verb phrases in (6). What differentiates these two groups of verb phrases is their internal structure, hetero- and homogeneous, respectively. When the idea of a final state is present in a conceptualization of a process, all the preceding states are viewed as evolving towards that final state. Therefore, they are heterogeneous in the sense that each consecutive state brings the whole process closer towards its final point. On the other hand, when we exclude the notion of the final state from our conceptualization, we view the process in question as a series of equivalent, homogeneous states. Therefore, I would like to claim that the sentences in (7) profile instances of process types which are internally homogeneous but temporally bounded. Since the temporal viewing frame adopted in these sentences encompasses an extended period of time it is full process instances, together with their temporal boundaries, that are put in profile. As illustrated in Figure 6 below, the fact that the endpoints of the respective process instances fall within the viewing frame makes the sequential interpretation possible.
VERBAL ASPECT AND CONSTRUAL
157
Figure 6. A series of imp effective VPs profiling instances of temporally bounded process types under an extended VF — sequential interpretation. t = conceived time, VF — viewing frame
It may be noted that the meaning of internal homogeneity, which in (7a) and (b) is signalled by the use of imperfective verb phrases, may also be conveyed by the use of perfective verb phrases, as illustrated by the examples under (8). (8) a. Wczoraj między 8 a 11 rano Piotr Yesterday between 8 and 11 in-morning Piotr pojadł kanapkę i poczytał list po(Perf.)-ate sandwich and po(Perf.)-read letter "Yesterday between 8 and 11 in the morning Piotr ate from a sandwich and read from a letter" b. Wczoraj między 8 a 11 rano Piotr Yesterday between 8 and 11 in-morning Piotr poczytał list i poreperował samochód po(Perf.)-read letter and po(Perf.)-repaired car "Yesterday between 8 and 11 in the morning Piotr read from a letter and worked at repairing a car" A detailed discussion concerning the semantics of the prefix poderiving the perfective verb forms in (8) on the one hand and, on the other, the prefixes z- and prze-, which derive the perfective forms in (6), deserves a separate study, since all these prefixes are highly polysemous. I would only like to note here that while the perfective verb phrases in (6) denote processes which proceed from their initial to their final state through a series of non-identical stages, the processes designated by the perfective verb phrases in (8) are conceptualized as internally homogeneous, just like the processes denoted by the imperfective verb phrases in (7). The only difference between the verb phrases in (7) and those in (8) is that when the
158
AGATA KOCHAŃSKA
latter are used in a sentence, they can only be interpreted as profiling full instances of temporally bounded process types, with their endpoints within the temporal viewing frame. On the other hand, the imperfective verb phrases in the examples under (7) may in other contexts, under a momenta ry viewing frame, be interpreted as designating only a portion of the inter nal states of a temporally bounded process, thereby excluding its temporal boundaries from profiling. However, not all imperfective verb phrases share the behaviour of those in the sentences under (7). As an illustration, let us consider the examples under (9). (9) a. Przez wieki iglica kościoła górowała nad Over centuries steeple of-church towered(Imperf.) over miastem i kierowała myśli wiernych town and directed(Imperf.) thoughts of-believers ku Bogu to God "Over the centuries the church steeple towered over the town and directed the thoughts of believers to God" b. *Przez wieki iglica kościoła górowała nad Over centuries steeple of-church towered(Imperf.) over miastem i ginęła w morzu innych zabudowań town and got-lost(Imperf.) in sea of-other buildings *"Over the centuries the church steeple towered over the town and was lost in the sea of other buildings" It seems that a sequence of imperfective verb phrases designating instances of true states is only acceptable when the profiled states may exist simultaneously. When they can't, the sentence is judged unacceptable. This is due to the fact that in the case of a series of verb phrases profiling instances of true states the possibility of a sequential interpretation is blocked. The difference in behaviour between the sequences of imperfective verb phrases in (7) and those in (9) seems to result from the fact that in (7) instances of internally homogeneous but temporally bounded process types are profiled. The sentences in (9), on the other hand, profile instances of process types which are normally thought of as temporally unbounded. As illustrated in Figure 7 below, in the case of such process instances their endpoints do not fall within the viewing frame. This is unproblematic in the
VERBAL ASPECT AND CONSTRUAL
159
case of process instances which may occur simultaneously, as in (9a). However, the absence of endpoints within the viewing frame precludes the possibility of a sequential interpretation. Hence sentence (9b) is unaccepta ble - it profiles two process instances which cannot occur at the same time and which, due to their temporally unbounded character, cannot be thought of as occurring in a sequence, even if the adopted viewing frame is tempo rally extended.
Figure 7. A series of imperfectlye VPs profiling instances of temporally unbounded process types — interpretation under an extended VF. t = conceived time, VF = viewing frame
3.
Conclusions Let me first sum up the main points made in the above analysis of the behaviour of Polish perfective and imperfective verb phrases under coordi nation. When a number of verb phrases are coordinated in a sequence, such a construction usually exhibits meanings that go beyond the purely compositi onal interpretation. One such constructional meaning is the meaning of some temporal relationship, such as simultaneity or sequentiality, between the profiled process instances. It should be stressed that this meaning of the coordinate construction cannot be attributed to any of its component parts but arises only at the level of the construction as a whole. Even if it is assumed that a conjunction such as and or the Polish i is polysemous and "and then" or "and at the same time" belong to its meanings, the claim on the non-compositionality of the coordinate construction's meaning still remains valid. First, the abovementioned meanings of conjunctions such as and appear only when verb phrases are coordinated, so any characterization of those meanings should make reference to the construction itself. Secondly, such meanings also
160
AGATA KOCHAŃSKA
arise when a number of verb phrases are simply juxtaposed without any conjunction. In this case their appearance clearly cannot be attributed to any of the components of the construction but only to the construction itself. The choice of the simultaneous or sequential interpretation of a num ber of verb phrases coordinated in a sentence depends, among other things, upon the temporal viewing frame adopted in that sentence. If the viewing frame is momentary, only the simultaneous interpretation is possible be cause any conceptualization of temporal sequentiality requires a conceptual ization of a span of time (and not a moment) in which the profiled process instances occur in a sequence. On the other hand, under an extended temporal viewing frame both interpretations may arise if, on the basis of our world knowledge, we judge a simultaneous occurrence of the profiled process instances to be possible. However, the sequential interpretation is usually favoured by virtue of its iconic character. The temporal viewing frame adopted in a sentence influences also the aspectual interpretation of the coordinated verb phrases. Under a momentary viewing frame sequences of perfective verb phrases are either judged unacceptable or else interpreted as designating only the transition from the last but one to the final state profiled against the base of a whole process instance. Sequences of imperfective verb phrases, on the other hand, are understood as profiling a portion of the internal states coextensive with the viewing frame. This portion is profiled against the base conception of a full process instance. The temporal boundaries of this process instance are excluded from the viewing frame and the internal states are construed as effectively homogeneous. When the viewing frame adopted in a sentence encompasses an extended span of time, sequences of perfective verb phrases are always acceptable. They are understood as profiling full instances of the respective process types, together with their natural endpoints. These process instances may be thought of as occurring either simultaneously or in a sequence, although, as already mentioned, the sequential interpretation is favoured whenever it is not excluded on the basis of our world knowledge. Profiling full process instances together with their endpoints is consistent with the extended character of the temporal viewing frame. The presence of these endpoints within the viewing frame makes the sequential interpretation possible, since to conceptualize one process instance as occurring after another process instance is to conceptualize the initial state of the latter as occurring after the final state of the former - hence the endpoints of both processes have to fall within the viewing frame adopted by the conceptualizer.
VERBAL ASPECT AND CONSTRUAL
161
When we consider the behaviour of sequences of imperfective verb phrases under an extended temporal viewing frame, it turns out that only sequences of verb phrases similar to those from (7) (such as jeść kanapkę "eat from a sandwich" or spać "sleep") are always acceptable. On the other hand, sequences of verb phrases similar to those from (9) (such as górować nad miastem "tower over the town" or ginąć w morzu innych zabudowań "be lost in the sea of other buildings") are acceptable only when a simulta neous occurrence of the profiled process instances is not excluded by our world knowledge. This difference in acceptability is due to the fact that the former sequences, unlike the latter, may receive a sequential interpretation. As already mentioned, the conception of two process instances occurring in a sequence requires that their endpoints fall within the adopted viewing frame. I suggest that the sequences of verb phrases in (7) profile full instances of internally homogeneous but temporally bounded process types. It is their temporally bounded character, as well as the fact that full process instances are profiled that makes the sequential interpretation possible. On the other hand, the verb phrases in (9) profile process instances which are conceptualized as temporally unbounded. We do not normally think of their endpoints as falling within the viewing frame, no matter how extended a period of time it encompasses. Therefore, the sequences of such verb phrases cannot be interpreted sequentially and they are only acceptable when the simultaneous occurrence of the profiled process instances is consistent with our world knowledge. On the basis of the data considered in the present study I would like to draw a few tentative conclusions of a more general nature. First of all, I would like to suggest that perfective and imperfective verb phrases are polysemous and, as such, cannot be reduced to a single semantic representation but rather constitute networks of interrelated senses. The same perfective verb phrase used in a sentence may designate a full instance of a temporally bounded and extended process type or an instance of a higher-order momentary process type. On the other hand, an imperfec tive verb phrase in a sentence may be interpreted as profiling, among other things, a portion of the internal states of a process which is coextensive with the adopted momentary viewing frame and construed as effectively homoge neous or, alternatively, a full instance of a process type which is internally homogeneous but temporally bounded.8 Furthermore, it seems to me that the meaning of a perfective or an imperfective verb phrase cannot be analysed as the sum of the meanings of a verbal stem and the perfective or the imperfective morpheme. This is
162
AGATA KOCHAŃSKA
because the aspectual interpretation of a verb phrase heavily depends on a sentential and discourse context in which this phrase is used. As illustrated above, this interpretation is influenced by the kind of temporal viewing frame adopted in a sentence. Also, the construction in which a number of verb phrases are coordinated tends to induce interpreting these phrases as profiling full instances of temporally bounded process types whenever an explicitly stated momentary viewing frame does not block such an interpre tation. This is because the favourite constructional meaning of a sequence of verb phrases is that of temporal sequentiality and such a meaning in turn requires that the endpoints of the profiled process instances fall within the viewing frame. Therefore, I suggest that the respective meanings of perfec tive and imperfective verb phrases cannot be characterized autonomously. Instead, they should be characterized relative to constructions and contexts in which the phrases in question may occur. I would also like to suggest that Langacker's classification of process types into temporally unbounded (and, by definition, internally homogene ous), temporally bounded and internally heterogeneous and, finally, tempo rally bounded but internally homogeneous may lead to revealing explana tions in the field of verbal aspect. This classification allows for explaining the behaviour of English verbs in progressive and simple tense constructions (cf. e.g. Langacker 1987b). In my view, the behaviour of Polish perfective and imperfective verb phrases under coordination may also be explained in terms of this classification. It should be noted that, contrary to what is claimed in many other analyses (cf. e.g. Antinucci and Gebert 1977, Klein 1994, Verkuyl 1993), the issue of the temporal boundedness of a process is, in Langacker's classification, kept distinct from the issue of its having a natural terminal point towards which the process evolves and beyond which it cannot continue. For Langacker, a process is temporally bounded simply if its component states are thought of as being distributed over a limited span of time and its temporal boundaries fall within the viewing frame. Under this understanding, internal heterogeneity and the presence of the natural terminal point are not crucial for temporal bounding. The internal states of a process may be construed as effectively homogeneous and yet this process may be thought of as temporally bounded if it is experienced as occurring in bounded episodes and if this experience is incorporated as a part of the conventional value of a verb phrase profiling the process in question. Therefore, following Langacker, I suggest that the notions of temporal bounding and internal homo- or heterogeneity should be kept distinct. What is more, in my view not only temporal bounding, but also internal homo- or heterogeneity are relevant in aspectual considerations - contrary to what is
VERBAL ASPECT AND CONSTRUAL
163
claimed by, for example, Verkuyl. The notion may prove relevant to explaining the behaviour of verb phrases in combinations with temporal adverbials of the for + NPspan of time or in + NPspan of time type, provided that it is understood in terms of construal, and not in terms of the same truth conditions either holding or not holding for a time interval, as well as for any of its sub-portions. Finally, I would like to suggest that for a revealing analysis of aspectual phenomena a truly subjectivist view of meaning in language is necessary. In claiming that aspectual contrasts may pertain to whether a speaker subjectively chooses to portray the same real-world situation as either ongoing or viewed in its totality, many linguists working in the field of verbal aspect in fact recognize the need for a subjectivist approach to aspectual meaning (cf. e.g. Comrie 1985:4; Laskowski 1984:129, Binnick 1991:148). I agree that speakers may construe the same real-world situation in many different ways, for example, by adopting differing viewing frames in their conceptualizations. But I would like to go even further in embracing a subjectivist approach to meaning. It seems to me that meaning in language has to be characterized relative to our world knowledge and that this knowledge does not reflect the objective structure of reality, but rather our subjective experience of this reality. Thus, despite the fact that all (or at least the vast majority of) real-world situations have their beginnings and ends, we subjectively conceptualize some of them as occurring in bounded episodes and others as continuing indefinitely. The decisive factor seems to be whether the temporal boundaries of the processes in question fall within any 'human-sized' temporal viewing frame or not. Conclusions drawn on the basis of such a fragmentary analysis cannot be treated as generalizations valid in the field of verbal aspect as a whole without further extensive investigation. I would like to suggest, however, that cognitive grammar offers a promising framework for analysing the semantics of verbal aspect, as well as its interplay with other phenomena, such as constructions in which aspectual verb forms are used, the discourse context in which they may appear, the meanings of prepositions in temporal adverbials that may combine with verb phrases under consideration, the influence of subject and object NPs on the aspectual behaviour of sentences, and many others. What is even more important, the cognitive framework seems to allow for considering aspectual phenomena in terms of cognitive mechanisms (such as e.g. bounding, adopting differing viewing frames, the ability to conceptualize the same objective entity as either internally homoor heterogeneous) which manifest themselves in a broad range of linguistic, as well as extralinguistic, phenomena.
164
AGATA KOCHAŃSKA
Notes * I would like to express my gratitude to Elżbieta Tabakowska for all her comments on the ideas discussed in the present study. Words of thanks are due to the editors of the present volume for their friendly attitude, their valuable criticism of an earlier version of this paper, and especially their bringing to my attention the lack of precision in my use of terminology. I would like to thank the two anonymous reviewers for all their comments (for example, for pointing out to me the necessity to consider in my analysis the role of the constructional meaning of coordinate structures as actually bringing about some of the results analysed in the present paper). I would also like to thank Emma Harris and Gretchen Siglar for checking and correcting my English. And last but not least, I would like to thank Andrzej and Piotruś, who made my work on the final version of this paper possible, in spite of all the difficulties. Of course, all the remaining flaws and errors in this paper are entirely my own. 1 It should be noted here that Langacker (1987b) does not yet operate with the type/instance distinction. Therefore he simply speaks of temporally bounded or unbounded processes. However, the introduction of the type/instance contrast does not seem to influence the present line of reasoning. 2 In fact, temporal bounding within the viewing frame introduces a minimal degree of internal heterogeneity into our conceptualizations of such process types - their internal states are viewed as effectively homogeneous, but the presence of temporal boundaries within the viewing frame results in conceptualizations of process types involving some change through time. This change, however, is "confined to the initiation and the termination of the process" (Langacker 1987b:86). 3 I have decided to replace Langacker's word perfectives with temporally bounded in this quotation to avoid confusion - throughout the present paper the perfective/imperfective distinction is used in reference to aspectual forms of a verb and this distinction is by no means equivalent to Langacker's classification of process types into perfective (temporally bounded) and imperfective (temporally unbounded). 4 To be more precise, in Verkuyl's analysis the terminative/durative classification applies to aspectual properties of whole sentences and not of verbs in themselves. Therefore, walk in itself cannot be said to be durative but a sentence such as, for example, John walked should be classified in this way. On the other hand, in Langacker's approach this sentence would be characterized as profiling a full instance of an internally homogeneous but temporally bounded process type. Therefore, even if the above qualificati ons are taken into consideration it is still clearly visible that Langacker and Verkuyl treat verbs such as walk and sentences with those verbs used in the activity sense in a quite different way. 5 It may be interesting to observe that this test, though commonly employed and very much taken for granted, is not accompanied by any detailed analysis of the semantics of for and in. 6 If a verbal stem profiles a process type and a finite clause profiles a grounded instance of this process type then the question is at which level of structural organization the conception of a process type is converted into the conception of a specific instance of this process type that is still ungrounded. This, in fact, is a point of some controversy. According to Langacker (cf. 1991:33f) the full specification of the direct object at the level of VP results in tying the conception of a type to a specific participant and "thereby converting it into the conception of an instance" {ibid. :33). This idea has been criticized by
VERBAL ASPECT AND CONSTRUAL
165
Heyvaert (the present volume). In a comment on her paper Langacker suggests that "the subject and grounding predication function together to derive a grounded instance, whereas an object, integrated at a lower level of organization, is more readily seen as deriving a higher-order type specification". 7 The verb phrases in (5e) denote unchanging properties of a building, that is, true states. In other contexts they may be used to designate dynamic processes and not states. 8 The former variant is an example of what may be called the quantificational function of a verb phrase's aspectual form - the choice of a perfective or an imperfective verb phrase in a sentence results in profiling, respectively, a full instance of the process type in question or only a portion of its internal states.
References Antinucci, Francesco and Lucyna Gebert. 1977. "Semantyka aspektu czasownikowego". Studia gramatyczne I ed. Zuzanna Topolińska, 743. Wrocław: Ossolineum. Binnick, Robert I. 1991. Time and the verb. A guide to tense and aspect. Oxford: Oxford University Press. Comrie, Bernard. 1985. Aspect. An introduction to the study of verbal aspect and related problems. (4th ed. [1st ed. 1976]). Cambridge: Cambridge University Press. Dahl, Östen. 1981. "On the definition of the telic-atelic (boundednonbounded) distinction". Syntax and semantics Vol.14. Tense and aspect eds. Philip J. Tedeschi & Annie Zaenen, 79-90. New York: Academic Press. Garey, Howard B. 1957. "Verbal aspect in French". Language 33.91-110. Givón, Talmy. 1985. "Iconicity, isomorphism and non-arbitrary coding in syntax". Iconicity in syntax ed. John Haiman, 187-219. Amsterdam: John Benjamins. Klein, Wolfgang. 1994. Time in language. London: Routledge. Kochańska, Agata. 1996. "Temporal meanings of spatial prepositions in Polish: the case of przez and w". The construal of space in language and thought eds. Martin Pütz & René Dirven, 491-508. Berlin: Mouton de Gruyter. Langacker, Ronald W. 1987a. Foundations of cognitive grammar. Vol.1. Theoretical prerequisites. Stanford: Stanford University Press. 1987b. "Nouns and verbs". Language 63.3-94. 1988a. "An overview of cognitive grammar". Rudzka-Ostyn 1988. 3-48. 1988b. "A view of linguistic semantics". Rudzka-Ostyn 1988. 4990.
166
AGATA KOCHAŃSKA
1991. Foundations of cognitive grammar. Vol.2. Descriptive application. Stanford: Stanford University Press. 1995. "Viewing in cognition and grammar". Alternative linguistics. Descriptive and theoretical modes ed. Philip W. Davis, 153-212. Amsterdam: John Benjamins. Laskowski, Roman. 1984. "Kategorie morfologiczne języka polskiego charakterystyka funkcjonalna". Gramatyka współczesnego języka polskiego. Morfologia eds. Renata Grzegorczykowa et al., 121-169. Warsaw: Państwowe Wydawnictwo Naukowe. Rudzka-Ostyn, Brygida, ed. 1988. Topics in cognitive linguistics. Amsterdam: John Benjamins. Talmy, Leonard. 1988. "The relation of grammar to cognition". RudzkaOstyn 1988. 165-205. Vendler, Zeno. 1967. Linguistics and philosophy. Ithaca, N.Y.: Cornell University Press. Verkuyl, Henk J. 1993. A theory of aspectuality. The interaction between temporal and atemporal structure. Cambridge: Cambridge University Press.
How I got myself arrested Underspecificity in Grammatical Blends as a Source for Constructional Ambiguity Nili Mandelblit École Polytechnique CREA Paris, France 1.
Gilles Fauconnier University of California San Diego, U.S.A.
Grammatical blending in the use of syntactic constructions * In Fauconnier & Turner (1996) and Mandelblit (1997), we develop an analysis of sentence processing as a case of conceptual and linguistic blending (what we refer to as grammatical blending or grammatical integration): sentence generation involves the blending of a conceived event with a syntactic construction; sentence interpretation starts with a reconstruction of the blending configuration. The syntactic constructions serve as integrating frames, allowing the linguistic representation of complex events as instances of simple clause constructions. The linguistic blending operation marks a conceptual integration operation, via which a rich complex event representation is integrated into a compact event schema. The central idea is that simple sentence structures can be used to linguis tically express a complex sequence of events by blending together elements from the event sequence with the simple sentence structure (the 'integrating syntactic construction'). By 'blending', we refer to a general cognitive opera tion studied by Fauconnier and Turner (1994, 1998). This operation includes a cross-space mapping between two input mental spaces, and selective pro jection from the two inputs into a blended space, which may then acquire emergent structure through completion and elaboration. This cognitive opera tion applies to a wide range of cognitive phenomena in language, science, and the arts.1 Grammatical blending, like any blending operation, is possible if a correspondence (cross-space mapping) is found between two conceptual structures. In the case of grammatical blending, the correspondence is between the structure of a novel conceived event (the one that is communicated) and the semantics of an integrating syntactic construction. An important function of grammar, in this view, is to mark various blending configurations of events in the world and chosen integrating con structions, thereby providing cues to the hearer in reconstructing (i.e., interpreting) linguistic blends. The analysis of constructions and blending operations in various languages (English, French, and Hebrew) has shown that similar
168
NILI MANDELBLIT & GILLES FAUCONNIER
blending configurations underlie these superficially different grammatical systems. These languages differ mainly in the formal tools that have evolved to mark conceptual blending configurations. In this paper, we focus on an in herent property of grammatical blending: the formal underspecification of the blending configuration which leads to multiple interpretations associated with a single construction. Fauconnier and Turner (1996) analyze the blending operations underlying English Caused-Motion sentences (studied by Goldberg, 1995). Goldberg, who follows the view of Construction Grammar (Fillmore, Kay & O'Connor 1988), argues for an independent existence of syntactic constructions in English pairings of syntactic forms with conceptual schemas. One syntactic construction that Goldberg analyzes is the English Caused-Motion construction: its form is [NP V NP PP] ( = SUB V OBJ OBL ), and its associated semantics is that of 'caused motion' (i.e., 'X causes Y to move Z'). Examples of this construction include: (1) The audience laughed the poor guy out of the room (2) Rachel sneezed the napkin off the table (3) She trotted the horse into the stable In the English lexicon, there are some verbs, like throw, which already specify in their semantics a caused-motion event (and are therefore referred to as 'lexical caused-motion verbs'). These verbs occur prototypically with the syntactic form [NP V NP PP], as in example (4): (4) Jack threw the ball into the basket Example (4) actually integrates a causal sequence of events: Jack acts on the ball and thereby causes the ball to move. The verb throw specifies both Jack's action and the ball's motion, and the fact that they are causally related. But, in English, the syntactic form [NP V NP PP] can also be used with verbs which themselves do not specify caused motion (as in examples (l)-(3)). In each of the examples (l)-(4), a novel conceived caused-motion se quence of events is blended with the caused-motion construction: partial as pects from the conceived event are conceptually and linguistically mapped onto the integrating caused-motion conceptual schema and the associated syntactic form [NP V NP PP]. In example (4), a single lexical item {throw) integrates in its semantics the whole causal sequence of predicates. In examples (l)-(3),
HOW I GOT MYSELF ARRESTED
169
Figure 1. The blending operation underlying the generation of the sentence Rachel sneezed the napkin off the table
the verb specifies only one predicate within the caused-motion event sequence: in examples (1) and (2), the verbs laugh and sneeze specify the agent's action only (while the motion of the grammatical object is left unspecified). In (3), the verb trot specifies the motion of the moving object (the horse), while the agent's (causing) action is left implicit. Figures 1 and 2 describe the blending operations underlying the generation of sentences (2) and (3), respectively. One input structure (INPUT 1 in the figures) is the novel conceived causal sequence of events (intended to be communicated). The second input structure (INPUT 2) is a schematic representation of the caused-motion construction. The blending between the two input structures is based on perceived structural similarity (i.e., the speak-
170
NILI MANDELBLIT & GILLES FAUCONNIER
Figure 2. The blending operation underlying the generation of the sentence She trotted the horse into the stable
er, we assume, perceives the correlation between the structure of the novel conceived event and the semantics of the caused-motion syntactic construction). The arrows in Figures 1 and 2 describe the mapping between the two input structures, and the blended space (at the bottom of the figure) depicts the actual sentence communicated in the language. The blend inherits its syntactic form from INPUT 2 (the integrating construction), and the lexical items from INPUT 1 (these lexical items schematically represent partial aspects of the conceived event). The blend also inherits the integrated conceptual structure of the integrating syntactic construction (this is not shown explicitly in the figures). For the detailed structure of these blending figures, see Fauconnier & Turner (1996), and Mandelblit (1997, chapter 2).
HOW I GOT MYSELF ARRESTED
171
Note that nothing in English grammar marks which predicate has been mapped onto the verbal slot of the integrating Caused-Motion syntactic construction (whether it is the causing predicate, as in Figure 1, or the effected predicate, as in Figure 2). It is up to the hearer to reconstruct the blending configuration, and elaborate its semantics to fit a possible event in the world. This grammatical underspecification of the blending operation leads to what is sometimes referred to as 'constructional ambiguity', or at other times 'lexical ambiguity' (if the underspecification of the blend is assigned to a particular lexical item). For example, a verb such as laugh is defined in some dictionaries as having multiple senses: one sense is the action of laughing; another sense is 'move by laughter' (Webster's Ninth New Collegiate Dictionary, 1991). Following Goldberg, we take the verb laugh to have only one sense, but this sense can be used to refer to complex event structures (e.g., causation by laughter) through a process of grammatical blending (as described in Figure 1). Since English grammar underspecifies the blending configuration, the verb laugh itself may refer to simple and complex event structures with no explicit marking. It is for this reason that the lexical item laugh itself is viewed as ambiguous. In the next section, we will discuss the Hebrew morphological verbal system of binyanim. Mandelblit (1997) shows that each morphological verbal pattern (binyan) in Hebrew marks a different blending configuration of predicates (and participants) from a conceived event onto an integrating syntactic construction. The conceptual blending operation is the same as the one observed in the English caused-motion sentences, but the elaborate Hebrew morphology explicitly marks the blending configuration, in a way which prevents the type of 'ambiguity' associated with English verbs (as discussed above). Other 'ambiguities' exist however in Hebrew verbs due to underspecification of other aspects of the blending process. 2.
Blending operations in the Hebrew morphological binyanim system. Hebrew verbs all consist of (a) a skeleton of consonants (the 'root'), slotted into (b) some vowel pattern, or prefix+vowel pattern {¡an - sg., binyanim - pl.). The consonants carry the 'core meaning' of the verb (or a certain 'semantic field', cf. Junger 1987:12), and the patterns provide the actual form of the word. There are seven major binyanim in Hebrew (the capital 's stand for the root consonants to be inserted): CaCaC, CuCaC, CiCeC, niCCaC, hiCCiC, huCCaC, and hitCaCeC. Consider, for example, the root r. ?.h. : this root occurs in Modern Hebrew in five principal binyanim (5a-e). The predicates in (5a-e) share the same consonantal root, and hence the
172
NILI MANDELBLIT & GILLES FAUCONNIER
same semantic field, but they differ in their actual semantics as a function of the changing binyan. (5) a. b. d. e.
CaCaC + r.?.h niCCaC + r. ?.h hiCCiC + r.?.h huCCaC + r. ?.h hitCaCeC + r. ?.h
[ra?a] [nir?a] [her?a] [hur?a\ [hitra?a]
"to see" "to be seen" or "to seem" "to show" "to be shown" "to see each other"
In traditional accounts, each binyan is associated with one or more grammatical functions, suchas 'causative', 'passive', or 'reflexive'. Mandelblit (1997) suggests that each binyan marks a blending schema between a representation of a conceived event and a chosen integrating syntactic construction. Each morphological binyan is therefore a construction: a pairing between a morphological pattern and a schema of blending operation. Consider, for example, the use of the causative binyan hiCCiC (termed hif'il), exemplified in sentence (6): (6) hamefaked heric (r.u.c-hif'il) et haxayalim the-commander RUN2-hif il past ACC the-soldiers "The commander made the soldiers run" Sentence (6) integrates a whole causal sequence of events: there is the causing event of the commander acting on the soldiers, and the resulting (or effected) event of the soldiers running. The causal sequence of events is expressed linguistically through a single syntactic structure: the Hebrew Transitive construction [NP V (et) NP].3 The basic assumption (following the Construction Grammar view) is that for Hebrew speakers, the syntactic pattern [NP V (et) NP] forms a construction and is associated with an independent semantic schema of an archetypal 'transitive' event: an agent who volitionally acts on and affects a patient (cf. for example, Givón (1984), Langacker (1991)).4 Figure 3 describes the blending operation underlying the generation of sentence (6).5 Note the similarity between the blending operation underlying the Hebrew hif'il sentence (Figure 3), and the English caused-motion sentences (Figures 1-2). As in the case of English caused-motion sentences (l)-(3), only one predicate from the conceived causal sequence of events is mapped onto the integrating syntactic construction in the Hebrew sentence (the effected motion predicate of running). In the blend (the generated sentence), the root of the
HOW I GOT MYSELF ARRESTED
173
Figure 3. The blending operation underlying the generation of a basic transitive hif il sentence in Hebrew (example 6)
main verb (r.u.c) depicts the effected motion predicate, and the morphological pattern (hif il - hiCCiC) marks the blending configuration (i.e., that the verbal root r.u.c, run, denotes an effected predicate in a causal macro-event). As shown in Mandelblit (1997), irrespective of the type of conceived causal event (and thus irrespective of the chosen integrating syntactic construction), the hif'il morphology marks the grammatical integration of a causal sequence of events, where what is mapped onto the verbal slot of the integrating construction is the effected predicate in the causal sequence.6 Example (7) is another Hebrew hif'il sentence, but this time the causal sequence is one of 'transfer' or 'enablement of possession', and thus the Hebrew Transfer construction [NP V et NP le-NP] is used as an integrating
174
NILI MANDELBLIT & GILLES FAUCONNIER
syntactic construction (this is the same syntactic pattern used with 'transfer' verbs such as give or send; evidence for the independent existence of the construction in Hebrew is discussed in Mandelblit, 1997, ch. 4). The verbal morphology (hif'il) marks again that what is mapped onto the verbal slot of the integrating syntactic construction (and expressed by the verbal root in the blend) is the effected predicate in the causal (transfer) event (i.e. the predicate inherit from the causal sequence [[Danny act] enable [Ruth possess(inherit) house]]). (7) dani horish (y.r.sh-hif'il) et habayit leruti Danny INHERIT-hif'il past ACC the-house to-Ruth "Danny bequeathed the house to Ruth" The analysis of all other six principal binyanim (verbal patterns) in Hebrew shows that each one is associated with a single generic blending configuration. Figure 4 summarizes the blending schemas associated with each binyan (for the complete analysis, see Mandelblit 1997, chapters 4-7). The schemas complement each other to form a unified system. The different functions associated with Hebrew binyanim (described in Indo-European based grammatical terminology as 'active', 'passive', 'reflexive') are produced directly by the different schemes. The Hebrew verbal morphology explicitly marks the mapping of predicates from a conceived event onto the integrating syntactic construction (the dotted arrows in the blending diagrams in Figure 4). However, since only partial aspects from the conceived event can be mapped onto a single syntactic construction, the blending configuration is still underspecified. This underspecification leads to particular types of 'ambiguities' associated with the Hebrew binyanim system. Consider the four 'intransitive' binyanim in Hebrew: nif'al, pu'al, buf'ai, and hitpa'el, principally associated with passive, middle, and reflexive functions. These binyanim are analyzed as a case of blending of a conceived event onto the Intransitive syntactic construction [NP V]. The Intransitive construction is intuitively associatcd with a generic semantic schema of [Theme Pred] — an entity (a Theme) undergoing some activity or change of state (Pred). This schema is extracted from the semantics of intransitive sentences in the language (see, for example, Kemmer 1993).7 As depicted in Figure 4, the blending schema associated with all these four binyanim involves the mapping of the principal affected entity (i.e., the agent or experiencer in the effected sub-event) onto the first NP (subject) slot of the intransitive construe-
HOW I GOT MYSELF ARRESTED
175
Figure 4. Λ blending characterization of the Hebrew binyanim system
tion. The four binyanim differ as to which predicates are mapped onto the verbal slot in the integrating construction. Now consider a 'grammatically ambiguous' binyan, such as nif'al (niCCaC), which is associated in traditional accounts with two principal grammatical functions: passive and middle.8 The passive and middle inter-
176
NILI MANDELBLIT & GILLES FAUCONNIER
pretations associated with nif'al are exemplified in (8b) and (9b) (in (8a) and (9a) are the corresponding active pa'al sentences): (8) a. raxel shavra(sh.b.r-pa'al) Rachel BREAK-pa'alpast "Rachel broke the vase"
et hakad ACC the-vase
b. hakad nishbar(sh.b.r-nif'al) the-vase BREAK-nif'alpast "The vase was broken" (by Rachel) or: "The vase broke" (9) a. raxel r(..r-pa'al) et habayit Rachel SELL-pa'alpast ACC the-house "Rachel sold the house" b. habayit nimkar(m. k. r-nif'al) the-house SELL-nif alpast "The house was sold" (by Rachel/to Dani) or: "The house sold"
(active)
(passive) (middle)
(active)
(passive) (middle)
Each of the active sentences (8a), (9a) integrates a whole causal sequence of events (an integration of a physical cause and effect in the verb break in (8), and a transfer causation of a patient to a recipient in the verb sell in (9); this is similar to the integration of a whole caused-motion sequence of predicates in the English verb throw, sentence (4)). The passive and middle sentences (8b) and (9b) integrate the same causal event into the Intransitive construction [NP V], and differ from the active sentences in the mapping of participants onto the Subject NP slot of the integrating construction: in the active sentences, it is the causal force that is mapped onto the subject slot, while in the passive/middle sentences it is the affected entity. Figure 5 describes the blending operation underlying the passive/middle nif'al sentence (8)b. The blending configuration underlying the passive and middle readings of thc nif'al sentence (8b) is exactly the same (Figure 5). The only difference between the passive and middle interpretation is in the implicit profiling of the causal force. In the passive reading, an external causal agent is understood to exist and be an inherent part of the event depicted by the verb (this implicit understanding of a necessary existence of an agent external to the Theme is often associated with passive constructions in the literature). The middle reading, in contrast, 'downplays' the role of an external causal agent, and pro-
HOW I GOT MYSELF ARRESTED
177
Figure 5. The blending operation underlying the generation of a nif'al sentence (example 8)
motes a construal of the event as something that 'just happened', probably due to properties (forces) internal to the theme (as Bolozky 1996:xii, notes, "typically the change-of-state [in middle-inchoative nif'al] happens on its own, i.e., involuntarily"). In the middle reading the reader assumes that 'the vase broke' (Example (8b)) or 'the house sold' (Example (9b)) because of some 'natural' (inherent) properties (e.g., the vase is fragile). Note however that the speaker and hearer in (8b) and (9b) arc aware at the same time of the role of an external causal force in bringing about the event indicated by the verbal root: i.e., the house would not sell and the vase would not break without an agent selling or breaking it. The very use of the stems (roots) BREAK and SELL (which are associated with transitive-causal semantics) promotes a recon struction of a causal event as part of the interpretation of the sentence. The implicit existence of a causal force in the conceptual representation of both the
178
NILI MANDELBLIT & GILLES FAUCONNIER
passive and the middle readings of nif'al is represented (in Figure 5) in the 'presence' of a causal force (X) in Input 1, a causal force which is not mapped onto the integrating construction (and therefore is not linguistically expressed). According to the blending analysis of nif'al, the grammatical passive middle 'ambiguity' associated with nif'al is an outcome of the under specification of the grammatical blending schema (underspecification of the causal force). This underspecification leads to (at least) two types of pragmatic elaboration on the blending configuration associated with the binyan: the passive and middle elaborations (one where the prominent causal force is associated with an external agent, and another where the prominent causal force is associated with internal properties of the theme). In Indo-European languages, these two pragmatic elaborations of the nif'al schema (the middle and passive readings) are marked distinctively in the grammar (note the different grammatical constructions used in the English translation of the passive and middle nif'al sentences in (8b)-(9b). The distinction in grammatical forms led Hebrew scholars to suggest that the Hebrew nif'al is grammatically ambiguous. But in fact, we suggest, within the Hebrew system of binyanim, the nif'al form is not grammatically ambiguous. It provides (like all other binyanim), a single set of grammatical cues (a 'blending scheme'), which can then be pragmatically elaborated in different ways in the interpretation process (i.e., in the reconstruction of the communicated event). From the grammatical blending point of view, the two alternative reconstructions of nif'al (as either passive or middle) are just like the various possible reconstructions of the sense of the main verb in English caused-motion sentences (where the verb may be linked for example to either the 'causing' or the 'effected' predicate in the causal event sequence).9 Both cases of 'disambiguation' require the conceptual reconstruction of a possible event, beyond the cues explicitly provided by the grammatical form. These two examples of 'disambiguation' differ only at the level at which the ambiguity is formally resolved: in the English caused-motion sentences, the ambiguity of the verb is resolved at the level of mapping reconstruction (i.e., which aspect of the communicated event is mapped onto the integrating construction); in the Hebrew nif'al sentences, it is resolved at the level of pragmatic elaboration (beyond the reconstruction of the mapping configuration).
HOW I GOT MYSELF ARRESTED
3.
179
Blending and underspecification in French causative-passive constructions The English caused-motion examples and the Hebrew binyanim examples reveal a very general scheme. The language possesses basic constructions (intransitive, transitive, or transfer), and blends them conceptually and formal ly with initial conceptual representations of conceived events (often having the general form of one event causing another). Since blending depends on a crossspace mapping between inputs, and since the conceptual structure of the perceived events to be blended is typically richer (has more predicates and arguments) than the target integrating construction, there are a priori a large number of possible blends, depending on which cross-space mappings are selected. A language must make it possible for speakers and listeners to construct blends systematically and coherently. Call this goal proper blending. A variety of means for achieving proper blending is found across languages. In Hebrew, the linguistic solution to proper blending is primarily morpholo gical; it is general and elegant. Verb morphology is itself a formal blend of a root and a binyan. Each of the seven binyanim signals a certain mapping configuration (a 'blending scheme') between the construction and the con ceptual causal structure. A conceptual (and formal) blend is constructed on the basis of that pattern, and additional variation of meaning follows from pragmatic underspecification. Different meanings described in Indo-European based grammatical terminology as 'causative', 'active', 'passive', 'middle', 'reflexive', are produced directly by the blending schemes. One does not have to consider some of the forms to be more basic than others. Now consider the solution to proper blending adopted by Romance languages such as French and Italian for the expression of causative and related constructions. It is argued in Fauconnier and Turner (1996) that causatives like (10)-(12) are the product of three different blends: (10) Pierre fait manger Paul NP V V NP "Pierre makes Paul eat" [Paul is the agent of manger] (11) Pierre fait envoyer le paquet (par Paul) NP V V NP "Pierre has the parcel sent (by Paul)" [the parcel is the object of envoyer]
180
NILI MANDELBLIT & GILLES FAUCONNIER
(12) Pierre fait manger la soupe à Paul NP V V NP à NP "Pierre makes Paul eat the soup" [Paul is the agent of manger] In example (10), the conceptual causal structure {[X do] CAUSE [Y pred]} is mapped onto the Transitive construction 'Agent act Patient' (syntax NP V NP). Example (12) is a case of blending with the Transfer construction 'Agent transfer Patient to Recipient' (syntax NP V NP à NP). The agent or experiencer of the effected event is mapped onto the Recipient role in the Transfer construction, and therefore shows up as an indirect object gramma tically. The two blends in question (Transfer and Transitive) are exactly the same as the ones at work in the hif'il construction in Hebrew (the hif'il transitive and transfer examples (6)-(7)). However, the linguistic solution to proper blending is different. Instead of reflecting the mapping scheme morphologically on the verb (as in Hebrew), French projects the underspecified causing predicate (faire) onto the blend, creating the double verb syntax in both cases ([NP faire V NP], example 10) and ([NP faire V NP à NP], example 12). Example (11) is an illustration of a third type of blend - called the Optional Transfer blend — which takes the Optional Transfer construction found in French for verbs like vendre. Syntactically, this construction has a subject, a verb, and optionally a direct object, and optionally an indirect object: [NP V (NP) (à NP)].10 But in the optional transfer blend (example (11)), there is no explicit agent/experiencer in the effected event. The blend is entirely focused on the effected predicate (envoyer), and on what we shall call the Theme, when there is one - in most cases an affected patient (in (11), le paquet). The analysis rests on the availability of a conceptual sub-structure of the form [Pred Theme], also found in unaccusatives and passives (the same schema associated with Intransitive constructions, as suggested for Hebrew, section 2). The schema is of an entity (the Theme) undergoing the event corresponding to Pred. In our example, this is the action of sending (envoyer), and the parcel is the Theme. The input conceptual structure to the blend in (11) thus has the form {[X do] CAUSE [pred Y]}. 'X' is mapped to 'Agent' (the subject NP slot) in the optional transfer construction, and 'Y' is mapped to 'Patient' (the object NP slot in the construction). As in examples (10) and (12), the causing predicate is projected as faire onto the blend, and the effected predicate (envoyer) is projected as the principal verb in the blend (Figure 6).
HOW I GOT MYSELF ARRESTED
INPUT 2
181
INPUT 1
French - Optional Transfer Const. [ NP V (NP') (à NP") ]
Figure 6. The French optional transfer blend (example 11)
Note that the agent of the effected sub-event (Paul in example (11)) is left unspecified in the blend. As in passives and some lexical causatives, it can be expressed externally by means of an instrumental by-phrase (in French 'par NP'). The optional transfer blend thus shows a mapping which we have not seen in Hebrew. Now consider the two meanings the Optional Transfer blend can take when its surface direct or indirect object is a reflexive pronoun, anaphoric to the subject (example 13):
182
NILI MANDELBLIT & GILLES FAUCONNIER
(13) Je me suis fait arrêter NP NP (AUX) V V I myself aux -'be' do arrest. "I got myself arrested" Example (13) has (at least) two equally possible interpretations. One is 'I caused myself to be arrested', implying that I deliberately did something that would (and that did) result in my being arrested. The other interpretation is essentially that of a passive 'I was arrested', and implies no action on my part or any intention to be arrested (note that the same double interpretation applies for the English parallel construction I got myself arrested). The two meanings appear quite distinct (a causative and a passive). Why does the causative form also have a passive interpretation? Under the usual generative or relational derivations of examples like (13), from a full-blown causative underlying form, there is no explanation for the ambiguity. It has to be considered a poorly motivated co-optation of the causative construction for other purposes. This is all the more surprising given that the language independently has a shorter and general passive construction. But if we take into account the results for Hebrew (section 2), we can offer a general explanation for this kind of superficially surprising data. Recall that blending operates only on selective projection from inputs, and that the causal conceptualization is pragmatically underspecified in systematic ways (see the English caused-motion examples, and the nif'al Hebrew examples). In the nif'al blend, only participants from the effected event were projected, leaving the causal force (in the causing event) implicit. This created a pragmatic ambiguity: if the causal force was attributed to an external agent, the result was a 'passive' interpretation for the nif'al blend. If on the other hand, the causal force was attributed primarily to internal ('inherent') forces within the affected patient, the interpretation was what we call 'middle' in Indo-European based terminology (examples (8b), (9b)). Construal is relevant here: we can talk of the vase breaking, and foreground the inherent properties of the vase that favor its breaking ('middle'), while still remaining aware of external forces that contributed to the breaking, or we can foreground the external forces (in English: The vase was broken), and get a 'passive' interpretation. In Hebrew, the nif'al blend can be used for both cases, because of the pragmatic underspecification of the selective projection into the blend (i.e., in nif'al, the causal force is not mapped onto the integrating construction). Returning now to example (13), we see that the case in French of the optional transfer blend is quite similar. The input conceptualization only reflects that there was a caus-
183
HOW I GOT MYSELF ARRESTED
INPUT 2 French - Optional Transfer Const. [NP V (NP') (à NP")]
INPUT 1
Figure 7. The French optional transfer blend (example 13)
al link (possibly quite weak) between 'my doing something' and 'my being arrested' (Figure 7), not that I did something in order to be arrested or that I deliberately engineered my arrest. In interpreting such causative blends, the hearer can fill in the implicit causing event and the force-dynamic (causal) relation to the effected event ('my being arrested') in different ways. This is confirmed by other data, which we now consider. First, let's look once more at non-reflexive cases (examples (14)-(16)): (14) La viande fait saliver Max "The meat makes Max salivate"
[transitive blend]
184
NILI MANDELBLIT & GILLES FAUCONNIER
(15) Les pyramides ont fait comprendre l'Egypte à Max [transfer blend] "The pyramids made Max understand Egypt" (16) Diana a fait pleurer Suzanne "Diana made Suzanne cry"
[transitive blend]
In (14), the meat is not doing anything special: we understand that it is its perception by Max which is the causing event. In (15), more than percep tion may be involved: for example an elaborate reconstruction by Max of the social conditions under which such edifices might have been built. Again, the pyramids are doing nothing, and the understanding requires elaborate schema completion. For the same reason, (16) could mean that Diana deliberately caused Suzanne to cry, but it could equally well mean that Suzanne cried because of Diana's death. The general point here is the underspecification of the causing event, already noted for the Hebrew nif'al blends. In the case of the (non-reflexive) optional transfer blend, where the agent of the causing event and the patient of the effected event are different, there is undoubtedly a tendency to choose a strong causative interpretation, as in (17). Example (17) is interpreted out of context as a deliberate action on the part of Pierre to have Jean arrested. It can also mean, however, that Pierre did something involuntarily which permitted Jean's arrest. (17) Pierre a fait arrêter Jean "Pierre caused Jean to be arrested" If the causing agent is not a separate individual, but a property of the affected patient (example (18)), then the blend can be understood with some thing like the inherent property interpretation of middle nif'al in Hebrew: (18) Son insolence a fait arrêter Max "Max's insolence caused him to be arrested" The availability of such interpretations suggests that in the case of a reflexive blend (as in (13)), the causative and passive interpretations are available on the basis of pragmatic underspecification. But there is more to this story. As Postal (1992) notes, there is distributional evidence for a syntactic (and not just pragmatic) ambiguity between the causa-
HOW I GOT MYSELF ARRESTED
INPUT 2 French - Optional Transfer Const. [NP V(NP')(à NP")]
185
INPUT 1
Figure 8. An alternative analysis of the French reflexive optional transfer blend. There is now only one instance of an agent/experiencer explicit in the input conceptual structure. The arrow can also go from Y to NP" (instead of NP') depending on the relation of Y to Pred2.
tive and passive interpretations of sentences like (13). The causative but not the passive allows a reinforcing pronoun with même: (19) Je me la ferai décrire à moi-même [causative interpretation only] "I will have her described to myself" This test distinguishes more generally between true reflexive pronouns and intrinsic reflexives: Il se la décrira (à lui-même) ("he will describe her to
186
NILI MANDELBLIT & GILLES FAUCONNIER
himself"), Il se tape un éclair (*à lui-même) ("he eats an eclair"). Postal views these facts as evidence for two distinct derivations of (13). Spang-Hanssen (1967) also opts for syntactic ambiguity, while Danell (1979) (along with others) calls the passive meaning a 'banal extension of the meaning of faire'. The blending account we have given (in terms of blending underspecification) is compatible with Postal's data, if his data reflects a pragmatic difference, namely if pro-même has the function of underscoring the agentivity of the coreferential subject. This is a plausible assumption, given other reinforcing uses of lui-même. There is also another interesting alternative for the blending configuration underlying example (13), an alternative which we mention here without further exploration. It is well known that syntactic change can be triggered by grammaticalization of pragmatics (Traugott, 1988). If the input conceptual structure leading to constructions like (13) does not have an explicit agent of the causing event, then the only available participant to be mapped onto the Agent position of the optional transfer construction will be the theme of the effected event. The Hebrew intransitive blends (huf'al, pu 'al and nif'al ) show that such association of the affected theme with a subject role in an integrating construction is natural in principle (in Intransitive constructions). In this alternative blending configuration (Figure 8), the theme is mapped onto both the subject and object positions of the integrating syntactic construction (resulting in a reflexive construction). This alternative amounts to a re-analysis of the French causative blend (a fourth blend in other words) with many of the same properties as the original optional transfer blend, but the 'meaning' of the subject NP in the blend becomes totally non-agentive (i.e., the passive interpretation). Within French, this type of blend (from one participant to both subject and object) may be underlying more generally unaccusative and passive reflexives, and perhaps intrinsic as well, but this proposal requires further exploration. Notes * This paper was written in part while Nili Mandelblit was a Chateaubriand Fellow at TALANA, Université Paris 7, Paris. She is grateful for financial support provided during that time by the French government Chateaubriand Fund. 1 See Coulson (1996), Liddell (1998), Moser & Hofstadter (ms.), Oakley (1995), Robert (in press), Turner (1996), and Zbikowski (ms.). For discussion of blending theory from a general cognitive science point of view, and as part of a cross-disciplinary conceptual change, see Fauconnier (1997), Mandelblit & Zachar (1998), Grush & Mandelblit (1998), and Turner (1996). 2 The glossing of Hebrew consonantal roots (e.g., r.u.c in example (6)) is done by using English stems (marked in SMALL CAPITALS). For example, the gloss RUN in (6) does not refer
HOW I GOT MYSELF ARRESTED
187
to any particular word in the English lexicon (such as 'run' v.i., 'run' v.t., 'run' n.) but rather to the stem (root) 'run', which represents (as best as possible) the common semantic component (schema) abstracted from all actual instances of this stem in the lexicon. 3 et is the Hebrew marker for definite direct objects; indefinite objects are not marked by a preposition. In the definition of the construction, et is marked as optional (enclosed in brackets). 4 This schematic event structure clearly represents only the most prototypical sense of the Basic Transitive construction. A full description of this grammatical construction involves a network of extensions to the prototypical sense as well as a list of idiomatic uses of the construction as discussed in Goldberg (1995), Lakoff (1987), and Langacker (1991). 5 Note that the blending analysis of sentence (6) (Figure 3) assumes independent existence (and semantic content) for consonantal roots in Hebrew as input entities to the blending operation. Positing roots as abstract entities (denoting a semantic stem) is compatible with many other analyses of Hebrew (going as far back as Gesenius, 1910, as well as contemporary word-based morphology accounts such as Aronoff, 1994). In actual language, there is no such entity as a consonantal root (since the roots never occur alone in the language, but only in combination with morphological patterns). We posit the existence of roots as schema abstractions in the sense of Langacker (1987): the root semantics is an abstraction over the semantics associated with instantiations of the root, and each instantiation is an elaboration of this schema. Polysemy, as well as the many idiosyncrasies in the Hebrew lexicon, sometimes pose a problem for abstracting the semantic schema of a root, and this led some Hebrew grammarians (e.g., Ornan (1971)) to suggest that neither the root nor the binyan alone can be taken to have an independent value. But as Berman (1978:91-95) convincingly argues, such accounts deny a clear regularity found in the system, both between words composed of the same root, and between words composed of the same binyan. 6 This characterization refers only to the group of hif'il verbs in the lexicon which are traditionally defined as causative. It does not reflect a more limited group of intransitive inchoative and reflexive hif'il verbs (for a discussion of this latter group and its link to the causative blending schema, see Mandelblit 1997, ch. 6.). 7 Kemmer (1993:73) characterizes root intransitives as denoting simple 'one-participant events', i.e., events of "one participant of which a state or action is predicated". 8 In addition nif'al is associated with several other less frequent grammatical functions, including inchoative (i.e. 'enter a state of...') counterparts for a few pa 'al verbs, reciprocal counterpart for few pa 'al, and (rarely, not productively) active-basic (Berman 1975, Glinert 1989, Bolozky 1996). 9 Consider, for example, the caused-motion sentence He ran his dog to the hospital. This sentence could mean either "he made the dog run to the hospital" or "he ran to the hospital with the dog". Each interpretation stands for a different mapping configuration. 10 As in Marie vend des livres à Paul ("Marie sells books to Paul").
References Aronoff, M. 1994. Morphology by itself. Cambridge: MIT Press. Bolozky, S. 1996. 501 Hebrew Verbs. New York: Barron's Educational Series.
188
NILI MANDELBLIT & GILLES FAUCONNIER
Berman, R. 1975. "The Morphological Realization of Syntactic Processes in the Binyanim System". Hebrew Computational Linguistics 9.25-39 [in Hebrew]. . 1978. Modem Hebrew Structure. Tel-Aviv: University Publishing Projects. Coulson, S. 1996. Semantic Leaps: The role offrame-shifting and conceptual blending in meaning construction. Ph.D. dissertation. University of California, San Diego. Danell, K.J. 1979. Remarques sur la construction dite causative. Stockholm: Almvist & Wiksell International. Fauconnier, G. 1997. Mappings in Thought and Language. Cambridge: Cambridge University Press. & M. Turner. 1994. Conceptual Projection and Middle Spaces. (Technical Report No. 9401). UCSD, Cognitive Science. . 1998. "Conceptual Integration Networks". Cognitive Science 22:2.133-188. . 1996. "Blending as a Central Process of Grammar". Conceptual Structure, Discourse, and Language ed. Goldberg, 113-129, Stanford: Center for the Study of Language and Information [distributed by Cambridge University Press]. Fillmore, C.J., Kay, P. & O'Connor. 1988. "Regularity and Idiomaticity in Grammatical Constructions: The Case of 'Let Alone'". Language 64: 501-538. Gesenius. 1910. Gesenius' Hebrew Grammar (A. E. Cowely, Trans.). Oxford: Clarendon Press. Givón, T. 1984. Syntax: a Functional-Typological Introduction, Vol 1. Amsterdam: John Benjamins. Glinert, L. 1989. The Grammar of Modern Hebrew. Cambridge: Cambridge University Press. Goldberg, A. 1995. Constructions: A Construction Grammar Approach to Argument Structure. Chicago: Chicago University Press. Grush, R. & N. Mandelblit. 1998. "Blending in Language, Conceptual Structure, and the Cerebral Cortex". The Roman Jakobson Centennial Symposium (Acta Linguistica Hafniensa, International Journal of Linguistics) eds. P. Brandt, F. Gregersen, M. Skov & F. Stjernfelt, Vol. 29.221-237. University of Copenhagen, Copenhagen: Hans Reitzel Forlag. Junger, J. 1987. Predicate Formation in the Verbal System of Modern Hebrew. Dordrecht, Holland: Foris Publications.
HOW I GOT MYSELF ARRESTED
189
Kemmer, S. 1993. The Middle Voice. Amsterdam: John Benjamins. Langacker, R. W. 1991. Concept, Image, and Symbol. New York: Mouton de Gruyter. Liddell, Scott. 1998. "Grounded Blends, Gestures, and Conceptual Shifts." Cognitive Linguistics 9:3.283-314. Mandelblit, N. 1997. Grammatical Blending: Creative and Schematic Aspects in Sentence Processing and Translation. Ph.D. Dissertation. University of California, San Diego. & O. Zachar. 1998. "The Notion of Dynamic Unit: Conceptual Developments in Cognitive Science". Cognitive Science 22:2.229-268. Moser, D. & D. Hofstadter. Errors: A Royal Road to the Mind. MS. The Center for Research on Concepts and Cognition, Indiana University. Oakley, T. 1995. Presence: the conceptual basis of rhetorical effect. Ph.D. dissertation, University of Maryland. Ornan, U. 1971. "Verb patterns and stems, inflections, declensions". Hauniversita 2.15-22. Postal, P. 1992. "Un passif sans morphologie spécifique". De la musique à la linguistique. Hommages à Nicolas Ruwet eds. L. Tasmowski & A. ZribiHertz, 475-484, Ghent: Communication and Cognition. Robert, A. 1998. "Blending in the interpretation of mathematical proofs". Conceptual Structure, Discourse, and Language, II ed. J.P. Koenig, 337350. Stanford: Center for the Study of Language and Information (CSLI). Spang-Hanssen, E. 1967. "Quelques périphrases passives du français moderne". Actes du 4ème congrès des romanistes Scandinaves dédiés à Holger Sten, 139-147, Copenhagen: Akademisk Forlag. Turner, M. 1996. "Conceptual Blending and Counterfactual Argument in the Social and Behavioral Sciences". Counterfactual Thought Experiments in World Politics. eds. P. Tetlock & A. Belkin, 291-295, Princeton: Princeton University Press. Traugott, E. 1988. "Pragmatic Strengthening and Grammaticalization". Proceedings of the Fourteenth Annual Meeting of the Berkeley Linguistics Society, 406-416. Berkeley: Berkeley Linguistics Society. Zbikowski, L. Conceptual blending and song. Unpublished MS, University of Chicago.
Konjunktiv II and Epistemic Modals in German A Division of Labour Tanja Mortelmans University of Antwerp, Belgium
1. 1.1
Subjectivity and 'grounding predications' * General characterization If one takes a look at the German literature on modality, one often finds that the topics of mood on the one hand and modal verbs on the other are treated separately, as if they did not have anything in common (for example in Drosdowski 1995, where pages 93-104 deal with modal verbs and pages 154-170 with mood). However, it can be shown that they share a number of properties, the most important of which — for our purposes at least — is the fact that they often evoke an element of the 'ground' as a reference point.1 The German moods Indikativ and Konjunktiv on the one hand and the modal verbs (especially when they are used epistemically) on the other can therefore be regarded as deictic expressions. The major difference, however, between German modal verbs and mood pertains to the degree of 'subjectivity' with which the ground is construed. Before we embark on the nature of this difference, let us first elucidate what Langacker means with 'subjectivity'. (1) a. girl b. the girl known to me this girl In the fully objective construal of (la) the noun 'girl' is said to profile a 'thing' without including the ground in its scope of predication. In (lb) the girl is identified relative to the ground 'me' who is put onstage and construed as part of the objective situation. Finally, in (1c) the ground figures as an implicit reference point with respect to the identication of the girl. So, in (lb) the ground belongs to the objective scene and must therefore be considered as a profiled part of the scope of predication, whereas in (1c) the ground is an unprofiled part of it. According to Langacker, (1c) exhibits the highest degree of subjectivity: despite its pivotal role in order to establish the meaning of 'this', the ground remains implicit, whereas in (lb) the ground functions as an explicit focus of attention.
192
TANJA MORTELMANS
These different types of subjectivity can also be identified in the historical evolution of the English modals, a development which can be characterized as a process of progressive 'subjectification'. Originally, the lexical antecedents of the present-day modals profiled the potency relation between the subject (trajector) of the verb and the landmark process, without including the ground in their scope of predication: magan, for example, indicated that its subject had the strength or physical ability to do something. A first type of 'subjectification' led to a reorientation of the potency relation from the trajector, i.e. the subject of the modal verb as an objective participant of the potency relation, to the ground, i.e. to a more subjective participant. The reoriented potency relation, however, remained in profile. In English, periphrastic modals like have (got) to, need to or be able to reflect this first type. In a second stage of subjectification the potency relation itself was "pulled offstage to become an unprofiled facet of the base" (Langacker 1991:270-271). The present-day English modals (can, may, must, will, shall and their so-called distal forms could, might, would and should) are therefore said to profile the process at which the potency is directed, and not the potency relation itself.2 Langacker stresses that, as far as conceptual content is concerned, periphrastic modals do not necessarily differ from the 'genuine' modals. However, only those grammatical elements that reflect the second subjectification stage qualify for the status of 'grounding predications', i.e. they effect the grounding (the epistemic location relative to the speaker, hearer and their spheres of knowledge) characteristic of finite clauses. One of the morphological characteristics of the English modals is the fact that they cannot appear as participles or infinitives: the existence of non-finite forms would be inconsistent with their grounding function. The German modal verbs, however, do not qualify as full 'grounding predications', as they allow non-finite forms. Langacker (1991) calls them 'periphrastic'; their construal reflects only the first subjectification stage whereby the modal relation of potency is profiled and not the process at which the potency is directed. So, in (2a) the necessity expressed by the infinitive müssen is profiled, whereas in (2b) the ability designated by gekonnt as a past participle stands in profile. (2) a. Er wird es sagen müssen "He will have to say it" b. Er hat das nicht gekonnt "He hasn't been able to do it"
KONJUNKTIV II AND EPISTEMIC MODALS IN GERMAN
193
The German moods Indikativ and Konjunktiv, on the other hand, seem to be perfect candidates for the status of 'grounding predications' in German, both formally (verb forms in the Indikativ or Konjunktiv are necessarily finite) and semantically (the semantic contribution of the German moods can generally be described relative to the speaker in terms of epistemic notions pertaining to (ir)reality). Increasing 'subjectification' is one of the processes accompanying grammaticalization (see Traugott 1995). It is not surprising, therefore, to find that 'grounding predications' tend to be highly grammaticalized. The English modals are a case in point, since they are generally regarded to possess a higher degree of grammaticalization than their German counterparts. In view of the correlation between subjectification and grammaticalization it can also be expected that an (ongoing) evolution from a periphrastic modal to a genuine 'grounding predication' will be reflected in an increased degree of gramma ticalization of the item involved. We will come back to this issue in a later stage. Finally, it must be remarked that Langacker applies the concept of (modal) 'grounding predications' both to deontically and epistemically used modal verbs. A deontic modal verb, i.e. a verb typically associated with the expression of obligation or permission, is said to be grounding if the locus of potency is a maximally subjectively construed speaker or someone associated with her. In this paper I will, however, concentrate on modal verbs with an epistemic value, that is on those verbs whose "sole import is to indicate the degree of likelihood of the designated process" (Langacker 1991:272). The English epistemic modals are considered to be 'grounding predications', since they typically express a speaker-oriented (and therefore grounded) assessment of the epistemic status of the process involved; they do not profile this assessment (the grounding relationship) but the verbal process (the grounded entity) itself. 1.2
'Grounding predications' in English and German As far as English is concerned, Langacker claims that tense and the modals ground a verbal process. For tense, the main distinction is that between present and past; for modality, between the absence or presence of a modal verb. The conceptual import of these 'grounding predications' is described relative to a number of cognitive models, the most fundamental of which is the so-called 'basic epistemic model' (Langacker 1991:242). Its main constituents are 'known reality', 'immediate reality' and 'irreality'. 'Known reality' refers to the set of situations (states of affairs) that are accepted by a particular conceptualizer (speaker) as being real, whether present or past. Part of 'known
194
TANJA MORTELMANS
reality' is 'immediate reality', i.e. reality at the latest stage of its evolution. Finally, 'irreality' comprises everything other than known reality. When tense is left unmarked, the process is said to be located in the epistemic domain of 'immediate reality'. A past tense-morpheme (termed 'distal' morpheme by Langacker), on the other hand, effects a location of the process in non-immediate reality. With respect to modal grounding, it is the absence of a modal verb that grounds the process in the epistemic realm of 'known reality', whereas a modal places it in 'irreality'. Let us now turn to German. As far as temporal grounding is concerned, we could claim that here again grounding is brought about by means of tensemorphemes. The main distinction is that between the {Indikativ) Präsens, the unmarked case, which locates the process in a region immediate to the speaker, and the (Indikativ) Präteritum, which locates it in a region temporally prior to the moment of speaking.3 The main distinction as far as modal grounding is concerned is not between the absence or presence of a modal verb, as is the case in English, but manifests itself as a distinction of mood between the unmarked Indikativ on the one hand and the Konjunktiv II on the other. As in English, however, we observe a second distinction as well, namely between the absence (again as the unmarked case) and the presence of an epistemically used modal verb. As will be shown, these two systems of epistemic location are complementary, so that we can speak of a division of labour between the Indikativ vs. Konjunktiv II distinction on the one hand and the absence vs. presence of an epistemic modal on the other. 2.
Mood as a 'grounding predication' in German There are three moods in German: Indikativ, Konjunktiv and Imperativ (the latter will be left undiscussed, as it falls under deontic modality). As far as the Konjunktiv is concerned, one has to distinguish between KonjunktivI (also called Konjunktiv Präsens, as it is built on a present stem) and KonjunktivII (also called Konjunktiv Präteritum, as it is built on a past stem). The main epistemic distinction relates to the opposition between the Indikativ — the unmarked case — and KonjunktivII. 2.1 Indikativ The German mood Indikativ "which, being the unmarked form, expresses that the state of affairs is, at least in the speaker's assessment, factual" (Diewald 1993:221, my translation) typically locates the process in the realm of known reality.
KONJUNKTIV II AND EPISTEMIC MODALS IN GERMAN
195
(3) a. Peter ist krank "Peter is ill" However, as in English, epistemic 'locations' can be overridden. Whereas in (3a) Peter's being ill is situated within (immediate) known reality, (3b) (3) b. Peter ist vielleicht krank "Perhaps Peter's ill" places it, at a higher level of conception, in the realm of potential reality, with vielleicht invoking a particular mental space in which the first conception is embedded (cf. Langacker 1991:247). A similar phenomenon occurs, when the Indikativ is used in a conditional context. In the following sentence (3c), the use of the conjunction wenn introducing the protasis indicates that the speaker sets up a hypothetical mental space. At the same time, the speaker makes clear — by means of the indicative form ist — that the truth of the protasis is consistent with her view of the world, without however asserting that the protasis actually holds. (3) Wenn jemand im Büro gewesen ist, dann hat er mich gehört "If somebody has been in the office, he (or she) has heard me" By contrast, when uttering (3d), in which the past subjunctive wäre figures in the protasis, the speaker expresses that she does not associate herself with the content of the protasis, i.e. that she is inclined to think that nobody was in the office. (3) d. Wenn jemand im Büro gewesen wäre, dann hätte er mich gehört "If somebody had been in the office, he (or she) would have heard me" In view of the fact that the indicative's default epistemic location in known reality can be overridden, it may be argued that it would be more appropriate to associate the Indikativ merely with the speaker's 'positive epistemic stance' (Fillmore 1990) towards the state of affairs. The concept of 'epistemic stance' (which Fillmore originally only applied to conditionals) refers to "the kind [of] commitment the speaker has to the proposition"
196
TANJA MORTELMANS
(Fillmore 1990:142). When the Indikativ is taken to express the speaker's positive epistemic stance, it merely signals that the speaker considers the proposition to be in line with her view of the world. This explains why the Indikativ can be used in a number of contexts that are clearly non-factual. However, when the Indikativ is used to mark the main verb in a simple declarative clause, the speaker's positive epistemic stance is asserted, which typically yields a location in known reality. 2.2 KonjunktivII The meaning of the KonjunktivII or past subjunctive is relatively difficult to describe, as it exhibits a great variety of subtle, context-driven meaning distinctions. However, there seems to be relative consensus among scholars (cf., for example, Diewald 1997, Durrell 1991, Thieroff 1992, Wilmanns 1906) to relate the meaning of the past subjunctive to 'irreality', as it "characterizes the statement as not in agreement with reality, be this because we perceive or know it to be contrary to reality, or because it is at least presently not actualized or not experienced as such. This mood is therefore a mood of irreality" (Wilmanns 1906:227, my translation). Drosdowski (1995), on the other hand, describes the KonjunktivII as a mood of both irreality and potentiality. At the same time, he stresses the counterfactual component by claiming that the use of the past subjunctive usually presupposes the negation of the corresponding indicative statement (Drosdowski 1995:157f.). So, according to Drosdowski, sentence (4a), in which hätte gekauft is a subjunctive form, 'presupposes' that he didn't buy the house, whereas such a presupposition is absent in (4b). Of course, the fact that (4a) presupposes a negated indicative statement whereas (4b) doesn't, can be linked to the tense form of the subjunctive: as a past perfect, hätte gekauft refers to the irreversible past, whereas würde loben (which can be considered as a periphrastic variant of a simple past subjunctive) relates to the present or even the future and therefore leaves the possibility of fulfillment open. (4) a. Er hätte das Haus gekauft "He would have bought the house" b. Sie würde deine Arbeit loben "She would praise your work" This semantic ambiguity between potentiality and irreality can be avoided if we assume that the KonjunktivII generally designates a speaker's 'negative
KONJUNKTIV II AND EPISTEMIC MODALS IN GERMAN
197
epistemic stance': when using the KonjunktivII, the speaker assumes that the proposition is not true. Let us take a look at some examples. In (5a) we are dealing with a counterfactual conditional ('irrealis'), whereas (5b) allows a 'potentialis'interpretation of both the protasis and the apodosis. However, we do not need to assume that the meaning of the KonjunktivII in these sentences is unstable, hovering between designating irrealis in (5a) and potentialis in (5b), if we consider it to express the speaker's negative epistemic stance. (5) a. Wenn er noch einmal gekommen wäre, hätte ich ihn rausgeschmissen "If he had come again, I would have kicked him out" b. Es wäre katastrophal, wenn er uns sehen könnte "It would be a disaster, if he could see us [perhaps he can see us]" In both cases, the speaker instructs the hearer to set up a hypothetical mental space which in (5a) is to be located in the past — due to the perfect tense forms (gekommen wäre) in the protasis. As the speaker expresses a negative epistemic stance towards this past, hence irreversible, event to show that it does not fit her view of the world, a counterfactual interpretation of the protasis is triggered. In (5b), the speaker sets up a hypothetical world ('wenn er uns sehen könnte') and again expresses a negative epistemic attitude towards it (by means of the past subjunctive könnte). The speaker has her doubts about the proposition's truth value and typically assumes that she is not being seen. However, since we are not dealing with perfect tense forms that would locate the hypothetical mental space in the past (wäre in the apodosis and könnte in the protasis are simple past subjunctives), a counterfactual state of affairs is not evoked. As a consequence, the possibility that the protasis holds ('that he can indeed see us'), cannot be excluded. We can therefore conclude that whether a conditional receives a counterfactual 'irrealis' or a hypothetical 'potentialis' interpretation is solely determined by context, and not by the meaning of the past subjunctive as such. In (6a), the past subjunctive hätte figures in a main clause containing fast 'almost'. The use of the KonjunktivII designating a negative epistemic stance in combination with perfect tense forms (hätte umgefahren) again evokes a counterfactual state of affairs that was about to occur, had reality in the end not taken another course. According to Lötscher (1991), the KonjunktivII in combination with fast expresses that "a development shortly before its expected
198
TANJA MORTELMANS
normal outcome is broken off and the end not reached" (Lötscher 1991:356, my translation). Interestingly, the Indikativ would be possible in this context as well (6b), simply expressing that a state of affairs almost arose. The implicature that the opposite outcome was imminent is not created, however. (6) a. Fast hätte ich ihn umgefahren "I very nearly knocked him over" b. Ich hatte ihn fast umgefahren "I nearly knocked him over" Highly problematic for standard accounts of the past subjunctive as marker of irreality is the fact that it can also be used for states of affairs that are undoubtedly real. Consider (7), where the KonjunktivII combined with a perfect tense form (hätten geschafft) evokes a counterfactual state of affairs ('We didn't manage') which, eventually, does not come about. The implicature it wasn 't easy arises precisely because of the evocation of the counterfactual state of affairs: it might well have ended the other way round (but it didn't). (7) So, das hätten wir geschafft "Well, we've managed, after all" [Implied: it wasn't easy] In the above examples the speaker's negative stance designated by the KonjunktivII seems to be motivated by more or less 'objective' circumstances that in each case pertain to the factuality of a state of affairs. In the following examples, however, the use of the past subjunctive can only be explained by referring to elements of the immediate speech context which can be considered as more subjective (since they pertain to the ground). Consider (8a), in which the past subjunctive hätte does not indicate a negative epistemic stance towards the proposition as such, but rather towards the possibility of fulfillment of the speaker's desire. (8) a. Ich hätte jetzt Lust auf ein Bier "I wouldn't mind a beer (if that's at all possible)" [but I know we are not supposed to drink here or I know that I've promised you to stop drinking or I know that there's no beer in the fridge, ...]
KONJUNKTIV II AND EPISTEMIC MODALS IN GERMAN
199
It must be remarked, however, that (8a) also allows a more positive interpretation as a standard indirect request for a beer, without necessarily indicating that the speaker knows that certain conditions are not fulfilled. In this interpretation, the negative stance meaning attributed to the past subjunctive is considerably weakened. It also seems to be very weak in examples (8b) to (8d): Such sentences are best interpreted as positive (albeit normally polite) recommendations or requests. Of course, at the origin of these constructions lies the expression of the speaker's negative stance towards the speech act she performs, but it seems as if the past subjunctive in utterances like these has evolved to a mere marker of a subjective attitude of the speaker that is not nessarily described in terms of politeness or deference. So, in (8d) ich würde empfehlen sounds rather like an intimidation instead of a genuine recommendation. (8) b. Ich würde Sie gern einmal sprechen "I would like to have a word with you" Ich hätte Sie gern einmal gesprochen4 "I would like to have a word with you" d. Der [...] unter Druck geratene Minister erklärte im ZDF, die Polizei werde hart [...] vorgehen. "Ich würde denen, die fried lich feiern wollen, empfehlen, nicht nach Oldenburg zu fahren. "5 "The minister, who had got under pressure, stated on ZDF that the police would take severe action. "I would recommend to all those who want to celebrate in peace, not to come to Olden burg."" 2.3. KonjunktivI The main function of the Konjunktivl or present subjunctive is to signal indirect speech, mainly in written language.6 By marking the information as coming from another source, as indirect information, the speaker disengages herself as to the epistemic status of what she is saying, as to the truth of the proposition. In other words, she can be said to express a 'neutral epistemic stance'. It must be remarked that the evidential category expressing reported evidence (the so-called quotative) is rather strongly entrenched in the German epistemic system: we do not only find the marking by means of the Konjunktivl·, the modal verbs sollen and wollen also have a quotative function, since they can be used by the speaker to mark information as hearsay. The
200
TANJA MORTELMANS
difference between the two modal verbs is related to the source of the information: with sollen, the source is neither the speaker nor the subject of the sentence, whereas wollen directs the attention to the subject of the sentence as the source of information. (9) a. Er soll das getan haben "He is said to have done it" b. Er will das getan haben "He claims to have done it" Interestingly, the KonjunktivII can also be used to mark indirect speech as in (10a): (10) a. Er sagt (e), er hätte Pech gehabt b. Er sagt(e), er habe Pech gehabt "He says (said) he has (had) been unlucky" It should be added that most German grammars claim that there is at best a stylistic difference between the use of KonjunktivII and Konjunktivl (in 10b) in indirect speech: the KonjunktivIIdoesnot signal the speaker's sceptical, i.e. epistemically negative attitude towards the content of what is said, but seems to be a less formal alternative to the KonjunktivI, so that it typically occurs in spoken language. With respect to the use of the KonjunktivII as a marker of indirect speech in written language, it is usually said to 'replace' the default use of the present subjunctive, when the latter formally coincides with the Indikativ and hence cannot mark the information as indirect. It would therefore seem appropriate to make a clear distinction between the use of the KonjunktivII as a marker of negative epistemic stance and its use in indirect speech, a position which is — amongst others — advocated by Thieroff (1992). 3.
The German epistemic modal verbs as periphrastic modals Langacker characterizes the meaning of the epistemically used modal verbs by means of the dynamic evolutionary model, at the basis of which lies the idea that a speaker can assess the possible factuality of future and unknown present or past states of affairs because she "conceives the world as being structured in a particular way, and reality as having a certain evolutionary momentum that constrains its future development" (Langacker 1991:547). In this model, present reality, in which the conceptualizer (identified as the
KONJUNKTIV II AND EPISTEMIC MODALS IN GERMAN
201
speaker) is situated, is distinguished from potential and projected reality. Given that reality is conceived of as continually evolving under the influence of a socalled 'evolutionary momentum' which prefers certain future courses and precludes others, we can distinguish between those future developments which, given the configuration of immediate reality, are not excluded and those that are. The former is referred to as potential reality (cf. Langacker 1991:277). The epistemic region of projected reality is situated within potential reality. It refers to those future events which can be anticipated (or 'projected') with considerable confidence. Thus, the difference between She may quit and She will quit must be related to the fact that the former situates the profiled process within (future) potential reality, whereas the latter places it in the realm of (future) projected reality. Langacker's model not only fits epistemic evaluations of future states of affairs but also allows for present-time interpretations of the modal verbs involved. The latter presuppose a more subjective construal of the notion evolutionary momentum, as "it is not reality, but the speaker's knowledge of present reality, whose continued evolution must be assessed and projected into the future" (Langacker 1991:275). Thus, when uttering the sentence she may be tired the speaker can be said to locate the process in the realm of present potential reality. This model can be applied to the German epistemic modals as well. It must be remembered, however, that the German modals — unlike their English counterparts — are periphrastic, so that they must be grounded themselves by mood and tense operators. The indicative or base forms of the German epis temic modals locate the verbal process in the epistemic regions of both potential and projected reality. They are associated with the expression of possibility, prediction and necessity, generally both for present (a) and future (b) states of affairs. As far as possibility is concerned, we find the verb können, locating the process in the realm of potential reality.7 (11) a. Sie kann in ihrem Zimmer sein "She may be in her room" b. Sie kann jeden Augenblick kommen "She may be here any moment now" Using the verb werden (the German counterpart of the English future auxiliary will) the speaker locates the process in the realm of projected reality.
202
TANJA MORTELMANS
(12) a. Er wird müde sein "He will be tired" b. Der Zug wird jeden Augenblick ankommen "The train will arrive any moment" The notion of 'projected reality' also figures in the characterization of epistemic müssen. For English must, Langacker notes that it is part of the present-time system and therefore cannot normally be used to locate processes in future projected reality, but rather in projected known reality, since "[...] must conveys immediacy, whereas will suggests a certain epistemic distance. [...] [M]ust [...] indicates that confirmation is regarded as virtually inevitable" (Langacker 1991:280). The same restriction seems to hold for German müssen. In a sentence like Er muß morgen hier sein ('He must be here tomorrow'), the modal is more readily interpreted as expressing deontic modality, whereas (13) is normally taken to be epistemic. (13) Er muß schon zu Hause sein "He must be at home already" Since the German modals are periphrastic, we must assume that the modal relationship is profiled and grounded by both tense and mood. We could say, then, that the form wird as Indikativ indicates the speaker's positive epistemic stance towards the epistemic location of the process in projected reality. The present tense morpheme in wird locates the epistemic relationship in present reality. In a way, however, it seems rather artificial to separate tense and mood markers from the import of the modal verb as such, in view of the fact that the main function of epistemic indicative werden is its epistemic location of a process in projected reality (in this respect it does not differ from the modal 'grounding predications' in English). I would therefore assume that the primary grounding of the modal verb is indeed effected by tense and mood, but that the combination of < indicative mood + present tense + epistemic modal verb > yields a periphrastic modal which, as a unit, grounds the verbal process in projected or potential reality. The plausibility of an analysis that attributes a certain amount of unit status to epistemic wird, muß and kann as complexes of three morphological categories is — amongst other things — corroborated by the fact that past tense-morphemes cannot normally ground the modal relationship in non-immediate (past) reality as reckoned from the ground anymore.8 In other words, one of the traditional 'grounding predications'
KONJUNKTIV II AND EPISTEMIC MODALS IN GERMAN
203
seems to have lost part of its typical grounding function when it is combined with an epistemic modal. According to Diewald (1993:227), German epistemic modals use the opposition between the present and the past indicative mainly to mark 'erlebte Rede' (quasi-direct discourse), as in the following examples, in which the epistemic act expressed by mußte in (14a) and konnte in (14b) takes the story's protagonist (and not the speaker/writer) as a vantage point. (14) a. Sie suchte schon zehn Minuten nach ihrem Schirm. Er war nicht da. Sie mußte ihn in der U-Bahn vergessen haben. "She had already been looking for her umbrella for ten minutes. It wasn't there. She had to have forgotten in the tube. " b. Sie erschrak so heftig, daß sie sich zurückfallen ließ. Jetzt wußte sie, wer das war! Das konnte nur die Frau sein, die sie einmal in Bernhards Begleitung gesehen hatte. Wieso kam sie hierher, und was wollte sie? "It gave her such a fright that she had to lie down again. Now she knew who that was. It could only be the woman whom she had once seen with Bernhard. Why did she come here and what did she want?" The following sentence provides an example of past epistemic mußte outside 'erlebte Rede'. However, here again it is not the speaker who reports on her own past epistemic assessment. The vantage point is shifted to the past (i.e. to the moment at which the stele was discovered) so that mußte expresses the assessment of the discoverers (who are not explicitly mentioned in the text). It is this deictic shift that the past-tense morpheme effects. (14) Einzig das Datum war in wohlvertrauten Mayaziffern vermerkt. Das allein jedoch machte den Fund zur Sensation: Auf der Stele von La Mojarra stand der älteste je gefundene Text Amerikas. Die Lettern mußten etwa 150 nach Christus in den Stein gehämmert worden sein, lange bevor die Mayas zur beherrschenden Macht in Mittelamerika aufstiegen. "Only the date was given in well-known Maya code. This alone made the find a sensation: the text on the stele of La Mojarra was the oldest one ever found in the Americas. The letters had
204
TANJA MORTELMANS
to have been hammered into the stone about 150 AD, that is long before the Mayas came into power in Central America." 4. The past subjunctive forms of the German epistemic modal verbs 4.1 Attenuation In her article on the grammaticalization of the German modal verbs, Diewald (1993) claims that the KonjunktivII has lost its typical meaning in combination with epistemically used modals. According to Diewald, the past subjunctive form of the epistemic modals merely expresses an attenuation, a weakening of the essentially positive epistemic assessment, which is to be regarded as an indication of an increased degree of grammaticalization of the modal. So, in (15a), the speaker expresses an epistemic judgement the force of which is weakened in comparison to its indicative counterpart (15b). The difference between epistemic muß vs. müßte can be described in more or less the same terms as the contrast between English epistemic must vs. should (cf. Coates 1983:64: "confident" vs. "less confident assumption"). One cannot say that, in (15a), the speaker dissociates herself from the epistemic evaluation she is making. The use of müßte still expresses an essentially positive assessment to which she is committed. The difference between (15a) and (15b) mainly pertains to the degree of likelihood, so that it seems rather problematic to describe the past subjunctive element in epistemic müßte in terms of 'negative epistemic stance'. (15) a. Er müßte inzwischen zu Hause sein "He should be home by now" b. Er muß inzwischen zu Hause sein "He must be home by now" As is pointed out by Coates (1983), core examples of epistemic must typically refer to the present, while examples of epistemic should typically refer to the future. The same distinction holds for German: epistemic müßte is most often found for evaluations of future states of affairs (15c). (15) Da wollen wir mal kräftig mitmischen. Ein Platz unter den ersten drei müßte drin sein. "We'll do our best to succeed. A ranking among the first three should be possible."
KONJUNKTIV II AND EPISTEMIC MODALS IN GERMAN
205
The same distinction between a stronger indicative judgement and an attenuated past subjunctive one holds for the pair kann vs. könnte. (16) a. Sie k auch einen anderen Weg genommen haben "She may have taken another road" b. Sie könnte auch einen anderen Weg genommen haben "She could/might have taken another road" Again, the utterer of (16b) cannot be said to express a negative epistemic stance towards the possibility designated by the kann-element in könnte. On the contrary, the speaker expresses a clearly positive epistemic evaluation. In these cases the combination of the KonjunktivII with the epistemically used modals seems to result in the Konjunktiv's giving up its typical meaning, merely expressing a diminished degree of likelihood. The past subjunctive can be said to have entered into a novel structure < epistemic modal verb — past subjunctive > which in its entirety — as a periphrastic modal — grounds the verbal process. This evolution reflects to a certain extent the highly subjective (i.e. speaker-oriented) use of the KonjunktivII illustrated by the examples under (8), in which the past subjunctive does not signal a negative epistemic stance anymore, but seems to have become a mere marker of a highly subjective attitude. An analysis of this kind is particularly relevant for the epistemic modal dürfte, which is generally associated with the expression of speaker-oriented probability. (17) a. Der Fall dürfte inzwischen vielen bekannt geworden sein "There's good reason to assume that the case has become known by many people" b. Er dürfte das gemerkt haben "I presume that he has noticed it" Morphologically, dürfte is the past subjunctive of the German modal verb dürfen, which in present day German is mainly used deontically to express permission. Epistemic dürfte is a somewhat isolated form, as the indicative of dürfen — unlike its English counterpart may — cannot be used for epistemic evaluations in German anymore. In other words, a contemporary link between an epistemically used indicative darf and an attenuated past subjunctive dürfte is absent. In view of the fact that a speaker by means of dürfte always
206
TANJA MORTELMANS
expresses a positive epistemic assessment of probability, it seems completely inappropriate to associate the past subjunctive element in this form with a negative epistemic stance. 4.2 KonjunktivII as a marker of negative epistemic stance Not all occurrences of epistemic müßte and könnte can be analysed in the above way. In the following examples the KonjunktivII retains its intimation of a negative epistemic stance, indicating that the speaker dissociates herself from the commitment that would have been associated with the use of the corresponding modal in the indicative. In (18a) to (18c) epistemic müßte appears in the apodosis of a conditional. Examples (a) and (b) differ with respect to the mood used in the protasis: in (18a) the mood of the verb wollte is KonjunktivII, whereas in (18b) the speaker uses the Indikativ in the protasis. In (18a), therefore, the speaker indicates that she is sceptical regarding the truth of the protasis. This scepticism is confirmed by the apodosis: the logical consequence of the protasis (i.e. the fact that the shriek would have to be short and shrill) stands in an apparent contrast to reality (the shriek is protracted), so that it must be concluded that the protasis is counterfactual. It should again be emphasized that the counterfactuality of the protasis is only contextually determined and not 'signalled' by the past subjunctive as such. By using KonjunktivII in the protasis, the speaker merely expresses a negative epistemic stance; in principle, the first sentence could be followed by and indeed, this is what we observe. The negative epistemic stance attributed to müßte in the apodosis is motivated by the apparent counterfactuality of the state of affairs that logically would have to obtain. It should be emphasized that the past subjunctive element does not signal an epistemic distance with respect to the logical necessity expressed by the muß-component in müßte: the inference as such is valid. (18) a. Wollte der Vogel seinen Fänger erschrecken, um so [...] den Zähnen und Klauen entweichen zu können, so müßte der Schrei kurz und schrill sein. In Wirklichkeit ist er aber langgezogen. "If the bird wanted to scare its aggressor to escape from its teeth and claws, the shriek would have to be short and shrill. In reality, however, it is protracted" In (18b) the speaker does not explicitly signal a negative epistemic stance in the protasis. The use of the indicative belagert werden is motivated by the fact that the speaker refers neutrally to the report of the Croatian radio.
KONJUNKTIV II AND EPISTEMIC MODALS IN GERMAN
207
However, as in (18a), the logical consequence of the truth of the protasis leads to a clash with reality: there aren't twenty times as many occupants in Cerska as in times of peace. The past subjunctive müßte in the apodosis explicitly draws attention to this contrast between what the protasis would lead us to conclude and reality. It can therefore be said to signal the speaker's negative epistemic stance. It should be noted again that the validity of the inference as such is not questioned. (18) b. Und wenn in Cerska "20 000 Frauen, Kinder und alte Leute belagert" werden, wie der kroatische Rundfunk gemeldet hat, dann müßte der Ort heute 20mal so viele Einwohner zählen wie in Friedenszeiten. "And if 20,000 women, children and elderly people are being besieged in Cerska, as the Croatian radio reported, then the town would have to have twenty times as many occupants as in times of peace." [In the following context it is made clear that this is not the case] Let us now consider the conditional in (18c), the protasis of which contains an indicative verb (ist). Again, müßte expressing logical necessity appears in the apodosis. In this case, the inference based on the protasis is confirmed by reality. However, one could still claim that the speaker signals a negative epistemic stance to the extent that she explicitly draws attention to the fact that her epistemic evaluation is conditional, i.e. that it depends on the correctness of the assumption that members of type A are indeed more susceptible to various coronary diseases. The speaker thus does not seem to be fully committed to the epistemic evaluation that is expressed by the modal verb. (18) Die beiden Forscher haben 1960 eine großangelegte Untersuchung begonnen. Sie haben 3500 gesunde Personen zwischen 39 und 59 Jahren herausgesucht und darunter diejenigen, die psychologisch dem Verhaltenstyp A entsprechen, von denjenigen geschieden, die diesen Typ nicht verkörpern. Wenn es nun richtig ist, daß Typ A besonders zu Herzkranzgefäßkrankheiten [...] neigt, dann müßte sich die Vorhersage bestätigen, daß die [...] Vertreter des Verhaltenstyps
208
TANJA MORTELMANS
A [...] statistisch häufiger an Herzkranzgefäßleiden erkranken als die Angehörigen der Kontrollgruppe. "Both researchers started a large-scale investigation in 1960. They selected 3500 healthy people between the age of 39 and 59, and separated those who fit psychologically into behavioural type A from those who do not. Now if it is correct that type A shows a special tendency toward coronary diseases, the prediction that type A representatives suffer coronary diseases at a statistically higher rate than the control group, ought to be borne out." [The rest of the text indicates that this prediction is indeed borne out] It is also possible to find examples for epistemic könnte in which the KonjunktivII signals a negative epistemic stance. In (19a) epistemic könnte appears in the apodosis of a conditional, the protasis of which sets up a clearly counterfactual state of affairs, as the preceding sentence explicitly states that the national character is indeed disappearing. Since the protasis doesn't hold, the consequence in the apodosis must be regarded as counterfactual as well. In (19b) the past subjunctive könnte receives a counterfactual interpretation which is triggered by our knowledge of the world: it would seem highly unlikely that the Croatians sing songs that are written by the people they besiege. (19) a. Der Nationalcharakter [...] wird allmählich abgebaut. Wenn das nicht der Fall wäre, könnte man in der Tat verzagen und verzweifeln. "The national character is gradually disappearing. If that were not the case, one could in fact become despondent and desperate." b. Manchmal spielen die Kroaten sentimentale Lieder, in Versen, die auch von den Belagerten komponiert sein könnten. "Quite often, the Croatians play sentimental songs, in strophes, which might as easily be verse written by the besieged." [but of course they aren't written by the besieged] I would claim that in cases like these, the modal verbs are clearly grounded themselves by the past subjunctive, signalling a negative epistemic stance. Both for müssen and können the negative epistemic stance is generally
KONJUNKTIV II AND EPISTEMIC MODALS IN GERMAN
209
determined by 'objective' factors that pertain to the apparent or possible nonfactuality of states of affairs. As the reader may have noticed, the above examples contain the forms müßte and könnte, but not dürfte. The observation that epistemic dürfte seems to be incompatible with a negative stance reading is not surprising in view of the high degree of grammaticalization the item possesses: using the grammaticalization parameters established by Lehmann (1985, 1995), we will be able to show that this verb form is one of the most strongly grammaticalized items within the paradigm of German epistemic modal verbs and therefore, as a non-decomposable unit, seems to move into the direction of a true 'grounding predication'. The rise of epistemic dürfte has to be situated in the sixteenth century, when the indicative of the verb could be used for the expression of epistemic possibility (this property was lost at a later stage) so that in combination with the KonjunktivII, dürfte expressed a kind of attenuated epistemic possibility. Some authors (e.g. Brünner & Redder 1983) favour an alternative devel opment: in view of the fact that the epistemic judgement expressed by dürfte nowadays is one of probability, i.e. is relatively strong, it is suggested that the epistemic meaning has to be related to an earlier language stage, in which dürfen expressed necessity, typically in negative contexts. According to Briinner & Redder, the attenuation effected by the KonjunktivII resulted in a weakened necessity meaning which gave rise to the probability meaning. This latter position, however, is highly problematic in view of the fact that the epistemic meaning arose at a moment when the necessity meaning of dürfen was not available (or at least not prevalent) anymore: dürfen in its necessity meaning had been more or less replaced by müssen by the beginning of the fifteenth century (cf. Lötscher 1991:352). 5. An analysis of the degree of grammaticalization of epistemic dürfte 5.1 Lehmann 's parameters Lehmann (1985, 1995) introduces three grammaticalization parameters and defines them at both a paradigmatic and a syntagmatic level. Firstly, the 'weight' of a sign is a property which renders it distinct from the other members of its class. The more distinct it is, the less grammaticalized. The second property, 'cohesion', can be tied up with the relational aspects of a given sign: if it systematically enters into relations with other signs, its autonomy decreases and it becomes more strongly grammaticalized. Finally, Lehmann introduces the parameter of 'variability', i.e. the mobility or shiftability with respect to other signs, which generally decreases with increased grammaticalization. Diewald (1993, 1997) makes use of these criteria
210
TANJA MORTELMANS
to show that the German modal verbs fall into two groups: the deontically used modal verbs are less grammaticalized, whereas the epistemically used modals are more strongly grammaticalized. Her analysis can, however, be taken one step further. The paradigm of German epistemic modal verbs does not seem to constitute a completely homogeneous group: some of the epistemic modals appear to possess a higher degree of grammaticalization than others. The epistemic modal dürfte is a case in point. To confirm this I will mainly make use of Lehmann's parameter of 'paradigmatic variability', i.e. "the freedom with which the language user chooses a sign" (Lehmann 1995: 137). The freedom with which dürfte can be chosen out of the paradigm of epistemic modal verbs is decreasing in con temporary German, whereas the other epistemic verbs (especially könnte and müßte) stand in a relatively free pattern of alternation and can be unproblematically selected for a variety of communicative purposes. 5.2 No alternation Indikativ - KonjunktivII It has already been pointed out that, in contrast to other epistemic modals, there is no possible alternation between an indicative form darf with epistemic meaning and a counterpart in the KonjunktivII. 5.3 Non-occurrence in past-time contexts Both epistemic müssen and können can — albeit to a restricted extent — be used in the past tense (see examples (14a - 14c)), this is impossible for dürfen. 5.4
Non-occurrence in counterfactual conditionals A third indication of the increased degree of grammaticalization of dürfte was provided by a survey I conducted among native speakers of German. I presented them with a number of sentences, some of which contained genuine counterfactual conditions, whereas others were open to a potentialis inter pretation. I then asked them to judge the acceptability of the verb forms würde, müßte, dürfte and könnte in the given contexts. Whereas könnte and müßte were generally possible in both (counterfactual) irrealis and potentialis contexts, dürfte was only selected in the latter. This can again be regarded as a reduction of the paradigmatic variability of dürfte, as the number of contexts in which it figures is reduced compared to that of the less grammaticalized items.
KONJUNKTIV II AND EPISTEMIC MODALS IN GERMAN
211
(20) a. Wenn er intelligenter wäre, er einsehen, dass seine Versuche umsonst sind "If he were more intelligent, he realize that his attempts are useless" b. Mit einer gründlicheren Vorbereitung sie die Prüfung schaffen "With a better preparation she pass the exam" c. Wenn die Lohnnebenkosten sinken würden, dies zu einer Verringerung der Arbeitslosigkeit führen "If the labour costs went down, this lead to a decrease in unemployment" In sentence (20a), where the condition is clearly counterfactual, only 3 out of 16 speakers approved of dürfte. In sentences (20b) and (20c), however, which differ in that they are future-oriented and therefore more open to a potentialis interpretation, 11 and 10 speakers respectively, found dürfte acceptable. 6.
Speaker orientation Many (if not all) cases of grammaticalization are accompanied by a semantic-pragmatic process whereby "meanings become increasingly based in the speaker's subjective belief state/attitude toward the proposition"(Traugott 1995:31). Traugott also terms this process 'subjectification', but defines it in a much broader way than Langacker. To avoid confusion between Traugott's and Langacker's use of the term, I will speak of 'speaker orientation' to refer to Traugott's 'subjectification'. As the previous sections have shown that dürfte is a highly grammaticalized item, we might expect it to be highly 'subjective', i.e. speaker-oriented, as well. This is corroborated by the observation in Matthews (1996), who notes for epistemic dürfte that "[n]ative German reactions to dürfte are that is it equivalent to something like: 'es ist anzunehmen', not just 'wahrscheinlich' — 'probably' [...]. This translates as something like: 'presumably' clearly within the subjective parameter" (Matthews 1996:377). The strongly subjective nature of dürfte can be regarded as one of the reasons why dürfte expresses a relatively high degree of likelihood. It is not unreasonable to assume that a highly subjective marker of epistemic possibility evolves to an expression of epistemic probability. For English, Coates (1983) notes that for many speakers the difference between epistemic may and might
212
TANJA MORTELMANS
has vanished. Both express an essentially subjective assessment of likelihood: "It [MIGHT] seems no longer to be used as the tentative form of MAY, but simply as an alternative form for the expression of the modality 'it is possible that'" (Coates 1983:153). For dürfte we may assume then that the original expression of attenuated epistemic possibility has gained force over time (because it is constantly backed up by the speaker) and therefore comes close to an expression of subjective epistemic probability nowadays. The strongly inferential character of epistemic müßte, on the other hand, seems to be compatible with both objective and subjective interpretations, 'objective' here referring to the strictly logical and hence generally accepted nature of the inferences involved. It is by no means coincidental that the negative stance reading of epistemic müßte in (18) occurs in contexts where the inference expressed by müssen is of a mathematical or strictly logical (and therefore rather 'objective') kind, so that it does not necessarily invoke any positive speaker commitment: since a logical inference is generally accepted, there is no special need for the speaker to commit herself to the validity of its consequence. Of course, inferential modals do not exclude a positive speaker commitment and can develop into highly subjective markers (e.g. epistemic must in English), but German müssen has not (yet) reached this stage. 7.
A constrained division of labour It should have become clear that there is a certain division of labour between the grounding brought about by the KonjunktivII and by the epistemic modal verbs in German. The former generally indicates the speaker's negative epistemic stance, which typically results in a location in (often counterfactual) irreality, whereas the modals locate the process in the epistemic regions of potential and projected reality. When the KonjunktivII and modal verbs are combined, the meaning of the resulting items cannot always be described in terms of full compositionality: epistemic könnte most often signals a weakened, but positive epistemic assessment rather than an epistemic judgement with negative implications. In these cases, an analysis that attributes an amount of unit status to the subjunctive form as such is advocated. However, epistemic müßte and könnte can also be used in contexts in which the strongly grammaticalized KonjunktivII as a 'grounding predication' clearly indicates the speaker's negative epistemic stance. Such a use is ex cluded for epistemic dürfte, which has moved so close to the status of a genuine 'grounding predication' that it resists a dominant grounding by the KonjunktivII and as a unit signals a highly subjective epistemic evaluation.
KONJUNKTIV II AND EPISTEMIC MODALS IN GERMAN
213
Notes * I would like to express my gratitude to Frank Brisard, Steven Geukens, Louis Goossens, Jan Nuyts, Gert Van Rillaer and two anonymous reviewers for their valuable comments on earlier drafts of this paper. Special thanks in this respect are due to Frederike van der Leek. Finally, I am also indebted to the Vlaamse Wetenschappelijke Unie for its financial support. 1 The concept 'ground' refers to the speech event, its participants (speaker and hearer(s)), and its immediate circumstances (Langacker 1991:548). 2 Goossens (1996) argues persuasively that the English modal can is not on a par with the other modals, since it can be grounded by tense (e.g. He could play the piano when he was six) and therefore must be said to profile the modal relation of ability. Moreover, modals expressing deontic modality can only taken to be grounding, if "the authority for the permission or obligation is clearly in the ground, as a rule, when the speaker has or assumes authority (Goossens 1996:28). Since the modal verb must can also be used to express objective obligation with which the speaker does not associate (e.g. Clay pots must have some protection from severe weather), must cannot be considered as a 'grounding predication' in all of its uses. In view of these possible refinements, the statement that all English modals are 'grounding predications' should be handled with care. 3 In German, the Indikativ Perfekt can sometimes be used as a pure equivalent of the Indikativ Präteritum without any aspectual components. In some regions of Germany, this periphrastic variant of the Indikativ Präteritum is even said to have replaced the original synthetic Indikativ Präteritum (cf. Drosdowski 1995:150). 4 It should be noted that the past perfect tense form in (8c) hätte gesprochen does not trigger a counterfactual interpretation. 5 The sentences of the examples 8 (d), 14 (b and c), 15 (c), 18 (a, b and c) and 19 (a and b) are taken from various existing textcorpora, made available for computer research by the Institut für deutsche Sprache (IDS). 6 In spoken language, the Konjunktiv II and especially the Indikativ take over (cf. Drosdowski 1995:164). 7 The modal verb mögen as a marker of epistemic possibility is gradually disappearing in present-day German (cf. Drosdowski 1995:103). As its English and Dutch counterparts may and mogen, mögen can be used in concessive clauses of the type Das mag zwar wahr sein, aber gutheißen kann ich das nicht ('That may be true, but I can't approve of it'), a use which according to van der Auwera & Plungian (1998:93) has developed out of the epistemic possibility meaning. 8 In order not to complicate matters I treat the three epistemic modals können, müssen and werden on a par. However, there are considerable differences regarding the possibility of using them epistemically for past situations. Epistemic werden does not allow a past tense form wurde at all, the epistemic use of past mußte is restricted to 'erlebte Rede' and the like, whereas können is the most flexible item in that it appears to allow — albeit marginally — an epistemic use in the past.
214
TANJA MORTELMANS
References Brünner, Angelika & Gisela Redder. 1983. Studien zur Verwendung der Modalverben. Tübingen: Narr. Coates, Jennifer. 1983. The Semantics of the Modal Auxiliaries. London & Canberra: Croom Helm. Diewald, Gabriele. 1993. "Zur Grammatikalisierung der Modalverben im Deutschen". Zeitschrift für Sprachwissenschaft 12:2.218-234. . 1997. Grammatikalisierung. Eine Einführung in Sein und Werden grammatischer Formen. (= Germanistische Arbeitshefte 36). Tübingen: Niemeyer. Drosdowski, Günther. ed. 1995. Duden. Grammatik der deutschen Gegen wartssprache 5, völlig neue bearb. u. erw. Aufl. Mannheim: Duden verlag. Durrell, Martin. ed. 1991. Hammer's German Grammar & Usage. London & New York: Edward Arnold. Fillmore, Charles J. 1990. "Epistemic stance and Grammatical Form in English Conditional Sentences". Papers from the 26th Regional Meeting of the Chicago Linguistic Society eds. Michael Ziolkowski, Manuela Noske & Karen Deaton, Volume 1.137-162. Chicago: Chicago Linguistic Society. Goossens, Louis. 1996. English Modals and Functional Models: A Confrontation. (=Antwerp Papers in Linguistics 86) Antwerp: University of Antwerp. Langacker, Ronald W. 1991. Foundations of Cognitive Grammar. Volume II. Descriptive Application. Stanford: Stanford University Press. Lehmann, Christian. 1985. "Grammaticalization. Synchronic Variation and Diachronic Change". Lingua e Stile. XX:3.303-318. . 1995. Thoughts on Grammaticalization. (=LINCOM Studies in Theoretical Linguistics 1) München & Newcastle: Lincom Europa. Lötscher, Andreas. 1991. "Der Konjunktiv II bei Modalverben und die Semantik des Konjunktiv IF. Sprachwissenschaft 16.334-364. Matthews, Richard. 1996. "Oblique' modals in English and German". Language Sciences 18.363-379. Thieroff, Rolf. 1992. Das finite Verb im Deutschen: Tempus, Modus, Distanz. Tübingen: Narr. Traugott, Elizabeth Closs. 1995. "Subjectification in grammaticalisation". Subjectivity and subjectivisation. Linguistic perspectives eds. Dieter Stein & Susan Wright. 31-53. Cambridge: Cambridge University Press.
KONJUNKTIV II AND EPISTEMIC MODALS IN GERMAN
215
Van der Auwera, Johan & Vladimir A. Plungian. 1998. "Modality's semantic map". Linguistic Typology 2.79-124. Wilmanns, W. 1906. Deutsche Grammatik. Gotisch, Alt-, Mittel- und Neuhochdeutsch. Dritte Abteilung: Flexion. 1. Hälfte: Verbum. Strassburg: Karl J. Trübner Verlag.
Subjectivity and Conditionality The Marking of Speaker Involvement in Modern Greek Kiki Nikiforidou & Demetra Katis University of Athens, Greece 1.
Introduction * It is by now commonly understood in cognitive linguistics that meaning is often a matter of construal (e.g. Fauconnier 1985, 1997, Lakoff 1987, Langacker 1987, 1990a, 1990b, Talmy 1988, Sweetser 1990, Goldberg 1996). "A foundational claim of cognitive semantics", says Langacker (1990b:5), "is that an expression's meaning can not be reduced to an objective char acterization of the situation described: equally important for linguistic semantics is how the conceptualizer chooses to construe the situation and portray it for expressive purposes". In this paper we wish to show how the speaker's construal of a situation enters crucially into the description of Modern Greek conditional constructions. We will examine conditionals introduced by ama, ean and na. Traditional and linguistic analyses (cf. Joseph & Philippaki 1987:31, Holton, Mackridge and Philippaki 1997:457) treat these markers as synonyms and free variants of an, which is widely assumed to be the commonest conditional marker; all four markers are normally glossed into English as 'if'. Our analysis of these constructions is based mostly upon corpus data as well as upon additional examples from different registers. The corpora include 14 hours of informal conversation on live TV shows (TVC), 20 hours of student and teacher discourse in a sixth grade classroom (SCD and TCD respectively), and a 159 pages anthropological linguistics textbook (ALT). The appendix shows the distribution of an and the other markers in the four registers of our corpora. We have been led by our data to the following arguments. First, the choice of one conditional marker over the other, far from being truth-conditionally controlled, reflects the speaker's construal of the situation expressed in the protasis as being of a particular nature: this is exactly what constitutes each marker's special meaning. We, therefore, refute previous treatments of these constructions as free variants, claiming instead that the speaker's assessment and portrayal of a situation should be part of linguistic meaning proper. Secondly, the construal of the protasis by all three markers falls under what Langacker (1990b:8) calls a 'subjective' construal or an 'egocentric
218
KIKI NIKIFORIDOU & DEMETRA KATIS
viewing arrangement'; all three markers code a relation of the protasis to the speaker, which is, in turn, fully consistent with their original etymology of highly subjective (and in some cases even deictic) semantics. Differences in their meaning and further specialization can be motivated through specific aspects of their lexical semantics. Semantic and pragmatic factors beyond truthconditions have already been found relevant to the description of conditionals. They include emotive aspects of meaning (e.g. Akatsuka 1986, 1997, Akatsuka and Clancy 1993), pragmatically motivated interpretations of conditionals (e.g. Sweetser 1990:ch. 5, Dancygier & Sweetser 1996), and the various pragmatic functions served by conditional protases (e.g. Van der Auwera 1986, Athanasiadou and Dirven 1996). However, the strict notion of a subjective construal in Langacker's (1990a, b) sense, which includes (albeit implicitly) the viewer (in our case the speaker) in the scope of the conditional marker (i.e. in the mental space set up by it), has not been exploited in the description of conditional semantics. What we aim to show is, precisely, that subjectivity in this sense allows a unified and systematic description of Modern Greek conditionals and perhaps of conditionals more generally. Finally, certain semantic/pragmatic extensions of the Greek conditional markers in discussion may be argued to represent moves towards greater subjectification. Such extensions do not involve further "realignment of some relationship from the objective to the subjective axis" (Langacker 1990b: 17) and are not therefore considered more subjective uses in Langacker's sense. They do however fall under Traugott's broader notion of subjectification as a tendency for meanings to become grounded in the speaker's world of reasoning, belief or metatextual attitude to the discourse (Traugott 1989, 1995). Whereas certain semantic developments can be characterized as instances of both types of subjectification (e.g. the development of go from its spatial to its temporal sense1), others (e.g. the shift from temporal to concessive while, the development of scalar particles, e.g. very, in uses like the very height of her career (see also Brugman 1984), and the acquisition of strongly epistemic meanings by the (already) epistemic modals) can only be seen as instances of subjectification in Traugott's sense. This is because such developments do not consist in the inclusion of part of the ground in the "onstage region" (Langacker 1990b: 7) — since the ground is already included in their original sense — but rather in their relying more (and perhaps more explicitly) on the speaker's personal evaluation of the situation. Subjectification as defined by Traugott is therefore a broad notion which affects large domains of the lexicon. The recurrence of such extensions in the meanings of the Greek conditional markers, which, starting with clearly subjective meanings (in Langacker's sense) develop more subjective (i.e. more explicitly speaker-
SUBJECTIVITY AND CONDITIONALITY
219
based) meanings, suggests that these developments are natural. In fact, we would like to claim that the shift towards meanings which actively involve the speaker's evaluation and judgement is clearly motivated in cases where the original meaning relies necessarily on the speaker/conceptualizer, even if implicitly, and depends on her vantage point and/or physical presence. We therefore aim to show that such data provide a link between the two notions of subjectivity and subjectification. 2.
Conditionals with ama: from simultaneity to speaker involvement We begin with conditionals introduced by ama, a marker derived from the Ancient Greek temporal adverb hama which meant 'at the same time as/simultaneously' (Liddell-Scott 1977), e.g. (1): (1) Hama de tauta elege kai At the same time PART these say:P:IMPF:3Sg and apedeiknye prove:NP:IMPF:3S "At the same time as saying these, he also showed (the baby)" (Herodotus 1.112)
Even in its original temporal sense hama reflected a subjective meaning in Langacker's sense (1990b). Simultaneity is temporal comparison which in turn presupposes a reference point, completely dependent on the concep tualizer, with whom the extrapolated relationship resides. Moreover, and this is where subjectivity really lies, hama (in accordance with its etymology from the Indoeuropean root sem- 'identical, together with' — cf. Watkins 1985) construes the speaker as being a 'viewer from within' of the event described in the temporal clause. Hama, in other words, sets up a space in which the speaker and the participants in the event co-exist and in which the former observes the latter from close-by. In an example such as (1), the events described in the two clauses are therefore subjectively construed (even if observable in the world), and the relationship of simultaneity, with the first event serving as the reference point for the temporal location of the second, presupposes a subjective (i.e. inside) observer who imposes this construal.2 The shift from temporal to conditional meaning which characterizes the subsequent history of ama, is common cross-linguistically (Traugott 1985). Already evident in medieval times (Kriaras 1969), the change is by now complete, since purely hypothetical uses are possible (e.g. (2)) alongside what might be deemed temporal-hypothetical ones (e.g. (3)) (cf. Schiffrin 1992 for
220
KIKI NIKIFORIDOU & DEMETRA KATIS
parallels with English then), whereas purely temporal uses are highly idiomatic and restricted to certain dialects. (2) ...jati ama ben ipince to perivalon, NEG exist:P:IMPF:3SG the environment, ...because if be θα ipirxe tipota pano sti zoi NEG FUT exist:P:IMP:3SG nothing on on-the life "...because if the environment didn't exist, there would be no life on earth" (SCD) (3) Eyo θα ton vro. Ama ton FUT him find:NP:PERF:2SG. If/when him I vro, esi θα paris tilefono find:NP:PERF:lSG, you FUT take:NP:PERF:2S telephone "I will find him. If/when I find him (and I win therefore the bet), you will call me on the phone" (SCD) We have mentioned above that in its temporal sense ama expresses a subjective construal in that the ground (in this case the speaker) is implicitly included in its scope as an inside observer, while at the same time being the 'judge' of the simultaneity relationship. The conditional meaning can be argued to represent an extension in which the temporal occurrence of the event marked by ama and the subsequent location of the second event with respect to it have been replaced by (or co-exist with) a mental occurrence of the first event and a subsequent (conditional) location of the second on the part of the speaker/ conceptualizer. The first clause is still the reference point for the 'location' of the second but this reference point is now a hypothetical one and the actual occurrence in time of the event described in the first clause (the objective part in the semantics of temporal ama) becomes occurrence in a mental space which includes the speaker's viewpoint (an aspect of the ama semantics inherited from the temporal use). In both cases however the speaker's vantage point is taken to be within the real-world space (in the temporal reading) and within the hypothetical space (in the conditional reading). In Langacker's sense, therefore, the shift from temporal to conditional is not an instance of subjectification, since there is no realignment of any relationship from the subjective to the objective axis. Both the conditional and the temporal meanings are subjective. It is however subjectification in Traugott's broader sense since it is a move towards a (more explicitly)
SUBJECTIVITY AND CONDITIONALITY
221
speaker-based meaning. Conditionality, unlike the original temporal meaning of ama, represents a relationship which is grounded in the speaker's world of reasoning; the very fact of setting up a hypothetical mental space marks the speaker's belief in relation to the epistemic status of the event in the protasis and the conditional dependence of the event in the apodosis. Indeed, much of the recent literature on conditionals (see for example Akatsuka 1986,1997), which treats them as subjective constructions, relies on this observation.3 As seems usual with conditionals deriving from temporals (cf. e.g. Tabakowska 1997 for Polish), ama appears frequently in conditionals which are closer to the realis pole of the epistemic continuum. In the oral registers where it is normally attested (as shown in Katis 1997 as well as in our data), ama seems to correlate significantly with 'general correlations' (cf. Fillmore 1990) involving a temporal 'whenever' meaning and a sense of co-occurrence, which further supports the appropriateness of ama in this context, (e.g. (4)). It also correlates with speech act conditionals (in the sense of Sweetser 1990) where the protasis is given (e.g. (5))4 (in our TVC register, at least 85.6% of the ama conditionals are of the above two types — 20.6% general correlations and 65% speech acts). For example in (5b), the protasis is clearly (contextually) given and the conditional in this case 'locates' the consequent with respect to the conversationally relevant space introduced by ama: (4) Ama kapios katastrepsi ena δasos If/whenever someone destroy :NP:PERF:3SG a forest olokliro, ine san na katastrefi enan whole, be:NP:3Sg like MOD destroy : NP:IMPF:3S a θr, mìa zoi human being, a life "If/whenever someone destroys a whole forest, it is like destroying a human being, a life" (SCD) (5) a. Ekini lei oti sas aγapai She say:NP:IMPF:3SG that you love:NP:IMPF:3SG para poli very much "She is saying that she loves you very much" (TVC)
222
KIKI NIKIFORIDOU & DEMETRA KATIS
(5) b. Ama m' ayapai, tin ayapao If me love:NP:IMPF:3SG, her love:NP:IMPF:1SG ki eyo and I "If (now that you are mentioning that) she loves me, I love her too" (TVC) A consequence of the 'inside the space' meaning of ama (whether temporal or conditional) is an additional aspect of its interpretation which is crucial for our argument. Ama often signals the speaker's personal interest in (and even empathy with5) the content of the protasis (and by implication of the apodosis). It codes, in other words, a kind of speaker involvement (involvement as opposed to detachment in the sense of Chafe 1982), in a way which is not possible for an. Chafe makes, in fact, a distinction among three types of involvement: involvement of the speaker with himself/herself, with what is being talked about, and with the listener. Although these three types may overlap, they are also distinguishable and it is involvement with the content of speech that is expressed by ama (e.g. (6), (7) as well as (4) above): (6) Ki an avvio to proi eyo arostiso? And if tomorrow the morning I get-ill:NP:PERF:1SG? Ama pedano, ti θα kanete? If die:NP:PERF:lSG, what FUT do:NP:PERF:2PL? "And if I fall ill tomorrow morning? If I die, what will you do?" (Katis 1997 corpus) (7) Afto ine sovaro jati ama xtipisi This be:NP:3SG serious because if strike:NP:PERF:3SG pali piretos, imaste xameni again the fever, be:NP:lPL lost "This is serious because if the fever strikes again, we are lost!" (TVC) In (4) and (7) ama marks an emotional attitude of the speaker towards the forest's destruction and someone else's illness respectively (which amounts to empathy with the forest and person) in a way that the corresponding neutral protases with an could not. This aspect is strongly present in (6) where the shift from an to ama takes place as the stakes in the antecedent get higher and
SUBJECTIVITY AND CONDITIONALITY
223
the question in the consequent urgently demands a response. In accordance with our analysis of ama as a subjective marker, we could say that the shift from an to ama marks exactly the speaker's viewpoint as shifting inside the mental space and, by implication, her urging of the listener to take an internal viewpoint also. We should make clear that ama, unlike other conditional markers discussed in the literature (e.g. Akatsuka 1997), does not code desirability/ undesirability as may be surmised from these three examples where strong undesirability is implied. Rather, it expresses involvement with the content of the conditional, one aspect of which (though not the only one) may be the desirability /undesirability dimension. Thus, in (8) we have involvement but the desirability /undesirability aspect is missing. (8) Jati to brio ke ti fantasia i ta 'xis Because the brio and the imagination or them have:NP:2Sg emfita i apios i fisi he su ta innate or simply the nature NEG to-you them δini. Na ta apoktisis, give:NP:IMPF:3SG. MOD them acquire:NP:PERF:2SG, to poδosfero sti iδika ama s' aresi especially if to-you like:NP:3SG the football in-the stratosfera, apokliete... stratosphere,exclude:NP:IMPF:3SG "Because brio and imagination are either innate or nature simply doesn't give them to you. To acquire them, especially if you like football in the stratosphere, is impossible..." (Katis 1997 corpus)
This is, in fact, the only ama utterance in a newspaper corpus, where a sports writer is so carried away with enthusiasm for what he is saying, that he feels the need to switch to a non-detached type of discourse. Ama's marking emotional involvement with the content probably explains its frequency in informal oral registers and its absence in formal detached ones. It is indicative that we found no instance of ama in the scientific text we examined, while Katis (1997) has shown it to be the most frequent marker in informal home conversations. Given the subjective character of an ama conditional, the extension from the marking of 'pure' or 'neutral' conditionality to the marking of personal involvement with the content of the conditional seems, as we have said, fairly
224
KIKI NIKIFORIDOU & DEMETRA KATIS
motivated. The speaker's concern for the antecedent and its result is more easily provoked when her vantage point is construed as being that of an inside observer and she may be therefore understood as directly or imminently affected. At the same time, this extension could be again described as an instance of Traugott's notion of subjectification as a tendency for "meanings to become increasingly based on the speaker's subjective belief state/attitude toward the proposition" (Traugott 1989:35). The shift to empathy, in other words, reflects an increase in the speaker's 'presence' and marked attitude with respect to the content of the protasis. We may therefore conclude that an originally subjective marker in Langacker's sense of the term becomes increasingly subjectified in Traugott's sense, i.e. its meaning becomes more explicitly speaker-based. As we show further on, this kind of development is evidenced in the semantics of other conditional markers as well. 3.
Ean conditionals: from concrete to discourse deixis We now turn to ean conditionals, which were already present in Ancient Greek, e.g. (9): (9) Ean eme apokteinete, blapsete hymas autous selves If me kill:PRES:2PL, harm:FUT:2Pl you "If you kill me, you will harm yourselves" (from Tzartzanos 1956). Ean was the contracted form of ei-an (Greenberg 1986). Ei was the typical means of introducing conditional clauses, while an was a marker of hypotheticality, used systematically in the apodosis of counterfactuals and occasionally in the protasis of non-counterfactuals as well. In the latter case, it was often preceded by ei, thus resulting in the ean contraction. Through a historical route that remains to be investigated, an has become the commonest means of introducing Modern Greek conditional clauses. Ean is, however, not only another possible choice but also a frequent one in some registers (in our scientific text it introduces 45.5% of all conditionals). We will claim that the choice of ean over the other markers is significant, as ean marks its own particular sort of speaker involvcmcnt. In our data ean systematically focuses at a particular point of the discourse, where the speaker wishes to raise the listener's involvement with what is expected of him on the speech-act level. It is important to clarify that ean does not so much focus attention upon the informational content of the conditional — though this aspect of focusing is never lost — as upon its pragmatic illocutionary value (see van Dijk 1981 and Tannen 1985 for the
SUBJECTIVITY AND CONDITIONALITY
225
distinction between the two types of focus). Given Chafe's (1982) distinction of involvement into three types, ean marks interpersonal involvement between the speaker and listener. This critical function of ean is most obvious whenever additional meanings are absent. In (10) the speaker shifts from an to ean when his plea to have an annoying mobile phone removed becomes urgent and the need for compliance very high. In fact, ean is the only marker to be systematically pronounced with emphatic intonation, a quite iconic means of calling for attention/involvement. (10) θα sas parakalusa an iparxi kapio FUT you plead:P:IMPF:lSG if exist:NP:IMPF:3SG some kinito. M' akui liγo, siγnomi, siγnomi, mobile. Me hear:NP.TMPF;3SG little, sorry, sorry, miso lepto! Ean iparxi kapio kinito, half minute! If exist:NP:IMPF:3SG some mobile, na to apomakrinete MOD it get-away :NP:2PL "I would ask you if there is a mobile (phone here). She can hardly hear me! Sorry, sorry, just a minute! If there is a mobile (phone here), you should remove it" (TVC) Ean's particular semantic contribution may be traceable to the originally demonstrative meaning of ei 'this', already a subjective one because of its deictic nature (Langacker 1990a, b). We may assume, based on Traugott (1985), that conditional ei developed from the topicalizing meaning of ei 'thus', topicalizers being a common historical source for conditionals, where the protasis serves as the frame for the apodosis. Traugott does not, however, concern herself with how the demonstrative may have turned into a topicalizer in the first place. We suggest that this shift represented a move from a concrete towards an abstract deictic meaning (of the sort proposed by Dancygier and Sweetser 1997 for English then); whereas ei previously highlighted a specific part of the concrete physical environment, it later highlighted a specific part of the discourse world by presenting it as given. In the further move to conditionality, ei highlighted a part in a hypothetical mental space. In both shifts, an increase in subjectivity (in Traugott's sense) also takes place, as the point of reference moves from the objective world to the world of discourse, becoming increasingly speaker-based. Modern Greek
226
KIKI NIKIFORIDOU & DEMETRA KATIS
ean inherits all these meanings from its ei component, although it may not necessarily be morphologically and semantically transparent to its speakers. It has nonetheless become even more abstract and subjective, as it highlights a particular point in the world of discourse in expressing the speaker's evaluative (and in cases emotive) stance (towards the listener and what is demanded from him/her). In fact, this focusing function of ean makes it most appropriate for speech acts calling for high interpersonal involvement. It is thus no accident that ean seems to correlate in the most natural site of speech, everyday mundane conversation, with deterrents, e.g. (11): (11) ..Jati ean hen δiefkolini ke ...because if NEG make easy:NP:3SG and afli i kalastasi, kat' anagi θα sinexisti continue:NP:3SG this the situation, upon need FUT proxorisi i piniki laviδa... proceed:NP:PFV:3SH the legal catch "...because if she doesn't make things easier and this situation continues, legal action will have to be undertaken..." (TVC) In warnings and threats the content of the protasis is normally highly undesirable to the speaker, because of its abhorrent consequences (often physical harm or socially imposed penalties for the listener), and thus in strong need to be averted (often in order to have a beloved person protected as in child-directed discourse). The association of ama with imminence would not, we suppose, make it appropriate for such uses, since the point in these speech acts is exactly to make the event non-imminent, while the neutrality of an would not suffice for getting the listener involved with the speaker's directive. This association of ean with deterrents probably motivates its extension to meanings relatively close to the irrealis pole of the epistemic continuum. Quite indicatively, the unlikelihood of the protasis is implicated in as high as 80% of ean conditionals in our TVC corpus (regardless of the tense pattern in the protasis and the apodosis which may be independently associated with the expression of the irrealis and the counterfactual or, conversely, the realis and the possible). This irrealis character of ean takes various shades: in the above corpus it predominantly functions as a marker of the speaker's lack of belief and/or desire regarding the situation denoted in the protasis — highly subjective factors which have been previously associated with conditionals (e.g. Akatsuka 1986, 1997). In example (12) the speaker expresses disbelief of an abhorrent
SUBJECTIVITY AND CONDITIONALITY
227
possibility as well as challenges the listener to take a position on this shocking matter. (12) Ean itan lipon singenis, θα ton afinan If be:Ρ so relatives, FUT him leave:P:IMPF:3PL na ftiaksun tetio prayma i ipolipi? MOD make:NP:PERF:3PL such thing the rest? θα on afinan eki pera? Ine FUT him leave:P.TMPV:3PL there over? Be:NP:3SG binaton ? possible? "So if they were relatives, would anyone else let him do such a thing? Would they leave him there? Is that possible?" (TVC) Ean's marking of interpersonal involvement seems, interestingly, its basic function even in detached written registers. In our ALT corpus it occasionally marks a particular and subtle meaning, which we will term 'deviation from the norm'. This seems an extension of exceptionality, suited to ean in the first place because of its highly irrealis character. In 'deviation from the norm' meanings, the deviation ranges from the nearly exceptional to the simply different from what has been said, while the norm is a matter of expectations given the overall discourse and the workings of the socio-physical world. In (13) for example we find a move to a different and contrasting point of view. (13) Den isxirizome oti i enia pu NEG claim:NP:IMPF:lSG that the concept that eksetazo ke efarmozo examine :NP: IMPF-1SG and apply :NP.TMPF:1SG apoteli ena anef elipseon erγalio ... consist:NP:IMPF:3SG a without lacks tool ... Omos ean xrisimopiiθi prosextika, bori However if use:NP:PERF:3SG carefully, can:NP:IMPF:3SG na mas katefθini se enδiaferuses leptomeries ... MOD us direct:NP:3SG to interesting details "I am not claiming that the concept which I examine and apply is a tool without drawbacks... If it is, however, employed carefully, it can direct us to interesting details..." (ALT)
228
KIKI NIKIFORIDOU & DEMETRA KATIS
Though in these cases ean seems to come closer to focusing upon informational content (as the writer's illocutionary intention is to get the listener involved with subtle yet important points of the text rather than with avoiding or carrying out some other act), it is, ultimately, even here a means of raising interpersonal involvement. In fact, the presence of ean makes even a detached scientific monologue approach the liveliness of an oral interactional register. We may conclude that ean is differentiated from the neutrally hypothetical an by marking points in the discourse where the speaker signals involvement with the listener. It may not, in fact, be fortuitous that it survived over its variant contraction en in Ancient Greek, as its difference from an is exhibited in a quite iconic manner, i.e. the extra meaning is signalled by the e- element. Further taking into account its special intonation, ean may be said to be an emphatic conditional marker. Interestingly, Akatsuka (1997) notes that subjectification in Japanese and Korean conditionals is accompanied by a strengthening of emphatic forms. The history of ean exhibits the same sort of semantic changes as ama, i.e. shifts from concrete to abstract meanings as well as from less to more subjective ones. The original demonstrative character of ei involves an egocentric view of a concrete physical reality (i.e. the marker starts with a subjective meaning in Langacker's sense), which in its topicalizing function becomes an egocentric view of a discourse reality (further subjectification in Traugott's sense), and finally in its conditional meaning becomes an egocentric view of a hypothetical mental reality. The critical function of today's ean involves a further shift from deictic reference to a mental space towards the marking of the speaker's evaluative stance within the mental space; its meaning becomes therefore more explicitly speaker-based in Traugott's sense.6 4.
Conditionals with na: grounded conditionality Of the three conditional markers discussed here, na codes the speaker's presence in the most direct way since it serves to deictically anchor the antecedent to the speaker, e.g. (14)-(16): (14) Na s' evlepe o pateras su, θα se If you see:P:IMPF:3SG the father your, FUT you skotone kill:P:IMPF:3SG "If your father saw you (now as I do), he 'd kill you" (from adultchild discourse)
SUBJECTIVITY AND CONDITIONALITY
229
(15) δio lepta pio prin ' fiji, θα Two minutes more before if have:P:3SG leave:INF, FUT prolavene make-it-on-time : Ρ : IMPF-3 SG 'If he had left two minutes earlier (from where we were), he would have made it" (from a recorded faculty meeting) (16) Na emfanisti ksafnika, θα jini If appear:NP:PERF:3SG suddenly, FUT become:NP:3SG meyali fasaria big trouble "If she suddenly appears (here), there will be big trouble" (from a recorded faculty meeting) In all these sentences na signals the speaker's physical presence at the scene or event described in the protasis (this being the crucial difference from ama which simply places the speaker in the same world/space as the event, attributing to her the position of an inside observer). In (14) the implication is that the speaker is watching the addressee, in (15) that she witnessed the departure referred to in the protasis and in (16) that she is actually present at the scene. Such a meaning vanishes once we replace na with neutral an, which carries no deictic implications. This specialized semantics of na probably explains its overall rarity as a conditional marker and its apparent exclusion from written registers. We have, in fact, found only one na conditional in our corpora, in the oral TCD register. The special meaning of na as a conditional may be motivated through its use as a deictic/presentative particle, e.g. (17): (17) na o janis ! DEMONS the John! "Here's John!" In particular, Christidis and Nikiforidou (1994) have argued that conditional na is a hybrid function between the presentative use of the particle illustrated above and its subordinating function exemplified in (18):
230
KIKI NIKIFORIDOU & DEMETRA KATIS
(18)
θelo na petixo want.NP.IMPF: 1SG MOD succeed:NP:PERF: 1SG "I want to succeed"
While there is no doubt that these two na 's are etymologically distinct (deictic na derives from the Ancient Greek particle en and subordinating na from the Ancient Greek conjunction hina), Christidis and Nikiforidou have argued that they are nevertheless synchronically related; subordinating na may be analyzed as an abstract metaphorical extension of presentative na, locating its complement in unrealized temporal space in a way analogous to deictic na locating its concrete complement. The synchronic polysemy may in turn shed some light on the diachronic developments which led to homophony of the two markers. As a grounding element in Langacker's (1990a,b) sense, na inherently represents a maximally subjective construal. At the same time, the deictic constraint of the speaker's physical presence at the scene may, on occasion, give rise to an implicature of personal involvement or interest in the outcome, not unlike the one we described for ama, although involvement arises in each case, as we have seen, via different routes. Example (16) conveys, for instance, a sense of worrying on the speaker's part, which is missing from the corresponding protasis with an (presumably because, in that case, there is no implication of the speaker's physical presence either). Similarly in (19), na, besides marking the speaker's physical presence, signals almost a wish on his part to have been present in the original scene he is referring to or to bring this scene from the past to the here-and-now. Because of this, (19) serves the teacher's attempt to enliven a writing lesson, by 'bringing' the grandmother he is referring to in front of his students. (19) Ande ke na mas eleje "Ax Go:IMP:2SG and MOD to-us say:P:IMPF:3SG "Oh mu, me tis sapunaδes", eprepe ti peδaki child-DIMIN my, with the soappings", should:P:3SG what na mas γrapsi? Grapste MOD to-us write:NP:PERF:3SG Write : IMP: PFV:2PL to kiolas it already
SUBJECTIVITY AND CONDITIONALITY
231
(19) "Well, if she (the grandmother) were (here) to tell us "Oh my child, with all this soap", what should she (the student) have written? Write it down immediately!" (TCD) To the extent that such subtle shades of meaning are possible, they may also represent steps towards further subjectification in Traugott's sense, that is extensions towards meanings situated in the speaker's (emotional) attitude and/or belief. As we argue below, the very recurrence of such extensions suggests a natural link between the two types of subjective meanings. 5.
Discussion and conclusions In this paper we have offered an analysis of three conditional constructions in Modern Greek, arguing that they can all be appropriately described as elaborations of the concept of subjectivity. More specifically, we have argued that the semantics of the three conditional markers ama, ean and na involve subjective construals of the situation depicted in the protasis, in the technical sense of the term 'subjective' proposed by Langacker (1990a,b). In taking this view we have departed radically from previous treatments of these markers as free variants and inconsequential synonyms of what is taken to be the basic conditional marker, i.e. an, showing instead that each has its own specialized semantics and conditions of use. As such, they have been shown to represent an important part of Modern Greek grammar. Through the analysis of the three markers, we have also shown that subjectivity, in the sense of an egocentric construal, pertains crucially to the description of conditionality. Ama conditionals are subjective through their placing the speaker within the space set up by the temporal or conditional clause. Ean conditionals are subjective because they represent an extension of an original concrete deictic to an abstract deictic in discourse, by means of which the speaker attempts to focus the listener's attention. Finally, na conditionals, are subjective because they function as deictic, space and time, anchors of the antecedent to the speaker. We have thus identified three different facets of subjectivity as it relates to conditionals, fully exploiting subjective semantics in the analysis of Modern Greek conditional constructions and arguing that subjectivity is really the unifying factor. Although subjective factors have been previously associated with conditionals (e.g. desir ability/affect in Akatsuka 1986, 1997), subjectivity in the technical sense has not, to our knowledge, been applied per se to the description of conditional semantics. Given the frequency of the ama and the tan markers in our corpora, subjectivity emerges as an important factor in the description of Modern Greek
232
KIKI NIKIFORIDOU & DEMETRA KATIS
conditionals and, by extension, perhaps of conditionals more generally. In agreement with previous work, an important conclusion of this paper is that conditionals, besides expressing relations between premises and conclusions, are also, and perhaps more importantly, expressions of subjective beliefs and attitudes. In addition to their basic conditional meanings, all three markers display further meaning extensions. Ama signals in certain cases the speaker's involvement and empathy with the content of the conditional. Na also occasionally signals the speaker's involvement and special interest, although the extension arises in each case via a different route: for ama through the direct observational 'post' of the speaker with respect to the event, while for na through the physical presence of the speaker at the event. Finally, ean extends towards marking interpersonal involvement and from there towards other related functions. It is important to clarify that these extensions are not equally entrenched and conventionalized for all markers; while for ama and na the implicature of speaker involvement may arise in certain contexts but be cancelled in others, for ean interpersonal involvement and/or one of the related extensions seem to be always present. Ean thus resembles more a complex polysemous category with conventionalized extensions — a fact probably related to its being a conditional since its first appearance in Ancient Greek. Finally, we have argued that these extensions to speaker involvement may be treated as instances of subjectification in Traugott's sense, that is as tendencies for meanings to become increasingly situated in the speaker's belief/attitude towards the situation. Akatsuka (1997) notes this tendency in Japanese and Korean conditional markers as well. In the case of the Greek markers, this amounts to an extension towards emotional attitude in the sense of empathy and interpersonal involvement. We have suggested that although these developments do not involve further realignment of any participant or relationship along the subjective axis, they do nevertheless mark the individual speaker's increased 'presence' in coding something highly personal and subjective like empathy or involvement. We further claim that such extensions are fairly motivated given the original egocentric semantics of all three markers. Conditional meanings which code the speaker's vantage point and even physical presence extend towards the expression of involvement, empathy and emotional attitude on the part of that speaker. In this, we suggest, lies a natural link between the different notions of subjectivity discussed in this paper.7
233
SUBJECTIVITY AND CONDITIONALITY
Appendix SCD
TVC
ALT
TCD
Ama
33
(23.1%)
46
(60.5%)
6
(4.7%)
An
78
(54.5%)
29
(38.2%)
121
(93.8%)
35
(58.3%)
Ean
32
(22.4%)
1
(1.3%)
1
(0.8%)
25
(41.7%)
1
(0.8%)
Na Total
143
76
129
60
Table 1. Frequency of markers across different registers
Notes * We wish to thank Eve Sweetser, the two reviewers and the editors for their comments and suggestions. 1 Indeed, both authors acknowledge overlap between the two concepts of subjectivity (cf. Langacker 1990b, Traugott 1995). Their work however tends to focus on data that seem to be clear instances of one or the other. The shift of go from its spatial to its temporal sense is nevertheless a case of subjectification both in Langacker's sense (since according to his analysis (Langacker 1990b), in the temporal sense it is no longer the subject but the speaker/conceptualizer who is mentally tracing a path in order to situate the process in relation to a reference point) and in Traugott's sense, since the temporal sense is obviously speaker-based. 2 The use of hama in example (1) forces in fact an interpretation of the author-historian taking the perspective of the speaker-narrator who is in turn construed as having direct access to the scene described. Hence the informal and at the same time lively style of the Herodotus history. 3 The shift from temporal to conditional may be possibly analyzed as an instance of metaphor. However, we do not here intend to argue for one analysis versus another, but simply to show that this shift, whatever else it may be, is also an example of subjectification in the broad sense. At the same time, we believe that the study of the relationship between metaphor and subjectification, or perhaps more appropriately of metaphors which involve subjectification, may yield important results if undertaken systematically. Similarly, our proposal is not in conflict with Traugott's (1985) suggestion that what motivates the shift from temporal to conditional is that the temporal, just like the conditional, 'frames' the event of the second clause. Indeed, our description may be seen as an elaboration of the idea of framing. 4 Speech act conditionals are those in which the performance of the speech act represented in the apodosis is conditional on the fulfillment of the state described in the protasis. In such conditionals,as shown in Sweetser (1990) and Nikiforidou (1990), the protasis tends to be given more often than in predictive conditionals. 5 Kuno (1993) defines empathy as the speaker's identification with a person or thing that participates in the event or state described in a sentence. Such identification inevitably involves emotional involvement.
234
KIKI NIKIFORIDOU & DEMETRA KATIS
6 The high subjectivity and the discourse-dependent character of ean's meanings may explain why it appears quite late ontogenetically (not at all in the child data examined by Katis 1997 and rarely even at the late age of 11 in our SCD register). 7 Kemmer (1995) and Carey (1995) also discuss possible relationships between the two notions of subjectivity. Kemmer suggests that the two notions are related through the speaker and her role as a "filter of consciousness". According to Carey, both notions of subjectivication highlight different facets of the same process, namely conventionalization of a conversational implicature. Our analysis is closer to Kemmer's, since we see these two notions as distinct (albeit related), arguing more specifically that one is a natural extension of the other. We have, moreover, shown different ways in which the speaker may be included in the scope of predication.
References Akatsuka, Noriko. 1986. "Conditionals and the Epistemic Scale". Language 61:3.625-639. . 1997. "Negative Conditionality, Subjectification, and Conditional Reasoning". On Conditionals Again eds. Angeliki Athanasiadou & René Dirven, 323-354. Amsterdam: John Benjamins. & Patricia Clancy. 1993. "Conditionality and Deontic Modality in Japanese and Korean: Evidence from the Emergence of Conditionals". Japanese/Korean Linguistics ed. Patricia Clancy, Vol.11.177-192. Chicago: University of Chicago Press. Athanasiadou, Angeliki & René Dirven. 1996. "Pragmatic Conditionals". Proceedings of the 9th International Symposium on Applied Linguistics ed. Michalis Milapides, 99-108. Thessaloniki: Aristotle University of Thessaloniki. Brugman, Claudia. 1984. "Very. History of a Hedge". Unpublished manuscript. Linguistics Department. University of California at Berkeley. Carey, Kathleen. 1995. "Subjectivication and the Development of the English Perfect". Subjectivity and Subjectivication eds. Dieter Stein & Susan Wright, 83-102. Cambridge: Cambridge University Press. Chafe, Wallace. 1982. "Integration and Involvement in Speaking, Writing and Oral Literature". Spoken and Written Language ed. Deborah Tannen, 3543. Norwood, N.J.: Ablex. Christidis, Anastasios P. & Kiki Nikiforidou. 1994. "Structural and Crosslinguistic Regularities in the History of Three Particles". Themes in Greek Linguistics eds. Irene Philippaki-Warburton, Katerina Nicolaidis & Maria Sifianou, 169-176. Amsterdam: John Benjamins.
SUBJECTIVITY AND CONDITIONALITY
235
Dancygier, Barbara & Eve Sweetser. 1996. "Conditionals, Distancing and Alternative Spaces". Conceptual Structrure, Discourse and Language ed. Adele E. Goldberg, 83-98. Stanford, Calif.: CSLI Publications. . 1997. "Then in Conditional Constructions". Cognitive Linguistics 8:2.109-136. Fauconnier, Gilles. 1985. Mental Spaces. Cambridge, Mass.: MIT Press. . 1997. Mappings in Thought and Language. Cambridge: Cambridge University Press. Fillmore, Charles. 1990. "Epistemic Stance and Grammatical Form in English Conditional Sentences". Papers from the Twenty-Sixth Regional Meeting of the Chicago Linguistic Society, eds. Michael Ziolkowski, Manuela Noste & Karen Deaton, 137-162. Chicago: University of Chicago. Goldberg, Adele. 1996. "Jackendoff and Construction-based Grammar". Cognitive Linguistics 7:1.3-19. Greenberg, Joseph H. 1986. "The Realis-Irrealis Continuum in the Classical Greek Conditional". On Conditionals eds. Elizabeth Closs Traugott, Alice ter Meulen, Judy Snitzer Reilly & Charles A. Ferguson, 247-264. Cambridge: Cambridge University Press. Haiman, John. 1978. "Conditionals are Topics". Language 54:3.564-589. Holton, David, Mackridge Peter & Irene Philippaki-Warburton. 1997. Greek: A Comprehensive Grammar of the Modern Language. London & New York: Routledge. Joseph, Brian & Irene Phillipaki-Warburton. 1987. Modern Greek. London: Croom Helm. Katis, Demetra. 1997. "Conditional Markers in Greek". Proceedings of the 2nd International Conference on Greek Linguistics eds. Gabrieli . Drachman, Angeliki Malikouti-Drachman, John Fykias & Cylla Klidi, Vol.1.287-296. Salzburg: University of Salzburg. Kemmer, Suzanne. 1995. "Emphatic and relexive -self, expectations, view point and subjectivity". Subjectivity and Subjectivication eds. Dieter Stein & Susan Wright, 55-82. Cambridge: Cambridge University Press. Kriaras, Emmanuil. 1969. Lexico tis neoellinikis kai dimodous ellinikis grammatias (A Dictionary of Medieval Greek Popular Literature), Vol.1. Thessaloniki: Kyriakidi. Kuno, Susumu. 1993. Functional Syntax: Anaphora, Discourse and Empathy. Chicago: University of Chicago Press. Lakoff, George. 1987. Women, Fire and Dangerous Things: What categories reveal about the mind. Chicago: Chicago University Press.
236
KIKI NIKIFORIDOU & DEMETRA KATIS
Langacker, Ronald. 1987. Foundations of Cognitive Grammar, Vol I Theoretical Prerequisites. Stanford: Stanford University Press. . 1990a. Foundations of Cognitive Grammar, Volli, Descriptive Applications. Stanford: Stanford University Press. . 1990b. "Subjectification". Cognitive Linguistics 1:1.5-38. Liddell, George H. & Robert Scott. 1940/1977. A Greek-English Lexicon. Revised by Henry S. Jones, Oxford: Clarendon Press. Nikiforidou, Kiki. 1990. Conditional and Concessive Clauses in Modern Greek: A Syntactic and Semantic Description. Ph. D. Dissertation. University of California, Berkeley. Schiffrin, Deborah. 1992. "Anaphoric Then: Aspectual, Textual, and Epistemic Meaning". Linguistics 30.753-792. Sweetser, Eve. 1990. From Etymology to Pragmatics: Metaphorical and Cultural Aspects of Semantic Structure. Cambridge: Cambridge University Press. Tabakowska, Elzbieta. 1997. "Conceptualization: Conditionals as an Instance of Figure-Ground Alignment". On Conditionals Again eds. Angeliki Athanasiadou & René Dirven, 323-354. Amsterdam: John Benjamins. Talmy, Leonard. 1988. "Force Dynamics in Language and Thought". Cognitive Science 12.49-100. Tannen, Deborah. 1985. "Relative Focus on Involvement in Oral and Written Discourse". Literacy, Language and Learning: The Nature and Consequences of Reading and Writing eds. David R. Olson, Nancy Torrance & Angela Hildyard, 124-147. Cambridge: Cambridge University Press. Traugott, Elizabeth C. 1982. "From Propositional to Textual and Expressive Meanings: Some Semantic-pragmatic Aspects of Grammaticalization". Perspectives in Historical Linguistics eds. Winfred P. Lehmann & Yakov Malkiel, 245-271. Amsterdam: John Benjamins. . 1985. "Conditional Markers". Iconicity in Syntax ed. John Haiman, 289-310. Amsterdam: John Benjamins. . 1989. "On the Rise of Epistemic Meanings in English: An Example of Subjectification in Semantic Change". Language 65.31-55. . 1995. "Subjectification in Grammaticalization". Subjectivity and Subjectivization in Language eds. Dieter Stein and Susan Wright, 31-54. Cambridge: Cambridge University Press. Tzartzanos, Achilleas. 1956. Syntactikon tis Archeas Ellinikis glossis (Syntax of the Ancient Greek Language). Athens: OESB.
SUBJECTIVITY AND CONDITIONALITY
237
Van der Auwera, Johan. 1986. "Conditionals and Speech Acts". On Conditionals eds. Elizabeth Closs Traugott, Alice ter Meulen, Judy Snitzer Reilly & Charles A. Ferguson, 197-214. Cambridge: Cambridge University Press. Van Dijk, Teun Α. 1981. Studies in the Pragmatics of Discourse. The Hague: Mouton. Watkins, Calvert. 1985. The American Heritage Dictionary of Indo-European Roots. Boston: Houghton Mifflin Company.
English Imperatives and Passives Hidemitsu Takahashi Hokkaido University, Sapporo, Japan 1.
Introduction This paper will discuss phenomena pertaining to English imperatives in the passive form such as those in (1) and (2) below: (1) a. *George, be taken to church by your sister b. *Be helped by Jill (2) a. Be checked over by a doctor, then you'll be sure there's nothing wrong b. Be flattered by what he says, it'll make his day The aim of this paper is to explain why passive imperatives like those in (1) are unacceptable, whereas those in (2) are better-formed. Early generative grammar work claimed that English imperatives disallow the passive form syntactically (cf. Lees 1964), based on facts like those in (1) above. However, later pragmatic and descriptive work found that this restric tion is far from absolute (cf. Bolinger 1977; Davies 1986). Given appropriate contextualizations, passive imperatives can be rendered perfectly acceptable. However, the previous analyses fell short of accounting for the full range of felicitous imperative sentences in the passive form in a unified way. This paper will argue that Cognitive Grammar notions, notably the degree of force, event model, prototype effect and constructional meaning, are useful in providing a unified treatment for the problem under investigation. First, the paper will illustrate the way in which English imperatives can be characterized within the framework of currently available cognitive grammar. Second, it will examine the extent to which such an analysis will explain the old problems pertaining to the acceptability of imperatives in the passive form. It will be argued that the passive may occur in the imperative provided that the former and the latter are conceptually compatible. This suggests that the imperative does not clash with the passive on syntactic terms but rather in terms of prototypes.
240
HIDEMITSU TAKAHASHI
2.
English imperatives and their four features In English, the 'imperative' refers to a construction which occurs only in the main clause, normally has no grammatical subject and contains the verb in its most basic form, as illustrated in Get up and go or Call Sharon tonight. In my previous work (Takahashi 1994), I characterized English imperatives in terms of four essential features, hypotheticality, non-past, second-person and force exertion.1 The first three notions are largely self-evident, so let me briefly explain only the last feature, force exertion. This is a mental notion, not directly observable from surface linguistic structure, as defined in (3) below (cf. Takahashi (1994: 375)): (3) Force Exertion: the degree of (directive) force that the speaker is applying (at the utterance time of an imperative) toward the addressee's carrying out the action. Here the term force is used in the sense of 'force dynamics' utilized and discussed in Talmy (1988). The idea of (3) is that the utterance of an imperative expresses the speaker's exertion of psychological, interpersonal force toward the addressee's performing an action, and that this force lends itself to a gradient, rather than an all-or-nothing, conception. Consider: (4) a. Sleep until noon b. Sleep until noon; you're tired Sleep until noon, and you'll miss lunch The preferred interpretation of (4a) is a positive sense such as command, advice, etc., as is the case in (4b). The identical form in (4c), however, nor mally conveys both a condition and warning. While the sentence conveys that 'If you sleep until noon, then you'll miss lunch', it also communicates the speaker's warning against the addressee's oversleeping. The aim for introdu cing the notion of force exertion resides in providing a unified treatment for these positive as well as negative readings. Figure 1 below, based on Figure 1 in Takahashi (1994: 377), might quantify the gradient notion of force exertion. This figure argues for the following points. First, an imperative may in principle be issued at any point on the continuum. The higher on the conti nuum, the more directive the imperative becomes, whereas the lower, the more conditional. Second, an imperative is potentially ambiguous concerning the val-
ENGLISH IMPERATIVES AND PASSIVES
moreconditional
241
moredirective
Figure 1. Imperatives and degree of Force Exertion
ue of force exertion — ranging between plus maximum [ + 1] and minus maximum [-1] including [0]. Third, an imperative like (4b) involves a high degree of force exertion; in a default-case context, an imperative like (4a) strongly evokes a high degree of force exertion, a sense considered proto typical of imperatives. Finally, the scale of force exertion accounts for why imperatives can be taken as commands or warning: (near) [ + 1] force exertion would yield a command while (near) [-1] exertion an admonition or a warning. Being highly schematic, the notion of force exertion is designed to cover a wide range of uses including 'pseudo-imperatives'. From this perspective, the 'command' sense can be considered the prototypical sense, rather than an intrinsic feature, of an imperative. Clean up this mess right now is a proto typical imperative involving a near-maximum exertion of directive force. Spare the rod and spoil the child is an atypical example involving a minus degree of force; the force is against, not on, the hearers' act of sparing and spoiling. 'Pseudo-imperatives' such as those in Put a proper Bostonian on the phone with a Texas oilman (and here comes miscommunication), which convey a straight condition, classify as peripheral, since hardly any force is exerted (see 3.2 below). The framework here also explains the humor in the bumper-sticker sign, Eat, drink and get fat; here the force exertion shifts from positive to negative.2 3.
A cognitive (image-schematic) model of imperatives This section will characterize the above four features within the framework of cognitive grammar. Beginning with an Action Chain analysis, we will come up with a Canonical Event Model of imperatives (both based on Langacker 1991), a model which I claim will permit a fuller analysis of imperatives.
242
HIDEMITSU TAKAHASHI
Figure 2. Action Chain and the imperative
3.1. Action chain model Cognitive Grammar work (for example Talmy 1988, Langacker 1987; 1991, Croft 1991, inter alia) contends that the semantic content of clause structure essentially resides in the conceptions of entities and their energetic interactions with respect to one another. Let me illustrate how one such model, which Langacker (1991: 283) labels as 'Action Chain (Billiard-ball) Model', will characterize the imperative with a transitive verb such as (5) below: (5) Call Sharon tonight. As is standard practice, bold circles and arrows are used for profiled conceptions, and light ones for unprofiled conceptions. The idea is that: (i)
The imperative is comprised of two separate (though interrelated) subevents, which combine to form a dynamic chain of actions; (ii) The speaker and addressee both participate in this action chain as indispensable entities — the speaker as head, the addressee as a second entity engaged in further action; (iii) The imperative makes explicit only the addressee's action and leaves implicit the two key entities; nor does it overtly code the application of directive force. It seems obvious that this action chain analysis captures some important facets of the imperative, namely its second-person nature and the concept of force exertion. However, the analysis is not capable of adequately handling two other features — hypotheticality and non-past. Moreover, it does not specify the semantic roles of participants (speaker and addressee). A fuller characterization will require a more elaborate and finely-grained model such as the 'Canonical Event Model' as introduced in Langacker (1991: 285).
ENGLISH IMPERATIVES AND PASSIVES
243
Figure 3. Canonical Event model (transitive)
3.2. Canonical event model The Canonical Event Model takes into account several semantic notions not dealt with in the Action Chain Model, namely, setting, semantic role and viewer, as sketched in Figure 3. The idea of this figure is that the simplest transitive clause contains two participants which play semantic roles, agent and patient, respectively. These participants are engaged in a kind of energetic interaction in a specific concep tual domain called a 'setting' chosen to be highlighted by a conceptualizer A transitive clause such as John ate the apple in the kitchen, where John acts as an agent and participates in an energetic interaction of eating an apple which is patient in the 'kitchen' setting, would be a straightforward illustration of this model. To obtain a Canonical Event Model of the imperative, let us specify the setting and the semantic roles of participants.3 As for the setting, the first subevent is restricted to an extremely narrow speech situation, the here-andnow of speaking, only populated by the speaker and addressee. It would be reasonable then to analyze this subevent as occurring in a deictic setting. On the other hand, the second subevent (or the addressee's action) is hypothetical in nature, regardless of whether or not the addressee's act will be realized in the objective world. For this reason, the second subevent may be treated as taking place in a hypothetical setting (or mental space).4 Next, let us consider the speaker and addressee's semantic roles. To begin with, it would not be a gross oversimplification to regard the speaker and addressee as both Agents, since after all both individuals are 'doers'; they typically initiate a process, engaged in some volitional activity. It must be emphasized that natural language affords multiple kinds of agency, and varying degrees of agency (cf. Delancey 1984; 1985), and that the grammatical subject (or 'an argument') may readily play more than one semantic role at once.
244
HIDEMITSU TAKAHASHI
Figure 4. Canonical Imperative Event Model (transitive)
Considering these factors, the speaker can be analyzed in terms of causeragent, and the addressee in terms of causee-agent. The rationale behind this treatment is that in the prototypical scene of an imperative, the addressee is triggered to act or undergo a substantial change in state by the utterance of an imperative.5 In light of these discussions, the Canonical Imperative Event Model would be something like Figure 4. This model incorporates and refines the Action Chain Model in Figure 2 above, by adding the notions of setting and semantic role. Two settings are assigned to two corresponding subevents as indicated in the thin squares, where the addressee belongs in these two settings at once. The time of event 1 coincides with the time of utterance, and event 2 occurs only after event 1. Although the figure appears to only illustrate the prototype command imperative, the analysis here is essentially flexible in nature, intended to accommodate the whole range of uses — central cases as well as departures from the norm. This model, which characterizes the prototypical imperative, can be stated as in (6) below. (6) The prototype IMPERATIVE i. The speaker exerts a high (near [ + 1]) directive force in deictic setting toward the addressee, who will thereby perform an action in hypothetical setting. ii. The speaker plays the semantic role of causer-like agent, and the addressee causee-like agent. More generally, the model in Figure 4 defines the imperative as in (7) below:
ENGLISH IMPERATIVES AND PASSIVES
245
(7) The schematic IMPERATIVE i. The speaker exerts a varying degree of force (ranging from [ +1] to [-1]) in deictic setting toward the addressee, who will thereby be engaged in a certain situation in hypothetical setting. ii. The speaker and addressee may play an agentive as well as nonagentive role. That is, there is intention/volition on the part of the speaker to cause the hearer of an imperative to make the propositional content true, and the pressure/force can be more or less strong, and positively as well as negatively exerted. One obvious advantage of the Imperative Event Model is that it captures the complete range of the four features of English imperatives discussed in Section 2 in a more integrated manner. Based on Figure 4 and (6) above, we can isolate the component parts constituting the notion of imperative prototypicality, identifying the following three parameters, each of which suggests a scale according to which imperative clauses can be ranked: (8) A. FORCE NATURE DEGREE B.
SUBJECT SEMANTIC ROLE IDENTITY
C. DYNAMICITY
PROTOTYPICAL
NON-PROTOTYPICAL
directive (near) [+1]
non-directive lower than (near) [+1]
agentive individuated high
non-agentive generic low
Considering the most obvious association between the imperative and the notion of force, the degree of force (in row A) can be taken as a primary parameter in the assessment of imperative prototypicality. As could be expec ted, the features in the rows are to a large extent interrelated. Strong force (prototypical in row A), for example, tends to vary with agency (prototypical in row B), which is also likely to covary with high dynamicity (prototypical in row C). However, the values do not necessarily covary within a given imperative. Weak force (non-prototypical in row A) may covary with low (non-prototypical in row C) as well as high dynamicity (prototypical in the same row):
246
HIDEMITSU TAKAHASHI
(A) Imperatives are atypical when the speaker's force is non-directive (or mixed with other kinds of force). In line with our pretheoretical conception of the imperative, I treat cases of the absence of force as most deviant from the norm, hence peripheral, and cases of negative force as more non-prototypical than peripheral. (B) When the subject is a non-agent such as an experiencer, theme or patient or when the understood subject is generic as in Shake before using, the imperative progressively deviates from its norm. (C) Events (highly dynamic situations) better fit the image schema of prototype imperatives than states as in Be sick (and they'llput you in bed). Compare: (9) a. Send me a fax tomorrow. OK? [prototypical] b. You want to get to the post office? Go straight ahead two blocks [non-prototypical] Spare the rod and spoil the child [non-prototypical] d. Regional accents can be a problem. Put a proper Bostonian on the phone with a Texas oilman and here comes miscommunication [peripheral] (9a) is a paradigm example of imperative, satisfying all the three criteria: the force is both directive and strong, the whole sentence deals with an event, with an individuated, agentive subject. The imperatives in (9b) and (9c) classify as non-prototypical on subtly different grounds. In (9b), in which the speaker is giving a direction, the presence of force is not obvious (nonprototypical in row A), whereas the subject is construed as an agent (nearprototype in row B) participating in an action (prototypical in row C). The pair of imperatives in (9c) are ambiguous between straight condition (the absence of force) and warning (minus force), hence non-prototypical in row A. The understood subject is non-individuated (non-prototypical in row B). In either case, the subject is agentive and the whole sentence deals with an event rather than a state (prototypical in row C). Finally, the purely hypo-thetical conditional Put a proper Bostonian on the phone with a Texas oilman in (9d) ranks very low in rows A and B. Crucially, the clause does not involve any force exertion, nor is much agency involved; the clause has a generic subject while dealing with an action. The imperative in (9d) classifies as peripheral on the grounds that the clause extremely deviates from the imperative prototype in the major parameters.
ENGLISH IMPERATIVES AND PASSIVES
247
4.
English imperatives and passives We are now in a position to turn to the central question: why are passive imperatives in (2) acceptable while those in (1) are odd? (1) a. *George, be taken to church by your sister b. *Be helped by Jill (2) a. Be checked over by a doctor, then you'll be sure there's nothing wrong b. Be flattered by what he says, it'll make his day
Many scholars have pointed out that passives with be occur more frequently in negative imperatives than positive (cf. Quirk et al. 1985: 827), and I focus more on positive imperatives in be-passives here. It is well established that an active clause more readily passivizes if it is more transitive, i.e. when the patient is directly affected by the activity in question (cf. Bolinger 1977, Shibatani 1985, Rice 1987, Langacker 1991, inter alia). In this connection, most previous works on passives are in general agreement concerning the following three points. First, the passive critically involves an agent (or something close to it such as an experiencer), which is invariably defocused either syntactically or lexically (cf. Shibatani 1985: 832). Second, the passive with be is in principle more static than its active counterpart, although this does not mean that the be-passive only refers to a state (cf. Givón 1981; 1990). Third, the subject tends to be a patient (cf. Jespersen 1924; Givón 1981, 1990). Taking these findings into account, I suggest the following as the prototype PASSIVE (the features indicated in italics constitute the main parameters): (10) The prototype PASSIVE i. The subject is directly affected by an external agent ii. The subject plays the semantic role of patient Now, we obtain the following parameters for passive prototypicality, according to which passive clauses can be ranked.6
248
HIDEMITSU TAKAHASHI
(11)
A. SUBJECT
PROTOTYPICAL
NON-PROTOTYPICAL
patient
non-patient or mixed with other role(s) non-physical non-agent (such as experiencer)
B. AFFECTEDNESS physical C. DEFOCUSED agent PARTICIPANT
The values in these three parameters are intimately interrelated. The subject's patienthood (in row A), the defocused participant's agency (in row C) and overall affectedness (in row B) normally go hand in hand. One almost automatically follows from another, since prototypical patient is a participant absorbing the energy transmitted from without and thereby undergoing a change in state (cf. Langacker 1991: 285). Conversely, the passive departs from the prototype when the subject is not a prototypical patient, when the defocused participant might be viewed as an experiencer rather than an agent, or when not much affectedness is involved in the designated situation. Compare the following passives with human subjects: (12) a. She was taken to a hospital b. She is well liked (by everyone) c. ?I was approached by the tram
[prototypical] [non-prototypical]
(12a) constitutes a paradigm example of passive, in which the subject entity is moved to a different location, which means that the subject is a patient, directly affected by the action via the force transmitted from an external agentive source. (12b) exhibits a non-prototypical passive; the subject is a non-patient, the defocused entity (everyone) seems an experiencer, and no substantial effect is produced. (12c) without context strikes us as bizarre because the sentence extremely deviates from the 'PATIENT DIRECTLY AFFECTED (BY AGENT)' conception characterizing the passive prototype. It only deals with the geometry of two positions.7 Careful examination of the characterizations of imperatives in Figure 4 and (8) and passives in (10) and (11) above reveals a serious incompatibility between the imperative prototype and the passive prototype. Crucially, the semantic role of imperative subjects and that of passive subjects differ. Prototype imperatives demand an agentive subject, whereas passives prefer a patient. Agent and patient are normally two mutually exclusive semantic roles. Less seriously, the degree of dynamicity is different. The imperative prefers a dynamic reading, whereas the passive with be inherently stativizes the
ENGLISH IMPERATIVES AND PASSIVES
249
conception of the event, although be-passives may also vary in the degree of dynamicity.8 Now, we come up with the following working hypothesis: (13) Hypothesis The passive construction does not clash with the imperative syntactically but on conceptual grounds, i.e. the clash occurs between prototypes. The data in (1) substantiate this claim. A semantic conflict occurs between the 'ideal' imperative vs. the 'ideal' passive. A linguistic construction has a strong tendency to evoke its most natural reading in isolated contexts, i.e. its prototype. Interpreted as a PASSIVE, the subject of the construction Be taken to church by your sister should be a patient. Interpreted as an IMPERATIVE, however, it should be an agent. The subject slot in (la) demands both a patient and an agent at the same time, two semantic roles which are normally at odds with each other.9 In addition, the lexical content increases the conflict. An individual who is taken to church by another individual is interpreted as a participant close to prototypical patient, since s/he is physically affected by an external source such as an agent (by your sister). That is, the sentence lexically evokes the 'PATIENT AFFECTED BY AGENT' reading unique to prototype passives. Overall, the prototypicality of PASSIVE is both constructionally and lexically emphasized here (cf. Goldberg 1995, Rice 1987).10 A similar account holds for (lb) as well. Next, let us consider why the passive imperatives in (2a) and (2b) are better-formed.11 Previous work employed the notion of '(self)controllability' (Davies 1986) or 'willfulness' (Bolinger 1977). Note that controllability and willfulness are among central conceptions constituting the notion of agency. The approach followed in this paper naturally incorporates these pragmatic accounts. In fact, the present analysis will succeed where previous accounts fail, as later discussion will reveal. As we have observed, PASSIVE and IMPERATIVE do not readily merge, since the two constructions share very few characteristics in their prototypes. In (2a), however, the lexical content along with contextual assist serves to alleviate the conflict. While an individual who is examined by a doctor appears to involve little agency, the additional context provided by the subsequent clause you'll be sure there's nothing wrong makes it apparent that the addressee is advised to act, i.e. to see a doctor with a clear purpose — for check-up. S/he will not be forced to be examined; s/he chooses to act on his/her own will instead, which means that a great deal of agency is
250
HIDEMITSU TAKAHASHI
contextually supplied. Similarly, (2b) is best paraphrased as 'Try/Appear to be flattered... ', a reading suggestive of agency, although be flattered itself is not an action and the state of being flattered is not something that can intentionally be brought about. Viewed as passives, therefore, both (2a) and (2b) are nonprototypical in that the subject is not a straight patient in (2a) or not a patient at all in (2b). It is an active agentive patient in the former and a subtly agentive experiencer in the latter. Viewed as imperatives, (2a) and (2b) might classify as near-prototype and non-prototype, respectively. (2a) can be treated as near-prototypical (if not prototypical) because the imperative evokes a conceptual structure close to the 'FORCE-CAUSES-AGENT-TO-ACT' schema characterizing the imperative proto type. (2b) departs somewhat more from the prototype than (2a) in that the force/ pressure seems not so strong as in (2a), with the subject's agency only subtly implied. Besides, (try/appear to) be flattered is not obviously an action to the extent of being checked over by a doctor. In a nutshell, both (2a) and (2b) are acceptable on the ground that non-prototypical exemplars of PASSIVE are acceptable within near-prototypical or non-prototypical IMPERATIVE. Not surprisingly, even in cases when be-passives are perfectly acceptable, speakers of English prefer to use get-passives if available: (14) Get checked over by a doctor, then you'll be sure there's nothing wrong Here the intrinsic dynamicity and agency of get (cf. Givón and Yang 1994: 139) makes the get-passive far better conform to the conception of the imperative, a construction which favors an active, (causee-like) agent subject. Groups , and below list at least three distinct patterns in which passives fare well within imperatives. Interpreted as passives, all the examples in these groups are non-prototypical, predominantly with an experiencer subject, except for (2a) and probably (15), whose subject is more patient-like. In most examples, the passive imperatives more or less deal with mental, (as opposed to physical) states, and not much affectedness is involved. Interpreted as imperatives, the three groups exhibit varying degrees of deviation from the prototype. Recall that imperative prototypicality is determined by three factors: the nature/degreeof force (row A), the agency/identity of the subject (row B), and dynamicity (row C) (cf. Section 3). Group A roughly satisfies all the three (hence, near-prototypical), Group to a lesser extent (hence, nonprototypical) and Group almost none (hence, peripheral). Group A includes set phrases such as (15) and negative forms like (16) below, cases frequently pointed out in the previous literature:
ENGLISH IMPERATIVES AND PASSIVES
251
Group A: near-prototype IMPERATIVE & non-prototype PASSIVE (2a) (15) Be prepared / Be warned (16) Don't be fooled by his act / Don't be taken in by the doctor, he's really a quack On the one hand, the examples in Group A exhibit near-prototypical imperatives, since they involve a strong, directive force causing the subject to act or behave in a certain way, although the conceived situation is not strictly an action in its truest sense of the word. On the other hand, the sentences in this group present non-prototype passives, since the subject is not obviously a patient; nor is there much affectedness involved. Note that the don't form in (16), for example, greatly reduces the subject's patienthood, as compared to its affirmative counterpart, and instills a great deal of agency. All this means is that non-central instances of the PASSIVE merge successfully with the nearprototype IMPERATIVE. The examples in Groups and present a serious challenge to previous pragmatic accounts, since they contain cases in which the imperatives are perfectly well-formed without involving any 'controllability' or 'willfulness'. Group is comprised of two subgroups, in both of which the passive impera tive needs contextual assist to be fully acceptable. In the two acceptable senten ces of (18), for example, the passive imperative typically occurs immediately after an ordinary command imperative: Group B: non-prototype IMPERATIVE & non-prototype PASSIVE (2b) (17) a. Oh, come on; be taken in just once — it isn't going to hurt you; do you think you're perfect? [Bolinger 1977] b. Be impressed by his stamp collection if you want him to like you! [Dixon 1994] (18) a. Come to Palm Court and be entertained by Joe Loss [Dixon 1994] and his orchestral a'. ?Be entertained by Joe Loss and his orchestra! b. Come to my office, be amazed at the mess I've made b'. ?Be amazed at the mess I've made Viewed as passives, all the italicized examples in Group are nonprototypical, exactly like Group A. As we have observed above concerning
252
HIDEMITSU TAKAHASHI
(2b), the passive imperative here involves an experiencer subject engaged in a nonphysical (mental) situation. To take an instance of (18a), the addressee is hardly conceptualized as a patient; s/he might be affected — but only mentally or emotionally, as suggested in the likely paraphrase Come to Palm Court and ENJOY Joe Loss and his orchestra! Viewed as imperatives, the clauses in question are also non-prototypical. However, the examples in (2b) and (17) and those in (18) exhibit differing degrees of non-prototypicality. The second subgroup is more deviant than the first. The imperatives in (2b) and (17) contain positive force exerted in order to cause the addressee to behave on his/her own will. To take an instance of Be impressed by his stamp collection in (17b), the immediately following //-clause suggests that the speaker asks the hearer to pretend to be impressed, a reading implying deliberate action, i.e. agency. In this respect, the subject retains control over the proposed act/situation, although be impressed itself is a state, and not something that can intentionally be brought about. In contrast, the sense of agency greatly decreases in (18). The italicized passive imperative conveys not so much a command as an outcome assumed to result from the action designated in the immediately preceding imperative. The passive imperative here, which might be termed CONSEQUENT IMPERA TIVE, does not involve any sense of controllability or willfulness, since the addressee is incapable of choosing to be entertained or be amazed. Quite unlike (17b), the hearers/readers are not asked to pretend to be entertained or be amazed either. Rather, they are simply asked to visit so that they will be entertained or amazed. In such a case, the subject is hardly an active, agentive subject. Nevertheless, a positive force still seems present, since the speaker is attempting to cause the hearer(s) to visit and be entertained or amazed. In this respect, the imperatives in (18) depart more from the prototype than those in (2b) and (17). Despite this difference in agency, the two subgroups might appropriately be subsumed under the rubrics of non-protototype IMPERATIVE in the important respect that the imperative contains the exertion of force. All the passive imperatives in Group fare well precisely because imperatives and passives divert considerably from their respective prototype. Finally, Group remains to be a non-prototypical passive with an expe riencer subject. As an imperative, however, this group departs further from the prototype than in Group B. While more than one reading is possible, the imperatives in (19a) and (19b) may obtain a pure conditional (IF-THEN) reading, a reading without any notion of force or agency:
ENGLISH IMPERATIVES AND PASSIVES
253
Group C: peripheral IMPERATIVE & non-prototype PASSIVE (19) a. Be pleased to see him, you'll make his day b. Be impressed by his stamp collection, you will make his day The examples in Group suggest that the imperative may permit a nonagentive, experiencer subject without containing any causal force only when it borders on the periphery. As a result, no serious conflict occurs in merger between IMPERATIVE and PASSIVE. 5.
Japanese imperatives and passives Finally, let me briefly discuss the imperative in Japanese to examine the cross-linguistic implications of the present analysis. In Japanese, a clause's category as an imperative is explicitly spelled out by an imperative marker, which invariably indicates the exertion of force as both directive and strong. As a result, Japanese imperatives are far more restricted as to the range of force exertion; the standard imperative morphology nasai is especially resistant to absence of force imperatives. Readings other than prototypical command are harder to obtain; only the conditional command, not the straight condition, is allowed, and the generic subject is not normally applicable except in somewhat archaic context. To illustrate, while sentences (9a) and (9b) above in Section 3 can be expressed in imperatives in Japanese, (9c) and (9d) translate into conditionals and/or declaratives, not imperatives. See also (20) below, in which (20a) and (20b) exhibit Japanese counterparts for the English imperatives in (4b) and (4c) in Section 2: (20) a. hiru made noon until "Sleep until b. ?hiru made noon until "Sleep until hiru made noon until "Sleep until
ne nasai, tsukarete iru kara sleep IMP, tired are because noon; you're tired" ne nasai, (soshitara) gohan nakunaru sleep IMP, then lunch be gone will noon, and you'll miss lunch" ne nasai, (soshitara) genkiniru sleep IMP, then feel better will noon, and you'll feel better"
While the nasai imperative is acceptable in (20a), it is only marginally acceptable in (20b) (in a sarcastic context) involving minus force but generally more natural-sounding in (20c) involving plus force.
254
HIDEMITSU TAKAHASHI
Passives in Japanese behave somewhat like their English counterparts as far as the examples in (12) above are concerned. Japanese counterparts for (12a) and (12b) are both acceptable, as prototypical and non-prototypical, respectively, whereas the one corresponding to (12c) is just as odd. This fact suggests that Japanese passives also center around the 'PATIENT AFFECTED BY AGENT' schema as its prototypical conceptual structure, although the mani festations of deviation may pattern differently. Not surprisingly, the occurrence of the passive form rare with the imperative form -nasai is extremely restricted, if not totally ungrammatical, mainly because Japanese imperatives do not readily exhibit non-prototypical instances. The majority of felicitous passive imperatives in English we have observed in Groups A, B, and translate into active sentences in Japanese. The only pattern in which the passive combines with the imperative seems to be the one in Group A: near-prototype IMPERATIVE & non-prototype PASSIVE. And the negative form is preferable, as the examples in (21) below illustrate: (21) a. sono isha ni damas -areru na the doctor by taken in be-PASS NEG-IMP "Don't be taken in by the doctor" b. (?)kanojo ni suk -are nasai her by liked be-PASS IMP "Be liked by her" The Japanese imperative in the passive form like (22a) below is simply ungrammatical, although the English counterpart of it can be rendered accep table if provided with an 'appropriate context' implicating agency and deli berate action: (22) a. *isha ni mirare nasai doctor by examined be IMP "Be checked over by a doctor" The only way to improve the sentence in (22a) is to employ the form -morau, a verb of receiving used to indicate the gaining of some benefit from the agentive action described in the -te verbs:
ENGLISH IMPERATIVES AND PASSIVES
255
(22) b. isha ni mite morai nasai doctor by examined have IMP "Have (yourself) examined by a doctor" The fact that the passive form rare cannot combine with the imperative form nasai can be explained in terms of a conflict in the subject's semantic role — between patient vs. agent. The phenomenon we obtain here might be viewed as just another instance in which the PASSIVE and IMPERATIVE clash between prototypes. This paper has attempted to offer a principled treatment for both felicitous and infelicitous passive imperatives as well as subsuming important insights from previous syntactic and pragmatic accounts. The main point of the analyses made here is that passives and imperatives clash in their prototypical conceptual structure but become conceptually compatible if they are used in a non-prototypical or peripheral sense. The once alleged 'syntactic' restriction can be viewed as the result of a mismatch between the prototypical conception of passives and that of imperatives. Notes 1 In Takahashi 1994, I used the term speaker commitment to capture the notion of force exertion. To avoid possible misunderstandings, I use the latter term in this paper. 2 The framework here is also flexible enough to capture the blend between the infinitive and the imperative: (i)
My advice to international travelers is: Never arrive in a foreign country without some ready cash or travelers checks in that country's currency
The italicized clause can read either with or without the conception of force. 3 The viewer is omitted here because this notion does not concern the problem under investigation. 4 Within a Mental Space framework (Fauconnier 1985), an imperative can be analyzed in terms of space builder, which evokes a hypothetical space containing a variety of entities including the addressee, and the parent space is a deitic space. Crucially, the actor in the hypothetical space corresponds to the addressee in the parent space. I believe the analysis in Figure 4 is fully compatible with, and incorporates, this mental space analysis. I thank Arie Verhagen (at the fifth ICLA conference) and Joseph Tomei (p.c.) for suggesting to me the implication of the present model for the Mental Space semantics. 5 Note that the case of double roles here is crucially different from the cases discussed by Schlesinger 1989, or Jackendoff 1990 where argument is assigned an action role and a thematic role in the same event. A reviewer suggests that it is closer to the cases of causative predicates having a composite argument structure (cf. Alsina 1992 or Matsumoto 1996: chap.6), in which a causee is a patient of the causing event and the agent (and others) of the caused event.
256
HIDEMITSU TAKAHASHI
6 The stativizing function of be-passives is not integrated here because it is something inherent rather than a matter of prototypicality. I am grateful to two reviewers for the suggestion that the parameters defining imperative prototypicality and passive prototypicality be explicitly stated. 7 Compare I was approached by a faculty member yesterday, which is perfectly acceptable with implied sense of patienthood and effectiveness. 8 For example, Croft (1994: 113) distinguishes among process passive, anti-passive, and stative passive. 9 According to Saksena (1980), the 'affected' agent may appear in some languages, a fact suggesting that Agent and Patient do not constitute single primitive categories which are mutually exclusive. 10 I assume that a construction carries an inherent semantic content independent of the information from lexical items (cf. Goldberg 1995). 11 I would like to thank Peter Grundy (at the ICLA conference) for informing me that sentence (2a) is not acceptable in British English.
References Alsina, Alex. 1992. "On the Argument Structure of Causatives". Linguistic Inquiry 23.517-555. Arce-Arenales, Manuel, Axelrod, Melissa & Barbara Fox. 1994. "Active Voice and Middle Diathesis: A Cross-Linguistic Perspective". In: Voice: Form and Function, eds. . Fox & P. Hopper, 1-23. Amsterdam & Phila-delphia: John Benjamins. Bolinger, Dwight. 1977. Meaning and Form. London: Longman. Croft, Bill. 1990. Syntactic Categories and Grammatical Relations: The Cognitive Organization of Information. Chicago: University of Chicago Press. --------. 1994. "Voice: Beyond Control and Affectedness". In: Voice: Form and Function, eds. . Fox & P. Hopper, 90-117. Amsterdam & Philadelphia: John Benjamins. Davies, Eirlys. 1986. The English Imperative. London: Croom Helm. Delancey, Scott. 1984. "Categories of Non-Volitional Actor in Lhasa Tibetan." Proceedings of the Conference on Participant Roles : South Asia and Adjacent Areas. eds. Ariene R.K. Zide, David Magier & Eric Schiller, 58-70. Bloomington: Indiana University Linguistics Club. ---------. 1985. "Agentivity and Syntax". Papers from the Parasession on Causatives and Agentivity. eds. William H. Eilfort, Paul D. Kroeber & Karen L.Peterson, 1-12. Chicago: Chicago Linguistics Society. Dixon, R.M.W. 1994. Ergativity. Cambridge: Cambridge University Press. Fauconnier, Gilles. 1985. Mental Spaces: Aspects of Meaning Construction in Natural Language. Cambridge, Mass. & London: MIT Press/Bradford.
ENGLISH IMPERATIVES AND PASSIVES
257
Fox, Barbara & Paul J. Hopper. 1994. Voice: Form and Function. Amsterdam & Philadelphia: John Benjamins. Givón, Talmy. 1981. "Typology and functional domains". Studies in Language 5.163-193. ---------. 1990. Syntax: A Functional-Typological Introduction. Vol. II. Amsterdam & Philadelphia: John Benjamins. --------- & Yang, Lynne. 1994. "The Rise of the English GET-Passive". In: Voice: Form and Function, eds. . Fox & P. Hopper, 119-149. Amsterdam & Philadelphia: John Benjamins. Goldberg, Adele. 1995. Constructions: A Construction Grammar Approach to Argument Structure. Chicago: University of Chicago Press. Jackendoff, Ray. 1990. Semantic Structures. Cambridge: Cambridge University Press. Jespersen, Otto. 1924 The Philosophy of Grammar. London: Allen & Unwin. ---------. 1961. A Modern English Grammar, Part V. London: Allen & Unwin. Langacker, Ronald W. 1987. "Nouns and Verbs". Language 63.53-94. ---------. 1991. Foundations of Cognitive Grammar, Vol.2, Stanford: Stanford University Press. Lees, Robert . 1964. "On Passives and Imperatives". Gengo 46.2841. Matsumoto, Yo. 1996. Complex Predicates in Japanese: A Syntactic and Semantic Study of the Notion 'Word'. Stanford: CSLI Publications & Tokyo: Kurosio. Quirk, Randolph et al. 1985. A Comprehensive Grammar of the English language. London: Longman. Rice, Sally. 1987. "Towards a Transitive Prototype: Evidence from Some Atypical English Passives". Proceedings of the Annual Meeting of the Berkeley Linguistics Society 13, eds. J. Ask et al., 422-434. Berkley Linguistics Society. Saksena, Anuradha. 1980. "The Affected Agent". Language 56:4.812-826. Schlesinger, LM. 1989. "Instruments as Agents". Journal of Linguistics 25.189-210. Shibatani, Masayoshi. 1985. "Passives and Related Constructions". Language 61.821-848. Sweetser, Eve. 1990. From Etymology to Pragmatics. Cambridge: Cambridge University Press. Takahashi, Hidemitsu. 1994. "English Imperatives and Speaker Commitment". Language Sciences 16:3/4.371-385.
258
HIDEMITSU TAKAHASHI
Talmy, Leonard. 1988. "Force Dynamics in Language and Cognition". Cognitive Science 12.49-100.
Lexical Causatives in Thai Kingkarn Thepkanjana Chulalongkorn University, Bangkok, Thailand
1.
Introduction * Causative constructions can be classified into two major types based on the productivity or regularity of causative forms, namely, productive causatives and lexical causatives. Productive causatives are realized by the use of caus ative verbs, such as English cause and make and Korean ha-ta, or with the use of affixes, such as Japanese sase. This means that productive causatives may be further classified into two subtypes: syntactic and morphological causatives. Syntactic causatives, which are alternatively called analytic causatives or periphrastic causatives, are generally defined as the causatives formed by means of specific verbs of causation such as make, have, cause in English, fare "make", lasciare "let, allow" in Italian, and hây "let, allow" in Thai. Morphological causatives, alternatively called synthetic causatives, on the other hand, are defined as the causatives that are realized by means of morphological devices applied to verbal forms, such as affixation in Russian and vowel alternation in Armenian. It is thus obvious that productive causatives are regularly formed by means of either syntactic or morphological devices. Lexical causatives refer to morphologically irregular, nonproductive causative forms. Lexical causatives are typically manifested in languages as a class of transitive verbs referred to as causative transitive verbs, such as cut, destroy, open, melt, kill, and boil in English. Causative transitive verbs designate the type of event that meets most of the criteria for our cognitive category of prototypical causation in the sense of Lakoff (1987:54-55). In prototypical causation, an agentive participant volitionally and directly transfers physical energy onto the second participant. The second participant then undergoes a change of state as a result of the agent's action. On this basis, it could be claimed that a lexical causative which is manifested in the form of a transitive verb, expresses two major subevents: the agent's activity and the patient's change of state. This paper aims to examine the semantic properties of lexical causatives in Thai and to classify these lexical caustaives into major types. The paper also addresses the issue of suppletive lexical causative forms, which are claimed to appear in so-called lexical gaps in a vocabulary inventory. Lexical causatives
260
KINGKARN THEPKANJANA
in the present paper are viewed from the theoretical perspective of the windowing of attention in language proposed by Talmy (1996), which will be discussed in detail in section 2. In section 3, I will review the differences between syntactic causatives and lexical causatives as postulated in a number of studies on causativity. Section 4 will deal with the classification of Thai lexical causatives and their semantic properties. In section 5, I will investigate the issue of suppletive forms and will argue that the distribution of lexical causatives is not random in Thai, and that the notion of lexical gaps in the domain of causativity is untenable. 2.
The windowing of attention in language As mentioned earlier, the lexical causatives in this paper are viewed from the cognitive framework of the windowing of attention in language set forth by Talmy (1996). According to Talmy, the windowing of attention is a cognitive process which takes place with respect to the coherent referent situation called an 'event-frame'. An event-frame refers to a generic unitary conceptual category resulting from the systematic segmentation of the occurrence of phenomena by human cognition. The event-frame evokes a set of conceptual elements and interrelationships which are felt to be central and to constitute a unitary coherent packet. One of the universal types of vent-frame treated by Talmy in his paper is a 'causal chain', which refers to a conceptualized sequence of linked subevents.1 The sequence of linked subevents results from conceptually chunking a 'causal continuum' into relatively discrete packets. The causal chain event-frame is demarcated by the initiating volitional act of an agent and by the final goal that the agent intends as a result of this act. In other words, the agent's volitional act and the goal mark the beginning and the end of the causal chain event-frame. It can be said that the causal chain must be encompassed within the agent's scope of intention. According to Talmy, the causal chain event-frame can be either prototypical or nonprototypical. In the prototypical conceptualization of the causal chain, the sense of causality occurs at the boundary marking the end of one subevent and the start of another. Furthermore, the prototypical causal chain tends to be of the direct causation type. In nonprototypical cases, the causality occurs not only at the boundary between each subevent and its linked successor, but also throughout the duration of a single subevent. With regard to causation type, nonprototypical causality encompasses both direct causation and allowance or enablement. The causal chain is initiated by the agent's act of volition which activates bodily motion. It then progresses through a sequence of intermediate causally chained subevents leading to the penultimate subevent, which constitutes the immediate
LEXICAL CAUSATIVES IN THAI
261
cause of the final result, and ends up in the final resulting subevent, which is the agent's intended goal. The windowing of attention in language is defined as a cognitive process which places a portion of an event-frame into the foreground of attention by explicitly mentioning it, while placing the remainder of the event-frame into the background of attention by omitting mentioning it. The portions that are foregrounded are said to be 'windowed' and the ones that are backgrounded are said to be 'gapped.' According to Talmy, the windowing of attention is only one part of the larger cognitive structural category in language called the 'distribution of attention', which is a system constituting the fundamental delineation of conceptual structuring in language. In the following sections of the paper, a semantic analysis of lexical causatives in Thai will be described in detail. Lexical causatives are regarded here as expressing sequential causal chain vent-frames in Talmy's sense. The sequential causal chain event-frame can have a window of attention placed over its beginning, medial or end portion. In other words, it may have 'initial, medial, or final windowing'. On the other hand, it can have a particular portion without a window upon it. That is, a portion of the vent-frame may be backgrounded by a lack of sentence constituents referring to it. On this basis, the event-frame can be said to have 'initial, medial, or final gapping' (Talmy 1996:237). The patterns of windowing of attention in the causal chain event-frame expressed by Thai lexical causatives will be investigated in the following sections. 3.
Semantic differences between syntactic causatives and lexical causatives In this section, I will review the differences between syntactic and lexical causatives in terms of event structure and the nature of the causation expressed by these two types of causatives. The formal distinction between syntactic and lexical causatives is concerned with the number of verbs expressing causality; a syntactic causative comprises two separate verbs one of which is a causative verb whereas a lexical causative, as the term indicates, is realized by means of a single transitive verb. However, both classes of causatives express the basic causative situation as viewed by Talmy (1976); this consists of two events: the causing event and the caused event or the effect. In the syntactic causative, the causing event is designated by the causative verb whereas the caused event is expressed by the other verb in the construction. As for the lexical causative, the causing and the caused events are expressed in a single verb. This means that each lexical causative corresponds to a conceptual structure composed of two subevents, namely, the agent's activity which represents the causing event, and the patient's change of state which cor-
262
KINGKARN THEPKANJANA
responds to the caused event or the effect. Notice that this fits in with Dowty's (1979) definition of an accomplishment verb.2 In the theory of force dynamics, which is a general theory of the linguistic treatment of the way humans conceptualize forces and barriers, hence causation, Talmy (1988) modifies his analysis of the structure of a lexical causative and claims that a lexical causative has a tripartite structure: a static prior state, a discrete statetransition, and a static subsequent state. This means there is an addition of a transitional state between the causing and the caused events in the lexical causative. According to Talmy, the lexical causative 'chunks' into a single unit certain portions of the continuum of interactions between entities. The lexical causative exhibits a type of force-dynamic pattern called a 'shifting' forcedynamic pattern, which expresses a change through time. I will show in section 4 that most lexical causatives in Thai have this tripartite structure except for one subtype, namely, the transitive verbs which alternate with activity verbs (see section 4.1.1). It has been extensively documented in the linguistic literature on causatives that the syntactic causative, which consists of a two-verb structure, names two separate events whereas the lexical causative expresses one single event with two subevents constituting the causative situation (Shibatani 1976). The crucial point is that the cause and the effect in the syntactic causative are felt to be two separate events whereas those in the lexical causative are perceived as jointly representing one single event. Haiman (1985) attributes this difference to the role of iconicity in language. He claims that the linguistic distance in any structure is generally motivated by the conceptual distance. One of the examples which is used to support this claim is causation. Haiman claims that the syntactic causative with two separate verbs has more linguistic distance than the lexical causative and thus signals more conceptual distance between the causer and causee. The conceptual distance and conceptual closeness expressed by the syntactic causative and the lexical causative respectively relate to the notions of direct and indirect causation which are discussed below. According to Lakoff (1987:54-55), causation is one of the most fundamental and interesting conceptual categories of human beings. Various kinds of causation are represented in the grammar of most languages and show prototype effects. Prototypical causation is understood in terms of a cluster of ten interactional properties, such as the presence of a human agent who does something, the presence of a patient which undergoes a change to a new state, the human nature of the agent, the volition of the agent, an overlap in time and space of the agent's action and the patient's change of state, a direct contact between the agent and the patient, etc. The type of causation that has the whole
LEXICAL CAUSATIVES IN THAI
263
cluster of the ten properties is categorized by Lakoff as prototypical causation. Nonprototypical varieties of causation are characterized in terms of deviations from that cluster. Prototypical causation is one of the most fundamental conceptual categories of human beings in that "it is used spontaneously, automatically, effortlessly, and often" (Lakoff 1987:55). For this reason, prototypical causation is directly represented in the grammar of languages either by means of grammatical constructions or grammatical morphemes. The constructions which are employed to code prototypical causation are transitive constructions, i.e. the constructions in which transitive verbs are present, such as in Lakoff's examples: Max broke the window and Brutus killed Caesar. On the other hand, words in different languages which are equivalent to the verb cause in English are used for expressing nonprototypical causation. It has been claimed by many linguists whose work touches on the notion of causatives that lexical causatives tend to signal direct, manipulative causation whereas syntactic causatives tend to express indirect causation (Shibatani 1973, Pinker 1989, Jackendoff 1990, Matsumoto 1996). In direct causation, the effect is brought about directly by means of direct contact by the causer's action whereas in indirect causation the effect is brought about through a mediating agent. The conceptual distance which is motivated by the two-verb structure in the syntactic causative materializes as indirect or mediated causation, in which the cause and effect are temporally separated from each other in such a way that they are perceived as two separate events. On the other hand, the conceptual closeness exhibited by the lexical causative materializes as direct causation, in which the cause and effect are tightly bound. The cause and effect in lexical causatives tend to occur at approx imately the same time and the same place. This does not always hold true for analytic causatives (Fodor 1970). According to Lakoff (1987), direct causation characterizes prototypical causation whereas indirect causation is the nonprototypical kind. However, lexical causatives do not necessarily express direct causation in the strictly objective sense, such as in the cases of the verbs build in I will build a house made of teak and clean in I cleaned this suit at the dry-cleaning store. In these two examples, the agents might not transfer their physical energy directly onto the direct object arguments in order to achieve the desired effects. There might be a number of intermediary agents intervening between the initiating agents and the patients. Talmy (1996) explains this phenomenon in terms of the theory of windowing of attention in language, which was mentioned in section 2. To recapitulate, the windowing of attention in language is defined as "the system with which languages can place a portion of a coherent referent situation into the foreground of attention by the explicit
264
KINGKARN THEPKANJANA
mention of that portion, while placing the remainder of that situation into the background of attention by omitting mention of it" (Talmy 1996:235). In the case of lexical causatives, he claims that some intermediary agents might be essential in the sequence leading to the effect, which is the final phase of the causal chain event-frame. In other words, there might be a number of interme diate causally chained subevents leading to the final resulting phase. However, the material referring to the intermediate subevents and to the intermediary agents is put into the background of attention or gapped from a sentence. Yet, causal continuity is conceived as progressing directly from the initiating agents through the intermediary agents to the final results. The cognitive process in which the medial portion of the causal chain is reduced in conscious concep tualization to the degree that the discontinuous initial and final phases may seem to be seamlessly linked to each other, is termed 'conceptual splicing' by Talmy (1996:249). What is at issue here is whether or not there is any motivation behind conceptual splicing. According to Goldberg (1995), lexical causatives can be used to imply 'conventionalized' causation that may actually involve an intermediate cause. In other words, activities which are conven tionally carried out in a particular way may be expressed as lexical causatives, or 'simple causatives' in Goldberg's terms, even when the causation is indirect. Such conventionally accomplished activities are termed by Goldberg 'con ventionalized scenarios'. Conventionalized scenarios "can be cognitively packaged as a single event even if an intervening cause exists" (Goldberg 1995:169). It can be concluded, therefore, that the cognitive process of conceptual splicing is motivated by the conventionality of carrying out certain activities. 4.
Types of lexical causatives in Thai If we adopt the definition of a lexical causative as a transitive verb which designates a situation involving an activity representing the cause and a subsequent change of state representing the effect, the lexical causative will, as pointed out in the previous section, count as an accomplishment verb as defined by Dowty (1979). However, I will show in this section that the above definition of a lexical causative encompasses a variety of transitive verbs in Thai not all of which are strictly accomplishment verbs. I will first classify transitive verbs designating causative situations into two major types based on whether they have identical intransitive forms or not. These two major types are termed alternating transitive verbs and non-alternating transitive verbs.
LEXICAL CAUSATIVES IN THAI
265
4.1 Alternating transitive verbs The term alternating transitive verbs refers to transitive causative verbs which have identical intransitive variants, such as khlûian (rot) "move (car)", phlík (kràdàat) "turn over (paper)", dàp (thian) "blow out (candle)", coot (rót) "stop (car)", pəət (pràtuu) "open (door)", and phan (bâan) "demolish (house)". In other words, these verbs can appear in both transitive and intran sitive constructions. These transitive verbs can be further classified according to the type of alternation they participate in, namely, the causative/activity alternation, the causativelinchoative/stative alternation and the causative/ stative alternation. In the first alternation, transitive causative verbs alternate with intransitive activity forms; in the second alternation, they alternate with inchoative forms and with stative forms; in the third alternation, they alternate with stative forms. These transitivity alternations exemplify a type of polysemy which is called 'regular polysemy'. According to Deane (1988), regular poly semy exhibits predictable semantic relations between alternants. Although the three subclasses of alternating transitive verbs express a causal chain eventframe, all of them have different internal structures as discussed below. 4.1.1 Transitive causative verbs alternating with activity verbs This class of transitive causative verbs has identical intransitive forms which indicate activities such as (bayphát) "turn around (blades of the fan)", dəən (khrûian càk) "operate (machine)", khlûian (rót) "move (car)", kràdəən (lûukbən) "bounce (ball)". In this subclass of transitive verbs, the final resulting subevent or the effect in the patient is realized as another activity acted out by the affected direct object argument. In other words, a dynamic event in addition to the agent's activity is embedded in the meaning of the causative verb. This class of alternating transitive verbs can thus be termed 'causative activity' verbs (Doron & Rappaport Hovav 1991). In analyzing the semantic composition of the causal chain event-frame expressed by this subclass of transitive verbs, we can see that the causal chain is initiated by the agent's act of volition and that the causal chain progresses through a number of intermediate causally chained subevents which are gapped from the sentence. This means that there are no linguistic elements in the sentence which indicate the specific action(s) the agent performs in effectuating the final intended goal (see section 4.1.4). The causal chain ends up in the final resulting subevent, which materializes as an activity acted out by the affected direct object argument. On this basis, it can be said that there is final windowing and medial gapping in the causal chain event-frame expressed by the transitive causative verbs which alternate with intransitive activity verbs. The sentences below exemplify the causative/activity alternation.
266
KINGKARN THEPKANJANA
(1) a. sùdaa kràdoon lûukbən kàp kamphææn Suda bounce ball against wall "Suda bounced the ball against the wall." b. lûukbon kràdoon ball bounce "The ball bounced." (2) a. sùdaa dəən khrûian) càk Suda operate,work machine "Suda operated the machine." b. khrûian càk dəən machine operate, work "The machine worked." (3) a. sùdaa ǔ bayphát Suda turn around blades of the fan "Suda turned around the blades of the fan." b. bayphát ǔ blades of the fan turn around "The blades of the fan turned around." (4) a. sùdaa khlûian rót Suda move car "Suda moved the car." b. rót khlûian car move "The car moved." 4.1.2 Transitive causative verbs alternating with inchoative and stative verbs The transitive verbs in this class alternate with inchoative and stative forms. The causative, inchoative, and stative alternants reflect the three com ponents making up the event and the resulting state which represent the three consecutive arcs of a causal chain in Croft's sense (Croft 1991). A causal chain according to Croft is a cognitive model of conceptualizing events in the world, which is based on the interactions between entities and the transmission of energy in the interactions. The causal chain is therefore used as a cognitive model to represent causation.3 According to Croft, a verb designates a segment of the causal chain. The causative, inchoative, and stative verbs designate three
LEXICAL CAUSATIVES IN THAI
267
prototypical event types and also provide the idealized cognitive model for the internal structure of simple events. The inchoative event type corresponds to the second and third arcs of the causal chain. That is, the inchoative verb designates an event in which a non-agentive participant 'acts out' a change up to and including its culmination point, abstracting away from whatever causes the change in question. On the other hand, the stative event type corresponds to the third, which is the last arc, of the causal chain. It designates a state that sets in when the event reaches its culmination point. The causative event type represents an idealized cognitive model for a single event structure, which is the causative-inchoative-stative structure represented in the full causal chain. This class of causative verbs include pît (pràtuu) "close (door)", pəət (pràtuu) "open (door)", phan [baan) "demolish (house)", dàp (thian) "blow out (candle)", lóm (khûutəəsûu) "make (opponent) fall down, beat (opponent)", khôon (tônmây) "fell (tree)". Notice that there is a noticeable time span starting with the onset of the change and ending when it reaches its culmination point. In other words, the change of state expressed in these causative verbs is not instantaneous. The change of state or the 'becoming' process, which occurs during such a noticeable time span, is designated by an inchoative verb.4 Such a process can be conceptualized as an autonomous event because it can be psychologically abstracted away from whatever energy that instigates it. Like the transitive verbs which alternate with intransitive activity forms, the transitive verbs in this subclass express the causal chain event-frame with final windowing and medial gapping owing to the fact that the final resulting subevent is put into the foreground of attention and that the intermediate causally chained subevents are put into the background of attention. As a result, the intermediate subevents leading up to the culmination point are vague, unspecified and have to be inferred by language users. The sentences below exemplify the causative/inchoative/stative alternation. (5) a. sùdaa dàp thian Suda put out,extinguish candle "Suda put out the candle." b. thian dàp candle burn out "The candle went out." c. thian dàp candle burn out "The candle was out."
268
KINGKARN THEPKANJANA
(6) a. sùdaa lóm khûutobsûu Suda make someone fall down opponent "Suda made her opponent fall down." b. khûutobsûu lóm opponent fall down "The opponent fell down." c. khûutobsûu lóm opponent be down "The opponent was down." (7) a. sùdaa əət pràtuu Suda open gate "Suda opened the gate." b. pràtuu pəət gate open "The gate opened." pràtuu əət gate open "The gate was open." 4.1.3 Transitive causative verbs alternating with stative verbs This class of transitive verbs has identical intransitive forms which designate states whose onset coincides with the culmination point expressed by the verbs when used transitively. Unlike the causative verbs in 4.1.2, the causative verbs in this class do not semantically encompass the 'becoming' process or the event in which a non-agentive participant 'acts out' a change leading to the culmination point. There is no noticeable time span between the agent's instigation of an action and the resulting change of state of the patient. In other words, the change of state seems to happen to the patient immediately or only a few seconds after the agent's activity has been completed. The instantaneous occurrence of the resulting change of state justifies the non existence of inchoative alternants of this class of transitive verbs. The causative verbs in this class include hàk (mày) "break (twig)", hăn (pâay) "turn around (sign)", phàp (kradàat) "fold (paper)", bit (chûiak) "twist (rope)", phlík (phâahòm) "flip over (blanket)". The changes that these transitive verbs indicate may be changes in the physical shape of the patient or changes in the position of the patient. Like the other two subclasses of alternating transitive
LEXICAL CAUSATIVES IN THAI
269
verbs, the transitive verbs in this subclass express the causal chain vent-frame with final windowing and medial gapping. The sentences below illustrate the causative/stative alternation. (8) a. sùdaa hàk íná Suda break twig "Suda broke the twig." b. kírjmáy hàk twig break "The twig was broken." [stative reading] (9) a. sùdaa pháp kràdàat Suda fold paper "Suda folded a piece of paper." b. kràdàat pháp (yùu) paper fold stative aspect "The piece of paper was folded." [stative reading] (10) a. sùdaa phlík phâahôm Suda flip over blanket "Suda flipped over the blanket." b. phâahôm phlík blanket flip over "The blanket was flipped over." [stative reading] 4.1.4 Semantic properties of alternating transitive verbs It is obvious from the analysis presented above that the three subclasses of the alternating transitive causative verbs in Thai are different from one another with respect to their internal structure despite the fact that the cause and effect are present in all of these verbs. The effect of a causative verb which alternates with an activity form is manifested in the form of motion acted out by an affected argument. On the other hand, the causative verbs of the two remaining types are similar to each other in that both of them encompass the culmination point but are different from each other in that one class of verbs also has a noticeable becoming process in which a non-causative participant acts out a change leading to the culmination point whereas the other class of transitive verbs lacks this phase. These two classes of verbs with the culmination point can thus be appropriately called accomplishment verbs. The
270
KINGKARN THEPKANJANA
changes of state in these accomplishment verbs are of various kinds, for example, the going out of existence of an entity such as in dàp "put out, extinguish", the change in the geometric shape of an entity such as in hàk "break", pháp "fold", and the horizontal and vertical change in position of an entity such as in khwâm "turn over", and phlík "turn over". There are three other important semantic properties of alternating transitive verbs which I will discuss below. First, the agent's action expressed in each of these causative verb types is vague or not specific. That is, we do not know exactly what specific action(s) the agent performs in order to effectuate a certain change in the patient. For example, in the verb pəət "open" as in (7a), we do not know what specific actions the agent performed in order to open the gate. The change of state, for instance, can be accomplished both by means of a physical, direct contact with the gate and by means of a remote control. These verbs highlight what change or activity the patient is caused to act out. It is for this reason that Denny (1983) speaks of patient-descriptive verbs. This means that the activity or change acted out by the patient in this class of transitive verbs is more salient and more vivid than the agent's action. The problem presented by these verbs can be dealt with given Talmy's proposal about the windowing of attention in language, which was mentioned in section 2. In the case of causative patient-descriptive verbs, the portions of the causal chain event-frame that are windowed correspond to the initial subevent and the final resulting subevent, which are, in other words, the agent's act of volition in effectuating a certain goal and the agent's intended goal, respectively. On the other hand, the portions of the causal chain eventframe that are gapped in patient-descriptive verbs correspond to the inter mediate causally chained subevents. Because of the medial gapping in the causal chain vent-frame, the means engaged in executing the agent's intended goal is left unspecified. The second property of the alternating transitive verbs is concerned with the temporal relation between the cause and effect. Jackendoff (1990:138-139) distinguishes between two types of temporal relation between the cause and effect, and explains the differences between them by giving two examples in English: Bill dragged the car down the road and Bill threw the ball into the field. In the first sentence, Bill's dragging, which constitutes the cause, is temporally coextensive with the motion of the car, which is the effect. In the second sentence, Bill's throwing only initiates the ball's motion. These two possibilities, which are referred to in the present study as 'causation type', are considered by Michotte (1954) as the principal variants in the perception of physical causation. Michotte calls the causation type exhibited by the verb drag
LEXICAL CAUSATIVES IN THAI
271
'entraining' and the one exhibited by the verb throw 'launching'. In the entraining causation type, after the agent's instigation of an action, there is no independent process unfolding on its own which leads up to the change of state or the accomplishment. In other words, the agent's active involvement must continue throughout until the change of state is accomplished. The entraining causation type is termed 'extent' causation by Talmy (1976). On the other hand, in the launching causation type, after the agent's instigation of an action, there is an independent process unfolding on its own leading to the change of state. The launching causation type corresponds to Talmy's point causation (Talmy 1976) and to Shibatani's ballistic causation (Shibatani 1973). Now we will consider the causation type that each class of alternating transitive verbs exhibits. It seems that there is no a priori restriction as to what causation type the first two subclasses of alternating verbs exhibit. These two subclasses of alternating verbs, namely, the transitive verbs which alternate with activity verbs and those which alternate with inchoative and stative verbs, can exhibit the entraining causation type in some circumstances and the launching causation type in some others. For example, in (4a), the agent could move the car by driving. In case the car moved downhill, the agent could push the car for a few moments, stop, and let the car move downhill by itself. The former case exemplifies the entraining causation type whereas the latter one exem plifies the launching causation type. The verb pəət (pràtuu) "open (door)", which is an example of a transitive verb alternating with an inchoative verb and a stative verb, also exhibits both causation types. In the case of the en training causation type, the agent can open the door manually with one hand on the knob right up to the end. In the case of the launching causation type, the agent can only kick the door and wait for the door to open. However, the situation is different in the case of the transitive verbs which alternate with stative forms such as hák (máy) "break (twig)", hăn (pâay) "turn around (sign)", pháp (kradàat) "fold (paper)", bit (chûiak) "twist (rope)", and phlík (phâahôm) "turn over (blanket)". These verbs exhibit only the entraining causation type. This is due to the fact there is no noticeable time span between the cause and effect. That is, the effect seems to be instantaneous. On this basis, the agent's involvement must be present until the effect is reached. Finally, notice that the resulting states of this class of causative verbs is 'externally caused' (Levin & Rappaport Hovav 1994). An external cause refers to an immediate control over bringing about an event, such as an agent, an instrument, a natural force, or a circumstance. The resulting states of this class of causative verbs must be brought about by an agent, an instrument, or some natural force. They cannot come about by themselves.
272
KINGKARN THEPKANJANA
4.2 Non-alternating transitive verbs The transitive verbs in this class do not participate in any of the alternations discussed above. In other words, they do not have identical intransitive forms. This class of verbs includes tàt "cut off, sever", bîi "crush", thúp "pound, smash", torn "boil, cook in liquid", sôm "repair", chon "collide with, bump against", lâak "pull", phlàk "push", tam "pound with a pestle in a mortar". These causative verbs are semantically classified into two subtypes based on whether the effects expressed in the verbs are inherent or potential. These two types of non-alternating transitive verbs are discussed below. 4.2.1 Non-alternating transitive verbs with inherent effects The non-alternating transitive verbs which semantically incorporate inherent effects include such verbs as chíik "tear", khâa "kill", tàt "cut off, sever", phàa "cut in half", chutan "slash, cut a small part off something", sák "wash (clothes)", láan "wash", sôm "repair", pàw "blow with the mouth", yôy "digest", tam "pound with a pestle in a mortar", bot "grind", law "sharpen (pencil)", sâay "build, create". The verbs of this type share a lexical aspectual property in that they consist of two internal phases, namely, the agent's activity and the agent's intended goal which may be defined by a certain resulting state of the direct object argument. Once the activity phase of these verbs has started, it will continue and will successively approach the goal. After the goal has been attained, the activity cannot continue beyond that point since the goal is the terminal stage or the boundary of the activity. These verbs are typically categorized as accomplishment verbs, which are charac teristically made up of an activity and a culmination point. After the culmination point has been reached, the final resulting state will set in and continue for a length of time. Therefore, the agent's intended goal is a changed state of the patient, whose onset coincides with the culmination point. For example, the agent's goal in the verb chîik "tear" is that the affected entity becomes torn apart; in khâa "kill", the affected entity is dead; in thamlaay "destroy", it goes out of existence or is disintegrated; in sâan "build", it comes into existence. Such inherent meanings of these verbs correspond more or less to the definition of telicness in aspectology.5 A telic situation refers to a situation which consists of an activity which leads up to a culmination point, beyond which the activity cannot continue. An atelic situation, on the other hand, refers to a situation which is realized as soon as it begins; it does not have to wait for a goal to be attained. According to the theory of aspectology, the telicness of situations can be designated either by verbs alone or by verbs combined with complements and/or modifiers. In the latter case, the telicness
LEXICAL CAUSATIVES IN THAI
273
is determined by the interactions of several factors such as the verb's inherent meaning, adverbials, tense, grammatical aspect, numerals, and determiners, and the nature of the verb's arguments, that is, of the subject and the object(s). Some telic situations include those described by John has made a chair, John has sung five songs, kill, cut. Some examples of atelic situations are expressed by John plays cards, John is singing songs, swim, walk. In this study, the factor that is relevant to the telicness of a situation is the inherent meanings of transitive verbs. Therefore, I will discuss only the inherent telic nature of transitive verbs without paying attention to the other factors listed above. Notice that the inherent telic nature of a transitive verb may change if the verb is combined with the other factors. Notice that the goals in some transitive verbs of this type, such as tàt "cut off, sever", chíik "tear", khâa "kill", phâa "cut in half", chǔian "slash, cut a small part off", and sâan "build" are more well-defined and more obvious than in some others, such as sôm "repair", láan "wash", and thôot "fry". The verbs with relatively more well-defined goals tend to be verbs of destruction and creation. Despite the fact that the non-alternating transitive verbs of this type are inherently telic, it is in practice possible that, once the agent's activity has begun, the activity is interrupted and the goal might not be attained due to a number of factors. For example, for the verb phrase chíik phâa "tear cloth", the cloth might be too tough for the agent to tear apart, or the agent might be interrupted by a person while he is passing energy onto the cloth, thus the goal of the cloth becoming apart might never be attained. Notice that we are allowed to use these telic transitive verbs to refer to situations in which the activities have begun and continue for a certain period of time but the goals are not achieved as in the sentences below. (11) sùdaa chïik phâa mây khàat Suda tear cloth not be torn "Suda tried to tear a piece of cloth but it did not come apart ." (12) sùdaa khâa plaa mây taay Suda kill fish not die "Suda tried to kill the fish but it did not die." (13) sùdaa chinan núia mây ?oòk Suda cut a small part off meat not out "Suda tried to cut a small part of meat off but was not successful. "
274
KINGKARN THEPKANJANA
Some aspectologists have noted that it is possible for telic verbs to be combined with the imperfective aspect and for atelic verbs to be combined with the perfective aspect (Garey 1957, Comrie 1976). The imperfective aspect "denotes a complete situation, with beginning, middle, and end" (Comrie 1976:18). The perfective aspect indicates a situation which is in progress and which lacks an inceptive stage and/or a terminal stage. In the case of the three sentences above, the telic verbs with the inherent culmination phase occur in the imperfective context in which the actualization of the culmination is not asserted. In other words, the sentences above do not indicate the successful completion of the situations although the main verbs are inherently telic. On the other hand, the actualization of the culmination of the telic verbs in the three sentences above can be asserted by modifying the main verbs with the verbs signifying the resulting states of the affected arguments as shown below. (14) sùdaa chïik phâa khàat Suda tear cloth be torn "Suda tore a piece of cloth apart." (15) sùdaa khâa plaa taay Suda kill fish be dead "Suda killed the fish successfully." (16) sùdaa chǔian núia ?oòk Suda cut a small part off meat out "Suda cut a small part of meat off successfully." Unlike the non-alternating transitive verbs, it is not possible for the alternating transitive verbs discussed in section 4.1 to be modified by verbs indicating the actualization of the culmination of the preceding activities, or by negated verbs signaling the agents' unsuccessful performance of the preceding activities. This is due to the fact that the alternating transitive verbs are patientdescriptive verbs. This means that the actualization of the culmination of the preceding activities is explicitly asserted by these verbs. In other words, the final resulting subevent of the causal chain event-frame is windowed or put into the foreground of attention. Therefore, it is logically impossible to find any verbs to modify these verbs which indicate that the culmination is attained or not attained. It can be said that the non-alternating transitive verbs in this subclass prototypically designate a telic event. However, given the appropriate type of discourse, this class of verbs can also be used non-prototypically to designate
LEXICAL CAUSATIVES IN THAI
275
the agent's actions on the patient without this including the endpoint of the change. In other words, the context dynamically deactivates the telic part of the verb's prototypical meaning in the language user's mind. 4.2.2 Non-alternating transitive verbs with potential effects This class of transitive verbs includes most 'surface-contact' or 'hitting' verbs such as tii "hit", t y "punch", tòp "slap", tè? "kick", krà?th "hit strongly", chon "collide with, bump against", th u "scrub" , khàt "polish". Some 'surface-contact' transitive verbs in Thai may designate both volitional and nonvolitional actions, such as chon "collide with, bump against", krà?th "hit strongly", and tè? "kick", whereas some others indicate volitional actions only, such as tòp "slap", and / y "punch". An action designated by this type of verb is characterized by the fact that it may or may not produce any effect in the patient; it only has the potential to have an effect in the patient. If it does produce an effect, the effect can be actualized in many forms depending on the context of situation. For example, we may hit a person without bringing about any effect in him at all. Therefore, it can be concluded that the effect is not inherent, but merely potential, in the semantics of the verb. To put it in Talmy's terminology, the final resulting subevent of the causal chain event-frame expressed by this type of verb is not necessarily present. Because of the potential nature of the effect in this class of verbs, these verbs are normally regarded as activity verbs in Dowty's sense, not accomplishment verbs typifying lexical causatives. Although this class of verbs are semantically different from the prototypical accomplishment verbs as far as the effected situation is concerned, we cannot deny that this class of verbs can also imply a change of state of some kind in the affected entities. However, the changes of state in these activity verbs, if they exist, are merely potential, vague, not well-defined and depend largely on the context of situation (Fillmore 1970). On this basis, these verbs are inherently atelic since they refer to situations which are realized as soon as they begin. However, it is still possible for the agent to set a goal in performing these actions even though the goal might never be attained. It can be concluded at this point that this class of verbs consists of nonprototypical lexical causatives since the verbs in this class lack a clear indica tion of an effect or a resulting event in their semantic description. The effect merely has potential to take place in some cases and might be even uninten tional on the part of the agent.
276
KINGKARN THEPKANJANA
4.2.3 Semantic properties of non-alternating transitive verbs Non-alternating transitive verbs display two semantic properties with respect to causation type and the roles of the agent and patient in the semantic description of the verbs. First, the non-alternating transitive verbs express the entraining causation type in which the agent's performance of an action must continue throughout until a change of state, if produced, is accomplished. There is no independent process leading up to the change of state. Secondly, the two subtypes of transitive verbs differ from each other with respect to the roles of the agent and patient in the semantic description of the verbs. Transitive verbs with potential effects, which are mostly surface-contact verbs, are characterized as 'agent-descriptive' because they primarily describe what the agent is doing to the patient rather than the nature of the resulting change of state which occurs to the patient. These verbs are characterized by the fact that the agents' actions embedded in these verbs have potential to effectuate certain changes of state in the patients. However, the resulting changes of state in the patients do not necessarily occur as discussed above. In the event that the changes of state do occur to the patients, we cannot visualize the exact nature of the changes of state in the patients which result from the agents' actions. The exact nature of the changes of state in the patients depends largely on pragmatic factors and on real-world knowledge of the interaction between the predicates and the patients. On the other hand, verbs with inherent effects, such as chïik "tear", khâa "kill", tat "cut", sák "wash (clothes)", and sâan "build, create", are partly agent-descriptive and partly patient-descriptive. In other words, in these verbs we know to a certain extent what specific action(s) the agents perform and what happens to the patients as a result of the agents' actions. For example, in the verbs chïik "tear", tàt "cut off, sever", phàa "cut in half", chǔian "slash, cut a small part off something", pàw "blow with the mouth", tam "pound with a pestle in a mortar", and sák "wash (clothes)", we can more or less visualize the agents' specific actions as well as the resulting states which may occur to the patients. It is also claimed by Vendler (1984) that in these cases there is a tight relationship between the initiator of the causative event and the rest of it. Viewing the semantic properties of the two subtypes of non-alternating transitive verbs from Talmy's framework of the windowing of attention in language, we can see different patterns of windowing of attention in the causal chain event-frames expressed by them. With regard to non-alternating transitive verbs with inherent effects, the attention is distributed more or less evenly over the causal chain event-frame. No portion in the event-frame is put into the foreground nor background of attention. In other words, no portion is win-
LEXICAL CAUSATIVES IN THAI
277
dowed nor gapped. This is due to the fact that this type of verb is partly agentdescriptive and partly patient-descriptive as discussed above. With regard to non-alternating transitive verbs with potential effects, more attention is placed over the medial portion than over the final portion of the causal chain eventframe. In other words, the medial part of the vent-frame is more windowed than the final portion. This is due to the property of agent-description displayed by this type of verb and to the potential nature of the effect brought about by the agent's action. On this basis, it can be said that the final resulting subevent of the causal chain event-frame expressed by this type of verb is optional, and is therefore considered a non-prototypical case. Notice that this pattern of windowing of attention differs sharply from that in the causal chain eventframe expressed by alternating transitive verbs in which the strongest attention is placed over the final resulting subevent. It is obvious that the role of the agent is relatively more dominant in nonalternating transitive verbs than in alternating ones. The dominant role of the agent in the former can be detected from their two semantic properties dis cussed above, namely, the entraining causation type and the agent-description which figures in the semantic description of the two subtypes of non-alternating transitive verbs. Correspondingly, the role of the patient is relatively less obvious in non-alternating verbs than in alternating ones. For non-alternating verbs with inherent effects, the resulting states of the preceding actions are inherent and are reached in the normal course of events; that is, if nothing interrupts the agent's performance. For alternating verbs, the actualization of the resulting states is asserted by the verbs themselves. This means that the resulting states in alternating verbs are reached in all possible courses of events. The role of the patient in surface-contact verbs is even less obvious since these verbs are characterized as agent-descriptive verbs. 5.
The so-called suppletive lexical causative forms A number of linguists working on causatives have noted that lexicalization of causatives is arbitrary and language-specific (Shibatani 1973, Givón 1976, Heringer 1976). Certain non-causative verbs in some languages may not have 'corresponding lexical causatives', which refer to transitive causative verbs that are identical or phonologically related to non-causative verbs. Some examples of such non-causative verbs in English include die, learn, wake up (intransitive verb), see, know, etc. It has been claimed that the nonexistence of the corresponding lexical causatives for some non-causative verbs gives rise to accidental gaps in the lexicon. In this case, the corre-sponding productive causative may be used to express the types of causation normally expressed only by the lexicalized causative (Heringer 1976:207). It has been claimed also
278
KINGKARN THEPKANJANA
that suppletive lexical causatives always appear in such lexical gaps as the examples in English below (Matisoff 1976:418). Non-causatives die eat see
Suppletive causatives kill feed show
However, from the analysis presented in this paper we may conclude that the distribution of suppletive lexical causative forms is not random in Thai. In other words, there are no such things as accidental lexical gaps in which suppletive lexical causative forms appear. It has been established that the alternating causative verbs are patient-descriptive verbs in which the effects are well-defined and their actualization is asserted by the verbs themselves. The nature of the effects embedded in these verbs is definite and does not depend on the context of situation or real world knowledge. On the other hand, supple tive causative verbs are found to correspond to non-alternating transitive verbs. These verbs are not purely patient-descriptive verbs; that is, they are either purely agent-descriptive as in the case of surface-contact verbs, or partly agentdescriptive and partly patient-descriptive at the same time as in the case of non-alternating transitive verbs with inherent effects. Unlike alternating tran sitive verbs, non-alternating verbs or, in other words, so-called suppletive causatives, are noncommitted as to the nature of the results of the actions. Furthermore, the resulting states semantically embedded in alternating tran sitive verbs are always externally caused as discussed earlier, whereas those which are expected or which are the potential outcomes of actions designated by some non-alternating transitive verbs, such as the states designated by taay "be dead", khàat "be torn", rúu "know", khâwcay "understand", and hen "see", are not necessarily externally caused. These states may be externally as well as internally caused. In the case of internal causation, these states may be able to come about and take hold of a human being owing to some inherent properties of the human being himself. Internal causation also holds true for the three English non-causatives cited by Matisoff (1976:418), namely, die, eat, fall. The events designated by these three verbs can be internally caused. We can thus see the following regularity governing the linguistic realization of lexical causatives in Thai: if the resulting states are always externally caused, i.e. parts of the causal chain, and are asserted by the verbs themselves, then the linguistic forms of the resulting states are identical with the causative verbs. On the other hand, if the resulting states are not asserted by the causative verbs, but merely expected or potential, and if they can be internally
LEXICAL CAUSATIVES IN THAI
279
caused, then the linguistic forms of the resulting states are not identical with those of the causative verbs. It is thus obvious that the identity in linguistic form between a stative verb and a causative verb takes place when the state designated by the former necessarily results from the preceding action performed by an agent. It can thus be said that the necessary interdependence and tight relationship between a causative action and its resulting state is iconically reflected by the identity in form between a causative verb and a stative verb designated by the two events respectively. On the other hand, if there is a relatively loose relationship between an action and a resulting event such as in the case of the non-alternating transitive verbs, the verb forms denoting these two events are not identical to each other. In this case, it is claimed that the causative verbs are linguistically realized as so-called supple tive forms. On this basis, it can be concluded that so-called suppletive lexical causatives in Thai do not occur in accidental gaps in the lexicon as previously claimed. Rather, their distribution in the lexicon is predictable. Because of the predictability of the distribution of suppletive lexical causatives, it is no longer appropriate to label these lexical causatives 'suppletive'. 6.
Conclusion In this paper, I have investigated the semantic properties of lexical causatives in Thai and classified them into two types, namely, alternating and nonalternating causative verbs. I have argued that these two types of verbs are semantically distinguished. Each subtype of lexical causative was further clas sified into subtypes. The notions of causation type, agent-description, and patient-description figure prominently in the semantics of lexical causatives in Thai. Finally, it was shown that the distribution of so-called suppletive lexical causatives in Thai is not random and therefore the notion of 'accidental lexical gaps' for suppletive lexical causatives appears to be untenable. Notes * I am grateful to Frederike van der Leek and Masayoshi Shibatani for their invaluable comments and suggestions on an earlier draft of the manuscript. I also thank Jordan Zlatev for helpful comments and textual improvements. 1 The other generic types of event-frame treated in Talmy (1996) are a path, a cycle, a participant interaction, and an interrelationship. 2 An activity is one of the four event types postulated by Vendler (1967) and revised by Dowty (1979), namely, state, activity, accomplishment, and achievement. An activity refers to a happening occupying a period of time which is inherently indefinite. An activity is homogeneous, that is, any part of it is of the same nature as the whole (Vendler 1967:101). 3 Talmy's notion of causal chain as discussed in Talmy (1996) is different from that postulated by Croft (1991) in that the former corresponds to a type of sequential event-frame
280
KINGKARN THEPKANJANA
which can be conceptually 'chunked' into a number of subevents all of which are within the agent's scope of intention. Some subevents in the causal chain event-frame may not be referred to by any linguistic elements in a sentence at all. On the other hand, Croft's causal chain is not as finely chunked by human cognition as Talmy's. Each segment of Croft's causal chain tends to be linguistically designated as a verb in the language. 4 An activity which is non-agentive and which lacks a patient is sometimes called a 'process'. 5 The term 'telic' was introduced by Garey (1957). It comes from the ancient Greek word telos meaning 'end.'
References Comrie, Bernard. 1976. Aspect. Cambridge: Cambridge University Press. Croft, William. 1991. Syntactic Categories and Grammatical Relations. Chicago: University of Chicago Press. Deane, Paul D. 1988. "Polysemy and Cognition". Lingua 75.325-361. Denny, J. Peter. 1983. "Semantics of Abstract Finals in Algonquian Transitive Inanimate Verbs". Canadian Journal of Linguistics 28.133-148. Doron, Edit & Malka Rappaport Hovav. 1991. "Affectedness and Externalization". Proceedings of the North Eastern Linguistic Society Annual Meeting 21.81-94. Dowty, David. 1979. Word Meaning and Montague Grammar. The Semantics of Verbs and Times in Generative Semantics and in Montague's PTQ. Dordrecht: D. Reidel Publishing Company. Fillmore, Charles J. 1970. "The Grammar of Hitting and Breaking". Jacobs & Rosenbaum 1970. 120-133. Fodor, Jerry. 1970. "Three Reasons for not Deriving 'kill' from 'cause to die'". Linguistic Inquiry 1.429-438. Garey, Howard B. 1957. "Verbal Aspect in French". Language 33.91-110. Givón, Talmy. 1976. "Some Constraints on Bantu Causativization". Shibatani 1976. 325-351. Goldberg, Adele . 1995. A Construction Grammar Approach to Argument Structure. Chicago: University of Chicago Press. Heringer, James T. 1976. "Idioms and Lexicalization in English". Shibatani 1976. 205-216. Haiman, John. 1985. Natural Syntax. Iconicity and Erosion. Cambridge: Cambridge University Press. Jackendoff, Ray. 1990. Semantic Structures. Cambridge, Mass.: MIT Press. Jacobs, Roderick A. & Peter S. Rosenbaum. 1970. Readings in English Transformational Grammar. Waltham: Ginn and Company. Lakoff, George. 1987. Women, Fire, and Dangerous Things. What Categories Reveal about the Mind. Chicago: Chicago University Press.
LEXICAL CAUSATIVES IN THAI
281
Levin, Beth & Malka Rappaport Hovav. 1994. "A Preliminary Analysis of Causative Verbs in English". Lingua 92.35-77. Michotte, Albert. 1954. La Perception de la Causalité. Louvain: Universitaires de Louvain. Matisoff, James A. 1976. "Lahu Causative Constructions: Case Hierarchies and the Morphology/Syntax Cycle in a Tibeto-Burman Perspective". Shibatani 1976. 413-442. Matsumoto, Yo. 1996. Complex Predicates in Japanese. A Syntactic and Semantic Study of the Notion 'Word'. Tokyo: CSLI Publications & Kurosio Publishers. Pinker, Steven. 1989. Learnability and Cognition. The Acquisition of Argument Structure. Cambridge: MIT Press. Shibatani, Masayoshi. 1973. A Linguistic Study of Causative Constructions. Doctoral dissertation. University of California, Berkeley. . 1976. "The Grammar of Causative Constructions: A Conspectus". Shibatani 1976. 1-40. , ed. 1976. Syntax and Semantics. The Grammar of Causative Constructions. New York: Academic Press. & Sandra A. Thompson, eds. 1996. Constructions. Their Form and Meaning. Oxford: Clarendon Press. Talmy, Leonard 1976. "Semantic Causative Types". Shibatani 1976. 43-116. . 1988. "Force Dynamics in Language and Cognition". Cognitive Science 12.49-100. . 1996. "The Windowing of Attention in Language". Shibatani & Thompson 1996. 235-287. Vendler, Zeno. 1967. Linguistics in Philosophy. Ithaca, N.Y.: Cornell University Press. . 1984. "Agency and Causation". Midwest Studies in Philosophy 9.371384.
Cognitive Models in Transitive Construal in the Japanese Adversative Passive Eijiro Tsuboi University of Tokyo, Komaba, Japan 1.
The Japanese adversative passive: basic f acts and previous analyses * The Japanese adversative passive has been one of the most well-discussed aspects of Japanese grammar, and one of the reasons for its popularity lies in the fact that it defies the usual characterization of passive and does not show the valency decrease which is usually regarded as one of the hallmarks of passive. Consider (1), which is based on an intransitive verb, as can be seen from (2) and (3): (1) Taro wa musuko ni shooshins-are-ta Taro TOP son by get=a=promotion-PASS-PAST (Lit.) "Taro was got promotion by son" "Taroi's son got a promotion, much to hisi chagrin" (2) Musuko ga shooshins-ita son NOM get=a=promotion-PAST "(Taro's) son got a promotion" (3) *Musuko ga Taro wo shooshins-ita son NOM Taro ACC get=a=promotion-PAST Here, the verb is shooshinsuru, an intransitive verb roughly meaning "get a promotion", and the passive morpheme is -are-.1 What is to be noted here is that the verb is an intransitive one, and unlike impersonal passives derived from intransitive verbs in some other languages, here the valency increases by one, contrary to the usual pattern of passive. As the ungrammatical transitive version in (3) shows, there is no active counterpart. This valency increase has been regarded as one of the most remarkable properties of the Japanese adversative passive, and in fact, some researchers like Nishio (1982) and Palmer (1994) have analyzed it as a causative construction due to this valency increase, despite its morphological identity and semantic affinity to the ordinary passive.
284
EIJIRO TSUBOI
Another distinct property of this construction is the obligatory adversity connotation, as is indicated by the English translation given in (1). (1) can only be taken adversatively even if Taro is known to be a fond father who tends to take too much pride in his son's success. In the sixties and seventies, when it attracted a lot of attention of generative linguists concerning its deep structure, many people assumed that there was a direct correlation between the adversity sense and the valency increase. For example, Kuno (1973) takes such a view, as expressed in the excerpt below: "What is peculiar syntactically ... is that these passive sentences have one extra noun phrase compared to the corresponding active sentences. What is peculiar semantically about these passives is that they mean that the subject of the main sentence is adversely affected." [Kuno 1973:24] "... this connotation of adversity passive arises only when there is an extra noun phrase ... that cannot be accounted for by simplex deep structure of the active version ..." [ibid.:302] However, as people gradually have come to notice the existence of examples where there is no such correlation, the general consensus now seems to be that there is no direct correlation between the existence of an extra noun phrase and the adversity sense. For example, (4) is clearly adversative, but it is derived from an ordinary transitive sentence, example (5), which has no adversative connotation, and there is no valency increase here: (4) Yoochien wa watashi no musuko ni yame-rare-ta kindergarten TOP I GEN son by quit-PASS-PAST (Lit.) "Kindergarten was quit by my son" "The kindergarten was adversely affected by my son quitting it" (5) Watashi no musuko wa yoochien wo yame-ta I GEN son TOP kindergarten ACC quit-PAST "My son quit the kindergarten" Thus, what seems to be accepted by many people now is the generalization expressed in (6): (6) "...the adversity reading obtains in a passive form when the event described does not directly affect the subject of the passive clause. The adversity reading obtains typically with an intransitive verb,
TRANSITIVE CONSTRUAL IN THE JAPANESE ADVERSATIVE PASSIVE 285
because, by definition, the subject of the passive of an intransitive verb is not directly affected. The notion of the directness and indirectness discussed here has been made more precise by Oehrle and Nishio (1981) in terms of the criterion of whether a described event can take place independently of (i.e. without involving) the subject of the passive clause." [Shibatani 1990:319-320]2 The adversative sense of (1) and (4) follows from this generalization, since in (1) Taro's son can get a promotion without involving Taro, which is why (2) is well formed in the first place, and the event is complete without involving Taro. Likewise, although the subject in (4), yoochien, "kinder garten", is involved in my son quitting it, it does not undergo any change in its state, and is not affected in the sense relevant here. Therefore, it meets the no-direct-affectedness condition for the adversative passive, and is interpreted adversatively. This view is further developed in Shibatani (1994), which takes a more typologically-oriented perspective and tries to provide a unified account of various phenomena in many languages that involve extra-thematic NPs. He claims that since the extra-thematic NP in the Japanese adversative passive is extra-thematic, it must be licensed constructionally, and is integrated into the meaning of the sentence by receiving the role of the adversely affected entity: (7) "When the body-part reading is available [to make sense of the extra-NP], no additional semantic support or augmentation is needed [to integrate it into the semantics of the whole sentence]. But when no such reading is possible, an additional meaning relationship must be sought in integrating the extra thematic argument in the described scene. The adversity reading is tied to the integration of its referent as an indirect participant being (negatively) affected by the described main event. ... [S]emantic augmentation is required that doctors up the relevance of the referent of the extra-thematic argument to the described scene." [Shibatani 1994:468, addition and emphasis by E.T.] But since being adversely affected is not the only way we are affected by an event, explaining the adversative reading along the lines of what Shibatani writes in (7) makes us wonder why it must be an adversative reading rather than a benefactive one in order for the extra-NP to be integrated into the whole sentence. Shibatani's solution is to attribute it to the blocking effect of an independently existing benefactive construction:
286
EIJIRO TSUBOI
(8) "Whether the extra-thematic argument construction is exclusively associated with the adversity reading or it permits the benefactive reading depends to a great extent on whether or not a given language has a distinct benefactive construction. Japanese, for example, has a distinct benefactive construction, and therefore its indirect passive is typically associated with the adversity reading." [Shibatani 1994:481] In this excerpt, the adversative passive is called 'indirect passive', but it is basically the same thing here. Also, Japanese has more than one construction that can be called 'benefactive', but what Shibatani has in mind here is probably the -temorau benefactive construction, though it makes no difference if he means otherwise. 2.
Issues to be addressed Although we must appreciate the comprehensiveness of the overall picture provided by Shibatani's (1994) typologically-oriented account, we should not fail to notice that it leaves the following three points unexplained. First, it fails to account for the constraint on the adversative passive. Shibatani (1994), and for that matter, Kuno (1983) as well, does not say any thing about this, but it is not the case that the adversative passive can be formed from any verb, which has been pointed out by researchers such as Mikami (1953). Although the constraint has not been adequately characterized so far, roughly speaking, it is only unergatives that enter into the adversative passive construction, and examples like (9) to (11) are unacceptable: (9) *Taro wa kyonen hidoi jishin ni oki-rare-ta Taro TOP last=year devastating earthquake by occur-PASS-PAST (Lit.) "Taro was occurred by a devastating earthquake last year" "*A devastating earthquake occurred last year, adversely affecting Taro" (10) *Tokyo no hitobito wa kinou kaminari ni Tokyo GEN people TOP yesterday lightning by ochi-rare-ta fall-PASS-PAST (Lit.) "People in Tokyo were fallen by lightning yesterday" "*Lightning occurred in Tokyo yesterday, adversely affecting the people"
TRANSITIVE CONSTRUAL IN THE JAPANESE ADVERSATIVE PASSIVE 287
(11) *Taro wa pan ni urikire-rare-ta Taro TOP bread by be=sold=out-PASS-PAST (Lit.) "Taro was sold out by bread" "*Bread was sold out, adversely affecting Taro" Shibatani's account would incorrectly predict that all these sentences were acceptable if they were interpreted adversatively, but in fact they sound odd even as adversative passive sentences.3 Second, if the extra NP is really extra-thematic and is licensed constructionally in order to be semantically integrated, then it is not clear why the same constructional rescue is not available in the active, making such sentences like example (3) grammatical by giving it an adversative reading. But in fact (3) is out, and there is no such thing as 'adversative active' in Japanese. Although construction-specific restrictions are generally expected in a constructional analysis, attributing the existence of the extra NP directly to the construction without considering the reason for its existence does not reveal much about the phenomenon, and we will argue below that there is some motivation worth considering for its existence in the adversative passive. Third, if the blocking effect of the -temorau benefactive construction was responsible for the impossibility of a benef active reading, then we would expect that a benefactive reading was possible with verbs that did not enter into the benefactive construction. Since -temorau attaches to verbs denoting controllable actions, verbs expressing uncontrollable events like shooshinsuru, "get a promotion", in (1) do not allow the benefactive -temorau. (12) *Taro wa musuko ni shooshinsi-temorat-ta Taro TOP son by get=a=promotion-BEN-PAST "*Taroi had his sonj get a promotion for hisi benefit (by asking him: to)" Now, what is to be recalled here is that (1) is clearly adversative, as we first saw, despite the fact that the benefactive -temorau is not expected to block the benefactive reading, as can be seen from (12). Attempts to derive the adversative reading from the blocking effect of the benefactive construction are in fact not new, and can be traced back at least to Alfonso (1971), but Wierzbicka (1979) has already pointed out that even when no blocking effect is expected, the adversative connotation is there. This suggests that the obligatoriness of adversativeness should be explained differently.
288
EIJIRO TSUBOI
In the following section, we will take a different perspective to deal with the above three points left unexplained by previous analyses, seeking to make a coherent picture of the data. 3. 3.1
Cognitive models underlying the Japanese adversative passive Thematicity of the subject of the adversative passive First, the reason for the privileged status of the extra-thematic NP in (1) as opposed to the unacceptable extra-thematic NP in (3) can be explained by the null hypothesis that it is intrinsic to the adversative passive itself, not externally endowed to sanction the subject NP. In other words, what is usually regarded as extra-thematic in the adversative passive, i.e. the subject NP, is not extra-thematic, as will be argued below. The account presupposing the extra-thematicity of the NP in question depends on a rather dubious assumption that the active and the passive are semantically the same in relevant respects, when in fact we can think that the semantics of the adversative passive is such that it naturally integrates into the meaning of the sentence what would count as extra-thematic in the active. Although far from being definitive, the characterization of the Japanese ordinary passive commonly made along the lines of (13) will be relatively uncontroversial and sufficient for our present purpose:4 (13) The Japanese ordinary passive: The subject is affected by the act done to it by the ni-marked NP Because of this function, passive examples like (4) whose subjects are not affected in the events described by the verbs cannot be instances of the ordinary passive, and can only be acceptable as adversative passive if the verbs are of the appropriate kind for the adversative passive as characterized below. Before going into the characterization of the adversative passive, it must be first noted that the adversative passive is a distinct extension from the ordinary passive. This point may seem to be a rather obvious and trivial one, but many people including Shibatani and Kuno do not seem to recognize it, and try to derive the adversity sense as something externally imposed to compensate for the lack of affectedness.5 However, as we saw above, sentences like (9) to (11) are not acceptable even as adversative passive, which suggests that the adversative passive is not something to take care of what is left unattended by the ordinary passive. Thus, we must recognize the independent status of the adversative passive with its inherent adversative semantics.
TRANSITIVE CONSTRUAL IN THE JAPANESE ADVERSATIVE PASSIVE 289
So our task is to come up with a plausible characterization of the adversative passive that is distinct from the ordinary passive while retaining the relatedness to the latter. We would suggest the following: (14) The Japanese adversative passive: The subject is annoyed by the act done to him or her by the ni-marked NP This is an extension from (13) since it differs from the latter in its specification of the nature of the affectedness as annoyance at the act done by the ni-marked NP, but it is also related to the latter in that both of them share the schema that the subject is affected somehow by something done to him or her by the ni-marked NP. However, the annoyance felt by the passive subject need not be meant as such by the ni-marked NP. For example, (1) can be used when Taro, being a jealous father, was not happy about the promotion of his son who had believed that his promotion would please his father. This is the reason for the prevalent analysis in which the subject is assumed to be affected not by the ni-marked NP but by the whole event denoted by the verb, as can be seen in the italicized part of Shibatani's statement in (7). Therefore, (14) alone incorrectly rules out those cases in which the ni-marked NP does not intend to cause annoyance to the passive subject. Filling the gap here is the following everyday logic, a cognitive model concerning event causation, which identifies an inadvertent causer of adverse effects with an intentional doer of an adverse act: (15) The person who could have prevented the occurrence of an undesired event and its consequences, and therefore can be criticized and blamed for his or her failure to do so, is the one who caused it to happen, and thus regarded as having done the adverse act to the affectee According to this model, the person who could have avoided causing annoyance by e.g. refraining from doing what he or she did and thus can be held responsible for the annoyance felt by the passive subject is construed as doing an act of annoyance to the passive subject through the act denoted by the verb. Put differently, what could have been construed as an intransitive event of someone feeling annoyance at something is construed as a transitive event of someone annoyed by someone else through the act of the latter, thus leading to the valency increase characteristic of the adversative passive. This construal is compatible with the meaning conventionalized for the adversative passive in (14), and is expressed by the adversative passive construction.
290
EIJIRO TSUBOI
Given (15), so long as the event described by the verb involves an actor who can be regarded as responsible for the occurrence of the event annoying to the passive subject, the adversative passive is possible. Actually, part of the model in (15) is presupposed by the notion of annoyance mentioned in (14), since the sense of annoyance expressed in the Japanese adversative passive is not merely the sense of being subjected to an annoying act but also incor porates an accusing feeling toward the person who could have avoided causing annoyance. This suggests that the causer, the ni-marked NP, is indispensable for the adversative passive to mean what it means, since accusing feelings cannot stand alone without the object of accusation. This may well be the reason for the well-known fact about the Japanese adversative passive that the ni-marked NP cannot be omitted, in contrast to the general omissibility of the agent NP in the passive. In the light of (14) and (15), the unacceptability of sentences like (9) to (11) makes sense. The verbs in these sentences are unaccusatives and the events described just happened, as it were, and the ni-marked NPs did not have the control over the course of the development of the events, and are not held responsible for the events. Thus, the first problem with Shibatani's analysis is solved. Moreover, since in this analysis the subject in the adversative passive is not affected by the event, but by the actor of the event, that is, the ni-marked NP, in the sense that the subject is annoyed by the ///-marked NP who is construed as doing an annoying act to the subject, the agent marker -ni on the actor of the embedded event follows naturally, while it does not in other accounts in which the subject is supposed to be affected by the event, not by the ni-marked NP. Many people assume that the subject of an adversative passive sentence is affected by the event, which is why the subject is usually regarded as extra-thematic in the first place, but examples like (9) to (11) indicate that the name 'adversative passive' is in fact a misnomer, since adversative effects are not the sufficient condition for the use of the Japanese adversative passive.6 It is also to be noted that since annoyance need not be meant as such and the construal of the ni-marked NP as someone doing the annoying act to the subject crucially depends on the perception from the subject's viewpoint, there is some motivation to the fact that the transitive construal of an intransitive event and the consequential valency increase are found in the passive rather than in the active, since the Japanese passive (except for the niyotte-passive mentioned in Note 4) is generally said to describe the event from the passive subject's viewpoint. Then, we can avoid stipulating the association of the presence of the extra-NP with the passive, thus solving the second problem with Shibatani's analysis.
TRANSITIVE CONSTRUAL IN THE JAPANESE ADVERSATIVE PASSIVE 291
3.2 Asymmetry between adversative and beneficial events The analysis given above thus solves the first and the second problems with Shibatani's analysis, but there is still one point about (14) and (15) that needs further consideration. (14) and (15) mention annoyance and undesired events, respectively, but as we pointed out as the third problem, we need to find out the reason for the special status of adversative effects as opposed to beneficial effects in the extended transitive construal in the Japanese adversative passive. Aside from the account based on the alleged blocking effect of the benefactive construction, the untenability of which we saw above, it has been suggested by Wierzbicka (1988:ch.2) and others that the adversity sense is a grammaticalized reflection of the pessimistic world view that it is rare for a beneficial event to take place without our effort while it is all too common for an unfavorable event to occur unexpectedly, i.e., that what happens around us that is not caused by our intentional action is adversative in nature. This is an interesting point, but, as we saw above, what is crucial with the Japanese adversative passive is not the adverse effect itself (cf. the unacceptability of (9) to (11)) but the presence of an actor who can be held responsible for the occurrence of the event and can be regarded as having done the annoying act denoted by the verb to the subject. This suggests the need for a different account, and it is interesting to note in this connection that the enhanced transitivity in event construal based on the model described in (15) is in fact not limited to the adversative passive, and can be found in non-prototypical, non-intentional transitive sentences like those in (16) as well: (16) a.
Taro wa kaze de boshi wo tobashita. Taro TOP wind with hat ACC flewvt (Lit.) "Taro flew hat with wind" "Taro got his hat blown off by the wind" [Nishimura 1993:509] b. He dropped his wallet (cf. His wallet dropped) c. She broke her leg in the accident (cf. Her leg got broken in the accident)
Non-prototypical transitive sentences have been studied by various researchers. In particular, Nishimura (1993) provides an insightful account of various phenomena involving agency, citing data from English and Japanese, and shows how the notion of responsibility is crucial for the subject status of non-agentive NPs in sentences like those in (16). In these sentences, the subject is not agent in the usual sense of the term. Take (16a) for example, the subject
292
EIJIRO TSUBOI
didn't do anything: the hat just flew away on the wind. Thus, the events described by these sentences could have been expressed as intransitive events, but Nishimura (1993) shows that the subject in this type of transitive sentence is construed as agent based on the construal that whatever happens due to the negligence of his responsibility for proper control of his belongings or body parts is caused by the negligent person himself. In our terms, the presence of a responsible person who should and could have prevented the occurrence of the adversative event evokes the model in (15), and causes the responsible person to be construed as the causer of the event, with the result that the event is construed not as an intransitive event but as a transitive one, with the valency increased by one. It is important to note here that this is parallel to the adversative passive, where the actor held responsible for the annoyance is construed as doing an act of annoyance to the passive subject, resulting in the valency increase as we saw above. What deserves more attention here, however, is that these transitive sentences made possible by the presence of a responsible actor usually express adversative events, as those in (16), and (17) and (18) below do. That is, just as the adversative passive requires an actor responsible for the adversative event, non-prototypical transitive sentences require actors responsible for the adversative events, and the asymmetry between favorable and unfavorable effects is not limited to the adversative passive, but in fact more general in extended transitive construal. In relation to the importance of the notion of responsibility in extended transitive construal, examples like (17) and (18) below are interesting in that they show that the notion of responsibility plays a greater role in Japanese than in English, where these events would not be expressed with the sufferers as the subjects of the transitive verbs. In these sentences, unlike in (16b) or (16c), not only do the sufferers not do anything but the adverse effects are caused by some other people. Nevertheless, the sufferers, despite the presence of the actual actors, are expressed as the causers of the events. In (17), the sufferer didn't do anything but the gangster hit the person and broke his teeth. Likewise, in (18), the loss of the house is actually caused by the bomber or the person to order the bomber to bombard the area. Still, these subjects are held responsible for failing to keep his or her property or body part intact, and for this reason they are construed as the agent doing the act denoted by the verbs.7 (17) Taro wa (yakuza ni nagur-are-te) ha wo ot-ta Taro TOP (gangster by hit-PASS-GER) tooth ACC breakvt-PAST (Lit.) "Taro (was hit by gangster and) brokevt tooth" "Taro (got hit by a gangster and) got his tooth/teeth broken"
TRANSITIVE CONSTRUAL IN THE JAPANESE ADVERSATIVE PASSIVE 293
(18) Watashi wa kuushuu de ie wo yai-ta I TOP air=raid with house ACC burnvt-PAST (Lit.) "I burntvt house in air raid" "(Due to my fault,) I had my house burnt down in an air raid" It is not the case, however, that Japanese extends the notion of responsibility to a greater degree than English with regard to transitive construal in general. Rather, the notion of responsibility in Japanese is slightly different from that in English. In English, responsibility is attributed even to inanimates, so it is quite common to say something like Too much fat is responsible f or many diseases, and probably because of this, inanimate subjects construed as actors based on their causal nature are not rare, whereas in Japanese they are generally impossible, and even sentences like The key opened the door cannot translate into Japanese with the key as the subject. As implied in (15), in Japanese the notion of responsibility is closely associated with animate or sentient beings capable of volitional control of what he or she does, and probably this is why example (19) is out while its German counterpart (20) is acceptable. Wierzbicka observes that the German datives of misfortune do not require a human causer unlike the Japanese adversative passive. In our terms, a house is not an animate entity which we can criticize or blame for the adverse effects we suffer from the burning down of the house, and therefore (19) is out. (19) *Kare wa ie ni moe-rare-ta He TOP house by burnvf-PASS-PAST "*His house burnt down on him" (20) Ihm ist das Haus abgebrannt to-him is the house burnt-down "His house burnt down on him" [Wierzbicka 1988 (=1979):279] Then, if we realize the prevalence of the asymmetry between adversative and beneficial events as well as the readiness Japanese exhibits with which an actor held responsible for an adversative event can be construed as the doer of the adverse act, it is not so surprising that the adversative reading is linguis tically entrenched to such a degree in Japanese that it is incorporated into the adversative passive.8 Although detailed research into the psychology will be necessary to confirm speculations, the reason for the asymmetry seems to be that adver-
294
EIJIRO TSUBOI
sative events contrast with beneficial ones in that they are something to be prevented whose occurrence highlights the responsible actor who has brought about what should not have been brought about. That is, when an event that should not occur turns out to have occurred, we conceive of the culprit who has caused it to occur, whereas when an event that is not prevented from occurring turns out to have occurred, we do not bother to think of the person who has caused it to occur, although objectively both adversative events and beneficial events equally require causers. Seen from a broader perspective, therefore, the predominance of the adversative reading can be made sense of as a manifestation of a general tendency in construing events, and we do not need to think that the adversative reading is arbitrarily conventionalized for the Japanese adversative passive, nor do we need to attribute it to the alleged blocking effect. This constitutes our solution to the third problem with Shibatani's analysis. 4.
Apparent counterexamples There are two types of apparent counterexamples to our account. One type consists of non-intentional verbs like shinu, "die", in (21). Since the subject of a non-intentional verb does not exercise his or her volitional control in the described event, it is apparently inappropriate to ascribe responsibility to the subject. To take (21) for example, it is perfectly acceptable if Taro's father didn't kill himself but just died a normal death. He probably didn't choose to die on his own will. However, we can nonetheless argue that he can still be held responsible for the consequences of his death. In the idealized Japanese cultural model of obligations, a father's premature death before finishing his paternal duty counts as an act of negligence done to his children. This can be seen in the fact that we will not use (21) if Taro is a fully independent adult who does not need his father's financial and moral support any longer.9 (21) Taro wa chichioya ni shin-are-ta Taro TOP father by die-PASS-PAST (Lit.) "Taro was died by father" "Taro's father died on him" Another similar example is (22).
TRANSITIVE CONSTRUAL IN THE JAPANESE ADVERSATIVE PASSIVE 295
(22) Wareware wa kanjinna toki ni shachoo ni We TOP important time at president by taore-rare-ta collapse-PASS-PAST (Lit.) "We were collapsed by president at important time" "The president got ill when we needed him/her badly" The verb taoreru, "collapse, or become sick", is a non-intentional verb, but a president of a company can be criticized and blamed for getting ill and its consequences. Another class of apparent counterexamples are weather verbs like ame/yuki ga furu, "rain/snow falls", in (23). (23) Taro wa ame/yuki ni fur-are-ta node Taro TOP rain/snow by fall-PASS-PAST because gaishutsu-deki-nakat-ta go=out-can-not-PAST (Lit.) "Because Taro was fallen by rain/snow, (he) couldn't go out" "It rained/snowed, forcing Taro to stay home" There is interesting contrast between rain and snow in (23) and earth quakes and lightning in (9) and (10). As we saw in (9) and (10), earthquakes and lightning cannot make acceptable adversative passive sentences in contrast to the raining and snowing in (23). Since rain and snow are inanimate, in order to maintain our account, we need to find out what aspects of raining and snowing make them agentive, if in a non-prototypical manner. In this connection, Van Oosten (1986:120) points out that raining and snowing normally last for some time, in such a way that we can recognize what is going on or what they are doing. Since dynamic events do not continue unless they are kept going on in some way, there is reason to construe raining and snowing as a kind of self-sustained activity as opposed to a static state. The unacceptability of (9) and (10) is due to the non-durative nature of the events denoted by the verbs. Earthquakes and lightning can bc devastating, but they occur out of the blue and last only for a very short time, and they disappear in an instant without leaving trace in the case of lightning, or it cannot be identified independently of trembling objects in the case of an earthquake.10 In a sense, raining and snowing are like a person running, while
296
EIJIRO TSUBOI
earthquakes and lightning are like a vase breaking. It will be interesting to note here that (24) sounds unnatural as compared with (23): (24) *Taro wa ame/yuki ni yam-are-ta node Taro TOP rain/snow by stopvi-PASS-PAST because kyooren ni de-ta drill to attend-PAST (Lit.) "Because Taro was stoppedvi by rain/snow, (he) attended drill" "*It stopped raining/snowing, forcing Taro to attend the drill" Yamu, "to stop", in (24) is an instantaneous event and is a kind of change of state rather than what rain or snow does. Klaiman (1991) argues that the Japanese adversative passive is a submissive construction whose subject must have some control over the situation denoted by the passive verb, and that the difference between earthquakes and lightning on the one hand and rain and snow on the other can be traced to the fact that the former kind of events are not so common as the latter, and are therefore unforeseeable and uncon trollable. (24) shows that such an analysis is untenable, since it stops raining or snowing just as often as it rains or snows. That the passive subject need not have control over the denoted situation is also shown by (21) and (22) where the subjects are presented as suffering from the events which are beyond their control. Also, although in (23) furu in ame/yuki ga furu, "rain/snow falls", is glossed as "fall", as it is usually glossed, actually furu is not just "fall", since furu is only used with rain drops, snow flakes, confetti, and the like, and seems to incorporate some notion of MANNER like "in small drops or pieces", as Ruwet (1991:110-111) observes. In contrast to furu, ochiru in kaminari ga ochiru, "lightning falls", in (10) is "fall", and seems to present lightning as something that 'falls' spontaneously. Examples (25) and (26) are cited as further evidence suggesting the action status of such weather phenomena as raining and snowing. In the dialogue in (25), "It's raining" or "It's snowing" can answer the question "What's it doing?" In (26), raining, snowing, and blowing can take hard, which usually modifies activities. (25) "What's it doing?" "It's raining/snowing" [Chafe 1970:101-102] (26) It is raining/snowing/blowing hard. [Kageyama 1996:76]
TRANSITIVE CONSTRUAL IN THE JAPANESE ADVERSATIVE PASSIVE 297
Examples involving ame/yuki gafuru, "rain/snow falls", have often been given as typical examples of the Japanese adversative passive, and have been giving us the wrong impression that the adversative effect is enough. However, they are by no means prototypical, in that there is no prototypical responsible actor. Therefore, if we can regard the weather verb examples like (23) as nonprototypical activity sentences, the adversative passive does not look so differ ent from the garden-variety passives as it has been traditionally made out to be, since it is related to the ordinary passive in that it also means that something is done to the subject. 5.
Conclusion We have examined the semantics of the Japanese adversative passive and shown how the typologically peculiar characteristics of the valency increase and the adversity sense are motivated by the cognitive models concerning event causation and responsibility, as well as how the models work to produce the characteristics in conjunction with the properties of the Japanese adversative passive and the readiness Japanese exhibits to extend transitive construal based on the presence of a responsible actor. We have also suggested that the pre dominance of the adversative meaning can be attributed to a general tendency in event construal. Although it is common practice to talk collectively about the adversative passives in the East Asian languages such as Japanese and Vietnamese, several important studies have recently been conducted that show how they differ.11 However, generally they still just deal with the distributional differences, and so far left unexplored is why they differ as they do. Such exploration is only possible through fine-grained analyses of the adversative passive in each language, to which this study is intended to contribute. Notes * This paper is based on the research conducted during my stay as a visiting scholar at the linguistics department of the University of California, San Diego, from 1994 to 1995. I am very grateful for the pleasant research environment and hospitality offered by Ron Langacker and other people at UCSD. I am particularly indebted to Ron Langacker, Gilles Fauconnier, Yoshiki Nishimura, and Masuhiro Nomura for their comments on earlier versions of the manuscript. 1 This is one of the two forms of the passive morpheme, and -are- appears when it is attached to a verb stem ending with a consonant, as in (1), and when the verb stem ends with a vowel, as in example (4), -rare- appears. 2 This quotation is from Shibatani (1990), but the basic idea expressed here, that adversative passive sentences are derived from sentences whose described events do not directly affect the passive subjects, is proposed by Wierzbicka (1979) and further discussed by Kuno (1983), and Shibatani's generalization cited here is actually a refinement on their ideas.
298
EIJIRO TSUBOI
3 In this connection, it is to be noticed that example (4) is derived from an ordinary transitive sentence (5), and does not involve any extra-thematic NP, but is clearly adversative and should be regarded as an instance of the adversative passive. Therefore, if the generalization made by Shibatani in (7) is to accommodate examples like (4), we need to interpret what he calls 'extra-thematic NP' in (7) as 'unaffected NP' as in his earlier generalization in (6), but the English counterpart of (4), which is the literal translation in (4), is unacceptable, regardless of whether we interpret it adversatively or not. Therefore, Shibatani's generalization in (7) should be taken as one that needs to be relativized and adjusted to individual languages. 4 This is not intended to apply to what is often called ' niyotte-passive', which is a relatively new construction in the history of Japanese which appeared through the influence of European languages and marks the agent with -niyotte instead of -ni. The adversative passive discussed in this paper does not allow -niyotte. According to Kinsui (1997:776), "[t]he ni-passive indicates that the derived subject is affected by the action of another participant", while the niyotte-passive "describes an activity of a human being or human beings from a completely neutral standpoint". See Kinsui (1997) for details about the Japanese -niyotte passive. 5 Kuno (1983:205) treats the ordinary passive and the adversative passive as if they were two complementarily distributed variant interpretations of the same passive morpheme conditioned by whether or not the referent of the subject is affected in the event described by the verb, and Shibatani (1994:483) mentions this analysis of Kuno's as one similar to his. His statement in (7) is actually quite similar to Kuno's original generalization. 6 This is one of the points that suggest that the adversative passive constructions in the East Asian languages are not necessary uniform and need finer-grained analyses before being lumped together. 7 Since treating one's house as if like one's own body part is closely associated with the traditional Japanese notion of one's house as something handed down from father to son that represents the keystone of the patriarchal, pre-war Japanese society, younger Japanese, who seem to be relatively free from the traditional norms and values, are expected to be less likely than the older generation to express the event in (18) using the transitive construction. As far as the author checked with about ten Japanese in each age bracket, this expectation was borne out and many of the younger consultants found (18) unnatural and all of them preferred to use the passive construction below: Watashi wa kuushuu de ie wo yak-are-ta I TOP air=raid with house ACC burnvt-PASS-PAST (Lit.) "I was burntvt house in air raid" 'T got my house burnt down in an air raid" The object-retained passive sentences like the above are not discussed in this paper. See Kim (1994) for discussion of the differences between the Korean and Japanese object-retained passive constructions. 8 Tsuboi (1997) discusses other phenomena exhibiting the asymmetry and provides a force-dynamic account of it. One of the constructions discussed there was the English 'fake resultative' construction, as exemplified by the following sentences, which is parallel to the Japanese adversative passive in its valency increase and typical adversative meaning although different in many ways:
TRANSITIVE CONSTRUAL IN THE JAPANESE ADVERSATIVE PASSIVE 299
(i) The joggers ran their Nikes threadbare (ii) The chef cooked the kitchen walls black The predominance of adversative interpretation in this construction was brought to my attention by Yoshiki Nishimura (p.c.), and is mentioned in passing in Goldberg (1991:72). 9 An anonymous reviewer disputed the presence of responsibility ascribed to the ni-marked NPs in (21) and (22), but the result of consulting about a dozen native speakers of Japanese confirmed that it indeed exists. As for (22), even in a situation in which the sudden absence of the president put his or her subordinates in deep trouble, if they think with sympathy that it is no wonder the president got ill (because he or she has been working too much for the company to keep their jobs) and do not have accusing feelings about it, they will not use (22). 10 An anonymous reviewer pointed out that it is debatable to say that an earthquake is non-durative, but okiru ("occur") in (9) does present an earthquake as non-durative, which can be seen in the impossibility of adding a duration phrase: Kinou jishin ga (*juu-byoo-kan) oki-ta yesterday earthquake NOM (ten-second-for) occur-PAST "An earthquake occurred yesterday (*for ten seconds)" When an earthquake is presented as durative, other verbs like tsuduku ("continue") will be called for. 11 See Shibatani (1990), Washio (1995), and Kim (1994), among others.
References Alfonso, Anthony. 1971. "On the 'Adversative' Passive in Japanese." Association of Teachers of Japanese, Journal-Newsletter 8. Chafe, Wallace L. 1970. Meaning and the Structure of Language. Chicago: University of Chicago Press. Goldberg, Adele E. 1991. "A Semantic Account of Resultatives." Linguistic Analysis 21.66-96. Kageyama, Taro. 1996. Dooshi Imiron. Tokyo: Kurosio Publishers. Kim, Kyunghwan. 1994. "Adversity and Retained Object Passive Constructions." Japanese/Korean Linguistics 4.331-346. Kinsui, Satoshi. 1997. "The Influence of Translation on the Historical Development of the Japanese Passive Construction." Journal of Pragmatics 28.759-779. Klaiman, M.H. 1991. Grammatical Voice. Cambridge: Cambridge University Press. Kuno, Susumu. 1973. The Structure of the Japanese Language. Cambridge, Mass.: MIT Press. . 1983. Shin Nihon Bunpoo Kenkyuu. Tokyo: Taishuukan Shoten. Mikami, Akira. 1953. Gendai Gohoo Josetsu. Tookoo Shoin.
300
EIJIRO TSUBOI
Nishimura, Yoshiki. 1993. "Agentivity in Cognitive Grammar." Conceptualizations and Mental Processing in Language eds. R.A. Geiger & B. Rudzka-Ostyn, 487-530. Berlin: Mouton de Gruyter. Nishio, Hiroko. 1982. Multiple Nominative and Adversity Constructions in Japanese. Ph.D. dissertation, University of Arizona. Oehrle, R.T. & H. Nishio. 1981. "Adversity." Coyote Papers 2.163-185. Palmer, Frank R. 1994. Grammatical Roles and Relations. Cambridge: Cambridge University Press. Ruwet, Nicolas. 1991. Syntax and Human Experience. Chicago: University of Chicago Press. Shibatani, Masayoshi. 1990. The Languages of Japan. Cambridge: Cambridge University Press. . 1994. "An Integrational Approach to Possessor Raising, Ethical Datives, and Adversative Passives." Berkeley Linguistics Society 20.461-486. Tsuboi, Eijiro. 1997. "Usage-Based Model as a Model for Dynamic Aspects of Language." Paper presented at the twenty-second annual meeting of Kansai Linguistic Society. Van Oosten, Jeane. 1986. The Nature of Subjects, Topics and Agents: A Cognitive Explanation. Bloomington: Indiana University Linguistics Club. Washio, Ryuichi. 1995. Interpreting the Voice. Tokyo: Kaitakusha. Wierzbicka, Anna. 1979. "The Japanese 'Adversative' Passive in a Typological Context." Papers in Linguistics 12: 111-162. (Reprinted in Wierzbicka 1988.257-292.) . 1988. The Semantics of Grammar. Amsterdam & Philadelphia: John Benjamins.
Caused-Motion and the 'Bottom-Up' Role of Grammar Frederike van der Leek University of Amsterdam
1.
Introduction The grammatical structures of sentences, in the words of Dennett (1991:300-1), "lay down some of the tracks on which 'thoughts' can then travel. " This statement raises a crucial question: What is the substrate to which grammar donates the structures that thoughts "can then travel" on?1 For advocates of the autonomy stance the above question does not arise, in that grammatical structure is at the beginning of things, so to speak. Thought in itself is, in line with Cartesian linguistics, looked upon as "unified and undifferentiated" (Chomsky 1966:47-48). Language is needed to provide thought with formal structure, and thereby make it accessible for under standing. The linguistic forms known as sentences — grammatically structured strings of words — thus constitute the autonomous foundation upon which semantic interpretation can proceed. Recently this conception, known as inter pretive semantics, has changed in that semantic structure is now seen as formed independently of grammar (Pinker 1989).2 Words have basic (or lexically transformed) semantic structure which is built out of innate primitive concepts, and lexical-semantic structure gets hooked onto independently generated grammatical structure by universal linking rules. The language system as a whole is a basic component of the human mind, a cognitive module that is independent of language-external cognition; this built-in language faculty leaves room for parametric variation so that individual languages can instantiate it in different ways. Thought is, in short, structured by language both form- and meaning wise, with grammatical as well as semantic structure autonomous relative to each other as well as to nonlinguistic cognition. Proponents of cognitive grammar locate grammatical structures, in a sense, at the other end of the spectrum. Thought has internal structure indepen dently of language. Humans, that is, have the cognitive ability to construct frames (Fillmore 1982, 1985) or, in Lakoff's (1987) terminology, Idealized Cognitive Models (ICMs), i.e. nonlinguistic schematic conceptual represen tations of, roughly, objects, and events.3 These ICMs, which arise directly (or indirectly) out of experience with the world, structure thought and form the background against which linguistic concepts are formed, understood, and
302
FREDERIKE VAN DER LEEK
organized. The earliest of such models, developing prior to language, are spatial sketches constructed out of perceptual images,4 simplified represen tations that globally parse the world in terms of animate/inanimate objects, biological/mechanical motion, self-motion/caused-motion, path of motion etc. (Mandler 1998:262ff). It is, as Langacker (1991), amongst others, argues, out of such prelinguistic 'image-schematic' ICMs that the prototypical senses of e.g. intransitive, transitive and caused-motion constructions arise. Words and grammatical structures thus symbolize, in this view, conceptual imagery (cf. Langacker 1987:110ff and Mandler 1998:263 on the notion 'imagery'; there should be no confusion with perceptual images); the images get imposed on nonlinguistic, but conceptually structured, thought, thereby foregrounding, or profiling in Langacker's terminology, certain of its facets and backgrounding others. The resulting meaning resides in the conceptual content of the nonlinguistic thought (Langacker's "base"; cf. 1987:183ff) in combination with the linguistically imposed profile. The alternate tracks, to return to Dennett's metaphor, that grammatical structures can lay down, thus allow thought to 'travel on' in conceptually transformed guises, guises that do not change the thought's content in essential ways but bring different facets of it to the fore. To recapitulate, the two camps differ as follows. Knowing a language is, in the first framework, viewed as solely coming about through linguistic cognition acting upon incoming language data, with grammar and semantics functioning as independent submodules.5 In the other approach, language (both grammar and words) is mapped onto prior conceptual structure, and linguistic cognition has the function of "tailor[ing conceptualizations] to the specifica tions of linguistic convention" (Langacker 1987:99). In the present paper I intend to investigate a third possibility, one that sides with Cognitive Linguistics in viewing language as an integral facet of human cognition but that allots a more powerful role to grammatical structure. I will, to put it briefly, hypothesize that grammatical structure, through 're entrant mapping' (Edelman 1989, 1992), finds its way back into perception and can, at least where "thinking for speaking" is concerned (Slobin 1996:76),6 structure thought 'bottom-up'.7 In what follows I will argue for and investigate this hypothesis relative to what is known as the caused-motion construction in English and the account of it proposed by Goldberg (1995). First, however, I will briefly survey two largely opposing cognitive linguistic views concerning verbal/constructional polysemy. 2.
Verbal and constructional polysemy: different cognitive approaches Though Cognitive Linguistics cannot be said to form a single unified framework (yet), it seems united with respect to two points: firstly, language
CAUSED-MOTION AND THE 'BOTTOM-UP' ROLE OF GRAMMAR
303
(words and grammar) structures nonlinguistic conceptualizations that are grounded in experience, and secondly, language underdetermines meaning. Opinions differ, however, on the question of polysemy relative to words and the grammatical constructions they feature in; in what follows I will concentrate on verbs featuring in subject-complement constructions (henceforth simply referred to as complement constructions). Langacker for instance argues that verbs are usually polysemous; if a verb is polysemous, each of its different uses has a semantic structure of its own (which, unlike what e.g. Fodor 1975, Jackendoff 1990 or Pinker 1989 advocate, arises out of nonlinguistic cognition that is ultimately grounded in experience). Each conventional verbal sense neatly fits in with the maximally schematic semantic structure of the particular complement construction that forms 'its' grammatical niche (cf. e.g. Langacker's 1995:32-33 account of two particular senses of the verb wash)} Langacker thus very much emphasizes how conventional polysemous verb senses become an entrenched part of the language user's knowledge of his language. Goldberg (1995), on the other hand, takes up a position against verbal polysemy and argues that verbs have, essentially, one basic sense, while the complement constructions verbs can occur in, are themselves polysemous, having a central sense as well as extended senses, independently of lexical content. Her basic argument is that, in their central sense, complement constructions "reflect scenes basic to human experience" (1995:5); the extended senses, all paired up with the same grammatical form, make up a family resemblance network around the central sense, to which they are related by linguistic convention. If a verb's (basic) sense, she further argues, is com patible with either the central sense of a construction or any of its extended senses, it is licensed in the construction, and the verbal sense gets integrated with the (relevant) constructional sense.9 The essential arguments that she offers in favour of her approach, are (i) that excessive lexical polysemy is avoided, and (ii) that the circularity inherent in a lexical polysemy approach is avoided as well. Now an essential problem of Langacker's approach is that it is not clear how truly novel verb senses, ones whose usage in a particular complement construction is not (yet) conventionalized, get licensed. Since that is a central issue Goldberg is concerned with, I will, in what follows, primarily concern myself with her arguments that it is complement constructions rather than verbs that exhibit polysemy. 3.
The caused-motion construction Goldberg (1995) shifts, as pointed out above, polysemy from verbal to constructional senses, this with respect to a number of complement construc tions, one of them the caused-motion type. Given, for instance, the caused-
304
FREDERIKE VAN DER LEEK
motion sentence Cindy blew the dust off the book, Goldberg's claim is that the complement form underlying this sentence, i.e. [NP V NP PP/ADVP] or, equivalently, 'SUBJ, V, OBJ, OBL', pairs up with an independent schematic semantic structure roughly definable as 'cause to move'. The basic sense of blow (as based on our real world knowledge) characterizes, again roughly, a process involving a force creating an air current, as in Cindy blew very hard (to get all the candles of her birthday cake out in one go).10 Fusion of the basic sense of blow with the constructional sense in the first sentence then yields the interpretation that this sentence is said to have, i.e. that, due to Cindy's blowing, the dust in question is caused to move off the book. There is, in other words, no need to postulate a second sense for blow, e.g. something like 'cause to move by blowing', to account for its use in the sentence Cindy blew the dust off the book. In a lexical approach like Pinker's, "rampant lexical polysemy" (Goldberg 1995:154) is inevitable; Goldberg points out, for instance, that the verb kick can occur with (at least) as many as eight different senses, senses that correspond to an equal number of different complement configurations (1995:11). The lexical viewpoint presents, in other words, a perspective of conventional usage, rather than of dynamic cognition. In fact, as Goldberg points out (same page), the approach is unavoidably circular as well: if verb χ has sense Y, it can occur in complement configuration z, and if verb x can occur in complement configuration z, it has sense Y. (Note that Goldberg only makes this point relative to Pinker's lexical rule approach. Langacker's position in this respect is more cautious and hence more difficult to evaluate; space limitations make it impossible to consider his views in any detail in this article, but see Beliën and van der Leek forthcoming on this issue.) Goldberg's approach relies, one may conclude, less on linguistic rules than does Pinker, thereby avoiding arbitrary lexical stipulations, and leaving more to what Fauconnier (1994:xviiff) so aptly calls "backstage cognition". However, there are both empirical and methodological problems with what she proposes as well. Before going into these, I will first briefly mention the constructional senses of the caused-motion construction that Goldberg proposes. 3.1
The constructional senses of the caused-motion construction described Once a central constructional sense has been posited for a particular complement construction, felicitous combinations of the complement construc tion with members of a class of verbs whose basic meaning does not fit in with this central sense, must be accounted for in terms of an extended constructional sense. Goldberg in fact assumes, with Lakoff (1987), that, while there is no
CAUSED-MOTION AND THE 'BOTTOM-UP' ROLE OF GRAMMAR
305
predicting an extended sense, the more semantic structure it inherits from the parent sense, the more 'motivated' it is (hence the less arbitrary). Faced with data requiring subsenses for the caused-motion construction, she therefore naturally attempts to provide optimally motivated subsenses. The following characterizations, taken verbatim from Goldberg (1995:16Iff), speak largely for themselves, with the exception of category VI; I will therefore simply list I-V, and briefly explain VI. Category I defines the central constructional sense, while categories II-VI represent extended con structional senses. The subsequent example sentences are, apart from minor alterations, due to Goldberg, except for sentence (3). The numbering is my own; sentence (1) is to be understood as exemplifying central sense I, sentence (2) as exemplifying subsense II, etc. I
' X CAUSES Y to MOVE Z '
II The conditions of satisfaction associated with the act denoted by the predicate entail: 'X causes Y to move Z' III ' X ENABLES Y to MOVE Z '
IV 'X PREVENTS Y from MOVING Z' V
' X HELPS Y to MOVE Z '
VI A Further Extension (1) Frank sneezed the tissue off the nightstand (2) Trenarrow gestured St. James into one of the room's wingback chairs. [Elizabeth George, Payment in Blood, 1991; slightly adapted] (3) Sam let Bill into the room (4) Harry locked Joe into the bathroom (5) Sam helped Bob into the car (6) Sam accompanied Bob into the room Category VI involves, as Goldberg points out, a few verbs (apart from accompany she mentions follow, trail and tail) that are perfectly acceptable in the complement construction under discussion, but that do not fit in with any of the other five senses, cf. example (6) above. This "group of exceptions", to use Goldberg's own words, "involve[s] a subclass of verbs which entail that the agent argument as well as the theme argument move along the specified
306
FREDERIKE VAN DER LEEK
path ... [and] can be recognized as a further extension to the pattern" (1995:164). Positing the caused-motion constructional senses outlined above, serves the purpose of generalizing, semantically, across individual verbs and thereby avoiding arbitrary lexical stipulations. From this viewpoint, Goldberg's way of dealing with verb alternation is definitely attractive. There are, however, various reasons to doubt its ultimate tenability. In fact, I will present argu ments in sections 4 and 5 suggesting that there is no need for constructional senses, and that lexical polysemy can be avoided by relying much more on dynamic cognitive selection (which comes down to compositionality) rather than on instruction (i.e. linguistic convention). However, in the following section I will first, taking Goldberg's theory as starting point, critically comment on the constructional senses Goldberg proposes as extensions of the caused-motion construction. 3.2
The caused-motion constructional subsenses: some problems The characterization Goldberg suggests for constructional sense II says, in a nutshell, that the satisfaction of certain conditions (as determined by the predicate) entails caused-motion. Note that this description is essentially different from the way the other senses are worded, and this description forms a problem. This way of defining the constructional sense may, to be sure, provide a motivational link with the parent sense, but it does not define what the constructional form itself means. Claiming that sentence (2), Trenarrow gestured St. James into one of the room's wingback chairs, entails that Trenarrow caused St. James to move into the said chair provided the relevant satisfaction conditions are met, is obviously correct. However, this simply does not tell us what the constructional sense underlying sentence (2) itself is. By Goldberg's own standards, we cannot understand the sentence unless we know (a) what the basic meaning of the verb is (plus its frame semantics) and (b) the constructional sense underlying its complement structure.11 Given there is to be such a constructional sense, it is more accurately described as 'X COMMUNICATES to Y to MOVE Z'. In other words, the constructional sense proper is less motivated than Goldberg would have us believe. And worse, this newly suggested constructional sense excludes a particular type of verb from those licensed by Goldberg's own constructional sense II (order, ask, invite, gesture etc.). Goldberg, that is, includes example (7) as an instance of the constructional sense in question: (7) Sam sent him to the market
CAUSED-MOTION AND THE 'BOTTOM-UP' ROLE OF GRAMMAR
307
However, it is hard to maintain that the basic sense of the verb send can be associated with a type of communication between the agent argument and the theme argument, as is obvious if we compare (7) with e.g. Biddy sent the letter to your home address. Moreover, though Sam sent him to the market does not, indeed, guarantee that the 'him' actually got to the market, it does guarantee that he went somewhere; the example is, in other words, not on a par with type (2) sentences. It seems, in short, as if there is, as it stands, no place in Goldberg's system of constructional senses for type (7) sentences. A problem not unsimilar to the one noted above for constructional sense II, crops up in a different guise with respect to sense IV, 'X PREVENTS Y from MOVING Z'. Consider sentence (4) again, here repeated for convenience: (4) Harry locked Joe into the bathroom This sentence conveys something like "Harry locked the door from the outside once Joe had gone into it". Notice therefore that the 'prevent from moving' sense, just as with sense II, characterizes what type (4) sentences entail rather than accounting for the semantic structure (i.e. constructional sense) they share as such. This is especially clear in cases like these, because the direction explicitly encoded in type (4) sentences is, of course, the opposite from the direction implied by the alleged constructional sense: Harry locked Joe into the bathroom, thereby preventing him from moving out. I can only conclude that 'prevent from moving' does not specify the (partial) semantic structure of type (4) sentences as such. With respect to what Goldberg, in connection with example (6), refers to as a "group of exceptions", it can, furthermore, be observed that the only reason to speak of 'exceptions' is the original claim that type (1) sentences have a constructional sense of 'cause to move'. The verbs making up class VI do not involve causation, motion is part of their basic meaning, and the agent argument (rather than the theme argument) is explicitly described as following the path mentioned in the sentence. Observe the special importance of this last fact: the oblique complement in category VI apparently functions as 'subject complement' rather than 'object complement' (the latter is the case for categories I-V). This actually means the syntax of the complement config uration under discussion is different for category VI. Note that the constructional sense Goldberg suggests for this class (i.e. that this type of sentences entails that both the agent argument and the theme argument move along the described path) represents an incorrect character ization of what sentences like (6) mean: entailment is not what is involved. What needs to be explained by Goldberg's constructional sense but is not, is
308
FREDERIKE VAN DER LEEK
how we know that the theme argument also moves along the path that is followed by the agent argument. Goldberg insists that the basic meaning of verbs must (pace Pinker 1989) be allowed "to be associated with [their] rich frame-semantic meanings" (1995:27). It seems to me, however, that it is precisely our understanding of the basic meaning of, say, follow, definable as 'go behind some entity χ in a certain direction', plus the associated framesemantic meaning that entity χ itself moves (must, in fact, move) in the same direction, that tells us what the (backgrounded, because frame-semantic) role of the theme argument in Sam followed Bob into the room comes down to. Grammatically, in summary, the goal complement functions as 'subject complement', while at the frame-semantic level, it informs us about the theme argument's path. I conclude that no constructional sense is needed to under stand type (6) sentences: the (basic plus frame-semantic) meanings of the verbs in question speak, as the saying goes, for themselves. In the above, it has become clear that, on closer inspection, a number of flaws show up relative to three of the caused-motion constructional subsenses that cannot be ignored and that are not easily mended either. Subsenses III and V seem alright, although even here doubts may arise as to their adequacy. Clearly, the ENABLE part of subsense III, 'X ENABLES Y to MOVE Z', is an accurate generalization over the meanings of the class of verbs participating (apart from let, Goldberg mentions allow, free and release). Observe, however, that 'enable to move' does not guarantee movement on the part of the theme argument (cf. Cindy enabled Bill to move into the house, but he stayed outside). Notice that caused-motion sentences featuring class III verbs do, on the other hand, guarantee actual motion on the part of the theme argument when used in the 'enable to move' construction type (cf. the unacceptability of *Cindy let Bill into the house but he stayed outside). The only conclusion that I can draw at this moment, is that constructional sense III does not account for the fact that type (3) sentences guarantee motion on the part of the theme argument, whereas this is what constructional sense III is supposed to account for. With respect, finally, to subsense V, 'X HELPS Y to MOVE Z', this is, as far as the predicate HELP is concerned, the correct gener alization over three out of the five verbs Goldberg mentions for this class, i.e. help, assist and guide. There are, however, two cases left that are somewhat more complex, show and walk, cf. examples (8) and (9) below (both from Goldberg 1995:162): (8) Sam showed him into the living room (9) Sam walked him to the car
CAUSED-MOTION AND THE 'BOTTOM-UP' ROLE OF GRAMMAR
309
It seems to me (i) that the sense of accompaniment that (8) and (9) both imply (but that is not inherent in the meaning of the verbs as such), is not accounted for by subsense V, (ii) that as far as show is concerned this signals enablement more than help, and (iii) that example (9) may, but need not imply help: I can walk someone to his car for the sheer pleasure of being in his company a bit longer. Note that the Longman Dictionary of Contemporary English (ed. Procter, 1978:1235; henceforth LDCE) characterizes the causedmotion use of walk simply as "to go on foot with (someone), usu. to a stated place: I walked her home". In fact, Langacker (1987:291) mentions a causedmotion usage of walk where the concept HELP is downright inappropriate, cf. sentence (10) below: (10) Sharon walked the ladder across the room All in all, I conclude that close scrutiny reveals that the constructional subsenses Goldberg proposes for the caused-motion construction are not nearly as convincing as seems to be the case when one takes them more at face value. This is not to say, however, that the patterns that she has unearthed play no role in language. It is, as I hope to show, just that they play a different role than Goldberg argues they do. What the above findings suggest, in my view, is that Goldberg is definitely on the right track by arguing against lexical polysemy for verbs and emphasizing the crucial role played by cognition in the form of Fillmorian frames. However, bringing polysemy in again through the back door so to speak, in the form of constructional senses, is not, it seems to me, the right solution. Below I will point out further problems inherent in Goldberg's constructional sense proposal. 4.
Arguments against additional CAUSE and MOVE verb senses In the following section, I will attempt to make clear that fusing a verb's basic meaning with 'X CAUSES Y to MOVE Z' does not yield the correct semantic results for caused-motion sentences, and to show why this is the case. In the subsequent section, I will describe what problems are actively created by positing the additional verb sense MOVE, and explain why the notion of motion is nevertheless somehow part of our understanding with respect to the construction in all its senses. 4.1 Fusing CAUSE to MOVE and basic verb meaning Consider the following two pairs of sentences, the a-version in each case an original sentence instantiating the caused-motion construction, the b-variant
310
FREDERIKE VAN DER LEEK
a rendering in the spirit of Goldberg's proposal (I use the causative verb make rather than cause so as to avoid having to use the infinitive particle to): (11) a. Roberta shrugged herself into the stripy coat [Dick Francis, Enquiry, 1971; slightly adapted] b. Roberta made herself move into the stripy coat by shrugging (12) a. I divided it [=the steak] onto two plates [Dick Francis, Nerve, 1965] b. I made the steak move onto two plates What examples (l1b) and (12b), which follow Goldberg's proposal in that the verbs shrug and divide are used in their basic meaning while being integrated into the 'cause to move' semantic structure, show, at least to me, is that when you really put the constructional sense proposal literally to the test relative to examples that are not as suggestive as, for instance, the sneeze examples, the effect is unimpressive, to say the least. To get to what is, in my view, really wrong with the constructional sense proposal, compare examples (13) and (14a-b) below: (13) Joe kicked the dog into the bathroom [Goldberg 1995:153] (14) a. Joe made the dog move into the bathroom by kicking it b. Joe made the dog move into the bathroom by kicking it there Notice that sentence (14a), which fuses the basic meaning of kick with Goldberg's central constructional sense, does not necessarily cover the same situations that sentence (13) covers. That is, (13) can only be used if Joe kicked the dog in the direction of the bathroom. (14a), however, only conveys that Joe kicked the dog, regardless the direction of the kicking, and that this happened to make the dog move into the bathroom. By adding there in (14b), we guarantee, of course, that Joe's kicking is in the right direction (as (13) does). Notice, however, that the moment there is explicitly added so that the correct meaning of (13) is actually guaranteed, the problem of infinite regress rears its ugly head. The reason is obvious: the constructional sense that is made explicit in (14b), at the same time underlies the by-phrase again. There is, in other words, no stopping the alleged constructional sense from having to be present again and again, ad infinitum. Notice that Langacker does not face the problem Goldberg does, since he assumes that verbs are polysemous. This means that a verb can occur in the construction in question if, and only if, it has, amongst others, one convention-
CAUSED-MOTION AND THE 'BOTTOM-UP' ROLE OF GRAMMAR
311
al semantic structure that maps onto the complement construction's semantic structure (be this prototypical or maximally general). However, as pointed out in section 2, Langacker does not seem to be able to account for novel usages. If I am right in saying that Goldberg's fusing proposal does not yield the correct semantic results, it follows that her proposal cannot satisfactorily deal with either conventional or novel usages. My conclusion is, then, that we are still in need of a proposal that can account for conventional as well as novel usages relative to the caused-motion construction proper. 4.2
The effect of the linguistic presence of MOVE In the present section I will argue that the absence of the verb move (or a verb close in meaning) in caused-motion sentences (ranging over all the classes, I through VI) has a particular semantic effect. Let me begin by dis cussing some instances of the central caused-motion sentences. When the verb move is absent in this type of sentence, where actual motion on the part of the theme argument is involved, this signals that the language user construes (cf. e.g. Langacker 1987:138ff on the notion 'construal') this theme argument as moving purely as a consequence of the agent argument's action, like an autom aton or, in the case of animate beings, almost zombie-like, will-lessly. This in turn helps explain the ungrammatical status of certain sentences whose unacceptability would otherwise remain a mystery. Let me begin by carefully considering two sets of three examples, (1517) and (18-20), involving an animate and an inanimate theme argument, and the main verbs trot and shoot, respectively. The first sentence of both three somes is an instance of category I ('cause to move'). The second features make as main verb, with trot and shoot embedded as explicit intransitive bare stem motion verbs. The third example in both trios simply instantiates both trot and shoot as intransitive motion predicates, but now used as main verbs. (15) She trotted the horse into the stable [Mandelblit & Fauconnier, this volume] (16) She made the horse trot into the stable (17) The horse trotted into the stable (18) Pat shot the ball right across the field (19) ??Pat made the ball shoot right across the field (20) The ball shot right across the field
312
FREDERIKE VAN DER LEEK
Notice that the above sets make an interesting pattern, in that only (19) is quite odd (as indicated by the double question mark). The question is, what this reveals about the meaning of the caused-motion construction used in (15) and (18). These latter examples draw, as observed above, all the attention to the action of the agent argument, leaving the theme argument's motion backgrounded. This has, to begin with (15), a noticeable effect on how we see the role of the theme argument in this sentence: the horse is seen as, in a manner of speaking, an automaton, merely acting out instructions. In other words, given that the agent argument is, for instance, actually riding the horse, she leg-pressures the horse into a trot, and uses the reins to signal the direction she has in mind for the theme argument, thereby determining the path the latter follows; merely following orders, the horse moves how and where it is 'told' to move by the agent argument. In a sense, then, the agent argument plays the role of 'trotter' (in the form of acting out a specific type of leg-pressure and rein-usage), even though, in the back of our minds, we know that the horse itself can also be said to be trotting, given the way its legs move. It is to get this specific relationship between rider and horse across that the caused-motion construction is used, because this has the effect that the horse's actual motion is not profiled (hence is not linguistically present). The situation is construed in an essentially different way when make is used, cf. example (16): the female in question is profiled as playing some (unspecified) causative role, with the horse directly reacting as a conscious animate being, by trotting into the stable. The objective situation may have been exactly the same as described earlier, be it that it is construed differently. Alternatively, the female in question may have done no more than, for instance, frighten the horse; the make-sentence is, in short, underspecified with regard to the factual role played by the agent argument.12 Sentence (17) is, in a sense even more neutral as to the background situation involved, but the construal is straightforward, in that only the horse's own action is profiled. As to sentences (18-20), it should now be clear why (19) is odd. (18) profiles the agent argument's part in the ball-shooting (the agent argument doing to the ball all it takes to make it count as 'shooting' and to guarantee the direction specified), and backgrounds the theme argument's motional role, while (20) only profiles the manner-motion role of the theme argument. (19), however, should potentially include a situation where the theme argument exhibits a reaction of its own; given the theme argument is a ball, this is hard to envisage. With regard to category II (involving verbs like ask, order, invite), the following observations can be made. Consider sentence (21):
CAUSED-MOTION AND THE 'BOTTOM-UP' ROLE OF GRAMMAR
313
(21) Sam invited Cindy into the room Note, first of all, that there are strict background conditions to be met for a sentence like (21) to be used felicitously: agent and theme argument must have direct communicative contact, the invitation to move along the path specified must be applicable 'then and there', and the motion must be agentargument-centered (in the sense of towards or away from the agent argument, depending on the latter's wishes). This, as I will argue in section 5.2, is a direct consequence of the grammatical configuration as such (cf. also note 17); the unacceptability of the following examples confirms the above observations: (22) *This morning Sam asked Cindy into his room later (23) *Sam asked Cindy, who was away in Paris, into his Berkeley office (24) *Sam ordered Cindy to the doctor All this makes sense, it seems to me, given that the language user picks sentence (21) to profile the agent-argument-centered, 'face to face' invitation and emphasize the passive role of the theme argument (she is merely the recipient of the invitation, and there is no indication whether she will, or even wants to, move into the room in question). It is the verb invite itself which evokes a frame expressing the agent argument's desire for the theme argu ment's company at the time of utterance. The notion of motion by the theme argument is, in short, only present in the agent argument's wishes — that is all the sentence gets across, at least. There is one verb in the class belonging to category II, bid, that can choose between the caused-motion construction or, alternatively, use a slightly different one, which adds an explicit bare stem motional verb (though this usage is becoming more and more old-fashioned), cf. the following examples: (25) Sam bade Cindy into the room (26) Sam bade Cindy come into the room Though judgments are very subtle, it seems to me that sentence (26), unlike (25), requires a background that makes it somehow clear that the theme argument is at least likely to go in (if asked). This explains why the relevant motional role is now profiled. With regard to category III ('enable to come'), we can note a similar pair of options relative to the verb let. Goldberg makes an interesting observation in this connection (without, however, providing an explanation). In the absence
314
FREDERIKE VAN DER LEEK
of an explicit embedded bare stem motion verb, the agent argument must actively remove the barrier that stops the theme argument from moving as desired. However, if a bare stem motion verb is present, enablement may be realized simply by "the failure to impose a potential barrier" (1995:162). The following examples, due to Goldberg (same page), nicely illustrate this difference: (27) *Sara let Bill into the room by leaving the door open (28) Sara let Bill come into the room by leaving the door open Enablement verbs like let (cf. Talmy 1985, 1988 on force dynamics) evoke a frame representing a scene involving a barrier that stops the object referent from doing something it wishes/is bound to do, and as doing it the moment the barrier is removed; this frame comes with such verbs' basic meaning. Notice that example (27) only profiles the agent argument's active role, leaving the theme argument's role unprofiled, thereby emphasizing its relative passivity. It therefore makes sense to understand this as saying that it is solely up to the agent argument to create the conditions such that the theme argument can move as it desires/is bound to. (27) would, not surprisingly therefore, have been perfectly fine if the by-phrase had said "by opening the door", since in that case no action on the theme argument's part would have been needed (apart from what the background frame already conveys). Because the second example does profile the theme argument's active role, it is no longer just evoked by the underlying frame. The situation is, in other words, construed differently, and the theme argument is acting explicitly (the dif ference is comparable to the difference between e.g. examples (15) and (16), which feature trot OBJ OBL and make OBJ trot OBL respectively). Leaving category IV to the last, I can be brief about category V ('help to move'); category VI has already been fully dealt with in section 3.2. The verb help can, as is well-known, allow the same options as earlier described for let and bid. The difference in construal is, in fact, becoming familiar, and is therefore not difficult to understand, witness (29) and (30) below: (29) Sam helped Biddy into the car (30) Sam helped Biddy get into the car Sentence (29) presents a construal of the situation where the theme argu ment need not actively do anything (except be, given the frame-semantics of help, co-operative, e.g. by just holding onto Sam); it is the agent argument's
CAUSED-MOTION AND THE 'BOTTOM-UP' ROLE OF GRAMMAR
315
role to move her into the car (say, by carrying her in). Sentence (30) presents a construal with the theme argument playing a much more active role, as is to be expected given that this role is now profiled. Notice that (31) is, therefore, quite dubious (it leaves no room for an active role by the theme argument): (31) ??Sam helped Biddy get into the car by carrying her in Category IV ('prevent from moving') presents an interesting puzzle. In section 3.2, I described sentence (4), Harry locked Joe into the bathroom, as objectively describing a situation with Joe going into the bathroom, and Harry locking the door on him. The verb lock, clearly, involves Talmy's force dy namics meaning, as also noted by Goldberg; what this means, in my view, is that, apart from lock's basic meaning (something like 'fasten something with a lock'), the combination of this verb with a directional phrase evokes a frame representing a scene involving a place that, being locked, forms a barrier for anyone/thing that has gone/is inside, from going/being taken out, and vice versa. It is, of course, the basic plus frame-semantic meaning of lock in combi nation with a directional phrase that is responsible for our understanding that the situation described by (4) prevents the theme argument from moving out of the bathroom. That the role of the theme argument in a sentence like (4) does not get profiled, makes sense therefore. The attention is not, after all, on the theme argument moving in but on a barrier getting formed. Notice that the opposition into I out of referred to in section 3.2 in connection with sentence (4) makes sense against the underlying frame involving a barrier stopping the theme argument from moving out of the bathroom. That is clearly why the combination lock + into is used: it profiles Joe's original direction (into the bathroom), which is, due to the barrier imposed, orthogonal to the direction the theme argument desires/is bound to move. This, it seems to me, is a very nice illustration of why Fillmore speaks of frames as not only motivating but also as forming a co-structure with the language expressions they underlie (1985:223). We in fact find another interesting confirmation of this costructuring function of frames in example (32) below, which uses in rather than into: (32) She locked her jewels in her jewel case [LDCE:643; slightly adapted] Given that jewels cannot themselves move into (or out of, for that matter) their case, it makes eminent sense to use locative in here rather than directional into.
316
FREDERIKE VAN DER LEEK
This issue of locative side by side with directional oblique complements in (various categories of) caused-motion sentences is one that requires further discussion, since Goldberg considers it a "major problem" (1995:158) for pro ponents of a compositional rather than a constructional account of the causedmotion construction. This issue is addressed in the following section. 4.3 Locative versus directional PPs in caused-motion constructions Lakoff (1987:42Iff) discusses the meanings of over in e.g. sentences like (33) and (34) below: (33) The bird flew over the yard (34) Sausalito is over the bridge His conclusion is that in (33) the preposition has a directional sense, whereas in (34) it has a locative sense. He argues that this latter locative sense is 'motivated' because it expresses 'end focus': the language user can link this usage with directional over because over, though locative in (34), can be understood to express the endpoint of a path; this extended sense thus has a conventional but motivated link with its directional source. Goldberg, whose caused-motion constructional senses all involveMOVEas an inherent compo nent of the constructions' linguistic meanings, faces a problem due to the fact that some caused-motion sentences take a locative oblique complement or can shift between a directional and a locative complement, cf. sentences (35) and (36) below. (36) should be ungrammatical, since 'cause too much salt to move in this food' is unacceptable: (35) She put the children to bed [LDCE:897, slightly adapted] (36) You put too much salt in this food [LDCE:897] Goldberg therefore suggests that a process of 'coercion' (cf. e.g. Croft 1991) is at work, which gives the locative preposition a directional reading. Goldberg adds that "coercion is not a purely pragmatic process; rather it is only licensed by particular constructions in the language", i.e. constructions that require the other reading (1995:159). The caused-motion construction, given that MOVE is an inherent part of any of its constructional senses, therefore allows such coercion.13 Talmy (1996), discussing cases like (36), argues, more perceptively to my mind, that end focus uses can be explained in terms of what he calls a "fictive" path, which is to be understood as referring to the "imaginai capacity of cognition" and the different uses are to
CAUSED-MOTION AND THE 'BOTTOM-UP' ROLE OF GRAMMAR
317
be seen as different perspectives from which the situation is presented (1996:212). Thus, (35) profiles, in Langacker's terms, the whole path, whereas (36), though still activating (in terms of nonlinguistic cognition) the path that the salt follows, profiles only the end result: too much salt in the food. Like Talmy, Langacker (1998) argues that sentences like (36) bring out a more subjective factor (we 'see' the path in our imagination), adding, insightfully, that this subjective factor is already "immanent in more objective conceptions" anyway (1998:87). Put differently, we also visually follow the objective path in cases like (35), but then literally rather than just in our imagination. As long as, in other words, one does not claim that MOVE is linguistically inherent to the caused-motion construction, the shifting between directional and locative prepositions in caused-motion sentences, is, after all, no problem. There is, in short, no need to resort to coercion. Another problem that advocates of compositionality cannot really solve according to Goldberg, concerns the grammatical status of oblique comple ments. Since this issue directly ties in, in my opinion, with how one views the notions 'basic meaning' and 'contextualized meaning', I will, in what follows, try to show that compositionality is no problem, given firstly my views of basic and contextualized senses, and secondly my own grammatical analysis of oblique complements and the consequences of this analysis for the interpre tation of sentences involving oblique complements. 5.
A compositional account In what follows I will first explain how I view what 'basic' word senses come down to, and what consequences this view has for actualized senses (section 5.1). I will furthermore argue, as I have done before in van der Leek (1996a, 1996b), that only NP complements function as verbal arguments. Oblique complements, in my view, are not arguments but function as subpredicates to the verb, which means that verb and subpredicate together form a complex predicate that may (due to its combined meaning) be transitive, even if the verb alone is intransitive. My arguments for this position will be given in section 5.2. Finally, in section 5.3 I will briefly touch upon the notion 'compositionality', which I interpret in a much broader sense than the way Frege, who is generally claimed to be the originator of the compositionality notion, interprets it. 5.1
Platonic concepts and senses in context In what follows, I intend to propose a(n experientially grounded) view of word meaning that can avoid lexical polysemy, while at the same time guaranteeing that words can semantically adapt to contextual needs. For con-
318
FREDERIKE VAN DER LEEK
venience, I will restrict myself to verbs. My assumption is, that underlying any verb there is a unified maximally schematic concept. Cognition, as I see it, conceptually constructs, for any individual usage of the verb, what process is going on in the world. From all these diverse conceptual constructs, it abstracts a unified concept; I will use the term 'Platonic concept' for this notion. 'Platonic' in this context is not to be seen in Plato's original metaphysical sense, but in the sense that any realization of this (mental) Platonic concept is by definition more concrete, hence no longer Platonic. (The term 'Platonic concept' is, by the way, due to Hofstadter & McGraw (1995:412), who apply it to concepts underlying letters rather than words.) What I have in mind, in other words, is one truly abstract but unified Platonic concept underlying (or symbolized by) a verb that characterizes what is truly unpredictable about the verb's meaning; any actualization of the Platonic concept underlying a verb is to be seen as a 'sense-in-context'.14 This sense in context is not the sense of the verb in its own right, but the sense that is symbolized through its association with other words (or phrases) that bring in their own meaning (which may already be contextualized themselves, and may include associated nonlinguistic frames or ICMs). To give a simple example, the sentences The boy swam to the other side of the pond and The fish swam to the other side of the pond both rely for their interpretation on a Platonic concept SWIM that represents a pro cess that can, in words, be fairly accurately described as 'self-propelling motion through water'. Clearly, however, the actual conceptualization of a fish swimming comes out differently than that of a human swimming. One could, in fact, at a stretch, imagine there to be languages that use two different words. It is precisely because English uses a single word that the language learner starts looking for the commonality in the two events. As Bowerman (1996:393) puts it: "hearing the same word repeatedly across differing events might draw childrens' attention to abstract properties shared by these events that might otherwise pass unnoticed". I'm told that the above view of word meaning is the 'monosemy' view: one Platonic concept out of which cognition constructs, depending on the local context, the appropiate and actual sense-in-context. Cognition, in brief, pulls out of memory the frame or ICM it has formed of how humans c.q. fish move self-propellingly through water and thus constructs two different senses-incontext for the verb including its different agentive companions. Most cognitive linguists argue differently, claiming that words are polysemous; they argue that a word has a prototypical sense, and that other senses get extended from this central sense. Thus the word swim as used in the above two sentences, symbolizes a prototypical (hence central) sense when the subject refers to a boy,15 and an extended sense when it refers to a fish.
CAUSED-MOTION AND THE 'BOTTOM-UP' ROLE OF GRAMMAR
319
My problem with this so-called polysemy approach is, that this view does not give due recognition to the role of cognition (cf. the way I illustrated this above for the coming about of the two senses-in-context for a boy swimming and a fish swimming). One of the main tenets of Cognitive Linguistics is that language underdetermines meaning (there is a lot of 'backstage cognition' constructing meaning in context, and "[l]anguage, as we use it, is but the tip of the iceberg of cognitive construction", Fauconnier 1994: xxii). The view that the word swim has, in itself, two distinct linguistic senses, is, it seems to me, inconsistent with the above tenet. A further, practical problem can be illus trated with the help of the following example: The boy swam, underwater, to the other side of the pond. The prototypical way of swimming for a human, is with his/her head above water. That certainly is the mental picture that presents itself by default in connection with The boy swam to the other side of the pond. The presence of the word underwater is a prompt for the language user to rely on a different frame (that of swimming underwater): the boy executes the swimming in a different way. Should we now say that the word swim as such has yet a third distinct sense? Where does this end? It seems much more natural to say that the process of a boy swimming underwater is associated with a particular sense in context that is constructed out of, and represents, this combination as a whole. I will return to this issue in section 6, where I discuss the difference between the conventional existence of partic ular usages (which is, indeed, unpredictable) as opposed to the sense in context of such conventional usages (which is, I will argue, by and large predictable). My point, to summarize, is that verbs only symbolize one abstract Platonic concept (whose content is, by definition, unpredictable) and that any actualized sense is never a sense of the verb in and of itself but, again by defi nition, a sense in context that includes whatever other (non)linguistic conceptu al elements helped create it; a sense, in other words, that gets compositionally constructed by cognition, and is predictable, given sufficiently tight contextual constraints (contextual in its broadest sense).16 Given the above exposition, it is now possible to explain my views about the grammatical status of oblique complements. 5.2
Oblique complements as subpredicates rather than arguments In this section I will, as already announced, argue that oblique comple ments of a verb do not have argument status, but form a subpredicate of the verb. Together they form a complex predicate, with the verb itself the head; this means, in Langacker's (1987:235-236) terminology, that the verbal pre dicate is the "profile determinant" of the complex predicate. Cognition builds a sense in context for the complex predicate which, therefore, can come to
320
FREDERIKE VAN DER LEEK
symbolize a transitive situation (if and when appropriate). I will illustrate what I have in mind with the help of example (3) of Mandelblit & Fauconnier (this volume), here numbered (15) and repeated below for convenience: (15) She trotted the horse into the stable In section 4.21 went to some lengths to try and explain what, to my way of thinking, the meaning is that gets profiled in sentence (15). What I tried to get across, partly by comparing (15) with its make variant and its intransitive counterpart, is that the agent argument's action can only be properly understood in terms of the complex transitive predicate symbolized by '(NP) trot (NP) into the stable'. This complex predicate has a sense in context that I tried to describe in terms of the two arguments involved (tried to, mind you: the conceptual content symbolized by and underlying language itself always does a much better job than any linguistic paraphrases employed). The agent argument's action, I said, comes down to acting out a directive type of leg pressure and rein usage on the theme argument. The latter's role is automatic, controlled as it is by the agent's actions; it moves, as I said in section 4.2, "how and where it is 'told' to move by the agent argument". To summarize, the 'complex predicate' claim is meant to ensure that the agent argument role is explicitly linked to not only the concept underlying the lexical verb, but also to that underlying the oblique complement (in the present context the goal complement), and that its relation with the theme argument consists in this complex predicate. After all, it is, as I already pointed out in connection with prototypical caused-motion sentences (13), (15) and (18), the agent argument's action as a whole (transferring energy onto the theme argu ment in the right direction) that guarantees what occurs to the theme argu ment.17 Notice that this way of viewing meaning is quite different from the logicbased objective presentation of trot as taking (in this type of example) three arguments: (i) the agent who somehow acts out (part of) what the verb symbolizes, (ii) the theme that is somehow involved in this action and (iii) the goal that expresses where the agent's action somehow leads the theme. Note my usage, thrice, of somehow: this logic-oriented approach, seeing trot as a ternary predicate, leaves out the 'glue' that is then added on artificially in terms of 'cause to move'. In my way of seeing things (which is, I suspect, very close to Langacker's),18 it is cognition that takes care of the glueing, thereby constructing the complex transitive predicate encoded as 'NP trot NP into the stable'; however, the glueing job is not done in terms of linguistic predicates such as CAUSE and MOVE, but in terms of (i) the grammatical
CAUSED-MOTION AND THE 'BOTTOM-UP' ROLE OF GRAMMAR
321
construction (as analyzed above), (ii) the lexical items' underlying (Platonic) concepts, and (iii) the relevant frame-semantic information brought in by the lexical items that 'fill' the construction. Understanding, finally, that (15) means what I described it as meaning, automatically explains the conceptual entail ment that is linguistically describable as 'cause to move'. In summary, I have argued, with Goldberg (1995) that — to stick to the verb under discussion — trot, apart from evoking rich frame-semantic in formation, encodes one basic meaning (a Platonic concept, in my terminology); I further argued, however, that, by turning up in a number of different complement constructions (including, additionally, different lexical material plus their frame semantic associations), it gets, at the conceptual level, to be part of many different senses in context.19 The above is, in a sense, my attempt to show that a compositional ac count of the caused-motion construction is, pace Goldberg (1995), feasible. In the following section I will briefly comment on the notion 'compositionality' as such. 5.3
Compositionality: a narrow view and a broad view In the above, I have used the term 'compositionality' and its derivatives without defining it in any way. Given the narrow sense allotted to the notion by logicians since Frege, some clarification is in order. Cognitive linguist Seana Coulson, who argues against compositionality in her Ph.D. dissertation, does so with the narrow Fregean definition in mind. She defines this as "[a] language is compositional if the meaning of a complex expression is system atically related to the meanings of its constituents" (Coulson 1997:3). She argues against this view of compositionality because it presupposes that only linguistic meaning is involved, and that lexical items have invariant meanings across different contexts. Given that actual language use amply shows that words have to be associated with multiple meanings, and that language usage is what Cognitive Linguistics aims to describe, compositionality in the above narrow sense cannot be maintained within this framework. However, one can also take a much broader view of compositionality if one simply drops the assumption that only linguistic cognition is involved in the understanding of language. Language understanding, be it viewed from the productive or the interpretive side, is based on systematically building (and rebuilding) of meaning out of all the information one can lay one's hands on (to use a rather inappropriate metaphor). Information is, to be a little more precise, partly prompted by language (in the form of (a) Platonic senses and (b) familiar or newly created frames), partly by the actual discourse situation and one's knowledge of the people participating in it, partly by what the
322
FREDERIKE VAN DER LEEK
discourse participants are jointly paying attention to, partly by one's general, social, and cultural knowledge of the world, and whatever else comes into it. A Frege-type narrow view of compositionality is, of course, intended to reduce the meaning of sentences to the (invariant) meanings of words and their grammatical arrangement. The broad view of compositionality that I am advocating here, is intended to emphasize the systematically and dynamically constructive aspect of our understanding of language in its context of use; communication can, and does, break down from time to time, but without compositionality in the broad sense, it would simply not be possible, it seems to me (however complex it all may be). At the end of section 5.1, I summarized my monosemy viewpoint, claiming that senses in context are, say, stochastically, predictable. This way of looking at things does, of course, go against the grain of what, as far as I understand, most cognitive linguists believe. I will therefore have to go into this matter, and will do so in the following section. 6.
Predictable or not, that is the question The view upheld by the majority of the cognitive linguistic community is (i) that words are polysemous, with a central (essentially prototypical) sense and others extending from the central sense or from another extended sense, and (ii) that extended senses are not predictable, only motivated. In section 5.1,I have expounded my monosemy view as an alternative for polysemy. My senses in context take over the role of extended senses in the polysemy account or, better, any sense in context is an extension (or more concrete conceptual ization of) the underlying Platonic concept.20 Supporters of the polysemy view claim, in the words of Lakoff (1987:65) that "the noncentral members [of a conceptual category] are not predictable from the central members, [but that] they are 'motivated' by it, in the sense that they bear family resemblances to it. " I agree that, to give a random exam ple, there was no predicting that, prior to its actual introduction into English, the combination of the word over with the expression turn the page would ever become part of the conventional language (in Dutch we do not use over here, for instance, but om "around", as in Hij sloeg de bladzijde om, literally "He struck the page around"). I do not agree, however, that the sense in context (and over's part in it) that evolves in the context of turn (the page) over is unpredictable (as Lakoff and most other cognitive linguists seem to claim). In fact, it cannot be: the first person to ever introduce this combination must, after all, have been able to 'recognize' its meaning. My point is, in other words: the introduction into, and, subsequently, its existence in the language of a particular usage is not predictable, but its meaning is (by and large)
CAUSED-MOTION AND THE 'BOTTOM-UP' ROLE OF GRAMMAR
323
predictable. (Hofstadter & McGraw 1995:442, in a different context, coin the felicitous phrase "The Predictable Unpredictability of the Creative Process". By this I understand them to mean that discovering a novel sense under pres sure of some structure that happens to be cognitively available, is a happen stance, hence unpredictable happening, but that this creative act of one's cognition strikes one as predictable in hindsight because the sense (in context) is so obviously 'there' once cognition has 'uncovered' it.) There is, at the same time, such a phenomenon as convention in lan guage, very pervasively so. Now it seems to me that the patterns Goldberg (1995) sees as constructional senses, are really general conceptual patterns shared by particular lexico-grammatical configurations, patterns that are unconsciously extracted from actual usage and stored as such by the language user's cognition. Thus, to limit myself to the caused-motion construction pro per, what all its instances have in common is, indeed, nonlinguistic conceptual caused-motion. The general cognitive linguistic belief is that this conceptual pattern feeds prototypical caused-motion sentences. My claim — to be sub stantiated in section 7 — is different: in the language of mature speakers/ hearers, this pattern may inspire, or serve as a model for, novel usages; put differently, it may license novel usages, without, however, lending them its meaning. It is now, clearly, time to turn to the final section and do two things: (i) go into the issue raised in section 1, i.e. my suggestion that bare complement constructions may, prior to conceptual structure, play the role of alerting the language user to the possibility of novel ways of giving conceptual content to particular complement constructions, and (ii) substantiate my claim that such patterns as the caused-motion one, have no more than a licensing function in the language of mature speakers (even though, originally, such patterns were what children took their first linguistic meanings from). 7.
Concepts are blind without percepts, percepts are vague without concepts The above title is the almost literal rendering of a statement made in the concluding section of Chalmers et al. (1995:192-193), who essentially attribute the insight to Kant. The authors argue emphatically that all mental representa tions are the "fruits" of perception, and that, therefore, high-level perceptual processes play as crucial a role in cognition as do conceptual processes; cognition, as they put it, "cannot succeed without processes that build up appropriate representations" (Chalmers et al. 1995:193). In a sense, what they are saying is that high-level perception and conceptualization involve two-way traffic: high-level perceptual processes are formative relative to conceptual
324
FREDERIKE VAN DER LEEK
processes (bottom-up) and conceptual processes influence high-level perception (top-down), with the two systems constantly feeding both on and into one another, through 'loops' (or through what Edelman 1989, 1992 speaks of, if I understand him correctly, as 're-entrant mapping'). It is because of this mutual feeding process that intentionality is guaranteed (intentionality being, of course, the 'aboutness' of meaning, i.e. its connectedness with the world including all that humans create in it, physically and mentally).21 A problem, from the above point of view, is that the cognitive linguistics 'story', as sketched in section 1, presents language meaning as involving one way traffic only (so-called 'conceptual semantics'). What the story, applied to the purely grammatical structure of the caused-motion construction, comes down to, is that the grammatical structure in question derives its (prototypical, or central) semantic structure from a caused-motion model, which prelinguistic cognition builds, schematically, out of prior perceptual images (see section 1). Out of its prototypical sense, the caused-motion construction then develops a number of extended constructional senses (Goldberg) or one maximally general semantic structure that is so abstract that it covers all the existing conventional uses of the construction (Langacker). However, there is no mention of either type of constructional sense(s) re-entrantly mapping back into perceptual categorization again: there is, so to speak, no touching base with, ultimately, the outside world. The theory does not, put differently, appear to account for intentionality. There is, I suggest, an alternative way of looking at things. This alternative story runs (hugely oversimplified) as follows. Children begin by trying "a fairly direct mapping of language they hear onto their already-formed [prelinguistic] conceptualizations" (Mandler 1996:377). For instance, preverbal children tend to master caused-motion conceptually, so they may begin by trying to map this model directly onto expressions like Jane roll ball (to) Mummy, thinking that the complement configuration they pick up on, itself symbolizes caused-motion. This gets confirmed by the fact that their usage often exhibits underextension (cf. Slobin 1985 on early uses of transitivity markings for only prototypical cases involving animate agents and inanimate object cases). However, the more they hear the complement configuration in question used in ways that disallow such direct mappings — and this is where the two stories split company — the more they may feel the need to 'bare' the complement configuration as such of the semantic content they formerly associated it with, and finally come to realize the non-lexicalized complement configuration's real function: to provide structure to encode conceptualization (after all: without morphemes, no concepts are introduced to activate conceptualizations).
CAUSED-MOTION AND THE 'BOTTOM-UP' ROLE OF GRAMMAR
325
Langacker's (1995:20) characterization of the maximally schematic semantic structure of a transitive, or subject-direct object, configuration, comes down to its consisting in an asymmetric temporal relation between a primary figure (the CLAUSE-level trajector, grammatically the subject) and a secondary figure (the CLAUSE-level landmark, grammatically the direct object). Remem ber that I analysed the caused-motion complement configuration as one that is of essentially the same transitive type, be it that the predicate in question is complex rather than simple. Thus, applying Langacker's characterization of a transitive configuration, we only need to add that this involves a complex predicate that is made up of (i) a temporal relation component functioning as the complex predicate's profile determinant (grammatically a verb) and (ii) a subpredicate component that consists in an atemporal relation (grammatically a PP or AD VP). Note two things about the above characterization. First of all, the notions used are all definable in terms of perceptual processing: asymme try, primary/secondary figure, (a)temporal relation, component; Langacker does, indeed, define all these notions in processual terms that make no appeal to concepts (cf. Langacker 1987, chs 5-7).22 Secondly, the characterization is flexible in two respects. On the one hand, the subpredicate may apply to the primary figure (grammatically the subject, with the subpredicate functioning as subject complement) or else it may apply to the secondary figure (gram matically the direct object, with the subpredicate functioning as direct object complement); given my observation in section 3.2 that Goldberg's categories I-V involve, at the linguistic level, the latter type, and category VI the former, these options are, in my analysis, both available prior to lexicalization. Which of the two is opted for, depends, clearly, on the choice of verb (cf. put as opposed to follow). On the other hand, the bare complement configuration does not, as such, involve a path notion: this only comes in when the PP/ADVP is lexicalized in terms of some preposition or adverb that symbolizes direction ality. My conclusion is that the bare complement configuration under dis cussion has no semantic import prior to lexicalization, while it does have a formative effect in that it constrains the possible encoding of semantic content to options such as those mentioned above. At the same time, and that is the constructive aspect of my story, bare complement configurations can also, bottom-up, make the language user aware of potentially novel applications. Given the language user's frequent experi ence of usages of the complement configuration at issue here, with an oblique goal complement, and given that these usages, apart from all their semantic differences, all share, at a superordinate level of conceptual categorization, the prototypical caused-motion pattern, then this language user's cognition au tomatically recognizes and extracts this pattern as one conventional option (or
326
FREDERIKE VAN DER LEEK
licensing pattern) for the complement configuration in question provided the oblique complement encodes a goal. This in turn tells our language user (if he/she is the creative type), that the complement configuration (including its goal complement) may be headed by a novel verbal predicate provided this can create — in combination with the goal complement — a complex transitive predicate expressing a sense in context that is an enriched version of the prototypical caused-motion pattern. What this effectively comes down to is, that it must be possible to associate the verbal predicate with a force of whatever kind that, transferred by the agent argument onto the theme argu ment, guarantees the latter following the path expressed by the goal subpre dicate. The following examples (nonce ones, I assume), nicely illustrate the imaginative powers needed on the part of the authors to 'discover' the neces sary force referred to above. Relative to sentences (37)-(39), I imagine the forces in question to have been attributed to religious faith, passion and such miracles as hope may achieve, respectively (I italicize the relevant verbs for convenience):23 (37) "Manuelito!" he cried, "for this darling mule I could almost pray you into Heaven! " [Willa Cather, Death comes for the Archbishop, 1927] (38) "I know you impulsive young fellows. You don't believe in polite preliminaries when you meet a girl who takes your fancy. You make a swan dive for her. You grab her. You start to kiss the stuffing out of her. And then what? She very properly scratches you to the bone, and serve you right." [P.G. Wodehouse, Company for Henry, 1967] (39) "My dear Mr Emerson," I said aloud. "The Robin bearing this letter is a former student. Please hope him to death, and keep him running. Your most humble and obedient servant, A.H. Bledsoe... " [Ralph Ellison, Invisible Man, 1952] My point is, in other words, that the bare complement configuration can, always at the level of Slobin's thinking for speaking, alert the speaker bottomup to its potential usefulness of structuring the situation that is to be linguis tically encoded, provided the right semantic concepts are (or can be made) available top-down. To return again, finally, to Dennett's quotation this paper started out with: maybe grammatical structure can, if my story points in the right direc-
CAUSED-MOTION AND THE 'BOTTOM-UP' ROLE OF GRAMMAR
327
tion, "lay down some of the tracks on which 'thoughts' can then travel" prior to their conceptualization, rather than just 'finessing' them. Notes 1 In what follows, the term 'grammar' and its derivatives is, unless otherwise stated, primarily used with reference to syntax. 2 Chomsky has not, to my knowledge, clearly articulated his own position in this respect. 3 As Lakoff (1987:68) points out, ICMs are of four types: Fillmorian frames (propositional in structure), Lang acker's image-schémas, and Lakoff & Johnson's (1980) metaphorical and metonymie mappings. I will use the term 'frame' for the first type, and otherwise speak of (types of) ICMs. 4 To avoid cumbersome terminology, I will, unless otherwise indicated, use the term 'perception' and its derivatives in a broad sense (covering not only the five regular senses but also proprioception and kinesthesis). 5 The innateness hypothesis that is inherent in this view is, it seems to me, an inescapable consequence of the modular view of language: acquiring language without the help of other cognitive faculties is mind-bogglingly difficult unless there is help from innate knowledge. 6 "Thinking for speaking", as Slobin (1996:76) puts it, "involves picking those characteristics of objects and events that (a) fit some conceptualization of the event, and (b) are readily encodable in the language". 7 In this paper I will consider this hypothesis only with regard to grammatical structures that are 'pure' in the sense that they are in no way phonologically realized. 8 Apart from their maximally schematic semantic structure, certain complement constructions also have, in Langacker's view, prototypical semantic values reflecting archetypical prelinguistic ICMs (cf. Langacker 1991:282ff). 9 Goldberg (1995) presents her proposal within the framework of Construction Grammar (cf. Fillmore & Kay, to appear). In this framework, a form-meaning pair is, very roughly, a construction, provided some aspect of it is not predictable from the parts it is built out of, or from any other construction. Goldberg argues that the senses she proposes for complement structures are nonpredictable, hence constructional senses. 10 Adopting Fillmore's frame semantic approach to lexical meaning (e.g. Fillmore 1982, 1985), Goldberg argues that a great deal of the meaning a verb has in context will be contributed by frames that get evoked by lexical material. For both Cindy blew hard and Cindy blew the dust ojf the book this means, for instance, that, given the frame semantic knowledge that Cindy is likely to be a person, the blowing involves 'forcefully creating an air current from the mouth'. The 'from the mouth' part need not, then, be assumed to be part of the basic sense of blow. 11 Only in one case is there no need to know the constructional sense underlying a basic simple sentence in order to understand it, namely if the verb itself is an instance of the constructional sense (thus give is an instance of the central constructional sense of the ditransitive construction). For gesture this is definitely not the case. 12 Mandelblit & Fauconnier (this volume) argue differently with respect to the meaning of sentence (15), numbered (3) in their paper. They argue that the English caused-motion construction underspecifies who/what acts out the main predicate (in Sam sneezed the napkin ojf the table, this is the agent argument, but in She trotted the horse into the stable, this is the
328
FREDERIKE VAN DER LEEK
theme argument). They propose that such differences can be accounted for in terms of blending (see their paper for details). They are, in fact, treating the English caused-motion construction on a par with the French faire construction (which, of course, makes the causative predicate explicit, perhaps comparable to the English make construction). In my view, the blending analysis of the English caused-motion construction misses out on the subtle distinctions I have been trying to point out in the above. I am, of course, no judge as to the authors' blending proposals in general, since my knowledge of the languages they are mainly concerned with in this paper, is insufficient (French) or even nihil (Hebrew). 13 This, frankly, sounds like a circular argumentation to me. 14 The term 'sense-in-context' is due to Pustejovsky (1995). 15 Humans are, as they say, anthropocentric. 16 This view does not necessarily conflict with Langacker's view of polysemy, which, as he emphasizes, involves decontextualization and is defined in terms of entrenchment. The only true difference is that Langacker starts from some (concrete) central sense. More on this in Beliën and van der Leek (forthcoming). 17 Note that, by the same argumentation, the complex predicate involved in nonprototypical cases like (21), Sam invited Cindy into the room, the agent argument's action merely guarantees that Cindy is the recipient of the invitation as characterized by the goal complement. 18 The question that I have avoided so far, is whether senses in context are to be seen as linguistic senses. Langacker's approach seems to suggest they are. I find this a truly moot question, which will be paid ample attention to in Beliën and van der Leek (forthcoming). 19 This view, which I developed independently but recognize as very close to the one expounded by Pustejovsky (1995), nevertheless differs from it in essential ways. Pustejovsky proposes that all potential senses in context are built out of the linguistic senses of lexical items. In the view I have tried to express in the above, I see nonlinguistic cognition as playing to a very large extent the role that linguistic cognition plays in Pustejovsky's framework. 20 In Beliën and van der Leek (forthcoming) we go into the notion of prototypicality and centrality of senses. 21 As Edelman (1992:110) puts it, "because concept formation is based on the central triad of perceptual categorization, memory, and learning, it is, by its very nature, intentional". 22 Langacker defines temporal and atemporal relations in terms of sequential and summary scanning processes respectively, not in terms of the concept TIME. 23 Notice that 'discovering' such a force is necessary even in Goldberg's framework: the caused-motion constructional sense is, when all is said and done, no help at all for producing/understanding the meaning of caused-motion sentences.
References Beliën, M. & F. Van der Leek. Forthcoming. "Polysemy versus Monosemy" (provisional title). MS, University of Amsterdam. Bowerman, M. 1996. "Learning How to Structure Space for Language: A Crosslinguistic Perspective". Language and Space eds. P. Bloom, Peterson, M., Nadel. L. & M. Garrett, 385-436. Cambridge, Mass.: MIT Press.
CAUSED-MOTION AND THE 'BOTTOM-UP' ROLE OF GRAMMAR
329
Chalmers, D., French, R. & D. Hofstadter. 1995. "High-level Perception, Representation, and Analogy". Fluid Concepts and Creative Analogies ed. D. Hofstadter, 160-193. New York: Basic Books. Chomsky, N. 1966. Cartesian Linguistics. New York: Harpers & Row. Coulson, S. 1997. Semantic Leaps: The role offrame-shifting and conceptual blending in meaning construction. Ph.D. dissertation. University of Calif., San Diego. Croft, W. 1991. Semantic Categories and Grammatical Relations. Chicago: University of Chicago Press. Dennett, D. 1991. Consciousness Explained. London: Penguin Books. Edelman, G. 1989. The Remembered Present. A Biological Theory of Consciousness. New York: Basic Books. . 1992. Bright Air, Brilliant Fire. On the Matter of the Mind. London: Penguin Books. Fauconnier, G. 1994. Mental Spaces. Cambridge: Cambridge University Press. Fillmore, C. 1982. "Frame Semantics". Linguistics in the Morning Calm ed. The Linguistic Society of Korea, 111-137. Seoul: Hanshin Publishing Co. Fillmore, C. 1985. "Frames and the Semantics of Understanding". Quaderni di Semántica 6:2.222-253. & P. Kay. To appear. Construction Grammar. Stanford: Center for the Study of Language and Information. Fodor, J. 1975. The Language of Thought. Cambridge, Mass.: Harvard University Press. Goldberg, A. 1995. A Construction Grammar Approach to Argument Structure. Chicago: The University of Chicago Press. Hofstadter, D. & G. McGraw. 1995. "Letter Spirit: Esthetic Perception and Creative Play in the Rich Microcosm of the Roman Alphabet". Fluid Concepts and Creative Analogies ed. D. Hofstadter, 407-466. New York: Basic Books. Jackendoff, R. 1990. Semantic Structures. Cambridge, Mass.: The MIT Press. Lakoff, G. 1987. Women, Fire, and Dangerous Things. Chicago: University of Chicago Press. Lakoff, G. & M. Johnson. 1980. Metaphors We Live By. Chicago: University of Chicago Press. Langacker, R. 1987. Foundations of Cognitive Grammar. Vol. I. Theoretical Perspectives. Stanford: Stanford University Press.
330
FREDERIKE VAN DER LEEK
1991. Foundations of Cognitive Grammar. Vol. 2. Descriptive Application. Stanford: Stanford University Press. 1995. "Raising and Transparency". Language 71:1.1-62. 1998. "On Subjectification and Grammaticization". Discourse and Cognition ed. J.P. Koenig, 71-89. Stanford: Center for the Study of Language and Information. Mandelblit, N. & G. Fauconnier. 2000. "How I got myself arrested. Underspecificity in Grammatical Blends as a Source for Constructional Ambiguity". This volume eds. A. Fooien & F. van der Leek, 167-189. Amsterdam: John Benjamins. Mandier, J. 1996. "Preverbal Representation and Language". Language and Space eds. P. Bloom, Peterson, M., Nadel. L. & ML Garrett, 365-384. Cambridge, Mass.: MIT Press. 1998. "Representation". Cognition, Perception, and Language tas. D. Kuhn & R. Siegler, 254-308. Vol. 2 of Handbook of Child Psychology ed. W. Damon. New York: Wiley. Pinker, S. 1989. Learnability and Cognition. The Acquisition of Argument Structure. Cambridge, Mass.: MIT Press. Procter, P. ed. 1978. Longman Dictionary of Contemporary English. London: Longman. Pustejovsky, J. 1995. The Generative Lexicon. Cambridge, Mass.: MIT Press. Slobin, D. 1985. "Crosslinguistic evidence for the language-making capacity". The crosslinguistic study of language acquisition. Vol. 2. Theoretical issues ed. D. Slobin, 1157-1256. Hillsdale, N.J.: Erlbaum. 1996. "From 'thought and language' to 'thinking for speaking' ". Rethinking linguistic relativity eds. J. Gumperz & S. Levinson, 70-96. Cambridge: Cambridge University Press. Talmy, L. 1985. "Lexicalization Patterns: Semantic Structure in Lexical Forms". Language Typology and Syntactic Description. Vol. 3. Grammatical Categories and the Lexicon ed. T. Shopen, 56-149. Cambridge: Cambridge University Press. 1988. "Force Dynamics in Language and Cognition". Cognitive Science 12:1.49-100. 1996. "Fictive Motion in Language and 'Ception' ". Language and Space eds. P. Bloom, Peterson, M., Nadel. L. & M. Garrett, 211-276. Cambridge, Mass.: MIT Press. Van der Leek, F. 1996a. "Rigid Syntax and Flexible Meaning: The Case of the English Ditransitive". Conceptual Structure, Discourse and Language
CAUSED-MOTION AND THE 'BOTTOM-UP' ROLE OF GRAMMAR
331
ed. A. Goldberg, 321-332. Stanford: Center for the Study of Language and Information. — 1996b. "The English Conative Construction: A Compositional Account". Chicago Linguistic Society 32.363-378.
Addresses
Athanasiadou, Angeliki School of English Aristotle University 540 06 Thessaloniki GREECE e-mail:
[email protected] Dancygier, Barbara Linguistics Department Simon Fraser University Burnaby, B.C. V5A 1SG CANADA e-mail:
[email protected] Dewell, Robert Loyola University Box 190 New Orleans, LA 70118 U.S.A. e-mail:
[email protected] Dirven, René Beekstraat 39 B-2800 Mechelen BELGIUM e-mail:
[email protected] Fauconnier, Gilles Department of Cognitive Science 0515 University of California, San Diego LaJolla, CA 92093-0515 U.S.A. e-mail : fauconnier@cogsci .ucsd. edu
Fooien, Ad Department of General Linguistics and Dialectology University of Nijmegen P.O. Box 9103 6500 HD Nijmegen THE NETHERLANDS e-mail:
[email protected] Fujita, Ikuyo International University of Health and Welfare 3-4 12-207 Asaka Ohtawara-shi, Tochigi 324 JAPAN Griffiths, Patrick Department of Literature and Language University of the South Pacific Box 1168 Suva FIJI ISLANDS e-mail:
[email protected] Hampe, Beate Institute of English and American Studies Department of Linguistics Friedrich-Schiller Universität Ernst-Abbe-Platz 4 D-07743 Jena GERMANY e-mail : beate. hampe@rzuni-j ena. de
334 Heyvaert, Liesbet Department of Linguistics Catholic University of Leuven Blijde Inkomststraat 21 B-3000 Leuven BELGIUM e-mail: liesbet. hey vaert® arts. kuleuven. ac. be
ADDRESSES
Nikiforidou, Kiki Department of English Studies University of Athens University Campus - Zografou Athens 15784 GREECE e-mail : nikiforidou@ath. forthnet. gr
Ihara, Hiroko Tokyo Zokei University 3-21-4-508 Fujimi-cho, Chofu-shi Tokyo 182-0033 JAPAN e-mail:
[email protected]
Takahashi, Hidemitsu Faculty of Letters Hokkaido University kitaku kita 10 nishi 7 Sapporo 060 JAPAN e-mail : htaka@letters. hokudai. ac .jp
Katis, Demetra Department of Preschool Education University of Athens Navarinou 13a Athens 10680 GREECE e-mail:
[email protected]
Thepkanjana, Kingkarn Department of Linguistics Faculty of Arts Chulalongkorn University Bangkok THAILAND e-mail:
[email protected]
Kochañska, Ágata English Institute Warsaw University ul. Nowy Swiat 4 00-497 Warszawa POLAND e-mail:
[email protected]
Tsuboi, Eijiro Language and Information Sciences Graduate School of Arts and Sciences University of Tokyo, Komaba 3-8-1 Komaba, Meguro-ku Tokyo 153-8902 JAPAN e-mail : e-tsuboi@boz. . u-tokyo .ac.jp
Mandelblit, Nili Department of Cognitive Science 0515 University of California, San Diego LaJolla, CA 92093-0515 U.S.A. e-mail:
[email protected] Mortelmans, Tanja UFSIA - D-128 Prinsstraat 13 B-2000 Antwerpen BELGIUM e-mail : tanj a. mortelmans@ufsia. ac.be
Van der Leek, Frederike Department of English University of Amsterdam Spuistraat 210 1012 VT Amsterdam THE NETHERLANDS e-mail:
[email protected]
Index A. accomplishment 262, 264, 269-272, 275 accusative object see object, accusative action chain 123, 132-135 action nominalization see nominalization, action activity 259, 261-262, 264-265, 267-273, 275 adversative construction see construction, adversative adversative passive see passive, adversa tive agent(ive) 36, 39, 47, 49, 50-51, 53-54, 76, 85-87, 117, 169, 172, 174, 176182, 184-186, 243-250, 252-255, 259265, 267-268, 270-279, 290-292, 295, 297, 305, 307-308, 311-314, 318, 320, 324, 326 agrammatism 123-125, 134-135 alienable possessive see possessive, alien able ambiguity constructional ambiguity 167, 171, 174-175, 178, 182, 184, 186 aspect verbal aspect 141, 162-163 attention distribution of attention 261 sequence of attention 47-48, 52-53, 55, 61, 64 windowing of attention 260, 261, 263, 270, 276-277 B. benefactive construction see construction, benefactive binyanim system 171, 174-175, 178-179 blend grammatical blend 167-168, 171, 178 optional transfer blend 180-186 proper blend 179-180
bounded(ness) see temporal (un)bounded(ness) canonical event model see model, canoni cal event case 27-33, 35-37, 39-41, 43 case marking 123, 135 case meaning 47 causal chain event-frame 260-261, 264265, 267, 269-270, 274-277 causative lexical causative 259-264, 275, 277279 syntactic causative 259-263 causative-passive construction see con struction, causative-passive caused-motion construction see construc tion, caused-motion cognitive model see model, cognitive complex predicate see predicate, complex compositionality 306, 316-317, 319, 321322 concept Platonic concept 317-319, 321-322 conceptual metaphor see metaphor, con ceptual conceptualization 35, 37, 39, 84-85, 91, 132-133, 142-144, 146-149, 153-154, 156-157, 160-161, 163, 182, 193, 200, 217, 219-220, 243, 252, 260, 262, 266-267 conditional conversational conditional 3-5, 23 counterfactual conditional 196-198, 206, 208, 210, 212, 224, 226 discourse conditional 4, 7, 10, 13-17, 22-23 identifying conditional 4-5, 7, 9 inferencing conditional 1, 5-7, 9, 11, 22-23
336
INDEX
logical conditional 3-5, 11 metacommunicative conditional 4, 7, 17-23 pragmatic conditional 1-8, 13, 15, 2224 speech act conditional 221, 226 construal 6, 27-28, 30-43, 47-50, 52-54, 56-59, 61-62, 85-86, 88, 92-93, 95, 98, 141, 148, 162, 177, 182, 191-192, 201, 217-220, 230-231, 283, 289-293, 297, 311-312, 314-315 construction 6, 8, 10-11, 13-14, 18, 22, 28, 31, 47, 49-57, 59-60, 62, 68-70, 72, 74, 77-78, 81, 83-85, 87, 91-94, 96, 98, 103-104, 106, 108, 110, 112, 116, 134, 137, 147-148, 159, 161163, 167-176, 178-180, 182, 186, 199, 217, 221, 231, 239-240, 249250, 259, 261, 263, 265, 283-287, 289, 291, 296, 302-308, 313, 316317, 319, 321, 323-324 adversative construction 286, 289 benefactive construction 285-287, 291 causative-passive construction 179 caused-motion construction 168-172, 176, 178-179, 182, 301-303, 306, 308-309, 311-313, 316-317, 320-321, 323-326 constructional ambiguity see ambiguity, constructional constructional polysemy see polysemy, constructional conversational conditional see conditional, conversational coordination 141, 147-148, 159-162 counterfactual conditional see conditional, counterfactual cultural model see model, cultural D. dative object see object, dative deixis/deictic 10, 86, 90-91, 114, 191, 203, 218, 224, 225, 228-231, 243-245 direct object see object, direct direction nouns see nouns, direction
discourse conditional see conditional, discourse distribution of attention see attention, distribution of E. epistemic epistemic modal verb 191, 193-194, 199-212 epistemic stance 195-200, 202, 204208, 212 event-structure metaphor see metaphor, event-structure experiencer 47-48, 174, 180, 185, 246248, 250, 252-253 expressivity 96-98 extended viewing frame see viewing frame, extended F. fictive motion see motion, fictive force exertion 240-242, 246, 252-253 G. gerundive nominalization see nominalization, gerundive goal 123-125, 127-133, 135 goal orientedness 123, 127-128, 131-133, 135 grammatical underspecification see underspecification, grammatical grammatical blend see blend, grammatical grammaticalization 193, 204, 209-211 grounding 103-106, 108-110, 113-118, 143, 191-194, 202-203, 209, 212, 230 H. heterogeneous process see process, hetero geneous homogeneous process see process, homo geneous I. identifying conditional see conditional, identifying
INDEX
imperative 239-242, 244-246, 248-255 imperfective 141, 147, 150-162 inalienable possessive see possessive, inalienable inchoative 265-268, 271 indicative 195-196, 201-207, 209-210 indirect object see object, indirect inferencing conditional see conditional, inferencing instantiation 103-106, 113-116, 118-120 instrumental 27-28, 31-35, 39-43 K. konjunktiv I 194, 199-200 konjunktiv II 191, 194, 196-198, 200, 204-206, 208-210, 212 L. lexical causative see causative, lexical lexical polysemy see polysemy, polysemy logical conditional see conditional, logical M. marked(ness) 83-84, 94-98 metacommunicative conditional see condi tional , metacommunicative metaphor 27-28, 41-42 conceptual metaphor 85-86, 89-90, 9293 event-structure metaphor 86-87, 89-93 model cognitive model 283, 288-292, 297 canonical event model 241-244 cultural model 294 momentary viewing frame see viewing frame, momentary monosemy 318, 322 mood 191, 194, 196, 201-202, 206 motion 32-37, 39-40, 42 fictive motion 39-40 N. nominalization gerundive nominalization 103-106, 108, 111-113, 115-119
337
action nominalization 103-106, 115119 nouns direction nouns 27-31, 34-39, 41, 43 O. object accusative object 33, 47-49, 52-54, 5657, 62 dative object 47-48, 57, 59-62 direct object 33, 47-48, 52-59, 62, 86, 88, 103, 112-115, 117-118, 124, 129-130, 133, 147, 163, 169, 179-181, 186, 263, 265, 272 indirect object 47-48, 114, 180-181 optional transfer blend see blend, optional transfer P. passive 239, 247-252, 254-255 adversative passive 283, 285-297 patient 47, 54-55, 117, 172, 176, 180, 182, 184, 243, 246-250, 252, 254255, 259, 261-263, 265, 268, 270, 272, 274-279 percept 302, 323-325 perfective 141, 146-150, 153-157, 159162 phrasal verb see verb, phrasal Platonic concept see concept, Platonic polysemy lexical polysemy 302-304, 306, 309310, 317, 319, 322 constructional polysemy 302-303, 309 possessive alienable possessive 67-68, 72, 78-79 inalienable possessive 67-68, 72, 7879 pragmatic conditional see conditional, pragmatic predicate complex predicate 317, 319-320, 325326 prepositions 27-32, 34-43 two-way prepositions 27, 31, 35
338
process heterogeneous process 143-146, 152153, 156, 162-163 homogeneous process 143-147, 152153, 156-158, 160-163 profiling 142-162 proper blend see blend, proper prototype 75, 239, 244, 246-255, 262 prototypical(ly) 47, 57-58, 61, 75, 89, 115-116, 133-135, 155, 168, 241, 244-246, 248-255, 259-260, 262-263, 267, 274-275, 277, 291-292, 295, 297 prototypicality 68, 76, 78, 134-135, 245, 247, 249-250, 252 S. schema 85, 92-93, 96, 98, 172, 174, 178, 180, 184, 246, 250, 254 schematic(al) 47-49, 52, 60-61, 95, 104, 146-147, 169- 170, 241, 245 schematicity 110 schematic meanings 47-48, 52, 60 semantic role 124, 131, 135 sense in context 318-320, 322, 326 sequence of attention see attention, sequence of source 47, 49-60, 123-125, 127-133, 135 space 27-30, 33-34, 40-42 speaker involvement 217, 219, 222-230, 232 speech act conditional see conditional, speech act Stative 265-269, 271, 279 subject 9-10, 22, 30, 32-33, 36, 38-40, 47-48, 52-56, 59-62, 68, 87, 104, 106, 113-114, 117-119, 123-125, 129131, 133-135, 147, 163, 176, 180181, 186, 192, 200, 240, 243, 246253, 273, 284-285, 288-294, 296-297
INDEX
subjectification 192-193, 211, 218-220, 224, 228, 231-232 subjectivity 191, 217-219, 225, 231-232 subjunctive 195-200, 204-208, 212 suppletive form 259-260, 277-279 syntactic causative see causative, syntactic T. temporal (un)bounded(ness) 143-147, 152153, 156-159, 161-163 text type 94, 96, 98 time 32-33, 41-42 two-way prepositions see prepositions, two-way type specification 103-108, 110, 112-116, 118-120 U. unbounded(ness) see temporal (un)bounded(ness) underspecification grammatical underspecification 167168, 171, 174, 178-180, 184, 186 V. verb phrasal verb 81-85, 88-98 verbal aspect see aspect, verbal viewing frame momentary viewing frame 148-153, 159-161 extended viewing frame 153-154, 156, 158-160 W. windowing of attention see attention, windowing of
