Subjectification: Various Paths to Subjectivity 9783110892970, 9783110185300

Citation preview

Subjectification

W G DE

Cognitive Linguistics Research 31

Editors

Dirk Geeraerts Rene Dirven John R. Taylor Honorary editor

Ronald W. Langacker

Mouton de Gruyter Berlin · New York

Subjectification Various Paths to Subjectivity Edited by Angeliki Athanasiadou Costas Canakis Bert Cornillie

Mouton de Gruyter Berlin · New York

Mouton de Gruyter (formerly Mouton, The Hague) is a Division of Walter de Gruyter G m b H & Co. KG, Berlin

© Printed on acid-free paper which falls within the guidelines of the A N S I to ensure permanence and durability.

Library of Congress Cataloging-in-Publication

Data

Subjectification : various paths to subjectivity / edited by Angeliki Athanasiadou, Costas Canakis, Bert Cornillie. p. cm. — (Cognitive linguistics research ; 31) Includes bibliographical references and index. ISBN-13: 978-3-11-018530-0 (hardcover : alk. paper) ISBN-10: 3-11-018530-X (hardcover : alk. paper) 1. Subjectivity (Linguistics) I. Athanasiadou, Angeliki. II. Canakis, Costas, 1 9 6 8 III. Cornillie, Bert, 1 9 7 5 IV. Series. P299.S89S83 2006 40Γ.43—dc22 2006006511

Bibliographic

information published by the Deutsche

Nationalbibliothek

The Deutsche Nationalbibliothek lists this publication in the Deutsche Nationalbibliografie; detailed bibliographic data are available in the Internet at http://dnb.d-nb.de.

ISBN-13: 978-3-11-018530-0 ISBN-10: 3-11-018530-X ISSN 1861-4132 © Copyright 2006 by Walter de Gruyter G m b H & Co. KG, D-10785 Berlin All rights reserved, including those of translation into foreign languages. No part of this book may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopy, recording, or any information storage and retrieval system, without permission in writing from the publisher. Printed in Germany

Table of contents

List of contributors Introduction Angeliki Athanasiadou, Costas Canakis, and Bert Cornillie

1

Section I: Large theoretical issues Subjectification, grammaticization, and conceptual archetypes Ronald W. Langacker

17

Logic, subjectivity, and the semantics/pragmatics distinction Frank Brisard

41

Toward a typology of linguistic subjectivity: A cognitive and cross-linguistic approach to grammaticalized deixis Satoshi Uehara

75

Section II: Case studies I - Modals and modality Subjectification in (expressions of) epistemic modality and the development of the grounding predication Peter Pelyvds

121

Langacker's 'subjectification' and 'grounding': A more gradual view Tanja Mortelmans

151

Conceptual and constructional considerations on the subjectivity of English and Spanish modals Bert Cornillie

177

vi

Table of contents

Section III: Case studies II - Adjectives Adjectives and subjectivity Angeliki Athanasiadou

209

Grammaticalization and subjectification of the English adjectives of general comparison Tine Breban

241

Subjectification in gradable adjectives Henk Pander Maat

279

Section IV: Syntax and semantics On subjectivity and 'long distance pj^-movement' Arie Verhagen

323

Subjective construal and factual interpretation in sentential complements Kiki Nikiforidou

347

Zero in syntax, ten in pragmatics: Subjectification as syntactic cancellation Conception Company Company

375

Author index Subject index

399 405

List of contributors

Angeliki Athanasiadou School of English Aristotle University of Thessaloniki Greece e-mail: [email protected]

Tine Breban Department of Linguistics University of Leuven Belgium e-mail: [email protected]

Frank Brisard Department of Linguistics University of Antwerp Belgium

e-mail: [email protected]

Costas Canakis Department of Social Anthropology and History University of the Aegean Mytilene Greece e-mail: [email protected]

viii

List of contributors

Conception Company Company Institut» de Estudios Filologicos Universidad Nacional Autonoma de Mexico Mexico e-mail: [email protected]

Bert Cornillie Department of Linguistics Universities of Leuven and Antwerp Belgium e-mail: [email protected]

Ronald W. Langacker Department of Linguistics University of California San Diego USA e-mail: [email protected]

Kiki Nikiforidou Department of English Studies University of Athens Greece e-mail: [email protected]

Tanja Mortelmans Department of Linguistics University of Antwerp Belgium e-mail: [email protected]

List of contributors

Henk Pander Maat Utrecht Institute of Linguistics OTS The Netherlands e-mail: [email protected]

Peter Pelyväs Department of English Linguistics University of Debrecen Hungary e-mail: [email protected]

Satoshi Uehara International Student Center Tohoku University Japan e-mail: [email protected]

Arie Verhagen Centre for Linguistics Leiden University The Netherlands e-mail: [email protected]

ix

Introduction Angeliki Athanasiadou, Costas Canakis, and Bert Cornillie

Subjectivity is a ubiquitous notion figuring prominently in the Social Sciences and the Humanities, yet it has only recently obtained a prominent place in linguistic theorizing. In a broad sense, 'subjectivity' has been traditionally contrasted with 'objectivity'. As the locus of scientific investigation, 'objectivity' is associated with the positivist tradition in the West while subjectivity has been related to the expression of personal opinions, beliefs, and attitudes supposed to be too ephemeral and idiosyncratic to deserve scholarly attention. The opposition between objectivity and subjectivity, broadly understood, has been accommodated within the predominant structural linguistic paradigms in various forms: both Saussure's langue/parole and Chomsky's competence/performance dichotomies can be understood as actually reflecting this opposition. Lyons (1982: 103) argues that "modern AngloAmerican linguistics [...] has been dominated by the intellectualist prejudice that language is, essentially, if not solely, an instrument for the expression of propositional thought", where 'propositional thought' is to be related to the objective description of reality. On the other hand, continental European linguists such as Breal (1964 [1900]), Bühler (1990 [1934]), and Jakobson (1957) have focused on subjectivity. But perhaps the clearest expression of this opposition is Benveniste's (1971 [1958]: 225) distinction between "sujet d'enonce" or the syntactic subject and "sujet d'enonciation" or the speaking subject, along with his understanding of subjectivity as one of the most pervasive functions of language (cf. Traugott 2003: 125). More recent accounts define subjectivity as "the way in which natural languages, in their structure and their normal manner of operation, provide for the locutionary agent's expression of himself and of his own attitudes and beliefs" (Lyons 1982: 102).

2

Angeliki Äthanasiadou,

Costas Canakis, and Bert Cornillie

In current linguistic theorizing, the most prominent exponents of subjectivity are probably Langacker and Traugott who have elaborated this notion and extended it to subjectification according to the premises of their respective theoretical frameworks. Therefore, in the area of cognitivefunctional linguistics, there are now two competing ways of exploring subjectivity and subjectification which are not incompatible although they may not be commensurable either. Whereas initially both authors considered each other's work as being in the same vein and referred to it in support of their respective arguments, recently more attention has been paid to the differences between the two approaches resulting in more fine-grained analyses. In what follows, therefore, these two approaches will be considered separately although possible points of convergence will also be suggested. Langacker, in his seminal 1985 paper on subjectivity, already sets the foundations of his approach to subjectification which he has further developed and refined over the past twenty years (1990, 1995, 1997, 1998, 1999a, 1999b, 2002, 2003). 1 He argues that qualifications such as subjective and objective are to be conceived in terms of viewing arrangement, i.e. they are "characterized relative to the asymmetry between the observer in a perceptual situation and the entity that is observed" (1985: 120, original emphasis). From this point of view, an entity is considered "subjective to the extent that it functions asymmetrically as the observer in a viewing situation, losing all awareness of SELF as it observes an OTHER" (1985: 121-122). Crucially, Langacker does not talk about the extent to which an expression or its meaning is subjective, but rather about the status of a particular element within the overall situation. In other words, an entity is construed with maximal subjectivity when it remains offstage and implicit, inhering in the very process of conception without being its target, and it is construed with maximal objectivity when it is put onstage as an explicit focus of attention (Langacker 1999a: 149). In his 1990 paper, the very shift towards maximal subjectivity is called subjectification (an objective situation is replaced by a subjectively construed relationship). More recently, Langacker, following Verhagen (1995) and Harder (1996), has suggested that subjectification should not be characterized in terms of replacement, that the subjective component is there all along, being immanent in the objective conception and simply left behind when the latter fades away (1999b: 298).

1. In fact, Langacker (1985) does not use the term subjectification yet.

Introduction

3

Although Langacker presents subjectification as a primarily synchronic phenomenon, he also mentions that it is the result of attenuation of the objectively construed subject, thus hinting at its diachronic character. Attenuation in subject control is a pervasive, multifaceted phenomenon that plays a major role in certain kinds of grammaticalization with important consequences for synchronic analysis and description. This driving force towards full subjectification can be observed with respect to at least four parameters: (a) a change in status (from actual to potential, from specific to generic), (b) a change in focus (the extent to which particular elements stand out as focus of attention, e.g. in The child hurried across the street the trajectory is profiled, in The child is safely across the street it is unprofiled, while Last night there was afire across the street is an extreme case of defocusing of the trajectory), (c) a shift in domain (from a physical interaction to a social or experiential one (the evolution of modals), and (d) a change in the locus of activity or potency (from a focused onstage trajector to an offstage addressee) (Langacker 1999a: 155-6). Attenuation and diffusion in the locus of control figure in many cases of grammaticalization, eventually resulting in "transparency". Another influential line of research on subjectivity and subjectification comes from a series of works by Traugott and her collaborators (Traugott 1989, 1995, 1997, 1999, 2003, Traugott and Dasher 2002). In this model, it is subjectification as a diachronic process which is at issue. From this point of view, subjectification is the process whereby "meanings tend to become increasingly situated in the speaker's subjective belief state or attitude toward the proposition" (Traugott 1989: 31). This is, in turn, identical with the last of three tendencies she puts forth as paths of semantic change, i.e., (a) the shift from externally based descriptions to internally grounded assessments (boor 'farmer'>'crude person'), (b) the extension to textual and metalinguistic uses (observe 'notice'>'state'), and (c) an increased involvement of the speaker (while 'during'>'although') (Traugott 1989: 3 4 35). Therefore, if the meaning of a lexical item or construction is grounded in the sociophysical world of reference, it is likely that over time speakers will develop polysemies that are grounded in the speaker's world, i.e., in reasoning, belief, or metatextual attitude to the discourse. In other words, subjectification is "the semasiological development of meanings associated with a form such that it comes to mark subjectivity explicitly" (Traugott 1999: 179). Although Traugott examines subjectification mainly as a diachronic phenomenon, synchronically it may also result in layerings of more or less

4

Angeliki Athanasiadou,

Costas Canakis, and Bert Cornillie

subjective meanings for the same lexical item or construction. Subjectification is a pervasive tendency in semantic change whose effects may be detected in areas as diverse as, for example, the development of lexical items, the development of evidential and epistemic domains, the development of performative uses of locutionary verbs {promise, insist), or the meanings of adverbs like in fact (Traugott 1999: 188-189). The common denominator in these processes is a tendency toward greater pragmaticization of meaning emphasizing the externalization of the speaker's subjective point of view. In this respect, for Traugott, subjectification cannot be seen as characterized by attenuation, but rather as involving '"pragmatic strengthening' and enriching of the form-meaning pair in question with the speaker's perspective" (Traugott 1999: 188). The above survey, albeit epigrammatic, already foreshadows some points of convergence and divergence between the two major approaches to subjectification. Langacker (1999a) suggests that the disagreement is terminological, whereas Traugott (1999) and Traugott and Dasher (2002) argue that they refer to considerably different phenomena, unlike Traugott (1989: 35) which foregrounds the similarities between the two approaches. In fact, it is not clear where the terminological difference lies, since both authors use the same terms. If anything, what seems to be the difference between them is their theoretical assumptions regarding how language is shaped. On the one hand, Traugott relies heavily on the pragmatic aspect of language, implicitly presupposing a distinct semantics, whereas Langacker's model does not draw a similar distinction (cf. Brisard, this volume). The importance of pragmatics in Traugott's notion of subjectification is primarily borne out from the crucial role of speaker/hearer (S/H) interaction and meaning negotiation in semantic change. By contrast, Langacker focuses on the speaker and his own conceptualization of a linguistic expression, leaving the hearer out of the picture. Besides, Traugott's subjectification has much wider applicability: for example, she claims, pace Langacker, that subjectification is not necessarily characterized by attenuation, in view of changes undergone by adverbs such as in fact or scalar particles like even. Therefore, it is hard to see how the differences between the main current approaches to subjectivity and subjectification could amount to a simple terminological divergence, especially since Langacker (this volume) also explicitly warns against confusing his notion of subjectification with that of Traugott's. What can be safely argued at this point is that at least the terminological and theoretical landscape with respect to the relationship of the two approaches is becoming clearer.

Introduction

5

The initial acceptance of Langacker's and Traugott's terms as analogous may be due to the fact that both authors began by anchoring subjectification to grammaticalization. A good example of how the two approaches deal with grammaticalization is provided by the English progressive going to construction in sentences such as An earthquake is going to destroy that town. For Langacker, the evolution from the movement verb to go to the aspectual going to construction involves a shift from a maximally objective to a more subjective construal, in which the "speaker/concepltualizer traces mentally along a path to situate a process with relation to a reference point" (Langacker 1990: 23). Similarly, for Traugott (1995: 34-35), who describes the history of going to in detail, this grammaticalized expression conveys epistemic modalization, i.e. a realignment strengthening speaker perspective. In the same work, Traugott cites Langacker without suggesting there is any incompatibility between the two notions, but simply as different approaches to a similar phenomenon. However, more recently, Traugott and Dasher (2002: 98) take a different stand on the aspectual construction. Specifically, they deny that the subjectivity in the aforementioned example is borne out by a construal account and suggest it is rather a function of a commitment to the likelihood that, the city, for example, shall be destroyed by an earthquake. Moreover, Traugott and Dasher (2002: 98) consider this commitment in terms of S/H interaction, or intersubjectification, which they see as a further elaboration on subjectification, the latter feeding the former. This new element in the debate allows them to clearly distinguish their notion from that of Langacker, who, in their opinion, can provide no principled account of intersubjectification. However, note that in Langacker's subjectivist model the question of intersubjectivity does not arise as there is no theoretical import attached to the S/H dyad. Thus, while Langacker focuses on subjectification in terms of construal relations, Traugott seems to be further reinforcing its pragmatic side. This difference, then, can be identified as reflecting distinct theoretical assumptions. In view of the above, Traugott and Dasher (2002: 100) clearly claim that, unlike Langacker's, their approach to subjectivity is discourse-based and thus contextualized. In so doing, they identify contextualization with discourse and imply that approaches focusing on conceptual structure are ipso facto decontextualized. However, construal takes places in communication, therefore necessarily in context, and is relevant to discourse issues. Indeed, discourse seems to have been an important locus for subjectivity and intersubjectivity for other authors as well. Verhagen (1995, 2000,

6

Angeliki Athanasiadou, Costas Canakis, and Bert Cornillie

2005, this volume), for instance, combines a conceptualist view originating in Langacker's work with the S/H dyad emphasized in discourse studies. Note that it is possible for Verhagen to evoke the notion of intersubjectivity without minimizing the importance of conceptual structure. Therefore, there is no inherent incompatibility between conceptual structure and contextualization (cf. Nuyts 2001). While the papers in this volume do not take up all of the issues discussed above, they do address questions that are still very much debated in the field of cognitive linguistics. Moreover, they add to the insights accumulated over the last two decades providing fresh empirical evidence which, in turn, highlights less thoroughly discussed aspects of subjectivity and subjectification. The papers in this volume do not exhaust the demanding issue of subjectivity, but by exploring its boundaries they seek to better define its main characteristics. The papers fall into three sections. The first three papers (Langacker, Brisard, and Uehara) address larger theoretical issues and deal with subjectivity and subjectification against the background of different theoretical frameworks focusing on theoretical archetypes, pragmatics, and typology. The second section includes six case studies which provide data-based analyses of modal verbs and adjectives. The first three (Pelyväs, Mortelmans, Cornillie) are descriptive studies on subjectification in modals further refining Langacker's system. The remaining papers in this section (Athanasiadou, Breban, and Pander Maat) concentrate on aspects of subjectivity and subjectification in adjectives. Finally, the three papers in the last section (Verhagen, Nikiforidou, and Company) probe into the relation between discourse and the syntactic organization of linguistic expressions exploring, among other things, the possibility of looking at construal relations in light of the S/H dyad (i.e., combining aspects of Langacker's approach with Traugott's). In his paper, Langacker deals with the subjectification and grammaticization of expressions designating conceptual archetypes. Conceptual archetypes refer to fundamental aspects of our everyday experience, function as gestalts and are expressed by basic lexical sources. More specifically, Langacker emphasizes the role of the cognitive operations inherent in the conception and subjectification of these lexical sources that undergo grammaticization. Subjectification, grammaticization and conceptual archetypes are exemplified with respect to a variety of linguistic phenomena, including modals, fictive motion, and quantification. Brisard, in his paper entitled Logic, subjectivity, and the semantics/pragmatics distinction, confronts Traugott's conception of "subjectiv-

Introduction

7

ity" and of the process of grammaticalization with received views on this topic in the cognitive semantics tradition. He argues that neither Traugott's approach to subjectification, nor her account of "pragmatic strengthening" lend themselves easily to straightforward implementation in cognitive terms. Brisard concludes that there are two reasons for cognitive grammarians to beware of uncritically adopting Traugott's framework. The first has to do with the (psychological) substance behind linguistic expressions of subjectivity, on which Traugott takes a process-oriented view. The second reason pertains to the problematic status of the inferences typically giving rise to subjective meanings and their relationship with speaker intentions. While most contributions to this volume contain in-depth analyses of the phenomenon of subjectivity in one language, Uehara examines linguistic subjectivity in breadth. His cognitive typology of subjectivity focuses particularly on the question raised by Lyons (1982), as to "whether different natural languages differ in respect of the degree of subjectivity that they impose upon their users." To this effect, it lays out Langacker's cognitive linguistic theory of subjectivity and examines its applicability to cross-linguistic studies. The data focus on deixis in three types of expressions: motion events, internal state predicates, and social deixis in expressions of nominal reference. Uehara concludes that the cognitive theory of subjectivity cannot only handle cross-linguistic patterns of variation, but it can also serve as a tool in providing a new dimension for the study of linguistic typology. In the first paper on modality in this volume, Pelyvas addresses the discrepancy between conceptual content and formal considerations in the definition of epistemic grounding. To this effect, he proposes to both restrict and extend the number of linguistic categories capable of functioning as grounding predications, suggesting a more gradual view on epistemic grounding. The restriction involves the revision of the status of deontic (English) modals, while the extension concerns the inclusion of (some of the) cognitive predicates and, probably, epistemic modals in German, Dutch, and other languages. The method suggested is a more fine-grained analysis of the conceptual structure of modals postulated in cognitive theory, while the modifications involved are in close correlation with cognitive grammar's changing view on the nature of subjectification. In her paper Langacker 's 'subjectification' and 'grounding': a more gradual view, Mortelmans looks at a number of properties which might enhance the degree of subjectification of a particular modal or a particular

8

Angeliki Athanasiadou,

Costas Canakis, and Bert Cornillie

form of the modal. Although she concentrates on German modals, she claims that a more gradual view on grounding could also be defended for English modals. She concludes that more and less subjectified uses of the modals reflect different construals of the modal relation and observes that each modal has a typical range of uses correlating with particular forms it may take as well as with particular syntactic contexts in which it occurs. Cornillie presents conceptual and constructional considerations on the subjectivity of English and Spanish modals. He argues that, despite their tense marking, Spanish epistemic modals subjectively ground the event expressed by infinitive, whereas the deontic modals rather profile the force structure belonging to the objective scene. Evidence for this claim is the observation that epistemic modals are not hindered by any event participants, which become the focus of attention with deontic modals. The subjectified epistemic modals are considered dynamic reference points directing attention to the infinitival process. The reluctance to show up as proforms and the lack of aspectual marking corroborate their dynamic function. As for the papers on subjectivity and adjectives, Athanasiadou's paper offers a cognitive grammar account of the meanings of adjectives in English. She suggests that their variety of meanings is partly due to the different positions they may occupy. The choice of one position over the other, namely the choice of premodifier, postmodifier or predicative position, reflects the speaker's construal of the situation expressed as being of a particular nature. She relates the construal of an adjectival expression by each one of the three positions with a 'viewpoint', i.e. with the coding of a unique relation to the speaker. Athanasiadou claims that the three positions are not free variants of adjectival expressions, but have specialized semantics and conditions of use. Thus, adjectives, apart from assigning properties or attributes to things, function as expressions of subjective beliefs and attitudes. In her paper Grammaticalization and Sub)edification of the English Adjectives of General Comparison, Breban discusses the extent to which the grammaticalization framework and the notions of subjectification used by Traugott and Langacker help determine three specific functions of pronominal adjectives. Specifically, she applies these notions to the semantic polysemy between lexical and textual readings characterizing adjectives of general comparison in present-day English. She argues that their textual uses, which comprise postdeterminer and classifier uses, can be explained

Introduction

9

as the result of concomitant processes of grammaticalization and subjectification of the lexical, attributive uses. Pander Maat examines Subjedification in gradable adjectives. First, he presents a short survey of the well-known properties of degree adjectives, concluding they can be coherently accounted for drawing on Langacker's notion of subjectification. Then, he provides two versions of a subj edification analysis of degree adjectives. The first one is based on the common assumption that degree adjectives refer to a degree of some property that exceeds the standard value for this property for a certain entity. This analysis reveals insurmountable problems. Second, he suggests an analysis combining the notion of subjectification with Ducrot's argumentative perspective. The author concludes that the second analysis should probably be preferred, as it avoids a number of problems inherent to the standard value approach. In the third section of the volume, the subjectivity of syntactic representations is examined. Verhagen's paper addresses the question of hearerspeaker interaction in terms of subjectivity and intersubjectivity and applies it to different syntactic constructions, such as w/z-movcmcnt. He begins with elaborating on a definition of subjectivity as having two dimensions: one in which a conceptualizing subject is opposed to an "object" of conceptualization and one in which the conceptualization of one subject is compared to that of another. These two dimensions jointly define what he calls the "basic construal configuration". Thus, one dimension of subjectivity in an expression resides in the construal relationship, while the second dimension relates to cognitive coordination between the very subjects of conceptualization themselves. Based on this definition, Verhagen then focuses on conceptual structure in relation to discourse. Nikiforidou, in her paper on Subjective construal and factual interpretation in sentential complements, shows that a complementation and an aspectual construction in Greek resemble each other in that both involve subjective construal of the reported event in Langacker's sense. She argues that both constructions construe the speaker as being inside or close to the reported/narrated event(s), therefore precluding a viewing arrangement in which the speaker/conceptualizer remains outside the maximal scope of the predication. This construal in both constructions correlates with a factive interpretation. She argues that both constructions conventionally code expertise arising from experience and/or first-hand knowledge, motivating in turn their factual interpretation.

10

Angeliki Athanasiadou, Costas Canakis, and Bert CorniJlie

In Zero in Syntax, Ten in Pragmatics: Subjectification as Syntactic Cancellation, Company looks more closely at the syntactic correlates of subjectification explaining why subjective expressions are usually syntactically impoverished. From a Traugottian point of view, she examines how subjectification affects Spanish syntax, analyzing the syntactic evolution of four discourse markers and focusing on fixation and autonomy of predication and loss of syntactic capacities. Company concludes that subjectivity might be labeled as "zero in syntax, ten in pragmatics". Finally, the analysis shows a strong parallelism between the semantic and the syntactic side of a subjectification change. There is a strict co-evolution in this kind of change, in that the internal, abstract, subjective semantic phase of the process goes hand in hand with a syntactic phase of diminished necessity for syntax. The present volume has made an attempt to define subjectivity, subjectification and intersubjectivity and has shown the way these notions may have reshaped our understanding of grammar. The crucial factors that have been taken into account is vantage point and the orientation and perspective assumed by the speaker, internal or external. One of its aims has been to contribute to a better understanding of the similarities and differences between the two major approaches to subjectivity and subjectification. Some contributions have managed to work towards integration, suggesting, for instance, that the two approaches are complementary to each other, while others have reached different conclusions being solely based on one of the two approaches. The papers in the volume represent a variety of languages and constructions. Thus, they bring us a step closer to a better understanding of the various aspects of this multifaceted phenomenon of semantic change or semantic extension, a phenomenon which requires rigorous analysis of seemingly diverse data from typologically unrelated languages if its ramifications are to be fully understood.

References Benveniste, E. 1971 Subjectivity in language. In Problems in General Linguistics, trans. Mary Elizabeth Meek. Coral Gables, FL: University of Miami Press, 223-30. [Original edition 1958] Breal, M. 1964 Semantics: Studies in the Science of Meaning, trans. Mrs Henry Cust. New York: Dover. [Original edition 1900]

Introduction Bühler, Κ. 1990

Harder, P. 1996

Theory of Language: The Representational Function of Language, trans. Donald Fraser Goodwin. Amsterdam: Benjamins. [Original edition 1934] Functional Semantics: A Theoiy of Meaning, Structure, and Tense in English. Berlin and New York: Mouton de Gruyter.

Jakobson, R. 1957 Shifters, Verbal Categories, and the Russian Verb. Cambridge, MA: Harvard University Russian Language Project. Langacker, R. W. 1985 Observations and speculations on subjectivity. In J. Haiman (ed.), Iconicity in Syntax. Amsterdam: Benjamins, 109-50. 1990 Subjectification. Cognitive Linguistics 1: 5-38. 1995 Raising and transparency. Language 71: 1-62. 1997 Consciousness, construal, and subjectivity. In Μ. I. Stamenov (ed.), Language Structure, Discourse and the Access to Consciousness. Amsterdam and Philadelphia: Benjamins, 49-75. 1998 On subjectification and grammaticization. In J.-P. Koenig (ed.), Discourse and Cognition: Bridging the Gap. Stanford: CSLI Publications, 71-89. 1999 a Losing control: grammaticalization, subjectification, and transparency. In A. Blank and P. Koch (eds.), Historical Semantics and Cognition. Berlin and New York: Mouton de Gruyter, 14775. 1999 b Grammar and Conceptualization. Berlin and New York: Mouton de Gruyter. 2002 Deixis and subjectivity. In F. Brisard (ed.), Grounding: The Epistemic Footing of Deixis and Reference. Berlin and New York: Mouton de Gruyter, 1-38. 2003 Extreme subjectification: English tense and modals. In H. Cuyckens, T. Berg, R. Dirven, and K.-U. Panther (eds.), Motivation in Language: Studies in Honor of Günter Radden. Amsterdam and Philadelphia: Benjamins, 3-26. Lyons, J. 1982 Deixis and subjectivity: Loquor, ergo sum?. In R. J. Jarvella and W. Klein (eds.), Speech, Place, and Action: Studies in Deixis and Related Topics. Chichester and New York: John Willey, 101-24. Nuyts, J. 2001 Subjectivity as an evidential dimension in epistemic modal expressions. Journal of Pragmatics 33: 383^-00.

11

12

Angeliki Athanasiadou, Costas Canakis, and Bert Cornillie

Stein, D. and S. Wright (eds.) 1995 Subjectivity and Subjectivisation: Linguistic Perspectives. Cambridge: Cambridge University Press. Traugott, Ε. C. 1989 On the rise of epistemic meanings in English: an example of subjectification in semantic change. Language 65: 31-55. 1995 Subjectification in grammaticalisation. In D. Stein and S. Wright (eds.), 31-54. 1997 Subjectification and the development of epistemic meaning: the case of promise and threaten. In T. Swan and O. J. Westvik (eds.), Modality in Germanic Languages. Berlin: Mouton de Gruyter, 185-210. 1999 The rhetoric of counter-expectation in semantic change: a study in subjectification. In A. Blank and P. Koch (eds.), Historical Semantics and Cognition. Berlin and New York: Mouton de Gruyter, 177-96. 2003 From subjectification to intersubjectification. In R. Hickey (ed.), Motives for Language Change. Cambridge: Cambridge University Press, 124-39. Traugott, Ε. C. and R. B. Dasher 2002 Regularity in Semantic Change. Cambridge: Cambridge University Press. Verhagen, A. 1995 Subjectification, syntax, and communication. In D. Stein and S. Wright (eds.), 103-128. 2000 'The girl that promised to become something': an exploration into diachronic subjectification in Dutch. In T. F. Shannon and J. P. Snapper (eds.), The Berkeley Conference on Dutch Linguistics 1997: The Dutch Language at the Millennium. Lanham, MD: University Press of America, 197-208. 2005 Constructions of Intersubjectivity. Oxford: Oxford University Press.

Editors' note The present volume is a compilation of selected contributions from a theme session we organized in the frame of the 8th International Cognitive Linguistics Conference held at the University of La Rioja, Spain, July 20-25, 2003. Despite ten year span separating it from Subjectivity and Subjectivisation (Stein and Wright 1995), the present volume establishes continuity with the 1995 collection of papers, and places subjectivity and sub-

Introduction

13

jectification within the wider framework of general processes of conceptualization. The reviewers of this volume are Michel Achard (Rice), Johan van der Auwera (Antwerp), Anastasios Phoibos Christidis (Thessaloniki), Kristin Davidse (Leuven), Nicole Delbecque (Leuven), Gabriele Diewald (Hannover), Rene Dirven (Duisburg), Teresa Fanego (Santiago de Compostela), Stefan Grondelaers (Leuven), Peter Harder (Copenhagen), Przemyslaw Lozowski (Lodz), Ricardo Maldonado (UNAM, Mexico), Juana I. Marin Arrese (Complutense, Madrid), Brigitte Neriich (Nottingham), Jan Nuyts (Antwerp), Carita Paradis (Lund), Günther Radden (Hamburg), Masayoshi Shibatani (Kobe/Rice), Dieter Stein (Düsseldorf), Arie Verhagen (Leiden), Jean-Christophe Verstraete (Leuven). Many thanks to all of them for their kind collaboration.

Section I Large theoretical issues

Subjectiflcation, grammaticization, and conceptual archetypes Ronald W. Langacker

The importance of subjectivity and subjectiflcation is more and more widely recognized among linguistic scholars. These terms, however, are used in different ways, and not always with the greatest precision. To make progress in this difficult area, we need a clear understanding of what is intended. Let me first point out that Traugott uses these terms (e.g. Traugott 1982, 1989) rather differently than I do. These uses are not unrelated, nor is it a matter of one being right and the other wrong - both pertain to important aspects of meaning and semantic change. It is however essential that they not be confused. For Traugott, the term subjectiflcation refers to the hypothesized tendency for meanings to become more subjective. This decomposes into three more specific tendencies (Traugott 1989: 34-35). Tendency I: Meanings based in the external described situation > meanings based in the internal (evaluative/perceptual/cognitive) described situation. Examples cited include such phenomena as pej oration, like the shift in meaning of boor, and the general direction of metaphorical extension from concrete to abstract domains. Tendency II: Meanings based in the external or internal described situation > meanings based in the textual and metalinguistic situation. This is illustrated by observe, which shifts from a "mental-state" to a "speech-act" verb. Tendency III: Meanings tend to become increasingly based in the speaker's subjective belief state/attitude toward the proposition. An example is the extension of while from temporal to concessive use, or the grammaticization of go to indicate "immediate, planned future". Traugott's definition of subjectivity and subjectiflcation pertains to the domain in which a situation resides (a matter of conceptual content). It therefore makes sense to talk about an expression's meaning becoming

18

Ronald W. Langacker

more subjective. For me, on the other hand, the terms pertain to vantage point (a matter of construal). In my usage it makes no sense to talk about the extent to which an expression or its meaning is subjective - we can only talk about the status of a particular element within the overall situation. A given meaning always comprises both subjectively and objectively construed elements. As I use them, the terms subjective and objective allude to the subject and object of conception (Langacker 1985, 1987a: 3.3.2.4, 1997, 2002a). An entity is said to be objectively construed to the extent that it goes "onstage" as an explicit, focused object of conception. An entity is subjectively construed to the extent that it remains "offstage" as an implicit, unselfconscious subject of conception. At issue, then, is the inherent asymmetry between the conceptualizer and what is conceptualized, between the tacit conceptualizing presence and the target of conceptualization. The asymmetry is maximal when the subject of conception lacks all self-awareness, being totally absorbed in apprehending the onstage situation, and the object of conception is salient, well-delimited, and apprehended with great acuity. These are of course matters of degree. But whether they are sharply distinct or somewhat blurred, the subject and object roles figure in every conceptualization. In principle, an expression's meaning always incorporates the conceptualizing presence who apprehends and construes the situation described. This basic arrangement is sketched in Figure 1. An expression's meaning always comprises both subjectively and objectively construed elements. Minimally, subjectively construed elements include the speaker, and secondarily the addressee, in their offstage role as the conceptualizers who employ the expression and thereby apprehend its meaning. Minimally, objectively construed elements include the expression's profile, i.e. what it designates (or refers to) within the conception evoked. Taken as a whole, therefore, an expression's meaning is neither subjective nor objective - as these terms are defined, it is only particular elements that are construed in a subjective or objective manner. By the same token, a semantic shift does not - in my sense of the term result in a global meaning becoming more subjective. Instead, we have to talk about the fate of individual conceptual elements, whether a particular element comes to be construed with a greater degree of subjectivity or objectivity. Consider the English modals (may, will, must, etc.). With apologies for grossly oversimplifying their long and enormously complex indi-

Subj edification,

grammaticization,

and conceptual archetypes

19

• maximal scope of conception

r\

• focused object of conception (profile)

τ

• onstage region (immediate scope)

1 [

• apprehension/construal by S

fa

• subject of conception (speaker/hearer)

Figure 1.

vidual histories, I summarize a major shared facet of their historical development in Figure 2. As main verbs, with meanings such as 'want', 'be able', etc., they profiled relationships in which the trajector (tr), expressed by the grammatical subject, manifests some tendency or potential for carrying out the occurrence expressed by the infinitival complement (a relational landmark (lm)). This latent force, or potency, is represented by a double dashed arrow - a double arrow to indicate a force dynamic relationship (Talmy 1988), and dashes because it is only potential. With such verbs, the main clause subject would often be the actor with respect to the subordinate clause event (hence the dotted correspondence line). Modern English Epistemic Modals

Old English Main Verbs

>

Figure 2.

20

Ronald W. Langacker

The Modern English epistemic modals exhibit the basic structure shown on the right (Langacker 1991: 6.3, 2002a, 2004). What was the subordinate verb now constitutes the profiled relationship in a single-clause structure. The role of the grammatical subject is now confined to its participation in that relationship - it is no longer the locus of the modal force or potency (hence the transparency of the modals, allowing as grammatical subject anything able to function as subject of their complement (Langacker 1995a)). It is instead the speaker who is the source of the potency. The speaker's epistemic judgment concerning the onstage process amounts to some degree of inclination toward accepting it as part of the speaker's conception of reality. The epistemic modals are distinguished from one another primarily by the strength of this inclination (cf. Sweetser 1982; Talmy 1988). Comparing the structures in Figure 2, I would say that the modal force has undergone subjectification. As main clause profile, it was originally the onstage focus of attention, hence objectively construed. Though essential to the meaning of English modals, it is now offstage and unprofiled, inhering in conceptualizing activity of the subject of conception (the speaker), hence subjectively construed. But if the force dynamic relationship is now construed more subjectively, other elements have shifted in the opposite direction. Originally backgrounded, the complement process is now the onstage focus of attention, the profiled main clause relationship. As such it is now construed with a higher degree of objectivity. It can also be argued that the speaker is now construed more objectively. In addition to its original role as the tacit subject of conception, the speaker now assumes the additional responsibility of making the epistemic judgment signalled by the modal. The speaker still remains offstage and implicit, hence subjectively construed, but her presence is now a bit more palpable owing to this force dynamic trace of her activity. I wish to emphasize that this description is perfectly compatible with characterizing the change as an instance of subjectification in Traugott's sense. It instantiates tendency I and is a prime example of tendency ΙΠ. It also represents subjectification in another sense, related to tendency ΠΙ, namely that greater weight is now given to the speaker's (subjective) judgment, in contrast to the idealized model of disinterested, objective reporting. The rationale for my own terminology runs as follows. Rather than treating an expression's meaning as an unanalyzed, undifferentiated whole, I am trying to say something substantive about the specifics of its internal conceptual structure. In this finer-grained attempt at describing

Subjectification,

grammaticization,

and conceptual archetypes

21

meaning and semantic change, it becomes possible and indeed essential to determine the status and change of status for individual conceptual elements. It is at this level of analysis that the terms subjective/objective construal, as well as subjectification/objectification, come into play. They pertain to the status of individual elements with respect to a fundamental dimension of conceptual organization, the asymmetry between the subject and the object of conception. This dimension is not the only one to be concerned with, but it is surely important and needs to be accommodated in a complete description. In earlier works (e.g. Langacker 1990, 1991: 5.2.3.2), I characterized subjectification as the realignment of some relationship from the objective axis to the subjective axis. An example is the subjectification of modal force illustrated in Figure 2. Originally, the exertion of force was part of the onstage situation described, residing in the interaction of focused elements, hence objectively construed. It winds up being subjectively construed, residing instead in how the subject of conception apprehends what is left of the onstage situation. More recently (Langacker 1998, 1999), I have come to believe that subjectification is better thought of as a kind of semantic "bleaching" or "fading away". The subjectively construed entity which remains as a vestige of an objectively construed counterpart was actually there all along, immanent in the latter (i.e. inherent in its conception). It simply becomes more evident when the objectively construed element is no longer there to mask it. The immanence of the subjectively construed entity in its objectively construed counterpart is perhaps least evident in the case of modals, but even here I believe it is valid. Let me suggest a way of thinking about epistemic modals which may be helpful in this respect. In an English finite clause, the absence of a modal indicates that the profiled relationship is accepted by the speaker as being real. For instance, (l)(a) portrays the vacation planning as part of the speaker's current conception of reality (the history of what has happened up through the present). The effect of a modal, as in (l)(b), is to indicate that the profiled relationship has not yet been incorporated in the speaker's reality conception. It is however a candidate for acceptance. It is under consideration, and the speaker inclines toward accepting it with varying degrees of force, reflected in the different modal choices (Langacker 1991: 6.3.2, 2004). We can think about this inclination in terms of mental extrapolation: the speaker extrapolates from the current view of reality, projecting its future development in such a way that it comes to incorporate the target situation. This extrapolation requires men-

22

Ronald W. Langacker

tal effort and engenders a force dynamic experience. However, since we are always striving to make sense of the world and anticipate future events (just by virtue of being alive and sentient), the effort expended in doing so lies in the background, as something taken for granted, much like the constant force of gravity. For this reason the mental extrapolation of reality toward a target event presents itself to us as a force inherent in the evolution of reality itself. We feel it as "pushing" us toward the target with varying degrees of strength. Hence a stronger force (as with must) translates into a greater likelihood of the target being reached. (1)

a. They are planning a vacation in Logrono. b. They {may / might / should / will / must} be planning in Logrono.

a vacation

So characterized, the modal force is subjectively construed. Primary focus is on the profiled relationship, and while the modals do make evident the strength of the speaker's inclination, that inclination is not itself the focus of attention or the topic of discussion - they merely register the speaker's epistemic struggle in assessing the target situation onstage. In regard to subjectification, therefore, the question is whether the speaker has a comparable force dynamic experience as an inherent aspect of apprehending the objectively construed relationships profiled by the historical antecedents of the modals. Is such an experience engendered by the very process of conceptualizing someone wanting or being able to do something, as shown in the first diagram in Figure 2? While I cannot prove that this is so, I strongly suspect that it is. Wanting, for example, is a force dynamic experience. If I conceptualize somebody else wanting something, and successfully apprehend the nature of that person's experience, I must to some extent - however minimally - vicariously have an experience of the same basic sort. This would conform to a general proposal, being seriously considered in various quarters, that the meaning of an expression describing an action or an experience involves a mental simulation of it (cf. Barsalou 1999). Be that as it may, the validity of the revised conception of subjectification is more obvious in other kinds of cases. It first became evident to me from the relation between the two senses of across illustrated in (2): (2)

a. A giant chicken marched angrily across the street. b. There's a KFC outlet right across the street.

Subjectification,

grammaticization,

and conceptual archetypes

23

In (2)(a), an onstage participant (the chicken) successively occupies all the positions vis-ä-vis the street which constitute the spatial relationship profiled by the path preposition. This sense of across is shown on the left in Figure 3. In conceptualizing it, the subject of conception necessarily conceives of the trajector occupying all these positions, in the proper sequence, through processing time (Τ). By contrast, the trajector of across in (2)(b), namely the KFC outlet, occupies just a single location, corresponding to the endpoint of the path in (2)(a). The conceptualizer does still invoke a path which traverses the landmark (the street). But here there is no onstage participant who moves along it. Instead, the conceptualizing subject scans mentally along the same path by way of computing the trajector's (static) location. Observe, however, that the conceptualizer scans along it in just the same way in (2)(a), in conceptualizing the trajector's path of motion. This mental scanning by the conceptualizer is therefore immanent in the conceptualization of the objective path of motion. It remains in (2)(b) as any notion of actual physical movement fades from the picture.

maximal scope

maximal scope

immediate scope lm

immediate scope lm

tr

TV"

Γ Ο ' Ά \ J I /

>

\!

\\\l

\ u / \ 11 / /

ι I 1/

—

τ

•P

λΛ>1L \ψ

τ

Figure 3.

The same holds true for the future sense of be going to (Langacker 1998, 1999). On the motion interpretation of (3)(a), the subject moves through space with the intent of mailing the letter upon reaching the endpoint of the path. This unfolds through time, so in conceptualizing the action, the subject of conception must also scan through time as well as space in tracking

24

Ronald W. Langacker

the spatial movement of the onstage trajector. The future sense results when the notion of spatial movement fades away. On the future interpretation {gonna), the onstage subject is not portrayed as moving; the idea of the subject intending to act may also disappear, as in (3)(b). What remains is the conceptualizer's mental scanning through time to situate the infinitival process with respect to the reference time. The conceptualizer's mental scanning is of course subjectively construed, inhering in the conceptualizing activity rather than the onstage configuration. And the same scanning through time is immanent in the conception of the objectively construed relationship of spatial movement. (3)

a. He was going to mail the letter. b. It is going to rain.

[motion or future intention] [future projection]

As I define it, subjectification is not the same as metaphor. I do not claim, for instance, that the subject of gonna is conceived metaphorically as moving through time. This is not to say that they are unrelated, or that metaphor might not be involved in cases of grammaticization where subjectification also figures (Sweetser 1988; Traugott 1988). For instance, metaphorical projection to abstract domains might be followed by a decrease in analyzability (awareness of the source domain) and eventual loss of the original objective content (Langacker 1987a: 12.1). Still, subjectification is not metaphorical per se. To see the difference, consider the verb rise. In (4)(a) it profiles actual - or what Talmy (1996) calls factive - motion by the subject. This is onstage and objectively construed. Through time (/), the trajector occupies successively higher positions with respect to the vertical axis (V). As shown in Figure 4(a), the subject of conception follows the trajector's spatial progress in apprehending the profiled event. The conceptualizer's scanning through the stages of this event unfolds through processing time (7). In (4)(b) rise is used metaphorically in reference to an increase in price. This is the same as (4)(a) except that the source domain of spatial motion is superimposed on the target domain of coffee pricing to form a blend (Fauconnier and Turner 2002). The trajector (the price of coffee) is still construed objectively in its movement along the vertical axis, and the conceptualizer still scans this event sequentially through processing time, as seen in Figure 4(b). The only difference is that this movement occurs in a blended space where the vertical axis is conflated with a quantity scale for price. Relative to the basic sense, the metaphorical sense is enriched rather than impoverished (unlike the coffee purchaser).

Subjedification, grammaticization, and conceptual archetypes (4)

25

a. The balloon rose quite slowly. [objective, "factive" motion] b. Last year the price of coffee rose steadily. [metaphorical motion (objective)] c. The trail rises steeply near the summit. [subjective, "fictive" motion]

Figure 4. Sentence (4)(c) exemplifies what is variously referred to as virtual, subjective, or fictive motion (Langacker 1986, 2003a; Matsumoto 1996; Talmy 1996). Here, objectively, there is no change through time - nothing in the onstage scene actually moves. (Rise is therefore imperfective, and occurs in the simple present tense, indicating constancy through time.) The trajector does still occupy a series of successively higher points along the vertical axis, but does so simultaneously, being a spatially extended object. The sense of directionality, and the motivation for using rise, reside in the conceptualizer mentally scanning along the trajector's expanse in building up to a full conception of the static configuration. This motion by the subject of conception is subjectively construed: the conceptualizer does not think of herself as moving through space, but merely apprehends the scene; the movement is inherent in the very conceptualizing activity, hence offstage and construed subjectively. Of course, the conceptualizer traces an analogous mental path in (4)(a), in tracking the balloon's objectively construed movement through space. The conceptual element of spatial movement therefore undergoes subjectification when rise is extended from factive to (imperfective) fictive motion. Consider next a basic possessive verb, like English have. Such predicates derive historically from verbs implying physical acquisition and con-

26

Ronald W. Langacker

trol: 'seize', 'catch', 'carry', 'hold', 'get', etc. (Heine 1997: 91). When grammaticized as general possessive predicates, these no longer profile specific actions, but rather a stable (hence imperfective) relationship in which the trajector controls the landmark. This steady-state control relationship is often multifaceted, implying some kind of exclusive privilege of access or potential for interaction, of a physical, social, and/or experiential nature. I suspect, in fact, that across the range of uses experiential (as opposed to purely physical) factors predominate. Some typical examples are given in (5). (5)

a. b. c. d. e. f.

I have a new computer. Alice has a very good job. Her brother has several grandchildren. Marvin has frequent headaches. They have a lot of hurricanes in Florida. Sherridan has brown eyes.

(a) Possessive Prototype

(b) Possessive Schema

- = connection/mental association R = possessOR (reference point) Τ = possessED (target) D = domain of control/access (dominion) S = subject of conception Τ = processing time Figure 5.

Subjectification,

grammaticization,

and conceptual archetypes

27

Prototypical possession is sketched in Figure 5(a). In the ways just indicated, the possessor (R) exercises control over the possessed (T). R's domain of control (D) is the set of entities to which it has privileged access. A verb like have profiles this control relationship, which is onstage and objectively construed. The diagram further indicates one aspect of what I believe to be involved in conceptualizing a control relationship of this sort. In the conceptualization of R controlling T, the subject of conception (S) traces a mental path from R to Τ - focusing first on R, then on Τ - by way of apprehending the direction of influence. I hold that this mental progression from R to Τ is immanent in the conception of the control relationship, even constitutive of its inherent directionality. This reflects the following working assumption: ... Any conception involving ordering or directionality at the experiential level implies some type of seriality at the processing level; an ordered conception necessarily incorporates the sequenced occurrence of cognitive events as one facet of its neurological implementation, and this sequencing is taken as being constitutive of the conceptual ordering. (Langacker 1986: 455)

This assumption is admittedly speculative but certainly plausible. Indeed, it could hardly be otherwise. The uses of have in (5) show varying degrees of subject control, which can be quite tenuous. In cases like (5)(e)-(f), the grammatical subject is only marginally a controller, even in the sense of having experiential access to the object. In such uses the onstage subject may function predominantly as a point of reference, serving to indicate the general location where the object can be found. In (5)(e), for instance, they can be identified as the inhabitants of Florida, most of whom may never individually experience a hurricane. And in many other examples, there is no real sense at all in which R controls or experiences T, e.g. in (6): (6)

a. This house has four bedrooms. b. The earth has a circumference of about 25,000 miles. c. The Grand Canyon has an interesting geological history.

It is for this reason that I proposed a reference point characterization of possessive expressions (Langacker 1993a, 1995b; Taylor 1996). As a schematic description of possessives (i.e. one valid for all instances), it can only be supposed that R functions as a conceptual reference point allowing the conceptualizer (S) to mentally access a target (T). Owing to

28

Ronald W. Langacker

some connection between them, the subject of conception is able to evoke R as a means of "reaching" or identifying T, i.e. Τ is mentally accessed via R. R ' s dominion (D) is defined as the set of potential targets accessible via R. This schematic characterization is sketched in Figure 5(b). Thus, starting from prototypical cases and moving toward more peripheral ones, the nature and degree of R ' s active control over Τ (onstage and objectively construed) is progressively attenuated. At the extreme, there is no sense at all in which R is a true controller; there is only some conceived connection between them. In such cases R is limited to the passive role of serving as conceptual reference point, allowing the offstage conceptualizer to mentally access a target. My point, of course, is that R also serves in this passive, reference point capacity even in prototypical instances. The mental progression whereby S evokes R, then T, through processing time (7) is immanent in the apprehension of prototypical possessive relationships. It remains even as all vestiges of active control by R fade away. Hence the profiled relationship in the most general description of possessives implies nothing more than the conceptualizer's ability to mentally access Τ via R by apprehending a connection between them. Active control by the onstage, objectively construed grammatical subject gives way to experiential control (mental access) on the part of the offstage, subjectively construed conceptualizer, the subject of conception. In short, the control relationship undergoes subjectification. The pattern that emerges across these examples is not an unfamiliar one. The expressions affected by subjectification tend to be frequent, basic, and fairly general. They usually represent conceptual archetypes, i.e. fundamental aspects of our everyday experience that we deal with as simple gestalts whatever their actual complexity. Among the archetypal notions which have figured in my examples are WANT, BE ABLE, INTEND, GO, RUN, RISE, HOLD, GET, OWN, CONTROL, and USE. Through a course of se-

mantic development which may be quite long and elaborate, meanings emerge that are more abstract and more tenuous in terms of their objectively construed content. Owing to their more schematic meanings, the descendant expressions are often considered grammatical rather than lexical (this is of course a gradation). I wish to be clear that subjectification is not the only factor involved in grammaticization, nor do all instances of subjectification result in grammatical elements (consider the alternate senses of across and rise). Let me offer another point of clarification, as well as a disclaimer. Subjectification should not necessarily be thought of as a mechanism of

Subjectification,

grammaticization,

and conceptual archetypes

29

semantic change. A more cautious approach is to view it as describing or summarizing one kind of relationship between original and extended meanings. The end result - complete disappearance of an objectively construed entity with retention of mental operations immanent in its conception - may come about through a long diachronic process involving numerous factors, including the gradual and multifaceted attenuation of conceptual content (Langacker 1999). I would never claim, for example, that the modern English epistemic modals derive from their Old English lexical sources in a single step or by means of a single kind of development. The discrepancy noted in Figure 2 summarizes an enormously complicated evolutionary process spanning many centuries, which - to be fully and correctly understood - has to be studied in its own right, in finegrained detail, by proper diachronic methods. My disclaimer is that I have not engaged in such research and thus have only limited understanding of the historical processes involved. Thus any picture I can give of historical developments is at best a coarse-grained sketch. Nevertheless, I find that comparing original and extended meanings in this manner is a very useful technique for fine-grained conceptual analysis. From the standpoint of both semantics and grammar, I think it is quite important to observe that the meanings of grammatical elements are plausibly characterized as residing in mental operations inherent in the conceptualizations constituting the meanings of their lexical sources. It is helpful to realize that grammatical meanings may lack objective content but still be conceptual meanings, the skeletal processing activity immanent in the conception of such content. To the extent that this is true, it reveals something significant about the nature and development of grammatical structure. It reveals that important aspects of grammatical structure reside in conceptual operations essentially devoid of any onstage conceptual content. Grammatical meanings may consist solely in the activity of the conceptualizing subject, activity which is immanent in the conceptualization of objectively construed situations but has come to be used independently, in abstraction from it. Such meanings are of course invisible or non-existent from the standpoint of an objectivist semantics. This a primary reason why the meaningfulness of grammar has not been generally recognized. This characterization of grammatical meaning is not necessarily limited to cases where it comes about through subjectification affecting a lexical source. Essential aspects of grammatical meaning may be subjectively construed and immanent in conceptual archetypes even in the absence of any evident diachronic connection.

30

Ronald W. Langacker

Consider the Japanese "double subject" construction, exemplified in (7). In a Cognitive Grammar analysis (Kumashiro 2000; Kumashiro and Langacker 2003), the overall expression is a clausal structure which profiles the relationship obtaining between the "outer" subject, in this case nihonzin-ga 'Japanese', and the following "nuclear" clause, kome-ga syusyoku-da 'rice is the staple food'. The profiled relationship is not coded by any overt morphological element; it is signalled merely by juxtaposing the subject nominal and nuclear clause. What does this profiled relationship amount to, semantically? I would argue that it does not have any significant objectively construed conceptual content. As shown in Figure 6(a), its essential import resides in the subjectively construed operation of the conceptualizer invoking the subject to mentally access the nuclear proposition. That is, the overall expression profiles a reference point relationship, just as in Figure 5(b), except that the target is a proposition (rather than a thing). If there is any objectively construed relational content, it is only the maximally schematic one of there being some connection between trajector and landmark. The essential factor is that the conceptualizer is able to exploit this connection to make a mental transition from reference point to target, so that the nuclear proposition is apprehended via and in relation to the reference point. (7)

Nihonzin-ga kome-ga syusyoku-da. Japanese-SUBJ rice-SUBJ staple: food-be 'The Japanese have rice as their staple food.'

(a) Double Subject Construction Clause

Figure 6.

(b) Topic Construction

Subjedification,

grammaticization,

and conceptual archetypes

31

Hence the grammatical meaning of this construction resides in the subject of conception moving subjectively from reference point to target. This is quite comparable to the most schematic meaning of have, which is why this verb is naturally adopted for the English translation. This construction does not however represent the grammaticization of any have-type predicate. Presumably it emerges historically through a clause-external topic construction being reanalyzed as a single-clause expression. Be that as it may, I would posit the same kind of reference point relationship as the essential meaning of a clause-external topic construction, as in (8), diagrammed in Figure 6(b). The topic specifies the dominion in which the conceptualizer can access the proposition expressed by the comment clause (i.e. the context for its interpretation). The difference from (7) is that the topic is external to the comment clause (not part of a single clausal structure) and that the reference point relationship is unprofiled. (8)

The Japanese, they have rice as their staple food.

Let us next examine locative prepositions. In a full characterization, it is not sufficient to say that a preposition profiles a particular spatial relationship between trajector and landmark. A third entity needs to be specified, namely what is called the search domain (Hawkins 1984). A search domain is the spatial region to which a locative expression confines its trajector, i.e. the set of trajector positions that will satisfy its specifications. The main point is that a preposition does not say in very specific terms just where its trajector must be vis-ä-vis the landmark - it can be anywhere within an extended region. This region, the search domain, plays a specific role in various grammatical constructions (Langacker 1993b). The search domain (D) for a couple of prepositions is very roughly sketched in Figure 7. I am going to suggest that the abstract organization of prepositional meanings - in terms of trajector, landmark, and search domain - reflects the conceptual archetypes of SEARCHING and FINDING. It is for this reason that search domain seemed like a natural term to Hawkins, and before him to Miller and Johnson-Laird (1976: 384), from whom he borrowed it. Searching through space in order to find something is a ubiquitous everyday experience. A typical way in which we do it is by first locating a salient reference object, which we know how to find, and then searching in its vicinity. Thus, I can direct you to a restaurant by telling you that it is very close to the post office; assuming that you know where the post office is,

32

Ronald W. Langacker

you can just go there and search in its vicinity until you find the restaurant. Or if you tell me that my car keys are on my desk, I can simply go to my desk and scan its upper surface. In schematized form, these are precisely the elements that figure in the description of prepositions, as sketched in Figure 7. The trajector can be found somewhere within a domain of search whose position is characterized in relation to a spatial landmark. Individual prepositions position it differently, but it is always located in reference to the landmark, which is coded by the prepositional object. (a) above

tr

D

Ä

v \ W \ v.

\\\\ \

F ^ v\ *

%

Figure 7.

Prepositions are thus well suited for giving instructions about an actual search: (9)

a. Go the the post office - the restaurant is quite near it. b. You can find your keys by looking on your desk.

However, this is only one possible use of prepositions, not necessarily even a privileged one. The actual notion of searching and finding is contributed by other linguistic elements (e.g. go, find, and look in (9)) or is simply apparent from the context. We use prepositional phrases all the time, for many purposes, even when we know exactly where the trajector is and do not have to search for it, as in (10):

Subjectification, grammaticization, and conceptual archetypes (10)

33

The lamp on the desk is more expensive than the one beside the table.

Moreover, prepositions are extended to all sorts of non-spatial uses, where searching and finding (at least in the archetypal sense) are not at issue. My suggestion is that prepositions reflect the conceptual archetypes of SEARCHING and FINDING, not that these are directly incorporated in their meanings. More specifically, certain conceptual operations immanent in these archetypes are also immanent in prepositional meanings, and are largely constitutive of their schematic meaning - the abstracted semantic commonality of prepositions as a class. As indicated by the labels S, R, and Τ in Figure 7, their schematic characterization specifies only that the subject of conception (S) evokes the landmark as a reference point (R) in order to mentally situate the trajector, the target (T) of search. The search domain is thus the reference point's dominion (D), i.e. the region accessible through it. Apart from trajector/landmark alignment (a matter of focal prominence imposed on the scene), this schema is essentially the same as the possessive schema in Figure 5(b). This is why locative expressions are often extended to possessive use, and conversely (Langacker 2003 c). This subjectively construed path of search, where the subject of conception traces a mental path via R to T, is immanent in the conceptualization of a physical act of searching and finding, where an onstage participant moves physically along the same spatial path. The schema for prepositions therefore has the same kind of relationship to the archetype of searching and finding that is characteristic of subjectification. It is not however the case that prepositions derive historically from verbs meaning SEARCH or FIND. (They have other sources, including body-part expressions (Rubba 1994; Langacker 2002b).) Rather than subjectification as a diachronic process, their relationship illustrates the more general point that grammatical meanings may often reside in the subjectively construed counterparts of conceptual archetypes (i.e. mental operations immanent in their conception). I do not know how consistently grammatical meanings relate to conceptual archetypes in this fashion. Here I can only mention a few examples that come to mind from the realm of nominal structure. There is not time to argue for the characterizations proposed, only to cite them as possible illustrations. The first example is my proposed conceptual characterization of nouns as a general category (Langacker 1987b). As a schematic definition - one

34

Ronald W. Langacker

valid for all instances - I suggest that a noun profiles a thing, abstractly defined. Since nouns refer to so many very different kinds of entities, the definition cannot pertain to any specific conceptual content. Instead, I define a thing in terms of two conceptual operations: the grouping of constitutive entities (which need not be individually salient), and reification, i.e. the manipulation of a group as a unitary entity for higher level cognitive purposes. These mental operations are fairly apparent in nouns like group, flock, stack, and team, which clearly do comprise constitutive elements conceptually treated as a single element for higher level purposes. More speculatively, I propose that these same operations are immanent in the apprehension of physical objects, a conceptual archetype which functions as the noun-class prototype. In the case of physical objects, the grouping and reification occurs so automatically (presumably at a lower level of processing) that we are generally not even aware of constitutive entities. If I am correct, the noun class schema consists in the subjectively construed mental operations immanent in the very apprehension of a physical object. My remaining examples pertain to nominal grounding (Langacker 2002a), i.e. definite and indefinite determiners. In a very preliminary vein, I would like to suggest that definiteness bears a special relationship to the common everyday occurrence of physically pointing to something, certainly an archetypal event. Demonstratives, of course, are frequently accompanied by a physical pointing gesture [-»], as in (11): (11)

This [ —>] one is more expensive than that [ —>] one.

I propose to take quite seriously the notion, entertained (I am sure) by many (e.g. Kirsner 1993), that demonstratives involve a mental pointing gesture even in the absence of a physical one, and that some attenuated form of mental pointing is characteristic of definite articles and definiteness in general. While this proposal may be vague, it is not intrinsically mysterious. In physically pointing to something, we both look at it and direct our interlocutor to do so as well. And in looking at something, we focus our attention on it. Thus, at the motor, visual, and mental levels, operations occur which have the effect of selecting or singling out a particular entity as the focused object of conception. The schematic notion of definiteness, common to all its manifestations, might then reside in the mental aspect of these operations, in abstraction from the visual and motor

Subjectification,

grammaticization,

and conceptual archetypes

35

levels. This mental process of selection and focusing attention is of course immanent in making or apprehending a physical pointing gesture. As for indefinites, I will simply note that certain grounding quantifiers lend themselves to descriptions based on Active (imagined) occurrences of archetypal events (Langacker 1991: 3.2, 2003a, 2003b). The proportional quantifiers all and most reflect the everyday experience of laying one object on top of another to compare their size. All corresponds to the experience of finding that the boundaries of the object being assessed "reach" those of the reference object (i.e. the boundaries coincide), and most to the case where its boundaries merely approximate those of the reference object. Each and any are representative instance quantifiers. Each reflects the basic experience of examining a set of objects sequentially, one by one, and any, that of choosing randomly from a collection. (12)

a. All linguistic theorists are quarrelsome. [comparison by superimposition; reaching] b. Most linguistic theorists are quarrelsome. [comparison by superimposition; nearing] c. Each linguistic theorist is quarrelsome. [sequential examination] d. Any linguistic theorist is quarrelsome. [random selection]

I am not saying that in using these quantifiers one is specifically thinking of an act of superimposition, sequential examination, or random selection. That would be to construe them objectively, to put them onstage as conscious objects of conception. The suggestion, rather, is that mental operations immanent in the conception of these actions are also inherent in apprehending the import of the quantifiers. I should note that these quantifiers, as well as the definite determiners, are grounding elements, constituting the grammaticized means for relating a nominal referent to the speech situation. (Likewise, tense and modals ground an English finite clause.) Elsewhere (e.g. Langacker 2002a) I have shown that a grounding element profiles only the grounded entity (the nominal referent). The grounding relationship and the speech situation (including the subject of conception) are offstage and subjectively construed. It follows that the operations mentioned in (12) do not inhere in the onstage content of the nominal expressions. Rather, they inhere in the very conceptualizing activity by means of which the interlocutors select a nominal referent.

36

Ronald W. Langacker

Finally, I should address the issue of how the kinds of subjectively construed mental operations I have been discussing relate to the notion of image schemas (Johnson 1987; Lakoff 1987). Despite its manifest insight and importance, I have always felt that this notion is vaguely defined and unclear in application. There is, however, no intrinsic conflict between the proposals made here and an image schematic account. If anything, the kind of subjective processing I contemplate has greater generality. For instance, the process of mentally progressing from a reference point to a target is not the sort of thing usually cited as an image schema. Nor is the act of directing attention or making a random choice. Still, I can readily imagine these being referred to as image schemas, and I would not object to this. The important thing is to specify in as much detail as possible precisely what sorts of mental operations are involved, however we label them. In this regard, I note that both accounts emphasize the dynamicity of conceptual structure (Langacker 2001). In my terms, the apprehension of subjectively construed entities resides in various kinds of processing activity, such as mental scanning, evoking a reference point to reach a target, the directing of attention, and the effort experienced in mental extrapolation. Likewise, Johnson (1987: 25-30) portrays image schemas in dynamic terms as mental operations. He says that image schemas involve operations that are analogous to spatial manipulation, orientation, and movement... A schema is a recurrent pattern, shape, and regularity in, or of, these ongoing ordering activities ... I am identifying the schema as a continuous structure of an organizing activity ... They are a primary means by which we construct or constitute order and are not mere passive receptacles into which experience is poured ...

In a similar vein, I have sometimes spoken of image schematic abilities (e.g. 1993a: 3). What would it mean to identify the mental operations I have been discussing with image schemas? It would mean that we should not think of image schemas as something we conceptualize (which the term image might suggest), but as cognitive abilities inherent in the conception of other entities. For instance, the source-path-goal image schema could be thought of instead as the capacity for mental scanning. The link schema could be thought of as the capacity to exploit a conceptual connection. The centerperiphery schema might be thought of as an asymmetry in mental access, adding directionality to the exploitation of a link. If this makes any sense, Mandler's (1991) notion of perceptual analysis - the redescription of sensorimotor experience in image schematic form - could be explicated as

Subjectification, grammaticization, and conceptual archetypes

37

the apprehension of primary experience by means of such processing capacities (presumably inborn). This may be what Johnson intended all along. If so, being explicit about the matter would at least constitute a significant clarification. The conceptual entities I have been discussing are offstage and subjectively construed. They reside in conceptualizing activity immanent in the apprehension of onstage elements, but they are not themselves focused objects of conception. In Johnson's words, they are regularities in or of an ongoing organizing activity. For this reason we are not aware of them per se. We experience them only indirectly, through their effect in shaping our conception of entities construed objectively. Whether image schemas should indeed be identified with subjective processing will have to be resolved by further investigation by all concerned. The pivotal issue is whether image schemas should be thought of as abstract objects of conception or as structuring activity immanent in the conception of other entities. In any case, if meaning resides in conceptualization, we can hardly avoid positing subjectively construed entities. A viable conceptual semantics has to recognize both the subject and the object of conception. By its very nature, however, the subject of conception is not itself conceived.

References Barsalou, Lawrence W. 1999 Perceptual symbol systems. Behavioral and Brain Sciences 22. 577-660. Fauconnier, Gilles, and Mark Turner 2002 The Way We Think: Conceptual Blending and the Mind's Hidden Complexities. New York: Basic Books. Hawkins, Bruce W. 1984 The Semantics of English Spatial Prepositions. San Diego: University of California doctoral dissertation. Heine, Bernd 1997 Cognitive Foundations of Grammar. New York and Oxford: Oxford University Press.

38

Ronald W. Langacker

Johnson, Mark 1987 The Body in the Mind: The Bodily Basis of Meaning, Imagination, and Reason. Chicago and London: University of Chicago Press. Kirsner, Robert S. 1993 From meaning to message in two theories: Cognitive and saussurean views of the Modern Dutch demonstratives. In Richard A. Geiger and Brygida Rudzka-Ostyn (eds.), Conceptualizations and Mental Processing in Language, 81-114. (Cognitive Linguistics Research 3) Berlin/New York: Mouton de Gruyter. Kumashiro, Toshiyuki 2000 The Conceptual Basis of Grammar: A Cognitive Approach to Japanese Clausal Structure. San Diego: University of California doctoral dissertation. Kumashiro, Toshiyuki, and Ronald W. Langacker 2003 Double subject and complex predicate constructions. Cognitive Linguistics 14. 1^-5. Lakoff, George 1987 Women, Fire, and Dangerous Things: What Categories Reveal About the Mind. Chicago and London: University of Chicago Press. Langacker, Ronald W. 1985 Observations and speculations on subjectivity. In John Haiman (ed.), Iconicity in Syntax, 109-150. (Typological Studies in Language 6) Amsterdam/Philadelphia: John Benjamins. 1986 Abstract motion. Proceedings of the Annual Meeting of the Berkeley Linguistics Society 12. 455-471. 1987 a Foundations of Cognitive Grammar, vol. 1, Theoretical Prerequisites. Stanford: Stanford University Press. 1987 b Nouns and verbs. Language 63. 53-94. 1990 Subjedification. Cognitive Linguistics 1. 5-38. 1991 Foundations of Cognitive Grammar, vol. 2, Descriptive Application. Stanford: Stanford University Press. 1993 a Reference-point constructions. Cognitive Linguistics 4. 1-38. 1993 b Grammatical traces of some "Invisible" semantic constructs. Language Sciences 15. 323-355. 1995 a Raising and transparency. Language 71. 1-62. 1995 b Possession and possessive constructions. In John R. Taylor and Robert E. MacLaury (eds.), Language and the Cognitive Construal of the World, 51-79. (Trends in Linguistics Studies and Monographs 82) Berlin/New York: Mouton de Gruyter.

Subjedification, grammaticization, and conceptual archetypes Consciousness, construal, and subjectivity. In Maxim I. Stamenov (ed.), Language Structure, Discourse and the Access to Consciousness, 49-75. (Advances in Consciousness Research 12) Amsterdam/Philadelphia: John Benjamins. On subjectification and grammaticization. In Jean-Pierre Koenig 1998 (ed.), Discourse and Cognition: Bridging the Gap, 71-89. Stanford: CSLI Publications. Losing control: Grammaticization, subjectification, and transpar1999 ency. In Andreas Blank and Peter Koch (eds.), Historical Semantics and Cognition, 147-175. (Cognitive Linguistics Research 13) Berlin/New York: Mouton de Gruyter. Dynamicity in grammar. Axiomathes 12. 7-33. 2001 Deixis and subjectivity. In Frank Brisard (ed.), Grounding: The Epistemic Footing of Deixis and Reference, 1-28. (Cognitive 2002 a Linguistics Research 21) Berlin/New York: Mouton de Gruyter. A Study in unified diversity: English and Mixtec locatives. In N. J. Enfield (ed.), Ethnosyntax: Explorations in Grammar and 2002 b Culture, 138-161. Oxford/London: Oxford University Press. Dynamicity, Activity, and scanning: The imaginative basis of logic and linguistic meaning. Korean Linguistics 18. 1-64. 2003 a One any. Korean Linguistics 18. 65-105. 2003 b Strategies of clausal possession. International Journal of English 2003 c Studies 3.2. 1-24. Aspects of the grammar of finite clauses. In Michel Achard and 2004 Suzanne Kemmer (eds.), Language, Culture and Mind, 535-577. Stanford: CSLI Publications. Mandler, Jean M. 1991 Prelinguistic primitives. Proceedings of the Annual Meeting of the Berkeley Linguistics Society 17. 414-425. Matsumoto, Yo 1996 Subjective motion and English and Japanese verbs. Cognitive Linguistics7. 183-226. Miller, George A. and Philip N. Johnson-Laird 1976 Language and Perception. Cambridge, MA: Harvard University Press/Belknap. Rubba, Jo 1994 Grammaticization as semantic change: A case study of preposition development. In William Pagliuca (ed.), Perspectives on Grammaticalization, 81-101. (Current Issues in Linguistic Theory 109) Amsterdam/Philadelphia: John Benjamins. 1997

39

40

Ronald W. Langacker

Sweetser, Eve Ε. 1982 Root and epistemic modals: Causality in two worlds. Proceedings of the Annual Meeting of the Berkeley Linguistics Society 8. 484507. 1988 Grammaticalization and semantic bleaching. Proceedings of the Annual Meeting of the Berkeley Linguistics Society 14. 389-405. Talmy, Leonard 1988 Force dynamics in language and cognition. Cognitive Science 12. 49-100. 1996 Fictive motion in language and "Ception". In Paul Bloom et al. (eds.), Language and Space, 211-276. Cambridge, MA and London: MIT Press/Bradford. Taylor, John R. 1996 Possessives in English: An Exploration in Cognitive Grammar. Oxford: Oxford University Press/Clarendon. Traugott, Elizabeth 1982 From propositional to textual and expressive meanings: Some semantic-pragmatic aspects of grammaticalization. In Winfred P. Lehmann and Yakov Malkiel (eds.), Perspectives on Historical Linguistics, 245-271. Amsterdam/Philadelphia: John Benjamins. 1988 Pragmatic Strengthening and Grammaticalization. Proceedings of the Annual Meeting of the Berkeley Linguistics Society 14. 406416. 1989 On the rise of epistemic meanings in English: An Example of Subjectification in Semantic Change. Language 65. 31-55.

Logic, subjectivity, and the semantics/pragmatics distinction Frank Brisard *

1.

Introduction

This paper deals with two separate yet related problems in semantic theory, and in the study of meaning change. It confronts Traugott's conception of "subjectivity", and of the processes of grammaticalization that lead to the establishment of subjective meanings, with received views on this topic in the paradigm of Cognitive Semantics (CS). It is argued that neither Traugott's approach to subjectification, nor her account of the "pragmatic strengthening" involved in its earliest stages (section 2), lend themselves easily to a straightforward implementation in cognitive terms. The problems, of subjectivity and of pragmatic strengthening, are related, at least in the outstanding functionalist and cognitive-linguistic literature, mainly because of Traugott's highly influential views in this respect. The first problem relates to the status of subjectivity in semantics, and particularly in the semantics of grammar. I will argue that Traugott's position in this, while semantically motivated, takes too much of a psychological turn, where the notion of "subject" is substantiated almost exclusively in relation to the actual speaker as a person, and to acts of selfexpression (section 3.1). In addition, it does not pay sufficient attention to the formal properties of certain highly grammaticalized elements, and confounds what specific linguistic expressions are actually about with what is needed to explicate their conceptual background (section 3.2). This last point was explicitly addressed by Langacker (1997: 70), whose approach to subjectivity is argued here to be preferable to Traugott's in many important respects. This is a very difficult issue, because there are obvious points of agreement between both authors, and Langacker (1999a: 150) has at one point suggested that the disagreement is terminological only, while

*

The author is a Postdoctoral Fellow of the Fund for Scientific Research Flanders (Belgium).

42

Frank Brisard

Traugott and Dasher (2002: 97-99) maintain that there is a significant difference. I will also try to demonstrate how Traugott's story of subjectivity crucially relies on a conception of pragmatics, as a separate domain of linguistic analysis, that seems to clash with the conceptualist and encyclopedic assumptions of standard CS. Traugott, in fact, conjures up a strictly modular notion of pragmatics to help her explain the transition from propositional and textual to expressive or nonreferential meanings that appears typical of grammaticalization (section 4.1). The increased expressivity/informativeness that results from this transition is then directly and exclusively related to the speaker's (explicit) stance, and thus to a rather restricted reading of linguistic subjectivity. The formalist streak in Traugott's work can be seen in her deployment of the term "implicature", as well as in the underlying adoption of a false dichotomy between a speaker-external, logical semantics and an internal pragmatics. In CS, in contrast, there is no inherent opposition between the linguistic function of expressing information and that of regulating communication. It is the latter function that is traditionally treated as theoretically marginal or derivative, but which surfaces in CS in the form of "construal", a central semantic notion that acknowledges the import of strategy in describing acts of meaning. I will finally suggest that some conception of strategy, or the exploitation of shared expectations, may serve as a nonarbitrary basis for defining the scope of pragmatics (section 4.2).

2.

The setting: Early processes of grammaticalization

Historically, Traugott's empirical work on semantic change may count as an important contribution to the analysis of polysemy.1 Her focus on the initial stages of subjectification reveals a plethora of dynamic and interactive patterns of behavior on the part of the (mostly grammatical) expressions at stake. The need to show systematically attested connections among meanings is far from evident in the theoretical climate of the time, the 1. In the early 1980s, one of the few contemporaneous endeavors comparable to Traugott's basic research program was Sweetser's (e.g., 1984) inquiry into the structural polysemy of expressions ranging over various worlds of reference, including the propositional, the epistemic, and the level of the speech act. It is not entirely clear where the dividing line between semantics and pragmatics, if any, should fall in Sweetser's work.

Logic, subjectivity, and the semantics/pragmatics

distinction

43

1980s, and the extant practice of representing ambiguous expressions in the form of discrete categories was accordingly perceived by functionalists and cognitivists alike as less than satisfactory. Moreover, in contrast to the relative lack of interest shown by much of formal semantics in issues of diachronic change, the mechanisms of polysemy cannot be fully understood without the assumption of a historical stage in which meaning categories that can be attributed to one form also overlap synchronically, such that "one meaning influences the other as they coexist in time" (Traugott 1989: 33). This stage, in any case, must be situated before "bleaching" occurs, when new meanings become more or less automatic choices, or routines. Traugott (1988) distinguishes two general types of process involved in early meaning change. One type covers processes of generalization over contexts, where grammatical restrictions are progressively loosened, as well as instances of metaphorization (e.g., the spatial marking of temporal categories). Both of these mechanisms are in principle semantic, according to Traugott, insofar as no account of pragmatic inferencing is needed to describe them. Metaphor, in fact, is explicitly portrayed as a mechanism of "semantic transfer through a similarity of sense perceptions" (Traugott 1988: 411; see also Traugott and König 1991), in spite of Traugott's simultaneously claiming an effect of strengthening (as opposed to bleaching), and thus some kind of pragmatic status, for metaphor. Now, it should be clear here that Traugott favors micro-analyses of subjectification in terms of "(re)alignment", rather than the coarser mechanism of bleaching that may go on in subsequent stages, and that for her the standard picture of metaphorical relations between polysemous meanings poses too few constraints in this regard - cf. Langacker (1997: 66) for a similar critique of metaphor as a tool of analysis for raising constructions and other cases of semantic "transparency". At the same time, however, Traugott keeps insisting on a separate treatment of metaphor qua informative strengthening, which she contrasts with metonymic implicatures. The latter are merely associative and not based on similarity, but on (perceptual) contiguity, and we should therefore conclude that metaphor as such does not give rise to implicature in Traugott's eyes. An implicature, then, is surely not just any type of (context-induced) inference, at least as Traugott uses the term, but it is difficult to reconstruct the rationale behind this decision (cf. section 4.1). We should also note in passing that insofar as conventionalized linguistic metaphors are based on "conceptual metaphors" (e.g., TIME IS SPACE; Lakoff and Johnson 1980), they are not at all pragmatic from a

44

Frank Brisard

cognitive point of view. This goes against pragmatic accounts of metaphor, as described for instance in Levinson (1983), and any a-priori conception of literal meaning that may be implicit in them.2 And so it is not evident how the "semantic" qualities of a cognitive theory of metaphor should be reconciled with the "pragmatic" orientation of a Gricean explanation. Yet Traugott cites both paradigms approvingly and seems to presuppose a unified basis for them. This is the first problem we encounter in examining Traugott's theoretical position, one to which I shall devote considerably less attention in the present paper (but see section 4 on the analytical use of implicature). The second type of process involved in meaning change essentially depends on acts of (initially defeasible) inferencing, usually at the level of metonymy (see also Dahl 1985 and Taylor 1989 for parallel examples of the conventionalization of implicatures in morphology and syntax, as well as several contributions, including the theoretical introduction, to Panther and Thornburg 2003). This kind of inference is inherently characterized as pragmatic by Traugott and is deemed more complex than the processes of metaphorization and generalization. Pragmatic inferencing is supposed to lie at the heart of the development of connectives {while), scalar particles (very, mere, just), evidentials, etc. from their respective lexical origins. It is argued that it produces a strengthening of informativeness or informational relevance, whereby the hearer is invited to infer more than what is directly communicated. On the part of the speaker, this kind of strengthening also implies that there is more of a subjective involvement in what is being communicated, i.e., that the speaker chooses to encode her own position or attitude with respect to what is said. The latter category, "what is said", is implicitly taken by Traugott in Grice's (1989: 25) sense of referring to the conventional meaning of words and the sentences they form, even if this reading is not wholly uncontroversial in the ensuing literature. In the following sections, I will deal with two problems which may arise in the confrontation of Traugott's views with a cognitive approach to 2. Hopper and Traugott (1993: 78) refer to Green (1989: 122) to support their claim that "metaphors often involve propositions that are intended to be recognized as literally false [...] but conversational implicatures do not". They proffer this claim "at a superficial level". Again, it is not obvious to me why, in a Gricean framework, metaphors should not be seen as producing conversational implicatures, when Grice (1989: 34) clearly does grant them this status. Moreover, there are certainly conversational implicatures to be found that are not metaphors yet that are not literally true, either (irony, hyperbole, etc.).

Logic, subjectivity, and the semantics/pragmatics

distinction

45

meaning. One, discussed in section 3, concerns the status and substance of subjectivity as the explicit expression of grounding in the speaker's perspective (Traugott and Dasher 2002: 6). The other scrutinizes Traugott's account of "invited inferencing" against the backdrop of a truly usagebased model of CS (section 4). Invited inferences, Traugott's theory claims, move from being properties of discourse understanding to being properties of encoded meaning.

3.

Subjectivity

3.1.

Speaker and linguistic perspective

Subjectivity in epistemic modality can be regarded as a gradable notion of which several subregions motivate a cluster of related analytical uses of the term. There is a real sense in which linguistic expressions can be analyzed as indicating more or less subjective meanings. Largely, the two ends of this cline are identified as representing "weak" vs. "strong" subjectivity, with the former pertaining to an assessment of reality with minimal intervention from the speaker, and the latter referring to the speaker's own beliefs (about some state of affairs obtaining in reality). Lyons (1982) uses "objective" vs. "subjective" modality to mark the same distinction, whereby we may observe that with objective modality, the speaker is wholly committed to the factuality of the information she is providing Lyons (1977) calls this the "I-say-so component" of an assertion - and is therefore not to be construed as simply absent from the conceptual background.3 Thus, the epistemic interpretation of example (1) is taken to be ambiguous between a strong and a weak subjective reading (cf. also Traugott 1989: 36):

3. Accordingly, objective statements can be denied, questioned, accepted as facts by the hearer, hypothesized in a real conditional, and embedded under factive predicates, which is by definition awkward with subjective ones. Lyons (1977: 797) points out that the difference between objective (weak) and subjective (strong) epistemic modality "is not a distinction that can be drawn sharply in the everyday use of language; and its epistemological justification is, to say the least, uncertain". I tend to agree with this statement and intend to show that the same holds for Traugott's treatment of the contrast, i.e., that it cannot be clarified from the explicit form of an utterance.

46

(1)

Frank Brisard

You must be very careful. (Lyons 1982: 109) a. 'It is obvious from evidence that you are very careful.' (weakly subjective) b. Ί conclude that you are very careful.' (strongly subjective)

Traugott presents this contrast primarily in terms of the instance that is deemed (by the speaker, presumably) to have the authority to warrant the epistemic status of the proposition at issue, with the appropriate modal force. In the case of (lb), it is the speaker herself who assumes this role, though implicitly so (for she is nowhere mentioned in the original utterance). In (la), the authority comes from an impersonal it, which may be interpreted as representing some setting relative to which an event is being construed, rather than one of the participants in this event.4 In the case of (la), too, this information remains implicit. In Langacker's (e.g., 1991b, 1999c) terms, the contrast relates to the "locus of potency" that is assumed for each usage. In fact, in moving from strictly deontic or dynamic ("root") modal uses to epistemic ones, the interpretation of this locus presents itself as the major semantic change involved. The difference between a deontic reading of (1) and an epistemic one resides in the alignment of the force that is targeted at the propositional content, which is itself construed as potential in both readings: a potential real-world fact in the case of a deontic use, and a potentially valid conclusion (following some inference chain) with an epistemic use. For deontics, that force is no longer located in the "objective scene" itself, as it was with the original, lexical uses of such modal verbs, but should be associated with some aspect of the ground (possibly the speaker), such that it is this very ground which is conceptualized as exerting the social force necessary for the actualization of the desired state of affairs. Epistemics take this process of subjectification one step further and locate the force at issue in the "evolutionary momentum" of reality itself, as it is assessed by the speaker. The modal force relevant to these uses is one where the speaker indicates the relative ease or difficulty with which she can "mentally ex4. One could describe this type of "active-zone" phenomenon (Langacker 1984) as a case of metonymy in its own right, whereby the subordinate that clause, the actual focus of the proposition, is metonymically represented by the grammatical subject. We should not forget, however, that we are dealing with a paraphrase here, and that it is not very revelatory to have aspects of the actual analysis of these examples depend on the relatively arbitrary grammatical form of the paraphrase (cf. Section 3.2).

Logic, subjectivity, and the semantics/pragmatics distinction

47

trapolate" the envisioned event, given what she knows about ongoing reality. However, it is not clear how, within these epistemic uses, a further strict differentiation can be made between weakly or strongly subjective meanings, except to say that the source of potency can sometimes be construed in a maximally diffuse and impersonal way, referring to "some nebulous, generalized authority" (Langacker 1999c: 308). It is exactly these cases which, counter to the claims made by Traugott regarding weak subjectivity, may arguably be regarded as extremely (strongly) subjectified. Traugott's emphasis in this on the speaker, as a psychological entity central to the notion of subjectivity, comes from a process-oriented view, in the psycholinguistic sense of online processing, of meaning change. 5 The move to more subjective meanings is typically meant to convey that an expression becomes "more anchored in the context of the speech act, particularly the speaker's orientation to situation, text, and interpersonal relations" (Traugott 1982: 253). On the whole, the resulting configurations can indeed be qualified as involving nonreferential or "procedural" (cf. Blakemore 1987) elements of meaning that may justifiably be called more pragmatic than their objective, "declarative" counterparts, but the point remains that even these highly pragmatic elements need not be substantively psychological in their conceptual import. That is to say that the pragmatic meanings that Traugott identifies as typifying subjectification are nevertheless not automatically to be associated with the speaker as a real and actual person, but rather with the evocation of a subjective conceptualizer. The position of this conceptualizer or " s e l f ' may naturally correspond with the speaker's own epistemic stance, possibly by default, but this particular association can be overridden at any time, given that the essence of linguistic subjectivity seems to be about expressing a point of view (not necessarily the speaker's, though). Moreover, inasmuch as the construal of maximal subjectivity involves a total loss of self-awareness on the part of the conceptualizer, who is focusing on the objective target of conceptualization, we cannot maintain that subjectification entails an increase in the explicitness with which the conceptualizer is linguistically represented, even if that role is to be equated with the speaker. Thus, we are faced with two directly conflicting suppositions about the effects of subjectification: one where the speaker assumes a central position that is taken to be explic-

5. Traugott does not, of course, mean to imply that meaning change itself is necessarily personal in the sense of "individualized", where one speaker should initiate such changes on her own.

48

Frank Brisard

itly encoded and that is treated as part of an "enriched" propositional content, vs. one where speaker and subject are not always identified with each other, and where they should in any case be conceived as part of an expression's conceptual background (unless one or the other, or both, are also portrayed as objective participants). This is particularly clearly illustrated in the following pair of examples, which Langacker (1991b: 328) uses to discuss the distinction between a subjective and an objective construal of the speaker: (2)

a. Vanessa was sitting across the table. b. Vanessa was sitting across the table from me.

In (2a), the speaker is implicitly present as the conceptual starting point from which the locative relation at issue is to be calculated, whereas in (2b), this same reference-point relation is brought onstage and, accordingly, is objectively construed. For Traugott, examples such as these would not even count as prototypical instantiations of the objective/subjective contrast, and in fact, sentence (2b) would, if anything, have to be much more subjectively than objectively oriented, since it introduces a subjective element (the speaker) explicitly within the proposition. The deeper reason for Traugott's difference of opinion is that her approach, whose specific subject is change of meaning, looks at how much speaker-oriented meaning is part of the explicit content. A major point in her theory is to show that over time, objective meaning - i.e., referential meaning that is a property of the "external described situation" - is replaced by subjective meaning i.e., meaning that describes speaker-internal features of the encoded situation (Traugott and Dasher 2002: 94). It seems that we are certainly in need of an additional distinction over and above the one between "syntactic subject" and "speaker" (as an actual person physically producing an utterance). For this third role, Benveniste (1966; and, largely following him, Langacker 1985) recognized a covert "subject of utterance", for which one could also propose such terms as conceptualizer or "viewer", since it refers to the instance that defines the (conceptual) viewpoint or perspective on a given scene. This type of subject is not psychological in any personalized sense (even when the perspective can be attributed to this or the other actual or fictive person); rather, it pertains to what Lyons (1995) refers to in terms of the "subjectivity of consciousness", as distinct from the "subjectivity of agency" (and in

Logic, subjectivity, and the semantics/pragmatics

distinction

49

particular, from that of speaking).6 It is about speakers accessing "worlds" (including the perception of "immediate reality", i.e., the ground) and referring to them "from the viewpoint of the world that they are in" (Lyons 1995: 341-342). The main reason why Traugott does not distinguish consciousness from the agentivity of speaking, which is also the chief cause of the antagonism vis-a-vis Langacker's proposals in this respect, appears to lie in an overly rationalist approach to semantics, in which an undivided self either operates dispassionately upon a mental proposition (objectivity), or explicitly judges a proposition (subjectivity as a form of tokenreflexivity).7 The apparent inadequacy of truth-functional semantics as a complete theory of meaning, which is the cause of this exclusive vision of linguistic objectivity/subjectivity, is nowadays being taken up through the conceptualist underpinnings of CS (cf. section 4.2), inscribing the notion of perspective, and thus of subjectivity, into the very fabric of both lexical and grammatical meaning. How, then, are we to reconcile the idea that subjectivity correlates with implicitness or at least a lack of conceptual salience with Traugott's claim that the speaker occupies a more prominent role in cases of subjectification? There is no evident way out of this conflict, I suggest, because it arises out of a fundamental clash between a rationalist ("formalist") and an empiricist ("usage-based") approach to language structure and use. Still, it is possible to revalidate Traugott's systematic focus on the speaker in matters of subjectivity, provided that we move to a different level of analysis, as Langacker does. In Cognitive Grammar (CG), the notion of subjectivity always invokes an entity (be it a process or a thing) that is "real" only at a processing level, for which the speaker is the obvious reference point. At this level of mental occurrences, the speaker is the only relevant locus

6. On a tentative note, I would propose that the proper philosophical conception of this kind of linguistic subjectivity be metaphysical, rather than psychological. The essential role of the subject(ive) would then consist in its definition of a "horizon" as the prerequisite to meaning. See Wittgenstein (1963: §§5.63-5.64) for a presentation of the subject as a viewer not unlike Langacker's. 7. While Traugott frequently denies the relevance of a purely objective epistemic modality to the study of natural-language semantics, she remains committed, I believe, to the theoretical position that there can be communicative styles, linguistic varieties, and even whole language systems which are less deeply imbued with subjectivity than others (cf. section 3.2). Such a typology would be completely absent in CS, since this model acknowledges the perspectival nature of meaning at all levels, and in every instance of language use.

50

Frank Brisard

of analysis, as the obvious origo of all acts of meaning. We could call this the neurological component of subjectivity, which I will illustrate below. There is also a phenomenological side to subjectivity, however, which describes aspects of an abstract viewing perspective (including the construction of utterly "virtual" worlds) that may or may not be the speaker's, but that in any case remains backgrounded or unprofiled by virtue of its procedural function.8 From the hearer's point of view, subjectivity in CG deals with the instructions needed to view some conceptual content under the right construal. Crucially, what is profiled (conceptually designated) in the message delivered by the speaker - what the message is invariably about - is the content, not the way in which it is construed. Let us look at an example of objective vs. subjective motion in English, as characterized in CG. Consider utterances (3a) and (3b), involving a directional and a locative use of the preposition across: (3)

a. Vanessa jumped across the table. b. Vanessa was sitting across the table.

In (3a), there is an objective path invoked by the meaning of across that takes an objective participant, the grammatical subject or trajector (tr), from a starting point, 'across' the landmark (lm) object, to an endpoint at the other side. This configuration, including the subjective process of scanning the objective path, is diagrammed in Figure 1(a). The objective path in its entirety describes the actual motion of the entity represented by the trajector (to the extent that we accept the proposition as true). In Figure 1(b), we see the diagram for the objectively static meaning of across in example (3b). There, only the conceptualizer's scanning operation immanent in the dynamic meaning of across is retained, in that it is the speaker as viewer who still performs a sequence of acts of mental scanning on an objectively presented content (starting with a refer-

8. "We can [...] characterize a semantic structure at either the phenomenological level - by describing essential aspects of a conceptualization in any way that proves revelatory - or else with respect to the neurological processing whose occurrence constitutes the mental experience." (Langacker 1991a: 513) While analyses at the two levels are not to be separated but are meant to inform one another (convergence), there are good indications to believe that Traugott's fixation on the speaker betrays a processing perspective that is only partially motivated by detailed phenomenological analysis.

Logic, subjectivity, and the semantics/pragmatics (a)

across (objective path)

(b)

distinction

51

across (subjective scanning)

lm

ψ tr

W \

\

W 1

/

. \ 1 / / \—X—ι—ι —

/

\\ I / / \\|//

δ

Figure 1. Objective and subjective construal (Langacker 2002a: 21)

ence point R, assessed in relation to the ground G). Accordingly, in the subjectified uses the motion component that is otherwise objectively profiled is lost: instead of the trajector construed as controlling the motion in Figure 1(a), it is the speaker who mentally controls the scanning process in Figure 1(b). In the latter case, the whole subjective motion component, with the speaker occupying a central role in it, is not profiled, i.e., it does not deserve enough attention to be encoded. As there are no dynamic elements as such left in the objective scene, in contrast with directional uses, the result, in terms of what is profiled, is a static scene. While in processing terms the speaker, as a subjective element, might be said to become more actively, or at least more centrally, involved in the construction of a meaning, in terms of reference and interpretation the element that is subjective gets backgrounded and attenuated.

3.2.

Complications and objections

Certainly, Traugott and Langacker share a keen interest in fairly subtle matters of semantic realignment, which seems to occur before, and might actually be a prerequisite for, the more patently visible changes in meaning that have typically been observed in historical studies. Such realignment tends to take place from features of the situation to features of the speaker's or some other party's position, and that makes it tempting to see the two theories as directly comparable. And hence at odds. Let us remember, though, that Traugott's first concern goes to whether expressions historically move towards an increasingly speaker-oriented content, and that

52

Frank Brisard

she can therefore rightly dismiss cases like in (2a/b) as largely irrelevant to her own goals. The least one could state, in response to this position, is that her theory simply does not capture the dimension of subjectivity that is Langacker's main interest. What further divides Langacker and Traugott in their discussion of subjectivity is that only the former, in describing semantic structure, seems to pay close enough attention to language form and to properties of grammar which, even if functionally motivated, may not be entirely predictable on the basis of choices (made by speakers) alone (Langacker 1999a: 18ff). Traugott's focus on the speaker as the main actor in subjectivity is perhaps warranted, but partially directed at the wrong level of analysis and not sufficiently backed by a formal means of representing linguistic meaning. If we return to example (1), and specifically to the paraphrases cited by Traugott, we are bound to notice that these can only be rough approximations of the two meanings of (1). For one, the paraphrases do not even indicate which units are to be associated with the various formal elements that constitute sentence (1). In addition, they provide us with no information as to the relative prominence that should be given to the different composing units (what is focused on? what is backgrounded? etc.). In short, this type of paraphrase may be useful in determining semantic ambiguities, but it does not constitute a valid tool of grammatical analysis, if it is the only one referred to. The problem is identical to Langacker's objections against Sweetser's (1990: 61) rendition of the meanings of various English modals. While these verbs are certainly force-dynamic in the way that Sweetser portrays them, including the epistemic use in (4), they cannot be meant to convey that the first part of their paraphrase (the main clause) should receive as much attention as the proposition that it qualifies: (4)

You must have been home last night. 'The available (direct) evidence compels me to the conclusion that you were home.'

As Langacker (1997: 70) notes, "this is problematic because the characteristic feature of grammaticized modals is precisely that the speech event participants and the modal force are offstage and subjectively construed". As a general point of methodology, we might add Lyons' (1982: 110) cautionary remark "that subjectively modalized utterances cannot be translated or paraphrased by means of unmodalized declaratives and that any attempt to do so (without making use of intonation or optional modal particles

Logic, subjectivity, and the semantics/pragmatics

distinction

53

which carry over the subjectivity, more or less satisfactorily) has the effect of adding to, or subtracting from, what is said".9 If, in epistemic terms, a single, wholly or nearly completely, objective utterance may be difficult, if not impossible, to find,10 there can be no wholly objective language as a rule, and not even as a particular stylistic variety. Such an objective code would require that the speaker is nowhere to be found "onstage" or "offstage" in the complex of meaning configurations that need to be constructed for each usage event. Any typology of natural languages, moreover, in which languages become increasingly (not cyclically!) "pragmatic" (see, e.g., Lyons 1995: 341) would have to presuppose that "true" semantic and grammatical functions are not originally pragmatic in any conceivable sense. But if pragmatics, dealing with the demands of communication, is inscribed into the principles of language use from the outset, as in a usage-based model, then semantics and grammar can be construed as the structural (nonautonomous) reflexes of these demands, and there can only be shifts in how languages, or one language at different points in time, manifest pragmatic motivation at various levels of organization. This would not make them more or less pragmatic, only more or less dependent on diverse (implicit, explicit) types of "conceptual grouping" (Langacker 1999b), in which languages may and do indeed differ typologically. Now, regardless of which linguistic typology one wishes to propose in this context, it cannot have anything to do with the presence vs. absence of the speaker's involvement, because, next to being the mouthpiece for her own utterances, the speaker as an "author" is also always involved in the construction of a perspective on what she says. This cannot be avoided - subjectivity resides in the use of every nominal and finite clause and is thus just a matter of grammatical habit - , and so it is not as if the speaker is investing much personal energy in her use of, for instance, a deictic tense marker (which may have subjective uses in both Traugott's and Langacker's accounts). It is not clear, moreover, in which sense such a tense marker might be more subjective than, say, a non-

9. See also Pelyvas (1996: 150ff) for an indication of how elements related to the speaker in so-called subjective usage types must already be immanent in earlier, more objective uses as well, suggesting that this cannot be what differentiates an utterance like (4) from its paraphrase. 10. Such an expression could in any case not be "grounded" and thus not occur in any reasonable stretch of ordinary discourse, for grounding always involves subjectivity in CG.

54

Frank Brisard

grounding aspect marker (e.g., English -ing), if we are to stick to Traugott's thesis of subjectivity as personal self-expression. This is not to deny that self-expression by the speaker motivates many constructions in language, but we should beware of linking this type of expressivity too easily to the core workings of grammar. Rather, all linguistic reflexes of subjectivity seem to be about the "ground", in Langacker's sense, and only some of them can be said to concentrate on the specific position of the speaker within that ground. What is more, from a CG perspective those predications that incorporate a general, impersonal notion of the ground can reasonably be taken as more, not less, subjectified than their expressly speaker-oriented counterparts (cf. above). Some of the specifics of examples given by Traugott and her colleagues themselves already point to the nonpersonal character of more evolved manifestations of subjectivity. Thus, Traugott (1995: 39) acknowledges that the loss of referential properties linked to the subject referent of (parenthetical) I think paves the way for the use of this construction to express a perspective, not necessarily the speaker's. I would not consider this an atypical or marked case, in contrast with Traugott's appreciation of the same example (in terms of a "realignment of the syntactic subject"). Similarly, Traugott and Dasher's (2002: 96) description of the "humiliative honofiric" uses of the Japanese verb agaru, meaning 'visit the place of one's/the speaker's superior', focuses excessively on the personal nature of the choices associated with such uses. In general, though, it would seem that such grammatical types as honorifics, precisely by virtue of their highly conventionalized status, do not call upon the speaker's individual decision to recognize or indicate someone else's symbolic position in some particular context, but instead bestow this position on the basis of social rules and norms (whether or not the speaker actually identifies with these on each and every occasion). By themselves, positing a typology of languages based on subjectivity, and studying the processes of change that may lead a language from one type to another, should not be problematic, if the relevant concepts are sufficiently clarified. The scientific merits of positing such typologies and then associating them with moral predicates (even if they are not the author's, who is merely mentioning them) are much less obvious, though. Traugott and Dasher's (2002: 3Iff) discussion of "objectification" as a language-external, deliberate intervention is particularly challenging in this regard. We are considerably removed from any strict grammatical conception of the subjectivity/objectivity contrast, if an objective language is as-

Logic, subjectivity, and the semantics/pragmatics

distinction

55

sociated with products of technological and cultural progress, such as science and education. We are definitely not talking about grammar, if objectivity is described as nonconnotative, neutral, transparent, even morally superior (because there is no concealment). Once more, these are not Traugott and Dasher's own terms, and they, for one, clearly position themselves on the side of "natural" linguistic subjectivity. Yet such partisanships, whichever side they may be on, are not to be considered in the first place, because the domains where these very broad uses of "objectivity" and "subjectivity" would actually make sense, are not situated anywhere in the study of grammar, but rather in literary and cultural studies and related disciplines. In the end, both Langacker and Traugott acknowledge the central import of subjectivity to grammar and lexicon. They probably also agree on the typically abstract nature of subjective meanings, comprising epistemic meaning types and, more generally, nonreferential or "expressive" elements of semantic construal. At the same time, their very real disagreements on a number of other, theoretically and methodologically significant aspects of the study of linguistic subjectivity appear virtually insurmountable, because, as I will argue next, they involve conflicting orientations towards even such basic analytical constructs as "context" and relevant levels of linguistic choice-making. Traugott and Dasher (2002: 97-99) list a series of objections to Langacker's treatment of subjectivity that is supposed to illustrate this rift. They include the following: -

-

Langacker's examples are usually constructed out of context and disregard situated qualities of linguistic style and genre; they focus on "choices" of form (e.g., syntactic subject), rather than content; they deal with event (and argument) structures of situation types, which is again a formal concern - as manifested in, e.g., the analysis of "raising" constructions; CG is rather silent on the matter of discursive "intersubjectivity";11 and

11. See Traugott and Dasher (2002: 40): there can be no non-truthconditional intersubjedification, i.e., the signaling of the hearer's (or of the hearer's and speaker's) point of view, without some prior degree of subjectification. Nuyts (2001) is a cognitive(-pragmatic) study of epistemic modality that makes extensive use of this notion of intersubjectivity as well, and defines it as relating to evidence that is accessible to a larger group of people sharing the same conclusions as the speaker. This is, however, a very consensual view of intersubjectiv-

56

-

Frank Brisard

it does not methodically zoom in on the historical mechanisms of semantic change (even if these can be seen as subsumed under CG's more general formulation of subjectivity).

Still, it should be clear that most of these objections do not primarily target the analysis of subjectivity in CG, but the very apparatus of that theory itself. It should be evident from this and the previous discussion that, for Traugott, Langacker's conception of subjectivity is too narrow in scope, as it only looks at a subset of purported subjectification processes, and that his conception of subjectification as attenuation should contradict her own critique of diachronic analyses of early semantic change in terms of a loss of meaning (and not a strengthening, as Traugott proposes). Problems like these cannot be remedied by adjusting a few terms or tinkering with definitions. In fact, I suggest that the rift only grows as we go into more elementary (pre-)theoretical assumptions and details of methodological orientation, both effectively separating the two models.

4.

Semantics and pragmatics

4.1.

Formal pragmatics

I will first demonstrate, in this section, that the subjectivity debate between Langacker and Traugott potentially triggers two very different notions of "pragmatics". On one hand, subjective expressions may have a pragmatic function to the extent that they symbolize those meanings that are typically called expressive, including attitudinal and affective types. Subjectivity, insofar as it also deals with epistemic modality, arguably comprises quite abstract meanings that are not so much about propositional contents, but about ways of presenting them, and which might be called pragmatic in a ity that does not seem to acknowledge elemental forms of disagreement or epistemic conflict (and their relevance to intersubjectivity). It would seem, furthermore, that the linguistic domain of intersubjectivity is not entirely constituted by the mere "adding up" of the opinions and beliefs of speaker, hearer, and other parties to a discourse; issues of face management and politeness, marking many instances of linguistic intersubjectivity, cannot be reduced to such quantitative interests alone. Importantly, both authors also seem to presuppose a contrast between strategic discourse and some conceptual structure that is properly decontextualized (see, e.g., Nuyts 2002: 452ff, and Traugott and Dasher 2002: 98), a claim that would be decidedly unwarranted in CG.

Logic, subjectivity, and the semantics/pragmatics distinction

57

broad sense as well.12 On the other hand, many instances of (grammatical) subjectivity crucially relate to indexicality, a domain traditionally relegated to pragmatics by formal semanticists. This style of analysis, however, mostly starts from explicit encodings of contextual features, calling them pragmatic but for reasons that may matter solely within a formalist framework.13 A formalist will call any meaning pragmatic that is not logical (or logically related, through entailment, to some literal content), and that is exactly why implicatures belong in this category. I will use the term "logic" to evoke the complex of truth-functional approaches to semantics in the predicate calculus, as well as quantificational theories of (definite) description, such as Russell's, together with the theoretical variants, elaborations, and challenges that have been proposed by analytic philosophers and linguists in the course of the past century or so. In Traugott's (e.g., 1989: 49; Traugott and König 1991: 194ff) account, subjectification implies an increase in the coding of speaker involvement or, in other words, an externalization of the speaker. (As we have seen, this goes directly against Langacker's discussion of the same phenomenon in terms of an attenuation of the speaker's prominence within a given scene.) Objectivity, to Traugott, then serves to name the kind of coding that requires the least amount of inferencing, whereby a strong correlation is noted between objective expressions and a mode of speaking that is characterized as literal. This literal quality is in turn linked to a conservative heuristic in semantic processing which Horn (1984) called the "Q Princi12. In CG, not all subjectivity is expressive in this sense, though. In an example like (3b), involving implicit or subjective motion, it would be very difficult to establish a strict division between propositional and expressive elements of meaning. 13. These are some of the central presuppositions, it seems, of any formal-cognitive approach to language: it assumes that the mind is, at least in part, a system of (subsystems of) representations which have syntactic and semantic properties, and that mental processes are computations driven by the formal (syntactic) properties of these representations, it recognizes a conceptual language of thought which is distinct from any particular natural language, and it adopts the view that the mind is to some extent modular in structure. (Carston 2002: 2)

Insofar as his Conversationalist Hypothesis, and its sub-hypotheses concerning the nature of implicated meaning, partake in this scientific project, Grice's program and its many elaborations are fully embedded within this generative ("Chomskyan") paradigm.

58

Frank Brisard

pie" of informativeness and which acts as a counterforce to the innovating pressures of subjectification.14 The latter is connected with increased levels of informativeness or expressivity, and correspondingly follows a complementary "R Principle": 'Make your contribution necessary; say no more than you must.' The process of pragmatic strengthening associated with this strategy is held responsible for effects of subjectification. [N]ew and innovative ways of saying things are brought about by speakers seeking to enhance expressivity. This is typically done through "deroutinizing" of constructions, in other words, through finding new ways to say old things. Expressivity serves the dual function of improving informativeness for the hearer and at the same time allowing the speaker to convey attitudes toward the situation, including the speech situation. This very process of innovation is itself typically based on a principle of economy, specifically the economy of reusing extant forms for new purposes[...]. (Hopper and Traugott 1993: 65)

What is being concretely suggested here, in a slightly different terminology, is that "conversational implicatures" that are particularly salient or stereotypical may be conventionalized15 and thus become part of the encoded meaning of a linguistic item (its semantics). Correspondingly, implicatures have a bridging role in semantic change, as follows: "particularized" invited inferences lead to "generalized" ones, establishing a temporary state dubbed "pragmatic polysemy", where a new meaning remains coexistent with an item's original meaning. Only then can this new meaning be "semanticized", when the meaning giving rise to the inference disappears in certain contexts. Semanticization thus depends on a

14. In the wake of Grice's own preliminary analyses, a number of authors have proposed to collapse his non-Quality maxims of conversation into two basic principles regulating the economy of linguistic information: a Q Principle of informativeness and brevity and an R Principle of least speaker effort and minimal form. Without exception, an author adopting this style and terminology will easily find herself entangled in the intricate web of meaning types and subtypes typical of post- and neo-Griceanism. 15.The suggestion is presented as following up on Grice's (1989: 39) tentative statement that "it may not be impossible for what starts life, so to speak, as a conversational implicature to become conventionalized". Grice's statement here, incidentally, was meant as a strictly methodological comment, not one of historical substance.

Logic, subjectivity, and the semantics/pragmatics distinction

59

reanalysis 16 (leading to semantic polysemy), whereby "inferences do become references" (Bolinger 1971: 522). As an example, consider the contrast between contemporary English since, indicating a temporal or an inferred causal relation (semantic), and while, possibly expressing both time and concession as these are available together (a case of "merger" in the words of Coates 1983, and presumably fully pragmatic for Traugott). In the rise of new, subjective meaning, this is supposed to be a typical path of evolution: from conversational over conventional implicature (or possibly presupposition), to semantic content. The first two stages in this process do not represent semantic mechanisms, mainly by definition. Firstly, all Gricean implicatures are "defeasible" (nonmonotonic) and consequently do not belong to the semantics of the inference-inviting item proper. (Presupposition is not semantic because it remains constant under negation, unlike genuine semantic entailments.) Moreover, conventional implicature is not semantic either because, in spite of its regular association with an item, the postulated meaning is still not part of "what is said". This is in practice either defined in terms of a logical meaning (for some expression types seen as the natural-language counterparts of logical operators, for instance), or it is simply identified with an utterance's literal configuration (contributing some referential substance to a propositional structure). Take the case of after (as a connective), which may, but need not, implicate a causal relation on top of its temporal meaning (Hopper and Traugott 1993: 74). It is rightly argued that this inference is not a conventional implicature, since it does not occur with each usage event of the item under consideration. Rather, the inference represents an instance of pragmatic polysemy for Traugott, because the second, causal meaning is not at all established (in contrast to what is the case for a word like since). However, to insist on calling this inference an implicature after all, if not con-

16. See Hopper and Traugott (1993: section 3.3). Reanalysis of an old form, unlike (deductive) analogy, is thought to be based on "abduction", which is probably the pragmatic/pragmatist principle par excellence in accounts of inferencing, if only because of its close association with the philosophy of Charles Peirce. Andersen (1973), for example, discusses instances of language change involving a variant interpretation of the same surface form as arising from the discontinuity of transmission between generations, thus allowing grammars, or aspects of the code, to be inferred on the basis of the context-specific usage phenomena constituted by speech (as outlined with reference to phonological change in Czech, but very relevant for syntax/grammar; McMahon 1994: 92-95 contains a useful summary).

60

Frank Brisard

ventional, would amount to an inflated use of the term, which would come to be synonymous with "inference" and lose all the special characteristics that are justifiably assumed to mark real implicatures. And this is the first major problem with Traugott's "semantic-pragmatic" account of meaning change: given the lack of formal constraints informing her story of indirect meaning, how are we to differentiate between implicatures and other, looser types of (pragmatic) inference? For example, how can we distinguish some of Traugott's more dubious "implicatures" (like the "pragmatic" causal reading of after) from, say, a plain gerund that is able to generate the same inference: Reading / After we read your novel, we felt greatly inspired? And why would any kind of sequencing between clauses not potentially do exactly the same? But if the mining of world knowledge that motivates these readings is sufficient to call such phenomena pragmatic, then perhaps we have left the strictly regimented realm of Grice's conversationalist proposals (where emphasis is put on specific uses of specific expressions in specific contexts, and their generalizations). Referring to Horn's heuristics as underlying the semantic changes that Traugott is interested in, could be a dangerous move if the changes in question are seen as the results of implicature. For it is not clear, first of all, whether Horn himself considers such inferences to be real implicatures:17 his Q and R principles are more like overarching principles that operate on language in its entirety and that are not confined to implicature (see Kearns 2000: 9). Now, the problem with calling just about any more or less systematic (and defeasible) inference an implicature is that no specific type of implicature unites all the features mentioned by Traugott throughout her many analyses. And so we are forced to determine which kind(s) of implicature may be relevant to all or most of Traugott's claims, and whether that tells us something about the division of labor that is presumed to hold between semantics and pragmatics. The crucial point to remember about Grice's discussion of implicatures is that they need an argument going from the utterance of proposition ρ to the assumption of a different proposition q, in order to safeguard certain basic expectations of discourse cohesion and relevance. Therefore, not any contextually driven 17. Horn (1989: 194) states that the interaction between his Q and R heuristics covers a wide span of diverse phenomena, "ranging from implicature and politeness strategies to the interpretation of pronouns and gaps, from lexical change to indirect speech acts [and so on, FB]". The implication here is clearly that implicatures are not routinely to be equated with other types of inference, including those pertaining to meaning change (narrowing and broadening).

Logic, subjectivity, and the semantics/pragmatics distinction

61

enhancement of lexical and grammatical meaning should count as an implicature. If, in fact, a new element of meaning emerges (through an expression's interaction with context or world knowledge, or simply through relatively automatic mechanisms of association) that does not need any such argumentation, i.e., that is not to be linked to the original meaning through a series of implicit (propositional) assumptions, then we might not be faced with implicature, but rather with a case of (cognitively grounded) semantic extension (see section 4.2). If, moreover, there is no real or apparent reason for the hearer to suspect the need for additional argumentation in the interpretation of utterance tokens (because the interpretation would be markedly deficient without it), then we might ask what is specifically conversational about possible (subpropositional, associative) inferences arising under these conditions. I believe that a majority of Traugott's examples of subjectification-through-conventionalization (of implicatures) suffers from both problems, and that this calls for a reappraisal of the data in a radically different theoretical light. Technically, these issues manifest themselves with respect to various levels of meaning analysis. There is, to begin with, the uncertain relation between Horn's principles and the status of an inference qua implicature. This is foreshadowed in Geis and Zwicky's (1971) original observations on invited inferences, which are restricted to generalized implicatures, i.e., to those (conversational) implicatures that apply regardless of any specific contextual conditions (see Traugott and Dasher 2002: 5). Hopper and Traugott (1993: 75) do also note that "for inferences to play a significant role in grammaticalization, they must be frequently occurring, since only standard inferences can plausibly be assumed to have a lasting impact on the meaning of an expression". But the recourse to "standard inferences" (see also Levinson 1983: 104) clashes, in a way, with Traugott's repeated insistence on the initial generation of implicatures leading to semantic change in local contexts. So there must be a stage of inferencing preceding the establishment of standard implicatures that is in effect local and that consequently leads to generalizable semantic change. Several candidates present themselves here. Conventional implicatures, for one, are attached to one, local form, but they do not arise in any historical sense within the standard Gricean picture - the conventional implicature of contrast associated with but (Grice 1961), for instance, is stipulated for reasons of logical analysis, and not as reflecting historical processes. Next, we might turn to particularized conversational implicatures, which augment an original meaning but are typically not attached to one form (by definition; see

62

Frank Brisard

Grice's feature of "nondetachability"). The only suitable remaining candidate, then, comprises the class of so-called generalized conversational implicatures (GCIs; cf. Atlas and Levinson 1981, Horn's 1984 discussion of Q-based scalar and clausal implicatures, and Levinson 2000). On the neoGricean view, GCIs are reconceived as contents that a speaker directly communicates, rather than being merely conversationally implicated. Levinson's own view is that GCIs are default inferences that will be drawn unless something unusual in the context blocks them. Still, this does not solve Traugott's problem of finding the originally local basis for such GCIs, and it certainly does not square with the observation that any good paraphrase of an utterance with an implicature (including a generalized one) will still carry that same implicature: why, then, would since initially carry a causal implicature while after would not? Incidentally, GCIs are generally called in to explain the conflict between logical and naturallanguage meanings (as with Horn's scalar implicatures), but the issue of logical meaning is not exactly foregrounded in Traugott's own work, or so it seems. A problem that is heavily correlated with the previous discussion concerns the "exploiting" nature of conversational implicatures. Let us first assume that the exact status (i.e., particularized vs. generalized) of such implicatures remains undecided for now. Still, what they are thought to share, at least in Grice's conception, is the idea that a hearer needs more information than what has been "said" in order to make sense of an utterance, and that she will therefore engage in a rational argument to arrive at that additional information. This is usually associated with Grice's (1989: 30ff) discussion of the conversational maxims elaborating his Cooperative Principle: it is the flouting of a maxim, i.e., the blatant failure to fulfill it, that characteristically leads to implicature, as the hearer must assume some extra bit of information q to maintain the supposition that the speaker is conversationally cooperating. None of this, now, is systematically available in the range of studies presented by Traugott. That is to say that some of the cases adduced do, while others do not, conform to Grice's descriptions. One might counter that Grice left room for implicatures that do not betray any violation of maxims: the class of GCIs would be a case in point, but there we are stuck with the local origins of Traugott's inferential meanings, which do not go well with the default status of GCIs. And even an example like (5), involving a particularized inference following an apparently unproblematic exchange, still assumes that there is a connection between A's

Logic, subjectivity, and the semantics/pragmatics distinction remark and B's reaction that is not present in the explicitly coded utterance parts: (5)

A: I am out of petrol. B: There is a garage around the corner, (implicature: the garage is, or may be, open, has petrol to sell, etc.)

Not to assume such a relation (e.g., of relevance) would amount to losing obvious pieces of information and would effectively render the exchange incomprehensible. So it is not so much the obviousness of inferences, but rather their necessity or inevitability in the course of interpretive work, that seems to determine whether or not such inferences might be labeled implicatures, at least in Grice's view. In the light of the fact that many of the semantic extensions identified by Traugott do not show this kind of necessity (e.g., it is not necessary to assume a causal relation between the two events in the temporal reading of Since we read your novel, we felt greatly inspired, although such an assumption could in many cases be useful), it might be reasonable to suggest that this type of additional information enters the picture regardless of what the hearer does or does not assume with respect to the speaker's intentions. It is, in other words, information that is relatively cheap in the automaticity with which it presents itself, and that should therefore be treated as somehow associatively related to an explicit content (rather than being discovered through something like a rational argument, as with canonical implicatures). Such associations may have greater or lesser degrees of prominence, depending on things like frequency and context, and thus they may vary in the necessity with which they are felt to accompany the use of a particular item. But that is no reason to call them pragmatic, if pragmatics is the study of particular meaning effects related to a speaker's intentional, strategic behavior. Two main conclusions can be drawn at this point. Either Traugott calls an implicature conversational if it is "not part of the meaning of any particular element in the utterance" (Traugott and König 1991: 194), but then the term is used as a synonym of inference and there is really no need to go into all of the technical subtleties marking present-day (formal) pragmatics. Or she is, in certain cases, effectively talking about standard implicatures that are generalized and do not involve any violation of maxims. But then the explicit analytic appeal to Gricean principles of relevance (and others) looks awkward and actually takes the focus away from the speaker's intentionality (e.g., in performing acts of conversational breaching), in favor of

63

64

Frank Brisard

the inference-drawing hearer - whereby it does not really matter what the source of an inference is: a speaker's decision to communicate or imply it, or its quasi-automatic association with a given item's use. This, in turn, seems to clash with Traugott's outspoken concern with the speaker as the prime locus of meaning change (cf. the discussion of subjectivity as personal self-expression, as well as the emphasis put on Horn's R Principle of least speaker effort and related claims about the speaker as a central force in semantic innovation). 18

4.2.

Conceptualist semantics

In what follows, I will very briefly sketch an alternative to the theoretical model adopted by Traugott as a background in her description of pragmatic enrichment and subjectification. This alternative issues from Cognitive Semantics and its appeal to usage-based mechanisms of linguistic organization (see especially Langacker 1987; other useful overviews can be found in Rudzka-Ostyn 1988, Tomasello 1998-2003, and Barlow and Kemmer 2000). In CS, there can be no contrast between an external (logical) semantics and an internal pragmatics, because the externalist ("objectivist") basis of semantics is basically not recognized. The dichotomy, as conjured up by Traugott and many of her colleagues in functionalist circles, is fallacious from a cognitive standpoint, because it assumes a fundamental difference between expressing information and regulating communication. Such a contrast is not provided for, however, in any conceptualist model of linguistic meaning, like CS, where the notion of construal serves to indicate an element of negotiation in communication that is always available, both in lexical and grammatical meaning and at all conceivable levels of structure. In this sense, there is no tension between linguistic representations and their (strategic) uses. Such a cognitively informed picture of pragmatics, furthermore, cannot possibly be reconciled with a notion of communication that is inherently severed from some decontextualized understanding of linguistic meaning. 18. Thus, Traugott and Dasher (2002: 5) do not take up Heine, Claudi and Hiinnemeyer's (1991) term "context-induced inference" because it would underexpose the speaker's intentional makeup, they explain: "the latter term suggests a focus on AD/Rs [hearers] as interpreters and appears to downplay the active role of SP/Ws [speakers] in rhetorical strategizing, indeed indexing and choreographing the communicative act."

Logic, subjectivity, and the semantics/pragmatics distinction

65

In a usage-based model with partial sanctioning of fully specified established units, there is always meaning beyond structure. If "pragmatic" is then identified with what has not been explicitly encoded, it will also have to denote purely associative aspects of meaning, which come with the use of a linguistic item regardless of whether the speaker intended them or not. This, in fact, would amount to a trivial use of the term. Nonconventional meanings that are mere associations of use are in principle covered by CS in its account of semantic extension. Such extensions of meaning are governed, according to cognitive linguists, by a set of readily identifiable conceptual mechanisms, including metaphor, metonymy, and various so-called image-schema transformations, but also subjectification/grammaticalization. These extensions may be labeled semantic simply because they belong to the realm of linguistic meaning; there is no need to devise a special term for them. And even if they are defeasible at some point in their evolution, that should not necessarily make them pragmatic. For instance, the causative meaning of since may have been lingering right below some threshold of unit status without ever being pragmatic in any formal sense. It was just part of the conceptual background and gradually became "semantic", i.e., conventionally associated or entrenched and thus less and less defeasible in certain contexts. Context, insofar as its relevance for a given meaning is frequently reestablished, may be represented in a predication's "base space",19 i.e., "those portions of active cognitive domains that a predication specifically invokes, providing the background against which some entity stands out as the [designating] profile" (Langacker 1991a: 544). In general, any type of implicit notion, traditionally a good candidate for pragmatic analysis, can be represented thus in CS. This constitutes its "encyclopedic" span. Any attempt to grant this type of meaning a distinctive status would have to

19. This solution holds for other phenomena as well that are typically considered pragmatic, like presupposition. In Langacker (2002b), the speaker's implicit acceptance of a proposition as real with the use of factive verbs like know is described in terms of the offstage and subjective construal of the speaker and her epistemic attitude. This represents a general analytic strategy in CG, where implicit or indirect meaning, including certain kinds of implicature, is said to be evoked in the background, as a "virtual" foil of the actual entities and relations at issue (see also Langacker 1999d). The idea of background conceptions is also essential for describing the workings of linguistic negation, including matters of scope, polarity, and the special behavior of presuppositions in this regard (Langacker 1991a: 134-135).

66

Frank Brisard

come up with principled ways of differentiating between essential (semantic) and nonessential (pragmatic) elements of meaning construction, yet such principles are nowhere to be found in CS. But is everything encyclopedic semantics, then, and, in particular, is subjectification, too, a mere process of "partial sanctioning" (the cognitive mechanism behind semantic extension)? The answer to the first question is negative, I believe: there is still room for elements of pragmatic meaning, which may comprise subjective ones. However, not every subjective meaning, at least in the sense of Langacker, is pragmatic. Certain meaning types may resort to the exploitation of conventionalized expectations with regard to "what is the case" (as expressed in the form of a proposition). This type of exploitation may be taken in the Gricean sense, without thereby co-opting the logical underpinnings of his conversationalist program (particularly noticeable in the explication of "what is said"). Other labels to refer to phenomena crucially linked to linguistic acts of exploitation include "accommodation" (Lewis 1979) and context creation, which depend on the availability of shared knowledge or, in its absence, on the readiness of discourse participants to behave as if there were something to be shared. Such uses may all eventually give rise to instances of semantic extension, but, importantly, the pragmatic element in them does not just disappear after they have been conventionalized. That is to say that conventionality, a very moot point indeed in any typology of meaning, is not and cannot be the decisive factor in calling something semantic or pragmatic. Instead, since exploitation is not a matter of adding or subtracting some conceptual content at the level of semantic representations, but of adopting a certain marked stance toward such representations, it constitutes a genuine qualitative (and hence not entirely fuzzy) cut-off point between different types of meaning, one which cannot be provided by the gradualist notions of convention or conceptual entrenchment. Semantics is thus not left to deal with "underdetermined" meanings that are pragmatically enriched. In CS, semantics is as rich as it gets and comprises those instances of "sense addition" that may be covered by general principles of cognitive organization. Exploitation, in contrast, is not a case of adding representational material, but rather of treating a given content conspicuously differently.20

20. This formulation might be seen by some to invite all sorts of methodological difficulties, related to issues of truth and interpretive authority. However, I do intend this claim to be understood from an "emic" perspective, that is, from the

Logic, subjectivity, and the semantics/pragmatics distinction

67

Consider the next two (rather unrelated) illustrations. It is not that I am adding anything to the meaning of block of ice when I call somebody that; I am rather using the term, with its conventional meaning, in a somewhat surprising way (where "surprise" itself is a relative and highly individualized notion, much like the idea of metaphorical "tension" in the extant psychological literature on metaphor comprehension). The idiomatic status of this expression, used figuratively, does not detract from its special character, most speakers of English would agree; probably, it is only with very strongly conventionalized, "dead" metaphors like the foot of a mountain that the pragmatic aspect of a figurative use, or its tension, can be lost, which would make the metaphor truly semantic - not on any grounds of frequency, though, but based on a category change, frequency-driven, in what speakers feel they are "saying" when they use the expression. Likewise, I am not changing (adding to or subtracting from) the established meaning of the English present tense, which is that an event coincides with the time of speaking ("is true right now"), when I use it to refer to a situation that is obviously nonpresent - say, in generic contexts, or with the historical present - , or when I ask a question about a present-time event that seems obviously observable - my asking Do you smoke? when I see someone enjoying their cigarette. What needs to be added, in the course of interpretation, to the conventional knowledge of the meaning of this tense, is the realization that its use in these particular contexts is less than felicitous unless the hearer mentally assumes a viewpoint from which she can make sense of the utterance. Characterizing this viewpoint is not a matter of adding representational content but of describing a certain attitude toward some "original" content: there is nothing that needs to be put into the semantics of the English present tense to explain how, in certain contexts of use, the phrase Do you smoke? can come to mean something like 'are you a smoker?'. Exploitation, thus defined, appears everywhere in language and completely redrafts traditional dividing lines between linguistic semantics and pragmatics: it includes typical examples of Gricean implicature but also forms of subjective meaning that rely on mental acts of footing and positioning, and it excludes those meaning types whose only perspective of discourse participants ostensibly generating meaningful categories and contrasts. It is they, and not the analyst, who decide what counts as "different" and thus pragmatic, and what does not. Evidence for the pragmatic status of a linguistic element should thus be found in the utterances, behaviors, and orientations of (real or idealized) discourse participants, not in any logical analysis of propositions.

68

Frank Brisard

claim to pragmatic fame lies in the fact that they happen to be somewhat less entrenched (within a polysemy network), or that they are presumed, purely on speculative grounds, to derive from an underlying logical meaning that is "timeless".21 Measures of semantic "distance", "networks" of polysemies, and other "maps" will not help establish the semantics/pragmatics distinction, because these all tend to be situated in, and describe, the semantic component of CS anyway. Any cut-off point within these models would be arbitrary, as there exists no absolute motivation to choose a degree of distance great enough to start talking about pragmatically related/derived meanings. One could conceivably wonder, then, to what extent Levinson's (2000: 243) rather uncharitable critique of cognitive-linguistic "pragmantics" may be warranted after all: "Pragmantics can be avoided if we can find a way of accounting for pragmatic intrusion into truth conditions while maintaining the modularity of a distinct pragmatics (built on nonmonotonic principles) and semantics (built on monotonic principles)." I have just tried to argue, though, that there might very well be a principled motivation - that is nonmodular - for differentiating between a cognitively based style of semantic analysis, and the pragmatic analysis of utterances in context (which is presumably also, if only partially, cognitively informed). It is true, of course, that terminology is only terminology. We should not be surprised to 21. Two important implications follow from this. In CG, the analysis of the alleged natural-language counterparts of logical operators (mostly involving so-called GCIs) and of conventional implicatures is handled by the usual cognitive mechanisms, deemed semantic. Thus, the exclusive (logical but counterintuitive) meaning of or is treated by Langacker (2003: 43) as a specialized, not basic, sense that may indeed be "implicated", but not conventionally. Secondly, it should be realized that, in the present proposal, pragmatic meaning is necessarily a feature of (quasi-)propositional structures (Gricean "utterances"), not of individual lexemes in abstraction. Although pragmatic effects can be located at the level of words, as in many cases of metaphor, it is only the use of such words in the utterances they are part of that calls for a pragmatic analysis. As indicated, principles of extension governing the meaning changes that occur in lexicon can arguably be called semantic in CS. This picture corresponds, though in very rough ways and with possibly minimal philosophical overlap, with Recanati's (2002) distinction between "subpropositional associative primary processes", driven only by considerations of cognitive effort and efficiency, and properly inferential "propositional secondary processes", which are nonlogical, utterly strategic, and guided by the standard Gricean maxims (and possibly others).

Logic, subjectivity, and the semantics/pragmatics

distinction

69

see Langacker (1995a) appear in a Handbook of Pragmatics, when many of the concerns of CG (its encyclopedic view of meaning, its interest in construal as a matter of perspective or position-taking) are also seen as valid research objects in different strands of contemporary linguistic pragmatics. If meaning is the object of study, it has not generally mattered in cognitive linguistics whether "semantic" or "pragmatic" was chosen as the appropriate adjective. I think this has been a healthy attitude. It is only when prompted by heated debates on the supposed semantics/pragmatics interface that cognitive linguists may take a more explicit stand, lest they be accused of only making a mess of things.

5.

Conclusion

There is every reason to retain a lively interest in Traugott's findings, as well as in her general story of semantic change as involving a conventionalization of (often abstract) inferentially or associatively generated meanings. That has not been the point of this paper. Rather, I have opted to present two good reasons for cognitive linguists to beware of uncritically adopting the theoretical mould into which some of these findings have been cast. The first of these reasons concerns the (psychological) substance to be suspected behind linguistic manifestations of subjectivity, where I suggest that Traugott takes a process-oriented view that is not entirely consonant with some of the phenomenology of linguistic subjectivity, as portrayed in CG. The second reason pertains to the status of the inferences typically giving rise to subjective meanings, and their relationship with speaker intentions. This is more of a methodological question, though with serious ties to underlying theoretical presuppositions regarding the "architecture" of linguistic meaning and its relation to context. The case to be made in this paper has only been directed at Traugott's work as a paradigm example, and nothing more, of what can go wrong in the scholarly interaction between cognitivism and various types of functionalism in linguistics. There is no denying that the two paradigms should be seen as largely compatible, but at the same time many functionalist models today are still tributary (particularly in some of their more implicit orientations) to formerly dominant logicist conceptions of language structure, and to referentialist assumptions about (basic) linguistic functions. Traugott illustrates this tendency most explicitly, I suppose. It also goes to show that Cognitive Grammar, and the collection of analytic tools and

70

Frank Brisard

protocols that might be distilled from it under the heading of Cognitive Semantics, constitute more of a radical linguistic approach than is often acknowledged. In its rejection of both grammatical formalism and semantic correspondence theories, the cognitive-linguistic paradigm presents new useful ways of dealing with old, insightful observations, provided that the truly usage-based character of the data is thoroughly thought through.

References Andersen, Henning 1973 Abductive and deductive change. Language 49. 567-595. Atlas, Jay D. and Stephen C. Levinson 1981 Λ-clefts, informativeness, and logical form: Radical pragmatics (revised standard version). In Peter Cole (ed.), Radical Pragmatics, 1-61. New York: Academic. Barlow, Michael and Suzanne Kemmer (eds.) 2000 Usage-Based Models of Language. Stanford: CSLI Publications. Benveniste, Emile 1966 De la subjectivite dans le langage. In Emile Benveniste, Problemes de linguistique generale, 258-266. Paris: Gallimard. First published Journal de Psychologie [1958]. Blakemore, Diane 1987 Semantic Constraints on Relevance. Oxford: Blackwell. Bolinger, Dwight 1971 Semantic overloading: A restudy of the verb remind. Language 47. 522-547. Brisard, Frank (ed.) 2002 Grounding: The Epistemic Footing of Deixis and Reference. Berlin: Mouton de Gruyter. Carston, Robyn 2002 Thoughts and Utterances: The Pragmatics of Explicit Communication. Oxford: Blackwell. Coates, Jennifer 1983 The Semantics of Modal Auxiliaries. London: Croom Helm. Dahl, Osten 1985 Tense and Aspect Systems. Oxford: Blackwell. Geis, Michael L. and Arnold M. Zwicky 1971 On invited inferences. Linguistic Inquiry 11. 561-566. Green, Georgia M. 1989 Pragmatics and Natural Language Understanding. Hillsdale, NJ: Erlbaum.

Logic, subjectivity, and the semantics/pragmatics distinction Grice, H. P. 1961 The causal theory of perception. Proceedings of the Aristotelian Society, Supplementary Volume 35. 121-152. 1989 Studies in the Way of Words. Cambridge, MA: Harvard University Press. Heine, Bernd, Ulrike Claudi and Friederike Hünnemeyer (eds.) 1991 Grammaticalization: A Conceptual Framework. Chicago: University of Chicago Press. Hopper, Paul J. and Elizabeth Closs Traugott 1993 Grammaticalization. Cambridge: Cambridge University Press. Horn, Laurence R. 1984 Toward a new taxonomy of pragmatic inference: Q-based and itbased implicature. In Deborah Schiffrin (ed.), Meaning, Form, and Use in Context, 11—42. Washington, D.C.: Georgetown University Press. 1989 A Natural Histoiy of Negation. Chicago: Chicago University Press. Kearns, Kate 2000 Implicature and language change. In Jef Verschueren, Jan-Ola Östman, Jan Blommaert and Chris Bulcaen (eds.), Handbook of Pragmatics 2000, 1-22. Amsterdam: Benjamins. Lakoff, George and Mark Johnson 1980 Metaphors We Live By. Chicago: University of Chicago Press. Langacker, Ronald W. Active zones. Proceedings of the Annual Meeting of the Berkeley 1984 Linguistics Society 10. 172-188. Observations and speculations on subjectivity. In John Haiman 1985 (ed.), Iconicity in Syntax, 109-150. Amsterdam: Benjamins. Foundations of Cognitive Grammar, vol. 1: Theoretical Prereq1987 uisites. Stanford: Stanford University Press. Foundations of Cognitive Grammar, vol. 2: Descriptive Applica1991 a tion. Stanford: Stanford University Press. Subjectification. In Ronald W. Langacker, Concept, Image, and 1991 b Symbol: The Cognitive Basis of Grammar, 315-342. Berlin: Mouton de Gruyter. Cognitive grammar. In Jef Verschueren, Jan-Ola Östman and Jan 1995 a Blommaert (eds.), Handbook of Pragmatics: Manual, 105-111. Amsterdam: Benjamins. Raising and transparency. Language 71. 1-62. 1995 b Consciousness, construal, and subjectivity. In Maxim I. Stamenov 1997 (ed.), Language Structure, Discourse and the Access to Consciousness, 49-75. Amsterdam: Benjamins.

71

72

Frank Brisard 1999 a

Assessing the cognitive linguistic enterprise. In Theo Janssen and Gisela Redeker (eds.), Cognitive Linguistics: Foundations, Scope, and Methodology, 13-59. Berlin: Mouton de Gruyter. 1999 b Conceptual grouping and constituency. In Ronald W. Langacker, Grammar and Conceptualization, 147-170. Berlin: Mouton de Gruyter. 1999 c Subjectification and grammaticization. In Ronald W. Langacker, Grammar and Conceptualization, 297-315. Berlin: Mouton de Gruyter. 1999 d Virtual reality. Studies in the Linguistic Sciences 29. 77-103. 2002 a Deixis and subjectivity. In Brisard, 1-28. 2002 b The control cycle: Why grammar is a matter of life and death. Proceedings of the Annual Meeting of the Japanese Cognitive Linguistics Association 2. 193-220. 2003 Dynamicity, Activity, and scanning: The imaginative basis of logic and linguistic meaning. Korean Linguistics 18. 1-64. Levinson, Stephen C. 1983 Pragmatics. Cambridge: Cambridge University Press. 2000 Presumptive Meanings: The Theory of Generalized Conversational Implicature. Cambridge, MA: MIT Press. Lewis, David 1979 Scorekeeping in a language game. Journal of Philosophical Logic 8. 339-359. Lyons, John 1977 Semantics, vol. 2. Cambridge: Cambridge University Press. 1982 Deixis and subjectivity: Loquor, ergo sum? In Robert J. Jarvella and Wolfgang Klein (eds.), Speech, Place, and Action: Studies in Deixis and Related Topics, 101-124. New York: Wiley and Sons. 1995 Linguistic Semantics: An Introduction. Cambridge: Cambridge University Press. McMahon, April M. S. 1994 Understanding Language Change. Cambridge: Cambridge University Press. Nuyts, Jan 2001 Epistemic Modality, Language, and Conceptualization: A Cognitive-Pragmatic Perspective. Amsterdam: Benjamins. 2002 Grounding and the system of epistemic expressions in Dutch: A cognitive-functional view. In Brisard, 433-466. Panther, Klaus-Uwe and Linda L. Thornburg (eds.) 2003 Metonymy and Pragmatic Inferencing. Amsterdam: Benjamins.

Logic, subjectivity, and the semantics/pragmatics distinction Pelyvas, Peter 1996 Subjectivity in English: Generative Grammar versus the Cognitive Theory of Epistemic Grounding. Frankfurt am Main: Lang. Recanati, Franfois 2002 Does linguistic communication rest on inference? Mind and Language 17. 105-126. Rudzka-Ostyn, Brygida (ed.) 1988 Topics in Cognitive Linguistics. Amsterdam: Benjamins. Sweetser, Eve E. 1984 Semantic Structure and Semantic Change: A CognitiveLinguistic Study of Modality, Perception, Speech Acts, and Logical Relations. University of California at Berkeley, doctoral dissertation. 1990 From Etymology to Pragmatics: Metaphorical and Cultural Aspects of Semantic Structure. Cambridge: Cambridge University Press. Taylor, John R. 1989 Linguistic Categorization: Prototypes in Linguistic Theory. Oxford: Clarendon. Tomasello, Michael 1998-2003 The New Psychology of Language: Cognitive and Functional Approaches to Language Structure, 2 vols. Mahwah, NJ: Erlbaum. Traugott, Elizabeth Closs 1982 From propositional to textual and expressive meanings: Some semantic-pragmatic aspects of grammaticalization. In Winfred P. Lehmann and Yakov Malkiel (eds.), Perspectives on Historical Linguistics, 245-271. Amsterdam: Benjamins. 1988 Pragmatic strengthening and grammaticalization. Proceedings of the Annual Meeting of the Berkeley Linguistics Society 14. 406415. 1989 On the rise of epistemic meanings in English: An example of subjedification in semantic change. Language 65. 31-55. 1995 Subjectification in grammaticalisation. In Dieter Stein and Susan Wright (eds.), Subjectivity and Subjectivisation, 31-54. Cambridge: Cambridge University Press. Traugott, Elizabeth Closs and Richard B. Dasher 2002 Regularities in Semantic Change. Cambridge: Cambridge University Press.

73

74

Frank Brisard

Traugott, Elizabeth Closs and Ekkehard König 1991 The semantics-pragmatics of grammaticalization revisited. In Elizabeth Closs Traugott and Bernd Heine (eds.), Approaches to Grammaticalization, vol. 1, 189-218. Amsterdam: Benjamins. Wittgenstein, Ludwig 1963 Tractatus logico-philosophicus. Frankfurt am Main: Suhrkamp.

Toward a typology of linguistic subjectivity: A cognitive and cross-linguistic approach to grammaticalized deixis Satoshi Uehara "A further question ... is whether different natural languages differ in respect of the degree of subjectivity that they impose upon their users." (Lyons 1982: 105-106)

0.

Introduction

Subjectivity, as an "intangible, seemingly nebulous concept" (Langacker 1985: 147), has played a rather minor role as the object of linguistic investigation.' Some, mostly functionally and cognitively oriented linguists

*

This is a revised version of the paper presented at the Paths of Subjectivity theme session of the 8th International Cognitive Linguistics Conference held at University of La Rioja, Spain in July 2003. Earlier versions of the paper were presented at the International Conference on Cognitive Typology (University of Antwerp, Belgium, April 2000) and at the first Annual Meeting of the Japanese Cognitive Linguistics Association (Keio University, Japan, September 2000). I am thankful to the audiences at the conferences, and my special thanks go to Ken Cook, Bill Croft, Susan Kemmer, Ron Langacker, Prashant Pardeshi, Matt Shibatani and three anonymous reviewers for their critical and helpful comments/discussions. I am also grateful to Bob Sanders for his suggestions on textual improvements. Needless to say, all remaining errors are mine. The following abbreviations are used in glosses: ACC = accusative; ART = article; AUX = auxiliary; CONJ = conjunctive; COP = copula; DAT = dative; DEC = declarative; HON = honorific (=subject honorific); HUM = humble (=non-subject honorific); NOM = nominative; OBJ = object; QT = quotative complementizer; SG = singular; TOP = topic. 1. Lyons (1982: 101) notes the pejorative connotations the English word 'subjectivity' has acquired, unlike its French and German counterparts, by virtue of its opposition with a positivistic interpretation of 'objectivity'. He goes on to say

76

Satoshi Uehara

(Benveniste 1971 [1966]; Ohye 1975; Lyons 1982; Langacker 1985; Iwasaki 1993, inter alia), however, have brought the issues of linguistic subjectivity to the fore. More recently, discussions of subjectivity have evolved around historical processes of subjectification, and Langacker (1985, 1990 and 1991a) and Traugott (1989 and 1995), in particular, have developed theories of subjectification in relation to the diachronic process of grammaticalization. The current research is typologically oriented, and takes a synchronic, but radically cross-linguistic approach to linguistic subjectivity. Particularly it will focus on the question raised by Lyons quoted above in the epigraph to this paper. To answer the question, we need a cross-linguistically viable definition of subjectivity and its theoretical constructs. I will argue in this paper that the cognitive theory of subjectivity (proposed by Langacker 1985, etc.) can handle this problem, because it can deal with 1) deixis and other subjectivity phenomena in a uniform fashion, and 2) the subjectivity of expressions denoting events and entities within the propositional content. Such proposition-internal expressions, according to Langacker's theory, can be subjective even before being "subjectified" into elements in the nonpropositional, expressive, attitudinal component of the language by the process of grammaticalization, which is more or less 'universal', thus yielding less cross-linguistic variation. 2 1 will demonstrate how this can be done, and propose a new dimension to linguistic typology, namely, 'subjectivity typology'. The organization of this paper is as follows: Section 1 lays out Langacker's cognitive linguistic theory of subjectivity and discusses its applicability to cross-linguistic study. Sections 2, 3 and 4 examine three types of linguistic expressions and the cross-linguistic variation in their degrees of subjectivity. Section 5 summarizes findings and discusses the implications of the current study.

that many influential works in "Modern Anglo-American linguistics, logic, and philosophy of language ... pay no attention at all to the [subjective] components of language or play down their importance." (Lyons 1982: 103-104). 2. Traugott and Dasher (2002: 40), for example, lists the following as two of the correlated paths of directionality in semantic change involved in grammaticalization (s-w = scope within; s-o = scope over): s-w-proposition nonsubjective

> >

s-o-proposition subjective

> >

s-o-discourse intersubjective

Toward a typology of linguistic subjectivity 1.

77

Subjectivity and cross-linguistic variation

Langacker explains 'subjectivity' as follows: "Subjectivity pertains to the observer role in viewing situations where the observer/observed asymmetry is maximized." (Langacker 1985: 109). Let us consider such a viewing situation, which he calls an optimal viewing arrangement, diagramed in Figure 1(a). In the diagram, ' S ' stands for the subject of conception (or observer), Ό ' for the object of conception (or observed), the arrow for the direction of conception, and the broken-line circle for the objective scene.

(a)

(b) Ο

Figure 1. (from Langacker 1985: 121) With respect to an optimal viewing arrangement in Figure 1(a), he notes, "S can be characterized as maximally subjective, and Ο as maximally objective."(p. 121). This optimal viewing arrangement is contrasted with what he calls the egocentric viewing arrangement, diagramed in Figure 1(b), where the locus of viewing attention is expanded to include the position of S and his/her immediate surroundings. The subject of conception S is no longer simply an observer, but to some degree an object of conception as well, and in this situation, S receives a more objective construal while the scene conceived becomes more subjective. Thus, the conceptualization diagrammed in Figure 1(b) represents the semantic structure of 'subjective' expressions, and Langacker defines a subjective expression "as one that includes the ground - or some facet of the ground - in its scope of predication (i.e. its base)." (Langacker 1985: 113, emphasis in the original). (The term 'ground' is used in Cognitive Grammar to indicate the speech event, its setting, and its participants.) As one can see, expressions like "maximally subjective" and "more subjective" indicate that subjectivity is a matter of degree, depending on how prominent the ground is conceived in the overall conceptualization. Thus,

78

Satoshi Uehara

Langacker introduces the notion of a subjectivity scale, along which linguistic expressions can be ranked.

1.1.

Subjectivity typology

Now, my purpose here is to examine the patterns of cross-linguistic variation within this construct of linguistic subjectivity. The idea here is not new; Lyons (1982: 106), for example, raises the question quoted in the preface of this paper and suggests that his intuitive answer would be in the affirmative (see also Ohye 1975 and Iwasaki 1993). But we need to consider how we can compare the degrees of subjectivity across languages. This is not an easy task to do, but I will be arguing that it can still be done, by employing a commonly used methodology in linguistic typology. That is, we take semantically/functionally same/similar expressions from different languages for comparison. As we will see below, the degree of subjectivity with which the linguistic expressions of some events/entities are conventionalized can vary from language to language. I propose two mutually related criteria in (1) for cross-linguistic comparison: (1)

two criteria for cross-linguistic comparison of linguistic subjectivity a. how subjectively certain events/entities are construed (i.e., how subjective the linguistic expressions of certain events/entities are) b. how obligatorily/preferentially/typically subjective expressions are used

For the first criterion (la), we need to know how to assess the degree of subjectivity of linguistic expressions. To do this, I propose three levels in the subjectivity scale. The three levels and how they are differentiated are shown in Figure 2 [">" means "is more subjective than"]: subjective implicit reference to the ground (e.g. He came.

objective > explicit reference to the ground > He came to me.

Figure 2. Three levels in subjectivity scale

>

non-reference to the ground > He went to Bill.)

Toward a typology of linguistic subjectivity

79

The three-level scale of subjectivity in Figure 2 is attained by combining two points of comparison. The first one is whether the expression is deictic or not, in other words, whether or not there is reference to the ground. When one of the two expressions is deictic, that is, it invokes the ground, while the other is not deictic, then the deictic expression is more subjective than the other non-deictic one (see the quote from Langacker 1985 above). For example, come/go up is more subjective than ascend (DeLancey 1985: 381), and Mary talked to me is more subjective than Mary talked to Ken. The other point of comparison is whether reference to the ground is implicit or explicit. When the two expressions make reference to the ground, they are both somewhat subjective, by the definition above. But the expression with implicit reference has a higher degree of subjectivity. The distinction in question is what Lyons calls "the distinction between the subjective experiencing self and the objective observing self." (1982: 107). He cites the sentences in (2): (2)

a. I remember switching off the light. b. I remember myself switching off the light.

He examines the two contexts in which each of the two sentences in (2) is most appropriately used (see his discussion of differences between the two contexts) and notes that (2b) "can be interpreted .... as reporting the memory of something observed, rather than experienced." (Lyons 1982: 107) Langacker (1985: 137-145) also discusses many such examples including the one reproduced in (3): (3)

a. Ed Klima is sitting across the table from me! b. Ed Klima is sitting across the table!

He explains the sentence in (3b) is possible when the speaker sits at a table and finds Ed Klima sitting in front, in other words, when s/he is experiencing the event/scene, while it is not possible when the speaker is looking at him/herself in the picture, i.e. objectively as the object of his/her perception, as in (3'): (3')

Look at this photograph! a. Ed Klima is sitting across the table from me! b. *Ed Klima is sitting across the table!

80

Satoshi Uehara

Notice here that his two contexts, 3a and 3b, are objectively (i.e., in the eyes of the third person) the same: they both involve the speaker (thus both are subjective), with respect to which Ed Klima is located, but that they differ in the degree to which the speaker is conceived as the object of his/her own perception. 3 He then argues that there is an "iconic relationship between objectivity and explicit mention." (p. 138) and that "(i)mplicit reference to the speaker correlates with the speaker being construed more subjectively" (p. 140). 4 Thus, according to Langacker (1991: 329), the implicit reference pattern in (3b) "adds another degree and type of subjectification to that already observed in" the explicit reference pattern in (3a). 5

3. See Uehara (1998b) for some pictorial representations of what these two sentences depict and a case study of subjective predicates in Japanese as an application of the subjectivity distinction in question. 4. Traugott and Dasher (2002: 98) argues against Langacker's analysis here observing the fact that the reference point is not necessarily the speaker, but it could be someone else for the sentence (3b) giving certain contexts for it. However, they seem to be confusing "maximally subjective" in the basic level and that in the secondary level or the cases of what I call "perspective transfer" (Uehara 1998a), where the speaker identifies himself with/takes the perspective of one of the participants of the event he is describing, and describes it from her perspective. Thus, the expression in (3b), if the implicit reference point is Martha, whose perspective the speaker is taking, is still maximally subjective because it should be compared with the sentence Ed Klima is sitting across the table from Martha, not with (3 a). 5. This point regarding the implicit reference to the ground (e.g. He came) being more subjective than the explicit reference to the ground (e.g. He came to me) appears to be confusing to some people, including one of the reviewers, who has disagreed with the proposed subjectivity cline in Figure 2 and claimed that "the correct hierarchy should be He came to me > He came". I would like to direct their attention to the existence of the observer role external to the observed scene for objectivity, as seen in Figure la, and the loss of the observer/observed asymmetry for subjectivity, as in Figure lb. Both He came to me and He came are subjective/deictic in that they both refer to the ground. However, the former (the speaker's explicit self-reference) implies an external vantage point, from which the speaker views him/herself on stage, while the latter (the speaker's implicit self-reference) implies no such external vantage point and has little basis for the observer/observed contrast. In this way one can clarify the confusion and find that, between the two subjective patterns, the latter (e.g. He came) is more subjective than the former (e.g. He came to me). The same applies to the

Toward a typology of linguistic subjectivity

81

Thus we can get the three-levels of subjectivity in Figure 2, by combining the two points of comparison, that is, whether or not there is reference to the ground, and when there is, whether the reference to the ground is implicit or explicit. And this three-level scale works to assess the degree of subjectivity of linguistic expressions cross-linguistically, i.e. for the first criterion in (la). The second criterion (lb) is rather self-explanatory. A basic motion verb come is a subjective verb (Langacker 1985), and those verbs with the same semantic function in other languages are equally subjective verbs. However, "come" verbs in some languages might be used in almost all occasions where basic motion events are expressed, while those in other languages might be seldom or less frequently used than subjectivity-neutral verbs like move. Thus, having subjective expressions alone does not tell us enough. We also need to compare them as to how preferentially they are used in the language. Obligatoriness/preferentiality of use of linguistic expressions/markers is naturally a matter of degree (see also Langacker's discussion of optional - preferential - obligatory deixis (1985: 118)). One typical example of grammatical markers which are "obligatory" in the sense of linguistic typology is an addressee-honorific marker of social deixis in Japanese. The masu form is attached to a verbal predicate form to indicate the polite style of the utterance in the language while the plain style is indicated otherwise. In other words, every verbal utterance is marked whether it is in the polite or plain style, and in that sense the speaker's psychological distance from the addressee is marked obligatorily in the language.

1.2.

Types of expressions to be examined in this study

To investigate the cross-linguistic patterns of variation in linguistic subjectivity, we will use the criteria in (1) and examine three types of expressions, which all tend to display cross-linguistically variant degrees of subjectivity. They are:

reviewer's other disagreement with the current analysis regarding Japanese A letter came and its English translation I received a letter in Section 2.1.2.

82

Satoshi Uehara a. expressions of motion events b. expressions of mental (internal) states c. sociological expressions of nominal reference

These expressions can be subjective in different ways, and the differences come from the role the speaker plays with respect to the conceptualization he/she entertains. The three types can be characterized in terms of the speaker's role (cf. Ohye 1975; Iwasaki 1993). Thus the speaker functions as the point of reference for motion deixis in a), as the experiencer of mental (internal) states in b), and as the point of reference for social deixis in c).

In the next three sections, we will examine the three types one by one.

2.

Deixis in expressions of motion events

Motion events here refer to all kinds of events where some object changes its location. They thus involve at least one moving object and the path/direction of its movement. Using a Cognitive Grammar-style representation, the semantic structures of motion events can be schematically represented in Figure 3.

Figure 3. Schematic semantic structure of movement expressions The bold line indicates the "profile", the most salient part of its meaning. The hatched line indicates that the moving object is only schematically specified, in other words, it is like an empty slot to be filled or "elaborated" by some concrete object in each instance of use of the expression.

Toward a typology of linguistic subjectivity 2.1.

83

Deixis in expressions of basic motion events

Basic motion events refer to those motion events which involve only one participant that is moving. Thus that participant functions as the trajector/subject of a motion event. Expressions of basic motion events are classified into two types in this section: basic motion verbs (e.g. move, come, go) and those other, somewhat less basic ones which rather indicate some other aspect of motion such as manner.

2.1.1. come/go A typical example of a subjective expression of this basic type is comet go in English (see a relevant discussion in Fillmore 1966). Langacker (1985: 115) notes The verbs come and go are ... strongly deictic, in that they invoke the ground as reference point on a preferential basis. Come in particular resists the construal of a non-ground element as its goal.... This is illustrated by the restriction that the verb come is subjected to as in (4): (4)

a. Let's move / go over there. b. * Let's come over there. (Fillmore 1966)

The semantic structure of the English verb come is thus represented as in Figure 4, where the arrow indicates the direction of the motion and G, the ground, is located at its endpoint (i.e. goal):

tr: trajector lm: landmark G: ground

Figure 4. Semantic structure of come in English.

84

Satoshi Uehara

Unlike the verb move, which reports the agent's motion objectively, i.e. without reference to the ground, the verb come has "the ground in its scope of predication" (see Langacker's definition above) and is properly called a subjective verb. Now moving away from English, we find that many other languages possess "come"/"go" verbs which behave similarly and appear to have the same subjective semantic structures as English come/go represented in Figure 4. Japanese kuru 'come', for example, has restrictions similar to those for English come as shown in (5), and therefore we can analyze it as having the same semantic structure as English come in Figure 4. 6 ikulkuru ('go'/'come') in Japanese (cf. (4) above): (5)

Saa, asoko e iki-masyoo. / *ki-masyoo. Well, there to go-let's come-let's 'Well, let's go/*come over there.'

Looking at "come" and "go" from a wider cross-linguistic perspective, however, we find that the lexical distinction cornelgo for the basic motion expression is not universal, and we encounter some languages without "come" (no lexical entry for the subjective basic motion expression). They are Jinghpaw, Rawang and Russian (DeLancey 1981, 1985).7 6. Some readers might point out a difference in use of coming verbs between some languages like English, where come is used when the addressee, instead of the speaker, is at the goal of motion, on the one hand, and other languages like Japanese, where such use of kuru 'come' is not available, on the other. I analyze this classification as representing a sub-classification of the typology that I present here, for the following two reasons: 1) both language types commonly possess a "come" verb, as opposed to the "other" types of languages introduced right below which do not; and 2) such use in languages of the former type can be analyzed as what I call "perspective shift" use of a subjective "come" verb, which can serve for some language-specific functions such as politeness, and except for the special use, the rest of usage patterns of such "come" verbs are in parallel with those of "come" verbs in the languages of the latter type (see Uehara, in preparation b). 7. One of the reviewers has pointed out that the honorific language of Japanese also does not distinguish between "go" and "come"; e.g. irrassyaru 'go/come/be.HON' and mairu 'go/come.HUM", and suggested that the ordinary language and the honorific language be typologized separately, t would rather account for the phenomena in question in terms of (behavioral) markedness in typology: Honorifics are a marked category, and therefore, the deictic distinc-

Toward a typology of linguistic subjectivity

85

DeLancey (1985: 367), in his discussion of "directive" verb systems in Tibeto-Burman, notes: In a language with an extreme version of the directive system, there are not two deictically specified lexical items 'go' and 'come', but a single unspecified motion verb ....

and mentions Jinghpaw and Rawang. Jinghpaw, where the deictically neutral motion verb string sa wa, which we can gloss for the present purposes as 'go', can be deictically specified (like any other motion verb) by means of the postverbal particles r- 'hither' and s'hence' (DeLancey 1985: 370, emphasis added).

His examples are reproduced here in (6) and (7): (6)

(7)

MaGam gat de? sa wa market to go 'MaGam went off to market.' MaGam gat de? sa wa market to go 'MaGam came to market.'

s-ai hence-DEC

r-a? hither-3rd

ai DEC

This seems to show that these languages need, to a much lesser degree, to refer to the speaker's position in describing the basic motion events, although of course it is still too early to draw any definite conclusion before we can also examine the degree of obligatoriness of such deixis specifying morphemes in these languages' discourse.

2.1.2. Other expressions of basic motion events and obligatoriness of motion deixis We saw in the previous section that the lexical come/go distinction exists in both English and Japanese, and the two do not seem to differ from each other in the degree of subjectivity they tend to exhibit in encoding basic tion in the unmarked category (plain/basic patterns) is neutralized, or "syncretized", in a more marked category (honorifics), just as English pronouns have the gender distinction (he/she/it) for the singular (unmarked), which is neutralized {they) in the plural (marked).

86

Satoshi Uehara

motion events. However, when we turn to other kinds of expressions of basic motion events, some differences between them in the obligatoriness of deixis begin to show up. Among such expressions are manner verbs and (non-subjective) directional verbs. In expressions of "goal-oriented motion" (DeLancey 1985), manner verbs in Japanese, such as, aruku, 'walk', hasiru 'run', unten-suru 'drive', unlike those in English, preferentially accompany the verbs of coming and going as in (8):8

(8)

(9)

Manner verbs in Japanese Kare wa koko e ?unten-sita / unten-site he TOP here to drove / driving 'He drove here.' (lit. 'He came here driving.')

kita. came

Directional verbs in Japanese [from Shibatani 2003: 260-261] Ken ga heya kara dete itta / kita. Ken NOM room from exit went / came 'Ken went/came out of the room.' Ken ga heya ni haitte itta / kita. Ken NOM room in enter went / came 'Ken went/came into the room.'

Directional verbs in Japanese, such as deru 'exit', hairu 'enter', display a similar strong tendency to accompany verbs of coming and going as shown in (9). In examining directional verbs in Japanese, Shibatani (2003) notes, one of the motivations for such behavior of directional verbs in the language is that Japanese, especially interactive colloquial speech, strongly prefers various kinds of coding of the speaker's stance. ... [those sentences without coming/going verbs] are felt to be not sufficiently revealing about the speaker' stance - in this case, the spatial orientation of the speaker with respect to the goal or source location of the directed motion. That is, these sentences do not give extra-propositional information that the hearer feels entitled to know (e.g., where were you when this happened?) (Shibatani 2003: 263).

8. The phenomena here have much to do with the "verb-framing" nature of the Japanese language (see Talmy 1985 for his "verb-framed" and "satelliteframed" language types). As we see below in the case study in Table 2, however, it alone cannot account for all occurrences of "come'V'go" verbs in the language.

Toward a typology of linguistic subjectivity

87

To see this difference in the obligatoriness between the two languages in a data-oriented manner, a quick search of kuru 'come' was made in a data base. The date base is comprised of 140 editor's daily essays titled "Tensei Jingo" (lit. 'heaven voice, people words') published in Asahi Shinbun, one of the leading Japanese newspapers during the first half of 1990, whose English translations titled "Vox Populi, Vox Dei" were also published in the paper's English version, The Asahi Evening News. Out of 51 occurrences of kuru 'come' for motion events9 in the Japanese original, 27 (52.9%) are translated into expressions without come in English as shown in Table 1. Some examples from the data are in Table 2. Table 1.

Japanese kuru and its English counterpart

Japanese original 51 kuru

Table 2.

English translation 24 (47.1 %)with come 27 (52.9%)without come

Some examples of kuru being translated into English without come

Japanese kaette kuru semete kuru tubame kuru tegami ga kita

Lit. translation 'come returning' 'come attacking' 'swallows come' 'a letter came'

=> => => =>

English translation return home attack swallows arrive10 I received a letter

The data clearly shows that, although the subjective coming verbs in the two languages may be equally subjective, the obligatoriness of the use of 9. This count does not include 11 occurrences of kuru for metaphorical, temporal aspects (e.g. haru ga kuru 'spring comes'), in 4 of whose English translation come was not used, and one occurrence of kuru for the inverse marking (see below), also found in the data. 10. Regarding the use of arrive and come discussed here, one of the reviewers has emphasized that "what one might call 'contextual subjectification' (e.g. the fact that arrive is understood as referring to the ground in the unfolding discourse) could result in the same degree of subjectification in the speaker's mind as the use of the more explicit (if uttered out of the blue) deictic verb come". It is a matter of course, and I am quite in agreement with him/her. However, the point in question is that the deictic verb come here necessarily invokes the ground as reference point in the speaker's mind, while the verb arrive does not.

88

Satoshi Uehara

kuru in Japanese appears to be higher than that of come in English. This indicates that the ground, or the speaker's position, is more typically invoked as a reference point in expressions of basic motion events in Japanese than it is in English. In fact, the phrase come running is seldom used in English although we all know that it is a possible sequence, while hasitte kuru 'come running' represents a normal use in Japanese. Therefore, although "some kind of use of deictic motion verbs to provide orientation for non-deictic verbs ... is probably nearly universal (cf. English come running, ...)" as DeLancey (1985: 381) notes in his study of motion deixis in Tibet-Burman, the current study shows that the tendency to use such deictic motion verbs for non-deictic verbs varies from language to language. With this cross-linguistic variation in the obligatoriness of deictic specification for basic motion events, let us look at come and go from a wider cross-linguistic perspective. We have seen some languages don't possess the verb "come" (i.e. the "come'V'go" lexical distinction). Most other languages seem to have verbs of coming and going for basic motion events. When we look at expressions of specific motion, such as flying, running, swimming, we easily find many languages rather like Japanese. In Chinese, Thai, and Yao (Matisoff 1991), for example, verbs of specific motion are typically or preferentially expressed with deictic verbs (e.g., Chinese: fei läi 'come flying', pao Ιάί 'come running'. Also, those in Korean require verbs of coming and going in expressions of goal-oriented motion, resembling the Japanese pattern as in (10):

(10)

2.2.

Korean: (Sangmok Lee, Ce salam-un yeki-ey the man-TOP here-to 'The man walked here.'

p.c.) lkel-ess-ta / kel-e wassta. walked / walking came (lit. 'He came here walking.')

Deixis in expressions of motion events in which the primary landmark (DO) is in motion

Motion events include not only those in which the trajector/subject of the event is in motion, but also those in which the landmark/direct object of the action is in motion. And such motion events can also encode deixis. They can be classified into two subtypes: transportation and transmission events, which are separately examined in the two sections below.

Toward a typology of linguistic subjectivity

89

2.2.1. Bring/take (deictic verbs of transportation) Verbs of transportation bring/take in English, for example, are very like come!go, as Fillmore (1966: 226) notes "(t)he characteristics of and go as revealed in these rules fit, in the author's speech, the pair and take as well, but apparently no other words in the language." He the restriction in the use of bring as shown in (11), similar to that for in (4) above:

(11)

much come bring notes come

? ?Let's bring these over there.

Thus, bring is also a deictic verb, whose semantic structure is schematically represented in Figure 5. Its semantic structure resembles that of come in that the ground is specified "in its scope of predication", namely, at the goal of motion (unlike carry, for example). The difference between bring and come is that bring presupposes two participants, namely, the action doer (trajector) or bring-er, which initiates the action of transportation, and the moving object (primary landmark) or bring-ee, which is transported. 11

Im

lm

Figure 5. Semantic structure of bring in English In Japanese, expressions of transportation events are not lexicalized as those in English, 12 but are bi-morphemically expressed in the form of verbs 11. Fillmore notes some difference in the degree of restriction for bring and come: "there are many who would accept Let's bring these over there who would not accept Let's come over there." (1966: fn. 15). This fact indicates that the ground is less obligatorily accessed, and the degree of entrenchment of the ground is less, in bring than in come. 12. The fact that the two morpheme sequences in Japanese "bring" are somewhat lexicalized is evidenced in their accentual contour: they together form a single accentual contour in Tokyo standard Japanese.

90

Satoshi Uehara

of holding obligatorily accompanied by deictic verbs of coming and going as in (12): Table 3.

bring!take in Japanese

English: Japanese: moving object things people

BRING =HOLD+COME motte kuru turete kuru

TAKE =HOLD+GO motte iku turete iku

(12)

Sore ο koko e *motta / motte kita. it ACC here to held / holding came '[He] brought it here.' (lit. '[He] came here holding it.')

(13)

Saa, asoko e kore ο motte iki-masyoo. / *ki-masyoo. Well, there to these ACC having go-let's /come-let's 'Well, let's take/??bring these over there.'

Japanese motte kuru 'bring', just like kuru 'come' in it, resists the construal of a non-ground element as its goal as shown in (13). Thus, although deictic verbs of transportation in Japanese take a bi-morphemic form, they seem to be used no less obligatorily than those in English (cf. 11).13 Table 4.

Bi-morphemic expressions of bringing/taking (from Matisoff 1991: 438)

Yao Thai Mandarin Japanese

BRING =HOLD+COME töo täay aw maa ηά läi motte kuru

TAKE =HOLD+GO töo miing [o=open o] aw paj ηά qü motte iku

13. There are many uses of bring which are not deictic and thus cannot be translated into motte kuru in Japanese, suggesting that the latter is more exclusively deictic: English: Japanese:

He brings good luck to everybody he touches. ... subetenohito ni ?motte kuru / motarasu everyone to bring / bring about

Toward a typology of linguistic subjectivity

91

Looking at other languages, we find that bring/take are expressed, especially in many serial verb languages, in the same bi-morphemic way, as observed in Japanese. This is shown in Table 4 above. Korean expressions of transportation events show the same pattern and probably the same degree of obligatoriness of deixis as those in Japanese as illustrated in (14): (14)

kukes-ul yeki-ey *kacyessta / kacye wassta (cf. (12)) it-ACC here-to held / holding came '[He] brought it here.' (lit. '[He] came here holding it.')

2.2.2. Verbs of sending (expressions of transmission deixis

events) and motion

Motion events which are similar to transportation events but slightly different from them are transmission events. Expressions of transmission events, or verbs of sending, resemble expressions of transportation, but differ from them in that the trajector/agent of transmission events only initiates the movement and does not change his location. Thus, the semantic structure of the English verb send can be schematically represented in Figure 6:

As Figure 6 shows, the ground is not specified in the semantic structure of sending verbs in English (e.g. Let's send these over there) and other languages, and so the speaker can be expected to be at any location of the transmission event without changing any form in the utterance (e.g. I sent it to them, or They sent it to me). In some languages, however, those trans-

92

Satoshi Uehara

mission events where the speaker is located at the goal of motion are typically or obligatorily marked differently from the other transmission events. Such languages can be said to possess the constructions whose semantic structure is that in Figure 6 but with the ground in its secondary landmark position, distinct from other expressions of transmission. Japanese is one such language, and Shibatani (2003) identifies, as a marker of the inverse system14 of Japanese, the kuru ('come') marking in some constructions with verbs of sending, where the kuru is "obligatorily used when the recipient is the speaker."15 His examples (Shibatani 2003: 273-274) are reproduced here in (15)—(17): (15)

a. Ken-ga Ken-NOM 'Ken threw b. Ken-ga Ken-NOM ki-ta

Hanako-ni booru-o nage-ta. Hanako-to ball-ACC throw-PAST the ball to Hanako.' boku-ni booru-o *nage-ta. / nage-te me-to ball-ACC throw-PAST / throw-CONJ

come-PAST

(16)

'Ken threw a. Ken-ga Ken-NOM 'Ken wrote b. Ken-ga Ken-NOM ki-ta

me the ball.' Hanako-ni tegami-o kai-ta. Hanako-to letter-ACC write-PAST a letter to Hanako.' boku-ni tegami-o *kai-ta. / kai-te me-to letter-ACC write-PAST / write-CONJ

come-PAST

(17)

'Ken wrote me a letter.' a. Ken-ga Hanako-ni denwa-o si-ta. Ken-NOM Hanako-to telephone-ACC do-PAST 'Ken made a phone call to Hanako.'

14. See Givon (1994) and Thompson (1994) for a cross-linguistic definition of the inverse system. 15. This phenomenon in Japanese seems to be less surprising if we realize that the pattern here for verbs of transmission events resemble that for verbs of specific motion (Section 2.1.2), which the language already has.

Toward a typology of linguistic subjectivity

93

b. Ken-ga boku-ni denwa-o *si-ta. / si-te Ken-NOM me-to telephone-ACC do-PAST/do-CONJ ki-ta. come-PAST 'Ken made a phone call to me.' If we turn to languages other than Japanese, the existence of similar constructions in Chadic languages (Afroasiatic family) is also noted by Shibatani (2003), to which he compares Japanese inverse constructions (See also Frajzyngier 1987). Korean (Choi Hyun Choel, p.s.) also allows the inverse pattern for transmission events similar to that in Japanese (e.g., ponay o-ta 'send-CONJ come-DEC' for 'send (letters, presents, etc.) to me'), although the pattern seems to be a little more restricted (e.g. In Korean, the translation equivalent of the Japanese sentence with "come" in (15b) is not acceptable). Also in Thai (Smyth 2002: 61), the verbs of coming and going (maa 'come' and pay 'go') are used after a number of verbs to indicate whether the action of the verb is directed towards or away from the speaker, and such verbs include thoo 'telephone' and song 'send', i.e., verbs of transmission here.

2.3.

Deictic expressions of transaction events (deictic verbs of giving)

Fillmore notes in his quote in Section 2.2.1 above that in English there exist only two deictic motion pairs (basic motion verbs cometgo and verbs of transportation bring!take). We have seen that in some other languages subjective expressions of specific motion and transmission events are conventionalized as well. In still rarer cases, even expressions of transaction events, namely, verbs of giving, can also be deictically marked in some languages. Verbs of giving are extensively analyzed in Newman (1996), and his representation of the semantic structure of giving verbs is reproduced here in Figure 7. Transaction events (Figure 7) resemble the transmission events that we examined in the previous section (Figure 6), but differ from them drastically: the motion in the former is metaphorical and does not have to be physical at all. Thus, when I say My father gave the pen to me, the pen can actually change its location from my father to me, but when I say My father

94

Satoshi Uehara

gave the land to me, the land does not. It changes its "location" from my father's "sphere of control" to mine.

someone's sphere of control

Figure 7. Semantic structure of give (from Newman 1996: 47) For English give, Japanese has two verbs, kureru 'give to the speaker' and ageru 'give to others' (Kuno 1973, 1978; Ohye 1975, inter alia), and they behave very much like verbs of coming and going as shown in (18): (18)

a. Mary ga boku ni NOM I s DAT 'Mary gave me flowers.' b. Boku ga Mary ni DAT lSG NOM Ί gave Mary flowers.

hana flower hana flower

ο ACC ο ACC

kure-ta / *age-ta. give-PAST * kure-ta / age-ta. give-PAST

Therefore, the semantic structure of kureru 'give to the speaker' in Japanese is the one which has the ground added to the secondary landmark position of that of English give in Figure 7. It has "the ground in its scope of predication" just like English come represented in Figure 4. Deictic verbs of giving are rather rare cross-linguistically, and Yamada's (1996) cross-linguistic study (over 30 languages) of benefactive constructions identifies only Japanese as the only example of a language having such a giving verb pair. Deictic verbs of giving are not present in Korean, the closest parallel to Japanese otherwise, or in Thai (Eda 1983). 16

16. Cross-linguistically, the deictic specification of direction of giving is extremely rare compared to basic motion deixis. I should also add here that there are some intra-linguistic variations in Japanese, that is, some dialects of Japanese lack the distinction in question (NLRI 1966, Hidaka 1994).

Toward a typology of linguistic subjectivity

95

Newman's (1996) work on give, however, discusses two other languages with some similar deictic distinction in expressions of giving, and they are Maori and Iku (Kuliak family, Uganda). In the following quote from Newman (1996: 22-23), he describes the situation in Maori: In Maori, directional morphemes mai 'hither' and a tu 'thither' have become integrated with a GIVE verb stem ho, resulting in the two verbs hömai 'give to speaker' and höatu 'give away from speaker'. Ho in Maori is not an independently occurring free morpheme anymore. His Maori examples are reproduced in (19): (19)

a. Kua hömai ia i te pukapuka PAST give he/she OBJ ART book 'She/he gave the book to me.' b. Kua höatu au i te pukapuka ki ä PAST give I OBJ ART book to ART Ί gave the book to him/her.'

ki ahau. to me ία. him/her

Thus, the Maori language possesses a lexicalized verb of giving that has the ground specified at the recipient position, exactly like kureru in Japa17 nese.

17. No references of Maori grammar that I consulted discusses in any systematic manner deictic expressions of motion events in the language. Ryan (1999), however, in the "directional adverbs" section of his modern Maori textbook, gives several sentences of motion events without directional adverbs, and writes that such sentences are "grammatically correct, but they would sound much better with what are called (directional) adverbs" (p. 127, my translation). Furthermore, the textbook lists in its glossary many verbs (e.g. hoki 'return', hou 'enter', hui(-a) 'gather', karero 'convey', mea(meinga) 'tell') directly followed by mai 'hither' or atu 'thither', suggesting that they and the deictic adverbs together are idiomaticized, if not lexicalized as homai 'give to me'. The pattern here suggests that the language strongly favors the use of the deictic adverbs with expressions of motion events in general, and also that the deictic verbs of giving alone may have been lexicalized because of the phonetic instability of the "give" verb stem.

96 2.4.

Satoshi JJehara Summary of deixis in expressions of motion events

The subjectivity cline in Figure 8 summarizes our discussion in this section on deixis in the expressions of motion events [the "?" mark next to language names indicates that the languages are yet to be conclusively placed on the subjectivity cline until more detailed, descriptive work on the relevant phenomena is made available]: subjective -4 Japanese, Maori? Ik? # of subjective patterns « degree of obligatoriness

m

Korean, Chinese? Thai? Yao? ^

objective English, Jinghpaw? Rawang? Russian?

Figure 8. The subjectivity cline in expressions of motion events It should be noted in passing that N e w m a n ' s description of the Maori case and DeLancey's work on the grammaticalization path of directive motions in Tibet-Burman suggest that we can, and need to, examine the diachronic aspect of this subjectivity cline. Such changes in the position on the subjectivity scale will be discussed in Section 5 together with similar phenomena in other types of subjective expressions as instances of subjectification/objectification processes. The discussion so far also leads us to see that the conventionalization of deictic specification is cross-linguistically more readily attested in expressions of basic motion events than in events of less basic motion. In such cases, typologists would often come up with a kind of generalization called "an implicational universal" which, in this case, reads as follows: The existence of (obligatory) deictic markers of less basic motion events (such as giving, transaction events) necessarily implies the existence of (obligatory) deictic markers of basic motion events (coming/going). This implicational universal is yet to be tested over typologically diverse samples of languages, but a tentative account for the tendency observed here can be suggested as follows: Basicness here means simplicity in terms of conceptual complexity of the event (the number of profiled participants, see Figures 4-7), salience in the figure/ground organization of event conceptualization (trajector is "the primary figure within a profiled relation", Langacker

Toward a typology of linguistic subjectivity

97

1991: 555), and/or physical in the physical/nonphysical distinction of motion.18 This is summarized in Table 5: Table 5.

Characteristics of motion events examined

kind of events basic motion specific motion transportation transmission transaction

# of participants 1 1 219 2 3

moving object trajector trajector landmark landmark landmark

motion physical physical physical physical non-physical

18. One of the reviewers has pointed out that this implication may be contradicted by the Japanese honorific language, where there is a distinction between sasiageru 'give to an honorified other' and kudasaru 'an honorified other gives me', but there is not distinction between going and coining expressed in irassyaru 'go/come.HON' and mairu 'go/come.HUM'). I would account for this phenomena in the following manner: First, as I argued earlier, honorifics do not form a separate language from the ordinary language of Japanese, but are a marked category of it with the plain/basic patterns as its unmarked category (just as the plural is a marked category vis-ä-vis the singular), therefore, it is not surprising that the deictic distinction in the unmarked category of plain patterns be neutralized in the marked category of honorifics. However, unlike basic motion events, honorification of "giving" events retains the deictic distinction. This is because of the difference in the number of participants between the two types of motion events: Unlike basic motion events, transaction events involve two (human) participants (see the next footnote). Because the speaker never becomes the target of honorification (at least in Japanese), the deictic directionality of giving always corresponds with the directionality of honorification. When the speaker is the recipient, only the giver can be the target of honorification (kudasaru 'give.HON to me', while only the recipient can be the honorification target when the giver is the speaker (sasiageru 'give.HUM' to an honorific other.) 19. The goals of transportation and transmission do not count as a participant, while those of transaction do. This is because the goals of the former are typically construed as a location (just as those of basic motion are, e.g. come), while those of the latter are typically construed as a person (the basic meaning of giving is transfer of possession between two people). Thus, give is a typical example of ditransitive verb while, for example, the verb throw is not. Interestingly, this conceptual departure of the latter from the former seems to iconically motivate the formal departure in languages like Japanese.

98

Satoshi Uehara

Deictic specification is extra information to be additionally encoded in expressions of motion events, and so, the more basic the motion events are, the less processing load is required in the construal of a single event. Thus, a tendency results from this. Accordingly, the conventionalization of deictic specification in expressions of non-basic motion events implies the normality, and the importance, of evoking the ground in the motion event conceptualization in the language.

3.

Deixis in internal (mental) state predicates

The second type of deictic expressions to be examined here are those of internal/mental states. It is widely known that Japanese possesses a group of predicates which we can call internal state predicates, which make clear distinctions in predicate morphology, depending on whether the experiencer of the mental state is the speaker or not (Kuroda 1973; Kuno 1973; Iwasaki 1993; Uehara 1998b, etc.). They denote, by default, the speaker's internal mental states, such as intentions, mental processes (e.g. I think), sensations (e.g. I am cold), emotions (e.g. I am sad), and desires (/ want), and, unlike their close equivalents in English, they require that their subject be in the first person if they are used in simple declarative sentences. An example is uresii 'glad', illustrated in (20): (Watasi wa) lSG TOP Ί am glad.' *Tanaka-san Ms. Tanaka 'Ms. Tanaka is

uresii glad

20

wa uresii TOP glad glad.'

20. It should be noted that the first person subject of internal state predicates is by default unexpressed or implicit in Japanese discourse (Uehara 1998a), which is indicated by the parenthesis in the example. Recall that those expressions with implicit reference to the ground correspond to the highest of the three levels in subjectivity scale in Figure 2. In contrast, in English the translation equivalent of (20b) as well as that of (20a) are acceptable, and the experiencer subject cannot be implicit.

Toward a typology of linguistic subjectivity

(20')

99

Tanaka-san wa uresi-sooda. / uresi-gatte irn/ uresii-yooda. Ms. Tanaka TOP glad-appear / glad-show signs of / glad-seem 'Ms. Tanaka looks glad/shows signs of being glad/seems to be glad.'

They do not take the third person subject, without being accompanied by evidential markers that indicate how the speaker acquired the information. The logic behind this is that the speaker has direct access to his/her own internal states but not to the internal states of others'.21 Thus, to express the proposition in (20) in Japanese, one has to make use of some kind of morphological marking and say as in (20'), which can be roughly translated as 'Tanaka looks glad', 'Tanaka is showing the signs of being glad', etc. Such grammatical behavior leads us to assume that Japanese internal state predicates have the speaker as the experiencer of such states in the semantics of them and are thus deictic, and their semantic structure can be schematically represented as shown in Figure 9:

OS: G:

onstage region (immediate scope of predication) Ground

Figure 9. Semantic structure of subjective predicates in Japanese (Uehara 1998b; slightly modified from Langacker 1985, 1991a)

21. The second person subject patterns after the third person subject since the speaker has no direct access to the internal state of others, the addressee or the third person. However, in interrogatives, the second person subject patterns after the first person subject, because with interrogatives, the speaker asks about the hearer's internal states to which the hearer has direct access. This patterning of the second person subject applies to all the internal state predicates with the restriction in question throughout the paper.

100

Satoshi Uehara

In other words, the argument we set forth here is that internal state predicates in Japanese are subjective expressions just like come in English (or kuru 'come' in Japanese, for that matter). The parallelism between the two types of expressions is obvious: The experiencer in the former and the goal in the latter are both specified as the ground, which can be made explicit if required by the context, as exemplified in (20a) and in (21) below, respectively: (21)

He came (here/to me).

And they both resist the construal of a non-ground element as the experiencer in the former and as the goal in the latter, as shown in (20b) and (4b), respectively. In sum, they are parallel in having the ground in their scope of predication,22 as represented in Figure 7 and Figure 4, respectively. English does seem to show some sensitivity in this respect as well and possess some predicates which exhibit similar behavior (Benveniste 1971, etc.),23 but in Japanese the person restriction seems to be more apparent and the number (of types) of predicates with such behavior seems to be larger.24 Table 6 shows the grammatical pattern which expressions of thought processes in Japanese, such as omou 'think', exhibit. It indicates that omou is also a subjective predicate in Japanese. Table 6.

Person restriction of a subjective predicate omou 'think' in Japanese

predicate subject 1 st person 3rd person

omou 'think' ok *

omotte-iru 'be thinking' ok ok

22. Of course, there are differences as well because they belong to expressions of different types of events, irrespective of their ground specification. For example, the former denotes a stative event while the letter a motion event 23. Benveniste discusses such meaning differences as I swear/he swears in English. Subjectivity is a matter of degree. 24.1wasaki (1993: 12-13) also notes, in his seminal work on Subjectivity in Grammar and Discourse, that "Japanese grammar is very clear in distinguishing between a situation in which the speaker explains his own internal states and that in which he explains others' such states."

Toward a typology of linguistic subjectivity

101

Some examples are given in (22): (22)

a. Watasi wa asu wa ame da to omoimasu. / I TOP tomorrow TOP rain COP QTthink.POL/ omotte-imsu. be.thinking-POL Ί think it will rain tomorrow.' b. Kare wa asu wa ame da to *omoimasu. / He TOP tomorrow TOP rain COP QT think.POL / omotte-imasu. be.thinking-POL 'He thinks it will rain tomorrow.'

Before looking at cross-linguistic patterns of variation of deixis in internal state predicates, let us note that there exist two confusing kinds of linguistic behavior of internal state predicates, which are similar to, but should be differentiated from, the phenomena that we are looking at here. They are: i) the so-called non-canonical case marking patterns of internal state predicates (e.g. Spanish: experiencer-object pattern) (cf. Croft 1991) Me gusta Maria. Ί [Obj.] like Maria.' (cf. Le gusta Maria. 'He/She [Obj.] likes Maria.') ii) perspective phenomena in sentence complement constructions (e.g. French: the complements of volition verbs, e.g., Achard 1996): a. *Je veux que je revienne.' I want that I come back.' b. Je veux revenir. Ί want to come back.' (Cf. Jean veut revenir.'John wants to come back.') As can be seen in the examples here, neither type of behavior concerns the person restriction of internal state predicates in their main predicate position. That is, they both concern the propositional content, or what is in the objective scene, alone, and have no direct relevance to the question of whether or not the speaker, the ground, is specified in it. Whether or not the speaker is specified is something we are very much concerned with here. When we look at internal state predicates from a wider cross-linguistic perspective, we will soon find that Japanese is not unique in this respect.

102

Satoshi Uehara

The facts of person restriction very similar to what is observed in Japanese, can also be found in internal state predicates in Korean, as shown in (23): Korean (No 1989, Sangmok Lee, p.c.) : (23) a. Na nun kippu-ta. 1SG TOP glad Ί am glad.' b. *Kim-ssi nun kippu-ta. Ms. Kim TOP glad 'Ms. Kim is glad.' There are also other languages, such as Angam-Naga (Tibet-Burman) and Newari (Tibet-Burman), that show similar person restriction, but in much less conspicuous ways. In these languages, such person distinctions are morphologically marked (i.e. prefixed) with internal state predicates only, while no such marking exists for the predicate in general. Thus, only such predicates are required to indicate clearly whom they are predicated of. The pattern in Angam-Naga (Giridhar 1975: 59-60, reproduced from Iwasaki 1993: 84) is shown in (24): (24)

a. ä ä-ηί bä I 1-happy AUX Ί am happy.' b. puö puö-ηί bä he 3-happy AUX 'He is happy.'

See also Newari in DeLancey (1987: 63-64), which has a similar marking for volition predicates. The expression of a thought process xiang 'think' in Taiwan Mandarin (Sanders, et al. 2000), unlike its counterpart in Mainland (i.e. Beijing) Mandarin, exhibits a person restriction very similar to that in Japanese expressions of thought processes, when it takes a complement clause. In examining the use/non-use of the verb shuo 'say' as a clause-initial complementizer - similar to the English complementizer that - for the verb xiang 'think' in Taiwan Mandarin (its complementizer use is not acceptable in Beijing Mandarin, cf. Cheng 1985), Sanders et al. (2000) found some person restriction: the use of the complementizer shuo is optional for

Toward a typology of linguistic subjectivity

103

the first person's thought process, while it is obligatory for that of the third person's. This is shown in (25): (25)

a. Wo xiang / xiang shuo yihuir keneng hui xiayu. I think / think that soon possibly can rain Ί think (that) it will probably rain soon, b. Ta *xiang / xiang shuo yihuir keneng hui xiayu. He think / think that soon possibly can rain 'He thinks (that) it will probably rain soon.

The usage pattern of the complementizer shuo according to the person of the subject is summarized in Table 7 (Compare the Taiwan Mandarin pattern with that in Japanese in Table 6): Table 7.

Person restriction of the verb xiang 'think (that)' in Mandarin (Sanders et al. 2000)

predicate form literal translation 1 st person subj. 3rd person subj.

Taiwan Mandarin xiang xiang shuo 'think 'think say' ok ok * ok

Beijing Mandarin xiang xiang shuo 'think' 'think say' ok * ok *

Now, by analyzing these cross-linguistic patterns in terms of the subjectivity scale, we can roughly classify the languages examined into three levels: First, those languages having a morphological person distinction for the subject/experiencer role of internal states (such as Japanese, Angam-Naga and Taiwan Mandarin) are higher on the subjectivity scale than those without that distinction. This is because internal state predicates in the former necessarily invoke the ground, the notion of speaker, as the experiencer of internal states. The morphological distinction is, thus, analogous to the deictic distinction in motion verbs discussed in the previous section, and these ground-sensitive internal state predicates are subjective expressions just as deictic motion verbs are. Among the languages with the subjective internal state predicate patterns, Japanese (and probably Korean as well) can be analyzed to be still more subjective than the rest of the languages. There are three reasons for this. Firstly, (the types of) such internal state predicates appear to be more

104

Satoshi Uehara

numerous in the former languages. Secondly, in terms of structural markedness criteria in typology, the predicate form for the speaker's internal states is structurally unmarked while that for other people is structurally marked in the former languages, as seen in the Japanese case in (20) and (20'). This means that in Japanese, the expression of the speaker's internal states is the norm, and in expressing other people's mental states, the speaker has to indicate that by adding extra morphemes. And lastly, the subjective pattern in the former languages is rendered even more subjective by the criterion in terms of the implicit/explicit mention of the ground. The first person pronoun subject of internal state predicates is usually omitted in Japanese discourse. Uehara (1998a) has examined all the personal pronouns in an English novel and their counterpart expressions in its Japanese translation, and shown that the first person subjects of internal state predicates are unexpressed in Japanese unless they are emphasized for focus or contrast. The logic behind this is simple: the unmarked form of internal state predicates necessarily means the first person for their subject, so there is no need for it to be mentioned in Japanese, a so-called zero anaphoric language.25 Thus, these facts about the internal state expressions in Japanese (and Korean), that they always invoke the ground and that the ground is implicit by default, make them even more subjective. The subjectivity cline of expressions of internal states in Figure 8, summarizes our discussion in this section:

25. In this connection, I should note another work of mine, Uehara, in preparation a, which compares the Japanese result in Uehara (1998a) mentioned above, with that of a Chinese translation of the same novel and shows that this tendency to omit the first person subjects for internal state predicates is greatly reduced in Chinese. This fact seems to parallel, and give a partial account of, a comment by Chinese students in my Japanese language class, which is, they often have trouble finding the correct subjects of certain types of Japanese sentences, which later turn out to be the first person of internal state predicates. Further in this connection, let us consider the non-omissibility of subject argument in English. It follows from the argument here that the configurationality, or the default grammatical pattern of formal presence or explicit mention of subject/object arguments in English is parallel to, and possibly contributes to, the language expressions' leaning toward the objective side of the subjectivity scale.

Toward a typology of linguistic subjectivity subjective -4 Japanese Korean? # of predicates distinction of Sp. implicitness of Sp.

Angam-Naga Newari Taiwan Mandarin

105

—objective English Mandarin Chinese?

Figure 10. Subjectivity cline of expressions of internal states

4.

(Social) deixis in expressions of n o m i n a l r e f e r e n c e

T h e third and last type of deictic expressions to be examined in this research involves expressions of social deixis in nominal reference. Let us begin with F i l l m o r e ' s discussion of honorific systems w h i c h is quoted below: The notion of deixis might be extended, for example, to include the socalled 'honorific systems' found in many East Asian languages, systems of categories by which the speaker reveals his relation of respect or his judgment of his social worth relative to the hearer or to the object of his speech. (Fillmore 1966: 220 fii. 2; underline added). T h e point most relevant to our discussion here in the quote is the underlined part, that is, "the s p e a k e r ' s relation of respect or his j u d g m e n t of his social w o r t h relative ... to the object of his speech." This is w h a t is called " r e f e r e n t h o n o r i f i c s " in the politeness/honorifics literature (Martin 1964; C o m r i e 1976; B r o w n and Levinson 1987; Tokieda 1941; T s u j i m u r a 1963; cf. " a d d r e s s e e honorifics"). 2 6

26. As one can see, this study does not deal with addressee honorifics, which encode, using Fillmore's words, "the speaker's relation of respect or his judgment of his social worth relative to the hearer" (underline added). This is because addressee honorifics are concerned rather with the issues of "intersubjectivity" (e.g., Benveniste 1971, Traugott and Dasher 2002), which are closely connected with (especially in this third type of subjective expressions), but are different from, the issues of subjectivity as defined here. In other words, unlike referent honorifics, addressee honorifics have as the target of the speaker's social/subjective attitude, the hearer, that is, something external to the proposi-

106

Satoshi Uehara

Since referent honorifics mark the speaker's subjective attitude toward someone he/she is talking about, they necessarily refer to that "someone", some participant involved in the denoted event, whether or not the referent is explicitly mentioned. Thus, the semantic structure of referent honorifics resemble that of nominal expressions, and the former differs from the latter in having the speaker (the ground) and his/her psychological attitude specified in it. By using SA (short for social/subjective attitude) to represent the conventionalization of the speaker's subjective attitude toward the referent, such as respect, those expressions of referent honorifics can be said to have the semantic structure as represented in Figure 11:

: correspondence Figure 11. Semantic structure of expressions of social deixis (referent honorifics)

Figure 12. Semantic structure of tu (2nd pers. pronoun) in Spanish (modified from Langacker 2001: 148)

One example of referent honorifics familiar to us is the so-called ΤΝ second-person pronouns in many European languages. Personal pronouns are deictic by themselves, but they become social-deictic as well, by conventionally expressing the familiarity/solidarity distinction, that is, the speaker's psychological attitude toward the referent, the hearer in this case. 27 Some T/V pronouns are listed in Table 8:

tional content, and in that sense differ from those subjective types discussed so far. 27. TV second-person pronouns are examples of referent honorifics, not of addressee-honorifics, because they cannot indicate the speaker's familiar-

Toward a typology of linguistic subjectivity Table 8.

polite familiar

107

European T/V languages referred to in Brown and Gilman (1968) Spanish usted tu

French vous tu

Italian Lei tu

German Sie du

English you {thou)

Figure 12 represents the semantic structure of such T/V pronouns, where the speaker's subjective attitude toward the referent (the hearer, in this case) is specified and conventionalized. The T/V distinction is grammaticalized and thus obligatory in these languages (except English); whenever the addressee is referred to in the denoted event, the speaker has to choose between the two forms, and there is no social-deictically neutral form for second person reference. 28 In contrast, English has lost one of the two forms, so the social deictic specification is not obligatory in English. Such grammaticalized social deixis in the forms of second person pronouns can be found in other languages like Japanese as well. Ide (1975) posits 5 forms (and thus 5 levels of politeness) for the second person pronouns in present-day spoken Japanese. Thus, Japanese has a social deictic pronominal system, whose distinction is much subtler. More notably, however, Japanese also possesses a rich system of verbal morphology, which marks those events in which the speaker shows his/her respect to the participants. 29 The predicate referent honorifics are more typical as referent honorifics than the T/V pronouns just discussed above in that they mark the speaker's subjective attitude toward any referent; not just the second person referent, but also the referent who is not present during the speech event. Thus, in (26) below, for example, the productive o- -ni naru verbal morphology indicates that the speaker holds his/her reity/solidarity toward the addressee without "referring" to him/her (see also Comrie 1976; Brown and Levinson 1987). This crucially differs from polite style morphemes in Japanese or polite style particles in Thai, typical examples of addressee-honorifics, which can be attached to every utterance, whether the utterance's denoted event involves the addressee or not. 28. To be precise, T/V distinctions are neutralized when there is more than one addressee (except for Spanish). In these languages, the T/V distinction is made only for the second person singular, not for the plural. 29. This would make some sense if we consider the so-called "pro-drop" nature of the language: pronouns are frequently dropped from Japanese utterances, so whether the pronominal form is present or not, the speaker's social/subjective attitudes are marked with the predicate form.

108

Satoshi Uehara

spect/gratitude toward the doer of the action denoted by the verb, the speaker's teacher in this case, who is not present at the time of his speech: (26)

Sensei ga hon ο o-kaki-ni nat-ta. teacher NOM book ACC write-HON-PAST 'My teacher honorably wrote a book.'

The parallelism between referent honorifics and deictic motion verbs should be made clear here. Referent honorifics are like the deictic verb go, and just as the latter cannot have the speaker at the goal of motion as shown in (27), the former cannot have the speaker at the target of honorification (the doer/subject of the action in the case of (26) above) as shown in (26'): (27)

*He went to me.

(26')

* Watasi ga hon ο I NOM book ACC Ί honorably wrote a book.'

o-kaki-ni nat-ta write-HON-PAST

Thus, the referent honorifics have the ground specified in their scope of predication just like the deictic motion verbs do. Before we see the ground specification in them in Cognitive Grammar style representations, it should be noted that Japanese referent honorifics are classified into subject honorifics and non-subject honorifics, depending on the role that the respected person plays in the denoted event. The two types of Japanese predicate referent honorifics are roughly characterized as follows: i) subject honorifics: indicating the speaker's respect toward the subject of the denoted action ii) non-subject honorifics (or subject humbling): indicating the speaker's respect toward some other participant directly or indirectly involved in the denoted action The semantic structures of these two types of social deictic expressions of nominal reference can be represented schematically in Figure 13.

Toward a typology of linguistic subjectivity

a. o-V-ηί naru (subject honorific)

b.

109

o-V-suru (non-subject honorific)

Figure 13. Expressions of social deixis (referent honorifics) in Japanese In Figure 13, the ground is specified somewhere in their scope of predication (the square box), inside or outside of its on-stage region (OS), - a n y where but the target of honorification (the SA arrow) i.e. the action doer (tr) in the subject honorifics, and someone who receives his action (lm) in the non-subject honorific. The conventionalized use and the restriction of the subject honorifics are shown in (26) and (26'), respectively. (28) below shows those of the non-subject honorifics, using the verb matu 'wait': (28)

a. Taro Taro b. Taro Taro

wa TOP wa TOP

sensei teacher *watasi lSG

wo ACC wo ACC

o-mati-si-ta. wait-HUM-PAST o-mati-si-ta. wait-HUM-PAST

For a cross-linguistic examination of honorifics, see Wenger's (1982) seminal typological work. 30 Let us limit our scope here to East Asian lan30. Wenger (1982) examines honorifics in several languages in South Asia, Southeast Asia and East Asia, and even comes up with some implicational universals. One of them is "nouns > pronouns > verbs", which means that honorifics are formed from the left to right, and so that the existence of honorifics in the category on the right implies the existence of each and every category to its left. He also argues that addressee honorifics are formed from referent honorifics (see also Tsujimura 1963).

110

Satoshi Uehara

guages (see Fillmore's quote above) and examine their honorifics from the perspective of the current study. Among languages geographically close to Japanese, Chinese possesses the plain/polite distinction (ni/nin 'you'/ 'you.POLITE' ) for the second person pronouns, 31 but is not complimented by a larger, systematic honorific predicate morphology similar to what is found in Japanese. Thus, the pattern in Chinese resembles that in European languages with the T/V distinction. Korean, on the other hand, resembles the Japanese pattern. It has six forms (and six levels of politeness) of the second person pronoun (Sohn 1999), and its subject honorific pattern is as productive as that of Japanese. However, its non-subject honorific pattern is not productive at all, and is rather a lexical phenomenon of a few verbs. The social deixis cline of expressions of nominal reference shown in Figure 14 summarizes our discussion in this section. subjective 4 Japanese, Korean

• T/V languages Chinese

objective English

# of subjective patterns Figure 14. Subjectivity (social deixis) cline of nominal reference (referent honorifics) A caveat is in order about the social deixis scale above. It should be noted that the scale reflects the languages' position in terms of the GRAMMATICALIZED subjective markers in them. Thus, it does not mean that languages on the right of the scale, such as English, are not a polite language or do not have "politeness strategies" (Brown and Levinson 1987). As polite expressions in English, for example, we might recall the patterns like 'Could you please ...?' as opposed to 'Can you ...?' or imperative forms used in requesting situations. Rather, the existence of grammaticalized referent-honorific patterns in the languages on the left of the scale, that in Japanese, for example, suggests that, in a variety of situations (not just requesting ones), and whether or not the respected person/referent is present at the time of speech, the speaker's social/subjective attitudes to-

31. Bob Sanders pointed out to me that the plain/polite distinction of the second person pronouns in Chinese do not exist for the plural, that is, whereas a singular/plural distinction exists for the plain form, i.e. ni vs. nimen, only the singular for nin is possible.

Toward a typology of linguistic subjectivity

111

ward the referent must have been frequently expressed to the extent that such (e.g. verbal morphological) patterns have become schematized and conventionalized in the language.

5.

Summary and implications

In this paper I have examined three types of subjective expressions crosslinguistically, and presented a very rough sketch, but some basic foundations, of research into linguistic subjectivity typology. Among the observations we can make from the current study are the following regarding the three subjectivity clines presented. Firstly, languages can and do vary in how much and what subjective expressions are conventionalized/grammaticalized. (Still additional subjectivity clines could be set up.) Secondly, each subjectivity cline is theoretically independent, and a language can be subjective in different degrees with respect to each cline, but each language seems to fall on more or less similar points of the three subjectivity clines (although whether or not one language can be placed at the most objective point of one cline and at the most subjective point of another cline is yet to be tested over a greater number of languages). And lastly, these three clines, accordingly, can collaborate to set up a new typology of languages, such as "subjective frame" vs. "objective frame" language types (a la Talmy's "verb-framed" and "satellite-framed" language types), according to which Japanese appears to represent the subjective frame (or speaker-oriented) language type and English the objective frame type. This type distinction is by no means binary and theoretically further subjective and further objective types are possible. The present study has a number of theoretical implications. Let us discuss three of them below and suggest some directions for further study. Firstly, the third observation above, that is, that there can be subjective frame languages, seems to suggest that we need to reexamine the canonical event structure model. The canonical event structure has been rather implicitly thought of as what is in the on-stage region in an optimal viewing arrangement diagrammed in Figure 1(a). Now that we know some languages by default lean rather toward the ego centric viewing arrangement diagrammed in Figure 1(b), the canonical event structure might not be so canonical at least for such languages, and some subjective factors might as well be called for in accounting for some linguistic phenomena (e.g. noncanonical case marking patterns) in them (cf. Uehara 1998b). In this re-

112

Satoshi Uehara

gard, we need to examine some possible correlations between the subjectivity phenomena in this study and phenomena called empathy and logophoricity. Secondly, in a related note, the current study has an interesting implication for linguistic typology. The case study of instances of kuru 'come' in Japanese and their counterpart expressions in English in Section 2.1.2 has shown that the default degree of subjectivity of a language has direct relevance to the canonical viewing arrangement for the language, and hence, to the canonical event structure for it. To be more specific, let us examine one example listed in Table 2, which is repeated here as (29): (29)

J. tegami ga kita letter NOM came (lit. 'a letter came')

Ε.

I received a letter

A Japanese expression of implicit reference to the ground (an intransitive verb expression) is frequently translated into an English expression of explicit reference to it (a transitive verb expression). This contrast in the typical valency structure is a well-known contrast between the two language types, with Japanese in one type and English in the other (e.g. Ikegami 1981 calls the former "become" language type and the latter "do" language type). The current study thus suggests that a language's membership into such language types can be, at least partially, ascribed to its default degree of subjectivity. (A similar observation can be made about the relation between the proposed typology in the current study and Talmy's verb-framed vs. satellite-framed language typology.) Lastly, although we have limited our discussion to the current, synchronic state of the languages' position on the subjectivity cline, diachronic changes in their position on the scale are also conceivable, and such instances can represent cases of subjectification and/or objectification. In fact, such cases are reported in the languages examined here. For example, Newman's (1996) description quoted in Section 2.3 of the Maori case, where a deictically neutral GIVE verb has become integrated with deictic directional morphemes and lexicalized, illustrates a case of subjectification. A similar case with expressions of basic motion events is examined in DeLancey (1985), which we discussed in Section 2.1.1. He argues that such processes of grammaticalization of directive motions in TibetBurman languages are "cyclic" processes, which means in this study that they are cases of subjectification and desubjectification/objectification.

Toward a typology of linguistic subjectivity

113

Sanders et al. (2000) suggests that Taiwan Mandarin possibly acquired its subjective inflectional paradigm for the THINK verb via their native Taiwan Southern Min helped along by language contact with Japanese. All such instances as these demonstrate that the three (and possibly more) types of expressions can change their position on the subjectivity cline (that is, change their degree of subjectivity) over the course of time, without changing any objective component of their meaning. Such historical processes also should be among the terms that subjectification and objectification refer to. Although this is a preliminary study, I hope to have shown that cognitive theory of subjectivity can not only handle cross-linguistic patterns of variation in subjectivity phenomena in a uniform fashion, but it can also work as a useful tool to provide a new dimension to the study of linguistic typology. With this tool, we can now provide an answer in the affirmative to Lyon's question posed at the beginning of this paper. Cognitivetypological approaches to subjectivity, however, have just begun.

References Achard, Michel 1996 Perspective and syntactic realization: French sentential complements. Linguistics 34: 1159-1198. Benveniste, Emile (Translation by Mary E. Meek) 1971 Problems in General Linguistics. Coral Gables: University of Miami Press. Brown, R. and Gilman, A. 1968 The pronouns of power and solidarity. In J. A. Fishman (ed.), The Sociology of Language, 252-275. The Hague: Mouton. Brown, P. and Levinson, C. S. 1987 Politeness: Some Universals in Language Usage. Cambridge: Cambridge University Press. Cheng, Robert L. 1985 A comparison of Taiwanese, Taiwan Mandarin, and Peking Mandarin. Language 61.2: 352-377. Comrie, Bernard 1976 Linguistic politeness axes: speaker-addressee, speaker-referent, speaker-bystander. Pragmatics Microfiche 1(7): A3. Department of Linguistics, Cambridge University.

114

Satoshi Uehara

Croft, William 1991 Syntactic Categories and Grammatical Relations: The Cognitive Organization of Information. Chicago: University of Chicago Press. DeLancey, Scott 1981 An interpretation of split ergativity and related patterns. Language 57.3: 626-657. 1985 The analysis-synthesis-lexis cycle in Tibeto-Burman: a case study in motivated change. In John Haiman (ed.), Iconicity in Syntax, 367-389. Amsterdam/Philadelphia: John Benjamins. 1987 Transitivity in grammar and cognition. In Rüssel S. Tomlin (ed.), Coherence and Grounding in Discourse, 53-68. Amsterdam/Philadelphia: John Benjamins. Eda, Sumire 1983 "Te yarn, te kureru, te morau" to Tai-go no hyogen: Hai no yöhö ni tyümoku shite ["Te yaru, te kureru, te morau" and Thai expressions: with a focus on the use of häi\. Nihongo Kyöiku [Japanese Language Education] 49: 119-132. Fillmore, Charles, J. 1966 Deictic categories in the semantics of 'come'. Foundations of Language 2: 219-227. Frajzyngier, Z. 1987 Ventive and centrifugal in Chadic. Afrika und Übersee 70: 3147. Giridhar, P. P. 1975 Angami Grammar. Central Institute of Indian Languages. Givon, Talmy 1994 The pragmatics of de-transitive voice: Functional and typological aspects of inversion. In Talmy Givon (ed.), Voice and Inversion, 3^14. (Typological Studies in Language 28.) Amsterdam/Philadelphia: John Benjamins. Hidaka, Mizuho 1994 Juyo döshi no zenkoku bunpu to tözai tairitsu [Differences in the Distribution of benefactive verbs of east and west dialects in Japan]. Osaka University Nihon Gakuhö 13: 65-82. Ide, Sachiko 1975 Onna no Kotoba, Otoko no Kotoba [Women's Speech, Men's Speech]. Tokyo: Nihon Keizai Tsüshinsha. Ikegami, Yoshihiko 1981 "Suru" to "Naru" no Gengogaku [Linguistics of "Do" and "Become"]. Tokyo: Taishükan.

Toward a typology of linguistic subjectivity

115

Iwasaki, Shoichi 1993 Subjectivity in Grammar and Discourse: Theoretical Considerations and a Case Study of Japanese Spoken Discourse. Amsterdam/Philadelphia: John Benjamins. Kuno, Susumu 1973 The Structure of the Japanese Language. Cambridge: MIT Press. 1978 Danwa no Bunpd [Discourse Grammar]. Tokyo: Taishükan. Kuno, Susumu and Etsuko Kaburaki 1977 Empathy and Syntax. Linguistic Inquiry 8.4: 627-672. Kuroda, S.-Y. 1973 Where epistemology, style, and grammar meet: a case study from Japanese. In S. R. Anderson and P. Kiparsky (eds.), A Festschrift for Morris Halle. New York: Holt, Rinehart and Winston, Inc. Langacker, Ronald W. 1985 Observations and speculations on subjectivity. In John Haiman (ed.), Iconicity in Syntax. Amsterdam/Philadelphia: John Benjamins. 1987 Foundations of Cognitive Grammar, Vol. 1, Theoretical prerequisites. Stanford: Stanford University Press. 1990 Subjectification. Cognitive Linguistics 1-1: 5-38. 1991 a Concept, Image, and Symbol: The Cognitive Basis of Grammar. Berlin: Mouton de Gruyter. 1991 b Foundations of Cognitive Grammar, Vol. 2, Descriptive application. Stanford: Stanford University Press. 2001 Discourse in Cognitive Grammar. Cognitive Linguistics 12-2: 143-188. Lyons, John 1982 Deixis and subjectivity: Loguor, ergo sum? In R. J. Jarvella and W. Klein (eds.), Speech, Place, and Action: Studies in Deixis and Related Topics. New York: John Wiley & Sons Ltd. Martin, Samuel E. 1964 Speech levels in Japan and Korea. In Dell Hymes (ed.), Language in Culture and Society: A Reader in Linguistics and Anthropology. 407^115. New York: Harper & Row. Matisoff, James A. 1991 Areal and universal dimensions of grammaticalization in Lahu. In E. C. Traugott and B. Heine (eds.), Approaches to Grammaticalization. Amsterdam/Philadelphia: John Benjamins. NLRI (National Language Research Institute) 1966 Nihon Gengo Chizu [Japan Language Map], Vol. 2. Tokyo: Okurasho Insatsukyoku.

116

Satoshi Uehara

Newman, John 1996 Give: A Cognitive Linguistic Study. (Cognitive Linguistic Research 7.) Berlin: Mouton de Gruyter. No, Yongkyoon 1989 Existential quantification of experiencer: Person constraint on emotion verbs and zero anaphora. In S. Kuno et al. (eds.), Harvard Studies in Korean Linguistics III. Seoul: Hanshin Publishing Company. Ohye, Saburö 1975 Nichieigo no Hikaku Kenkyü: Shukansei ο megutte. [A Contrastive Study of Japanese and English: On Subjectivity], Tokyo: Nan'undö. Ricca, Davide 1991 Andare e venire nelle lingue romanze e germaniche: dall'Aktionsart alia deissi ['Go' and 'Come' in the Romance and Germanic languages: From Aktionsart to deixis]. Archivio Glottologico Italiano 76: 159-92. Ryan, P. M. 1999 Maori-go Kyohon [Modern Maori Book], Vol. 1. Fukushima: Fukushima Prefectural Association for International Exchange. Sanders, Robert, Satoshi Uehara and Chien-ling Chiang 2000 A case of cognitive restructuring in the Southern Min and Mandarin of Taiwan. IsCLL VII Proceedings. Shibatani, Masayoshi 2003 Directional Verbs in Japanese. In Erin Shay and Uwe Seibert (eds.), Motion, Direction and Location in Language: In Honor of Zygmunt Frajzyngier, 259-286. (Typological Studies in Language 56.) Amsterdam/Philadelphia: John Benjamins. Smyth, David 2002 Thai: An Essential Grammar. London: Routledge. Sohn, Ho-Min 1999 The Korean Language. Cambridge: Cambridge University Press. Talmy, Leonard 1985 Lexicalization patterns: semantic structure in lexical forms. In T. Shopen (ed.), Language Typology and Syntactic Description 3: Grammatical Categories and the Lexicon, 56-149. Cambridge: Cambridge University Press. Tokieda, Motoki 1941 Kokugogaku Genron [Principles of Japanese Linguistics], Tokyo: Iwanami Shoten.

Toward a typology of linguistic subjectivity

117

Thompson, C. 1994 Passives and inverse constructions. In T. Givön (ed.), Voice and Inversion, 47-63. Amsterdam/Philadelphia: John Benjamins. Traugott, Elizabeth C. 1982 From propositional to textual and expressive meanings: some semantic-pragmatic aspects of grammaticalization. In Winfred P. Lehmann and Yakov Malkiel (eds.), Perspectives on Historical Linguistics, 245-71. Amsterdam/Philadelphia: 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. In Dieter Stein and Susan Write (eds.), Subjectivity and Subjectivisation: Linguistic Perspectives, 31-54. Cambridge: Cambridge University Press. Traugott, Elizabeth C. and Richard B. Dasher 2002 Regularity in Semantic Change. Cambridge: Cambridge University Press. Tsujimura, Toshiki 1963 Keigo no bunrui ni tsuite [On classification of honorifics]. Gengo to Bungei [Language and Literature], 5,2. Uehara, Satoshi 1998 a Pronoun drop and perspective in Japanese. In N. Akatsuka, H. Hoji, S. Iwasaki, S.-O. Sohn and S. Strauss (eds.), Japanese/ Korean linguistics, Vol. 7, 275-289. Stanford: CSLI. 1998 b Subjective predicates in Japanese: A cognitive approach. Paper presented at the 4th Australian Linguistic Institute Workshop: Cognitive Research Issues in Cognitive Linguistics. Wenger, James Rodney 1982 Some universale of honorific language with special reference to Japanese. Doctoral Dissertation, University of Arizona. Yamada, Toshihiro 1996 Some universal features of benefactive constructions. Osaka University Nihon Gakuho 15: 27^-5.

Section II Case studies I - Modals and modality

Subj edification in (expressions of) epistemic modality and the development of the grounding predication Piter Pelyvds

1.

Introduction

The appearance of the grounding predication in Langacker's holistic cognitive grammar is a major development in the elaboration of a grammatical theory capable of dealing with a number of factors connected with language use. It had become apparent by the 1980's that traditional linguistics was imposing too many restrictions on what was to count as part of the language system and that many of these restrictions were a consequence of a formal theory's inherent inability to give a principled account of the issue at hand.1 It had also become apparent that although there might perhaps have been some remote theoretical advantage in describing the linguistic system first, without reference to its use, this was an impossible task simply because language was not structured that way: in a number of cases no systematic description of the former is conceivable without taking the latter into consideration.2 One significant property of language use is the speaker/conceptualizer's attitude to what is being communicated, a relevant factor in the linguistic formulation of the utterance. Since this factor is not marked as consistently in languages like English as, for instance, in the languages that have evidentials (cf. Palmer 1986), it was often disregarded, sometimes to the extent that even the existence of the speaker's epistemic commitment in the

1. Cf. e.g. Lakoff s (1991) discussion of generative grammar's commitments along these lines. 2. A good example for this is the 'irregular' syntactic behaviour of most cognitive predicates in English, first described in Kiparsky and Kiparsky 1970 and elaborated on within the framework of epistemic grounding in Pelyväs (1996). We shall return to a discussion of this issue in 3.2.

122

Peter Pelyväs

declarative sentences of such languages was brought into question (Palmer 1986: 86-87). In cognitive theory these aspects of speaker involvement are manifested in the introduction of the notion of epistemic grounding, a product of the reference-point construction and subjectification. The grounding predication relates (the linguistic expression of) a process or thing (a verb or a noun) to the situation of its use: speaker/hearer knowledge, and time and place of utterance. Langacker (1987, 1991a) defines epistemic grounding as follows: An entity is epistemically grounded when its location is specified relative to the speaker and hearer and their spheres of knowledge. For verbs, tense and mood ground an entity epistemically; for nouns, definite/indefinite specifications establish epistemic grounding. Epistemic grounding distinguishes finite verbs and clauses from nonfinite ones, and nominals (noun phrases) from simple nouns. (Langacker 1987: 489) [Grounding is] a semantic function that constitutes the final step in the formation of a nominal or a finite clause. With respect to fundamental "epistemic" notions (e.g. definiteness for nominals, tense/modality for clauses), it establishes the location vis-ä-vis the ground of the thing or process serving as the nominal or clausal profile.3 (Langacker 1991a: 549)

As the definitions also suggest, epistemic grounding can be approached from the conceptual or from the formal side. What they do not reveal is that there may be a conflict between the predictions of the two sides that the principles of cognitive grammar do not find easy to resolve. The aim of this paper is to explore the nature of this conflict and to suggest solutions to them. After a brief discussion of the relevant aspects of epistemic grounding as described in Langacker's system in Section 2, modifications will be suggested on both sides. On the formal side, it appears that the reference-point nature of the construction leads to a perhaps unnecessary restriction on the lexical items that can count as grounding predications. Since the grounding process leaves only the grounded head (the complement) in profile, the latter has to be a process if the clause is to be regarded as finite. In the present cognitive system, this condition excludes all forms that are clearly non-finite (since they are seen as summarily scanned and thus incapable of profiling a process) and also the ones that are clearly finite (since Tense can only come 3. As the definitions suggest, epistemic grounding is relevant both at the NP and the clause level. In this paper we will concentrate on the clause level.

Subjectification in (expressions of) epistemic modality

123

from a grounding predication already present in the structure). This leaves only the English modals as grounding predications, excluding modals in a number of other languages (e.g. German) on the one hand, and also cognitive predicates with meanings that are very close to those of the modals. In Section 3 I will argue that - non-finite forms are not necessarily summarily scanned; - finite forms need not be necessarily grounded. The overall effect of the two suggested changes would be a system capable of regarding (epistemic) cognitive predicates as grounding predications. On the conceptual side, the question arises whether epistemic and deontic modals should both be equally regarded as grounding predications. The formal side (bare infinitival complements in both cases) would certainly warrant this, but a more fine-grained analysis of the conceptual structures of the two kinds of modal meanings reveals substantial differences in the nature and degree of subjectification occurring in them. Langacker's (1999) version of subjectification, which describes the process in terms of attenuation of subject control, provides a solid background for the hypothesis presented here that only epistemic modals should be regarded as establishing (epistemic) grounding. In Section 4 I will argue that - the conceptual structures of root and epistemic modals are different to the extent that justifies regarding only epistemic modals as grounding predications. This is a restriction on the original system; - the conceptual structures of modals and some cognitive predicates are very similar, which can be seen as further evidence in favour of regarding the ones with epistemic meanings as grounding predications.

2.

The nature of the grounding predication

Conceptually, the grounding predication is a combination of the referencepoint construction and subjectification. Being both subjective and a reference-point (the speaker's vantage point to a situation), this kind of predication shifts out of the centre of attention once the target is reached. As a result, the grounding relationship will profile the grounded entity (grounded head) rather than the grounding predication or the relationship itself-to the extent that in the majority of cases there will be no overt indi-

124

Peter Pelyväs

cation (except for Tense) that a process of grounding has taken place (cf. Palmer 1986, referred to in Section 1). This relationship, which could not be part of a system grammar owing to the property that it changes in (processing) time, may account for phenomena of 'non-propositionality' attributed to certain structures in traditional theories, cf. the status of the question tag in sentences with cognitive matrix predicates, cf. (1) and (2): (1)

(2)

a. b. c. d. a. b.

I don't think that Christine is clever, is she? ??I don't think that Christine is clever, do I? *John doesn 't think that Christine is clever, is she? John doesn't think that Christine is clever, does he? *I don't regret that Christine is clever, is she? I don't regret that Christine is clever, do I?

Question tags tend to leave cognitive predicates out of their scope when the subject is first person. This is not possible with non-cognitive matrix predicates. In cognitive terms, we could say that what is in profile in (la) is not the cognitive predicate think, seen here as a grounding predication,4 but the grounded head (the subordinate clause). For different reasons this could not apply to (lc,d) or to the sentences in (2). The formal side of the grounding predication's grammatical properties appear to follow directly from its conceptual structure. Pelyväs (1996) summarizes Langacker's (1987, 1991a) position as follows: since the grounding predication profiles the grounded head rather than the grounding relationship itself, the grounded head has to be a form that is neither finite (as Tense is given by the grounding process) nor non-finite.5 The reason is that the whole clause profiles a process, and since this must come from its 4. Assuming that cognitive predicates have properties relating them to grounding predications may appear to be problematic at this point in the light of what was said about the form required of the grounded head in Section 1, but we will bring further arguments in 3.2. in support of their grounding predication status. 5. This may sound like a contradiction at first sight, but, as it turns out, the contradiction is in the names that linguists have given these forms rather than in their properties. Since finite means having Tense and non-finite is to be interpreted as being summarily scanned, the two properties need not be contradictory. There may be room for relationships (at least partially) sequentially scanned (= processes) that are nevertheless non-finite forms. We will provide a more detailed analysis and suggest a solution in Section 3.

Subjedification in (expressions of) epistemic modality

125

profile determinant, the grounded head, it will also have to profile a process. Such forms are not easy to find, and this seems to determine what can count as a grounding predication. The only English form that can qualify as neither finite nor non-finite is the bare infinitive (or simple verb), and this is in good agreement with the fact that the English modals, the prototype for the grounding predication, are followed by this form. Symbolicity of conceptual and grammatical structure can be said to have been achieved, especially since hardly any other predicate takes its complement in this form. Unfortunately, this solution has some undesired consequences as well. (i) The bare infinitive is highly language-specific. Even in German, a language closely related to English, modals are followed by a form of the verb that is no different from an ordinary infinitive. If what can count as a grounding predication is determined by this criterion, German modals are not members of this class and have to be given different analyses from their English counterparts (cf. Section 3.1. and also Tanja Mortelmans 2002, in this volume, who argues, among other things, that such a conceptual distinction is not warranted). (ii) Deontic English modals also take the bare infinitive, with the result that they will also be analyzed as forms that establish an epistemic relationship. There may be a deontic relationship similar or equivalent to epistemic grounding (cf. Hare's [1971] neustic and tropic), but it seems to be secondary to its epistemic counterpart: deontically modalized sentences can (and must) be epistemically grounded.6 In Section 4 we shall also argue that there are substantial differences between epistemic and deontic conceptual structures that may warrant regarding only the former as grounding predications. The conceptual side seems to require that we restrict the notion to epistemics.

6. The nature of this relationship and its compatibility with epistemic grounding will certainly require further study, but apparently the 'matrix' grounding relationship need not be restricted to Tense but should also include an (optionally unmarked) epistemic 'slot'. A deontic modal cannot follow an epistemic one, but the sequence is excluded for formal rather than conceptual reasons. There is nothing conceptually wrong with (i) I think Christine must leave now. (We are again assuming that cognitive predicates can also function as grounding predications, cf. Footnote 4.)

126

Piter Pelyväs

(iii) Since on the conceptual side epistemic grounding was given a broad semantic definition, it would be an advantage for the theory if all forms that are compatible with that definition could be considered for grounding predication status. I have argued extensively (cf. Pelyväs 1996, 2001 a,c, 2003) that cognitive predicates, at least in their epistemic uses, could qualify: they have conceptual structures that are very similar to those of the epistemic modals: the differences are in what is highlighted rather than in the relationships themselves. Some (e.g. expect) even share with modals the root to epistemic extension described in Sweetser (1990). Unfortunately, since the complements of cognitive predicates are clearly finite,7 they cannot even be considered for formal reasons in Langacker's framework: a finite clause is seen as grounded by definition (Langacker 1991b: 244). But, since in sentences like (3) (3)

It is likely that Christine is clever.

the complement clause is nevertheless not to be regarded as CERTAIN, Langacker (1991b: 247) postulates an overriding mechanism that can reassess the validity of the complement.8

3.

The formal side: flniteness and the modes of scanning

For the formal reasons given above, grounding predication status remains one of the very few categories in cognitive grammar that are still based on a strict dichotomy although, arguably, this may not be justified on the conceptual side. We will now examine the problems emerging from the form required in grounded head status. As we have seen, postulating a form restricted to only a small set of English structures has undesirable consequences in both directions: it includes cases that may not be relevant on a conceptual basis 7. or clearly non-finite after 'raising', a reference-point construction (Langacker 1995, 1999) which owes its existence to the semantic transparency of the (typically) cognitive matrix predicate. We will return to a discussion of this in 3.2. 8. With this solution we have two different components that do essentially the same job. In addition, as Pelyväs (1996) argues, the mechanism can 'override' the default value, which is unmarked (except by Tense) but also the strongest. Weaker but marked values cannot be overridden, which sounds to some extent counterintuitive.

Subjectification in (expressions of) epistemic modality

127

and excludes some that very probably are. The differences between the conceptual structures of deontic and epistemic modals will concern us in Section 4. In this section we will concentrate on the issue of the form that the grounded head can take. As we have referred to it in Section 2, the postulation of a form that is neither finite nor non-finite is not a contradiction in terms. We could even say that the two terms are to some extent orthogonal, since the term finite means 'having Tense' while non-finite does not simply refer to the opposite in Langacker's system but to the properly of 'being summarily scanned'. I will argue in this section that non-finite forms of the verb need not be summarily scanned. The addition of the derivational affix -ing or to of the infinitive may bring about changes in conceptual structure that may to some extent conflict with or decrease the significance of sequential scanning without actually turning it into summary scanning. This would mean that these non-finite forms could still profile processes and thus function as grounded heads, i.e. complements of grounding predications.

3.1.

Non-finite forms

The essence of the atemporal relation vs. process dichotomy remains the difference between summary and sequential scanning, but transitions between those two modes of scanning are not inconceivable in the grammatical categories that have been analyzed as transitional even in traditional grammars. Langacker (1991b: 26) describes the effect of adding the affix ing to a verb stem as threefold: it confines (immediate) scope to exclude the end-points of the situation, neutralizes the differences between the internal points (homogenizes the situation), and construes the situation holistically by suspending sequential scanning. On closer scrutiny, each change appears to be a consequence of the one listed before it. Excluding the beginning and the end necessarily has an adverse effect on the conceptualizer's view of the internal development of a situation. This is well illustrated by the semantic differences between the perfect and perfect progressive forms of a telic verb phrase. In this situation, when the event proceeds, as it were, 'from nowhere to nowhere', time naturally loses at least some of the crucial role that it plays in sequential scanning, the mental operation underlying the conceptualization of a process. But this need not mean that the situation is downgraded all the way down to the status of atemporal relation (summary scanning). After all,

128

Peter Pelyväs

even though non-finite forms of the verb do not have Tense, they are still capable of expressing temporal relations (simultaneity, anteriority9), form a continuum from most verb-like participle, through non-factive and factive gerund to verbal and deverbal noun. Most of them retain at least some of the original verbal complementation as well, which may be symbolic of the survival of some relationships originating in the verbal conceptual structure. This is almost totally lost at the nominalization end of the scale in the often discussed ambiguity of the shooting of the hunters}0 We may have enough evidence in the case of -ing to argue for a transitional status where the situation is still (basically) sequentially scanned but the significance of this is reduced by other factors in the evolving conceptual schema. We could say in this way that even a non-finite form can to some degree profile a process, to the extent to which sequential scanning is preserved in it. I cannot offer at present a similar analysis for the toinfinitive, but since it is also a transitional form (even closer to verb status than the participle), and since to highlights path and in an abstract sense intention and potentiality, its conceptual structure may not be incompatible with a similar analysis.11 If the proposal that non-finite forms may (to some extent) profile a process is accepted, it will solve - the problem of the German modals (as well as of modals/auxiliaries in a number of languages including Dutch, Russian or Hungarian12); - the problem of English ought to, have to, etc. - the problem of cognitive predicates like likely in 'raising' constructions. Since the association of the 'bare infinitive' form with grounding predication status is weakened, there is less reason now to attribute this status to English deontic modals. It can also explain why the 'bare infinitive' can sometimes occur in a structure where a grounding relationship is clearly not relevant (as often with help).

9. In Russian and also, arguably, in totally unrelated Hungarian, what were initially participial forms expressing anteriority were reanalyzed as past tense forms after the loss of the original narrative past tense. 10. This argument is in close connection with the problems of grounding on the NP level, which we cannot discuss here (cf. Pelyväs 1986: 176-181). 11. The fact that there is relatively little semantic difference between should and ought even though the latter takes the to-infinitive can serve as an argument in favour of this view. 12. We will see some Dutch and Hungarian examples in Section 4.

Subjedification in (expressions of) epistemic modality

129

If non-finite forms of the verb can be both summarily and sequentially scanned, the question may now arise how (if at all) a distinction can be made between them. It is clear that occurrence as the complement of a grounding predication as a criterion for sequentially scanned status would lead to circularity. One answer could be that they cannot (and perhaps need not) be clearly distinguished - it is enough to think of the different types of factive and non-factive gerunds in the literature (e.g. Langacker 1991: 34). This is what I referred to as a probably untenable strict dichotomy in the first paragraph of Section 3. However, if we need independent evidence, we can find some in the conceptual structures involved (transparency of the matrix predicate, cf. Langacker 1995, 1999) and, on the formal side, in the syntactic structures in which (more) sequentially scanned non-finite forms occur. They prototypically alternate with finite i/jai-clauses in what would be called the Raising or the ECM transformation in Generative terms. This behaviour links this problem to that of the grounded status of some finite forms, our topic for the next section.

3.2.

Finite forms

In this section we will examine the clausal complements of cognitive predicates. As we have referred to it in the previous section, many of them appear with to-infmitival complements as well, a complement type that could in our hypothesis function as the complement of a grounding predication. In fact, we will be using this ambivalence between the two types of complement as evidence in our argument that rather than permitting the overruling of established values of grounding, cognitive theory should regard cognitive predicates as grounding predications as well. Two arguments for this are concerned purely with the conceptual side. (i) Since the respective roles of the Tense and of the modality element of the grounding predication are not fully understood in the theory of epistemic grounding, finiteness, contra Langacker (1991a: 244), need not be a sufficient condition for grounded status. I will not venture a detailed analysis of this point here, but the argument in 3.1. could be extended in this direction as well. (ii) As was mentioned in Section 2, the conceptual structures of cognitive predicates are remarkably similar to those of the modals (including perhaps similar diachronic development), cf. Figure 3, the conceptual schema for deontic MAY, which would serve permit equally well, with the

130

Peter Pelyväs

difference that some of the participants and relationships not named in MAY are made explicit in permit, which does not require speaker/ permission giver identity. This argument is discussed in detail in Pelyväs (2001c). Three arguments concern more directly the formal side: (i) In some languages a number of the predicates relevant here would take subjunctive complement clauses, clearly indicating non-grounded status. In English the role of the subjunctive has been decreasing steadily (owing perhaps to the similarity of indicative and subjunctive forms), with its role often being taken over, in addition to indicatives, by non-finite forms. (ii) Non-cognitive matrix predicates do not usually permit alternative finite and non-finite complement structures. This is called 'Raising' or 'Exceptional Case Marking' in Generative terms. Langacker (1995, 1999) analyses the phenomenon as the creation of a reference-point construction and states that this is only possible with a limited set of predicates that are fully transparent semantically with respect to the 'raised' element: The governing predicates include verbs of need, desire, or intention with respect to a situation ... Many of them express some degree of commitment to the reality of a situation ... Others ... are primarily verbs of communication serving to inform the addressee about a situation or to effect it by virtue of the speech act named. Strikingly, all of these verbs imply a conceptualizer ... who conceives of a situation and assumes some stance or attitude in relation to it. More importantly, they do not specify any direct interaction between the conceptualizer and a participant in that situation - the profiled relationship links the conceptualizer to the conceived situation per se.13 (Langacker 1995: 48)

13. Clearly, this includes bouletic, deontic and speech act predicates as well as epistemic ones, which supports the assumption that deontic (and perhaps also the other listed) relationships could be analyzed in terms of some sort of grounding structures as well, even though we would not like to identify them with epistemic grounding. It is not clear whether the second paragraph applies equally to all categories listed here: in (i) (i) I expect Christine to be home at 11 sharp, if it is given a prototypical 'narrow scope' deontic interpretation, Christine can be seen as having a more or less direct relationship with the speaker as conceptualizer, which would of course not be part of an epistemic reading. For details, see 4.3.

Subjectification in (expressions of) epistemic modality

131

The conceptual properties of these predicates are identical with those of the grounding predication except for the fact that in Langacker's view the profiled relationship here is the link between conceptualizer and the situation (as would be the case with the German modals,14 cf. 4.1), whereas in a truly grounding relationship it is the 'conceived situation' alone that would be in profile. (iii) The 'irregular' syntactic behaviour of cognitive predicates can be seen as quite regular with our hypothesis that they are grounding predications rather than true matrix predicates in mind as an organizing principle, since the structure often resembles a simple clause in its behaviour. Consider the following examples: (4)

(5)

(6)

(7)

The 'Raising' or 'ECM' construction: a. I believe Christine to be a criminal. b. * I resent Christine to be a criminal. Long distance Wh-movement: a. Who do you believe to be a criminal? b. * Who do you resent to be a criminal? Negative Raising: a. I don't think that Christine is a criminal = I think that Christine is not a criminal. b. I don't resent that Christine is a criminal ΦI resent that Christine is not a criminal. Topicalization from subordinate clause: a. Christine I think (that) you'll like. b. *Christine I resent that you 7/ like.

14. The fact that this situation also describes German modals, which we would also like to see as grounding predications, does not count as evidence but is perhaps part of what we could call 'cumulative evidence'.

132 (8)

Peter Pelyväs

Time adverb movement: a. Tomorrow I think that Christine will come. b. * Tomorrow I resent that Christine will come.

These phenomena are apparent counterexamples to the generative statement that nothing can be moved out of a maximal projection. If we assume that a 'true' maximal projection is a grounded clause, then the subordinate clause complements of non-cognitive predicates in (4b) to (8b) provide evidence for their integrity as grounded structures. Since the 'subordinate' structures in (4a) to (8a) do not show this integrity, they could be naturally analyzed as grounded heads attached to their grounding predications. Within these complex structures movement is not impossible, since the grounded head in itself is not a 'true' maximal projection. These properties relate cognitive predicates to modal auxiliaries in their syntax as well. In fact, in Pelyväs (2001a) I argue that the two categories have reached grounding predication status through similar diachronic processes. The major difference between these two representatives of a relatively recently developed relationship, the linguistic expression of which may not (yet) be fully developed, is that in their present form modals look more like members of a simple clause, while cognitive predicates appear in some structures, but not in significant aspects of their syntactic behaviour, to be still full-fledged matrix predicates. Further evidence is provided for this hypothesis in (9). These sentences suggest that with non-cognitive matrix clauses both the matrix and the subordinate clauses can (and must) be grounded independently, whereas this is not possible in the case of cognitives: (9)

a. It must be surprising that Christine cannot be a criminal. b. * It cannot be likely that Christine must be a criminal. c. It is likely that Christine may be a criminal.

Whereas various (if not any?) combinations of main clause and subordinate clause grounding appear to be possible in (9a), the grounding elements 'cancel each other out' in (9b). (9c) owes its acceptability to the corroborating effect of compatible grounding values. Postulating ungrounded status for complement clauses of cognitive predicates would give a systematic account of these phenomena in contrast

Subjectification in (expressions of) epistemic modality

133

to generative grammar, where the transparent S-bar and ECM provide at best a makeshift solution. 15 - Extraposition is obligatory with cognitive predicates, optional otherwise: (10)

a. *That Christine is a criminal is likely. b. That Christine is a criminal is shocking.

If likely functions as a grounding predication, it has to be stated initially in neutral contexts (cf. Kiparsky and Kiparsky 1970). Since the default value is unmarked, this must be done to avoid the undesirable consequences of reanalysis in parsing. - A related phenomenon occurs in gerundial constructions and nominalizations, which would also concern grounding on the NP level: (11)

a. *The shooting of the hunters was likely. b. The shooting of the hunters was terrible.

Ungrounded structures cannot be easily condensed into concise structures that can elaborate larger structures. - The last example is that of the question tag discussed in Section 2. The examples are repeated here for convenience. (Γ)

(2')

a. b. c. d. a. b.

I don't think that Christine is clever, is she? ??Idon't think that Christine is clever, do I? *John doesn't think that Christine is clever, is she? John doesn't think that Christine is clever, does he? *Idon't regret that Christine is clever, is she? I don't regret that Christine is clever, do I?

15. There are (at least) two factors that could make this solution less appealing. One is that the syntactic behaviour is not fully regular, e.g. possible does not occur in 'raising' constructions. The other is that this syntax appears to be indicative of a potential to function as a grounding predication (or perhaps reporting grounding values attributed to some other conceptualizer) rather than actual grounding predication status, since the rules would not change with non-first person subjects, where the predicates in question clearly cannot function as (primary) grounding predications. Sentences like (Γ) appear to be the only exception in this respect. These questions will certainly require further study.

134

Peter Pelyväs

This can illustrate the reference-point-like nature of the grounding predication. Attention has shifted from the reference-point to the target, the grounded head. It is interesting that this phenomenon is sensitive to the person of the subject and will only work when the matrix predicate actually functions as a grounding predication. I have no ready explanation for this: my best hypothesis is that shift of attention is a psychological phenomenon that does not occur when the actual situation does not hold.

4.

Conceptual structures: deontic vs. epistemic modals

In the previous section we have argued for an extension of grounding predication status to the class of cognitive predicates. We have argued that the form of the complement may not be a decisive factor and that syntactic behaviour as well as conceptual content provide evidence that at least some of these predicates can function as grounding predications. The main advantage is that this saves the inconvenience of first establishing and then overriding the newly established grounding values - an implausible procedure from the conceptual point of view. In this section we will argue for a restriction: we will turn to a more detailed analysis of conceptual content through examining the deontic and epistemic senses of the English modals. A detailed examination of the conceptual structures of may, must and should/ought suggests that the deontic and epistemic senses are rather more different in terms of the nature and extent of subjectification occurring in them than envisaged by Sweetser (1990) or Langacker (1987, 1991a, 1999). We will argue that the nature of the differences justifies the view that only epistemic modals are to be regarded as grounding predications. Two factors are seen as crucial in the development of the epistemic sense: one is a restriction of immediate scope (cf. Nordlinger and Traugott 1997) to exclude essential factors of the deontic domain not compatible with the extension. The other is subjectification in Langacker's (1987, 1991a, 1999) sense. As considerable syntactic evidence suggests, the latter is only characteristic of epistemic modals. I suggest that we accept Langacker's (1999: 307-308) description of the modal meaning as a proper characterization of the epistemic senses but would like to argue that a more fine-grained analysis of the deontic senses reveals the latter to be rather different in nearly all respects that are relevant in a grounding predication.

Subjectification in (expressions of) epistemic modality

135

The analysis is based on cognitive grammar's changing view on subjectification.

4.1.

The early version: subjectification as replacement

Subjectification, the mental operation that is probably the most important in the development of the grounding predication, can occur in steps even in Langacker (1991b), a relatively early version, which envisages the process essentially as replacement: some facet of an initially objectively construed situation is 'replaced by a comparable but subjectively construed relationship in the very process of conception' (Langacker 1999: 298). Figures l a to c (Langacker 1991b: 334) illustrate the subjectification of the modals in two steps:

lm Figure la. Objective construal In this arrangement the speaker/conceptualizer does not play a specific role. The profiled relationship is between a trajector and some potential situation to be performed by the same trajector. This could describe the main verb phrase of the diachronic development of the modals, or contemporary subject-oriented modality (will of intention or can of ability).

t

PAST

PRESENT

Figure lb. Subjectification 1

FUTURE

136

Piter Pelyväs

In this first type of subjective construal the trajector of the objective construal is replaced by the ground (speaker/conceptualizer). Thus a previously objectively construed relationship is now construed (more) subjectively. This arrangement still profiles the relationship between the conceptualizer and the conceptualized situation and is therefore not regarded as a grounding predication. In Langacker's (1991b) view it would describe German modals. In the second type of subjectification the relationship between the ground and the situation is no longer in profile. The only profiled entity is the situation itself. This is the schema of the grounding predication: attention has shifted from the reference point to the target (grounded head) and the relationship is fully subjectified. This Figure would describe the English modals either in their root (mainly deontic) or epistemic senses.

t

PAST

PRESENT

FUTURE

Figure 1c. Subjectification 2 How subjectification as replacement was supposed to work can be illustrated on Sweetser's (1990) account of metaphorical extension in the English modals, with the difference that replacement here occurs in the extension from the root to the epistemic meanings. In the case of MAY, Sweetser suggests the following paraphrases: (12)

(13)

Root: John may go "John is not barred (by my or some other) authority (in the sociophysical world) from going" Epistemic: John may be there "I am not barred by my premises from the conclusion that John is there"

Subjectification in (expressions of) epistemic modality

137

In the epistemic sense the speaker/conceptualizer simply replaces the trajector in essentially the same schema. This account appears to be unsatisfactory for a number of reasons (cf. Pelyvas 1996, 2000). There are two problems directly related to this topic: (i) The scope of negation changes in MAY in the extension into the epistemic domain. This makes Sweetser's 'forces and barriers' approach inapplicable in the epistemic sense. (ii) Replacement of the trajector with the speaker appears to violate the Invariance Hypothesis, which states, among other things, that the semantic relationships between (remaining) participants may not be confused in the extension. For a detailed analysis, cf. Pelyvas (2000). This second problem, i.e. that participants' roles get confused in the extension will only become apparent if we can provide a more fine-grained analysis of the conceptual structure of the modals,16 which is fully compatible with the more recent version of subjectification put forward in Langacker (1999).

4.2.

Subjectification as attenuation of subject control

Langacker (1999) takes a different view on subjectification. On this view the subjective component is inherent in the situation (part of the objective viewing arrangement) and can gradually become relatively more prominent as some objectively construed component, in our case subject control, 'fades away' (Langacker 1999: 298). The endpoint is full transparency, where the trajector (subject) is no longer in a traditional 'logicalgrammatical relationship' with the matrix predicate (reference point). Langacker (1999: 307) emphasizes two properties of the modals: - they are force-dynamic; - the event marked by the complement remains potential rather than actual. The profiled relationship involves some kind of effectiveness or potency tending toward realization of the type of action expressed by the complement but no actual instantiation ofthat action is implied. (Langacker 1999: 307-308).

16. We will be assuming that the trajector of Figure la remains part of the schema.

138

Piter Pelyväs

Thus the modal meanings develop as a result of the attenuation of subject control (progressive diffusion in the locus of potency). Then he argues that already in the root senses ... the source of potency is no longer associated with the subject, but is implicit and subjectively construed. It may be the speaker but it need not be ... It is not necessarily any specific individual, but may instead be some nebulous, generalized authority. In other words, the source of potency is highly diffuse. Nor is the subject necessarily the target of potency, which is also diffuse... (Langacker 1999: 308)

Epistemic modals are seen as ... widely diffuse in regard to the source and target of potency, hence transparent. I have described their potency as inhering in the evolutionary momentum of reality itself, as assessed by the speaker/conceptualizer (Langacker 1999: 309)

I find the statements quoted above to be an essentially valid summary of the conceptual properties of the epistemic modals, maintaining however, that root modals may be different in significant ways. In what follows I propose to examine in finer detail the differences in the diffuse nature of potency in root and epistemic modals. I will argue that potency is a direct consequence of the force dynamics and that in the prototypical root modals all major components of the force dynamics can be grasped and associated with participants, which may also include the speaker. I believe this analysis to be fully compatible with Langacker's: the main difference is that it pays a little more attention to the initial stages of the process. Langacker's observations on the diffuse nature of potency apply more to the less prototypical 'wide scope' senses of the root modals (cf. Nordlinger and Traugott 1997) and to the epistemic senses. On this account I also argue that only epistemic modals should be regarded as grounding predications.

4.3.

Root modality - 'narrow scope'17

I assume the conceptual structure of the prototypical root modal to be richer in participants/forces than the previous accounts believe. The as17. The terms 'narrow scope' and 'wide scope' will be interpreted in 4.4.

Subjedification in (expressions of) epistemic modality

139

sumption that apart from the force associated with the subject some other forces may be present can help solve the problems referred to in connection with Sweetser's analysis, since the forces associated with the subject (DOER) need not be replaced (e.g. with speaker-related ones).18 A new force can easily find its place in the conceptual structure of MAY, the epistemic and deontic senses of which can both be derived (in that order19) from a now extinct ability sense: 'be strong enough'. In the ability sense I identify Sweetser's original force with the subject's (DOER's) physical strength. In place of the barrier I postulate a counterforce that can be matched against but is less than equal to the subject's strength. This has to be a direct consequence of any transitive use of this verb, since in being strong enough to lift a rock the DOER's strength is estimated in the context of the weight of the rock. For the purposes of this analysis, however, the nature of this counterforce may remain quite diffuse, cf. Figure 2 (Pelyvas 2000). In this structure the speaker/conceptualizer appears only in the grounding role that would be part of any conceptualization, so the situation is construed objectively. In the deontic sense the original force associated with the DOER remains intact in the form of an intention to perform the purposeful action-a difference attributed to the extension into the sociophysical domain.20 It is the newly introduced counterforce that is identified with the speaker/conceptualizer. In this interpretation the speaker is weak or chooses to remain

18. If DOER and SPEAKER are both part of the source domain (Pelyvas 1996, Langacker 1999), Sweetser's extension from (12) to (13) leads to a violation of Transparency. 19. Contra Sweetser (1990), who derives epistemic MAY from the deontic sense. Diachronic evidence strongly suggests that MA 7 is a rare exception in that the epistemic sense appeared before the deontic one. Both derive from the now extinct ability meaning, cf. Traugott (1989). 20. We have reason to assume that this intention is present in the typical deontic context. Permission is usually given to perform acts that the permittee would like to do. If this intention is not present, the action will probably not take place in spite of the permission (even formal logic acknowledges that in a deontic system Pp does not imply p. There may be exceptional cases still compatible with this analysis, cf. Pelyvas (1996: 142-143).

140

Peter Pelyväs

weak - relinquishes authority by not preventing some purposeful action intended by the DOER that would [perhaps] be in his power to prevent. 21

potential counterforce ) matching ^ DOER'S strength

: Ort>. Ο 1 )OI κ

>0

potential action

OS •

( S/G) Figure 2. may - ability: the direct source of epistemic may We have reason to believe, however, that identification of the counterforce with the speaker is not direct replacement. In Pelyväs (1996) I referred to this as correspondence - probably an initial stage of subjectification, cf. Figure 3:

scope

OS

,Ο DOER

/

. . . . . . ο :\ speaker relinquishes aullioriti

/,,,

> ο

DOER 'sjjufpaSeful

S/Q

a/ction

(conceptualizer)

Figure 3. Deontic may

21. In some situations the speaker is really seen as weak, cf. concessive MAY, or in the epistemic sense, which we will discuss in 4.5.

Subjectification in (expressions of) epistemic modality

141

Arguments for the view that this is correspondence: - Correspondence occurs elsewhere in the conceptual structure of root modals, cf. DOER as permittee and DOER as agent of the potential purposeful action in Figure 3. This may not look a very strong argument since it is part of the hypothesis offered here, but its validity can be proved through its grammatical consequences. The Hungarian nearequivalent of MUST, KELL can occur in two syntactic constructions, each grammaticalizing one of the two roles linked by correspondence: (14)

a. Jänosnak ki kell tisztitani(a) a [John-Dat. (out) must clean-inf-(3rd.sing.) the shoes cipöjet. -Acc.-3rd.sing. poss.acc.] b. Jänos ki kell, hogy tisztitsa a [John-Nom. (out) must that clean-3rd.sing.imp. the shoes cipöjet. -3rd.sing. poss.acc.] 'John must clean his shoes'

The dative case in (14a) indicates that in this construction the DOER's subordinate {permittee) role is grammaticalized (the tail end of the action chain in terms of Langacker 1999: 27-34), while the nominative in (14b) suggests that in this construction, as in English, it is the agent role in the embedded structure (the head of the profiled part of the action chain) that finds grammatical expression. This provides a close analogy to our case in which the two roles can also be seen as distinct. - S/G retains the conceptualizer role (epistemic judgement) while assuming another role of a rather different nature. Even though this role could only be grammaticalized through a cognitive predicate, it is easy to see that deontic modals themselves can (and must) be epistemically grounded if we accept the hypothesis that cognitive predicates are also grounding predications. - The newly assumed role in OS is distinct from the conceptualizer role because it can be taken by another participant (Langacker 1999: 308). Even in MAY and MUST the speaker is only prototypically source of the permission or obligation, in contrast to SHOULD/OUGHT, where the (weak) obligation is prototypically attributed to social norms/expectations. In the case of Dutch MOET a source different from the speaker

142

Peter Pefyväs

can be inserted in the deontic sense by means of the equivalent of a byphrase (Sanders and Spooren 1997): (15)

Jan moet van Klaas thuisblijven. 'Jan must [by order of Klaas] stay at home' (Sanders and Spooren 1997: 97).

Even if we do not regard Dutch modals as grounding predications, there remains the striking difference that this option is not available in the epistemic sense to mark a conceptualizer different from the speaker. This suggests correspondence in the deontic and direct introduction of the speaker as conceptualizer in the epistemic sense: (16)

Jan moet (*van Klaas) thuisgebleven zijn. 'Jan must [*by order of Klaas] have stayed at home' 22

Lack of space does not permit a full analysis of the other modals expressing 'narrow scope' deontic modality: MUST (cf. Pelyvas 2000) and to some extent SHOULD/OUGHT (cf. Pelyvas 2001b), but our findings fully support the view that in this type of meaning, when the speaker/(conceptualizer) indirectly enters the scene (through correspondence), one facet of the relationship, i.e. one of the forces is redirected to the subjective axis. But it is not the force originally identified with the subject (DOER). Subject control is not significantly attenuated in this process. The sentences containing this type of modal will prototypically have an animate subject and, as mentioned above, the potential situation involved will have to be purposeful action. (Hence the unacceptability of ??You must grow another 5 inches.) If subject control is attenuated at all, that is attributed only to the existence of the counterforce (arguably present in the premodal meanings as well), which is part of the force dynamics responsible for the potency.

22. Although the examples in (14) to (16) are not taken from English, I find them relevant here as evidence for alternative grammaticalizations of options that I assume to be present in conceptual structure.

Subjectification in (expressions of) epistemic modality

4.4.

143

Root modality - 'wide scope'

With some of the root modals, most prominently with deontic SHOULD/ OUGHT, a different interpretation can also occur, which I will call 'wide scope' after Nordlinger and Traugott (1997), who give the name on the basis of the formula that would describe such sentences in terms of formal logical systems. I use inverted commas because I hold the view that what in fact happens is exactly the opposite: in this sense the immediate scope/objective scene (OS) is restricted to exclude any forces associated with the subject, cf. (17) and (18): (17)

(18)

You ought to know better than to suggest putting money on that horse again. [You OUGHT TO (know better...)], 'narrow scope' Mothers who have been raped, ... , by God, they have rights, too. And the bill ought to say that. (4 July 1990, UPI) (' those who are responsible for the bill have a moral responsibility to ensure that it says that') OUGHT TO [the bill (says that...)], or OUGHT TO [those responsible for the bill (ensure that...)] 'wide scope' (Nordlinger and Traugott 1997: 302)

The differences between the two types of deontic modality are seen in Figures 4 and 5. Since this is obligation rather than permission, participants of the situation and the forces associated with them are arranged differently in the conceptual structures. In obligation the force introduced in my analysis to replace Sweetser's barrier(s) represents the DOER's resistance and the original force associated with the speaker (through correspondence) in MAY (or deontic MUST, but as we shall see, not necessarily in deontic SHOULD/OUGHT) will work towards the realization of the action).23

23. Figure 4 could serve as a conceptual structure for MUST as well, with small modifications. In MUST the IMPOSER role would be identified with the speaker (through correspondence) and the counterforce associated with the reluctance of the DOER would be smaller to the limit of being negligible. Incidentally, these two factors may be spotted as being responsible for the greater politeness of SHOULD/OUGHT.

144

Piter Pelyväs

OS

ο — Ö [D \

IMPOSER DOER (e,.g. social norms, expectations).

purposeful

action,.

scope

S/Gj

(conceptualizer)

Figure 4. 'Narrow scope' deontic should/ought In 'wide scope' deontic modality the forces and consequently the semantic restrictions associated with the DOER are no longer highlighted. This is the net effect of the restriction in immediate scope (or objective scene-OS) to exclude this factor. The change is true attenuation of subject control in Langacker's (1999) sense, since the situation is now highly transparent and there is no constraint on the type of the potential situation involved. The force representing the IMPOSER is left without a counterpart—no relationship is established with a participant as was the case in the 'narrow scope' sense, which may be the conceptual factor behind the vagueness that is often felt about this meaning. It can also be seen as perhaps the first significant symptom of the developement of the modal meanings towards 'progressive diffusion in the locus of potency' referred to in 3.2, a state which, according to Langacker (1999), characterizes all modal meanings. What is particularly interesting with both senses of deontic SHOULD/ OUGHT is that the type of weak subjectification through correspondence associated with deontic MAY or MUST does not occur at all. The force driving towards the realization of the act is prototypically associated with social norms and expectations. This is a case of attenuating subject control without concomitant subjectification. The properties of 'wide scope' deontic SHOULD/OUGHT appear to be the exception rather than the norm or the prototype in deontic modality, in the development of which factors like passivization, the frequent appearance of general subjects or the factor of social norms as imposer could

Subjectification in (expressions of) epistemic modality

145

have played a significant role. These factors could have contributed to the development of more diffuse attenuation of potency on this facet as well. At this point I am in disagreement with Nordlinger and Traugott (1997), who see the development of the 'wide scope' sense as a natural step in the development of deontic meanings in general.

OS

o

—

•

o

f

IMPOSER (e.g. social norms. expectations)

> o

any situation scope

S/G

(conceptual izer)

Figure 5. 'Wide scope' deontic should/ought SHOULD and OUGHT can be seen as exceptional because in their epistemic meanings they often retain the remnants of a 'wide scope' deontic meaning (deontic overtone): (19)

a. They must all be dead by now b. They should all be dead by now.

This does not occur with MAY ox MUST. Since all epistemic modals are 'wide scope' by definition (restriction of the OS is a necessary condition for them), this fact can be seen as evidence that the latter are prevented because they do not have a 'wide scope' deontic meaning that could be compatible with the ('wide scope') epistemic sense, so this change in OS in the root meaning is not characteristic of other root modals. Please note that the deontic 'positive expectation' is now associated with the speaker, even though no facet of conceptual structure was originally associated with him/her in the root meanings (cf. the behaviour of Dutch M O E T in 4.3).

146

Piter Pelyväs

This analysis of deontic modality suggests that subjectification, though a weak form of it (through correspondence) often does occur, is not a crucial factor in the deontic meanings. Since there is no or little subjectification in comparison with the epistemic senses, no epistemic grounding predication status should be claimed for deontic modals.

4.5.

Epistemic modality

Little needs to be said about the epistemic sense here, since Langacker (1999: 307-308, quoted in 4.2.) neatly summarizes its important characteristics. All I can add is that what he says about modals in general may in some cases apply only or mostly to the epistemic senses. This is apparent in a comparison of epistemics with 'narrow scope' deontics, in which subject control is still quite marked and subjectification is weak at best. All epistemic modals are consistently 'wide scope', i.e. they restrict immediate scope to exclude properties of the deontic domain that are not compatible with the epistemic domain. This is predominantly those aspects of the source domain that will not be identified with the speaker (subjectified) in the extension. As the conceptual schemas indicate, the excluded forces are mainly the ones originally identified with the subject. In this way restriction of immediate scope seems to equal attenuation of subject control, which would not be compatible with the properties of the epistemic domain. Full subjectification occurs only in the epistemic sense. The speaker/ conceptualizer in this role is included in overall scope as a reference point. This has to be direct inclusion, since in all cases where another participant could appear in the deontic sense(s), only the speaker/conceptualizer can now serve as a participant (cf. Dutch deontic vs. epistemic MOET in (15) and (16), or the 'deontic overtone' in epistemic SHOULD/OUGHT). Immediate scope (OS) is always restricted to the profiled situation (the grounded head), an essential property of the grounding predication. In the case of the epistemic modals, e.g. in MAY (Figure 6), I identify the speaker/conceptualizer role with that of the potential counterforce in MAY of ability:24 the speaker is too weak to commit him/herself as to the epistemic status of the situation.

24. As referred to in 4.3, the direct source of epistemic MAY, the earliest epistemic meaning is the now extinct MAY of ability, although in the development of the

Subjedification

in (expressions of) epistemic modality

147

Epistemic meanings are clearly far more subjective than deontic ones: ... the force dynamics are inherent in the conceptualizer's mental activity, hence subjectively construed in the strong sense. (Langacker 1999: 309)

As a result, they fully qualify as grounding predications.

OS

Ο

· > ο qiiv-situation

Λ κ S/G) scope

speaker relinquishes epistemic commitment (grounding)

Figure 6. Epistemic may

Conclusions and further issues The aim of the paper was to show that there is a discrepancy between conceptual content and formal considerations in the definition of epistemic grounding. The category may require redefinition in two directions, since the semantic definition of epistemic grounding would probably exclude deontic modals on the one hand but would include German modals and (at least some) cognitive predicates on the other. Both changes seem to be made impossible by the formal constraint formulated in Langacker (1991b) that the complement of the grounded predication must be a form that is neither finite nor non-finite, since it must profile a process. In the paper I presented evidence that - clearly non-finite forms (the complements of German modals and of cognitive predicates in the 'Raising' or reference-point construction)

fully subjectified meaning the deontic sense may also have had some role (also cases o f D O E R - S / G identity, as in 7 may change my mind').

148

-

-

Piter Pelyväs

may (to some extent) profile a process, since some reduced properties of sequential scanning may be retained in them; some clearly finite forms (the complements of cognitive predicates) may not be fully grounded, since their exceptional syntactic behaviour resembles single clause structure and thus suggests that the apparently complex sentences are in fact grounding predication + grounded head complexes. the conceptual structures of deontic (root) and epistemic modals are rather more different than envisaged in Sweetser (1990) or Langacker (1999), both in terms of attenuation of subject control (diffusion of potency) and in the nature and extent of subjectification. The differences found all point in the direction that only the epistemic senses of the modals should be regarded as grounding predications.

References Hare, R. M. 1971 Practical Inferences. London: Macmillan. Kiparsky, P. and Kiparsky, C. 1970 Fact. In M. Bierwisch and Κ. E. Heidolph (eds.), Progress in Linguistics. The Hague: Mouton. Lakoff, G. 1987 Women, Fire and Dangerous Things. Chicago: University of Chicago Press. 1991 Cognitive versus generative linguistics: how commitments influence results. Language and Communication 11, 1/2: 53-62. Lampert, G. and Lampert, M. 2000 The Conceptual Structure(s) of Modality: Essences and Ideologies. Frankfurt am Main: Peter Lang. Langacker, R. W. 1991 a Foundations of Cognitive Grammar I, II. Stanford, CA: Stanford University Press. First published 1987. 1991 b Concept, Image, Symbol: The Cognitive Basis of Grammar. Berlin and New York: Mouton de Gruyter. 1995 Raising and Transparency. Language 71,1: 1-62. 1999 Grammar and Conceptualization. Berlin and New York: Mouton de Gruyter. Mortelmans, T. 2002 'Wieso sollte ich dich küssen, du hässlicher Mensch!' Α study of the German modals sollen and müssen as 'grounding predications'

Subjedification in (expressions of) epistemic modality

149

in interrogatives. In F. Brisard (ed.), Grounding. The epistemic footing of deixis and reference, 391-432. Berlin and New York: Mouton de Gruyter. this vol. Langacker's 'subjectification' and 'grounding': a more gradual view. Nordlinger, R. and Traugott, Ε. C. 1997 Scope and the development of epistemic modality: evidence from ought to. English Language and Linguistics 1(2): 295-317. Palmer, F. R. 1986 Mood and Modality. Cambridge: Cambridge University Press. Pelyvas, P. 1996 Subjectivity in English. Frankfurt am Main: Peter Lang. 2000 Metaphorical extension of may and must into the epistemic domain. In Antonio Barcelona (ed.), Metaphor and Metonymy at the Crossroads, 233-250. Berlin and New York: Mouton de Gruyter. 2001 a On the development of the category modal: a cognitive view. How changes in image-schematic structure led to the emergence of the grounding predication. In Piroska Kocsäny and Anna Molnär (eds.), Wort und (Kon)text, 103-130. Frankfurt am Main: Peter Lang. 2001 b Extension of should and ought into the epistemic domain. In J. Andor, T. Szücs and I. Terts (eds.), Szines eszmek nem alszanak ... (Szepe György 70. Születesnapjara) [Colorful Ideas Do Not Sleep ... (For the 70th Birthday of György Szepe)], Lingua Franca Csoport, Pecs: 932-945. 2001 c The development of the grounding predication. Epistemic modals and cognitive predicates. In Τ. E. Nemeth and K. Bibok (eds.), Pragmatics and the Flexibility of Word Meaning, 151-174. Amsterdam: Elsevier. 2003 The relationship of conceptual structure and grammatical structure in modals. In L. Komlosi, P. Houtlosser, and M. Leezenberg (eds.), Communication and Culture. Argumentative, Cognitive and Linguistic Perspectives, 101-113. Amsterdam: Sic Sat. Sanders, J. and Spooren, W. 1997 Perspective, subjectivity, and modality from a cognitive linguistic point of view. In Wolf-Andreas Liebert, Gisela Redeker, and Linda Waugh (eds.), Discourse and Perspective in Cognitive Linguistics, 85-112. Amsterdam: Benjamins. Sweetser, Eve E. 1990 From Etymology to Pragmatics. Metaphorical and cultural aspects of semantic structure. Cambridge: Cambridge University Press.

15 0

Peter Pelyväs

Traugott, Ε. C. 1989 On the Rise of Epistemic Meanings in English: An Example of Subjectification in Semantic Change Language 65: 31-55.

Langacker's 'subjectification' and 'grounding': A more gradual view Tanja Mortelmans

1.

Introduction

As is well-known, the English modal verbs are generally considered to be 'grounding predications', i.e. highly grammaticalized linguistic elements the function of which consists in specifying how the designated entity (in this case: the complement process) relates to the ground (defined as the speech event, its participants, and its immediate circumstances, cf. Langacker 1991a) in regard to basic epistemic notions like time and reality. A crucial feature of grounding predications is the subjective construal of the ground, i.e. the fact that despite its pivotal role, the ground remains off-stage: grounding predications by definition do not profile the grounding relationship, but the grounded entity (represented in the case of the modals by the complement process expressed in the infinitive). Assuming that the English modals function as grounding predications, they cannot profile the modal relationship (anchored in the ground), as this would imply that the ground is not construed in a maximally subjective way. So, the modal must, for instance, is said to profile the process deemed necessary in the complement and not so much the modal relation of necessity (Langacker 1991b: 335). On the whole, the German modals1 are less grammaticalized than their English counterparts.2 This is the main reason for Langacker to claim that 1. I will consider the German modals sollen ('shall'), wollen ('will'), müssen ('must'), dürfen ('may'), können ('can') and mögen ('like'). I will also pay some attention to the future/modal auxiliary werden, which is to some extent integrated in the German paradigm of modal verbs (compare for instance Vater

1975, 1997) 2. In contrast to the English ones, the German modals allow for (ungrounded) nonfinite forms (e.g., zu können vs. *to can). More generally, the German modal verb category shows a greater paradigmatic variability as far as modal and temporal forms are concerned than the English ones (the German modals appear in

152

Tanja Mortelmans

they have not acquired the status of grounding predication, but that they are grounded themselves - by tense, person and mood inflections (see Langacker 1991b: 335 for the German modals, and Achard 1998 for a similar claim pertaining to the French modals). This implies that in the German equivalents of the English sentences (la), (2a) and (3a), it is the relation expressed by the modal verb that is grounded by the mood inflection, such that the modal relation is situated within immediate reality (by the grounding predication of indicative mood) or somewhere in non-reality (by the grounding predication of the past conjunctive). On this analysis, the modal relation (as expressed by the modal verb) is considered to be grounded itself, and thus (to some extent) profiled. As grounding predications, the English modals, on the other hand, by definition profile the complement process in all their uses. (1) (2) (3)

a. b. a. b. a. b.

You must come right away. Du musst (= IND+müssen) gleich kommen. You should be more careful. Du solltest (= KONJII+sollen) vorsichtiger sein. That might be true. Das könnte (= KONJII+können) stimmen.

In view of the highly similar function and the intuitive similarity of these specific English and German utterances, this analysis seems somewhat unsatisfying. Moreover, various analyses of German data (see particularly Diewald 1999) have shown that the grammatical properties ascribed to them by Langacker do not hold for every single item to the same extent. The formation of non-finite forms, for instance, is more or less impossible when the modal has an epistemic or a speaker-oriented root meaning. By the same token, some of the epistemically used German modals do not easily form preterite forms,3 which casts doubt on the fact whether they can still be grounded by tense. the present and past tense, have perfect forms and can be combined with werden to form a future tense; as for mood, only the imperative is ruled out - which can be accounted for on semantic grounds). 3. The formation of past tense forms is ruled out for epistemic dürfte (formally a past conjunctive). In the case of epistemic müssen and können, preterite forms typically indicate a shift in perspective, from the current speaker to an embedded one in the text. See for clarification the examples under (20) and the discussion there.

Langacker's 'subjectification' and 'grounding'

153

In this paper, a more gradual view on grounding and subjectification will be presented, which takes local and constructional factors of the specific modal verb into account. This entails that I prefer not to grant or deny the status of 'grounding predication' to a class of verbs, but to look at a number of properties which might enhance the degree of subjectification of a particular modal or even a particular form of the modal. I will concentrate on the German modals, but I believe that a more gradual view on grounding could also be defended for the English ones (see also Pelyvas this volume). One well-known element that warrants such an approach is the observation that a clear-cut dichotomy - with outspoken differences regarding grammatical behavior - has been shown to exist between the root and the epistemic uses of modal verbs (for German, see Diewald 1993, 1999; Helbig 1995; for Spanish, see Cornillie, this volume), whereby the epistemic uses show a higher degree of formal grammaticalization than the root ones. But it is not only the semantic distinction between root and epistemic meaning (reflected by particular grammatical properties) that opens up the possibility to consider some modals in some uses as more subjectified (and therefore closer to the grammatical class of grounding predication) than others.

2.

Parameters of subjectification

In the main part of this paper, I will discuss five parameters that to some extent influence the possible grounding status of a modal. These parameters include the syntactic environment in which the modal occurs (i.e. its preferred types of subject and complement verb (2.1)), the sentence type (2.2), the presence of explicit speaker-oriented expressions like ich denke (2.3), the influence of negation (2.4) and the morphological flexibility of the modal itself (2.5). These parameters were detected on the basis of extensive corpus research (see Mortelmans 1999), the aim of which was to establish to what extent the German modals could be considered as 'grounding predications' (contra Langacker's (1991a-b) claim that they don't function as such at all). It should be noted that I do not intend to provide an exhaustive list of factors influencing the possible grounding status of a particular modal. Moreover, in a number of cases (e.g. negation) more investigation into the exact nature of the relationship between subjectification/grounding and the

154

Tanja Mortelmans

factor under discussion seems to be called for. In fact, one of the aims of this paper is to show that the inclusion of natural language data necessitates a discussion of additional elements - which in more theoretically oriented work are often neglected or simply overlooked.

2.1.

The (typical) nature of the subj ect and the complement verb

In Langacker (1999), the evolution of the English modals - in their root and epistemic uses - is described in terms of a process of attenuation of subject control, both with respect to the source or locus of potency and the target at which this potency is directed. It is claimed that neither source nor target of the modal verb's potency are typically linked to the subject (anymore), which is taken to be indicative of a high degree of subjectification. Moreover, the source of potency is no longer identified with the subject, but is implicit and subjectively construed. It may be the speaker, but it need not be . . . . It is not necessarily any specific individual, but may instead be some nebulous, generalized authority. In other words, the source of potency is highly diffuse. ... Nor is the subject necessarily the target of the potency, which is also diffuse. Although the modal force may be directed at a specific individual - be it the subject, the addressee, or some third party - we see ... that this is not always the case. The force is simply directed toward realization of the target event. (Langacker 1999: 308)

Pelyvas (this volume) relativizes this statement by pointing to the fact that the above description of the semantics of the English root modals mainly fits in with the semantics of 'wide scope' deontic should/ought to, but is less suitable to account for the more prototypical uses of the English root modals (must and can, for instance), which favor a narrow-scope reading.4 Pelyvas is probably right in pointing out the fact that the individual modals differ with respect to their scope preferences, and I believe this is an observation of major importance. It could be questioned, however, whether can and must are indeed to be considered as better (i.e. more prototypical) representatives of the modal category than e.g. should or ought to. Why should this be so? For one thing, the mere frequency of should (which in 4. Note, however, that both root can and must are compatible with passive complements, or with expletive subjects, both leading to a wide-scope reading of the modal.

Langacker's 'subjedification' and 'grounding'

155

present-day English occurs considerably more often than must) calls the alleged prototypical status of must into question. 5 On the basis of their frequency, can and will ought to be considered as prototypical modals, whereby it should be noted that the modal verb will has posed some problems for standard force dynamic accounts (Sweetser 1990: 55, see also Langacker 1991: 273ff.). In view of these observations, it would be wiser just to focus on the differences between the individual modals with respect to scope preferences, and to keep Langacker's observation in mind that also within the root domain - attenuation of subject control (which is typical of a wide-scope reading) is indicative of an increased degree of subjectification. If we then turn to the German modals, more in particular to sollen and its past conjunctive form sollte, we notice that scope differences are indeed an important factor with respect to the degree of subjectification of a particular modal. As its English counterpart shall, the original meaning of sollen is connected with an obligation that arises out of some kind of debt on the part of the subject (compare present-day English owe). Typically, then, both source and target of potency are clearly identifiable in the early uses of this verb, and can both be related to the subject, as it is the subject who has somehow created the debt (hence functions as locus of potency) and at the same time has to pay (hence functions as target of potency). Both locus and target appear on-stage, and are thus objectively construed, as in example (4). (4)

ich I

sol owe

im him

einen one

schillinc (DWB 1984) shilling

In present-day German, the indicative of sollen has to some extent retained this original usage, whereby both source and target of the modal relation appear on-stage. It has to be noted, however, that the locus of potency is no longer the grammatical subject, but typically has to be equated with a third 5. Based on findings in the British National Corpus (BNC), can and will/would are the most frequent modals in present-day English, should ranks in between, and must is considerably less frequent than the latter. So, must tagged as VMO [modal auxiliary verb, i.e. + infinitive] occurs 69 332 times (in 3 544 texts), should as VMO 109 156 times (in 3 732 texts), can as VMO 231 722 times (in 3960 texts), will as VMO 244 050 times (in 3854 texts) and would as VMO 245 802 times (in 3863 texts). Thanks are due to Richard Kastein for looking these figures up for me.

156

Tanja Mortelmans

person that is neither subject nor speaker. So, in example (5), the grammatical subject ich functions as target, whereas the previously mentioned participant die mini is the locus of potency. (5)

ahja die mini hat mir ne sms geschrieben und hat gemeint ICH soll sie anrufen aber ich hab ihre nummer nichtlkannst du sie mir geben du hast eh meine nummer sende einfach!danke gogy (http://xxxfreshgxxx.uboot.com/) 'well yes mini has written me an sms and said I had to call her, but I haven't got her number, can you give it to me, because you've got my number just send it! thanks gogy'

The past conjunctive of sollen, however, does not normally refer to an identifiable third element as locus of potency. Rather, (implicit and thus subjective) social norms or expectations (with which the speaker typically sympathizes) are to be considered as the implicit force behind the obligation. Hence, the locus of potency can be claimed to receive a more subjective construal with the past conjunctive than with the indicative. (6)

Nomags Gedanken: "So ein beschissenes Wochenende. Die arme Caroline. ICH sollte sie anrufen, und mit ihr über die ganzen Ereignisse reden, vielleicht hilft ihr das. " (http://www.stud.tu-muenchen.de/~andre.balibey/kamp/7v/run6 nomag.html) 'Nomag's thoughts: "Such a lousy week-end. Poor Caroline. I should call her, and talk to her about everything that has happened, it'll help her perhaps'"

In fact, with past conjunctive sollte, it can be observed that not only the locus of potency, but also the target of potency is typically attenuated, as sollte often occurs with passive complements or active complements with inanimate subjects (see table 1). Also, animate, but indefinite subjects (e.g. man 'one') are frequently found with past conjunctive sollte, again causing the subject to be vague and undetermined. It can be concluded that the modal relation expressed by sollte is thus not typically directed at the overt subject of the clause, representing a well-delineated participant. Instead, the speaker seems to profile the state of affairs which - in her opinion should be realized, and this 'wide-scope' reading of the modal (with clear

Langacker's 'subj edification' and 'grounding'

157

attenuation of subject-control) points to an increased degree of subjectification of the modal relation in the case of sollte. Table 1.

Subjects and complements with past conjunctive sollte

Sollte main clause past conjunctive H85 (written) WK (spoken) Total

comple- complement: ment: active; passive subject: inanimate 16 16

Complement: active; subject: indefinite, animate6 12

complement: active; subject: animate, definite 5

Total number

13

22

29

13

77

29

38

41

18

126

49

The following examples illustrate the various categories: dummy pronoun es with passive complement verhindert werden in (7a), dummy es with active (state) verb geben in (7b), indefinite human subject man with active complement blicken in (7c), and the rather infrequent combination of a definite human subject with an active complement anhören in (7d). (7)

a. Es_ sollte jedoch im Interesse der französischen Wirtschaft verhindert werden, daß die nötige Privatisierung zerredet wird. Die Gefahr besteht. Denn die Materie ist kompliziert. (Mannheimer Morgen, 25.11.1985, p. 3, Entscheidung im März) 'For the benefit of the French economy it should be avoided that the necessary privatization is flogged to death. The danger exists. Because this issue is complex' b. Auf der Ebene der Regierungsbezirke sollte es mindestens einen solchen Fachberater geben. (Mannheimer Morgen, 31.05.1985, p. 22, Gegen die Kasernierung in Elite-Kader schmieden) 'There should at least be one such expert adviser at the level of the governmental districts' c. Dennoch sollte man über seinen eigenen Fächerzaun blicken. Warum nicht Sinologie machen neben dem Wirtschaftsstudium?

6. The pronouns man 'one', jeder 'everyone' wer 'he/she who', niemand 'no one', manch einer ('many a person') refer to indefinite, animate subjects.

158

Tanja Mortelmans (Mannheimer Morgen, 31.08.85, p. 12, Belegen: mehr Lust oder Frust?) 'Still, one should look over one's own fence. Why not study sinology besides economy?' d. Hinsken: Das sollte sich die Frau Saibold auch anhören! (Bundestagsprotokolle, 2. Hj. 1989, Sitzung Nr. 178, Bd. 151, p. 13685-13689, 89.11.29, p. 13687) 'Mrs Saibold should listen to that as well!'

Note also that in large corpora past conjunctive man sollte occurs more than twice as often as its indicative counterpart man soil. [PUBLIC-Corpus, 7 accessed in May 2003] a. man soil 3586 occurrences b. man sollte 8596 occurrences An estimation of a modal's degree of subjectification should therefore also include an analysis of the syntactic patterns in which the modal typically occurs.

2.2.

Sentence types

Often neglected is the fact that the German modals behave differently in different sentence types (Diewald 1999, for instance, hardly pays any attention to this factor). As the unmarked 'default' sentence type, the declarative main clause presents a syntactic environment in which the modal typically conveys its widest variety of uses and meanings, some of which are more grammaticalized and/or subjectified than others. Here, then, properties of the subject and the complement might influence the possible grounding status of the modal. Let's consider the modal müssen ('must'). In the following main clauses (8a and b), the verb is combined with a human subject (functioning in both cases as the agentive target of potency) and a dynamic complement verb. The locus of potency can be easily reconstructed: it is to be equated with identifiable judicial authorities in (8a) and with the subject itself (responding to some kind of internal need, possibly 7. The PUBLIC-corpus contains all the written texts made electronically available by the Institut für deutsche Sprache (Mannheim).

Langacker's

'subjectification' and 'grounding'

159

provoked by outside circumstances) in (8b). In both cases, there is force interaction between target and locus of potency, and this is clearly profiled. Both situations are objectively construed; in both cases, the modal relation can be said to be grounded by tense. (8)

a. Drogenverdacht: Friedman musste [= müssen/TND PRÄT] Haarprobe abgeben. (www.spiegel.de/panorama/0,1518,252618,00.html) 'Suspicion of drug abuse: Friedman had to provide a hair sample' b. Dr. Lulatsch muss [= müssen!IND PRÄS] auch lachen. (www. triticea. de/mikrobiologie .html) 'Dr. Lulatsch has to laugh as well'

In the following example, müssen appears in combination with a second person subject (du), referring to the addressee. Typically, the speaker wants the addressee to do something for him; the target of potency is therefore the addressee,8 whereas the locus of potency corresponds to the speaker. Following Pelyvas (this volume), the speaker/conceptualizer indirectly enters the scene (via so-called correspondence), which is regarded as an instance of (albeit minor) subjectification. (8)

c. ein dialog zwischen einem box-champion und seinem manager lieber du musst [= müssen/ IND PRÄS 2Sg] mir den gefallen tun du musst [= müssen/ IND PRÄS 2Sg] mich gegen ricky nelson kämpfen lassen. (DSK/DAS.00000, swf, november 1966) 'it's a dialogue between a boxing-champion and his manager my friend you got to do me a favor you must let me fight against ricky nelson'

Example (8d) exemplifies epistemic müssen, again in a declarative context. The modal clearly has wide-scope here, and there is clear attenuation of subject control (the subject das refers to a state of affairs, and not to a welldelineated participant). Note, however, that the modal relation can still be

8. Note that for directive uses like in (8c) above, the target could also be understood in terms of "realization of the target event" (Langacker 1999: 308), i.e. the speaker does not so much focus on the subject as goal of the directed potency, but wants the subject to act. On this analysis, the modal relation would receive a slightly more subjective construal than suggested above.

160

Tanja Mortelmans

regarded as being profiled to some extent: the speaker objectifies the epistemic necessity by explicitly presenting the evidence upon which her assessment is based (denn er schrieb später 'because he wrote later on'). This evidence, then, can be regarded as an on-stage locus of potency. (8)

d. Julius wurde 1845 ... aufs Gymnasium geschickt. Das muß ihm sogar Spaß gemacht haben, denn er schrieb später: "Die fünf Jahre, die ich hier zugebracht, sind die glücklichsten meines Lebens gewesen " (http://home.t-online.de/home/zerries/jr.htm) 'In 1845 Julius was sent to grammar school. He must have enjoyed it, because he wrote later on: "The five years that I have spent here have been the happiest of my life'"

We find, however, particular usages of the modals which are tied to particular syntactic environments, in which the modals have developed clearly subjectified uses. It is therefore important not to concentrate on declaratives, but to look at other clause types as well. Again, the behavior of the modal verb sollen presents an interesting case. First, when the speaker uses indicative sollen in questions, the addressee is typically implicitly construed as locus of potency of the modal relation: soll ich etwas tun, thus typically means: 'do you want me to do something for you'? 9 The following excerpt nicely illustrates that the identification10 of the locus of potency with the addressee is implicit and might therefore also give rise to misunderstandings in this particular situation (for a more detailed discussion of sollen in interrogative contexts, see Mortelmans 2002). (9)

dieses "soll ich dich nach hause fahren " konnte ich noch nie leiden, entweder heisst es "fahren bzw. gehen WIR? " oder aber

9. This could be considered as an instance of intersubjectification as introduced by Traugott (1999), i.e. "a mechanism whereby meanings become more centered on the addressee" (see also Traugott and Dasher 2002: 19-24). Note, however, that intersubjectification is assumed to occur after subjedification has taken place (Traugott and Dasher 2002: 24), an assumption that in the case of indicative sollen does not seem to hold, as the addressee-oriented meanings have originated independently of a prior speaker-oriented meaning. 10. Again, the identification of the locus of potency with the addressee can be regarded as a case of correspondence, which therefore presents only a minor instance of subj edification.

Langacker's 'subj edification' and 'grounding'

161

"ok, war lustig, ich ruf mir ein taxi. "... männer, lasst doch diese Spielchen, macht aus dem "soll ich dich bringen" ein "fahren wir zu dir" und die sache ist klar. (http://schmutzig.twoday.net/stories/31635) Ί could never stand this 'shall I take you home' [said by a man to a woman]. Either one says 'let's go' or 'okay, it was nice, I'll call me a cab' Men, please, stop these games. Say instead of 'shall I take you home' 'let's go to your place' and everything is clear.' Note in this connection that the modal verb müssen (which in its root meaning shows some semantic overlap with root sollen, at least in declarative contexts) has not developed an addressee-oriented use in interrogative environments. In questions, müssen typically profiles the necessity expressed by the modal, as in Muss sie sterben? ('Will she have to die?'). If we look at past conjunctive uses of sollen in questions, a different image emerges. For one thing, ,so///e-questions are not normally used to obtain an answer from the addressee (real questions hardly occur), but are highly speaker-oriented, in that the speaker - very often by means of a rhetorical question - typically recommends a particular way of action. It could be claimed for cases like (10) that not so much the modal relation, but the complement process is profiled: the speaker does not question the desirability as such, but presents a state of affairs (expressed in the infinitive complement) for action. Note, by the way, that rhetorical questions containing sollte are typically negated - we will come back to the relation between grounding and negation in section 2.4. (10)

Sollte sich die Gewerkschaft nicht erst einmal finden ..., bevor zum letzten aller Streikmittel gegriffen wird? [Berliner Zeitung, 02.02.90, p. 1] 'Shouldn't the trade union first come to terms with itself, before it turns to the ultimate strike action?'

Particularly interesting is the question type illustrated in ( l l a - b ) . Here, sollte patterns in a yes-no question, which I have labeled the deliberativemirative so//fe-question (see Mortelmans 2002 for extensive discussion of this question-type). This use of sollte provides a case of high grammaticalization (for one thing, there is a clear reduction of syntagmatic and paradigmatic variability (see Lehmann 1985), as the meaning of sollte seems to be tied to main clause yes-no questions, and only the past conjunctive form

162

Tanja Mortelmans

of the verb can appear) and subjectification: on the basis of observation of the surrounding world, the speaker arrives at a particular supposition (which often causes surprise, hence the label 'mirative' see Delancey 1997). In this particular use, modal verb and sentence type seem to reinforce each other, which also implies that one cannot dissociate this use of sollte (and its possible grounding status) from the sentence type in which it occurs. (11)

a. Erst vor 6000 Jahren zog sich die letzte große Eisdecke vom kanadischen Festland zurück. Wodurch werden diese Kälteperioden ausgelöst ? Sollte es einen festen Zyklus geben, eine Art kosmischen Puls, der alle 250 Millionen Jahre einmal schlägt? [Mannheimer Morgen, 15.03.85, p. 3: Kosmischer Puls] O n l y 6000 years ago the last big layer of ice disappeared from the Canadian continent. What causes these periods of cold weather? Is there perhaps a fixed cycle, some kind of cosmic pulse that beats every 250 million years?' b. Subject: Sollte ich die Liste abbestellt haben!? Hallo zusammen! In der letzten Woche gingen bei mir ganze 0 (in Worten: NULL) eMails von dieser Liste ein... Ist das normal (z.B. weil keiner was schickt) oder sitze nur ich auf dem Trockenen? (http://lists.suse.com/archive/suse-programming/2002-0ct/0057. html) 'Subject: Have I unsubscribed from this list? [Is it possible that...]' 'Hi there! Last week I received 0 emails from this list. Is this normal (for instance because no one is mailing anything) or am I the only one to receive nothing?'

This use of sollte, then, can be taken to be maximally subjective, both in terms of the source of potency (the implicit speaker) and the target of the modal relation (which is always a state of affairs). Conceptually, the above usage is clearly linked to the use exemplified in (12), in which sollte shows up in a conditional protasis. Again, the verb has no (deontic or epistemic) meaning comparable to the normal meanings of the modal (in terms of obligation or necessity); its semantic contribution seems to be limited to

Langacker's 'subjectification' and 'grounding'

163

opening up a mental space - typically within potential reality" - with respect to which the conditional apodosis is situated. (12)

sag mal ich glaub ich stehe im wald. Sollte ich dich in die finger bekommen kannst dich auf was gefaßt machen. Was fallt dir eigentlich ein dich unter meinem namen und meinen bildern irgendwo breit zu machen? (http://hexenluda.uboot.com/) Ί am absolutely flabbergasted. If I ever get hold of you, you'd better be prepared. How dare you use my name and my images?'

It should not come as a surprise that these highly subjectified uses of the modal occur in yes-no questions (or in conditional subclauses like (12) whose structure is derived from the prototypical yes-no question; see e.g. Hentschel 1998), which - in contrast to question-word questions - do not presuppose the state of affairs. That is, the epistemic status of the proposition as coded by the (interrogative) sentence type is in line with the semantics of the modal, which does not presuppose the factuality of the coded event either.

2.3.

The presence of hedges and other 'subjective' features

Another factor to be considered is the frequent presence of other 'subjective' (in the sense of speaker-oriented) markers, like hedges (/ think) and phrases explicitly referring to the speaker (like I'm sure). Of course, here the speaker (as an element of the ground) is brought on-stage (and thus clearly objectified), but it seems to be a typical characteristic of highly subjectified modals to sometimes objectify the speaker-involvement, resulting in a subtle alternation between higher and lower modes of subjectification of the speaker-role.12 11. The conditional protasis containing sollte can only refer to potential states of affairs, not to counterfactual ones (* Sollte ich jünger sein, würde ich mein Leben anders organisieren 'Should I be younger, I would organize my life in a different way'). 12. This is fully in line with Langacker (this volume), who notes that with epistemic modals in English, the speaker is construed somewhat more objectively, as she "assumes the additional responsibility of making the epistemic judgment signalled by the modal" resulting in a presence which is "a bit more palpable". It thus does not come as a surprise that this slightly increased degree of speaker

164

Tanja Mortelmans

A first observation in this respect stems from Coates (1983), who notes for English epistemic must that 23 out of 92 occurrences in the (spoken) Survey-corpus are accompanied by 'subjective' hedges, which in Coates's view "underline the fact that Epistemic modals are essentially subjective, that is, for the most part they focus on the speaker's attitude to the proposition expressed in the main predication" (Coates 1983: 46). Table 2.

Speaker-oriented hedges with epistemic must (Coates 1983: 46)

hedges with must I think I mean Isuppose Ifancy I lake it I would guess (13)

occurrences 15 3 2 1 1 1

NSC: (laughs) I think it's nice to have a few sort of 3m - pinpoints in the future.. Β: I think it must be very nice (example from Coates 1983: 41)

By the same token, the absence of speaker-oriented hedges with seemingly subjective meanings like the epistemic ones could be taken to indicate a lesser degree of subjectification than one would normally expect. Let's take the case of German epistemic müssen, which - when compared to English epistemic must - is hardly accompanied by hedges like ich meine or ich denke,13 This observation, then, suggests that German epistemic

objectivity is linguistically mirrored by a tendency towards an explicit mentioning of the speaker/conceptualizer entity. 13. The epistemic meaning of German müssen is on the whole rather infrequent, and combinations with hedges are difficult to come by. In Mortelmans (1999), I have reported on a number of corpus analyses (all on written material taken from present-day German texts) containing müssen. Three of them are relevant here: in a first general sample containing 426 müssen-utterances, I found 35 epistemic instances, none of them being accompanied by a hedge. A second sample contained 118 utterances with müssen combined with the state verb haben ('to have') in the infinitive; 60 of them were epistemic. Again, in none of the utterances, müssen was accompanied by a hedge. Finally, in a sample of 200

Langacker 's 'subjectification' and 'grounding'

165

müssen in the majority of its occurrences seems to underscore the relatively objective inferential moment (typically based on accessible evidence) more strongly than the epistemic evaluation in which the inference results (and to which the speaker is implicitly or explicitly committed). (14)

Und immer wieder fragte man mich nach Eva Braun. Ich wußte damals gar nichts von ihr. Sie muß sehr unscheinbar ausgesehen und immer geschwiegen haben, daß sie mir nicht aufgefallen war. 'Again and again people asked about Eva Braun. I didn't know anything about her at the time. She must have looked very inconspicuous and have been silent all the time, as I didn't notice her.'

So, I would claim that German 'epistemic' müssen also profiles the modal relationship of necessity based on typically accessible evidence (which might even be taken to function as locus of potency)14 apart from the complement process. On the whole, therefore, I take epistemic müssen in German not to be as subjectified as its English counterpart. The mere infrequency of the epistemic reading in German, by the way, also points into the same direction (in Coates's material, the epistemic reading accounts for about 50 % of all uses of must in the oral corpus, and about 30 % in the written one).15

müssen + sein ('to be') utterances, 62 of which were epistemic, not a single instance with a hedge was found. 14. Of course, it is the speaker who draws the inference based on the evidence at hand. One could thus alternatively claim that with epistemic müssen, both evidence and speaker (as an element of the ground) function as loci of potency, whereby subtle differences between more objectified (i.e. linguistically available) and less objectified (implicit) construals of the evidence at hand can be discerned. Another factor possibly leading to a more subjectified use is the presence of modal particles like wohl, which often combine with the German modals. The interaction of these particles with the individual modals is an intricate matter, though, which would go beyond the theme of this paper. 15. In the oral Survey corpus sample, 92 out of 200 instances were classified as epistemic, whereas 74 out of 236 instances of the Lancaster corpus are given the label 'epistemic' (Coates 1983: 32).

166

Tanja Mortelmans

Table 3. Percentage of epistemic readings of the German modals (based on Diewald 1999: 217) German modals dürfen mögen sollen können müssen wollen

percentage 15,8% 14,6 % 8% 7,2 % 4,9 % 0%

There are other modal verbs, however, which frequently occur in combination with hedges explicitly referring to the speaker. Again, root sollte is one of them. Note that deletion of the hedges, as in (15a', b' and c'), does not affect the inherent speaker-orientation of these utterances. (15)

a. Deshalb meinen wir, daß man [..] im mittelständischen Bereich ansetzen sollte [...]. (Wendekorpus) 'That's why we think that one should start in the middle class.' a'. Man sollte im mittelständischen Bereich ansetzen. O n e should start in the middle class' b. Ich denke, das sollte man aufmerksam prüfen. (Wendekorpus) Ί think that should be carefully investigated.' b'. Das sollte man aufmerksam prüfen. 'That should be carefully investigated.' c. Ich denke, daß man deshalb von dem Beginn einer zweiten Etappe [...] sprechen sollte. (Wendekorpus) Ί think that therefore one should speak of the beginning of a second phase.' c'. Deshalb sollte man von dem Beginn einer zweiten Etappe sprechen. 'Therefore one should speak of the beginning of a second phase.'

Compare in this respect the following utterances containing root soli in the indicative. Here, deletion of the phrase marking the speaker-involvement (ich bin der Ansicht lit. 'I am of the opinion', nach unserem politischen Willen 'in accordance with our political will') leads to an utterance for which it is hard to decide whether the speaker reports on intentions or

Langacker's

'subjectification'

and 'grounding'

167

plans that originated in a third party or whether she expresses her personal point of view. (16)

a. Ich bin der Ansicht, Deutschland soll wiedervereinigt werden. (Wendekorpus) Ί am of the opinion that Germany should be reunited.' a'. Deutschland soll wiedervereinigt werden 'Germany should/is to be reunited.' b. Man kann gar nicht genug Klarheit darüber schaffen, daß die Westgrenze Polens nach unserem politischen Willen dauerhaften Bestand haben soll. (Wendekorpus) O n e can't be too clear on the issue that Poland's western border in accordance with our political will should have long-lasting existence.' b \ Die Westgrenze Polens soll dauerhaften Bestand haben. 'Poland's western border should/is to have long-lasting existence.'

Like root sollte, epistemic werden is often accompanied by explicit markers of speaker-involvement. Their deletion, however, does not diminish the inherent speaker-oriented character of the verb. Note also that the rhetorical sollte-question illustrated in (17d) - a parallel 'hedge-less' example was presented in (10) - is often accompanied by hedges. (17)

a. Dies wird, nehme ich an, auch Walter Momper schaffen. Ί believe Walter Momper will also manage this.' b. Ich denke, wir werden da zu einem guten Ergebnis kommen. Ί think we will arrive at a good result there.' c. [...] daß [...] die Bundesregierung Gelegenheit hat und, wie ich vermute, nehmen wird, sich zu diesen Fragen zu äußern '[...] that the federal government has and, as I suspect, will also take the opportunity to give its view on these matters' d. Ich meine, sollte so was nicht von Herzen kommen? [http://www.planetliebe.de/vbb/archive/index.php/t-2288] Ί mean, shouldn't something like that come from the heart?'

The above should not suggest that I regard the appearance of hedges as a decisive criterion for subjectification. This would go against the standard (Langackerian) interpretation of subjectification, in which the notion of

168

Tanja Mortelmans

implicitness is crucial. Still, a look at language data shows that speakers like some kind of variety in their construal of the ground, and that speakeroriented markers tend to pop up with subjectified modals. So, it is not so much the presence of speaker-oriented markers with inherently subjective predications, but rather their absence with seemingly subjectified ones (like the epistemic modal müssen) which is of importance here.

2.4.

The presence of negation

Langacker (1991a: 134) has stressed the close association of clausal negation in English with the grounding predications of a finite clause (tense and modals). In fact, in a number of cases, the modal's combination with a negation marker seems to enhance the modal's subjectivity. Let's take the case of German dürfen, meaning 'may' (see also Mortelmans 2003 for discussion). Diewald (1999) attributes a reactive component to the semantics of dürfen·, when a speaker uses dürfen to express a permission, she is taken to respond to a previous wish of the addressee of the permission (typically the subject): "dürfen [...] reacts to a previous communicative move, which is aimed at receiving a directive (the permission) to perform the action" (Diewald 1999: 132; my translation). In a force dynamic account (see e.g. Talmy 2000) one could say that the original initiative with dürfen lies in the hands of the addressee of the permission, typically the sentence subject, who has somehow asked to receive it. The original locus of potency is therefore not to be identified with the speaker, but is onstage (and thus objective). Negating permission, however, can be regarded as a clear intervention by the (off-stage) speaker and is therefore likely to effect a rearrangement of the force distribution. A noteworthy observation in this respect pertains to the generally high amount of negated utterances with German dürfen: the modal figures considerably more often in negated than in affirmative environments. Table 4. negation with dürfen dürfen /INDICATIVE (various forms) + negation marker

150 (see Mortelmans 1999: 714) 126

du darfst (= dürfen PRES IND 2nd sg) + negation marker

66 (see Mortelmans 2002) 41

Langacker's

'subj edification'

and 'grounding'

169

In directive environments,16 negated instances of dürfen typically invite a speaker-oriented, directive interpretation, in which the speaker urges the addressee not to perform a particular action. So, rather than focusing on the absence or presence of a permission, it is the final result (the event not taking place) which is in the center of attention (and could thus be regarded as target of the directed potency). So, in the examples under (18), the speaker wants the addressee to comply (instead of merely describing the absence of permission). (18)

a. Geh weg, du darfst meiner Mami nicht weh tun! 'Go away, you mustn't hurt my mummy!' b. Nein, Emanuel, da darfst du dich nicht einmischen. 'No, Emanuel, you mustn't interfere in that.' c. Nicht sprechen, Michael, du darfst jetzt nicht sprechen. 'Don't speak, Michael, you mustn't speak now.' d. "Das darfst Du nicht riskieren", riet der Stürmer eindringlich ab. '"You mustn't risk that", the striker insistently advised.'

Affirmative instances containing du darfst, on the other hand, seem to be more compatible with non-directive, descriptive interpretations, in which the speaker role is less prominent. An interesting observation in this light is the fact that in the corpus at hand, 5 out of 25 affirmative instances appear in a conditional context - whereas only 1 of the 41 negated instances features in the apodosis of a conditional. In the conditional cases, the permission expressed by darfst depends on a condition, the realization of which lies in the hands of the sentence subject. In terms of force distribution, one could say that in (19) both the on-stage subject participant and the speaker are somehow involved as loci of potency in the modal relation (there is therefore less attenuation of subject control), whereas in (18) the main force is clearly with the (off-stage) speaker. By the same token, the permission as expressed by darfst seems to be more in focus (more strongly profiled) in the affirmative conditional cases in (19) than in the

16. A directive environment refers to a speech situation that easily invites a directive interpretation. Expressions in which the subject equals the addressee (you) functioning as agent of the action specified in the infinitive qualify as such (e.g. you must do that).

170

Tanja

Mortelmans

negated utterances in (18): the condition pertains to the granting of permission, and not to the realization of the state of affairs in the infinitive. (19)

a. Die ... Abgeordnete ... erklärte, die Koalition gehe bei der Diskussion um den Ladenschluß nach dem Motto vor: "Wenn Du mir am Samstag eine Stunde gibst, darfst Du am Donnerstag abend bis 20.30 Uhr ausweiten " ... . [Mannheimer Morgen, 11.05.1989] 'The member of parliament stated that the coalition in its discussion of the shop closing time proceeded according to the motto: "If you give me an hour on Saturday, you can stay open till 8.30 p.m. on Thursday ... ."' b. Natürlich darfst Du heute das Fernsehen ausfallen lassen, wenn Dir Deine Hausaufgaben wichtiger sind! [.Mannheimer Morgen, 14.11.1995] O f course you can drop television today, if your homework is more important to you.' c. Wenn du mir das Rätsel lösen kannst, darfst du meine Tochter heiraten. [Mannheimer Morgen, 24.12.1995] 'If you can solve the riddle, you are allowed to marry my daughter.'

A similar link between negation and a stronger degree of subjectification holds for German können·, according to Diewald (1999: 219), the indicative of können strongly prefers negation in its epistemic use, a preference which is absent in the dynamic, non-epistemic uses of the verb. How negation affects and interacts with the other modals and their degree of subjectification, is a matter of further investigation.

2.5.

Formal flexibility of the modal

Even within the more grammaticalized epistemic paradigm, differences between the individual modals regarding formal restrictions can be discerned. Epistemic müssen and können, for instance, are found to have preterite forms, in which the modal verb could be claimed to be grounded by the tense predication creating an alternative ground (identified with a figure in the text) from which the epistemic evaluation is effected.

Langacker's 'subjectification' and 'grounding'

171

Therefore, transformation to past tense implies perspectivization, a shift to the perspective of another, embedded current speaker, whose moment of speaking is positioned in the past. In other words, when the epistemic modification itself is situated in the past, as in had to have, this causes a double manifestation of subjectivity: the epistemically modified information (subjectification) is interpreted as bound to a subject in the text (perspectivization). (Sanders and Spooren 1997: 103)

So, in both cases under (20), the epistemic evaluation expressed by past tense müssen must be traced back to a subject in the text rather than to the current speaker. [examples taken from Milan 2001: 141] (20) a. Den Gerüchen aus ihrer Küche nach mußten sie sich an fetten Speisen ... überfressen haben. 'Judging by the smells from their kitchen they had to have gorged themselves on fatty dishes' b. Sie entstammte einer Familie, die verarmt und wieder zu Reichtum gekommen war. Sie mußte eine natürliche Beziehung zum Geld haben. ' She came from a family that had become poor and then got rich again. She had to have a natural relationship with money' This perspectivization, however, seems to be impossible for the highly grammaticalized epistemic dürfte,17 which is more strongly tied to the conceptualizer (ground). (21)

Im vergangenen Jahr wurde dieses Material auf CD-rom angeboten. Schon damals dürfte (* durfte: IND PRÄT) diese CD-rom eine der bestausgestatteten Einleitungen in das Alte und Neue Testament gewesen sein. (http://de.groups.yahoo.com/group/mikra/message/473) 'Last year this material was compiled on CD-rom. Already at that time this CD-rom was probably one of the best introductions to the Old and New Testament'

17. The fact that epistemic dürfte does not allow past-tense forms anymore, can be regarded as an instance of loss of paradigmatic variability, one of Lehmann's (1985) six grammaticalization parameters.

172

Tanja

Mortelmans

In other words, modal verbs with a high degree of formal grammaticalization (mirrored in the case of the German modals by a decrease of their flectional properties) are expected to score higher with respect to their degree of subjectification.

3.

Summary

As was already mentioned at the beginning of this paper, the above list is probably not exhaustive or even unproblematic. 18 Still, I believe that a number of the parameters that have been introduced affect the degree of subjectification of a particular modal in a way that previous research has not dealt with in sufficient detail. First, the preference of a particular modal (form) for wide-scope readings (with clear attenuation of subject control) can - even in root meanings - be taken to reflect a tendency towards higher subjectification. Second, the sentence type might influence the degree of subjectification, to the extent that a particular sentence type (e.g. the interrogative) 'joins in' with the modal verb's semantics to construe the state of affairs in different and potentially more subjective ways than in the 'ordinary' declarative uses of the modal. Third, the alternation between higher and lower modes of subjectification of the speaker-role (the latter achieved by means of explicit speaker-oriented hedges like I think and I mean) seems to be typical of the highly grammaticalized English modals. The absence of such speaker-oriented hedges with seemingly subjectified modals (e.g. epistemic müssen in German) is taken to point to the fact that German epistemic müssen generally effects a less subjective construal than its English counterpart. Fourth, negation is taken to interact with the modal verb such that with some modals (e.g. German dürfen and können) negated instances increase the degree of subjectification of the modal relation. Finally, differences between individual modals with regard to their formal flexibility also point to differences with regard to their general degree of subjectification.

18. In a standard Langackerian account, the onstage-appearance of the speaker, which is effected by hedge-like formulae, would rather be an argument against subjectification than an argument in favour of it. Still, the frequent cooccurrence of hedges with seemingly highly subjectied modals cannot simply be ignored.

Langacker's 'subj edification' and 'grounding'

173

A number of obstacles still remain, though. For one thing, how does subjectification ä la Langacker relate to the notion of subjectification presented by Traugott (1989, 1995)? I believe that especially in areas in which modals have not acquired a maximal degree of subjectification (root modality, German modals in general), both notions are helpful in that the onset of the subjectification process (in Langacker's terms) can be observed when meanings become indeed more centered on the speaker (and/or addressee). The emerging image is one of a subtle and rich spectrum, in which more and less subjectified uses of modals reflect different construals of the modal relation. From a synchronic perspective, every modal seems to have its typical range of uses, which - at least for the German ones - are tied to particular forms of the modals and to particular syntactic contexts in which these can occur.

References Achard, Michel 1998 Representation of Cognitive Structures. Syntax and Semantics of French Sentential Complements. (Cognitive Linguistics Research 11.) Berlin/New York: Mouton de Gruyter. Coates, Jennifer 1983 The Semantics of the Modal Auxiliaries. London/Canberra: Croom Helm. Cornillie, Bert this vol. On subjectification. Formal and conceptual criteria for epistemic modal grounding predications in Spanish. Delancey, Scott 1997 Mirativity. Lhe grammatical marking of unexpected information. Linguistic Typology 1: 33-52. Diewald, Gabriele 1993 Zur Grammatikalisierung der Modalverben im Deutschen. Zeitschrift für Sprachwissenschaft 12, 2: 218-234. 1999 Die Modalverben im Deutschen. Grammatikalisierung und Polyfunktionalität. Tübingen: Niemeyer. Heibig, Gerhard 1995 Kontroversen über die deutschen Modalverben. Deutsch als Fremdsprache 32, 4: 206-214.

174

Tanja Mortelmans

Hentschel, Elke 1998 Negation und Interrogation. Studien zur Universalität ihrer Funktionen. (RGL 195.) Tübingen: Niemeyer Langacker, Ronald W. 1991 a Foundations of Cognitive Grammar. Volume II. Descriptive Application. Stanford: Stanford University Press. 1991 b Concept, Image, and Symbol: The Cognitive Basis of Grammar. Berlin/New York: Mouton de Gruyter. 1999 Grammar and Conceptualization. (Cognitive Linguistics Research 14.) Berlin/New York: Mouton de Gruyter. this vol. Subjectification, Grammaticization, and Conceptual Archetypes. Lehmann, Christian 1985 Grammaticalization: Synchronic Variation and Diachronic Change. Lingua e Stile 20: 303-318. Milan, Carlo 2001 Modalverben und Modalität. Eine kontrastive Untersuchung Deutsch-Italienisch. Tübingen: Niemeyer. Mortelmans, Tanja 1999 Die Modalverben 'sollen' und 'müssen' im heutigen Deutsch unter besonderer Berücksichtigung ihres Status als subjektivierter 'grounding predications'. Ph.D. dissertation, University of Antwerp. 2002 'Wieso sollte ich dich küssen, du hässlicher Mensch!' Α study of the German modals sollen and müssen as 'grounding predications' in interrogatives. In Grounding. The epistemic footing of deixis and reference, Frank Brisard (ed.), 391-432. Berlin/New York: Mouton de Gruyter. 2003 The 'subjective' effects of negation and past subjunctive on deontic modals: The case of German dürfen and sollen. In Deictic Conceptualisation of Space, Time and Person, Friedrich Lenz (ed.), 153-182. Amsterdam/Philadelphia: John Benjamins. Pelyväs, Peter this vol. Subjectification in (expressions of) epistemic modality and the development of the grounding predication. Sanders, Jose and Wilbert Spooren 1997 Perspective, subjectivity, and modality. In Discourse and Perspective in Cognitive Linguistics, Wolf-Andreas Liebert, Gisela Redeker and Linda Waugh (eds.), 85-112. Amsterdam/Philadelphia: Benjamins.

Langacker's 'subjectification' and 'grounding'

175

Sweetser, Eve 1990 From etymology to pragmatics: metaphorical and cultural aspects of semantic structure. Cambridge: Cambridge University Press. Talmy, Leonard 2000 Toward a Cognitive Semantics. Cambridge (Mass.): MIT. Traugott, Elisabeth C. 1989 On the Rise of Epistemic Meanings in English: An Example of Subjectification in Semantic Change. Language 65.1: 31-55. 1995 Subjectification in grammaticalisation. In Subjectivity and subjectivisation. Linguistic perspectives, Dieter Stein and Susan Wright (eds.), 31-54. Cambridge: Cambridge University Press. 1999 From subjectification to intersubjectification. Paper presented at the Workshop on Historical Pragmatics. Fourteenth International Conference on Historical Linguistics, Vancouver, Canada, July 1999 (http://www.Stanford.edu/~traugott/ect-papersonline.html). Traugott, Elisabeth C. and Richard B. Dasher 2002 Regularity in semantic change. Cambridge: Cambridge University Press. Vater, Heinz 1975 Werden als Modalverb. In Joseph Calbert and Heinz Vater (eds.), Aspekte der Modalität, 71-148. Tübingen: Narr 1997 Hat das Deutsche Futurtempora? In Heinz Vater (ed.), Zu Tempus und Modus im Deutschen, 81-104. Trier: WVT Wissenschaftlicher Verlag.

Conceptual and constructional considerations on the subjectivity of English and Spanish modals* Bert Cornillie

Introduction This paper deals with the degree of subjectification attested in dynamic, deontic and epistemic modal readings of the Spanish modals poder 'can', deber 'must' and tener que 'must/have to' and their respective counterparts in English. The analysis presented here is developed against the background of two different ways of understanding subjectification. Cognitive Grammar (CG) refers to subjectification as a shift towards "the full disappearance of any objective basis for the conceptualizer's mental scanning" (Langacker 2000: 299). In Grammaticalization studies, by contrast, subjectification is understood as "the historical pragmatic-semantic process whereby meanings become increasingly based in the speaker's subjective belief state, or attitude toward what is said" (Traugott 1989: 35).1 Both approaches see epistemic modals, i.e. those modals that express the speaker's judgment of possibility, as being derived from dynamic and deontic modals, which express capacity and obligation/permission, respectively. However, these two approaches interpret the semantic shift from dynamic/deontic modality to epistemic modality in a different manner.

*

A previous version of this paper was presented at the "Paths of Subjectivity" workshop held at the 8th International Cognitive Linguistics Conference (Logrono, July 2003). I would like to thank the audience for their interesting comments. I am greatly indebted with Frank Brisard, Nicole Delbecque, Dagmar Divjak, Tanja Mortelmans, Ricardo Maldonado, Jan Nuyts and Peter Pelyvas for the lengthy conversations on the topic of subjectification and grounding predications. Their advice and encouragement have proved very stimulating. I especially want to thank Costas Canakis, Nicole Delbecque, Tanja Mortelmans and Peter Pelyvas for their revision of previous versions of the paper. All remaining errors are mine, of course. 1. The main differences between the two interpretations are discussed in Brisard (this volume).

178

Bert Cornillie

From a diachronic point of view, Traugott and Dasher (2002: chap. 3) argue that a deontic modal such as must has both a subjective (performative) reading (la) and an objective (descriptive) reading (lb). According to them, the epistemic modal reading (lc) derives from the objective deontic reading (lb) through invited inference (see Brisard, this volume, for a discussion on inference). (1)

a. You must play this ten times over. (Coates 1983: 34) b. We must all die. (Traugott and Dasher 2002: 127) c. You must have been home last night. (Sweetser 1990: 61)

Traugott and Dasher (2002: 148) present the shift from pre-modal to epistemic modal readings in the following way: premodal

>

nonsubjective >

deontic

>

epistemic

subjective nonsubjective [objective]

>

subjective

Figure 1. From premodal to epistemic verbs. (Traugott and Dasher 2002: 148) Yet, the shift from subjective deontic readings over objective deontic ones to subjective epistemic readings is problematic. First, it appears that the intermediate stage of subjective and non-subjective deontic modality preceding subjective epistemic modality does not fit Traugott's definition of unidirectional subjectification (cf. Traugott's 1989, 2003). That is, splitting up the subjectification process into two independent processes (i.e. premodal » subjective, nonsubjective » subjective, illustrated in Figure 1.), contrasts with the emphasis Traugott puts on steadily increasing speakerorientedness. Second, one should prefer a straightforward account of the modal shift that is both diachronically and synchronically valid. Such a unified account of the process from pre-modal to epistemic modals, I argue, can be found in Langacker's Cognitive Grammar. This model does not differentiate between objective and subjective deontic (and epistemic) readings on the basis of performativity nor does it rely on invited inference to account for the polysemous nature of the modals. Instead, Langacker (2000) analyzes the subjectification process of modals in terms of the increasingly subjective grounding by the modal and the steady attenuation of

Considerations on English and Spanish modals

179

its force structure, which lead to a more subjective construal. Since such an account holds both synchronically and diachronically, Langacker's model is more appropriate to account for the different degree of subjectification of deontic and epistemic modals in Spanish 2 than Traugott and Dasher's (2002) account, although some elements of the latter can be integrated in the former. The paper is structured in the following way. I will first present Langacker's analysis of the English modal auxiliaries and discuss the relation between different types of subjectification and tense inflection (section 1). I will argue that, rather than the different tense marking, the deontic and epistemic modal types correspond to different subjectivity. In section 2, I show that the basic epistemic model and the dynamic evolutionary model represent the difference between epistemic and deontic modals. In section 3, the two types of modality correlate with a wide scope and a narrow scope reading, respectively. I account for the different modal readings on the basis of the shift from an objective construal of deontic force over a subjective construal of force to absence of force with epistemic modality. Section 4 deals with the availability of pro-forms with epistemic and deontic modals. In section 5, the reference point function and the subjectification of force will further underpin the specific conceptual content of epistemic modals.

1.

Subjectification, grounding and tense inflection

As far as the English modals are concerned, Langacker (1990, 1991, 2003) distinguishes between Subjectification Type I and Subjectification Type II, as illustrated in Figure 2.

2. This is not the first CG analysis of modals in Romance: Achard (1998: 123172) gives a detailed account of the French modal verbs, which are somewhat similar to the Spanish ones. Achard (1998: 161-172) pays special attention to the realignment of modal force and the impact of speaker's control, but does not discuss the different types of subjectification that Langacker (1990, 1991) proposes for the English modals. Moreover, Achard's (1998) analysis also includes savoir 'know/can' in the modal paradigm.

180

Bert Cornillie

. tr

lm

o^o •·,..

OS

scope

OfO X'" •1 II

x'i|

Θ ώ (a) objectively construed relation

'•·... OS ..·••·' scope

(b) Subjectification Typ« 1

ά

scope

(c) Subjectification Type 2

Figure 2. Types of Subjectification (Langacker 1991: 216)

In Langacker's theory, modal verbs belonging to Subjectification Type I have tense inflection, e.g. the English periphrastic modals such as have to, need to, which profile the grounding relationship between the modal verb and the infinitival process. Grounding specifies a designated entity with regard to "fundamental notions such as time, reality and referential identity" (Langacker 2003: 7). In Figure 2 (b), the ground and its relation with the process are both put in boldface since both are part of the objective scene (OS). By contrast, English modal auxiliaries such as must and may correspond to Subjectification Type II because they lack tense marking. Since only the grounded entity, i.e. the infinitival process, is profiled (the ground is not put in bold, being situated outside the OS), Langacker considers these expressions "grounding predications". Langacker claims that German deontic and epistemic modals necessarily belong to the first type of subjectification. According to him, these verbs do not function as grounding predications, since they "themselves bear inflections for tense and person that [he] would analyze as grounding predications" (Langacker 1990: 27). By extension, the same would then hold for the Spanish modals since they also inflect for tense and person. In other words, the difference between the first and the second type of subjectification consists in the different temporal grounding as expressed by tense marking or the lack thereof. However, one would expect that the conceptual differences between epistemic modality and deontic modality led to different types of subjectification. The question then arises whether tensed modals cannot be further differentiated in terms of different types of

Considerations on English and Spanish modals

181

subjectification. Moreover, there is also the question whether tense marking really blocks modals from reaching the most advanced type of subjectification. In Cornillie (2005), I tested the hypothesis that in modal constructions the conceptual import of temporal grounding is less crucial than that of modal grounding. Since most Western European modals are characterized by temporal grounding, emphasis should be put on the different modal readings and the degree of subjectification they involve. As a consequence, the central point of the paper is that Spanish epistemic modals have undergone more subjectification than deontic modals. In line with Pelyvas's (1996, 2001, this volume) study of English modals, I will argue that Spanish epistemic modals exhibit a stronger type of subjectivity because they are not hindered by event participants that become the focus of attention, as is the case with deontic modals. The epistemic modals will be considered dynamic reference points, which only shift attention to the infinitival process (the latter being the grounded entity).3 This view on the Spanish modal verbs will be underpinned by distributional evidence.4 Tense inflection is not as problematic for advanced subjectification as one may think (see also Pelyvas 1996: 171 for English mental state verbs). Interestingly, Langacker (2003) has already taken a first step in this direction by describing extreme subjectification in present tense finite forms. Indeed, Langacker (2003: 16) revises his former 'naive' characterization5

3. As far as terminology is concerned, the ungrounded verbs following the modal will simply be referred to by means of the traditional term "infinitive" or the CG term "infinitival process" (for a more detailed discussion on the grounded head and sequential or summary scanning see Pelyvas (1996: 172-173, this volume)). Other theoretical notions are extensively explained in other papers in this volume (in particular Brisard, Langacker, Mortelmans and Pelyvas). 4. I do not use the term "formal" for the analysis of modals, since in Cognitive Grammar it refers to the (absence of) formal grounding of the modal auxiliaries. This has been shown to be "too inclusive" in that no distinction is made between epistemic and deontic modals (Pelyvas 1996: 171). 5. From the 'naive' point of view, present tense stands for "a full instantiation of the profiled process [which] occurs and precisely coincides with the time of speaking" (Langacker 1987: 82), meaning that the state of affairs is 'out there in the world'. The speaker conceptualizes the proposed process while producing the utterance, but (s)he can also mentally construe more dimensions of the process than those actually verbalized in the speech event. The ability of the speaker to do so is reflected in the use of the present tense: it is well known that a lin-

182

Bert Cornillie

of the English present tense by arguing that, when the present tense does not correspond to the realization of the occurrence, "the temporal unfolding of the process is construed with extreme subjectivity". The viewing of the state of affairs expressed in a formally grounded linguistic expression may involve a 'Active' or 'virtual' dimension, since "not every instance of a thing or process type is an actual one. We very often deal with arbitrary instances, which are 'conjured up' for some special local purpose" (Langacker 2003: 18). In (2a), for example, the arrival is planned for next week, although the verb is in the present tense. This utterance is made possible by a subjective construal through virtual extension, which includes future reality. (2)

a. My mother arrives next week, (from Langacker 2003: 22) b. *My mother arrives right now. c. My mother is arriving right now.

Different verb types behave differently as to the instantiation or virtual extension they allow for. A perfective verb such as arrive needs the progressive tense for full instantiation at the moment of enunciation, as illustrated in Langacker's (2003: 22) examples (2b-c). By contrast, imperfective stative verbs, which do not stand for a bounded process, have a valid instance in the present tense, as shown for to have in (3a), whereas they are infelicitous in the progressive tense (3b). As a matter of fact, imperfective verbs do not necessarily have a specific instantiation. (3)

a. My mother has her own opinion. b. *My mother is having her own opinion.

The foregoing discussion on the tense and aspect of perfective action verbs and imperfective stative verbs can also be applied to the tensed Spanish modal verbs. Interestingly, just as the imperfective verbs, epistemic deber 'must', in (4a), and epistemic poder 'can/may', in (4b), cannot appear in the progressive tense.

guistic expression with present tense inflection sometimes includes future reference. Since present tense often does not lead to instantaneous accomplishment of the process expressed, a mere conceptualization of the process takes place.

Considerations on English and Spanish modals (4)

183

a. *Estä debiendo estar de vacaciones. be-3sg.PRES.IND must-GER be-INF of holidays * 'He is musting be on holidays' b. *Estd pudiendo haber hecho un error. be-3sg.PRES.IND can-GER have-INF make-PART a mistake * 'He is being able to have made a mistake'

By contrast, in line with the perfective verbs, dynamic poder 'can' expressing ability in (5a), deontic deber 'must' (5b) and tener que 'have to' (5c) are not prevented from appearing in the progressive tense. (5)

a. Los dos partidos estdn pudiendo realizar the two party-PL be-3pl.PRES.IND can-GER reach-INF un pacto. a pact. 'The two parties are being able to reach a pact' b. El plazo de presentacion de ofertas finalizarä a the term of presentation of bid-PL fmish-3sg.FUT at las 14 horas del dia 10/01/04, debiendo ser the 14 hours of the day 10/01/04, have-GER to be-INF entregadas en el Departamento de Compras. submit-PART in the Department of Purchase-PL 'The call for bids will be closed on January 10th at 2 p.m.; [lit.] having to be submitted to the Department of Purchases' c. Estä teniendo que terminar la tarea be-3sg.PRES.IND have to-GER fmish-INF the task räpidamente. quickly-ADV

'He is having to finish the task quickly' Thus, when they have dynamic and deontic readings, the Spanish modals can appear in the progressive tense, whereas their epistemic counterparts do not allow such a tense. The differences between epistemic and dynamic/deontic modals observed in (4) and (5) can be accounted for by the fact that the latter ones stand for a process that involves a certain force structure, whereas the former do not. Dynamic modals such as poder in (5a) display a potency relation between the subject and the infinitival process; and deontic modals such as deber (5b) and tener que (5c) convey interplay of forces, usually from the speaker to the addressee. In other words,

184

Bert Cornillie

one can argue that dynamic and deontic modals allow a progressive tense construction because these verbs stand for a process that can be instantiated, hence, be profiled (i.e. made prominent). Epistemic modals differ from their deontic counterparts in that they lack a prominent force structure. The fact that a progressive tense construction does not fit epistemic modal verbs can be accounted for by the observation that their force structure, unlike that of deontic modals, is not prominent enough to receive further instantiation. In other words, epistemic modals cannot be processes themselves, but may only give a specific view on the process: they express a version of reality which is subjectively construed by the speaker. Thus, despite the tense marking, the distribution of deontic and epistemic modals highlights a different conceptual content. Two conclusions can be drawn from the foregoing discussion: (i) tense inflection is not to be seen as a major problem that precludes Spanish modals from attaining different degrees of subjectification, (ii) epistemic, dynamic and deontic modal verbs should not be lumped together on the basis of the presence or absence of tense inflection and can be differentiated on the basis of distributional criteria.

2.

Basic epistemic model and dynamic evolutionary model

The subjectivity of the present tense constructions and the lack of force with the epistemic modals can be further specified by means of Langacker's basic epistemic model and his dynamic evolutionary model. In the epistemic model presented in Figure 3., reality represents the level at which events are directly "observed, dreamed, imagined, ..." (Achard 2002: 208). The model is based on three theoretical notions: known reality, immediate reality and irreality. These three ways of looking at the world characterize the speaker's basic conception of reality. Known reality contains all situations that a given speech participant accepts as being real, including immediately available reality. The notion of immediate reality acknowledges the fact "that reality is generally conceived as dynamic, as constantly evolving and growing" (Mortelmans 2003: 154). Immediate reality constitutes the last stage of evolution of reality before irreality, which is then the realm beyond known and immediate reality. The speaker can think of events that are likely to happen in the future and can categorize them in terms of degrees of likelihood. Thus, our

Considerations on English and Spanish modals

185

Irreality (Known) Reality j lmmediate\ Reality

\

@

/

Figure 3. Basic epistemic model (Langacker 1991: 242) knowledge of reality is not restricted to the directly grasped state of affairs, but also includes "the capacity to evaluate the position of those events with respect to reality" (Achard 2002: 208). Epistemic modals clearly convey this qualification, while deontic modals comment on a concrete state of affairs by expressing a moral judgment about the event evoked (or some element or some participant of it). The immediate reality expressed by epistemic modals can be further refined by means of Langacker's dynamic evolutionary model (see Figure 4). This model presents elaborated reality in terms of potential and projected reality, i.e. the speaker's knowledge of the possible (future) evolution of reality. Potential reality includes all events that may take or may have taken place. Projected reality, on the other hand, stands for developments in the future or judgments on the past which the speaker is very committed to.

Reality

/

/

Present \ Reality \

Projected Reality

/

Figure 4. Dynamic evolutionary model (Langacker 1991: 277)

186

Bert Cornillie

Langacker (1991: 278-279) seems to account for both deontic and epistemic modals on the basis of the dynamic evolutionary model. However, as shown by Pelyvas (1996: 192), this model lends itself better for describing epistemic readings than deontic ones. The two above-presented models further support the view that the conceptualization of the epistemic lack of force differs from that of deontic force. In other words, as argued before, Langacker's (1990, 1991) equal treatment of epistemic and deontic modal should be replaced by a clear division of labor between deontic and epistemic modals. In the next section, I will discuss in which respects epistemic modality, involving wide scope over the process as a whole, differs from dynamic and deontic modality that corresponds to narrow scope over parts of the process. I will rely on Pelyväs's (1996, 2001, this volume) proposal, which will provide proof for the claims that different scope has consequences for syntactic representation (Section 4), the result of which testifies to different degrees of subjectification (Section 5).

3.

'Narrow' versus 'wide' scope

As illustrated in Figure 2., Langacker's Subjectification Type I and Type II have the same scope, and only differ as to the extension of the objective scene (OS). Pelyvas (2001b, this volume), by contrast, argues that deontic and epistemic modals display a different scope. Pelyväs's account is a Cognitive Grammar adaptation of Nordlinger and Traugott's (1997: 309) use of the terms 'narrow' and 'wide' scope. On the one hand, Pelyvas uses the term "scope" as it is commonly understood in Cognitive Grammar, i.e. as "the 'stage' or general locus of attention in a predication" (Brisard 2002: 269). On the other hand, Pelyväs's use draws on Nordlinger and Traugott (1997: 309) and Traugott and Dasher (2002) in which 'wide scope' and 'narrow scope' are used to refer to different types of both epistemic and deontic modality. Wide scope corresponds to descriptive, objective deontic and epistemic modality, whereas narrow scope very much relies on the speaker's performative deontic force or epistemic view. I will apply Pelyväs's adaptation of scope to the Spanish modals. From the construal-based point of view promoted by Pelyväs, the interplay of forces and counterforces involved in the deontic modals belongs to the narrow scope, while the epistemic elaboration of reality has a wide scope. The reprinted image schema exemplified in Figure 5 illustrates that the 'narrow scope' of deontic poder focuses on the objective scene (OS)

Considerations

on English and Spanish modals

187

and its participant(s), who is/are supposed to be able to accomplish the infinitival process. The most common deontic relation is the one between the so-called emitter of force and the participant, on the one hand, and the process in which the latter normally participates, on the other. Now, deontic poder does not display a very complex force structure, since the emitter relinquishes the possible counterforce. In other words, the permission boils down to giving green light to a force after having omitted possible counterforces.

,

OS

scope

relinquishes mitliorilv

\ o

*

o

> o

\

] /

Γ f

s (ι

(conceptualizer)

Figure 5. Deontic maylpoder (from Pelyväs this volume)

Although the 'narrow scope' is broader than the immediate scope (OS) and the forces involved, the speaker and the ground (S/G) are not included.6 Nevertheless, with a 'narrow scope' construal, the speaker's interference in the interplay of forces is more prominent than the epistemic modal profile. Epistemic poder, for its part, renders a 'wide scope' reading, as represented in Figure 6. With epistemic modals the 'narrow scope' is broadened to an overall 'wide' scope which includes the speaker and the ground (S/G). A 'wide scope' reading stands for a global view on the state of affairs in which the speaker does not play a role. Since the speaker is not part of the interplay of forces, the infinitival process can be construed in a maximally subjective way.

6. I want to thank Peter Pelyväs for explaining this point to me.

188

Bert Comillie

/

OS

Ο ->o ajw situation /\

/ \

/

/ \

\ S/Grf scope

speaker relinquishes epistemic commitment (grounding)

Figure 6. Epistemic may/poder (from Pelyväs this volume) The case of deber and tener que is similar, although these verbs present a different interplay of forces in the deontic reading. Here the counterforce is associated with the participant, whereas the force is emitted by the speaker or a third participant (see Pelyväs, this volume). This means that in their respective epistemic readings, deber and tener que stand for a stronger commitment to the likelihood that the proposition is true than epistemic poder. There is, hence, a correlation between the force structure and the kind of epistemic reading. Since the viewing of the deontic force structure contrasts with epistemic viewing, further discussion about the shift from deontic to epistemic modality is needed, however. It can be argued that it is the decreasing prominence of the force that broadens the scope of predication in epistemic readings and leads to an "evolutionary momentum" of immediate reality (Langacker 1991: 276). In other words, the shift from deontic to epistemic modality corresponds to an attenuation of force. It can also be expected that the different degrees of attenuation of the interplay of forces correlate with different distributional properties. In the following section, I will address the availability of pro-forms and the (in)felicitous insertion of supplementary aspectual marking.

4.

Pro-forms and interplay of forces

Before discussing the Spanish pro-forms, it is useful to review Langacker's (1990, 2003) comments on the availability of modal pro-forms in English.

Considerations on English and Spanish modals

189

Langacker (1990: 14; 2003: 7) states that "the modals can serve as proforms for finite clauses (as is usual in English, a subject is required as well): They should, I may. This suggests quite strongly that grounding elements have the same referent (i.e. profile) as the structures they ground". According to Langacker (2003: 8), the fact that grounding predications with the highest subjectivity can function as a pro-form is evidence for the claim that they "profile the grounded entity, and not the grounding relationship". For the English modals, this means that the infinitival process and not the tense inflection is the focus of attention. Yet, Langacker does not say whether these pro-forms license dynamic, deontic or epistemic readings, nor does he comment on pro-forms in languages other than English. Langacker's pro-form argument leads me to add two comments. First, it can be expected that the availability of pro-forms differs according to the dynamic, deontic or epistemic sense the modal verbs express in a particular context. My hypothesis is that the relative conceptual independence of modal pro-forms emanates from the interacting forces between participants. A second question arises: does Langacker's account of pro-forms suggest that cross-linguistically the availability of modal pro-forms always characterizes the high subjectivity of modals and systematically implies a special profile for the grounded infinitival process? In other words, do proforms by default give a special profile to the infinitival process, which is then the grounded head? Let us look at a series of Spanish examples. The examples in (6) show that an epistemic modal verb with an overt grammatical subject does not readily show up without an infinitive when it yields an unambiguously epistemic reading. (6)

a. Juan puede

haber

John can-3sg.PRES-IND

terminado

la tarea.

have-lNF finish-PART the task.

'John may have finished his task' Si, puede ('0/ ser}. Yes.

can-3sg.PRES-IND

be-INF

'Yes, he may/it can be' b. Puede queyahaya can-3sg.PRES-IND

that have-3sg.PRES-IND

'It may be that he has already finished' Puede (ser). can-3sg.PRES-IND

'It can be'

terminado. finish-PART

190

Bert Cornillie c. Esto puede crear this can-3sg.PRES-IND create-INF 'This can/may cause problems' Puede.

problemas. problems

can-3sg.PRES-IND 'It c a n / m a y '

In (6a), the utterance cannot easily be converted into a construction whithout infinitive. The lack of an infinitive seems to lead to a dynamic reading of capacity. When a copula such as ser 'to be' is added, an epistemic reading is most probable. Note that in this case, it is not the grammatical subject, Juan, but the whole utterance that agrees with the verb poder. In the same vein, epistemic puede que in (6b) does not have another grammatical subject than the gwe-clause of the construction. Although poder could appear alone here, the ser infinitive is preferably added. The addition of ser shifts the profile from the possible modal force to the infinitival process. Yet, when there is no risk of subject agreement with a forcebearing entity, epistemic poder can show up as a pro-form. There are also other cases in which pro-forms are felicitous. In (6c), for example, epistemic poder seems to have a pro-form. The possible omission of the infinitive may be attributed to dynamic overtones, however, given that in this utterance both an epistemic and a dynamic reading is possible. In the latter case, the subject can also have a potency relation with the infinitive. Native speaker judgments differ as to whether epistemic pro-forms are allowed. Gomez Torrego (1999: 3360) mentions that all kinds of modal constructions allow for a pro-form. Yet, among the examples he mentions there is only one epistemic modal reading (7b), while the deontic reading of permission is represented twice (7a-c). (7)

a. lLa The 'Can - No,

policia puede torturar? police can-3sg.PRES-IND torture-INF the police torture?' no puede.

- No, no

can-3sg.PRES-IND

- 'No, they cannot' b. ^ Puede estar tu padre en casa ahora? can-3sg.PRES-IND be-INF your father in house now 'Can your father be at home now?'

Considerations on English and Spanish modals

- Si,

191

puede.

- Y e s , can-3sg.PRES-lND

- 'Yes, he may/it may' c. iSe puede REFL can-3sg.PRES-IND

entrar? enter-INF

'May I enter' - Si, se puede. - Yes, REFL can-3sg.PRES-IND

- 'Yes, you may' Moreover, poder in (7b) is not a prototypically epistemic use, since epistemic modifiers do not readily show up in interrogative sentences (cf. Nuyts 2001). The interpretation of the question is the following: "is your father usually at home around this time?" The question bears more prominence to the father's time management than the answer suggests, which in turn evokes the possibility that he is at home. Rather than epistemic modality, example (7b) expresses situational dynamic modality. By contrast, in examples (7a) and (7c), the pro-form of poder 'can' is more common, which suggests a correlation with the fact that the modal in the main clause displays interacting forces. The letting force of permission is prominent and makes the finite modal verb have some conceptual independence. So far the hypothesis that the conceptual independence of the pro-form would differentiate between epistemic and other readings does not appear to apply to all examples of poder. This may be due to the fact that the verb poder, unlike other modal verbs, bears a rather weak force structure without counterforces in dynamic and deontic readings. In other words, the difference between dynamic/deontic and epistemic modality is small (capacity vs. potential reality). In what follows, I will show that the difference between these readings is much larger for deber. Deber can yield both an epistemic-evidential and a deontic reading. That is, this verb can convey projected reality and obligation. In (8a), the epistemic-evidential reading does not allow for a construction without an infinitive. In (8b), the epistemic reading in the first utterance receives a strong deontic dimension in the pro-form construction, which seems to render the utterance felicitous. (8)

a. Juan debe haber terminado la tarea. John must-3sg.PRES-IND have-lNF finish-PART the task. 'John must have finished the task'

192

Bert Cornillie Si,

debe

{*0/ser}.

Yes,

must-3 sg.PRES-IND

'Yes, he must/it must be' b. Deben haber

llegado

ya.

must-3pl.PRES-lND have-INF arrive-PART already

'They must have arrived by now' Deberian. must-3pl.COND

'They should' The difference between the infelicitous pro-form of epistemic-evidential deber in (8a) and the felicitous pro-form of the conditional form with deber is in line with the reduced force with epistemic deber, on the one hand, and the prominence given to the interplay of forces conveyed by deontic deber, on the other. In other words, the pro-form of deber in (8a) gives too much prominence to the force to be able to function as an epistemic marker. In addition to deber, deontic modality can also be expressed by means of tener que 'have to' in Spanish, as shown in (9). (9)

a. [Tiene / tenia / tendrä / tendria] que have

to-3sg-IMPERFECT/FUT/COND

ir

α Estados Unidos.

gO-INF to U S

'She {has/had/will} have to/should go the United States' b. Si, ? tienel *ίεηίαΓ tendrä! tendria que. Yes, have to-3sg- IMPERFECT/FUT/COND 'Yes, she has to/had to/will have to/should' Of all tenses used in (9), the present and the conditional most easily allow for a pro-form construction of tener que, whereas the imperfect and the future tense do not fit very well. Tener que can yield a speaker-oriented deontic reading or a participant-oriented one; that is, the deontic force is attributed to the speaker or to the participant (or another kind of entity). Interestingly, with tener que the speaker often bases his/her deontic deontic claim on moral laws or volition by other people. In other words, the speaker's opinion is then based on forces from "out there, in the world" (Langacker 2003: 10). Hence, it is interesting to review to what extent this verb qualifies well for the so-called objective deontic reading (cf. Traugott and Dasher 2002).

Considerations on English and Spanish modals

193

Besides a default deontic reading, tener que 'have to', can also yield an epistemic-evidential reading, as in (10).7 However, the presence of "objective" deontic modality does not imply that the verb easily leads to epistemic readings. It is a case in point that of all three verbs tener que yields an epistemic reading least often (see Cornillie 2004b: chap 2). (10)

a. Este cientifico tiene que This scientist must/have to-3sg.PRES-IND realmente honesto. really-ADV honest. 'This scientist must be really honest' b. * Si, tiene que. yes,

ser be-INF

must-3sg.PRES-lND

'Yes, he must' As with deber in (8), example (10) shows that a pro-form is infelicitous with epistemic-evidential tener que. The fact that deontic forms of deber and tener que show up alone more easily than their epistemic counterparts corroborates the argument for distinguishing between epistemic modal readings, on the one hand, and deontic ones, on the other. For deontic modals with a strong force structure such as deber and tener que, the availability of an independent pro-form can be seen as an indication of how the speaker profiles the deontic relationship with the participants, rather than how (s)he profiles the infinitival process as a whole. Hence, for deber and tener que the pro-form hypothesis can be confirmed. The fact that a subjective construal expressing epistemic modality lacks participants competing for prominence contrasts with the prominent force structure of the proform. This is in line with the fact that the epistemic modality consists in the speaker's general attitude toward the state of affairs, without any speaker intervention in this state of affairs. The question still remains why epistemic modal auxiliaries in English have pro-forms, whereas their Spanish counterparts do not regularly have them. I believe this is due to the morphosyntactic characteristics of the English modals: the untensed auxiliaries in English have a tendency to behave increasingly as modal particles, while Spanish modals are morphologically more like main verbs. Furthermore, it should be noted that not

7. In Olbertz' Functional Grammar terminology, tener que 'have to' stands a.o. for Extrinsic Inherent Modality (Olbertz 1998: 385).

194

Bert Cornillie

all English pro-form constructions are equally felicitous. For example, the pro-form must shows up almost exclusively in deontic contexts, while epistemic-evidential must usually does not appear alone. Since the English pro-forms exceed the scope of this analysis, other studies should shed new light on the different modal pro-forms and their frequency. Importantly, the differences between deontic deberltener que and epistemic-evidential deberltener que are not dichotomous. The two readings are endpoints on a continuum of semantic change. Recall that Traugott and Dasher (2002) account for the epistemic modality of must in terms of a shift away from objective deontic modality. Let us now look into the question of whether deber also expresses this so-called objective deontic modality.8 In example (11), deber conveys an abstract modal force, which is directed to the subject los ciudadanos 'the citizens'. Instead of expressing his/her subjective view, the speaker refers to the moral laws that are expected to be complied with in a certain context. According to Traugott and Dasher (2002), this reading is more objective than clearly performative deontic ones, i.e. those with instant speaker commitment. However, although the emitter of force may be less strongly involved than in other constructions, (s)he can certainly be a supporter of the moral principle s/he is appealing to. It is hence difficult to consider utterance (11) strictly in terms of subjective and objective modality. (11)

a. Los ciudadanos the citizen-PL

deben

las

leyes.

must-3pl.PRES-IND respect-INF the

respetar

laws.

'Citizens must obey law' b. Deben. Los ciudadanos ??

must-3pl.PRES-IND

The citizen-PL

'They must' Deberian. must-3pl.COND 'They should'

'The citizens must'

deben. must-3pl.PRES-IND

8. Kronning (1996) argues for French devoir that such a construction conveys alethic modality, which is, according to him, situated in between deontic and epistemic modality. Note, however, that this notion is not commonly accepted to have a place in linguistics, but is often referred to in logic (see Nuyts 2001: 28). Nevertheless, alethic modality could fit in the discussion on the force structure of modal constructions.

Considerations on English and Spanish modals

195

As an alternative, the reading of deber in (11) can easily be accounted for on the basis of weak counterforce. The ciudadanos 'citizens' are not characterized by a strong counterforce that resists the deontic force emitted by the speaker. Interestingly, deber also expresses another kind of modality which does not involve participants and which lacks counterforce altogether. In (12), the force is limited to a logical principle. In such an utterance, the speaker and the hearer rely on encyclopedic knowledge about rectangles. That is, the speaker has recourse to general principles of the world we live in. This reading may be considered to be more in line with Traugott and Dasher's objective deontic modality. (12)

a. Un rectängulo debe tener 4 ängulos. A rectangle must-3sg-lND have-INF four angles Ά rectangle must have four angles' b. * Debe. *Un rectängulo debe. Deberia. must-3sg-IND A rectangle must-3sg-IND must-COND 'It must' Ά rectangle must' 'It should'

Now, the question is whether the construction in (12) is really more objective than the one in (11), as proposed by Traugott and Dasher (2002). At first sight, there is no evidence to state that the construal in (12) is more objective. The only difference is that there is no force by an emitter or counterforce by another entity involved. In other words, it is possible to discover a continuum from constructions involving force applied from the speaker to participants to constructions without such a force. In Langacker's (1999: 308) terms, it could be said that the force displayed in the so-called objective deontic modality becomes "transparent" since the locus of potency only refers to "some nebulous, generalized authority", i.e. without participant involvement. This process is referred to as "attenuation" of the deontic force structure, which ends up in epistemic modality that lacks any force whatsoever. What is left in profile in the latter type of modality is the 'evolutionary momentum' of reality, as stipulated in Langacker's (1991: 275-281) dynamic evolutionary model. To some extent, the examples in (11) and (12) also corroborate the above suggested link between decreasing prominence of force with deber and the unavailability of a bare pro-form. Since the deontic modality that Traugott and Dasher (2002) call objective only has a weak force structure (with a less prominent counterforce), it can be seen as a bridge between deontic modality with a prominent force structure and participant involvement and epistemic modality without such force structure or participant

196

Bert Cornillie

involvement at all. Moreover, the lack of pro-form constructions for Spanish epistemic deber and tener que indicates that the epistemic modals themselves are not salient enough to appear alone. The case of poder is different, however, since the dynamic, deontic and epistemic readings of this verb do not differ as much as the deontic and epistemic readings of deber and tener que do. The attenuation of the force structure is in line with the analysis of scope presented in Section (3). The 'widening' of the scope corroborates the gradually less prominent force structure from deontic modal readings to epistemic modal readings. This shift from immediate scope, over 'narrow scope' as in (9) and (11), to 'wide scope' as in (8) and in (10) involves subjectification. These above-mentioned differences point to different degrees of subjectification. In the next section, I will show that the analysis of Spanish modals presented so far underpins the idea that modals function as reference points and score high on the subjectification scale (cf. Langacker 2000 and Pelyvas 1996, this volume).

5.

Reference point and subjectification

In this section I link the low availability of epistemic pro-forms to the dynamic nature of the reference point (cf. Langacker 1993: 24). According to Langacker, a reference point construction implicitly establishes the link between the discourse participants, i.e. the hearer and the speaker, and a specific grounded instance of a type. Determiners and modals are typical reference points in language. As illustrated in Figure 7, a reference point implies a shift of prominence from the reference point (R) to the target (T), i.e. the noun or the process. The dynamic nature of the reference point is crucial for a better understanding of the conflicting relation between deontic force and epistemic readings of deber and tener que. Even if the determiner or the modal has initial prominence in the construal (see the bold R in Figure 7), it ultimately profiles the target (see the bold T), although the construction may "occasionally retain some prominence of the reference point" (Pelyvas 2001a: 5). In Spanish, the deontic modal pro-form gives a full profile to the potency relation and can thus remain fully prominent. By contrast, a Spanish epistemic modal verb only functions to profile the infinitival process. Due to its reference point function, the epistemic modal immediately gives

Considerations on English and Spanish modals

197

more prominence to the infinitival process, being the main focus of attention. That a reference point is never left in profile correlates with the scarce epistemic pro-forms. This certainly holds for deber and tener que, but certainly should be modified for poder. Since poder has a weak force structure, there is no strong competition between its different readings. This may explain why epistemic poder sometimes can retain some profile without activating a deontic reading.

Figure 7. Reference point construction (Langacker 1993: 6)

However, just as with deber and tener que strong prominence of epistemic poder is not possible, while this is no problem with its dynamic and deontic counterparts. This is illustrated in (13), where the inchoative verb empezar 'begin' can readily precede the infinitival form of poder 'can/may', which has an ability or permission reading. (13)

Buenos Aires estä atestada de autos y si la Buenos Aires be-3SG chock-full of cars and if the genie empieza a poder usar la bicicleta en people

begin-3SG

to can-lNF

use-INF

the

bike

in

los trayectos cortos y medianos, es posible the trajects short and middle, be-3SG possible que reduzcamos el caos en el träfico. that reduce-1 PL the chaos in the traffic (.LaNacion, Argentina, 19-05-1997) 'Buenos Aires is full of cars and if the people begin to be able to take bikes for short and medium distances, it is possible to reduce traffic chaos.'

198

Bert Cornillie

In this construction, both the deontic (permission) and the dynamic (ability) reading display clear involvement of participants. The permission in (13) is based on a letting schema with the beneficiary la gente 'the people', whereas the focus of the ability reading is centered on the capacities of the subject la gente to use the bike. Importantly, deber and tener que in (14) cannot express epistemic modality when these verbs are preceded by an aspectual verb, either. (14)

a. ...la gente va a deber the people go-3SG to must-INF la bicicleta. [*epistemic] the bicycle '*people will must use the bike' b. ...la gente va a tener que the people go-3SG to must-INF la bicicleta. [*epistemic] the bicycle '*people will must use the bike'

usar use-lNF

usar use-INF

The modals poder, deber and tener que, in (13) and (14), cannot render an epistemic reading, which means that an epistemic modal cannot be profiled by another verb, i.e. cannot be grounded further. This finding is in line with the assumption that an epistemic modal verb stands for the subjective grounding of aspect, as a part of the dynamic elaboration in terms of potential or projected reality. Silva-Corvalan (1995: 69) mentions that deontic deber can be preceded by ir a, as shown in (15a). Although I have not found infinitival forms of deontic deber in the corpus nor on the internet, Silva-Corvalän's native speaker judgment indicates that a deontic reading of an aspectually marked infinitive is not considered problematic in principle, which corroborates my claim of the felicitous insertion of an aspectual verb preceding the deontic modal. 9

9. Another construction I found on the internet shows that deontic deber can be preceded by evidential parecer. (i) No puede ser ninguno de los elementos, de modo que parece deber concluirse que tiene que ser algo asi como la masa indiferenciada de los elementos previa a toda "formacion" (Jose Ferrater Mora 1979: 2135— 2143)

Considerations on English and Spanish modals (15)

199

a. Juan va a deber cruzar el rio. John go-3SG to must-INF cross-INF the river 'John's going to have to cross the river' [taken from Silva-Corvalän 1995: 69] 11 b. la gente va a deber usar la bicicleta. [deontic] the people go-3SG to must-INF use-INF the bicycle ....'people will have to use the bike' c. la gente va a tener que usar la bicicleta. [deontic] the people go-3SG to must-INF use-INF the bicycle ....'people will have to use the bike'

The restrictions on deber in (16b) do not mean that all speaker-oriented deontic readings necessarily disappear with an aspectual construction. In corpus example (16) tener que is preceded by the inchoative vamos a 'we are going to'. Here, tener que can yield both a speaker-oriented and a participant-oriented deontic reading. (16)

....entonces me lastime las dos rodillas.. /Que then me hurt-lSG the two knees.. What barbaridad! iY ahora ya no haces ningün atrocity! And now already not do-2SG none deporte? Pues no pero lo vamos a tener que hacer sport. Well no but it go-lPL to have that do ahora tambien por los chamacos para que now also for the guys for that sigan el mismo paso que nosotros porque es follow-3PL the same step than we because be-3SG una cosa muy.... Mi suegro por ejemplo a thing very... My father-in-law for example ahorita corre tres kilometros. now run-3SG three kilometers (Habla Culta: Mexico: M5) 'Then I hurt both my knees. What bad luck! And now you don't do sports? Well, no, but now we will have to do it, too, because of the guys so that they follow our steps because it is a very ... thing. My father-in-law for example runs three kilometers' 'It cannot be any of the elements, subsequently he seems to have to conclude that it has to do with something as the indifferentiated mass of elements previous to whatever "formation"'

200

Bert Cornillie

With its flexible distribution tener que fills the gap in the deontic field. The relative unavailability of deber in an aspectual construction seems to broaden the use of tener que. In combination with an aspectual verb, the speaker-oriented deontic readings of deber seem to be handed over to tener que. This is a way of making the deontic force expressed by tener que more speaker-oriented in the sense that the aspectual dimension becomes an expression of deontic compulsion or insistence. Moreover, another plausible explanation is that deontic tener que is a more recent development and is thus less grammaticalized than deontic deber. The fact that epistemic modals cannot receive any supplementary profile themselves, together with the advanced attenuation of the force structure witnessed by epistemic modals, paves the way to higher subjectification with epistemic poder, tener que and deber than with their deontic modal counterparts. In order to explain the gradual shift from a prominent force structure to absence of force, I will first describe the weak subjectification witnessed in deontic modals. When a normal, transitive process is construed objectively, "it is held within the objective scene [OS]" (Pelyväs 2001a). The deontic force leads to subjective realignment from the process to the participants, but the latter still belong to the objective scene (OS). The speaker remains outside the 'narrow' scope of the predication since the attention is directed to the participants in the interplay of forces (see Pelyväs, this volume). In other words, the speaker relies on an objective construal of deontic force based on "narrow scope". The transition from 'narrow' scope to 'wide' scope readings is an example of subjectification. In epistemic readings, the lack of force broadens the scope of predication, which now includes the speaker. The fact that the ground is part of the 'wide' scope leads to a subjective construal of the process. There is a shift of the profile from the participant with a prominent position in the interplay of forces, to the speaker and, last but not least, to the process itself. Hence, with epistemic modals there is subjectification from the objective construal of the interplay of forces to the subjective construal in which no other force is involved. Since no participant other than the speaker actively interferes in epistemic modals, the infinitival process introduced by the epistemic modal is more salient than the deontic modal interplay of forces: the only element in profile is the (infinitival) process itself. In such a construction, the epistemic modal verb profiles the infinitival process through subjectification. As for the difference between deontic and epistemic modality, a "breakdown mechanism"

Considerations on English and Spanish modals

201

(Pelyvas 1996: 167-168) or "attenuation" (Langacker 2000) reduces the profile of the agentive force structure step by step. It has been shown that the weaker force structure of poder more easily allows for a shift to a subjective construal than deber and tener que with a strong interplay of forces.

6.

Conclusions

The fact that Spanish modals, unlike English ones, are a heterogeneous group of verbs with tense inflection, infinitive and past participle has not prevented me from showing that the Spanish epistemic modals stand for higher subjectification than their deontic counterparts. I have illustrated that the epistemic readings of the Spanish modal verbs poder, deber and tener que distinguish themselves from the deontic ones in terms of the lack of force structure, the rare availability of pro-forms and the infelicity of constructions with aspectual verbs. In the first section, I have argued that tense inflection does not prevent Spanish epistemic modals from expressing high subjectification. Langacker's (2003) analysis of present tense English verbs in terms of extreme subjectification paves the way towards granting German or Spanish modals the status of subjectification type II. In a second section, I have shown that, like English modal auxiliaries, epistemic modals fit in the basic epistemic model and the dynamic evolutionary model, whereas deontic ones do not. In the third section, it is shown that Spanish epistemic modals function as a modal profile with 'wide scope'. In the fourth section, I have discussed Langacker's argument for pro-forms as indicative of grounding predications and have applied it to Spanish. Pro-forms, although available in English, do not readily represent the infinitive introduced by the Spanish epistemic modals. By contrast, pro-forms seem to be available when there is a prominent interplay of (deontic) forces. In section 5,1 have shown that the epistemic modal itself is just a reference point that profiles the more salient infinitival process, as a grounded entity. The impossibility of adding an aspectual auxiliary verb corroborates this assumption. The reference point function is related to the wide scope and is presented as a typical result of subjectification in line with the "breakdown mechanism" of deontic force first proposed by Pelyvas (1996: 167-168) and later formulated as attenuation. Epistemic modals stand for maximal subjectification in that they completely attenuate the force directed to the participants involved in deontic modal readings.

202

Bert Cornillie

The analysis, broadly based on Langacker's model, qualifies as a systematic account of varying degrees of subjectivity. I have preferred this model as an alternative to the vaguer qualifications of objective and subjective deontic and epistemic modality used by Traugott and Dasher (2002) because it yields a straightforward account in terms of the attenuation of the objective construal of force to a subjective construal without force and, hence, is a valid alternative to the more problematic shift from subjective deontic over objective deontic to subjective epistemic modality.

References Achard, Michel 1998 Representation of Cognitive Structures. Syntax and Semantics of French Sentential Complements. Berlin/New York: Mouton de Gruyter. 2002 The meaning and distribution of French mood inflections. In Frank Brisard (ed.), Grounding. The Epistemic Footing of Deixis and Reference, 197-250. Berlin/New York: Mouton de Gruyter. Brisard, Frank 2002 The English present tense. In Frank Brisard (ed.), Grounding. The Epistemic Footing of Deixis and Reference, 251-298. Berlin/ New York: Mouton de Gruyter. Coates, Jennifer 1983 The Semantics of the Modal Auxiliaries. London: Croom Helm. Cornillie, Bert 2004 a Lhe shift from lexical to subjective readings in Spanish prometer 'to promise' and amenazar 'to threaten'. A corpus-based account. Pragmatics 14.1: 1-30. 2004 b Evidentiality and epistemic modality in Spanish (semi-) auxiliaries. A functional-pragmatic and cognitive-linguistic account. Doctoral dissertation K.U. Leuven. 2005 Reference point and subjectification in grounding predications: the case of the Spanish modals. Annual Review of Cognitive Linguistics 3: 56-77. Gomez Lorrego, Leonardo 1999 Los verbos auxiliares. Las perifrasis verbales de infmitivo. In Ignacio Bosque and Demonte Violeta (eds.), Gramätica Descriptiva de la Lengua Espanola Vol 2, 3323-3389. Madrid: Espasa.

Considerations on English and Spanish modals

203

Heyvaert, Liesbet 2003 A Cognitive-Functional Approach to Nominalization in English. Berlin/New York: Mouton de Gruyter. Kronning, Hans 1996 Modalite, Cognition et Polysemie: Semantique du Verbe Modal devoir. Uppsala: Acta Universitatis Upsaliensis. Langacker, Ronald W. 1987 Nouns and verbs. Language 63: 53-94. 1990 Subjectification. Cognitive Linguistics 1.1: 5-38. 1991 Foundations of Cognitive Grammar. Volume II. Descriptive Application. Stanford, California: Stanford University Press. 1993 Reference point constructions. Cognitive Linguistics 4.1: 1-38. 1995 Raising and Transparency. Language 71: 1-62. 1999 Grammar and Conceptualization. Berlin/New York: Mouton de Gruyter. 2000 Subjectification and Grammaticization. In Ronald W. Langacker (ed.), Grammar and Conceptualization, 297-315. Berlin/New York: Mouton de Gruyter. 2001 Estructura de la clausula en la gramatica cognoscitiva. In Ricardo Maldonado (ed.), i.c.w. Monica Sanaphre. Estudios Cognoscitivos del Espanol. Special issue of Revista Espanola de Lingüistica Aplicada. Logrono: 19-66. 2002 Deixis and subjectivity. In Frank Brisard (ed.), Grounding. The Epistemic Footing of Deixis and Reference, 1-28. Berlin/New York: Mouton de Gruyter. 2003 Extreme subjectification. English tense and modals. In Hubert Cuyckens, Thomas Berg, Rene Dirven and Klaus-Uwe Panther (eds.), Motivation in Language. Studies in Honor of Günther Radden, 3-26. Amsterdam/Philadelphia: John Benjamins. Mortelmans, Tanja 2000 Konjunktiv II and epistemic modals in German: A division of labour. In Ad Foolen and Frederike van der Leek (eds.), Constructions of Cognitive Linguistics, 191-215. Amsterdam/Philadelphia: John Benjamins. 2002 "Wieso sollte ich dich küssen, du hässlicher Mensch!" Α study of the German modals sollen and müssen as "grounding predications" in interrogatives. In Frank Brisard (ed.), Grounding. The Epistemic Footing of Deixis and Reference, 391^132. Berlin/New York: Mouton de Gruyter. 2003 The "subjective" effects of negation and past subjunctive on deontic modals: The case of German dürfen and sollen. In

204

Bert Cornillie

Friedrich Lenz (ed.), Deictic Conceptualisation of Space, Time and Person, 153-182. Amsterdam/Philadelphia: John Benjamins. Nordlinger, Rachel and Elizabeth C. Traugott 1997 Scope and the development of English modality. English Language and Linguistics 1: 295-317. Nuyts, Jan 2001 Epistemic Modality, Language, and Conceptualization. Amsterdam/Philadelphia: John Benjamins. 2002 Grounding and the system of epistemic expressions in Dutch: A cognitive-functional view. In Frank Brisard (ed.), Grounding. The Epistemic Footing of Deixis and Reference, 433-466. Berlin/ New York: Mouton de Gruyter. Olbertz, Hella 1998 Verbal Periphrases in a Functional Grammar of Spanish. Berlin/ New York: Mouton de Gruyter. Pelyvas, Peter 1996 Subjectivity in English. Generative Grammar Versus the Cognitive Theory ofEpistemic Grounding. Frankfurt am Main: Peter Lang. 2000 Metaphorical extension of may and must into the epistemic domain. In A. Barcelona (ed.), Metaphor and Metonymy at the Crossroads, 233-250. Berlin/New York: Mouton de Gruyter. 2001 a On the development of the category modal: a cognitive view. How changes in image-schematic structure led to the emergence of the grounding predication. In P. Kocsäny and A. Molnar (eds.), Wort und (Kon)text, 103-130. Frankfurt am Main: Peter Lang. 2001 b The development of the grounding predication: Epistemic modals and cognitive predicates. In Enikö Τ. Nemeth and Karoly Bibok (eds.), Pragmatics and the Flexibility of Word Meaning, 151-173. Current Research in the Semantics/Pragmatics Interface 8. London: Elsevier. Silva-Corvalan, C. 1995 Contextual conditions for the interpretation of "poder" and "deber" in Spanish. In J. Bybee and S. Fleishmann (eds.), Modality in Grammar and Discourse, 67-105. Amsterdam/Philadelphia: John Benjamins. Taylor, John 1996 Possessives in English. An Exploration in Cognitive Grammar. Oxford: Clarendon. Traugott, Elizabeth C. 1989 On the rise of epistemic meanings in English: An example of subjectifcation in semantic change. Language 65: 31-55.

Considerations on English and Spanish modals 1996

205

Subjectification and the development of epistemic meaning: The case of promise and threaten. In Swan Toril and Olaf Jansen Westvik (eds.), Modality in Germanic Languages, 185-210. Berlin/New York: Mouton de Gruyter. 2003 From subjectification to intersubjectification. In Raymond Hickey (ed.), Motives for Language Change, 124-139. Cambridge: Cambridge University Press. Traugott, Elizabeth C. and Richard B. Dasher 2002 Regularity in Semantic Change. Cambridge: Cambridge University Press.

Section III Case studies II - Adjectives

Adjectives and subjectivity* Angeliki

1.

Athanasiadou

Introduction

Subjectivity has b e e n established as a m a j o r topic of cognitive linguistics as it plays a very important role in h o w m e a n i n g is created and construed. 1 Its operation has b e e n applied to verbs, concerning the extension f r o m lexical to m o d a l verbs or auxiliaries, (Langacker 1990, 1997, 1998, 1999, 2003, this volume; Traugott 1989, 1995; Carey 1995; Sanders and Spooren 1996, 1997), to prepositions, conjunctions and discourse markers (Langacker 1990, 1998, 1999; Traugott 1995, 1999), to grammatical relations coded b y reciprocals ( K e m m e r 1995) or b y w o r d order (Stein 1995), to areas of syntax ( C o m p a n y this volume; V e r h a g e n 1995, 2001, and in this volume), to domains such as conditionality (Akatsuka 1997; N i k i f o r i d o u and Katis 2000), to n a m e j u s t a f e w contributions to the field. In this p a p e r I will b e concerned with adjective uses, and with the different positions they occupy in the conceptual and syntactic spectrum.

*

I would like to thank Rene Dirven, the two anonymous reviewers, the two editors of the volume and the series editors for constructive suggestions and comments on an earlier version of this paper. I am solely responsible for the views which follow. 1. Benveniste (1971[1958]: 225) was the first to characterize subjectivity as the ability of speakers to view themselves as subjects and examines how common grammatical categories contribute to this capacity of speakers. Lyons (1982: 107) claims that a subjective point of view is conveyed when it reveals the speaker's consciousness of himself/herself (I remember switching off the light) whereas an objective point of view is conveyed when it can be interpreted as a report of an event in which the speakers mention themselves as participants (/ remember myself switching off the light). Lo put it another way, the former example illustrates the speaker's offstage personal experience while the latter reports the onstage memory of something observed rather than experienced. Lhis still intuitive distinction made by Lyons concerns the contrast between "the subjective, experiencing, internal self' and "the objective, observing, external self', which is given a principled definition by Langacker.

210

Angeliki

Athanasiadou

Adjectives and the correlation between position, as a criterion for their semantic analysis, and subjectivity expressed by adjectives has been treated in both pre-cognitive and cognitive approaches (Adamson 2000; Bache 2000; Bolinger 1967; Dirven 1999; Ferris 1993; Hetzron 1978), particularly focusing on the order of adjectives in the premodifier zone. The choice of an adjective use in a certain position (and not only in the premodifier but also in the postmodifier and the predicative position) seems to be associated with the viewpoint of a speaker. The various interpretations adjectives may have are due to the conceptualizer's judgement of property ascription on the basis of built-in norms and values as well as to the position they occupy. An investigation will, thus, be attempted into the complex problem of the positional classification of adjectives in English. This is complex because the use of adjectives in different positions is partly dependent on the types of adjectives used. But, moreover, the choice of an adjective use in a certain position which, as will be claimed, is associated with the viewpoint of a speaker, qualifies for subjectivity. We actually have two interacting parameters: position and type of adjective use. As for the position occupied by an adjective, my analysis will be based on a dimension in which a conceptualizing subject is opposed to an object of conceptualization adopting a subjective vs. objective construal of an entity or, in this case, a relation (Langacker 1990). As for the use of a given type of adjective, I will focus on Langacker's revised characterization of subjectivity according to which "an objective relationship fades away leaving behind a subjective relationship that was originally immanent in it." (Langacker 1998: 75, 1999).

2.

Types of Adjective Uses and Subjectivity

2.1.

Types of Adjective Uses

The interpretation of adjectives is determined by their inherent properties, the meaning of the noun the adjectival properties are assigned to, the manner in which these different meanings are related, and also the linguistic and pragmatic context. Sweetser (1999: 147) summarizing Langacker's (1987: 271-4) treatment of active zones notes that "the noun referentially profiles some entity as a member of the appropriate category, while the adjective elaborates some active zone of the entity profiled by the noun." In Sweetser's elaboration of the above treatment (1999: 138, 147), in "red ball, the noun ball profiles the concept of a ball, and the adjective red

Adjectives and subjectivity

211

elaborates the color of the surface of the ball." This is the simplest interpretation that comes to the speakers' minds. However, depending on the context, red ball can be a ball filled with red paint or a ball with red marks (Sweetser 1999: 13 7-8).2 Taylor (1992: 20) discusses the different meanings the adjective old may have depending on the noun with which it makes a composite expression and says the adjective is being used in a synthetic sense {old friend is 'a friend of long standing', old girlfriend is 'a former girlfriend', old regime is 'a regime which no longer exists'), in addition to the absolute reading of old friend, which is 'a friend advanced in years'. 3 In fact, a variety of mechanisms may be involved in the semantic composition of adjectival modification. According to Sweetser (1999: 129), even 'simple' cases like red apple or red ball require cognitive mechanisms such as metaphor, metonymy, frames, mental spaces, active zones, profiling. The interpretation of expressions like old friend has been discussed as a case of metonymy (reference points) and also blending. As Sweetser says, "instead of each word always representing the same rigid and stable semantic chunk or building-block, the same word can represent very different complex meaning structures in different contexts and may alter flexibly depending on the meanings surrounding it" (Sweetser 1999: 136). Thus, speakers on the basis of context are able to determine 'simple' readings or readings requiring the evocation of other spaces or active zones. This interpretation of Adjective-Noun compositional meaning is a specific case of construal which will not be treated here as it does not involve subjectivity.

2. In Sweetser's expanded sense, 'active zone' may include "not only parts or aspects of the entity itself, but parts or aspects of the frames associated with it in the complex context of the particular utterance, and even counterparts of the entity in another mental space (whether via metaphorical mappings, or by other inter-domain counterpart relationships such as depiction, belief, time, and so on). The reading of red ball as the ball used by the team with the red uniforms takes the team's uniforms (linked to the ball by a cognitive frame of competitive ball-playing) as an accessible active zone of the ball" (1999: 147). 3. Absolute vs. synthetic readings of adjectives were discussed by Vendler (1967, 1968). A beautiful dancer is 'a dancer who is beautiful' in its absolute reading. In its synthetic reading she is 'a dancer who dances beautifully'. This distinction between absolute vs. synthetic readings of an adjective is somehow equivalent to the distinction scalar vs. event-like adjectives by Radden and Dirven, which will be discussed shortly.

212

Angeliki

Athanasiadou

It is taken for granted that adjectives express properties in relation to the things expressed by the nouns to which their properties are assigned. However, for their semantic description different classifications have been suggested which reflect rich insights, but which either capture different phenomena or lack any conceptual principles that would motivate the proposed classification. This issue can be illustrated even by a limited overview of the classification of adjectives in the literature at large. Bache (2000: 230-231) in his grammar distinguishes between the uses of adjectives in semantic oppositions: (i) gradable vs. non-gradable, (ii) inherent vs. non-inherent, and (iii) restrictive vs. non-restrictive. The gradable use of adjectives involves scalar properties (beautiful); the non-gradable use involves categorial properties (linguistic). The inherent use of adjectives involves the direct assignment of properties (beautiful woman) while in the non-inherent use, properties are assigned by association (old friend 'a friend of long standing'). It needs to be noted that gradable and inherent properties may coexist as in beautiful woman. Finally, an adjective like brave is characterized by a restrictive use if a group of soldiers is isolated in The brave soldiers ran forward ('some of the soldiers, the brave ones, ran forward'). If it refers to all the soldiers, it is non-restrictive ('the soldiers, who were all brave, ran forward') (Bache 2000: 231). The fact that the same phrase, e.g. the brave soldiers can simultaneously belong to each of the two values of the opposition does not imply that the above taxonomy is not satisfactory. In fact, Bache's contribution to the study of adjectives and his account of the position of adjective uses to be treated in Section 3, is highly valued. Bache's binary dimensions may capture different aspects of modification; in this way, many adjective uses can be characterized in terms of perhaps distinct values, yet values assumed simultaneously within each dimension. Paradis (2001: 51) adopts a cognitive approach and extends the notion of boundedness, which typically applies to things, to relations such as adjectives. She provides finer distinctions for gradable adjectives in particular, namely a distinction between scalar adjectives {long, good), extreme adjectives (terrible, brilliant), and limit adjectives 4 (dead, true, identical). These distinctions are based on the link between degree modifiers and adjectives. Scalar adjectives combine with scalar degree modifiers (fairly

4. There seems to be an oxymoron here as limit adjectives cannot be gradable

(*very dead).

Adjectives and subjectivity

213

long) and extreme 5 adjectives combine with reinforcing totality modifiers (iabsolutely terrible). The difference between the two is that the latter express the ultimate point in a scale. Limit adjectives combine with totality modifiers (completely dead) and they are different from the other two types in that they are conceptualized in terms of 'either.. .or' (Paradis 2001: 512). So *very dead is impossible, since being dead is not a question of more or less, but of being or not. Although we can have He was left behind half/almost dead there is no problem since half dead may entail 'not dead'. There have been prominent studies that set the basis for the different classifications of types of adjectives such as Bolinger's and Dixon's, among many others. Bolinger (1967: 15-16) deals with the opposition between attributive and predicative adjectives in English. Within attribution he posits a distinction between reference modification (criminal lawyer 'practitioner of criminal law') and referent modification (criminal lawyer 'criminal practitioner of law'); this distinction, however, deserves to be further elucidated. In this example, reference modification can be equated with the classifying function and referent modification with the characterizing function of adjectives. This case shows that the linguistic or the pragmatic context do not provide sufficient grounds to determine how an adjective characterizes a noun or, better, its referent. Dixon (1982) in his attempt to explain the ordering of adjectives (*four all collars vs. all four collars or ?dog leather collars vs. leather dog collars) tries to find a semantic principle and suggests a division of adjectives into seven semantic sub-classes: value (good, nice, horrible), dimension (small, long, thin), physical property (hard, heavy, smooth), speed (fast, quick, slow), human propensity (gracious, kind, proud), age (new, young, old), colour (black, green, red). The types of classification discussed above depend on the various ways adjectives can be described, a fact which already proves how hybrid this particular category is. Dixon groups them in seven categories because he aims at finding plausible ways of ordering them; Paradis claims that degree modifiers are fundamental for the classification of adjectives and that "there has to be a harmonious relationship between a degree modifier and its adjective" (Paradis 2001: 51); Bache distinguishes types of adjectives in terms of their properties in binary sets of semantic opposition. However,

5. I would prefer the noun extremity to Paradis's use of the adjective extreme because extreme constitutes a member of the categories to be described.

214

Angeliki Athanasiadou

we find a large number of adjectives which do not enter a binary system and he calls them indeterminate, such as economic, linguistic, religious. Using some of the insights discussed so far, I will mainly follow the analysis suggested by Radden and Dirven (to appear). It is based on the way adjectives can be conceptualized by speakers emphasizing their hybrid character in either being closer to nouns or to verbs. In particular, adjectives, on the basis of their meaning, may fall into four categories: scalar, event-like, thing-like, determiner-like. Scalar uses of adjectives can be conceptualized as denoting a range on a scale. Adjectives like long, fast, heavy, beautiful, intelligent designate variable properties such as length, speed, weight, values, personality traits, etc. These properties range from more to less, that is the adjectives that designate properties of entities tend towards opposing poles of a scale such as its maximum or minimum, while never reaching these endpoints. There is, however, a subgroup of scalar adjectives (as discussed by Paradis) such as terrible, excellent, brilliant which only represent the ultimate points of a scale and not the in-between values. Many scalar adjectives, as they range on a scale, may characterise an event or a state. They can be called event-like uses of adjectives as they specify a relationship between a thing and an action, or more generally, an event. So a beautiful novel is a novel written in a beautiful way, an elegant dancer is a person who dances elegantly. An adjective use, then, characterizes a noun in different ways. These two types of adjective uses are radically different from thing-like uses of adjectives which evoke participants within situations. There is no association with a scale nor with the way an event takes place. So, linguistic papers, presidential decision, doctoral dissertation, plastic bottle, musical box illustrate the thing-like character of these adjectives. Finally, determiner-like uses of adjectives such as same in the same person serve like additional determiners in the noun phrase. They have a referential function and, in addition to this, they have a grounding function in stressing the uniqueness of the referent. It must be noted here, that the same adjective may be cross-classified or may exhibit behaviour typical of other types (as in Bache's classification). Taking the adjective old again, as in a very old friend 'advanced in years'; this has, in Radden and Dirven's terms, a scalar use. In an old friend 'of long standing' and an old girlfriend 'a former girlfriend' the adjective has an event-like use. However, the latter case, namely an old girlfriend, could be contested to involve the determiner-like use which identifies the instance in both a definite NP (his former girlfriend) and an indefinite NP (a former girlfriend).

Adjectives and subjectivity

215

Consequently, adjectives cannot be rigidly claimed to belong to any of these types of use. They are characterized by a great deal of freedom in construal. Although speakers may agree on the meaning (as if this is a fixed thing) of adjectives, they do not necessarily agree on their application so that in the four types of use there is semantic variation. The variability of adjectives, their freedom of use and the flexibility in their semantic make-up lead to their hybrid character. In fact, any adjective can give rise to a variety of interpretations of the noun phrase; even those which have absolute readings like colours; cf. Langacker's example (1987: 273), a red pencil, which may be a pencil which is red from the outside, but it may also be a pencil that makes red marks or Sweetser's example of red ball. Yet, we are able to identify the reading our hearers wish to communicate because we view the two senses as more specific instantiations of a single, more abstract sense. According to Taylor (1992: 20), certain specifications of one sense are suspended or modified in the extended sense. He follows Langacker (1988: 51) in that semantic extension involves "some conflict in value." Thus, although the three senses of old are distinct: 'advanced in age', 'of long standing', 'former', yet, we view them as more specific instantiations or "elaborations" of a "schematic" sense called OLD 0 by Taylor. The commonality in the three senses of old, which can be seen in the form of a network, 6 is the entity's existence at some time preceding reference time (R), i.e. the time at which the trajector of old is characterized with respect to its oldness (Taylor 1992: 20). From the way adjectives have been categorized in the different approaches, it becomes evident that they constitute a very heterogeneous category. Their scalar as well as their event-like character tends to be conceptually close to verbs while their thing-like character and even their unique attribution of a property to a noun's referent tends to be closer to 6. In Taylor's network the meanings of old are represented as extensions and this is shown by broken lines, while elaborations are represented by continuous lines: OLD 0

OLD

OLD'

OLD"

216

Angeliki

Athanasiadou

nouns or a noun phrase's essential constituents. This tendency is actually what led Radden and Dirven to the above classification of adjective uses. For instance, size adjectives like big that describe a thing or an event imply more time stability and from this perspective they are more like nouns, whereas evaluative adjectives like nice or adjectives of emotion like sad are less time-stable and from this perspective they designate relations and can be compared to verbs. 7 Moreover, certain adjectives even become more like function words than content words. Paradis (2001: 60) claims that in Bill is a perfect husband and Bill is a perfect idiot, husband may be associated with perfection, but in idiocy the adjective is employed, according to her, as a degree modifier for reinforcement; 'perfection', thus being left in the background. I find that this kind of semantic extension qualifies for an analysis in terms of subjectivity.

2.2.

Adjective Uses and Subjectivity

In Langacker's earlier publications (1990, 1991), subjectivity was characterized as "the realignment of some relationship from the objective axis to the subjective axis." More recently, he has come to believe that "the subjective component is there all along, being immanent in the objective conception, and simply remains behind when the latter fades away" (Langacker 1999: 151). So he offers a revised characterization of subjectivity: "an objective relationship fades away, leaving behind a subjective relationship that was originally immanent in it (i.e. inherent in its conceptualization)" (Langacker 1998: 75-76). For the type of adjective used and the way it functions as a content or function word, I will be based on this revised characterization of subjectivity. An attempt will, then, be made to define adjectives as values ranging on a scale. As adjectives are assignments of values by speakers, their assignment takes place on the basis of norms necessary for their conception and usage.8 Let us discuss the property of completeness as seen in (1). 7. These differences between adjectives are reflected in their ordering possibilities. An evaluative adjective like nice is put before a descriptive adjective like big. So we have a nice big house rather than *a big nice house. Nice designates a highly subjective attribute whereas big designates a more stable attribute and, for this reason is located closer to the noun. 8. In Cognitive Grammar, the categories are distinguished according to the nature of their profiles; adjectives profile atemporal relations (Langacker 1987: Ch. 5

Adjectives and subjectivity

(1)

217

a. The complete works of Shakespeare. (LDCE: 271) b. He is a complete stranger to me. (LDCE: 1426)

In (la) complete as a property presupposes either a plural (the complete works), or a collective noun (the complete staff), or a mass noun (the complete harvest). I even dare to argue that complete in this example describes a 'spatial configuration' rather than a property so that it could be termed a gestalt adjective (following Broccias's (2004) terminology). Gestalt adjectives refer to some aspect of an object as a whole. In (lb) there is a shift from the measurable, rather objective configuration, towards a different type of quantification with a more subjective value scale. In this case, complete describes a property of any arbitrarily chosen part of an object and Broccias calls it a part-whole adjective. This change is motivated because complete as a property refers to a norm which also implies a range. The quantification contained in this meaning of complete applies to a range of the adjective strange (completely/absolutely/totally strange) or to the noun derived from it (a complete/an absolute/a total stranger). In spite of the difference between complete\ and complete2 there is also some commonality between them: both describe a quantity. Completeι describes a quantity of a set of things or substance, and complete2 describes a quantity or a measure in the presence of a given quality, i.e. someone can be a bit strange to you, may have become half strange, or may still be a complete stranger.9 The meaning of complete in (lb) resides in the activity of the conceptualizing subject. This activity is immanent in the conceptualization of the rather objectively construed situation of (la), but in (lb) it is used in abstraction from any objective spatial configuration. In other words, we can draw a distinction between objective aspects of objects accessible to any observer and independent of the conceptualizer, and subjective propand 6). Dirven and Taylor (1988: 386) identify two types of tall, an extensional and a positional: tall| as in high window extending from bottom to top and tall 2 in which the trajector is at the highest point in the vertical extension. In Taylor (2002) the discussion on tall continues. Tall profiles the relation between its trajector (it is overtly stated as a nominal entity, in this case an entity claimed to be tall) and its landmark which is not stated in a separate expression; its landmark is incorporated in the semantic structure of the word; it is a region in excess of a norm with respect to 'tallness'. Thus, an entity counts as 'tall' if its vertical extent exceeds an implicit norm. 9. The terms 'quantity' or 'measure' do not imply 'degree'. This is reserved for cases like very strange applied to behaviour but not *it is completely strange.

218

Angeliki Athanasiadou

erties which imply an assessment of the state of an entity on the part of the conceptualizer. In both examples, Paradis's and in (1) above, we find a shift from property to quantification in terms of subjectivity: attribution of a property in the complete yvorks/the works are complete; the perfect husband/the husband is perfect, but only modifiers in complete stranger and perfect idiot.

3.

Position of adjectives as a reflection of their conceptual content and subjectivity

The role of iconicity in the word order of adjectives was already taken up by Whorf (1956: 93) who proposed a general principle according to which adjectives denoting inherent properties are placed closer to the head noun than those denoting non-inherent properties. Dixon (1982: 25) elaborates on Whorf s notion of iconicity and proposes a scale of inherence in which, properties of VALUE and DIMENSION are taken to be less inherent than properties of AGE and COLOUR. His account is opposed to Quirk et al.'s 'natural' order of recursive qualification. Quirk et al. (1972: 924) basically focus on how adjectival modification is a form of scope-restriction in which the addition of adjectives progressively narrows down the class of referents. In fact, Dixon contrasts adjectives with other classes of premodifiers and claims that pre-adjectival modifiers do recursively restrict scope (hence, for him, the cleverest two men is not synonymous with the two cleverest men). But adjectives are not recursive. A clever brave man has, in Dixon's words, similar cognitive meaning as a brave clever man (Dixon 1982: 25); each adjective directly qualifies the head noun. He posits, though, a semantic distinction (which is similar to Quirk et al.'s 1972: 924-925) within the adjective group: all his categories are descriptive or referent-oriented except for the VALUE category which is affective or speaker-oriented. The function of the VALUE category is not to describe properties but to express speaker-response to it. So saying a nice book is as if implying Ί approve' and your knowledge of me allows you to infer what properties in the book have evoked that response. For Dixon VALUE adjectives are subjective; their referential meaning depends on the speaker's identity. The subjective/objective polarity is also suggested by Quirk et al. (1972: 924-925), who say

A djectives and subjectivity

219

...modifiers relating to properties which are (relatively) inherent in the head of the noun phrase, visually observable, objectively recognizable or assessible, will tend to be placed nearer to the head and be preceded by modifiers concerned with what is a matter of opinion, imposed on the head by the observer, not visually observed and only subjectively assessible.

The position an adjective may occupy is a seemingly formal criterion which may evoke its different interpretations. This issue is treated by Bolinger (1967, 1977) who notes the contrast between attributive adjectives that tend to characterize the noun, or the referent of a noun phrase, in terms of a permanent, typical, or enduring property, and predicative adjectives which tend to ascribe a temporary or incidental characteristic. Bache (2000: 238-239) examines the way in which pre-head attributive adjectives occupy the modificational zone between determination and categorization. Presenting the chart below Determination

The

Modification

Categorization

Specification

Description

Classification

same

beautiful

French

actress

he comments that there is no strict separation between the three zones but a continuum of values from determination to categorization via modification. The order adjectives may have in relation to the noun their properties are assigned is one of their fundamental characteristics. Moreover, the same adjective may appear in different positions depending on its subfunctions: In Scottish popular ballads, popular subclassifies ballads with respect to genre, while in Popular Scottish ballads, popular describes ballads (Bache's example 2000: 240). The correlation between the position of adjectives, in the premodifier zone, and degrees of subjectivity expressed by adjectives is discussed in typological studies, such as in Seiler (ed.) (1978). In this collection, Hetzron (1978) claims that the more objective and collective adjectives are closer to the head noun than the more subjective and individual. A general subjective-objective gradience is also viewed from the functional point of view, namely Halliday (1985), McGregor (1997), Bache (2000), stating that adjectives move from the more subjective and individuating uses located closer to the determiner end of the noun phrase to the more objective

220

Angeliki Athanasiadou

uses located closer to the head noun. Halliday, in particular, adds that the properties indicated by adjectives positioned towards the left hand end of the N P are more ephemeral ones, they are restricted in time, while the properties indicated by adjectives positioned towards the right hand end of the premodifier zone are more permanent ones. A very interesting point put forth by Tucker (1997) and Vandelanotte (2002) is that these ordering principles may be overridden if premodifier adjectives are not structured in terms of recursive modification, but in terms of coordinated, independent modification of the head noun (bright, classy, cheerful restaurant). The above-mentioned contributions concerning degrees of subjectivity may not be cognitively oriented ones but they do deal with speaker-evaluation and speaker-related meanings expressed by adjectives. Langacker (1991: Ch. 2) distinguishes four functions which are realized progressively from left to right in the nominal group: grounding, quantification, instantiation, and type specification. The grounding of an instance anchors it in the speech event taking place between speaker and hearer. As this is structurally situated at the far left, we can see a cline in the overall build-up of the nominal group not only from instantiation to type specification but also from subjective to more objective. W e can, then, say that subjective adjectives appear more to the left, i.e. closer to the subjective grounding end. Similarly, Adamson (2000: 60) identifies a link between subjectivity and leftness in the noun phrase. As the item de-subjectivises it moves to the right. So, because of this, according to Adamson, there is a difference between criminal old lawyer and old criminal lawyer, a difference which is overlooked by Dixon. This is, in fact, expected as he is based on types of adjective uses whereas Adamson basically focuses on position and subjectivity. Summing up, in these studies we have a group consisting of several adjectives each one either directly qualifying the head noun or the adjective to the left being subjective compared to the one closer to the noun which is objectively construed. These belong to the category attributive. Teyssier (1968: 242) calls this position 'intraposition'; attributive adjectives are situated between determinatives and nouns and "they are part and parcel of the process of nomination." They have a direct identifying or classifying function "to substantiate the extension of a subsequent noun." Teyssier contrasts attributive adjectives with predicative ones by saying they are in 'extraposition'. These adjectives are extraposed both to the noun and its determinative; these are independent of the process of nomination and they have an indirect qualifying or characterizing function "to substantiate the

Adjectives and subjectivity

221

comprehension of a preceding noun." In other words, adjectives in predicative position "describe an attribute peculiar to a previously determined nominal element" 10 (Teyssier 1968: 242). There is, however, another position, namely the postmodifier, which partly belongs to the category attributive, and which Teyssier calls appositive. Teyssier contrasts juxtaposition with intra- and extra-position. "The appositive adjective juxtaposed to a noun can be considered as an (explicit) form of attributive or an (implicit) form of predicative;" it shares either of the two functions depending on its position (Teyssier 1968: 243). By functioning as an explicit attributive the attributive value of an interposed adjunct is emphasized. Teyssier's example, Matters philosophical, is an emphatic way of classifying and categorizing whereas Alexander the Brave has a resumptive function in the sense of renaming an identified object. Functioning as an implicit predicative, the adjective is used in a non-determining, merely descriptive way as in A man young and handsome. Teyssier's contribution is interesting in that he posits a hierarchy of positions (Teyssier 1968: 242):

Intraposition Attributive

Juxtaposition Appositive

Extraposition Predicative

Juxtaposition constitutes a complex pattern, in-between an adjective-noun pattern (attributive) and a noun-attributive pattern with copula (predicative). Teyssier's tri-partite functional schema did not receive much attention as it appeared a year after Bolinger's account of adjectives and also due to the linguistic climate of the 1960s. Its implications, however, for the

10. Teyssier uses the terms extension and comprehension as defined in La Logique ou I' Art de Penser by A. Arnault and P. Nicole: "J' appelle comprehension de Γ idee, les attributs qu' eile enferme en soi et qu' on ne peut lui oter sans la detruire. J' appelle etendue de Γ idee les sujets ä qui cette idee convient. La comprehension marque les attributs contenus dans une idee et Γ extension les sujets qui contiennent cette idee" (Teyssier 1968: 242-3). [Comprehension of an idea concerns the attributes it contains in it which cannot be excluded without abnegating it. Extension of an idea concerns the subjects to which this idea conforms. Comprehension highlights the attributes contained in an idea and extension highlights the subjects which contain this idea]. The only reference cited is that it was first published in 1662.

222

Angeliki Athanasiadou

hierarchy of positions an adjective use may assume are particularly interesting and especially useful for the purposes of the present analysis. Radden and Dirven (to appear) distinguish three types of position for adjectives: (i) adjectives which describe stable properties and which occur in premodifier position, (ii) adjectives which describe temporary properties and which occur in postmodifier position, and (iii) adjectives which describe assigned properties and which occur in predications. The choice of any of the three positions adjectives may occupy, namely the choice of premodifier, postmodifier and predicative position, reflects the speaker's construal of the situation expressed as being of a particular nature; this constitutes each position's special meaning. I will attempt to demonstrate that the choice of different positions is not random but shows different degrees of subjectivity. In what follows, the different positions will be examined in relation to the types of adjective uses. The criterion of position is interesting because many adjectives are restricted to one or another of the three positions to be discussed. Many scalar uses of adjectives ('descriptive' in Bache's terminology) are found in both attributive premodifier and predicative position but not in attributive postmodifier position: The happy couple/the couple is happy/*the couple happy. Thing-like and determiner-like adjectives (in Bache's terminology 'classifying' and 'specifying', respectively) are found only in attributive premodifier position: The doctoral dissertation/*the dissertation is doctoral/*the doctoral The other man/*the man is other/*the man other.

dissertation

However, attributive adjectives may also be found in postmodifier position: He had no patience with problems hypothetical. (Bache 2000: 236) The property adjectives express in attributive postmodifier position may be of a temporary character: Stars visible, rivers navigable,

... (Bolinger 1967: 12; Bache 2000: 236; Radden and Dirven to appear)

Adjectives and subjectivity

223

When the same adjectives are in attributive premodifier position, they describe a more permanent, generally valid property. 11 Adjectives starting with the prefix a- (afraid, alone, aware,...) occur in predicative or in attributive postmodifier position: The children were afraid of the dark/Children afraid children...

afraid of the

dark/*The

The above overview of a limited number of studies dealing with the position of adjectives, though detailed, lacks a particular aspect. I would like to approach this issue, namely the choice of a particular position an adjective use may take in terms of Langacker's work (1990, 1997) on subjectivity, arguing that the use of each one of the different positions codes a difference in construal. Langacker (1990: 7) defines subjectivity and objectivity as viewing relations between a perceiver and an object of perception. A viewing situation is characterized as subjective when the perceiving subject is implicit (offstage) whereas in the objectively construed viewing situation the observer is salient and the focus of attention (onstage). For Langacker, an entity or an event is said to be objectively construed to the extent that it goes 'onstage' as an explicit, focused object of conception. Subjective construal of an entity entails that it remains 'offstage' as an implicit, unselfconscious subject of conception. Langacker (1990: 7) represents the optimal viewing arrangement in which the conceptualizer (C) is external to both the profiled, onstage region and the maximal scope of conception. In this kind of arrangement, the construal of C is maximally subjective while that of the conceptualized object (O) is maximally objective. He represents the egocentric viewing arrangement in which C has moved to the expanded, egocentrically defined onstage region. Each step along this path, the end-point of which is that C becomes O, increases the objectivity of C's construal and diminishes the objectivity of O. In what follows, I will show that the choice of different positions reflects a difference in construal that will be analyzed accordingly, with the different types of adjective uses (i) in the premodifier position being marked for a gradation of subjectivity starting from the optimal arrangement and step by step moving to the ego-

11. Bache (2000: 237) notes that adjectives in -able or -ible occur in postmodifier position in case a superlative precedes (They had the greatest difficulty imaginable getting there in time), or in case a word like only precedes (The only room suitable is the one on the third floor).

224

Angeliki Athanasiadon

centric region, (ii) the different types of adjective uses in the predicative position being marked for an arrangement where the conceptualizer is external to the onstage region and (iii) the different types of adjective uses in the postmodifier position being compatible with arrangements where the conceptualizer may either be distant or close to the object of conceptualization.

3.1.

Premodifier position

Adjectives in this position are used for identifying and delimiting an instantiation of the referent of the noun. It could be claimed that in performing the above functions, they tend to designate typical, permanent or enduring properties. With respect to time stability, their properties are either timeless or without any reference to time. This observation was first made by Bolinger (1967: 12-14) who showed that adjectives in this position modify the reference by specifying the extent to which a properly is applicable to a particular referent. Degrees of stability of a property depending on the different types of adjective uses can, however, be traced. From the four types of adjective uses in terms of the stability of a properly, it is the thing-like type which comes first in the hierarchy of stability. Thing-like uses of adjectives work in combination with the function of the noun and restrict entities to smaller subcategories. Examples of thing-like adjective uses are: economic programme, linguistic paper, presidential decision. Due to their thing-like character they are close to noun-noun compounds and they are characterized as being conceptually autonomous and independent. The next two types of adjective uses, namely scalar and - due to their verb-like characteristics - event-like ones, follow the thing-like type in terms of stability. As noted before, scalar uses of adjectives are conceived of as occupying a range or an extreme point on a scale of a norm for a property. When scalar and event-like adjectives are used in premodifier position, they indicate a speaker's personal experience as to what is considered, for instance, beautiful or intelligent. There is, however, a delicate difference between these two types of uses. (2)

Mary is a beautiful

woman.

Adjectives and subjectivity

225

(2) suggests a permanent attribute of Mary. Moreover, the conceptualized s built-in norm, 'she is beautiful for a woman', is disclosed. In the case of the use of adjectives as event-like, the characterization of the event in terms of stability is now totally different. The manner of the event cannot last longer than the event itself. In (3)

This is an intelligent

solution.

the event evoked is 'a solution found in an intelligent way'. This use, due to limited stability, is true of manner modification. Determiner-like uses of adjectives which are combined with nouns as determiners relate to an even more temporary instantiation and stress the uniqueness of a referent, as the former, the last, the previous, the present, the current, the future, the next, the very, the only,... These adjectives are typically used with nouns denoting occupations or special offices. They only appear in premodifier position, at least with their determiner-like meaning. One of their uses12 is that they specify the unique character of the noun which exists at a particular point in time, namely past {the former/the last president), present (current president) or future (next president), but in the designated time-period they are 'stable'. As they are temporally bounded for a particular period, their stability lasts for as long as the association between the property and the noun holds. We can see that there is a cline of degrees of stability and permanence starting from thing-like uses which characterize autonomous and independent properties, then scalar uses denoting permanent attributes of an entity, then ranging to event-like uses characterized by limited stability, and ending with determiner-like uses with even more temporary instantiations. This order is inversely analogous to the degree of subjectivity exhibited by each use. One might ask how the four different types of adjective uses are linked via time-stability to subjectivity. The thing-like character {presidential decision) as well as the unique attribution of a property to a noun's referent by determiner-like uses {only solution) both designate relations closer to nouns. This closeness to nouns adds to their stability, and on this basis the object of conceptualization is clearly apprehended with the thinglike uses, or it becomes the focus of the viewing attention in the determiner-like uses. Scalar and event-like uses that designate relations closer to

12. Another determiner-like use of adjectives is when it occurs in indefinite NPs, as another problem, a former girlfriend, ...

226

Angeliki

Athanasiadou

verbs are more subjective than the former set. The fact that they rest on the conceptualizer's evaluation makes one wonder why an intelligent solution is less time-stable than a beautiful woman. Intelligent solution means that a certain solution evokes 'intelligence' in the conceptualiser's mind. It may not be a relatively permanent feature of the person who came up with the solution (He is not brilliant at all but came up with an intelligent solution that time), but it may also refer to a time-stable property of the process evoked by 'solution' with respect to shared standards of intelligence (Pythagoras ' Theorem is a solution to a specific problem which we still find intelligent). Conversely, 'beautiful' may be a permanent attribute in a beautiful woman but it can also be a stage-level predicate in She is a beautiful woman when she is dressed like that,13 But at least in usual circumstances, event-like uses of adjectives, due to their event-like character, express less time-stable relations than scalar uses. They differ in the degree of subjectivity: event-like uses are closer to the object of conception whereas scalar uses are distinct from it. When the speaker uses an adjective in premodifier position, s/he presents it as experienced or observed from an insider's perspective. The import of the premodifier position is to mentally transport the listeners to another space, allowing them to observe the characterization of the entities as they unfold through the eyes of someone who witnesses or experiences them and moreover to make them take what is said for granted. However, in the case of the four different types of adjective uses, I find a gradation of subjectivity and objectivity of the perceiving individual and of the entity perceived. We can start with the thing-like uses of adjectives that express properties that are autonomous and time-stable: The speaker is external to both the maximal scope of predication and to the onstage region of the conceptualized entity. The construal of the conceptualizer is subjective and that of the entity objective (Figure 1 (a)). The scalar use of an adjective of evaluation, such as beautiful refers to an aesthetic judgement. It conforms to a norm for beauty on the part of the conceptualizer. In (2) Mary is a beautifid woman, beauty is experienced and attributed to a person by the conceptualizer. The person herself may or may not find herself beautiful. The conceptualizer is now moving towards the inner part of the maximal scope region viewing the attribution from within and thus diminishing the objectivity of the conceptualized object (O) (Figure 1 (b)). This explains, in part, why scalar uses of adjectives of

13.1 owe this observation to the series editors' comments.

Adjectives and subjectivity

227

evaluation are located at the greatest distance from the noun when other scalar uses are present as well. 14 Similarly, with event-like uses of adjectives expressing properties that last for a limited period of time ((3) This is an intelligent solution), the speaker is a step closer to the object of conception (Figure 1 (c)).

CeD OS/

r

MS

MS

MS

Θ (a)

(b)

(c)

(d)

Figure 1. MS: maximal scope of conception OS: onstage region O: focused object of conception S: subject of conception (speaker) ^ : apprehension/construal by S

With determiner-like uses of adjectives that identify more closely the instance relevant to the referent of the noun, the speaker is implicitly coded as an experiencer viewing this individuating use from within a privileged vantage point (as in, for instance, The only solution to the problem...) (Fig-

14. In a nice large round old brown Chinese table, nice is at the greatest distance from table followed by adjectives of size, shape, age, colour in exactly this sequence. Last, adjectives of provenance and origin come immediately before the noun they modify (R. Dirven (ed.) 1989: 128).

228

Angeliki

Athanasiadou

ure 1 (d)). As determiner-like uses occur closest to the determiner-end in the NP, this confirms the subjective-objective gradience from left to right, which is accepted for the NP (Bolinger 1967, Halliday 1985). There are, of course, properties such as dimensions and colours which are routinely and naively regarded as being out there in the world independently of the conceptualizer. Still, when these properties are in premodifier position they are presented from the insider's perspective conforming to the norms and values for dimensions or colours on the part of the conceptualizer. For the premodifier position I could say that the speaker goes, step by step, onstage, within an expanded, egocentrically-determined onstage region. Becoming the focus of viewing attention or moving towards it increases the objectivity of the conceptualizer's construal and diminishes that of the perceived object which is construed more or less subjectively (Langacker 1990: 8).

3.2.

Postmodifier position

Adjectives used after the noun, though still belonging to it, are said to be in postmodifier position. Following Teyssier's classification (1968), when functioning as an explicit form of attributive they classify, categorize or subcategorize entities in an emphatic way; when functioning as an implicit form of predicative they simply describe entities. Thus, adjectives used in the postmodifier position either classify, categorize, subcategorize emphatically or describe entities. I will suggest that in the former function the speaker is inside or close to the entity whereas in the latter function the speaker is distinct from the object of conceptualization. Thing-like and (due to their use in particular position) determiner-like uses of adjectives are normally not found there. We find event-like uses of adjectives emphasizing the event causing a result or scalar uses of adjectives assigned to things (though to a lesser degree). The event-like use of the adjective navigable in premodifier position in (4a) refers to a river with the stable, permanent, enduring attribute of navigability. The adjective identifies and classifies the referent of the noun it modifies. (4)

a. A navigable river. b. A river navigable in winter. c. The only river navigable is the Nile.

Adjectives and subjectivity

229

But in (4b) and (4c) the 'temporary' markings allow for setting up further subcategories of things in even smaller subgroups. These contextual devices such as in winter, and only strengthen the temporary and specific character of the use of the adjective navigable. This specific and unique characteristic of the river Nile is presented from the insider's perspective and is hardly questionable. The viewer is within the onstage region, very close to the object of conceptualization.

Figure 2. This example also shows once more that adjectives may never belong to a fixed class as they can be used in different ways. As Langacker (2003: 1) says "it is not an expression's meaning that is subjective or objective but it is rather the individual elements it comprises that are construed subjectively or objectively; and an expression's meaning always comprises both subjectively and objectively construed elements." So it may be the case that adjectives may occur in all positions. Each single position imposes its meaning on the 'meaning potential' of the adjective. 15 The special character of the postmodifier position is further strengthened by the fact that certain forms of adjectives, like past participles, basically occur in this position. They also need to be further specified by time, manner or circumstantial adjuncts. (5)

a. ?A built house. b. A house built near the river.

15. Here we have to do with Goldberg's 'constructional' meaning (1995) which overrides word meaning. But this issue will not be treated in the present paper.

230

Angeliki Athanasiadou

(5 a) sounds odd in a neutral context and if it is to exist in premodifier position at all it needs a temporal (recently built house) or a manner adverb (icarefully built house)}6 In (5b) the occurrence of the completed event at some past time foregrounds the unique character of the event. Similarly, taken items sounds odd and it is dispreferred in this position. It is acceptable in postmodifier position as items taken home must always be checked. Or given funds vs. funds given when in need are welcome. The past participles built, taken, given in postmodifier position profile a resultant endstate, which, however, is not sufficiently informative. For this reason, it necessitates various types of adjuncts which specify it further. The general after-the-fact report of the event is consistent with the external to the onstage region perspective of the conceptualizer. It is a post event general report providing information. The speaker has no involvement in the event; s/he only describes it as an event that took place at some past time. Past participles may occur in both positions: either denoting a property of the noun's referent, such as stolen jewels, in premodifier position or jewels stolen in postmodifier position denoting an action.17 The speaker may provide information on Jewels stolen from an external perspective; the event is still activated as at some point will allow the police to identify the robbery. But in Stolen jewels are very difficult to sell or to wear there is no event, but a category of its own with a permanent property, namely 'jewels not freely usable'. This is consistent with an egocentrically viewing arrangement. We can only be led to the view that certain adjectives are unique in terms of the position they occupy. Built, taken, given are past participles and, as we know, they profile the final, resultant state of a process. However, some of them can express a more or less stable property of the referent of the noun and occur in premodifier position and certain others occur in postmodifier position contributing to a more unique or specific instance of the referent of the noun. (Cf. some of Radden and Dirven's examples such as water drinkable when boiled, products marketable in winter, woman lying on the floor, ...). This again is clearly an instance of the difference between lexical meaning and constructional meaning. 16. As noticed by Ackerman and Goldberg (1996: 22), participles of verbs of creation in premodifier position require some adverb in order to be acceptable. So

we have half-baked cake and not *baked cake. 17. This is an example discussed in Bolinger (1967: 3) who claims that past participles are unambiguous in premodifier and postmodifier positions. This is in conformity with what has been said so far about the two positions reflecting specific properties.

Adjectives and subjectivity

231

Figure 3.

3.3.

Predicative position

Premodifier and postmodifier adjective positions are crucial for the following functions: for modifying a thing by identifying its referent and for classifying things by delimiting their instantiation, that is, restricting their class of referents. When adjectives are used in predicative position they predicate a property of a thing. Speakers qualify or characterize the thing they refer to without modifying its value. They describe entities with attributes peculiar to the ones previously determined. From the four types of adjective uses, scalar and (to a lesser degree) event-like ones can predicate a property of a thing. Certain scalar adjectives are even exclusively used in predicative position. For instance, ill, glad, asleep, alone, aware never occur in premodifier or postmodifier position. They have their counterparts that occur in premodifier position, namely sick, happy, lonely,.. (6)

a. The man is drunk. b. He is glad to see me. c. This man has been ill for quite some time. (Radden and Dirven, to appear)

232

Angeliki

Athanasiadou

The instances above are typical examples of this position. Adjectives like these designate a transient state of the body or of the mind and they describe properties by only naming them. Scalar uses of adjectives like beautiful found in predicative position can designate a permanent or a temporary property. (7)

a. She is naturally beautiful. b. She is beautiful tonight in that new dress.

(permanent) (temporary)

In (7a) the property is associated with her. In other words, beauty is inherent in her; a striking property of hers is that of being beautiful. This property can be restricted to special occasions and now the 'constructional' position or the context imposes a non-permanent interpretation, as in (7b). Similarly, navigable is ambiguous in denoting a more lasting and permanent or a more transient and restricted property of an entity: (8)

The river is navigable

a. all year round. b. only in winter.

(permanent) (temporary)

(vs. the unambiguous the only river navigable (occasion) and the only navigable river (characteristic)). 18

Figure 4.

18. The properties 'characteristic' and 'occasion' are used by Bolinger in his discussion of -able adjectives (1967: 3).

Adjectives and subjectivity

233

Thus, the speaker may describe an entity as either having a permanent or a restricted property. In any case, s/he codes his/her presence as a nonexperienced reporter, external to the onstage region. The speaker is within the maximal scope of predication but is external to the onstage region of the conceptualized entity. The subjectivity of the speaker and the objectivity of the entity described do not exhibit complete polarization; the distance between them diminishes slightly. The speaker may not have any involvement or direct access to the entity, but s/he presents information with a perspective external to the entity. Behavioural adjectives having an evaluative use 19 are less inherent in an entity, and may denote a temporary interpretation: (9)

a. You are being extremely kind, thank you. b. Don't be so noisy. I can't hear anything. c. He had been very rude to his aunt. (Radden and Dirven, to appear)

In (9), the construction 'X is being ' requires an adjective which denotes some form of non-stative, active conduct. It describes an evaluation, designating a person's behaviour; here again, the construction imposes its meaning on the 'meaning potential' of the adjective. The predicative position seems to be kept for predicating a property which is considered as such, by the conceptualizer for a certain entity. The conceptualizer evaluates an entity as beautiful, ill or navigable, from a general, global viewing and reporting, and assigns it to a noun's referent as having either a permanent or a restricted character. In other words, speakers do not have any involvement or direct access to the qualification or characterization of an entity. This is consistent with an external perspective of simply focusing on providing information. When the assignment of the properties is there but there is no particular grounding device, like in winter in The Nile is navigable (in winter), or when she laughs in She is beautiful (when she laughs), then the interpretation can only be inferred. The construal of the speaker is subjective and the hearer may implicate meanings that are not linguistically coded. Speakers, in this case, in an attempt to assign properties to things, choose a more 'reserved' and perhaps less obliging variant either because they do not have the power to do so or be-

19. We find such uses of adjectives in evaluative constructions as: Ifind it strange, It is kind of her to say so, It was very cruel of you to say such things.

234

Angeliki Athanasiadou

cause they do not wish to draw attention to their power. The predicative position is appropriate for a general viewing and reporting correlating with external non-experienced readings. It is actually the speaker's memory of something observed rather than experienced.

4.

Concluding Remarks

In this paper an attempt has been made to show that subjectivity appears to shed light on the way speakers employ the conceptual category of adjectives. Notions of VALUE and N O R M have been employed and properties may figure on a scale along them. Original and extended meanings have been compared in example (1). They were characterized by residing in conceptual operations immanent in the conception of such values. The original meanings have been objectively construed whereas the extended, more abstract, meanings have been construed independently of any onstage content. They are offstage and subjectively construed and we only become aware of their semantic extension through their effect in shaping our conception of relations construed objectively. To account for the use of different types of adjectives being located in any of the three positions, a particular dimension of subjectivity has been adopted, namely, one in which a conceptualizing subject is opposed to an object of conceptualization. The asymmetry between the two is maximal when the conceptualizer lacks self-awareness, focusing exclusively on apprehending the onstage relation, and the object of conceptualization is salient and clearly apprehended. This optimal viewing arrangement is expressed by the thing-like uses of adjectives in the premodifier position. Apart from this arrangement of extreme polarization between the conceptualizer and what is conceptualized, there is the other extreme situation, with the egocentric viewing arrangement expressed by the determiner-like uses of adjectives in the premodifier position, where S becomes the focus of the viewing attention (S=0). The subject and object roles are not sharply distinct but figure with a greater degree of subjectivity or objectivity in the case of scalar and event-like uses of adjectives. They are found in all three positions (premodifier, postmodifier, predicative) with matters of degree, figuring in the construal of the different situations described: event-like uses being closer to the object of conception than scalar uses which are more distinct from it.

Adjectives and subjectivity

235

The positional classification of adjective uses in English, proposed here, on the basis of subjectivity is just one step that leads to the understanding of the conceptual category of adjectives and of its limits.

References Ackerman, Farrell and Adele Ε. Goldberg 1996 Constraints on adjectival past participles. In Adele Ε. Goldberg (ed.), Conceptual Structure, Discourse, and Language, 17-30. Stanford, Calif.: CSLI Publications. Adamson, Sylvia 2000 A lovely little example. In Olga Fischer, Anette Rosenbach and Dieter Stein (eds.), Pathways of Change, 39-66. Amsterdam/Philadelphia: John Benjamins. Akatsuka, Noriko 1997 Negative conditionality, subjectification, and conditional reasoning. In Angeliki Athanasiadou and Rene Dirven (eds.), On Conditionals Again, 323-354. Amsterdam/Philadelphia: John Benjamins. Bache, Carl 2000 Essentials of Mastering English. A Concise Grammar. Berlin/ New York: Mouton de Gruyter. Benveniste, Emile 1971 Subjectivity in language. In Problems in General Linguistics, 223-230, trans. Mary Elizabeth Meek. Coral Gables, FL: University of Miami Press. [Original edition 1958] Bolinger, Dwight 1967 Adjectives in English: Attribution and predication. Lingua 18: 1 34. 1977 Meaning and Form. London: Longman. Broccias, Cristiano 2004 The cognitive basis of adjectival and adverbial resultative constructions. Annual Review of Cognitive Linguistics 2: 103-126. Carey, Kathleen 1995 Subjectification and the development of the English perfect. In D. Stein and S. Wright (eds.), Subjectivity and Subjectivisation, 8 3 102. Cambridge University Press. Company Company Concepcion this vol. Subjectification and syntactic cancellation.

236

Angeliki

Athanasiadou

Dirven, Rene (ed.) 1989 A User's Grammar of English: Word, Sentence, Text, Interaction. Peter Lang. 1999 The cognitive motivation for adjective sequences in English. Journal of English Studies 1(1): 57-67. Dirven, Rene and John Taylor 1988 The conceptualization of vertical space in English: The case of tall. In B. Rudzka-Ostyn (ed.), Topics in Cognitive Linguistics, 379-402. (CILT 50.) J. Benjamins Publishing Company. Dixon, R. M. W. 1982 Where Have All the Adjectives Gone? Berlin: Mouton de Gruyter. Ferris, Connor 1993 The Meaning of Syntax: A Study in the Adjectives of English. London, New York: Longman. Goldberg, Adele 1995 Constructions: A Construction Grammar Approach to Argument Structure. Chicago: The University of Chicago Press. Halliday, Michael A. K. 1985 An Introduction to Functional Grammar. Edward Arnold. Hetzron, Robert 1978 On the relative order of adjectives. In H. Seiler (ed.), 165-184. Kemmer, Suzanne 1995 Emphatic and reflexive -self: expectations, viewpoint, and subjectivity. In D. Stein and S. Wright (eds.), Subjectivity and Subjectivisation, 55-82. Cambridge University Press. Langacker, Ronald 1987 Foundations of Cognitive Grammar, vol. I, Theoretical Prerequisites. Stanford: Stanford University Press. 1988 A view of linguistic semantics. In B. Rudzka-Ostyn (ed.), Topics in Cognitive Linguistics, 49-90. Amsterdam/Philadelphia: John Benjamins. 1990 Subjectification. Cognitive Linguistics 1-1: 5-38. 1991 Foundations of Cognitive Grammar, vol.2. Descriptive Application. Stanford: Stanford University Press. 1997 Consciousness, construal, and subjectivity. In Maxim I. Stamenov (ed.), Language Structure, Discourse and the Access to Consciousness, 49-75. Amsterdam/Philadelphia: John Benjamins. 1998 On subjectification and grammaticization. In Jean-Pierre Koenig (ed.), Discourse and Cognition: Bridging the Gap, 71-89. Stanford: CSLI Publications.

Adjectives and subjectivity

237

Losing control: Grammaticization, subjectification, and transparency. In Andreas Blank and Peter Koch (eds.), Historical Semantics and Cognition, 147-175. (CLR 13.) Berlin/New York: Mouton de Gruyter. Grammar and Conceptualization. (CLR 14.) Berlin/New York: 2000 Mouton de Gruyter. 2003 Extreme subjectification: English tense and modals. In Hubert Cuyckens, Thomas Berg, Rene Dirven and Klaus-Uwe Panther (eds.), Motivation in Language: Studies in Honor of Günter Radden, 3-26. Amsterdam/Philadelphia: John Benjamins. this vol. Subjectification, grammaticization, and conceptual archetypes. Longman Dictionary of Contemporary English (LDCE) 1995 Longman Dictionaries. Third Edition. Lyons, John 1982 Deixis and subjectivity: Loquor, ergo sum?. In R. J. Jarvella and W. Klein (eds.), Speech, Place, and Action: Studies in Deixis and Related Topics, 101-124. New York: John Wiley and Sons. McGregor, William B. 1997 Semiotic Grammar. Oxford: Clarendon. Nikiforidou, Kiki and Demetra Katis 2000 Subjectivity and conditionality. The marking of speaker involvement in Modern Greek. In Ad Foolen and Frederike van der Leek (eds.), Constructions in Cognitive Linguistics, 217—237. Amsterdam/Philadelphia: John Benjamins. Paradis, Carita 2001 Adjectives and boundedness. Cognitive Linguistics 12-1: 47-65. Quirk, R., S. Greenbaum, G. Leech and J. Svartvik 1972 A Grammar of Contemporary English. Harlow: Longman. Radden, Günter and Rene Dirven to appear Cognitive English Grammar. Amsterdam/Philadelphia: John Benjamins. Sanders, Jose and Wilbert Spooren 1996 Subjectivity and certainty in epistemic modality: A study of Dutch epistemic modifiers. Cognitive Linguistics 7: 241-264. 1997 Perspective, subjectivity, and modality from a Cognitive Linguistic point of view. In W. A. Liebert, G. Redeker and L. Waugh (eds.), Discourse and Perspective in Cognitive Linguistics, 85112. Amsterdam/Philadelphia: John Benjamins. Seiler, Hansjakob (ed.) 1978 Language Universals. Tübingen: Narr. 1999

238

Angeliki

Athanasiadou

Stein, Dieter 1995 Subjective meanings and the history of inversions in English. In D. Stein and S. Wright (eds.), Subjectivity and Subjectivisation, 129-150. Cambridge: Cambridge University Press. Sweetser, Eve 1999 Compositionality and blending: semantic composition in a cognitively realistic framework. In Th. Janssen and G. Redeker (eds.), Cognitive Linguistics: Foundations, Scope and Methodology, 129-162. Berlin: Mouton de Gruyter. Taylor, John R. 1992 Old problems: Adjectives in Cognitive Grammar. Cognitive Linguistics 3—1: 1-36. 2002 Cognitive Grammar. Oxford: Oxford University Press. Teyssier, J. 1968 Notes on the syntax of the adjective in Modern English. Lingua 20: 225-249. Traugott, Elizabeth Closs 1989 On the rise of epistemic meanings in English: An example of subjedification in semantic change. Language 65: 31-55. 1995 Subjectification in grammaticalisation. In D. Stein and S. Wright (eds.), Subjectivity and subjectivisation, 31-54. Cambridge: Cambridge University Press. 1999 The rhetoric of counter-expectation in semantic change: A study in subjectification. In Andreas Blank and Peter Koch (eds.), Historical Semantics and Cognition, 177-196. (CLR 13.) Berlin/ New York: Mouton de Gruyter. Tucker, Gordon H. 1997 A functional lexicogrammar of adjectives. Functions of Language 4.2: 215-250. Vandelanotte, Lieven 2002 Prenominal adjectives in English: Structures and ordering. Folia Linguistica XXXVI/3-4: 219-259. Vendler, Zeno 1967 Linguistics in Philosophy. Ithaca: Cornell University Press. 1968 Adjectives and Nommalizations. The Hague: Mouton. Verhagen, Arie 1995 Subjectification, syntax, and communication. In D. Stein and S. Wright (eds.), Subjectivity and Subjectivisation, 103-128. Cambridge: Cambridge University Press. 2001 Subordination and discourse segmentation revisited, or: Why matrix clauses may be more dependent than complements. In Ted Sanders, Joost Schilperoord and Wilbert Spooren (eds.), Text

Adjectives and subjectivity Representation. Linguistic and Psycholinguistic Aspects, 357. Amsterdam/Philadelphia: John Benjamins, On subjectivity and 'long distance Wh-movement'.

239 337-

this vol. Whorf, B. L. 1956 Language, Thought, and Reality: Selected Writings of Benjamin Lee Whorf, J. B. Carroll (ed.). Cambridge, Mass.: M.I.T. Press.

Grammaticalization and subjectiflcation of the English adjectives of general comparison1 Tine Breban

1.

Introduction

In recent years, study of the shift from 'lexically full' to 'grammaticalized' meanings has become well established for verbs and nouns. With regard to verbs, there is, for instance, the well-known, growing body of investigations into the shift from lexical verb to modal auxiliary (amongst others Traugott 1989; Langacker 1990, 1991, 1998, 1999, 2003; Heine 1993; Krug 2000; Traugott and Dasher 2002: chapter 3; Cornillie this volume; Mortelmans this volume; Pelyvas this volume). Concerning nouns, increasing attention is being devoted to nouns losing their status as head of a noun phrase (NP) and coming to function as a - re-analysed - part of another type of structure such as a conjunctive item and/or discourse marker (Traugott 1995: 39^12, 1999, 2003b), or a modifying (Tabor 1994; Denison 2002) or quantifying expression (Langacker 1991: 88; Brems 2003a, 2003b). However, less attention seems to have gone so far to processes of delexicalization and grammaticalization affecting adjectives. There have been a few studies focusing on the development of adjectives into degree modifiers (Paradis 2000; Nevalainen and Rissanen 2002; Pandermaat this volume) and the relation between different adjectival functions and subjectiflcation, in the NP (Adamson 2000) and in general (Athanasiadou this volume). In this article, I will focus on a class of adjectives which displays grammaticalization processes that are revealing about the possibilities of semantic and grammatical re-analysis within the adjectival field, viz. ad1. The research reported on in this article was funded by the Fund for Scientific Research - Flanders (Grant FWO G.0218.01, "Categorization and instantiation in the nominal group"). I would like to thank the anonymous referees for their interesting and inspiring comments. Special thanks are due to Kristin Davidse for her many insightful remarks on an earlier draft of this article and for her contribution to the discussion on adjectives of comparison in general.

242

Tine Breban

jectives of general comparison such as same, identical, different, other and similar. These are adjectives which, in their lexical uses, do not express a comparison with regard to particular properties, as comparative forms of ordinary adjectives do, but simply express likeness or unlikeness between entities (Halliday and Hasan 1976: 76-77). As I have argued earlier (Breban 2002, 2002/2003; Breban and Davidse 2003), these adjectives are, in present-day English, polysemous between two types of readings, fully lexical readings and grammaticalized, textual readings. Compare for example (1) and (2). (1)

(2)

"This southern milk is nowhere near as nice as green top milk. No way. " "A different taste altogether isn't it?" "Completely different." (CB)2 If you have problems once you arrive at the cottage, the agency may be able to move you to a different house or solve the difficulty. (CB)

In (1), α different taste compares the taste of southern milk to that of green top milk and expresses that the former has few features in common with the latter. Different in (1) is concerned with the expression of gradable '(un)likeness' between two entities. (Note that in the subsequent phrase completely different, the degree of 'unlikeness' is graded explicitly by the adverb completely.) In (2), by contrast, different does not convey that the new house has few features in common with the cottage, but rather indicates that 'a different', in the sense of 'not the same instance o f , house is involved. Different hence signals the 'referential' status of the NP a different house: it indicates that a 'new' instance is involved of a type of which other instances are present in the discourse context. These two types of uses correspond to different grammatical functions which adjectives can realize in the NP (Breban 2002; Breban and Davidse 2003). The lexical uses are prototypically associated with the function of 'attribute'. Attributes always express a quality of the instantial set referred to by the NP (Halliday 1994: 184; McGregor 1997: 176-178). In the case of the adjectives of comparison, they attribute the quality of being like or

2. The examples marked CB are taken from the COBUILD Corpus (The Bank of English) and reproduced here with the kind permission of HarperCollins. The examples marked INL are Dutch data extracted from the 38 million words corpus of the INL.

English adjectives of general comparison

243

unlike another entity and express how much one entity is like or unlike the other entity. In other words, they indicate how many qualitative features different entities have in common, ranging on a gradable continuum from none or few (adjectives expressing difference), over many but not all (adjectives expressing similarity), to all (adjectives expressing identity). In example (3), for instance, different attributes the quality of being considerably unlike Tel Aviv to Haifa in the sense that it shares only few or no qualitative features with Tel Aviv; for example the features that are enumerated in the rest of the sentence, flat and open, only characterize Tel Aviv. The modifier very explicitly grades the unlikeness as being large. (3)

Again, the weather report in Haifa is not my expertise at this exact moment, but Haifa's a very different city from Tel Aviv which is very flat and open and the dissipation of chemical agents will be much swifter. (CB)

The grammaticalized, textual, uses of the adjectives of comparison can link up with two functions that premodiiying adjectives can realize. Firstly, in examples such as (2), (4) and (5), they give referential information about the instantial set denoted by the NP. In (4) and (5), for instance, they indicate whether the instance is respectively the same instance as or a different one from other instances present in the discourse context. (4)

(5)

Popescu injured knee ligaments during Saturday's 2-1 defeat by Wimbledon and said last night: "I won't be able to play for three or four weeks. " Campbell tore a hamstring in the same game and could be sidelined for up to six weeks. (CB) Mr. Plante has now graduated from cooking school and is working as a chef. His former wife is contesting the papers she signed giving custody of their daughter to Mr. Plante's parents at the behest of her mother. But Kayla's natural mother has not appeared in court on any of the three occasions her case has been scheduled. The court has been informed that she has had a second child with another man, this time a son named Michael. And according to Mrs. Plante, the parents of Michael's father now have custody of their son's son. (CB)

In (4), same indicates that the specific game referred to is 'Saturday's 2-1 defeat by Wimbledon'. (An)other in (5), by contrast, signals that another

244

Tine Breban

instance of the type 'man' than the previously mentioned 'Mr. Plante' is involved. These textual uses, which give extra information about the instance referred to, function as 'postdeterminers', a sort of secondary determiners which add further to the identification or quantification of the instantial set (Halliday 1994: 183; Davidse 2001: 2). A second, less frequent, group of textual uses do not give additional information about the instances referred to, but about the subtypes derived from the general type. In (6), for instance, other indicates that the problems involved not only cover political ones but also include other types of problems to be filled in by the reader, such as e.g. social problems, cultural problems, etc. (6)

Foreign trade strengthens co-operation between nations, eases mutual understanding, makes the solution of political and other problems easier, and creates an atmosphere of trust, security and peace. Trade is after all considered to be the harbinger of peace. (CB)

These textual uses are 'classifiers', i.e. adjectives which derive a subtype of the general type denoted by the head noun (Halliday 1994: 184-186). They are, however, different from the cases of classifiers discussed by Halliday (1994) and Halliday and Hasan (1976) so far, in that those were all lexical classifiers, like political in (6). But, as illustrated here, classifiers can also be realized by adjectives of comparison used in a delexicalized, 'textual' sense. In example (6), other only indicates that different types of problems from the political ones are involved, without lexically predicating which subtypes they are. Grammaticalized adjectives of comparison can thus be used as classifiers, indicating that the subtype which they derive from the general type is the same one as or a different one from another subtype.3 It was further argued in Breban (2002) and Breban and Davidse (2003), that the correspondence between lexical reading and attribute use and textual reading and postdeterminer or classifier use is also confirmed by the formal behaviour of the lexical versus the textual uses. As, amongst others,

3. It has to be noted that there are also classifiers such as identical in identical twins which have a lexical value as they indicate that the two persons involved look identical or have the same genetic material, but these fall outside the scope of the present discussion.

English adjectives of general comparison

245

Quirk et al. (1972) and Bolinger (1967) observe, the fully lexical uses have formal alternates typically associated with attributes which the two grammaticalized textual uses do not have, viz. they can be graded 4 (7) and allow alternation with a pragmatically equivalent predicative use (8). (7)

(8)

a. Haifa's a very different city from Tel Aviv b. She has had a second child with *a very other man c. Foreign trade strengthens co-operation between nations, eases mutual understanding, makes the solution of political and *very other problems easier a. The city of Haifa's very different from Tel Aviv b. She has had a second child with *a man who is other c. Foreign trade strengthens co-operation between nations, eases mutual understanding, makes the solution of *problems that are political and other easier

The lexical uses hence behave as attributes or "central" adjectives, while the textual uses function as "peripheral" adjectives, as is prototypically the case for postdeterminers and classifiers (Quirk et al. 1972: 234; Bache 2000: 235-236). The previous paragraphs summarized the semantic analysis given in previous work, in which the concept of grammaticalization was invoked to account for the present polysemy of the English adjectives of comparison. In this article, I will investigate this grammaticalization hypothesis in more detail, focusing on the specific processes of grammaticalization and subjectification that have led to the polysemy of the adjectives of comparison. I will try and explicate in more detail the semantic processes that the adjectives have gone through and are still going through, causing the original lexical attribute uses to develop textual - postdeterminer and classifier uses. With regard to subjectification, I want to show that the semantic change affecting the adjectives of comparison can be analysed as involving both subjectification as defined by Traugott (1989, 1995) and, for some of the postdeterminer uses in particular, subjectification in Langacker's interpretation of the concept (Langacker 1990, 1998, 1999).

4. The gradability of the adjectives of comparison is however more limited than that of ordinary qualitative adjectives: they cannot be graded by comparative and superlative forms and only allow a restricted set of submodifiers.

246

Tine Breban

From a theoretical point of view, the investigation of these processes of grammaticalization and subjectification is not only interesting because, as already indicated, it charts relatively new territory, viz. adjectives, but also because it takes up areas of grammaticalization and subjectification theory that fall somewhat outside the scope of the current research on grammaticalization and subjectification. Firstly, within the framework of grammaticalization and subjectification theory established by Traugott (1988, 1989, 1995, 2003a; Traugott and König 1991; Traugott and Dasher 2002), the process of semantic change involved in the development of textual uses of adjectives of comparison instantiates the general path from propositional to textual uses. But this particular path of change has only been investigated in the context of 'discourse markers' in the narrow sense, which display processes of subjectification involving the gain of meanings expressing speaker attitude. As I will show, the postdeterminer and classifier uses of the adjectives of comparison display another type of subjectification, viz. the development of 'text-creating' meanings, which were subsumed under subjectification in Traugott (1995). Then, there is the work on subjectification in Langacker's sense. So far, attention in the literature has been focused on prepositions (Langacker 1990, 1998, 1999) and especially (modal) verbs (Langacker 1990, 1991, 1998, 1999, 2003; Pelyvas 2001; Brisard 2002b; Mortelmans 2002), i.e. elements which take part in the grounding of relations rather than entities.5 In his own work, Langacker (1990, 1998, 1999, 2002a, 2004) includes the development of nominal grounding elements in the discussion of subjectification as a concept and remarks on the subjectified status of determiners in the broad sense. I will interpret the shift from attribute to phoric postdeterminer use of comparative adjectives as a process of subjectification in the Langackerian sense, thus exploring a specific case of this process. The argument presented in this article will be developed in the following way. In the next section, I will show in what way the postdeterminer and classifier uses are the result of a process of semantic change and more specifically of a process of grammaticalization affecting the attribute uses. In the third section, it will become clear that the semantic change also involves processes of subjectification in the sense of Traugott (1989, 1995), while Section 4 will focus on the phoric postdeterminer uses and claim that

5. Brisard (2002a) devotes equal attention to clausal and nominal grounding, but the contributions on nominal grounding, although some invoke Langacker's concept of subjectivity, do not deal with subjectification as a process.

English adjectives of general comparison

247

they also involve subjectification in the sense of Langacker (1990, 1998, 1999). In section 5, I will sum up the argument and assess to what extent the concepts of grammaticalization and subjectification are useful explanatory tools in the context of adjectives of comparison and, on a wider plane, of adjectives in general.

2.

Postdeterminer and classifier uses as grammaticalizations of attribute uses

As indicated in the previous section, the semantics of the adjectives of general comparison are in present-day English characterized by a polysemy between lexical and textual readings, which are linked up with different functional categories in the NP. In this section, I will interpret this synchronic polysemy as the result of a process of grammaticalization that the adjectives of comparison have gone through and are still going through. Besides semantic arguments, further evidence of a diachronic, formal and comparative nature will be given. From a semantic point of view, it is rather obvious that the attribute uses distinguish themselves from both the postdeterminer and classifier uses, and that there are important analogies between the latter two. The attribute uses express degrees of likeness, in terms of the sharing of features, while the postdeterminer and classifier uses express relations of identity or non-identity between different instances or subtypes. What seems to have happened is that the adjectives of general comparison, because they had the necessary conceptual potential, came to be used in the 'peripheral' functions of postdeterminer and classifier by a process of semantic extension, resulting in a semantic shift. In my view, the relations of identity and non-identity expressed by the postdeterminers and classifiers are more schematic interpretations of the concept of likeness as it is expressed by the attributes. When they come to be used as postdeterminers or classifiers, the adjectives of comparison cease to indicate whether an instance or subtype is like or unlike another one, and how much it is like or unlike this other one. They go through a semantic shift in which the original meaning of gradable likeness is replaced by an inference often associated with this use. For adjectives of difference, this inference is that instances which do not share any qualitative features are ipso facto distinct instances. For adjectives of identity, the relevant inference is that instances that share all qualitative features are likely to be the same instance. As a

248

Tine Breban

result of this semantic shift, the new uses of the adjectives of comparison only retain the more fundamental idea that the instance or subtype is the same or not the same one as another instance or subtype that is activated in the mind of the hearer. Together with the evolution from semantically specific to more abstract meanings, the semantics of the adjectives at the same time become more grammatical or functional. Depending on the specific formal context in which the adjectives occur, two different types of grammatical content can be distinguished. In this way the adjectives of comparison illustrate the point made in recent contributions to grammaticalization theory that grammaticalization does not affect lexemes in se, but rather lexemes in specific contexts and constructions (cf. Heine 2003; Traugott 2003b). The crucial factor distinguishing the two contexts involved here is whether the comparison denoted by the adjective involves a separately expressed comparans and comparandum (i.e. the comparandum is expressed either by a different NP or by a prepositional phrase postmodifying the NP) or whether comparans and comparandum are referents of the same NP (i.e. comparans and comparandum are denoted by the same head noun). These distinct structural environments are represented in figure 1.

In the context of a separately expressed comparans and comparandum, the adjectives display a shift from expressing lexical comparison to indicating text-cohesive or 'phoric' relations between NPs. More specifically, they indicate relations of identity or non-identity between the instances or subtypes denoted by the distinct NPs. Examples (9) and (10) illustrate this text-cohesive use. In (9), same expresses that the specific city referred to can be identified as New Orleans, the city that has already been talked about in the discourse. Same hence signals co-referentiality with the comparandum, or anaphoric reference to 'the same' discourse referent. Another in (10), by contrast, conveys that the boy referred to is not the boy that was

English adjectives of general comparison

249

mentioned earlier. Another hence indicates that the NP denotes 'a different' discourse referent than the comparandum. (9)

(10)

The Pearl Archery Club was organized by women in New Orleans. It was not the first such organization, for the Crescent City Female Archery Club had already been founded in the same city during the 1870s. (CB) A boy aged 14 drowned in a canal yesterday after being chased from a stolen car. David Boyle, of Newcastle upon Tyne, jumped into the freezing water at Ulverston, Cumbria, while being chased by a member of the public. Another boy was able to swim to the bank and was taken to hospital. The teenagers had been disturbed while inside a Vauxhall Astra in a car park adjacent to the canal. (CB)

In the context in which comparans and comparandum are denoted by the same NP, the lexical semantics shift to a general reference-supporting function. The grammaticalized adjectives provide information concerning the internal make-up of the instantial set denoted by the NP. Adjectives expressing relations of non-identity convey that the NP refers to more than one instance. Different in (11), for instance, emphasizes that several instances of 'lead singers' are referred to. Adjectives expressing relations of identity, on the other hand, signal that one instantial set is associated with different circumstances. In (12), same expresses that one night is associated with different events, viz. several different balls. (11)

(12)

Tim Simenon is Bomb The Bass but he uses different lead singers on his records as he likes to maintain a shifting pool of talent. (CB) This year for the first time, the Red Cross is staging all its big Desperate and Dateless Balls on the same night, which is expected to draw a total of25,000 merrymakers nationally. (CB)

The specific grammaticalization process involved in the development from attribute uses to these two types of textual postdeterminer and classifier uses clearly instantiates one of the patterns of grammaticalization that were recognized by Traugott (1988, 1989), viz. the change from a propositional to a textual, text-cohesive meaning or "the shift ... from meanings grounded in more or less objectively identifiable extralinguistic situations

250

Tine Breban

to meanings grounded in text-making (for example connectives, anaphoric markers, etc.)" (Traugott and König 1991: 189).6 That is to say, the original attribute uses describe a qualitative relation in the extralinguistic world, whereas the postdeterminer and classifier uses function either as phoric markers helping the hearer to identify and keep track of the different entities involved in the discourse or as general determining elements clarifying reference. Either way, they contribute to the creation of 'text'. The argumentation so far has largely been based on present-day English language material, 7 and therefore constitutes only a synchronic approach to the grammaticalization process. However, various uses with different possible readings in the synchronic material can help us give an analytical description of the diachronic process of grammaticalization resulting in the development of postdeterminer and classifier uses. The following description sketches the different steps in the grammaticalization of the adjective different from fully lexical item to phoric postdeterminer and the further grammaticalization of the adjective other from phoric postdeterminer to phoric classifier. Schematic representations, following the conventions of Langacker (1987, 1991) are added to clarify the specific processes involved. Originally, the semantics of different are fully lexical. As represented in figure 2, different, when used in this lexical qualitative meaning, expresses that two referents share no or few qualitative features or, in other words, that the two referents are characterized by distinct sets of qualitative features. As attribute in a NP, therefore, different conveys that the instance of a certain type denoted by the NP (i2) has no or few qualitative features in common with another instance of that type mentioned earlier in the discourse (ij), cf. figure 3.

6. The formulation of the specific path of grammaticalization given here, which is taken from Traugott (1988, 1989) and Traugott and König (1991), is, in my opinion, consistent with the revised version of the distinct paths of grammaticalization and in particular their ordering which is assumed in Traugott's more recent work (cf. Traugott 1995). 7. The material on which the present analysis is based was extracted from the COBUILD Corpus as part of the corpus study on adjectives of general comparison described in Breban (2002) and Breban and Davidse (2003).

English adjectives of general comparison

251

Quality domain

Domain of instantiation

Figure 2. Qualitative meaning of different (two instances sharing no qualitative features) (based on Langacker 1991: 80) Type specification plane

Quality domain

Domain of instantiation Spatial instantiation plane Figure 3. Attribute meaning of different (two instances of a certain type Τ sharing no qualitative features)

252

Tine Breban

The trigger for the development of the text-cohesive postdeterminer meaning then involves, as is shown in figure 4, a shift in focus: whereas the original meaning profiles the different sets of qualitative features that characterize the two instances, the new meaning profiles the fact that the two instances are two distinct instances of the same general type. Type specification plane

Quality domain

Domain of instantiation Spatial instantiation plane Figure 4. Change in focus from attribute to postdeterminer meaning of different Figure 5 illustrates the resultant postdeterminer meaning, which has the schematic meaning of indicating that the instance denoted by the NP (i2) is another instance of the same type instantiated earlier in the discourse. For the adjective other, which has the same phoric postdeterminer meaning as different, the COBUILD corpus contains examples of a second grammaticalized phoric use, the phoric classifier use as illustrated in figure 6. This phoric classifier meaning seems to be the result of a process of analogy. By analogy with the postdeterminer use, which expresses that an instance is a different one from a previously mentioned instance, the classifier use expresses the same phoric relation but with respect to subtypes rather than instances. Thus, the phoric pattern realized by the postdeterminer use is copied onto a different level of organization of the NP, the

English adjectives of general comparison

253

type specification plane, which covers the relations between types and subtypes. The representation of the phoric classifier meaning of other, figure 6, hence conveys that the instances denoted by the N P belong to another subtype than other instances of the same type that were already present in the discourse.

Figure 5. Postdeterminer meaning of different (another instance of the same type T)

Figure 6. Classifier meaning of other (the instances referred to by the NP belong to another subtype than the other instances of the same Τ that have already been mentioned in the discourse) In the previous paragraphs, I have given the semantic reasoning behind the claim that the postdeterminer and classifier uses of the adjectives of com-

254

Tine Breban

parison are grammaticalizations of the corresponding attribute uses. In the remainder of this section, I will show that this grammaticalization analysis is further corroborated by various kinds of evidence. A first, semantic, piece of evidence for the grammaticalization analysis of the different uses of the adjectives of comparison is the occurrence of examples which Evans and Wilkins (2000) refer to as 'bridging contexts'. 8 These are examples in which the adjective can be interpreted in more than one meaning, each of which is supported by elements from the context. In other words, one form is associated with two or more distinct meanings both of which make, in their own way, sense in the particular context. This is, of course, made possible by the fact that the meanings in question share some significant semantic features, while at the same time departing from one another by the loss or addition of features. Evans and Wilkins view 'bridging contexts' as a stage preceding polysemy, and hence as an indicator of ongoing semantic change. An example of such a bridging context encountered for the adjectives of comparison is (13). (13)

Prince Saud declined to mention Yemen by name, but referring to the catastrophe of Iraqi aggression, he said: One of the saddest elements of the crisis was that there were voices in the Arab world trying to justify the premise that Arabs lived by different standards from the rest of the international community. (CB)

In this example, the context allows either the attribute or the postdeterminer meaning to be the profiled meaning of different. In the former case, the profiled meaning of different standards is that of referring to qualitatively different standards or standards that are characterized by other qualitative features. This interpretation is present in a veiled way because it carries the politically incorrect implication of a value judgement with respect to the standards of the Arab world versus the non-Arab community. The postdeterminer meaning allows the speaker to maintain the appearance of political correctness as its foregrounded meaning is that of other rather than the same standards, without attributing any qualitative values to the different sets of standards.9 Examples of this kind confirm the semantic 8. A similar concept, viz. the "overlap stage" or "bridging stage", is used by Heine in his overlap model (e.g. Heine 2003: 589-590). 9. If a submodifier is added (i) or the adjective is changed into predicative position (ii), only the former, lexical reading of different is available.

English adjectives of general comparison

255

shift sketched above, viz. that the inference that instances which are characterized by different sets of qualitative features are ipso facto distinct instances, which can be referred to separately, is the basis for the development of a new reading for the adjective different. As the corpus studies in Breban (2002/2003) and Breban and Davidse (2003) showed, these bridging contexts are relatively frequent and are found with all currently polysemous adjectives. Secondly, the grammaticalization analysis is confirmed by diachronic information concerning the different adjectives of comparison available in dictionaries such as the Oxford English Dictionary (OED). The OED indicates that the grammaticalized meanings of the adjectives of comparison are generally more recent than their attribute meaning, and that for some of the adjectives the attribute meaning was gradually lost. Other, for example, displayed the two uses from the earliest stages mentioned in the OED, but the attribute meaning is no longer available in present-day English (OED Vol. 7: 229). With respect to different, the OED states that the lexical item was borrowed from French around about 1400 expressing the attribute meaning. The first example cited of a postdeterminer reading dates from 1651 (OED Vol. 3: 229). In the third place, some of the adjectives of comparison also show certain formal characteristics associated with the process of grammaticalization. The two adjectives that have grammaticalized furthest, same and other (the quantitative corpus study presented in Breban 2002 has shown them to be semantically fully grammaticalized), display the formal characteristic of bonding or "coalescence" (Lehmann 1985: 308). As noted in the OED (Vol. 9: 74), same has from its earliest citings on always been used in combination with the definite article, or occasionally a demonstrative, and they functionally form one fixed chunk. The postdeterminer use of other takes the process of bonding a step further: the indefinite article and the adjective have become fused into one word form, another.10

(i) .. .the premise that Arabs lived by very different standards from the rest of the international community (ii) ...the premise that Arabs lived by standards that are different from those of the rest of the international community 10.1 thank one of the anonymous referees for pointing out that another manifests further grammaticalization into the pronoun one another. Another here no longer functions as a determiner but as a reciprocal marker. Similar processes of grammaticalization resulted in the equivalent combination each other, and af-

256

Tine Breban

Table 1.

Current degree of grammaticalization in English and Dutch (quantitative results of the corpus study presented in Breban 2002/2003)

English

Grammaticalized uses

Attribute uses

Bridging contexts

Same Identical Other Different

100% 30.5% 100% 68.5%

0% 68% 0% 22%

0% 1.5% 0% 9.5%

Dutch

Grammaticalized uses

Attribute uses

Bridging contexts

Zelfde Identiek Ander Verschillend Verscheiden

100% 33% 87.5% 93% 97%

0% 62.5% 4.5% 4% 3%

0% 4.5% 8% 3% 0%

In the fourth place, the grammaticalization hypothesis receives further support from comparative evidence. That is to say, the grammaticalization process displayed by the adjectives of comparison is not limited to the English adjectives, but also characterizes the different uses of their Dutch counterparts. In Breban (2002/2003), I argue that the Dutch adjectives zelfde,

identiek,

ander,

verschillend

and verscheiden

have been going

through the same general process of grammaticalization, leading from lexical attribute uses to grammaticalized postdeterminer and classifier uses, as their English equivalents same, identical, other and different respectively. The - more subtle - semantic differences between the English and Dutch adjectives pertain on the one hand to a difference in current degree of grammaticalization (see table 1 above) and on the other to the different options that are chosen along the general path of grammaticalization. One such example of different grammaticalization options is, for instance, the absence of the text-cohesive postdeterminer meaning for the Dutch adjectives verschillend and verscheiden in contrast to their English counterpart different. English different has both the text-cohesive postdeterminer meaning, which indicates that reference is made to a different instantial set, as illustrated in (14), and the second general referential fected other related adjectives such as self, which combined with the personal pronouns to form emphatic markers such as myself, etc.

English adjectives of general comparison

257

meaning introduced above. In examples of this second type, e.g. (15), different does not construe a phoric link with other instances of the same type in the discourse, but simply conveys that the instantial set referred to by the NP consists of several distinct instances. (14)

(15)

The benefits of housing segregation are obvious in the case of groups who speak a different language from that of the surrounding society. (CB) A United Nations official leading a team of experts investigating Iraq's nuclear capability says that during visits to different sites in Iraq today the authorities have been very co-operative. (CB)

For the Dutch adjectives verschillend and verscheiden, only this second, general referential postdeterminer reading is possible. In (16), for instance, verschillend expresses that several organs and functions of the human body can be seriously damaged by lead pollution. (16)

Over de gevolgen van loodverontreiniging voor de mens is al veel geschreven. Lood werkt schadelijk op verschillende organen en functies en kan ook de voorplanting nadelig be'invloeden. (INL) Much has been written about the consequences of lead pollution for man. Lead can harm different organs and functions and it can also negatively influence procreation.

Notwithstanding these fine-grained semantic differences, one cannot fail to remark the obvious parallel development demonstrated by the Dutch adjectives, which confirms the explanatory power of the grammaticalization interpretation. A final piece of evidence concerns the specific path of grammaticalization from lexical to textual uses discussed here. Inherent in the semantic categorizations of these two types of uses, the process of grammaticalization involves a change in semantic organization of the field of comparison: a change from a semantic continuum, ranging from the sharing of all features, over many, few, to no shared features (see figure 7), to a strict binary semantic opposition, identity or non-identity. The development of the adjectives of identity and difference is straightforward in this respect, they naturally grammaticalize into postdeterminers and classifiers indicating relations of identity and non-identity respectively. But what happens to the third group of adjectives of comparison, the ad-

258

Tine Breban

jectives of similarity, which, as figure 7 shows, complete the semantic continuum of the sharing of features? In Breban and Davidse (2003), it was demonstrated that the process of grammaticalization caused the intermediate semantic area of similarity to be divided into grammaticalized uses that instantiate the semantic schema of identity (uses expressing relations of identity) and others that instantiate the schema of non-identity (uses expressing relations of non-identity). For the classifier uses of the adjectives of similarity, for instance, the following two classifier uses, instantiating the two different schemata, can be distinguished. In examples such as (17), the classifier, in this case comparable, signals that the subtype of 'Whites' involved is the same subtype that was derived from the general type 'African Americans' earlier in the discourse, viz. 'middle-class or affluent'. In examples such as (18), by contrast, the classifier, similar, functions in the same way as the classifier other, it indicates that next to 'greenfood' other types of 'fresh foods' are being referred to. shared qualitative features all

many

few

none

identity

similarity

difference

difference

adjectives of comparison

Figure 7. The lexical fields expressed by adjectives of comparison: continuum of shared qualitative features (17)

(18)

Indeed, studies, even with children, show that when the selfimages of middle-class or affluent African Americans are measured, their feelings of self-esteem are more positive than those of comparable Whites. (CB) The most common condition encountered is Vitamin A deficiency. This is because this vitamin is only present in seed at low levels: much richer sources are present in greenfood and similar fresh foods. (CB)

Breban and Davidse (2003) further indicate that these two kinds of uses not only apply distinct semantic schemata, but also involve distinct lexicogrammatical configurations, which again illustrates the theoretical claim that grammaticalization affects lexical items as part of specific construc-

English adjectives of general comparison

259

tions. In examples such as (17), in which the classifier derives 'the same' subtype, the phoric relation typically connects two general types that are of the same level of generality, viz. 'African Americans' and 'Whites'. The other classifier use, which derives 'a different' subtype, always involves a phoric relation between a NP consisting of a head noun designating a more specific type ('greenfood') and the NP with the similarity classifier + a more general 'supertype' ('fresh foods'), which are syntactically coordinated (in this example by and). These two types of classifier uses can be referred to as the 'subclassification' and the 'superclassification' pattern respectively. Thus, the grammaticalization process manifested by the adjectives of similarity confirms the specific path of grammaticalization from lexical uses situated on a gradable scale with three lexicalized areas (identity, similarity and difference) to grammaticalized uses dealing with either identity or non-identity without a meaning area in between the two. So, we can conclude that the proposed grammaticalization analysis not only captures the semantic polysemy of the adjectives of comparison, but also takes into account the attested diachronic development of the different adjectives and offers an explanation for certain formal characteristics of the adjectives and for the occurrence of equivalent adjectives with similar semantic patterns in related languages such as Dutch.

3.

Postdeterminer and classifier uses as subjectifications of attribute uses: Traugott's subjectification interpretation

In this section, I will argue that the process of grammaticalization displayed by the adjectives of comparison goes together with an increase in subjectivity as defined by Traugott (1989, 1995). The concept of subjectification that I will use here, is the broad interpretation of subjectification found in Traugott (1995), viz. the development of any meaning that can be characterized as 'speaker-involved' or "speaker-based" (Traugott 1995: 32). In her later articles (Traugott 1999, 2003a; Traugott and Dasher 2002), Traugott puts forward a more restricted interpretation of subjectification, which only covers the development of meanings expressing speaker attitude. Traugott (1995) defines subjectification in the context of grammaticalization as "the tendency to recruit lexical material for purposes of creating text and indicating attitudes in discourse situations" (Traugott 1995: 47). This definition of subjectification not only covers the development of

260

Tine Breban

meanings expressing speaker attitude, but also includes a second type of subjectification, viz. the development of meanings with which the speaker "creates text". The dual character of Traugott's (1995) definition of subjectification is also emphasized by Finegan (1995). He points out that these two types are the natural consequence of two fundamental "needs" at the basis of subjectification, viz. the "cognitive need" to increase informativeness, and the "social need" to express speaker attitudes (Finegan 1995: 9).11 The two types of subjectification subsumed under this broad definition involve two types of 'additional', non-propositional meaning, both of which can be characterized as more 'speaker-involved'. The first type of subjective meaning, which ties in with the use of the opposition subjectiveobjective in ordinary, non-linguistic language, covers meanings that add some expression of the ' self of the speaker to the propositional content. The second type refers to meanings concerned with the formulation of propositional content in context, both the extra-verbal context of the speech event and the verbal discourse context. These meanings are speaker-involved in the sense that it is the speaker who decides how to link the propositional content with the context.12 Traugott's later interpretation 11. The choice of the two terms with which Finegan refers to the two needs is somewhat unfortunate. The term "cognitive", as pointed out by one of the anonymous referees, is too general and can better be seen as an overarching principle. The second term, "social" need, is better avoided in the light of later articles on subjectification, in which 'intersubjectification', a possible subsequent development of subjectification, is explained as the "process whereby meanings come over time to encode or externalise implicatures regarding SP(eaker)/W(riter)'s attention to the 'self of AD(dressee)/R(eader) in both an epistemic and a social sense" (Traugott 2003a: 129-130), and is hence characterized as "interpersonal" (Traugott and Dasher 2002: 23) or social in nature. 12. It should be noted, however, that the distinction between the two types of subjectified meaning is not always clear-cut. Some examples of connectives and discourse markers, for instance, display both types of subjectification. Anyway, for instance, not only directs the flow of the discourse, but also signals that the speaker wants to elaborate on or justify what has been said (Traugott 2003b: 641). Traugott herself (2003b: 633) refers to this issue when she indicates that the term 'textual' in the original 1982 formulation of the grammaticalization cline propositional (> textual) > expressive meaning is "ambiguous between: (i) (what was originally intended) the development of meanings signalling cohesion, especially intra-clausal truth-conditional connections made by the same speaker in the same turn;

English adjectives of general comparison

261

of subjectification (Traugott 1999, 2003a; Traugott and Dasher 2002) seems restricted to the attitudinal type of subjective meaning. Subjectification is characterized as the process that causes lexical items to "develop polysemies that are grounded in the speaker's world, whether reasoning, belief, or metatextual attitude to the discourse" (Traugott 1999: 179).13 Traugott's broad interpretation of subjectification, consequently, covers a number of phenomena which are excluded from her later, more narrowly defined, concept of subjectification. In Traugott (1995), for example, the development of be going to from a construction expressing motion to a quasi-auxiliary expressing near future (Traugott 1995: 34-36) is one of the main examples illustrating subjectification. With respect to the topic at hand in this article, it is instructive to note that the developments of both the definite and the indefinite article, the from the demonstrative that and a from the numeral one, are also mentioned as involving subjectification (Traugott 1995: 39). Matters are further complicated because the two interpretations of subjectification are not always clearly distinguished in the literature. This can be blamed partly on the ambiguity of the metalanguage, as the same terms, such as "speaker's perspective", and "speaker's point of view", are often used in the definitions of both interpretations. The process of grammaticalization displayed by the adjectives of comparison can only be characterized as subjectification in the broad sense, as it involves the text-creating type of subjectification. As described above, the original lexical attribute meaning of the adjectives either develops into postdeterminer and classifier meanings which by means of phoric relations of identity/non-identity help the hearer identify and keep track of the different instances or subtypes in the discourse or into grammaticalized uses that clarify the internal make-up of the instantial set denoted by the NP. Both grammaticalized uses involve the speaker's taking control of the or(ii) (what later came to be of focal interest among practitioners of pragmatics and discourse analysis) the development of meanings signalling strategic interaction." (Traugott 2003b: 633) 13. In her latest articles (Traugott 2003b, 2004), Traugott no longer simply refers to the speaker as locus of subjectification, but rather to the SP(eaker)W(riter)/ AD(dressee)R(eader)-dyad (2004: 550) or "the cognitive and communicative pragmatics of speaker-hearer interactions and discourse practices" (2003b: 634). It would hence be more straightforward to include the postdeterminer and classifier uses of the adjectives of comparison, which crucially negotiate speaker-hearer interaction, into the present approach to subjectification.

262

Tine Breban

ganization of discourse in order to reduce the processing effort of the hearer. They clearly fit in with the definition which Traugott (1995: 45) gives of stance adverbs, degree modifiers, etc. which, she says, "are subjective in that they are the devices by which speakers take responsibility for success in communication and seek to meet the hearers' attempts 'to integrate new information with information that is already accessible' (Blakemore 1990: 364)". Having pointed out Traugott's change of mind with regard to her definition of subjectification, I would like to indicate my own position with regard to this issue. I see three arguments in favour of the broad subjectification interpretation and hence the inclusion of the second type of subjectification. The first argument concerns the fact that even in the more recent articles, which deal with subjectification in the restricted sense, examples are included that can only be explained as subjectified in the broad sense. In Traugott (2003a: 125), for example, space and time deictics taking speaker/writer as reference point and performative speech acts are mentioned as obvious examples of subjective elements in language, although they are only speaker-involved in the sense that they relate the proposition explicitly to the speech event and its participants. Secondly, the association of subjectivity with the speech event and its circumstances was part of the foundational works on subjectivity. In the early works to recognize the role of subjectivity in language, such as Benveniste (1966), Lyons (1977, 1982), subjectivity is not only associated with the speaker as a person, although it can be (cf. Lyons' account of subjectivity in epistemic modality (1977)), but also with the speaker in his role of speaker, as part of the speech event. It is precisely the involvement of the speaker and the speech event which accounts for the subjectivity of personal pronouns, deictic elements, the expression of temporal relations, performative uses of verbs (cf. Benveniste 1966). In a metaphorical way, the postdeterminer and classifier uses can be argued to function as deictic elements within the text, as they crucially depend on the here-and-now of their occurrence within the text. The third argument concerns the development of the adjectives of comparison in particular. With respect to the occurrence of processes of subjectification in the NP, Adamson (2000) argues that subjectification goes

English adjectives of general comparison

263

together with a leftward movement in the NP.14 Only the broad subjectification interpretation allows us to maintain this generalization, as the more restricted definition does not include determiners and postdeterminers as subjective elements.15 In the light of the present volume, I would like to note that it is exactly the second, text-creating type of subjectification that is to a certain extent connected with subjectivity as defined by Langacker (1990, 1999). I will return to the relation between Traugott's and Langacker's concepts of subjectification in the conclusion.

4.

Phorie postdeterminer uses as subjeetifieations of attribute uses: Langacker's subjectification interpretation

Langacker's interpretation of subjectivity and subjectification (especially Langacker 1990, 1998, 1999) pertains to the status of an entity as subject or object of conception, whereby an entity is "construed objectively to the extent that it is distinct from the conceptualizer and is put onstage as a salient object of conception" (Langacker 2002a: 17), while an entity construed with maximal subjectivity remains implicit, "inhering in the process of conceptualization without being its target" (Langacker 1998: 71). Against the background of these basic assumptions, he defines subjectification as the attenuation of an original objectively construed relation, revealing a subjectively construed relation inherent in the process of conceptualization of the original objective relation.16 Langacker has illustrated his definition of subjectification mainly with modal verbs and prepositions. 14. The grammaticalized classifier uses seem to argue against the identification of subjectification in the NP with a leftward movement, but they are, as pointed out on the basis of figure 6, presumably the result of a process of analogy with respect to the grammaticalized and subjectified postdeterminer uses rather than the result of a process of subjectification in their own right. 15. Athanasiadou (this volume), for instance, analyses the adjectives in the NP on the basis of the restricted interpretation of subjectification, and hence only recognizes certain types of attributes, such as value adjectives, as subjective, while the postdeterminers are marked as objective. 16. The characterization of the concept of subjectification given here (cf. Langacker 1998 and later works) is the revised interpretation of the original characterization of subjectification as the realignment of an objectively construed relation to the subjective axis (cf. Langacker 1990).

264

Tine Breban

Here, I will discuss how subjectification manifests itself in the development of the adjectives of comparison. As will be shown, the process of subjectification involved here is more complex because it involves fusion of the adjective with the primary determiner of the NP that it occurs in. The attribute uses of the adjectives of comparison constitute an objective relation as they compare a profiled entity with another entity in terms of the number of shared features. Although this objective relation does not invoke the ground as a reference point, there is some subjective involvement on the part of speaker and hearer. That is to say, the conceptualization of the two instances involves their relation to the ground and as indicated in Langacker's articles on grounding (Langacker 2002a, 2002b, 2004), a grounding relation is necessarily subjectively construed (2002a: 11, 2004: 86). In these attribute examples, this grounding relation is expressed by the primary grounding elements, viz. articles and quantifiers. But, in contrast to other adjectives, the adjectives of comparison have an inherent element in their semantics which naturally fits in with this grounding relationship, viz. the semantics of identity/non-identity. This semantic element is potentially subjective because it is 'epistemic': it concerns the fundamental cognitive notion or "epistemic domain" (Langacker 2001: 166) of identification. When the objective relation based on the sharing of features becomes less salient and disappears, the adjective undergoes a semantic shift from expressing qualitative likeness to expressing exactly this relation of identity/non-identity. What happens then is that this epistemic, potentially subjective meaning is tied to the primary grounding relation conveyed by the grounding element of the NP to express a more complex grounding relation, which crucially involves the identity of the instance denoted by the NP in relation to other instances of the same type. More particularly, this new use of the adjectives of comparison fits in with the characterization of grounding as "a primary means of specifying how the content of an expression is to be integrated in the CDS (current discourse space) and connected to the conceptual structure already in place" (Langacker 2001: 167). The adjective, in this combination with a primary determiner, designates a (grounding) relationship that "resides exclusively in the conceptualizer's activity" (Langacker 1999: 153) and that is hence a subjectively construed relationship. Concretely, for the adjective different in an indefinite NP, this process of subjectification can be described as follows. When the objective attribute meaning (illustrated in figure 3) conveying that two instances are characterized by different sets of qualitative features, fades away, the gram-

English adjectives of general comparison

265

maticalized postdeterminer meaning (represented in figure 5) becomes the profiled meaning of different. But this grammaticalized meaning is not present as such, rather, it combines with the grounding relation expressed by the indefinite article (figure 8) into the subjective grounding meaning that is represented in 9. Following Davidse (2004), the semantics of the indefinite article can be characterized in terms of the 'correspondence relation' between type specification Τ and instantiation ip as represented by the arrow emerging from the ground in figure 8, the speaker instructs the hearer to conceptualize an instance as corresponding to the type specification.

Figure 8. Indefinite identification (based on Davidse 2004: 523) As shown in figure 9, which combines figures 5 and 8, the grammaticalized meaning of different, indicating a relation of non-identity with respect to another instance of the type, naturally links up with the correspondence relation expressed by the indefinite article. The combination a different then expresses that the hearer should be able to conjure up a new instance i2 as corresponding to a type Τ that has already been instantiated by other instances (represented here by ii) in the discourse. The adjective different, expressing this epistemic, potentially subjective meaning, has hence become a true 'secondary determiner' or postdeterminer. Crucial to this interpretation is that the adjectives of comparison, when they express the postdeterminer meaning, do not become grounding elements as such, but rather combine with the primary determiner to form one determiner or grounding complex. As already indicated above, the combination primary determiner + postdeterminer is, especially for the adjectives which have grammaticalized uses only, felt to be one fixed chunk, to the

266

Tine Breban

extent that one combination, another, even reflects this 'unity' in its orthography. For the combination the + other, the bonded nature of the two elements makes it possible to use the other as a sort of proform in contexts where two instances of one type are evoked. Moreover, this new grounding complex displays several of the special properties associated with grounding elements (cf. Langacker 2002a: lOf, 2002b: 29-30). As is characteristic of grounding predications, the determiner complex does not profile the ground or the grounding relationship, but only the grounded entity, viz. the instance referred to. Its semantics are, as illustrated by figure 9, schematic in nature and are concerned with a fundamental cognitive notion or "epistemic domain" (Langacker 2001: 166), viz. identification, in the sense of mental contact with the denoted instance.

Figure 9. Subjectively construed relationship denoted by a different (recognition of an instance as a new instance of a previously instantiated type) The semantic make-up of the grounding complex primary determiner + postdeterminer is in a certain way similar to that of demonstratives, which also have a dual semantic make-up signalling mental contact by speaker and hearer and making a further specification regarding proximity (Langacker 2002a: 10). (The notion 'proximity' not only covers proximity in concrete spatial terms, but also in 'metaphorical' terms, for example proximity in the discourse (cf. Langacker 2004: 98).) The determiner complex primary determiner + postdeterminer of comparison functions in a similar way, it indicates whether mental contact has been made or not while directing the hearer to previous mentions in the text in terms of identity and non-identity.

English adjectives of general comparison

267

The dual semantic value of the determiner complex results in three possible determining combinations, dependent on the type of primary determiner (definite or indefinite) and the type of phoric postdeterminer (displaying relations of identity or non-identity). The three resulting combinations are: 1. definite determiner + postdeterminer expressing identity (e.g. the same) 2. definite determiner + postdeterminer expressing non-identity (e.g. the other) 3. indefinite determiner + postdeterminer expressing non-identity (e.g. another) Let us first look at the combination definite determiner + postdeterminer expressing identity. The determiner complex as a whole signals that the designated entity is accessible for the hearer because he has already established mental contact with this specific instance in the current discourse space.17 In (19), for instance, the same expresses that the particular knee referred to is the knee that is talked about in the previous sentence. (19)

Ian Botham is to consult a specialist this week to find out whether he needs another knee operation. His knee became badly swollen during Saturday's Benson and Hedges Cup final in which Worcestershire were beaten by Lancashire but in which Botham was the top scorer with 38. He's already been sidelined for three weeks this season after an operation on the same knee. (CB)

Figure 10 gives a schematic representation of this grounding combination. The hearer can establish mental contact with the specific instance referred to, because it is i ls an instance that speaker and hearer had already established mental contact with, rather than a new instance, i2.18 17. As noted by Langacker (2001: 168), the definite article only signals that the referent of the NP is in some way already available, and not that it is previously mentioned or that coordinated mental contact with it has already been established. The latter meaning is hence added by the postdeterminer of identity. 18. In Langacker (2004), the difference between the definite article and the demonstratives is explained on the basis of the different number of possible candidates for the identification of the instance denoted by the NP: for the definite article, there is "only one instance of the specified type within the relevant scope of consideration" (Langacker 2004: 99), while demonstratives also need "other

268

Tine Breban

Figure 10. Definite determiner + postdeterminer expressing identity

The second determiner complex, definite determiner + postdeterminer expressing non-identity, invokes the second natural reference point besides the ground, viz. a mass consisting of all the instances of the type Τ available in the present discourse context (Langacker 1991: 91), the instantial mass M T . 19 It indicates that the hearer can establish mental contact with the designated entity because it is that instance of the instantial mass that has not yet been introduced into the discourse itself. For example, the other in (20) signals that the particular instance of the type 'semi-final' can be identified because it is that one of the two semi-finals that has not been discussed yet. (20)

In today's men's semis, Frenchman Cedric Pioline will be carrying home hopes as he attempts to prevent a Spanish whitewash. Pioline, who reached the Wimbledon semi-finals last year after beating Tim Henman in the last eight, faces clay court specialist Alex Corretja. The other semi-final is an all-Spanish clash between Carlos Moya and Felix Mantilla. (CB)

potential targets" (Langacker 2004: 97). The determiner complex definite determiner + postdeterminer of identity is again similar to the demonstratives; as shown in figure 10, the same conjures up a second possible target. 19.1 prefer to use the term 'instantial mass' to Langacker's term "reference mass" (Langacker 1991: 91), because the latter, although more generally used, intuitively evokes the concept of relative quantification. In addition, M T should also be distinguished from Langacker's more recent interpretation of the concept 'reference mass', the "maximal extension of a type", "E t " (Langacker 2004: 84). Crucially, M T is a dynamic concept: it is not a predetermined mass, but one that changes as the discourse is proceeding.

English adjectives of general comparison

269

The representation of this determiner combination in figure 11 shows that the hearer can establish mental contact with the instance denoted by the NP, i2, because it is the remaining instance of the instantial mass M T .

Figure 11. Definite determiner + postdeterminer expressing non-identity The third and last combination is indefinite determiner + postdeterminer of non-identity. Here the determiner complex conveys that the hearer has not established mental contact with the specific instance in the current discourse space. But, as discussed above for different, he can recognize it as an instance of the type Τ which is identifiable (cf. Gundel et al. 1993: 276), in this case, because it is accessible in the current discourse space by means of other instantiations. So, the determiner complex indicates that the type to which the new instance belongs has already been instantiated in the discourse. The relation established by the postdeterminer can hence be interpreted as one of 'type-anaphora' (Davidse 2001: 21). The determiner complex another in (21), for instance, signals that a new instance of the type 'Australian', distinct from the previously mentioned instance of the type, 'the Australian Michelle Martin', is referred to. (21)

New Zealand's world number one SQUASH player, Susan Devoy, dropped a game before beating the Australian, Michelle Martin in the quarter finals of the Hong Kong Open. She now meets the third seed, Lisa Opie of Britain, who put out another Australian, Robyn Lambourne. (CB)

As represented by figure 12, the determiner complex consisting of indefinite determiner + postdeterminer of non-identity expresses that the hearer

270

Tine Breban

has not yet established mental contact with i2, the specific instance referred to, but that he can recognize it as an instance of the type T, because previously mental contact has been made with a different instance of the same type, viz. i].

We can conclude that, besides processes of grammaticalization and subjectification in Traugott's sense, the phoric postdeterminer uses of the adjectives of comparison also display subjectification in Langacker's (1990, 1998, 1999) sense. This causes the postdeterminer uses to function as part of the determiner complex, expressing a grounding relation of the instance in terms of its relation of identity/non-identity with other instances in the discourse.

5.

Conclusion

In this article, I have invoked the grammaticalization framework and the notions of subjectification as defined by Traugott and Langacker to look at three functions of prenominal adjectives which had not been looked at in terms of grammaticalization hitherto. More specifically, I have applied these notions to the semantic polysemy between the lexical readings and the textual readings that characterizes the adjectives of general comparison in present-day English. I have argued that their textual uses, which comprise postdeterminer and classifier uses, can be explained as the result of concomitant processes of grammaticalization and subjectification of the lexical, attribute uses.

English adjectives of general comparison

271

In the first part of the article, the postdeterminer and classifier readings were argued to be the result of grammaticalization of the 'original' attribute reading. I identified the semantic change they went through with one of the patterns of grammaticalization recognized by Traugott (1988, 1989, 1995; Traugott and König 1991), viz. the development of lexical, propositional meanings into grammatical, textual meanings. To further enforce the argumentation, the grammaticalization analysis was then shown to accommodate various diachronic, formal and comparative properties of the adjectives of comparison. Besides a process of grammaticalization, the semantic change from attribute to postdeterminer and classifier meaning was also claimed to entail processes of subjectification. The increase in subjectivity displayed by the adjectives of comparison fits in with the two main interpretations of subjectification in the literature. On the one hand, it can be analysed as a process of subjectification as defined by Traugott in her earlier articles (Traugott 1989, 1995), viz. as meanings gaining in speaker-involvement. In the case of the adjectives of comparison, the original lexical attribute meaning has evolved into a means for the speaker to create and organize discourse and in this way reduce the processing effort of the hearer. On the other hand, the development of phoric postdeterminer meanings also instantiates subjectification in the specific sense defined by Langacker (1990, 1998, 1999). That is to say, the development of the phoric postdeterminer use goes together with the fading away of the objectively construed relation of (un)likeness between instances, leaving behind the epistemic, potentially subjective meaning of identity/non-identity. The subjective potential of the adjective is realized in the grounding complex which it forms together with the primary determiner and which expresses a subjectively construed relation inherent in the conceptualization process of the NP, or more specifically a grounding relation. On a more general level, it has again become clear that the concepts of subjectivity and subjectification can and have received many different interpretations. This article has specifically focused on the two main interpretations in current linguistic theory, Traugott's and Langacker's, which can be characterized as fundamentally different. The core of the difference between the two approaches to subjectivity lies, in my opinion, in the different interpretations of the notion of 'subject' underlying the concepts of subjectivity. For Traugott, the subject can be identified as the 'speaker', while for Langacker it is the 'conceptualizer'. Although this amounts in both approaches to the same 'person behind the utterance', the role of this

272

Tine Breban

person is substantially different: for Traugott, the activity of the subject is to 'express' (and in second place, to organize what he/she expresses), whereas for Langacker, the subject 'construes'. This is also what Langacker (this volume) explains when he calls subjectification as defined by Traugott "a matter of conceptual content", while his own account is referred to as "a matter of construal". Although the approaches are thus different with respect to the most basic assumptions, they can sometimes be applied to the same lexical items or constructions. To a certain extent, this can, in my opinion, be accounted for in terms of Langacker's concept of 'grounding', which plays an important role in many of the instances of subjectification that he analyses. For instance, when he discusses the development of new auxiliaries and new determiners, the role of the conceptualizer as part of the ground is crucial to the subjective construal, and the conceptualizer is as part of the ground more explicitly identified with the speaker. This is why developments such as that displayed by the postdeterminer uses of the adjectives of comparison can be interpreted as manifesting subjectification in both interpretations. But even when the two interpretations of subjectification can be applied to the same lexical item or construction, they constitute opposed perspectives on the same phenomenon. When in Langacker's terms the construal becomes more subjective, the conceptualizer, i.e. the ground, necessarily becomes more objective (Langacker, this volume), and it is exactly this process of 'objectification' that characterizes subjectification in Traugott's sense, according to which the speaker becomes more visible in the construction. In conclusion to this investigation, we can say that analysing the postdeterminer and classifier uses in terms of grammaticalization and subjectification of the original attribute uses provides a coherent explanation for the current polysemy of the adjectives of general comparison. It not only accounts for the semantic make-up of the adjectives, but also chimes in with their formal, diachronic and comparative properties. From a functional perspective, Langacker's subjectification analysis in particular enhances our understanding of the origin and the functioning of the postdeterminer uses as 'secondary' determiners. In the domain of adjectival functions in general, it is, in my opinion, to be expected that processes of grammaticalization will prove to be involved in the development of all postdeterminers, as they, in contrast to classifiers, have a typical 'grammatical' function. How precisely adjectives belonging to other semantic

English adjectives of general comparison

273

domains than general comparison shift to postdeterminer or classifier function will, of course, have to be established in future case studies.

References Adamson, Sylvia 2000 A lovely little example. Word order options and category shift in the premodifying string. In Olga Fischer, Anette Rosenbach and Dieter Stein (eds.), Pathways of Change. Grammaticalization in English, 39-66. Amsterdam/Philadelphia: John Benjamins. Bache, Carl 2000 Essentials of Mastering English. A Concise Grammar. Berlin/ New York: Mouton de Gruyter. Benveniste, Emile 1966 De la subjectivite dans le langage. In Emile Benveniste (ed.), Problemes de Linguistique Generale, 258-266. Paris: Gallimard. Bolinger, Dwight 1967 Adjectives in English: Attribution and predication. Lingua 18: 1 34. Breban, Tine 2002 Adjectives of comparison: Postdeterminer, epithet and classifier uses. Unpublished M.A. Thesis. Department of Linguistics. University of Leuven. 2002/03 The grammaticalization of the adjectives of identity and difference in English and Dutch. Languages in Contrast 4-1: 167-201. Breban, Tine and Kristin Davidse 2003 Adjectives of comparison: The grammaticalization of their attribute uses into postdeterminer and classifier uses. Folia Linguistica XXXVII-3/4: 269-317. Brems, Lieselotte 2003 a Measure noun constructions: An instance of semantically-driven grammaticalization. International Journal of Corpus Linguistics 8-2: 283-312. 2003 b Measure noun constructions: Degrees of grammaticalization and delexicalization. In Karin Aijmer and Bengt Altenberg (eds.), Advances in Corpus Linguistics. Papers from the 22nd International Conference on English Language Research on Computerized Corpora (ICAME 22) Göteborg 22-26 May 2002, 249-265. Amsterdam: Rodopi.

274

Tine Breban

Brisard, Frank (ed.) 2002 a Grounding: The Epistemic Footing of Deixis and Reference. Berlin/New York: Mouton de Gruyter. 2002 b The English present. In Frank Brisard (ed.), Grounding: The Epistemic Footing of Deixis and Reference, 251-298. Berlin/New York: Mouton de Gruyter. Davidse, Kristin 2001 Postdeterminers: Their Secondary Identifying and Quantifying Functions. Preprint nr. 177. Department of Linguistics. University of Leuven. 2004 The interaction of identification and quantification in English determiners. In Michel Achard and Suzanne Kemmer (eds.), Language, Culture and Mind, 507-533. Stanford: CSLI Publications. Denison, David 2002 "History of the Sort of Construction Family". Paper presented at ICCG2 (Second International Conference on Construction Grammar), Helsinki 7 September 2002. Available on the WWW: , accessed November 2003. Evans, Nicholas and David Wilkins 2000 In the mind's ear: The semantic extensions of perception verbs in Australian languages. Language 76: 546-592. Finegan, Edward 1995 Subjectivity and subjectivisation: An introduction. In Dieter Stein and Susan Wright (eds.), Subjectivity and Subjectivisation: Linguistic Perspectives, 1-15. Cambridge: Cambridge University Press. Gundel, Jeanette, Nancy Hedberg and Ron Zacharski 1993 Cognitive status and the form of referring expressions in discourse. Language 69: 274-307. Halliday, Μ. A. K. 1994 An Introduction to Functional Grammar. 2nd edition. London: Arnold. Halliday, Μ. A. K. and Ruqaiya Hasan 1976 Cohesion in English. London: Longman. Heine, Bernd 1993 Auxiliaries: Cognitive Forces and Grammaticalization. Oxford/ New York: Oxford University Press. 2003 Grammaticalization. In Brian D. Joseph and Richard D. Janda (eds.), The Handbook of Historical Linguistics, 575-601. Oxford: Blackwell.

English adjectives of general comparison

275

Hopper, Paul J. 1991 On some principles of grammaticization. In Elizabeth Closs Traugott and Bernd Heine (eds.), Approaches to Grammaticalization. Volume I: Focus on Theoretical and Methodological Issues, 17-35. Amsterdam/Philadelphia: John Benjamins. Hopper, Paul J. and Elizabeth Closs Traugott 1993 Grammaticalization. Cambridge: Cambridge University Press. Krug, Manfred G. 2000 Emerging English Modals: A Corpus-Based Study of Grammaticalization. Berlin/New York: Mouton de Gruyter. Langacker, Ronald W. 1987 Foundations of Cognitive Grammar. Volume I: Theoretical Prerequisites. Stanford: Stanford University Press. 1990 Subjectification. Cognitive Linguistics 1-1: 5-38. 1991 Foundations of Cognitive Grammar. Volume II: Descriptive Application. Stanford: Stanford University Press. 1998 On subjectification and grammaticization. In Jean-Pierre Koenig (ed.), Discourse and Cognition: Bridging the Gap, 71-89. Stanford CA: CSLI Publications. 1999 Losing control: Grammaticization, subjectification, and transparency. In Andreas Blank and Peter Koch (eds.), Historical Semantics and Cognition, 147-175. Berlin/New York: Mouton de Gruyter. 2001 Discourse in cognitive grammar. Cognitive Linguistics 12-2: 143-188. 2002 a Deixis and subjectivity. In Frank Brisard (ed.), Grounding: The Epistemic Footing of Deixis and Reference, 1-28. Berlin/New York: Mouton de Gruyter. 2002 b Remarks on the English grounding systems. In Frank Brisard (ed.), Grounding: The Epistemic Footing of Deixis and Reference, 29-38. Berlin/New York: Mouton de Gruyter. 2003 Extreme subjectification: English tense and modals. In Hubert Cuyckens, Thomas Berg, Rene Dirven and Klaus-Uwe Panther (eds.), Motivation in Language: Studies in Honor of Günter Radden, 3-26. Amsterdam/Philadelphia: John Benjamins. 2004 Remarks on nominal grounding. Functions of Language 11-1: 77-113. Lehmann, Christian 1985 Grammaticalization: Synchronic variation and diachronic change. Lingua e Stile 20: 303—318. Lyons, John 1977 Semantics. Volume 2. Cambridge: Cambridge University Press.

276

Tine Breban 1982

Deixis and subjectivity: Loquor, ergo sum?. In Robert J. Jarvella and Wolfgang Klein (eds.), Speech, Place, and Action: Studies in Deixis and Related Topics, 101-124. New York: Wiley and Sons. McGregor, William 1997 Semiotic Grammar. Oxford: Clarendon Press. Mortelmans, Tanja 2002 "Wieso sollte ich dich küssen, du hässlicher Mensch!" Α study of the German modals sollen and müssen as "Grounding Predications" in interrogatives. In Frank Brisard (ed.), Grounding: The Epistemic Footing of Deixis and Reference, 391^432. Berlin/New York: Mouton de Gruyter. Nevalainen, Terttu and Matti Rissanen 2002 Fairly pretty or pretty fair? On the development and grammaticalization of English downtoners. Language Sciences 24: 359380. OED: Murray, James A.H., Henry Brodly, W.A. Craigie and C.T. Onions (eds.) 1933 The Oxford English Dictionary. Oxford: Oxford University Press. Paradis, Carita 2000 Reinforcing adjectives: A cognitive semantic perspective on grammaticalisation. In Ricardo Bermüdez-Otero, David Denison, Richard M. Hogg and C.B. McCully (eds.), Generative Theory and Corpus Linguistics. A Dialogue from 10 ICEHL, 233-258. Berlin/New York: Mouton de Gruyter. Pelyväs, Peter 2001 On the development of the category modal: A cognitive view. How changes in image-schematic structure led to the emergence of the grounding predication. In Nemeth T. Enikö and Käroly Bibok (eds.) Pragmatics and the Flexibility of Word Meaning. London: Elsevier. Quirk, Randolf, Sidney Greenbaum, Geoffrey Leech and Jan Svartvik 1972 A grammar of contemporary English. London: Longman. Tabor, Whitney 1994 The gradual development of degree modifier sort of and kind of. A corpus proximity model. Papers from the Chicago Linguistics Society 29: 451-465. Traugott, Elizabeth Closs 1988 Pragmatic strengthening and grammaticalization. In Shelley Axmaker, Annie Jaiser and Helen Singmaster (eds.), Proceedings of the 14th Annual Meeting of the Berkeley Linguistics Society. General Session and Parasession on Grammaticalization, 406—416. Berkeley CA: Berkeley Linguistics Society.

English adjectives of general comparison

277

On the rise of epistemic meanings in English: An example of subjectification in semantic change. Language 65: 31-55. Subjectification in grammaticalisation. In Dieter Stein and Susan 1995 Wright (eds.), Subjectivity and Subjectivisation: Linguistic Perspectives, 31-54. Cambridge: Cambridge University Press. The rethoric of counter-expectation in semantic change: A study 1999 in subjectification. In Andreas Blank and Peter Koch (eds.), Historical Semantics and Cognition, 177-196. Berlin/New York: Mouton de Gruyter. 2003 a From subjectification to intersubjectification. In Raymond Hickey (ed.), Motives for Language Change, 124-139. Cambridge: Cambridge University Press. 2003 b Constructions in grammaticalization. In Brian D. Joseph and Richard D. Janda (eds.), The Handbook of Historical Linguistics, 624-647. Oxford: Blackwell. 2004 Historical pragmatics. In Laurence R. Horn and Gregory Ward (eds.), The Handbook of Pragmatics, 538-561. Oxford: Blackwell. Traugott, Elizabeth Closs and Richard B. Dasher 2002 Regularity in Semantic Change. Cambridge: Cambridge University Press. Traugott, Elizabeth Closs and Ekkehard König 1991 The semantics-pragmatics of grammaticalization revisited. In Elizabeth Closs Traugott and Bernd Heine (eds.), Approaches to Grammaticalization. Volume I: Focus on Theoretical and Methodological Issues, 189-218. Amsterdam/Philadelphia: John Benjamins. 1989

Subjectification in gradable adjectives Henk Pander Maat

1.

Introduction

Adjectives are commonly divided into non-gradable and gradable ones. Typical non-gradable adjectives include adjectives denoting nationality, provenance or style {American, Gothic) and other denominal adjectives with meanings such as 'consisting o f , 'involving', or 'relating to' (annual, medical, political). Among gradable adjectives we need to distinguish between scalar adjectives (e.g. big, warm and fat), limit adjectives like identical and true and extreme adjectives such as terrible or brilliant (Paradis 2001). Scalar adjectives are fully gradable, that is they allow comparative and superlative degrees and can be modified by items such as very, rather or somewhat. Limit and extreme adjectives only allow totality modifiers (e.g. completely identical, absolutely true). We will confine ourselves here to prototypical gradable adjectives such as big. This article discusses how 'bare' gradable adjectives, i.e. adjectives outside comparative and superlative constructions, are understood. I will first present a short survey of ten distributional and semantic properties of gradable adjectives, concluding that these properties have not yet been accounted for in a coherent way. I will propose such an account, drawing on Langacker's (1990, 1999) notion of subjectification. I suggest that the crucial property of gradable adjectives is not simply scalarity, but their combination of scalarity with subjective construal. I present two versions of a subjectification analysis of gradable adjectives. The first one starts from the common assumption that gradable adjectives refer to a degree of some property that exceeds the standard value for a certain class of entity. Upon closer inspection, this analysis reveals problems that appear insurmountable. Hence, a second analysis will be presented that combines the notion of subjectification with the argumentative perspective of Ducrot (1980, 1996).

280

Henk Pander Maat

2.

Ten properties of gradable adjectives

2.1.

Comparative and superlative degrees

It is important to realize that we may only talk about "gradable adjectives" on a high level of schematicity. The members of this class of lexemes always appear in particular constructional schema's; these instantiations exploit the class-related semantic potential in different ways (see Taylor 2002, 164 ff. for general discussion). Fully gradable adjectives allow both bare forms (attributive or predicative) and comparative or superlative constructions. The last two types of construction can in turn be instantiated by inflected forms with -er and -est or by periphrastic equivalents with more, most, less and least. In all these forms, one entity is being compared to another entity or a set of other entities. This other entity may be left implicit; that is, the comparison may be 'elliptic', see (1). (1)

2.2.

Sheila is taller (than her sister).

Comparison classes

It has often been indicated that the interpretation of bare gradable adjectives requires the use of so-called comparison classes (e.g. Lyons 1977, Ludlow 1988, Kennedy 1999). For instance, big in 'a big elephant' invokes the comparison class of elephants, while in 'a big mouse' the comparison class of mice is relevant. The comparison class can be made explicit by adding specific prepositional clauses or phrases, see (2). (2)

The house was fairly big (as laborer's cottages go).

Interestingly, the interpretation of comparative and superlative constructions does not require the use of general comparison classes. That is why we can understand sentences such as an elephant is larger than a mouse.

2.3.

Expressivity

Gradable adjectives can be used in sentence formats carrying some kind of expressivity, such as exclamations and exclamatory questions (see Bolinger 1972: 281-292), illustrated in (3) and (4) (stressed syllables are indi-

Subjectification in gradable adjectives

281

cated by capitals). Gradable adjectives can also be combined as predicate adjuncts with verbs referring to personal impressions such as consider (Verspoor 1997), see (5). (3) (4) (5)

What a WARM welcome they gave us. AmIHUNgry! Michael considers the table clean.

2.4.

Gradable adjectives precede other adjectives

In the structure of the nominal group, premodifying gradable adjectives tend to precede non-gradable ones: we say big American car instead of *American big car, and impressive annual report instead of *annual impressive report (compare Dixon 1982: 15-26; Halliday 1985: 163-165; Quirk et al. 1985: 1338-1341).

2.5.

Degree modifiers go both ways

A striking properly of degree modifiers of scalar adjectives is that they may both reinforce and weaken the extent to which some quality holds: we can both say that a cupboard is very heavy and that it is a little bit heavy. This raises the question of what exactly is being upgraded and downgraded here.

2.6.

Antonymy

A number of gradable adjectives form pairs of symmetrical, contrary antonyms: big-small, high-low, warm-cold, rich-poor, young-old, good-bad. The existence of antonyms seems a natural thing, given that the adjective refers to a scale and scales can be bipolar. All adjectives with contrary antonyms are gradable. However, this relation is not reversible: a fair number of gradable adjectives has no antonym, for instance many adjectives referring to character traits and emotional states: angry, scary, jealous.

282

Henk Pander Maat

2.7.

Unmarked terms

Some antonym pairs have an unmarked term that can be used in howquestions (how old is he?) and can be accompanied with measurement phrases (he's three months old). Examples include deep-shallow, thick-thin (unmarked terms are underlined). Other adjectives can be pre-modified by interrogative how (how heavy is it?) but are not used with measurements (*It is two pounds heavy). But again, only a subset of gradable adjectives allows this kind of unmarked use (compare *How kind is he and *How jealous is he?).

2.8.

Correlative (consequential) grading

Apart from simple adverbial degree modifiers such as very, gradable adjectives also allow correlative degree constructions with enough (for/to), too (for), so (that) and such (that), see (6), (7) and (8) below. Constructions with enough and too can also be used without infinitival complements, and may express causal relations across sentence boundaries, see (9). All these degree constructions are not understood in relation to a general comparison class, but in terms of a particular consequence attached to the degree of some quality. Hence we will further speak of consequential grading. The consequence may be mentioned in the construction itself, but can also be referred to in neighboring utterances. But often, even this is unnecessary since the consequence needs not to be stated at all, see (10). Here, the consequence is provided by a general expectation regarding some quality associated with some kind of entity: tea must not be too hot to drink. (6) (7) (8) (9) (10)

2.9.

The grass is too short to cut. The grass isn't long enough to cut. The grass is so short that it can't be cut. I stopped doing voluntary work. My job got too demanding. The tea is too hot.

Gradable adjectives have non-gradable uses

Gradable adjectives have well-entrenched uses in which grading is impossible:

Subjectification

(11)

in gradable adjectives

283

we had a hot meal /? a very hot meal we ate some cold meat / ? very cold meat

Compare also short vowels, long trousers, long weekend, dark beer, light artillery, light beer and heavy traffic. Entrenched as these uses may be, they still seem to be rather specialized; the number of these non-gradable combinations appears to be limited. And our first hunch when encountering a gradable adjective modifying an unknown noun (e.g. a light fonk') is that it is gradable.

2.10. Non-gradable adjectives have gradable uses Conversely, standard non-gradable adjectives may be accompanied by degree modifiers: (12) (13)

Harry Potter is a very English film. Teacher Education programmes are becoming too vocational.

Of course, the gradable use of non-gradable adjectives is a marked use in that it requires the presence of degree modifiers. Paradis (2001: 58-59) speaks of 'contextual coercion' toward scalar construal in these cases. She also observes that some adjectives are indeterminate between scalar and non-scalar construal. For instance, true and empty may construed in either way. Croft and Cruse (2004: 185-187) discuss different construals of clean and dirty. The observations 9 and 10 require some additional discussion, since they show that gradability is a properly of adjective uses, not of adjectives. Now what distinguishes gradable and non-gradable uses semantically? My suggestion is that non-gradable combinations categorize the nominal referent, while the gradable combinations qualify it. Non-gradable adjective-noun combinations tend to denote unified entity concepts, i.e. categories. For instance, polar bear denotes a category, just like bear does. Although the bear category encompasses the polar bear category, both are kinds of entities (animals, in this case). Similarly, dark beer refers to a kind of beverage, just like beer does. Like Wierzbicka (1986: 356-361) has stressed, categories are unified, multi-dimensional wholes evoking a positive image, not arbitrary collections of features. In gradable combinations however, speakers add a uni-dimensional quality

284

Henk Pander Maat

concept to a multi-dimensional categorical concept (see Wierzbicka 1986: 373-374). The quality concept does not entirely fuse with the nominal category creating a new category, but it remains distinct. A tall man is not a kind of individual, but an individual qualified in terms of length. Let us assume that we agree about the 10 properties of gradable adjectives discussed so far. Our problem then is: what is the underlying unity in this set of properties? For instance, why should a part of speech that allows degree modification also allow comparative inflections? Why should this same part of speech also carry an element of expressivity? And why should it require interpretation by means of comparison classes? And what does it mean that gradable adjectives sometimes do not allow degree modifiers, while non-gradable adjectives sometimes do? In section 4 we will try to answer this kind of questions by an account of grading as invoking subjective reference points. First we need to introduce the two basic notions of reference points and subjectivity.

3.

Reference points and subjectivity

Quite a number of natural language expressions require interpretation in terms of an operation on contextually given reference points. The best known categories in this respect are probably person, time and place deictics (see Anderson and Keenan 1985 and Hanks 1992 for overviews). The members of a deictic paradigm differ with respect to the distance between the entity referred to and the speech situation, conceived of as an ensemble of person, time and place coordinates. For instance, the first person is responsible for speaking and as such he is more closely involved in the speech situation than the hearer; the hearer, being the addressee of the utterance, is again closer involved in the speech situation than third persons (Janssen 1995). Similarly, the temporal distance between the moment of speaking and the time point indicated by now is minimal, and smaller than the temporal distance between the moment of speaking and the time point indicated by then·, and the same kind of difference holds between the locations indicated by here and there. In some deictics it is not the distance to the reference point that is being expressed, but the direction of movement with regard to the reference point: examples are the contrast between come and go (see also the contribution by Uehara to this volume) or that between pull and push.

Subjectification in gradable adjectives

285

To sum up, deictic terms impose an ordering of persons, time points, locations and events in terms of a relation (distance or direction) to a contextually given position, viz. the reference point. Without further notice, the reference point is taken to have the person, time and place coordinates of the speech situation. For instance, in Peter is sitting left in the first row the standard interpretation is ' left as seen from the position of the speaker'. A second notion we need to introduce is subjectivity, that is the degree to which an expression invokes implicit reference points that are related with a salient participant, most often the speaker. Compare (14), (15) and (16). (14) (15) (16)

Paula's house is opposite to the school. Paula's house is across from mine. Paula's house is just opposite. OS IS

lm

—χ

tr

—χ

&-Γ® \\ \

4 ιt ι // x v 1; «! \ | /

Figure 1. Paula's house is opposite to the school. OS = overall scope; IS = immediate scope; LM = landmark; TR = trajector; S = school; PH = Paula's house; C = conceptualizer All three sentences refer to the location of Paula's house. Still, there are important differences in the role of the speaker in the sentence and its interpretation. In (14), the speaker is outside the explicit content of the sentence - its immediate scope - and his position is not involved in the interpretation. According to Langacker (1985, 1990, 1999) the speaker is maximally subjectively construed·, the speaker is in no way part of the ob-

286

Henk Pander Maat

ject of conception and functions entirely as the subject of conception without being itself conceived (see Figure 1). While the speaker is maximally subjective, the linguistic utterance is maximally objective in the sense that it does not contain anything subjective. By contrast, in (15) the speaker is objectively construed since he is referred to in the sentence; his house is the landmark in terms of which Paula's house is being located. At the same time the speaker is part of the overall scope, that is the entire conceptual complex evoked by the sentence, because his identity is needed for the interpretation of the deictic mine (see Figure 2). Finally, in (16) the speaker is construed as moderately subjective: he is outside of the immediate scope and hence is construed subjectively, but not completely so, because his position is an implicit reference point with regard to which Paula's house is being located. While the speaker is outside of the immediate scope, he is inside the overall scope (see Figure 3).

OS

Figure 2. Paula 's house is opposite to mine. CH = conceptualizer's house Subjectification is understood by Langacker (1999) as a process of semantic change in which the objective content of some linguistic unit gradually fades away, while the activity of the conceptualizer remains in place. In (14) above, there is an objective spatial configuration involving two buildings; the conceptualiser is taken to mentally follow the path leading from the landmark to the location of the trajector. The dashed arrows

Subjectification in gradable adjectives

287

coming from the conceptualizer symbolize this mental activity. In the subjectified counterpart (16), there is no objective spatial configuration involving two profiled locations anymore. An implicit reference point has replaced one of the locations. But still, the conceptualizer is taken to mentally track the imaginary path leading from the reference point to the located entity. In section 2, ten properties of grading and gradable adjectives were mentioned. In the next section, I use the notions of reference point and subjectification to account for the co-occurrence of these properties.

4.

Grading, standard values and subjectification: the first analysis

It was observed in section 2 that gradable adjectives have non-gradable uses and non-gradable adjectives may be used with degree modifiers. Our discussion needs to exclude the first phenomenon and to include the latter one. Hence the rest of this article is concerned with grading, not with a particular class of adjectives. Most of the discussion will focus on grading uses of prototypical gradable adjectives however, since only these adjectives may be used for grading purposes in their bare form, while nongradable adjectives require explicit modification in order to realize grading.

OS IS

\\ χvV

x

ιI I II

V / V \\ 11 /// vl/

Figure 3. Paula 's house is just opposite. R = Reference point.

/ /' / '/—

288

Henk Pander Maat

Now what exactly is grading? As we saw, it is commonly assumed that a big elephant refers to an elephant bigger than the normal elephant, according to the conceptualizer. In other words, the interpretation of big involves a standard value associated to the comparison class that is relevant to the modified entity. The relevance of comparison classes is mentioned by virtually every author on gradable adjectives (e.g. Lyons 1977, Cruse 1986: 206, Ludlow 1988, Kennedy 1999: 117-124, Paradis 2001: 54, Taylor 2002: 220). However, the notion of standard values has an interesting consequence that has been discussed less often, except that it has been already mentioned by Lyons (1977: 273): "The use of a gradable antonym always involves grading, implicitly if not explicitly". In other words, adjectives in comparative constructions constitute instances of explicit grading of entities in terms of other entities, but bare gradable adjectives are also cases of grading, in the sense that they comparatively qualify entities in terms of a standard, though this standard remains implicit. This may be the underlying unity of the different constructional instantiations of gradable adjectives. Let us take a closer look at the similarities and the differences between explicit and implicit grading.

4.1.

How explicit comparisons differ from bare gradable adjectives

Let us start with Pete is taller than Frank. In their explicit comparative uses, gradable adjectives do not require comparison classes for their interpretation; this also holds for elliptic comparisons in which the second entity is not mentioned because it can be discovered from the context. The meaning of Pete is taller than Frank is schematically represented in Figure 4. The two compared entities are objectively construed participants and hence appear in the immediate scope. Their relation is such that the trajector participant Pete has a higher value on the variable of length than the landmark participant Frank. The speaker is outside the overall scope. Now compare the analysis of Pete is tall for his age, displayed in Figure 5. This sentence involves a comparison between the length of Pete with a standard value that functions as a reference point. The standard value here is derived from the comparison class indicated in (forj his age. Because of this, the reference point is in the immediate scope. Now compare Pete is tall, displayed in Figure 6. The only structural difference with Figure 5 is that the reference point remains outside the immediate scope. It is involved as an unprofiled, 'offstage' entity, since it is part of the implicit assump-

Subjectification in gradable adjectives

289

tions brought to bear on the interpretation by the conceptualizer. Because of this, the conceptualizer is also present in the overall scope. Hence, though the dimension of length is explicitly referred to, the exact length of Pete is subjectively construed. This analysis offers several advantages. First of all, it explains why gradable adjectives have comparative and superlative instantiations: these degree constructions only bring out into the open the comparative configuration that remains implicit in bare uses of gradable adjectives (at least when these are used to realize grading; here and below we do not mean to discuss categorizing uses of gradable adjectives). But besides this, the subjectification analysis explains a number of semantic differences between comparative constructions and bare gradable adjectives.

OS lm

tr

Figure 4. Pete is taller than Frank. OS = overall scope; IS = immediate scope; LM = landmark; TR = trajector; C = conceptualizer; L = length; R = reference point

For one thing, it explains why the explicit comparative constructions do not invoke standard values: these are only needed when a single entity is to be characterized, with the standard value replacing the second entity. Furthermore, the subjective nature of the standard value helps us to understand its specific characteristics. This value is not only subjective in the Langackerian sense that it remains offstage, but also in the sense that it requires complex mental activities of a subject of consciousness. More specifically, this subject needs to identify the comparison class involved in a certain utterance, and to construct a general representation of this class in

290

Henk Pander Maat

terms of the specific variable that is involved. The general character of the standard value may explain why bare gradable adjectives appear to assign qualities to entities as if they were qualities of the entities themselves. Comparative statements such as A is older than Β do not allow the inference that A is old when taken by itself. Why then does A is old allow this inference? It is intuitively plausible that the location of an entity with regard to a general standard is a more reliable basis for assessing its qualities than a comparison with a particular other entity could be.

OS IS

ΊI Γ ^'

\

>-L

Figure 5. Pete is tall for his age.

OS IS

Figure 6. Pete is tall. At the same time, the connection between the standard value and the activities of a subject of consciousness explains why gradable adjectives display a certain affinity with expressive sentence formats and verbs that

Subjectification in gradable adjectives

291

explicitly refer to personal judgments. This 'personal' element can be elucidated by comparing the way non-gradable and gradable adjectives conceptually combine with nouns. As discussed in section 2 above, nongradable adjective-noun combinations tend to denote unified categories, while gradable adjectives are used to add a uni-dimensional quality concept to a multi-dimensional categorical concept (see Wierzbicka 1986: 373-374). That is, the quality is non-intrinsic to the qualified entity; its stems from a qualification, a subjective assessment by some speaker. This also seems to explain the preferred order of adjectival premodifiers in the nominal group. Why do we say big American car and not *American big carl As the principle accounting for this kind of preferences, Quirk et al. (1985: 1341) suggest a subjective/objective polarity: "Modifiers relating to properties which are (relatively) inherent in the head of the noun phrase, visually observable, and objectively recognizable and assessible, will tend to be placed nearer to the head and be preceded by modifiers concerned with what is relatively a matter of opinion, imposed on the head by the observer, not visually observed, and only subjectively assessible". That is, adjectives that are conceptually fused with the nominal concept are closer to the noun than additional qualifications of this newly formed complex concept. In the nominal group, linear distance mirrors conceptual distance (see also Dirven and Verspoor (1998: 10) on the iconic distance principle). The reference points invoked by gradable adjectives differ in some respects from the situational reference points invoked by deictic terms. Deictic reference points are generally unrelated to the conversation topic, while the standard values invoked by gradable adjectives can only be reconstructed with the help of prior knowledge on the category under discussion. A similarity between standard values and deictic reference points is that they may both be located with some other participant than the speaker, provided that this participant is contextually salient: (17)

Arthur looked out of the window. An enormous truck drove by.

The adjective in the second sentence needs to be understood in terms of the standard value holding for the ^ri/zwr-participant. When we know that Arthur is three years old, we might infer that the truck may not be exceptionally big from an adult point of view. Another implication of the reference point analysis concerns the correlation between the amount of linguistic coding and the distance from the reference point. In verbal paradigms, the first person present tense form

292

Henk Pander Maat

often has zero-inflections, that is it uses a minimum of linguistic material; presumably, this is because the first person and the present tense constitute the reference points in the person and temporal system of the language (compare Frawley 1992: 94-99 and 280-281 on the person system and 340-370 on the deictic structure of tenses). Similarly, when referring to entities or situations without information concerning their location, we assume that the distance between the situation and the place of speaking is minimal. When talking about unemployment, we assume that the situation in our own country is being discussed, except when indications to the contrary are given; and generally, the village refers to the village most nearby. Likewise, a gradable adjective is only added to a nominal when grading is asked for, that is when the degree of some quality needs to be comparatively assessed, because it is remarkable in some way. The next sentence implies that the man had a remarkable age, but not a remarkable length: (18)

An old man with a suitcase entered the hotel lounge.

Now it could be argued that this is nothing more than the Gricean maxime 'be relevant' at work. But there is a difference between different kinds of modifications in this respect: the additional information that the man has a suitcase, while possibly relevant in order to enrich the entity representation under construction, does not imply that carrying a suitcase is somehow unusual in the world being described. In order to create expectations concerning suitcases, we would need to say something like remarkably enough he was carrying a suitcase. By contrast, gradable adjectives by themselves already convey the idea that something differs from the norm. In combination with the fact that adjectives are the most efficient linguistic device of adding information to nominal referents (see Croft 1991: 52-53), this justifies the default inference that in the absence of a gradable adjective, the value of the variable in question for the entity involved does not significantly differ from the standard. To conclude this section I will comment on recent suggestions by Taylor (2002), who proposes a different analysis of gradable adjectives in a Cognitive Grammar framework. According to Taylor, gradable adjectives profile a relation between an entity (the trajector) and a region "in excess of some norm", this region functioning as the landmark that is "incorporated in the semantic structure of the word" (2002: 220). I do not follow this analysis. First, the notion of an incorporated landmark does not make sense to me. Langacker (1997: 66) has been quite explicit on this: "Since

Subjedification

in gradable adjectives

293

trajector and landmark status is a matter of focal prominence, the entities selected must by definition be onstage and objectively construed". But more importantly, the subjectivity analysis of standard values provides a more satisfactory explanation of the specific semantic and distributional features of gradable adjectives, as I have tried to show above.

4.2.

Why degree modifiers may both reinforce and weaken interpretations

In section 3 we saw that the interpretation of deictics may involve two parameters: the distance between the located entity and the reference point, and the orientation of the entity with regard to the reference point, more specifically its movement or direction. Our analysis of gradable adjectives has so far only used the parameter of orientation, stating that the actual value of the entity is taken to exceed the standard value. In bare gradable adjectives the distance between the value and the reference point remains unspecified. But the notion of distance clearly invites interpretations in terms of scales, admitting several values. Hence, it is only to be expected that languages offer all kinds of devices to modify the distances involved in the interpretation of gradable adjectives. This is why gradable adjectives are commonly encountered in constructions with degree modifiers. But the modifiers may work in two directions. First, they increase the distance. For instance, he is very fat means that the difference between his weight and the standard weight for persons of his length is larger than is the case for persons simply called fat. Other adverbial modifiers serve to decrease distances, such as a little (bit), slightly, rather and somewhatThis explains why these terms weaken the interpretation of the adjective, but do not cancel the assignment of the quality involved. For instance, he is a little bit fat means that the difference between his weight and the standard weight is smaller than for persons just called fat, but the difference is still there: he is still fat to some degree. This analysis implies that the distance to the reference point associated with a bare gradable adjective functions as a secondary reference point (see RP2 in Figure 7), with regard to which intensifies and weakeners are un1. Of course there are differences between the various modifiers within the two classes (e.g. somewhat is often felt as expressing a negative evaluation) but a discussion of these falls outside the scope of this article.

294

Henk Pander Maat

derstood. A weakener moves the distance below the secondary reference point; intensifiers indicate that the distance exceeds the secondary reference point (see Figure 7; a little and very are only used as examples of modifiers here). RP2 = X - RP1

a little A

very A

Figure 7. The interpretation of a gradable adjective A and its combination with weakening and reinforcing modifiers. SV = standard value; W = the variable associated with adjective A; RP1 = the primary reference point; RP2 = secondary reference point. In section 2 we asked ourselves what exactly is being graded by degree modifiers. Figure 7 provides the beginnings of an answer: gradable adjectives can be graded because they are already implicit grading devices by themselves, that is they refer to the distance between a subjectively construed reference point and the actual value of a variable. This distance in turn can be secondarily graded by the addition of modifiers. In short, gradability presupposes grading, not the other way round. Secondary grading can also be used in comparative and superlative constructions, in which it modifies the distance between several objectively construed entities. For instance, we can not only say that this stick is longer than another one, but also that it is just a little bit longer, somewhat longer or much longer. Likewise, somebody may be by far the best or the very best of his class. Some degree modifiers (e.g. rather) only allow implicit grading contexts, others only allow explicit contexts (e.g. much, by far), but it is important to note that others appear in more than one environment (somewhat, a little bit, very). So far, our analysis accounts for the first five properties of gradable adjectives mentioned in section 2. We will now proceed to discuss the properties 6 and 7: gradable adjectives may come in antonym pairs and some of them have unmarked uses.

Subjectification in gradable adjectives

4.3.

295

Antonyms: relative scales sharing a reference point

So far we have spoken of values exceeding the standard values, but this is a simplification. In antonym pairs, one of the members indicates that a certain value falls below the standard value. Although only a minority of gradable adjectives comes in pairs (see section 2), semantic work on gradability has almost entirely concentrated on antonymy. Hence it is important to discuss antonymy from the present perspective. Older work on gradable antonyms (see the overview in Hoffman 1987: 98 ff. and Cruse 1986: 211) presents bipolar scales with the antonyms at the extremes. However, more recently Croft and Cruse (2004: 169-192; see also Cruse and Togia 1995) have discarded the notion of one single scale for antonyms. Let us review their proposals. To begin with, Croft and Cruse distinguish two scale schema's: absolute scales, most typically representing quantitative measuments (25 cm), and relative scales in which a quantity is seen as more or less than some reference value. They then proceed to discuss the following types of antonym scale: - Monoscalar systems', these systems involve an absolute scale (such as LENGTH) along which the relative scales of short and long can be found (see Figure 8). Monoscalar systems have a 'supra' term (e.g. long) which is more prone to absolute uses than the 'sub' term (e.g. short). For instance, the question how long? is 'open-minded', while how short? suggests that the length is below the reference point. Likewise, twice as long can be straigthforwardly understood, while twice as short is at least more complex. Biscalar systems come in three types: - Disjunct equipollent systems are exemplified by the pair cold - hot, the two items behave symmetrically in all respects; they both have their scales, and the two scales meet at their zero points (see Figure 9). Both items can be used in absolute and relative construals (how cold / hot is it?). - Parallel equipollent systems in which the scales do not meet at their zero points but overlap each other completely. Croft and Cruse cite hard-soft as an example (see Figure 10), but they note that this kind of system is rarely found. - In overlapping systems there is only partial overlap between the two scales. Since the only example of this scale seems to be the evaluative adjective pair good - bad, we will further ignore this system type.

296

Henk Pander Maat

LENGTH (X-ness)

1