Issues in Cognitive Linguistics: 1993 Proceedings of the International Cognitive Linguistics Conference 9783110811933, 9783110152197

Table of contents :
Preface
Theoretical issues
Bridges between generative and cognitive linguistics
Lexical semantics and morphology
A modular approach to the semantics of space in language
Beer and semantics
Is everything black and white in conceptual oppositions?
Categorization and analogical change: The case of athematic 1sg -m in the Slavic languages
Contrast and schemas: Antonymous adjectives
Figurative giving
Nominalizations, metonymy and lexicographic practice
Langacker semantics for three Coeur d'Alene prefixes glossed as ‘on’
Aspects of prepositions and prepositional aspect
Some aspects of relational nouns
Prototype marker or reflexive marker: Russian -sja and categorical change
Metaphor
The strengths and weaknesses of the left/right polarity in Russian: diachronic and synchronic semantic analyses
‘Unnatural barriers’: Why metaphor matters (or, linguistics meets the geopolitics of law)
Through as a means to metaphor
Is metaphor really a one-way street? One of the basic tenets of the cognitive theory of metaphor put to the test
Syntax and semantics
Samoan as an active zone language1
Two transitive construction frames in Spanish: The prepositional and the non-prepositional accusative
Swedish abstract transitional phrases: An in-between phenomenon in the linguistic system
Grammatical relations in Spanish triactant clauses
Degrees of subjectivity in epistemic modals and perspective representation
Biactant Spanish clauses. Syntactic markedness and semantic prototype
To infinitives
Adverbial placement and iconicity
Pragmatics
Nominal vs. temporal interpretation in discourse
Markedness and prototypical speaker attributes
Computational Linguistics
Implementing cognitive semantics – overview of the semantic composition processes and insights into the grammatical composition processes
Subject index

Citation preview

Issues in Cognitive Linguistics

1749

1999

Cognitive Linguistics Research 12

Editors Rene Dirven Ronald W. Langacker John R. Taylor

Mouton de Gruyter Berlin · New York

Issues in Cognitive Linguistics 1993 Proceedings of the International Cognitive Linguistics Conference Edited by Leon de Stadier Christoph Eyrich

w DE

G Mouton de Gruyter Berlin · New York 1999

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

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

Library of Congress Cataloging-in-Publication Data International Cognitive Linguistics Conference (3rd : 1993 : Louvain, Belgium) Issues in cognitive linguistics : 1993 proceedings of the International Cognitive Linguistics Conference / edited by Leon De Stadier, Christoph Eyrich. p. cm. - (Cognitive linguistics research ; 12) Includes bibliographical references and index. ISBN 3-11-015219-3 (alk. paper) 1. Cognitive grammar Congresses. I. De Stadier, L. G. (Leon G.) II. Eyrich, Christoph, 1964. III. Title. IV. Series. P165.I58 1993 415-dc21 99-26546 CIP

Die Deutsche Bibliothek - Cataloging-in-Publication Data Issues in cognitive linguistics : 1993 proceedings of the International Cognitive Linguistics Conference / ed. by Leon De Stadier ; Christoph Eyrich. - Berlin ; New York : Mouton de Gruyter, 1999 (Cognitive linguistics research ; 12) ISBN 3-11-015219-3

© Copyright 1999 by Walter de Gruyter GmbH & 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. Printing: Werner Hildebrand, Berlin Binding: Lüderitz & Bauer, Berlin Printed in Germany

Preface This volume contains contributions in the field of Cognitive Linguistics originally presented as papers at the Third International Cognitive Linguistics Conference held in Leuven, Belgium, during the summer of 1993. Due to its long incubation period, many of the contributions were revised and the references updated.

The contributions Although it lies at the very heart of the cognitive linguistic paradigm that one should not have the linguistic level approach traditionally associated with a school such as generative linguistics, I have opted for some kind of thematic organization in this volume. Even though I acknowledge the fact that many of the contributions may fit into more than one slot in the organizational pattern, I have placed them in six broad categories according to the main focus in each contribution, namely • Theoretical Issues • Lexical Semantics and Morphology • Metaphor • Syntax and Semantics • Pragmatics • Computational Linguistics The variety of the contributions give a clear indication of the diversity and vigour to be found in the cognitive linguistic paradigm. With the following prelude I wish to give a first glimpse of this diversity and vigour as it is portrayed in the pages to follow. In the theoretical section, Frederick Newmeyer addresses the possibility of a convergence between cognitive grammar and generative approaches to language. He explores the extent to which the differences between the theories are merely the result of differences in focus or even terminology and the extent to which they are substantive, showing that at least some of the stumbling blocks to mutual understanding are the result of pseudo-issues. The field of lexical semantics, and lately related fields such as morphology, drew quite a substantial number of contributions.

vi

Preface

The first paper in this section, by Michel Aurnague and Laure Vieu, aims to elaborate some formal tools for representing the semantic content of French expressions referring to space. They discuss the topological prepositions sur On' and dans 'in', as well as the projective propositions dessus 'above', dessous 'below', devant 'in front of \derriere 'behind', in the schema "Ntarget est prep Nj^^^". In the category of spatial referents they also took into account several internal localization nouns such as haut 'top', has 'bottom', devant 'front extremity', which are all lexical items pointing out the different portions of an object. From an empirical point of view, the whole study rests on a detailed semantic analysis which distinguishes the different spatial configurations each of these lexemes allow us to refer to. This linguistic study also identifies the different inferential schemata that appear to combine spatial expressions in discourse. On the basis of these observations, they propose a formal representation of the semantic content of these expressions. The contribution by Dirk Geeraerts offers an onomasiological analysis of Belgian beer names. He argues that two important factors influence the selection of a name for a particular type of beer. On the one hand, salience effects reveal that some referential features of the beers (the target domain of name giving) are more prominent onomasiological motifs than others. On the other hand, an analysis of the source domains of the figurative names reveals the metaphorical connotations that are attributed to beer. Various languages possess opposites that have developed historically from the same root. The paper by Gabor Gyori and Iren Hegedus examines the cognitive processes underlying this kind of semantic change in the case of some basic perceptual oppositions. It is argued that such oppositions can only be conceptualized as gestalts, i.e. as two inseparable poles of a unity. The simultaneous lexicalization of these oppositions probably reflects this feature. They show that the semantics of the original form must have been characterized by an internal polarization, which split up at the lexical level and caused the opposite semantic developments to proceed in parallel. The paper by Laura Janda deals with the phenomenon of irregular inflectional affixes in inflected languages. What were once forms related to productive and systematic linguistic distinctions often seem to fade into oblivion. However, Janda makes the point that not all of these affixes necessarily wither and die. Languages are capable of recycling nearly extinct morphemes and using them either to restore distinctions that have eroded or even to build entirely new systems of distinctions. She illustrates this exaptative analogical extension with a chapter from the history of the Slavic languages: the spread

Preface

vii

of Isg -m from a handful of athematic verbs to much or all of the verbal lexicon in the West and South Slavic languages. In his paper, Arthur Mettinger looks at the cognitive basis of the traditionally well-established semantic phenomenon of antonymy and complementarity in adjectives. The claim is made that two image Schemas, the SCALE schema and the CONTAINER schema, are responsible for our understanding of these adjectives. Moreover, the contribution shows the necessity of integrating image-schematic structures (grounded in bodily experience) and word-class specific schematic structures (grounded in linguistic experience). Finally, suggestions as to the poly schematic nature of CONTRASTIVITY (opposition) are made. John Newman shows that 'give' verbs are part of basic vocabulary as well as being complex in their internal semantic structure. These properties are relevant to understanding the impressive range of semantic extensions which these verbs show cross-linguistically. These extensions, including grammaticizations of 'give' verbs as prepositions, auxiliaries etc., are documented and classified into major subgroups, and some attempt is made to motivate each of these figurative extensions. Kiki Nikiforidou examines the semantics of action/state nominalizations in English and Modern Greek and proposes that at a cross-lexemic level, nominalizations have meanings such as the product or result, the manner (in which an action was performed), the degree (to which a state holds) and the fact (that the action/state occurred) which are fairly regular and systematic. She suggests that these meanings are related by metonymy to the basic action or state meaning of a given nominal. Although some of these meanings (product and result) represent semantic extensions for some nominals while others (manner, degree, fact) should be best viewed as pragmatic variants, they are all regular enough to warrant description through general metonymic principles which turn out to be quite productive. She shows the relevance of her research for lexicography, proposing that the lexicographic treatment of nominalizations, which appears to have been rather unsystematic, should take into account such metonymic mappings in order to achieve consistency. Roy Ogawa and Gary Palmer take the concepts of Langacker (1991) and create a methodology for constructing the semantics of morphemes of spatial relations from the Coeur d'Alene language. The semantics are described by Langacker's networks of Schemas. The results of their analysis are applied to the prefixes /*it/, /t-/ and/*-/, all of which are glossed with some usage of the English word On'. Their methodology consists of the following. For

viii

Preface

each prefix, a list of words that contain it are collected. A spatial relation is extracted from the meaning of each word and is schematized. The words are then subcategorized according to these spatio-relational Schemas. These subcategories, each associated with a common spatio-relational schema, are arranged in a network where the Schemas are related by extension or elaboration. Each network may include more than one connected component, each of which represents a meaning-complex of the prefix made up of related meanings of the prefix. The separate components represent unrelated meanings of the prefix. In the case of /*it-/ two meaning-complexes make up the whole group; there are four subcategories for /t-/ with two meaningcomplexes; and, there are five subcategories with two meaning-complexes for /*-/. Results show all of the previous glosses to be either incorrect or incomplete and better glosses are given. The paper by Sally Rice forms part of a large-scale empirical investigation into the syntax and semantics of the English prepositions, the theoretical implications that their multivariate behaviour has on models of lexical representation, and, given their polysemy, whether the prepositions form a homogeneous or heterogeneous word class. In the paper she focuses on aspectual properties of prepositions using concepts from Cognitive Grammar. In Japanese a limited number of morphemes, called "relational nouns" (RN's), occupy a central place in locational expressions. RN's occur as the head of a "noun + particle, no + RN" noun phrase, followed by the locational particle ni 'at'. They also occur without the first two morphemes. In her paper, Yoshiko Tagashira examines the characteristics of the seven RN's and those of the aforementioned structures built around them. The following findings, which may shed new light on the question of how people conceptualize locations and on a few location-derived concepts, follow from her research: (1) RN's denote regions or areas in our basic conceptualization of the environment and this conceptualization is mapped onto locational relations within an object or within the abstract world of human society, etc. (2) In most cases, the noun which precedes the no Of preceding the RN thereby modifying it, represents PLACE rather than OBJECT. (3) The ground of a locational expression designates a PLACE and the RN specifies a particular dimension or facet within that PLACE. (4) Grammaticalization has taken place and some RN's are used to express such abstract concepts as TIME, QUALITY, and CAUSE. (5) The function of the RN stated above in (3) explains occurrence or non-occurrence of the RN's. (6) Two related but distinct lines in which the use of RN's is expanded should be recognised: (i) the orientational opposi-

Preface

ix

tions existing in our environment are mapped onto abstract areas; (ii) RN's are used with a PLACE noun and specify a particular dimension or facet within that PLACE. In a paper crossing the boundary between syntax and morphology, Adger Williams presents an analysis of the Russian reflexive marker -sja. The paper shows that this analysis has interesting implications for the analysis of reflexives in particular, but also for grammatical theory within the cognitive framework in general. He outlines an experiment to verify the hypothesis that the Russian marker -sja is, in certain cases, used to mark a predicate that presents a situation prototypical of the verb that the marker is affixed to. He then argues that this implies that schematic networks of the categories symbolized by verbs in Russian must have two different kinds of nodes, and that the grammar recognizes this distinction by affixation of -sja to a verb that is used to present a situation symbolized by a marked node. Finally, he shows that this development represents another instance of grammaticalization of a former reflexive marker. Research in the field of metaphor has been central to the development of the cognitive paradigm, as the contributions by Cienki, Delaney and Emanatian, Hilferty, and Jäkel attest. Alan Cienki's discussion on left/right polarity mediates neatly between the sections on lexical semantics and metaphor. In Russian, as in many languages of the world, there is a very strong semantic differentiation between non-spatial meanings of the terms for "left" and those for "right". "Right" is often used for reference to what is correct or proper, authority, a straight line; while "left" has negative connotations, often referring to something being wrong, illegal, crooked, or weak. However, the left-right spatial axis is very weakly polarized in terms of the human body and our daily conscious functioning. Research in the field of cognitive linguistics has shown that the metaphorical usage of spatial terms is normally based on the transformation of an 'image-schema', some structure or pattern which organizes our experience. The question is, what is the basis for the strong semantic polarity of the two terms, given their relatively weak markedness in the spatial realm? Cienki argues that left and right metaphors rely heavily on force-dynamic concepts of strength and weakness. Using Russian for a case study, he considers diachronic semantic evidence that shows an association of positive and negative concepts (primarily of strength and weakness) with the right and left hands respectively, and demonstrates how this can inform a synchronic semantic analysis of the Russian roots -lev- (left) and -prav- (right).

X

Preface

The paper by David Delaney and Michele Emanatian explores the social significance of metaphorical reasoning. They examine a set of judicial decisions concerning racial residential segregation in America in the 1930s and 1940s. Legal reasoning is conventionally considered to be a rather rigorous and highly constrained mode of thought, one with obviously significant and consequential products: judicial opinions. They show that a select, coherent system of metaphors shapes thinking and influences judgment. In the U.S. throughout the 20th century different people have sought to reinforce or dismantle spatial patterns of racial segregation by bringing claims to the legal arena. Analysis of numerous legal cases reveals the repeated reliance on and manipulation of complex metaphors from the domain of Plant Ecology to model and explain urban form and process. Racial segregation and racism itself are 'naturalized' through the application of ecological metaphors such as "adaptation", "competition", and "succession". Their study also illustrates what they call the 'filter effect'. This idea describes the successive winnowing of conceptual richness that results when the immediate source domain (in this case, Plant Ecology) is itself a target domain of metaphors derived from elsewhere (here, geo-politics and militarism). They show in their paper that the selection of plant ecological metaphors not only constrains legal reasoning, but conditions how judges conceptualize causation, the validity of contracts, and responsibility. Joseph Hilferty discusses various metaphorical uses of the English preposition through. He first describes through^ prototypical spatial sense, denoting TRAVERSAL, then showing how some of through^ metaphorical uses signifying MEANS are grounded in its principal spatial sense. Hilferty examines two interesting special cases indicating CAUSAL MEANS and PROCEDURAL MEANS, and discusses the semantic constraints that these two uses obey. Finally, he concludes that prepositional metaphor is clearly not beyond the pale of linguistic inquiry, contrary to what some linguists have intimated in the past. This conclusion follows from the remarkable coherency of through*s metaphorical manifestations, which in turn are motivated by its prototypical spatial meaning. The Cognitive Theory of Metaphor, as developed by George Lakoff and Mark Johnson, maintains as one of its basic tenets the unidirectionality of metaphorical projection. Given that metaphor permits the understanding of one conceptual domain in terms of another, the claim is that in this metaphorical junction of concepts, the source domain is in general more concrete and physical than the target domain. Even if exceptions are found, this unidirec-

Preface

xi

tionality hypothesis would be a powerful statement of a very strong general preference. In fact, it represents one of the fundamental pillars of Lakoff's and Johnson's theory of "Experientialism". Olaf Jäkel's investigation focuses on an empirically testable hypothesis which can be derived from the more general unidirectionality hypothesis: Metaphors deviating from the generally preferred direction should prove harder to understand. This empirical hypothesis is tested in a combined quantitative and qualitative investigation. Using a detailed questionaire, a group of 39 informants (all native speakers of German) were asked to pass judgment on a sample of ten metaphors. The paper explains the make-up of this investigation and discusses some of its results, the most important of which is the confirmation of the hypothesis that metaphors deviating from the generally preferred direction of transfer should prove harder to understand. Quite a number of contributions deal with aspects of grammar. The first paper in this section, by Kenneth Cook, demonstrates that Samoan (an ergative Polynesian language) codes active zones in the syntactically prominent position of absolutive. This is in contrast to the general tendency in languages like English to code cognitively salient entities in syntactically prominent positions. This Samoan coding preference and its ergative case marking system are claimed to be manifestations of a tendency in Samoan to code events "from the inside out." Nicole Delbecque proposes that Spanish has developed two transitive construction frames, one prepositional, the other not, each of which has its own semantics. The point of departure for her study is that the meaning of a construction does not only result from the interaction between the semantics of the verb and that of the event participants, but that it also, most crucially, hinges upon the meaning imposed by the construction frame as a holistic cognitive unit. This principle of cognitive grammar, according to which abstract meaning is integrated in syntactic constructions, leads us to believe that the presence vs. absence of the Spanish preposition a cannot be satisfactorily explained in terms of determinisms which operate at the level of the direct object complement. She suggests that these two constructions yield a different conceptualization of the global argumentative structure. In the non-prepositional configuration, the accusative is plainly integrated in the relational predication, instantiated by the VP, whereas the prepositional configuration signals the relevance of a proper reference domain, which is not bounded by the verbal predication. In Swedish a particular type of morpho-syntactic construction, such asfalla

xii

Preface

I sömn 'fall into sleep', is used productively to express the transition into a state or an activity. At the same time these constructions, referred to as Abstract Transitional Phrases (ATPs), behave like more or less syntactically frozen idioms. In her paper, Lena Ekberg discusses this construction type, showing that they also have properties in common with productively derived words. In each of the larger semantic domains where ATPs occur, there is at least one prototypical phrase consisting of an invariable combination of verb and preposition, and an archilexeme. Semantically, the verb-preposition combination behaves like a derivational affix. There are also gaps in the formation of ATPs, resembling those found in word formation, which seem to be due to the lack of prototypical patterns, which in practice determine the coining of new ATPs. The study of ATPs shows that a linguistic expression simultaneously may be productive and syntactically frozen, and simultaneously may have the status of a lexical unit and a semantically compositional syntactic structure, supporting the assumption of Cognitive Grammar that traditional linguistic levels, such as syntax and lexicon, are placed along a continuum, rather than being distinct. Jose Garcia-Miguel reiterates the point that grammatical relations like subject and object assign a special status to certain constituents of the clause. Some conceptual frameworks for linguistic analysis allow only two places for the participants which are given special prominence. This is the case, for example, of subject and object 'syntactic functions' in Dik's Functional Grammar. Within Cognitive Grammar, Langacker has regarded subject and object as, respectively, the most prominent clausal participant and the secondmost prominent clausal participant (Langacker 1991: 321), but, in his opinion, "indirect object ought not to be considered a grammatical relation of the same type as subject and direct object" (Langacker 1991: 326). In his paper, Garcia-Miguel opts for an analysis that supports the idea that subject, direct object and indirect object are all three central participants in Spanish clauses. In the paper by Jose Sanders it is argued that variation in degree of subjectivity, as expressed in speakers' choices in epistemic qualification and speech representation, can be explained by the concept of subjectification. A model of subjectification (Langacker 1990) is applied to epistemic predication and perspective predication. Perspective predicates, especially various means of representation of speech and thought, such as the direct and indirect mode, express different degrees of subjectivity. In a similar way, modal predicates can be distinguished with respect to the amount of subjectivity they express. It is claimed that variation along the scale of subjectivity in both phenomena

Preface

xiii

can be clarified by describing them in terms of mental space theory (Fauconnier 1985). This approach places subjectivity in a framework of similar linguistic phenomena such as temporal and local modifications. According to Victoria Väzques Rozas, the analysis of the variation in the syntactic coding of the object in Spanish biactant clauses supports the idea that, in the expression of events with two participants, the transitive semantic prototype is correlated with an unmarked (or less marked) syntactic construction, whereas a certain degree of departure from the prototype is matched with a marked (or more marked) pattern. To verify this, certain semantic features of transitive clauses are compared with those of SUBJECT-INDIRECT OBJECT clauses, and systematic differences of meaning coherent with the prototype-model of the transitivity notion are found. In the second place, it is shown that the pattern SUBJECT-INDIRECT OBJECT is syntactically marked in contrast with the unmarked construction SUBJECT-DIRECT OBJECT Marjolijn Verspoor's paper deals mainly with to infinitives used as verbal complements of verbs like order and believe as in / order him to go and related constructions such as that clauses and small clauses as in 1 believe that he is honest. She shows that both epistemic and deontic verbs may take that clauses, to infinitives and small clauses as their complement. For both types, the that clause expresses some sort of non-directness. With an epistemic verb it expresses that an opinion is not based on direct personal or experiential knowledge. With a deontic verb, a that clause expresses an order that is not directly given to the subject of the that clause. A to infinitive expresses some sort of directness. With an epistemic verb it expresses that an opinion is directly based on some personal or experiential knowledge and the to expresses that X is moving towards a categorial state Y. With a deontic verb it expresses that there is some direct contact between the subject of the main clause and the subject of the complement clause. The to expresses the goal of an intention. A small clause expresses some sort of direct and immediate causation. The speech act, the thought, or the action causes a categorial state of affairs tobe. The paper by Tuija Virtanen deals with adverbial placement in written English from the perspective of iconicity. The focus is on clause-initial adverbials of time and place signalling a temporal or locative text strategy. Adverbial placement is first considered in terms of 'experiential iconicity', which refers to instances where the linear ordering in the text forms an icon of our experience of the world. This discussion is briefly related to text types. Secondly,

xiv

Preface

adverbials of time and place signalling a temporal or locative text strategy are examined in terms of what might be called 'structural iconicity', i.e., reflections of the hierarchy of text structure in the number and/or size and information status of the material used to signal textual boundaries. Finally, the notion of 'iconicity' is given some attention, as it may easily prove too broad, encompassing such a wide range of phenomena that it risks losing its explanatory value in the analysis of authentic texts. The central question in the paper by Alfons Maes and Leonoor Oversteegen is to what extent nominal and temporal interpretative phenomena in discourse are analogous. They first discuss three apparent similarities between nominal and temporal reference. They then deal with the ontological differences between temporal and nominal referents and their linguistic realisations in discourse. Eventually these differences are used as a kind of search light in discovering, describing and explaining three discourse phenomena in which nominale and temporals behave differently. Willy van Langendonck deals with the notion of markedness in an experiential framework. Following Mayerthaler (1980, 1988), he derives unmarked categories from prototypical speaker attributes. Adapting and extending this concept, he applies it to phonology and to the nominal area of personal pronouns and proper names, while also paying attention to the much discussed phenomenon of markedness assimilation. Cognitive Grammar has also made an inroad into the field of computational linguistics. The contribution by Kenneth Holmqvist presents an on-going research project concerned with the development of an experimental computer implementation in order to investigate and improve a processual language understanding model based mainly on Langacker's Cognitive Grammar. The article contains two main parts. In the first part Holmqvist sketches the computer representation of schemata and the superimposition processes for constructing and evaluating composite schemata. In the second part, he closely examines a grammatical valence suggestion model based on Behaghel's principle, showing how it segments the incoming morpheme stream and describing how it can be easily integrated with the semantic superimposition processes.

Preface

xv

A word of appreciation I wish to thank the members of the editorial committee who had the major task of refereeing the contributions to this volume. They are Eugene Casad, Kristin Davidse, Nicole Delbecque, Dirk Geeraerts, Adele Goldberg, Rufus Gouws, Laura Janda, Arthur Mettinger, Fritz Ponelis, Brygida RudzkaOstyn, Piet Swanepoel, and Eve Sweetser. A special word of thanks goes to Rene Dirven and Eugene Casad for all their advice and support, and to Anke Beck and Katja Huder of the publishers for technical and other assistance, and to Christoph Eyrich who, in spring 1998, joined the project as technical editor. Last, but not least, I need to thank all the contributors to this volume. The project was plagued by major infrastructural problems, taxing a lot of people's patience, and I can only respond with a sincere word of appreciation for all the support that I received. Stellenbosch

Leon G. de Stadier

Contents Preface

v

Theoretical issues Bridges between generative and cognitive linguistics Frederick J. Newmeyer

3

Lexical semantics and morphology A modular approach to the semantics of space in language Michel Aurnague and Laure Vieu

23

Beer and semantics Dirk Geeraerts

35

Is everything black and white in conceptual oppositions? Gabor Györi and Iren Hegedüs

57

Categorization and analogical change: The case of athematic Isg -m in the Slavic languages Laura A. Janda Contrast and Schemas: Antonymous adjectives Arthur Mettinger

75

97

Figurative giving John Newman

113

Nominalizations, metonymy and lexicographic practice Kiki Nikiforidou

141

Langacker semantics for three Coeur d'Alene prefixes glossed as On' Roy H. Ogawa and Gary B. Palmer

165

xviii

Contents

Aspects of prepositions and prepositional aspect Sally A. Rice

225

Some aspects of relational nouns Yoshiko Tagashira

249

Prototype marker or reflexive marker: Russian -sja and categorical change Adger Williams

277

Metaphor The strengths and weaknesses of the left/right polarity in Russian: diachronic and synchronic semantic analyses Alan Cienki

299

'Unnatural barriers': Why metaphor matters (or, linguistics meets the geopolitics of law) David Delaney and Michele Emanatian

331

Through as a means to metaphor Joseph Hilferty Is metaphor really a one-way street? One of the basic tenets of the cognitive theory of metaphor put to the test Olaf Jäkel

347

367

Syntax and semantics Samoan as an active zone language1 Kenneth William Cook Two transitive construction frames in Spanish: The prepositional and the non-prepositional accusative Nicole Delbecque

391

407

Contents

xix

Swedish abstract transitional phrases: An in-between phenomenon in the linguistic system Lena Ekberg

425

Grammatical relations in Spanish triactant clauses Jose M. Garcia-Miguel

447

Degrees of subjectivity in epistemic modals and perspective representation Jose Sanders

471

Biactant Spanish clauses. Syntactic markedness and semantic prototype Victoria Vazquez Rozas

491

To infinitives2 Marjolijn Verspoor

505

Adverbial placement and iconicity Tuija Virtanen

527

Pragmatics Nominal vs. temporal interpretation in discourse Alfons Maes and Leonoor Oversteegen

549

Markedness and prototypical speaker attributes Willy Van Langendonck

567

Computational Linguistics Implementing cognitive semantics - overview of the semantic composition processes and insights into the grammatical composition processes Kenneth Holmqvist Subject index

579

601

Theoretical issues

Bridges between generative and cognitive linguistics Frederick J. Newmeyer

Langacker (1991: 532) addresses the possibility of a convergence between cognitive grammar and generative approaches to language. While he feels that certain trends in generative linguistics have narrowed the gap between the two, basic philosophical differences remain. The most serious is the autonomy thesis of generative grammar, which stands in contrast to 'the idea that grammar might reduce to symbolic relationships' (1991: 533). I am sure most generative grammarians welcome Langacker's willingness to open a dialogue. Actually, before I had read the second volume of his Foundations of Cognitive Grammar, I wrote a piece that has since appeared in Language (Newmeyer 1992) that is very much in the spirit of his ecumenical appeal. I took the concept of 'iconicity', a notion central to functionalist and cognitive approaches, and demonstrated that the structure of the generative model is well designed to capture iconic relationships between form and meaning. I even argued that the abstract levels of structure in a 'GB-style' grammar provide a direct means of representing iconic relations in language. In short, certain form-meaning parallels are as congenial to generative linguistics as to cognitive linguistics, and therefore not a point by which one might be judged as superior to the other. In this paper I will explore the extent to which the differences between the theories are merely the result of differences in focus or even terminology and the extent to which they are substantitive. I hope to demonstrate that some of the greatest stumbling blocks to mutual understanding are the result of pseudo-issues rising to the fore. One can imagine four distinct ways that generative and cognitive linguistics might be more in alignment than is generally believed: (A)

At an abstract level, generative and cognitive linguistics are simply notational variants of each other.

(B)

Generative and cognitive linguistics are not notational variants, but are considerably closer than is generally thought.

4

Frederick J. Newmeyer

(C)

Some theoretical constructs of cognitive grammar interface in the same overall theory with some posited by generative grammar.

(D)

Generative and cognitive linguistics are both true, describing different aspects of cognitive functioning. This could be effected in two ways: 1. The child acquires the structures and principles of both theories 'side-by-side'. 2. Generative and cognitive linguistics describe different levels of the same reality.

I will discuss each of the four options in turn. Option (A) is clearly false, since the generative computational system has no direct counterpart in cognitive linguistics. I will now argue that Option (B) is true; many of the differences that separate generative from cognitive linguistics are more apparent than real. Perhaps the most pervasive of pseudo-issues revolves around what Lakoff has called the 'cognitive commitment', namely 'the commitment to make one's account of human language accord with what is generally known about the mind and brain from disciplines other than linguistics' (Lakoff 1991: 54). In Lakoff's opinion, cognitive linguistics is defined by the cognitive commitment, while generative grammar rejects it. If what he means is that generativists reject the idea that linguistic theory must be situated in a neuropsychologically real overall theory of mind-brain, he is mistaken. In Chomsky's words, 'a grammar is a cognitive structure interacting with other systems of knowledge and belief (Chomsky 1975: 86). It has been suggested that the cognitive commitment is incompatible with grammar being a formal system. But grammar as a formal system no more requires the separation of syntax from cognition than a formal phonological theory prohibits the phonetic grounding of phonological constructs. More than one cognitive linguist has suggested that model-theoretic semantics is entailed by generative grammar. But many generativists also reject model-theoretic semantics - in particular it has been rejected by Chomsky and Jackendoff. As a matter of fact, in a series of books and articles, Jackendoff has developed a theory of conceptual semantics, with features remarkably similar to the semantic conceptions of cognitive linguistics (see especially Jackendoff 1983; 1990; 1992; Jackendoff and Aaron 1991). Most importantly, Jackendoff sees the goal of a semantic theory to provide a mental representation of the world in relation to language. The fact that Jackendoff

Bridges between generative and cognitive linguistics

5

adopts an autonomous generative syntax shows that a nonobjectivist semantics and an autonomous syntax can go hand-in-hand. Cognitive linguists, quite properly, cite results from neuropsychology, concept acquisition studies, and so on to bolster their view of language. But there are at least as many studies that favor the generativist orientation. To take one example, event-related brain potentials elicited during comprehension distinguish between syntactic anomaly and semantic anomaly remarkably along the lines suggested in much current generative work (Osterhout and Holcomb 1992). The numerous examples of dissociations that have been reported to exist between grammatical abilities and other cognitive abilities provide strong support for the autonomy hypothesis. Among aphasic patients, examples have been found, for example, of those who have lost the ability to encode thematic roles within the grammatical system, but who maintain knowledge of the conceptual relations themselves (Caplan and Hildebrandt 1988); those with lexical and semantic deficits, but with relatively little grammatical loss (Riddoch, Humphreys, Coltheart, and Funnell 1988); and those with massive pragmatic deficits leaving the grammatical system intact (Kaplan, Brownell, Jacobs, and Gardner 1990). Analogously, numerous cases of developmental dissociations between syntax and other aspects of cognition have been reported. The best known, of course, is the case of Genie (Curtiss 1977); more recently Gopnik (1990) has not only identified a specific grammatical impairment, but has pointed to a possible genetic basis for it. More and more cases of the converse phenomenon have also been reported, namely of children with full grammatical abilities, but massive developmental deficits in other aspects of cognition (Cossu and Marshall 1990; Yamada 1990; Smith and Tsimpli 1991). Dissociations of cognitive abilities are not limited to language. For examples, studies of autistic children (Leslie and Thaiss 1992) has suggested the existence of a variety of cognitive mechanisms, which, in pathological cases, are capable of individual breakdown. And take the visual system, whose representational and computational structure is remarkably similar to that hypothesized for the linguistic system by generativists (see Marr 1982 and subsequent work). It contains distinct 'modules' analyzing color, form, motion, depth, and so on, which then interact to yield the mental image of the object perceived. Along the same lines, facial recognition is quite autonomous with respect to general visual perception. Other domain-specific innate cognitive abilities include perception of caus-

6

Frederick J. Newmeyer

ality and objects as unified entities, categorical perception of speech sounds, physical shapes, facial expressions, and a variety of imitative abilities. Now I am well aware that cognitive grammarians also point to studies from neurology and psychology that lead them to precisely the opposite conclusion. The problem is that the fields to which we linguists often look for confirmation of our hypotheses are as riddled with disputes, unclarities, and difficulties of interpretation of experimental data as is the field of theoretical linguistics itself. We are naturally tempted to find some seemingly congenial result from work in these fields and raise it in support of our favorite linguistic hypothesis. But we have to bear in mind that controversial results can only with the utmost care be applied in the support of controversial hypotheses. In short, generative and cognitive linguistics stand united in their devotion to the 'cognitive commitment'. What stands in the way of a reconciliation between the two approaches is not the principled question of whether a theory of language has to be 'psychologically real', but the empirical one of how the results of linguistics are properly situated within a broader theory of mind/brain. Of course cognitive linguistics rejects the 'autonomy thesis' of generative grammar. According to this thesis there is a separate component of our knowledge, the grammar, which is not reducible to other forms of knowledge. While one can hardly dispute the fact that this hypothesis is the major point of contention between cognitive and generative linguistics, autonomy is often assumed to entail incorrect subsidiary hypotheses that serve to magnify unjustly their differences. These hypotheses are stated in (1): (1)

a. b.

Autonomy entails that there can be no commonalities between grammar and other cognitive faculties. Autonomy entails a high degree of arbitrariness between form and meaning.

Neither of these hypotheses are correct. As far as (la) is concerned, Jackendoff, who explicitly defends the autonomy thesis and Leonard Talmy, who takes no position on it, agree that there is a fundamental core to conceptual structure that is common across cognitive domains, and both have pointed in particular to structural commonalities between the grammatical and the visual system. Talmy even writes of grammar as 'the determinant of conceptual structure within one cognitive domain, language' (Talmy 1988: 200), a wording thoroughly in keeping with the spirit of the autonomy hypothesis.

Bridges between generative and cognitive linguistics

1

A great deal of the unnecessary invective between generative and cognitive linguistics has revolved around (Ib), the question of the closeness of the fit between form and meaning. Many of the arguments devoted to supporting autonomy have been based on the full or partial arbitrariness of distributional classes with respect to meaning. On the other hand, works devoted to demolishing autonomy have primarily focused on showing the generally close fit between form and meaning, in particular to showing that the occurrence of formal elements tends to be linked to particular semantic effects. Both strategies are illegitimate for one simple reason: the autonomy thesis entails no claim whatever about the closeness of the fit between form and meaning. Suppose that the fit between the two were utterly perfect, that every grammatical element had an invariable meaning. Would that render autonomy incorrect? Not necessarily, since there still might be evidence for a grammatical system governed by nonsemantic principles linked in a trivially simple fashion to the semantic system. Or suppose that the relationship between form and meaning were utterly chaotic. Would that support autonomy? Not necessarily; even so, the optimal description of form-meaning relations might involve a vast number of individual Schemas along the lines that Langacker proposes to represent irregular English past tense forms. In short, it takes more than pointing to a fit, or lack of a fit, between form and meaning to impact the autonomy thesis. A case in point is Goldberg's study of the English ditransitive construction (Goldberg 1989), which has been cited as demonstrating the nonautonomy of syntax. Goldberg argues that there is a central sense to the V NP NP construction, represented by such sentences as Jo gave Bill an apple, which involves transfer of a physical object to a recipient. Additionally, there are five major classes of extensions, based on different types of metaphorical transfer. There is no problem preserving Goldberg's basic generalizations within an autonomous syntax framework, should one wish to. The principles of the syntax overgenerate structures with two noun phrases following the verb. Give and verbs of its class are specified as occurring in such structures along with a characterization of the semantic properties of their arguments. Principles of metaphorical extension at work in the semantics derive the lexical entries of the verbs in nonbasic classes (along with their relevant semantic properties). In other words, simply showing that there is an intimate relation between form and meaning or even that metaphors have grammatical consequences is not sufficient to defeat autonomy.

8

Frederick J. Newmeyer

Let us take another example. On the basis of work on the preposed negative adverb construction with Claudia Brugman (Lakoff and Brugman 1987), Lakoff (1991) has argued that autonomy is refuted by the fact that you find auxiliary inversion only when the nonoccurrence of the main clause event is entailed: (2)

a. b.

For no money would she sky-dive. (She wouldn't.) For no money, she would sky-dive. (She would.)

Suppose that this were the extent of the relevant data. How would it bear on the autonomy thesis? The answer is 'Not in the least'. The rules linking syntactic and semantic structures would simply specify the relevant generalization. For facts such as these to challenge the autonomy thesis one would minimially have to demonstrate that this particular syntax-meaning pairing is a nonarbitrary one, that is, to show that there is some reason rooted in grammar-external facts why there should be a relationship between inversion and entailment. One thing to keep in mind is that in investigating the nature of the fit between form and meaning, we need a neutral vocabulary, that is, one that does not presuppose any particular degree of relationship between the two. Unfortunately, the question has been posed within cognitive linguistics in such a way as to inevitably lead to the conclusion that form and meaning are inseparable. This has been achieved by taking the notion '(grammatical) construction' as the focal point. But within cognitive linguistics, constructions are defined as 'pairings of form and meaning'. Clearly, if constructions are defined in such a way, it will hardly be a surprising discovery that the bond between the two is tight. Let us examine the inversion data instead in terms of the more neutral entity 'structures', instead of 'constructions'. The first thing that we notice is that the inverted auxiliary structure is not restricted to preposed negative adverbs: (3)

a. b.

Have you been working late? What have you been eating?

(4)

a. I wondered whether you had been working late, b. *I wondered whether had you been working late.

(5)

a. I wondered what you had been eating, b. *I wondered what had you been eating.

(6)

a.

What has been bothering you?

Bridges between generative and cognitive linguistics

9

b. *Has what been bothering you? (7)

a. Under no circumstances will I take a day off. b. *Under no circumstances I will take a day off.

(8)

a. *Given any possibility will I take a day off. b. Given any possibility I will take a day off.

(9)

a. Had I known the dangers, I would have kept my distance, b. *If had I known the dangers, I would have kept my distance.

(10)

a. So tall is Mary, she can see into second story windows, b. *So tall Mary is, she can see into second story windows.

Does this one-to-many relationship between form and meaning support autonomy? In and of itself, the answer is 'no'. Cognitive linguistics would presumably posit a number of constructions, each specifying the inversion structure. Generativists would posit principles that generate the structure and link that structure to a number of different interpretations. What gives the autonomist account greater plausibility is the fact that the inversion constructions share formal idiosyncracies. For example, each allows only one auxiliary element to occur in inverted position, as (11) illustrates: (11)

a. b. c. d.

*Have been you working late? *What have been you eating? *Under no circumstances, will be I taking a leave of absence. *Had been I thinking about the dangers, I never would have done that. e. *So competent has been Mary, she'll surely get the promotion.

Anyone taking the cognitive linguistics-construction-based approach has to explain why the same restriction occurs on each construction. Clearly, the language learner is not required to learn this restriction anew each time a new meaning for the structure is learned. The obvious generalization is that the restriction is linked to the structure itself, and is not part of each construction in which the structure participates. Thus the autonomy thesis, which characterizes structures, not constructions, receives support. The kind of grammar-internal evidence that bears on the validity of the autonomy thesis, then, is evidence that knowledge of a language consists in part of internalized generalizations about linguistic form. These are incompatible with what Langacker has called the 'central claim of cognitive grammar [...]

10

Frederick J. Newmeyer

that language is fully describable in terms of semantic structures, phonological structures, and symbolic links between the two' (Langacker 1991: 514). While autonomy might be the bigggest theoretical stumbling block to a rapproachement between cognitive and generative grammar, the biggest methodological stumbling block is the question on whether to focus on individual constructions. Let me give an example from Lakoff's challenging treatment of deictic there (Lakoff 1987). In his view, there is not just one such construction, but a whole family of them, more than half a dozen, centered around the prototypical construction (the 'central deictic') in which initial deictic there (as in example 12) designates a location in physical space: (12)

There's Harry with the red jacket on.

From a succinct characterization of how the noncentral constructions differ from the central one, the appropriate 'metaphors' are triggered, which in turn serve to predict the distinct syntactic properties of each construction. One of the distinct constructions that Lakoff posits is the 'perceptual deictic', at work in sentences such as (13a-d): (13)

a. b. c. d.

There's the beep. There goes the bell now! Here comes the beep. There goes that throbbing in my head again.

This construction is given the representation in (14), which triggers the application of the metaphors in (15a-c) and a metonymy (15d): (14)

The Perceptual Deictic Based on : The Central Deictic About: NONVISUAL PERCEPTION

(15)

a. b. c. d.

NONVISUAL PERCEPTUAL SPACE IS PHYSICAL SPACE; PERCEPTS ARE ENTITIES REALIZED IS DISTAL; SOON-TO-BE-REALIZED IS PROXIMAL ACTIVATION IS MOTION THE THING PERCEIVED STANDS FOR THE PERCEPT WHILE THE PERCEPTION IS IN PROGRESS

Hence sentences such as (13a-d) are captured: There is understood in a

Bridges between generative and cognitive linguistics

11

nonvisual perceptual sense, go is used to denote activation, rather than motion, and so on. How would a generativist approach this problem then? In the absence of evidence to the contrary, the simplest assumption is that independently motivated principles of grammar, meaning, and use interact to derive the formal and semantic properties of sentences involving there. What about the meanings of the lexical items involved, in particular that of there/here and come/got The simplest assumption for the former is that they are deictic locatives, and that the latter are verbs of motion. But, then, what about the fact that the deictics in (13) do not refer to literal physical location, but rather a metaphorical location in 'perceptual space'? I would follow Jackendoff (especially Jackendoff and Aaron 1991) in arguing that we do not have true metaphor here at all, but rather thematic parallelism. Jackendoff argues that the structural parallels between our encoding of physical and perceptual space (for example) are part of our underlying mental resources, not a contingent by-product of metaphorical extension. In other words, Jackendoff's hypothesis allows us to eliminate the need for a distinct perceptual deictic construction and to purge the syntax of the need to make appeal to metaphorical extension. While Lakoff does not address Jackendoff's work directly, he does put forward several arguments which would appear to argue directly against such a solution. The first is based on the contrast in (16) and (17): (16) (17)

a. There's Harry, b. Harry is there. a. There's the beep, b. *The beep is there.

Lakoff concludes that this shows that the perceptual sense of there occurs only in the perceptual deictic construction. If so, meanings of words are not independent of the grammatical constructions they occur in, and the analysis I sketched above would not work. But in fact there is a construction-independent reason for the impossibility of (17b). Harry and the beep in (16a) and (17a) are in focus position; roughly, they serve to call attention to the hearer of a significant event (the location of Harry, the awaited sound of the beep). In the (b) sentences, however, the noun phrases are in discourse topic position; Harry has been the subject of conversation, perhaps, in (16b) and the speaker wishes to point out his location to the hearer. The oddness of (17b) comes from the fact that transitory

12

Frederick J. Newmeyer

beeps make poor discourse topics - for reasons having nothing to do with deictics per se. One can create a scenario in which the acceptability of (17b) is dramatically improved - say if one had finally identified the source of an annoying beeping sound. Consider the following exchange: (18)

a. b.

Have you figured out where inside the wall that beep is coming from? As far as I can tell, the beep is there.

Lakoff further argues that it is necessary for the construction specifically to refer to 'nonvisual perception' because of the contrast in (19): (19)

a. There's the sound of the bell, b. *There's the sight of the bell.

I would suggest that the difference between (19a) and (19b) follows from independent facts about what can be conceptualized as occupying a location in one's spatial or perceptual field. In folk perceptual psychology, sounds are conceived of as existing independently (in space and time) from the perception of them, while sights are not. Notice that one can say Where is that sound coming from?, but not Where is that sight coming from! Therefore (19a) is fine, and (19b) deviant. Memories of sounds do occupy (mental) space, and so can be described using deictic there, as in (20). On the other hand, auditory perceptions, as opposed to the sounds themselves, are not conceived of as occupying space, and so (21) is impossible. (20) (21)

There's a sight that always gives me pleasure. *There's the auditory perception of the bell.

In short, the representation of the perceptual deictic does not have to refer specifically to nonvisual perception. Lakoff's last argument for a specific perceptual deictic construction is a little bit more complicated and is based on the metonymy (15d). Consider (22a-b): (22)

a. b.

There goes the beep. There goes the alarm clock.

The alarm clock stands for the sound it makes. Or, similarly, consider (23a-b):

Bridges between generative and cognitive linguistics (23)

a. b.

13

There goes the pain in my knee, There goes my knee.

The knee stands for the pain in it. But, Lakoff argues, the perception has to be in progress for this to work. So, according to Lakoff, (24a) is fine if the beep is about to happen, but not (24b): (24)

a. Here comes the beep. b. *Here comes the alarm clock

Likewise, according to Lakoff, (25a) is good if we sense that the knee pain is on its way, but not (25b): (25)

a. Here comes the pain in my knee, b. *Here comes my knee.

So Lakoff concludes that a special construction is needed embodying metonymy (15d) - this in his view is the only way to rule out (24b) and (25b). Again, Lakoff fails to consider that there is a simple discourse explanation for the strangeness of these two sentences. Under normal every-day conditions, both alarm clocks and knees are in our field of awareness before they go beep or cause us pain. So (24b) and (25b) have readings where they are factually contradictory - something can't be 'coming' if it is there all along. But notice how much better (24b) is than (25b). An alarm clock can be hidden from view - if it is, then (24b) is as good as (24a). But a knee can't be so dismissed; it is part of you! It is almost impossible to contextualize (25b) without a contradictory reading. It is important to stress that there is nothing inherent to the foundations of cognitive linguistics that would make impossible the modular treatment of the perceptual deictic that I just sketched. But two factors seem to conspire to divert a cognitive linguistic analysis away from it. One is its focus on the properties of individual constructions, which has the effect of downplaying any attempt to derive their complexity from the interaction of simple principles governing different domains. The other factor is particularly interesting, in that it might represent in cognitive linguistics the repetition of a stage in the history of generative grammar. Early generative grammarians were so enamored of transformational rules that they typically ignored obvious nonsyntactic - in particular discoursebased - explanations for unacceptability. Perhaps in cognitive linguistics we have an analogous over-appeal to cognition-based explanations, again, with

14

Frederick J. Newmeyer

a downplaying of the role of principles grounded in discourse and communication. Pushing modular explanations to their limit actually increases the possibilities for convergence between generative and cognitive linguistics. Let me illustrate. Contrasts like (26a-b) have always proved troublesome for purely syntactic treatments of extraction: (26)

a. Who did Mary write a book about? b. *Who did Mary destroy a book about?

The syntactic principle of subjacency predicts that both sentences should be ungrammatical. Only ad hoc unnatural treatments resulted from the attempt to provide a purely syntactic explanation for these sentences, as long as subjacency, in its 1970s manifestation, was assumed to be correct. More recent developments in generative grammar, however, have enabled the possibility of a more adequate modular account. In the framework of Chomsky's Barriers (Chomsky 1986) both (26a) and (26b) are grammatical. Briefly, extraction is permitted as long as the extracted element and those elements dominating it, are lexically governed. The problem then becomes to explain the deviance of (26b). But, in fact, explanations have been put forward not involving grammatical principles per se. Kuno 1987 and Deane 1988 (the latter working explicitly withing cognitive linguistics) attribute its strangeness to the semantic relations between the verb, the head noun, and the prepositional object. In other words, in generative terms, (26b) is unacceptable for nonsyntactic reasons. The issue, as Kuno and Deane stress, is not the existence of purely syntactic locality conditions - both accept that they exist but the proper balance between them and semantic and pragmatic principles. A modular account allows for the formulation of fully general principles from all the relevant domains. I think that many syntactic prototypicality effects, as well, can be reanalyzed in modular fashion - most of Ross' squishes are examples. Less prototypical NPs, for example, are those with more pragmatic conditions affecting their use. In other cases, say, the failure of prototypically active verbs to passivize in many languages, the possibility of passivization is an automatic consequence of principles needed elsewhere to link syntactic and semantic structure. Other apparent differences between generative and cognitive linguistics seem to rest more on differences of terminology and focus than on underlying theoretical differences. Consider, for example, the rejection by cognitive

Bridges between generative and cognitive linguistics

15

linguistics of the distinction between between grammar and lexicon. As I read the literature, this is more a point of emphasis than of theory. A lexicon is even defined in Langacker (1991) as 'the set of fixed expressions in a language', a definition that few generativists would find fault with. What about the distinction between semantics and pragmatics? Jackendoff's conceptual semantics also rejects the distinction, or, at least, assigns it little theoretical importance. Nevertheless, we can base a rough and ready distinction between semantics and pragmatics in the following manner: the former deals with those aspects of interpretation linked either to individual forms or to predicate-argument relations; the latter with those derived from broader extralinguistic principles. The usefulness of a distinction is not threatened by the fact that there are borderline cases, though, again, it's not clear that the distinction should be attributed any theoretical weight. Finally, it is worth calling attention to recent developments within generative grammar itself that have led it to resemble that of cognitive linguistics in ways that I suspect few cognitive linguists are aware. In Chomsky's recent 'minimalist program' (Chomsky 1992), all central grammatical principles apply directly on the surface or at the level where grammar interfaces with meaning; language particular variation is restricted to the lexicon. The de-emphasizing of the computational system in favor of the lexicon is a step in the direction of cognitive linguistics. Let us now consider Option (C), namely that the structures posited by cognitive linguistics interface in the same overall theory with those posited by generative grammar, in effect making cognitive linguistics the 'semantic component' of a generative grammar. That is, an autonomous syntax would be linked to the lexicon and meaning representations, preserving as much of the structure posited by cognitive linguistics as is possible. My sketch of how the Goldberg paper on ditransitives might be compatible with autonomy gives the flavor of this option. My feeling is that few on either side of the issue would find this an attractive alternative, and for reasons of space constraints, I will not discuss it. Option (D 1), the idea that cognitive and generative linguistics might share the mental limelight, is by no means a priori absurd. For example, in a number of publications, Jerrold Sadock has argued that there exists massive redundancy between the lexical and the computational part of the grammar (Sadock 1983; 1984; 1991). That is, the internal structure of stored lexical entries to a great extent matches the output of the generative rules and principles.

16

Frederick J. Newmeyer

He is careful to point out that a vast stock of complex lexical entries, many of phrase or sentence length, does not threaten generative grammar. Rather, they exist side-by-side; the proper characterization of a speaker's knowledge makes reference to both systems. It is easy to reinterpret the evidence in support of Sadock's conclusions in terms of the hypothesis at hand, namely, that both cognitive and generative linguistics have a mental reality. Pinker and Prince (1991) present interesting evidence that something along these lines might be right. They show that the psychological status of regular morphological processes is very much along the lines suggested in traditional generative treatments. But irregular morphology is quite different. The linkages between irregular pairs of related words are stored in an associative memory structure with certain connectionist-like properties, along the lines suggested in Bybee (1988). But they acknowledge that there is nothing stopping certain regular forms from being stored in the same way, and discuss certain circumstances in which just that might happpen. The effect, again, is in a certain sense a side-by-side generative and cognitive grammar. Option (D 2), cognitive and generative linguistics occupying different levels of the same reality, would involve deriving the principles of generative grammar from those of cognitive linguistics, much in the way that the BoyleCharles Law of classical thermodynamics has been derived from principles of statistical mechanics. However, the obstacles to this are absolutely formidable: neither theory has really been formulated clearly enough to make this practical; also there is the fact that the explananda of the two theories differ in ways that might prove daunting to the task. Furthermore, there is hardly unity among philosophers and cognitive scientists about whether theory reduction in psychology is desirable or, if desirable, in principle possible. So, for example, the program for reducing theories of psychology to theories of neuroscience, defended at length in Churchland (1986), has been challenged by Fodor (1974) and Pylyshyn (1984). Actually, there is not even agreement that Mendelian genetics can or should be reduced to molecular genetics. Any serious investigation of the intertranslatibility of generative and cognitive linguistics will have to be undertaken in the context of the interrelationship between connectionist models and models involving symbolic processing. Now there is at least plausibility to the idea that connectionist models can be regarded as implementations of classical (i.e. symbolic) models of cognitive architecture. This raises the question then whether cognitive linguistics models are interpretable as implementations of models of generative gram-

Bridges between generative and cognitive linguistics

17

mar. The fact that the models are not in general isomorphic is not in itself a barrier to that. After all, as Fodor and Pylyshyn (1988: 63) note, 'No one expects the theory of protons to look very much like the theory of rocks and rivers, even though, to be sure, it is protons and the like that rocks and rivers are "implemented in"'. The degree of lack of isomorphism between the two models might well be an incidental feature of analysis, rather than one based on inherent differences. For example, Tanenhaus, Dell, and Carlson (1987) have shown how a modular structure to language and a connectionist neural architecture can be made compatible. Certainly, there is the possibility that in order to accomodate such natural language phenomena as (partial) compositionality, multiple embedding, long-distance dependency, and so on, that symbol-based systems are well designed to handle, connectionist models might have to be altered in ways that the essential properties of symbol-based systems are (covertly) incorporated. To conclude, we have seen that there is good reason to believe that generative and cognitive linguistics are closer than is generally thought: they both espouse the cognitive commitment, the autonomy thesis allows the statement of direct form-meaning relations, both theories permit a lexicon, and the foundations of generative grammar do not require the adoption of model-theoretic approaches to semantics or to a theoretically significant separation of semantics and pragmatics.

References Bybee, Joan L. 1988 The diachronic dimension in explanation. In: Hawkins, J. A. (ed.). Explaining language universals. Oxford: Blackwell. 350-379. Caplan, David and N. Hildebrandt 1988 Disorders of Syntactic Comprehension. Cambridge, MA: MIT Press. Chomsky, Noam 1975 Questions of form and interpretation. Linguistic Analysis 1: 75-109. 1986

Barriers. Cambridge, MA: MIT Press.

1992

A Minimalist Program for Linguistic Theory. MIT Occasional Papers in Linguistics 1.

Churchland, Patricia S. 1986 Neurophilosophy: Toward a Unified Science of the Mind-brain. Cambridge, MA: MIT Press.

18

Frederick J. Newmeyer

Cossu, G. and John Marshall 1990 Are cognitive skills a prerequisite for learning to read and write? Cognitive Neuropsychologyl: 21-40. Curtiss, Susan 1977

Genie: a psycholinguistic study of a modern day 'wild child'. New York:

Academic Press. Deane, Paul D. 1988 Which NPs are there unusual possibilities for extraction from? Chicago Linguistic Society 24: 100-111. Fodor, Jerry A. 1974 Special sciences (or: disunity of science as a working hypothesis). Synthese 28: 97-115. Fodor, Jerry A. and Zenon W. Pylyshyn 1988 Connectionism and cognitive architecture: a critical analysis. In: Pinker, S. and J. Mehler (ed.). Connections and Symbols. Bradford: MIT Press, 3-72. Goldberg, Adele E. 1989 A unified account of the semantics of the English ditransitive. Berkeley Linguistics Society 15: 79-90. Gopnik, Myma 1990 Feature-blind grammar and dysphasia. Nature 344: 715. Jackendoff, Ray 1983 Semantics and Cognition. Cambridge, MA: MIT Press. 1990 Semantic structures. Cambridge, MA: MIT Press. 1992

Languages of the Mind: Essays on Mental Representation. Cambridge, MA:

MIT Press. Jackendoff, Ray and David Aaron 1991

Review of G. Lakoff and M. Turner, More than Cool Reason: a Field Guide to Poetic Metaphor. Language 67: 320-338.

Kaplan, J. A., H. H. Brownell, J. R. Jacobs, and H. Gardner 1990 The effects of right hemisphere damage on the pragmatic interpretation of conversational remarks. Brain and Language 38: 315-333. Kuno, Susumu 1987

Functional Syntax: Anaphora, Discourse, and Empathy. Chicago: University

of Chicago Press. Lakoff, George 1987

Women, Fire, and Dangerous Things: what categories reveal about the mind. Chicago: University of Chicago Press.

1991

Cognitive versus generative linguistics: how commitments influence results. Language and Communication 11: 53-62. Lakoff, George, and Claudia Brugman 1987 The semantics of aux-inversion and anaphora constraints. Unpublished paper delivered to the Linguistic Society of America. Langacker, Ronald 1991

Foundations of Cognitive Grammar, Volume 2: Descriptive Application.

Stanford: Stanford University Press.

Bridges between generative and cognitive linguistics

19

Leslie, Alan M. and Laila Thaiss 1992 Domain specificity in conceptual development: neurophysiological evidence from autism. Cognition 43: 225-251. Marr, David 1982 Vision: a computational investigation into the human representation and processing of visual information. San Francisco: Freeman.

Newmeyer, Frederick J. 1992 Iconicity and generative grammar. Language 68: 756-796. Osterhout, Lee and Phillip J. Holcomb 1992 Event-related brain potentials elicited by syntactic anomaly. Journal of Memory and Language 31: 785-806. Pinker, Steven and Alan Prince 1991 Regular and irregular morphology and the psychological status of rules of grammar. Berkeley Linguistics Society 17: 230-251. Pylyshyn, Zenon W. 1984 Computation and Cognition: towards a foundation for cognitive science. Cambridge, MA: MIT Press. Riddoch, M. J., G. W. Humphreys, M. Coltheart and E. Funnell 1988 Semantic systems or system? Neuropsychological evidence reexamined. Cognitive Neuropsychology 5: 105-132. Sadock, Jerrold M. 1983 The necessary overlapping of grammatical components. In: Richardson, J. F., M. Marks and A. Chuckerman (eds.). Papers from the Parasession on the Interplay of Phonology, Morphology, and Syntax. Chicago: Chicago Linguistic Society. 198-221. 1984 The poly-redundant lexicon. In: Testen, D., V. Mishra and J. Drogo (eds.). Papers from the parasession on lexical semantics, Chicago: Chicago Linguistic Society. 250-269. 1991 Autolexical syntax: a theory of parallel grammatical components. Chicago: University of Chicago Press. Smith, Neil and I. M. Tsimpli 1991 Linguistic modularity: a case study of a 'savant' linguist. Lingua 84: 315351. Talmy, Leonard 1988 The relation of grammar to cognition. In: Rudzka-Ostyn, B. (ed.). Topics in cognitive linguistics. Amsterdam: John Benjamins. 165-205. Tanenhaus, Michael K., Gary S. Dell and Greg Carlson 1987 Context effects and lexical processing: a connectionist approach to modularity. In: Garfield, Jay L. (ed.). Modularity in Knowledge Representation and Natural-language understanding. Cambridge, MA: Bradford. 83-110. Yamada, Jennie 1990 Laura: a Case for the Modularity of Language. Cambridge, MA: MIT Press.

Lexical semantics and morphology

A modular approach to the semantics of space in language Michel Aurnague and Laure Vieu

1. Introduction The aim of our work is to elaborate some formal tools for representing the semantic content of French expressions referring to space. We considered the topological prepositions sur On' and dans 'in', as well as the projective prepositions dessus 'above', dessous 'below', devant 'in front of, derriere 'behind', in the schema "Ntarget est prep N^^^". In the category of spatial referents we also took into account several internal localization nouns (or ILNs) such as haut 'top', has 'bottom', devant 'front extremity', which are all lexical items pointing out the different portions of an object. From an empirical point of view, the whole study rests on a detailed semantic analysis which distinguishes the different spatial configurations each of these lexemes allow us to refer to (Aurnague 1989), (Borillo 1988, 1992). What we think is more specific of our methodology is that this linguistic study also identifies the different inferential schemata that appear to combine spatial expressions in discourse. On the basis of these observations, we proposed a formal representation of the semantic content of these expressions. This formal system yields inferences whose results are similar to deductions made by human beings. Such a tool can be used as a theoretical basis in the automatic understanding of natural language spatial expressions.

2. A semantics in three levels According to some important linguistic works on spatial preposition semantics (Herskovits 1986; Talmy 1983; Vandeloise 1986), we claim that the semantics of space in NL cannot be represented with geometric notions alone. The description of the shape of some spatial entities and of the geometric relations holding between them is not sufficient to determine the applicability of spatial expressions.

24

Michel Aurnague and Laure Vieu

Otherwise, it would not be possible to explain, for example, why one cannot use both sous 'under' and dans 'in' to describe some situations as: (1)

La poire est sous I *dans le hol The pear is under / *in the bowl'

Neither why we make a difference between some uses of sur On' and the uses ofcontre 'against': (2)

L'qffiche est sur I *contre le mur. The poster is on / * against the wall.'

(3)

La planche est contre I *sur le mur. The board is against / *on the wall.'

In the same way, one could not account for the impossibility of stating le livre est dans le gant 'the book is in the glove' to describe a situation for which we can yet state the two sentences: (4)

Le livre est dans la main. The book is in the hand.'

(5)

La main est dans le gant. The hand is in the glove.'

The first three examples (adapted from Vandeloise 1986) show that the notions of containment and support contribute to the semantics of dans and sur in a decisive way. Actually, in many cases such "functional" notions have to be taken into account on top of the mere geometry. Taking an extreme position, Vandeloise even argues that functional notions (such as containment, anthropomorphism, accessibility ...) fully explain the semantics of French spatial prepositions. In example (4—5), the part of a pragmatic principle of "fixation" shows up. One can consider as the inside of the glove only the space portion corresponding to the typical use of a glove, that is, where one can put his hand. As a consequence, it is impossible to describe the book as being in the glove, even though the shapes of the hand and the glove have a similar concavity in which the book is (partially) included. These remarks together with several others of the same kind, have led us to adopt the methodology of analysing and representing the meaning of French spatial expressions on three levels (Vieu 1991b). First comes a geometric

A modular approach to the semantics of space in language

25

level representing objective space as described by the text analysed; this level is the base of the system. Second, a functional level represents every relation between entities introduced by the text, in particular non-geometric ones. Third, a pragmatic level takes into account some conventions and principles of "good" communication (such as Grice's principles (Grice 1975)), and for this purpose it also uses information that is not present in the text: context (as the intentions of the speaker) and world knowledge (as the typical use of the entities). These three levels are organized into a hierarchy: the second level introduces functional data over the geometric data of the first level, so it allows to represent what we call the "crude" semantics of spatial expressions. For its part, the pragmatic level alters the second level's result in order to adapt this semantics to the actual situation. The elements handled at the geometric level and at the two others are of a different nature. Functional and pragmatic phenomena affect the entities considered with all of their features, properties, and functional links between them, whereas geometric relations are independent of the entities' colour, age, substance, use . . . The importance of this difference lies in the fact that at the functional level, we can distinguish several entities which describe the same concrete object. In such a case, these entities have exactly the same geometric relations with respect to every entity: they cannot be distinguished at the geometric level. For instance, an object and the portion of matter making it up are distinct entities: they may have different ages, different structural properties, etc.; there are also several ways to form collective entities from the same collection of basic entities.1 For this reason, the functional and pragmatic levels deal directly with the entities introduced in the text, while the geometric level deals with what we call the spatial referents of those entities. As a result, the geometric level is completely independent of the kind of entities described, therefore it is independent of the domain treated.

3. The geometrical level As we underlined it before, in the geometrical level, we handle the spatial referents of the entities that is to say the space portions determined by their matter at a determined moment. These elements are also called here individuals. At this level we deal with various topological aspects like inclusion, contact, boundaries, etc., and with concepts related to projective geometry such

26

Michel Aurnague and Laure Vieu

as straight line, distance, order on a straight line, etc. The actual use of prepositions like sur On' and dans 'in' which allow us to situate an entity called trajector with respect to another entity called landmark shows the relational nature of the structures handled in the language as opposed to the absolute spaces used in robotics (where entities are localized by means of coordinates). In order to reflect these characteristics, topological data is represented in our system by means of B.L. Clarke's individual calculus (Clarke 1981, 1985), which we have modified and completed so as to take into account some important spatial concepts in language. This calculus, which lies on the sole primitive of connection between two individuals (noted C(x,y)), is used to define some mereological operators as well as Boolean and topological ones. As regards mereology we can mention the inclusion (P(x,y)), the overlapping (O(x,y)) or the external connection (EC(x,y)) between two individuals χ and y. For instance, the inclusion is defined stating that χ is included in y if every ζ connected with χ is also connected with y: P(x,y)= def Vz(C(z,x)^C(z,y)). It must be noted that the part-whole relations (also called meronomies) used in the language (e.g., cette roue est unepartie de ma voiture, 'this wheel is a part of my car') are much more complex than the mere geometric inclusion because they involve many functional factors depending directly on the entity (and not only on its spatial referent). Consequently we define this type of relation at the functional level of the system. In the Boolean part of the calculus, the operators sum (+), product (prod) and complement (comp), by means of which a lattice structure can be introduced, are defined. Because of the lack of a null element, this part of the theory is called pseudo-Boolean. As for topological aspects, the interior of an individual, its closure or the properties of being closed and open can be stated. Let us stress that the topological interior of an individual does not correspond with what we usually call inside in the language. For example, when people refer to the natural inside of a glass they are usually pointing out the portion of space enclosed between its sides (where one can put water) and not the very interior of those sides which contains only glass substance and corresponds in fact to the topological interior. As the functional aspects play a great part in determining the natural interior of an entity we introduce this notion at the functional level of the system.

A modular approach to the semantics of space in language

27

This topological part of the calculus is called pseudo because of the impossibility of handling boundary elements present in classical topology. The principal reasons of such an impossibility are the lack of a null element and the distinction between external connection and overlap configurations. In order to make up for this inconvenience we define the notion of empty individual (an individual which has only open parts) and we regard a limit as a tangential part having an empty interior. In fact we introduce three types of limits (limits of type 1, 2 or 3) by means of which we can characterize an element as a surface, a line or a point. Those concepts of limits are very important for the formalization of ILNs like dessus 'top extremity', bord 'edge', angle 'corner', etc. At this geometrical level we also add two types of contact, a strong one (the individuals in contact are assumed to "share some boundary points") and a weak one (the individuals do not have any common point although they are touching together), which seems to be very close to common sense. Let us indicate that contact plays a great role in the semantics of the relation sur On'. In spite of the fact that we used this intuition in order to describe some predicates, the individuals handled in B.L. Clarke's calculus are not interpreted as sets of points. On the contrary the author (Clarke 1985) shows a way to introduce points in its system as sets of individuals. This definition based on the filters technique consists actually in characterizing points through the individuals (two by two connected) "containing them". We proved in (Vieu 199la) that this definition in terms of filters worked correctly for interior points but was not right for boundary ones. Consequently we proposed an alternative definition for boundary points, and this enables us to represent the limit concept from those boundary points. The definition of a limit in terms of individuals and this last one based on points provide identical results. As we said before, at this geometrical level we do not only take into account topological data but we also integrate some important concepts of projective geometry. Having defined "points" in this theory as sets of two by two connected individuals, we specify the notions of straight lines, oriented straight lines, through various definitions based on the primitive relation (linking points) "is situated between". We also introduce distance by means of the primitive relation "is closer to" and this relation, together with the other one, makes it possible to define concepts of parallelism and perpendicularity. In order to formalize correctly the orientational process we have to complete our ontology by introducing the basic concept of direction. Then we define a set of thirteen predicates constituting an extension to multidirectional space

28

Michel Aurnague and Laure Vieu

of Allen's relations (Allen 1984). Each formula R(x,y,D) indicates the configuration holding between the maximum intervals filled by the individuals χ and y in the direction D. Finally, we define what is the extremity y of an individual χ in a direction D and we also grasp the concept of the direction D generated by the extremities y and z of an individual x. We can conclude the presentation of this level saying that we obtained a complete relational geometry.

4.

The functional level

One of the most important processes which take place at the functional level concerns orientation. The notion of "direction generated by the extremities of an object" defined at the geometrical level as well as various concepts introduced at the functional level among which the canonical use of an entity allow us to model the different aspects of intrinsic orientation in a quite detailed way (for vertical and frontal orientations). In particular the part of gravity is explicitly taken into account in the cases of intrinsic vertical orientation and the various modalities underlying the frontal intrinsic orientation (fixation through general orientation, tandem orientation or mirror orientation) are clearly differentiated. All those orientational tools are used in the semantic definitions of ILNs as haut 'top', has 'bottom', devant 'front extremity' etc., expounded in Aurnague (1991) and allow us to model the component of the semantics of sur On' relative to the position of the entities on the vertical. This latter characteristic is useful in particular to distinguish three configurations of sur On' according to the position of the localized object in relation to the referent object. If the object localized or trajector is higher than the referent object or landmark (e.g., le livre est sur la table 'the book is on the table') we speak of sur\. The case in which the localized object is at the same level as the referent object (e.g., Vaffiche est sur le mur 'the poster is on the wall') is called sur2. At last sur^ applies when the localized object is lower than the referent object (e.g., la mouche est sur le plafond 'the fly is on the ceiling'). Apart from geometric notions and the functional notion of orientation (for sur, as we just saw), the semantics of the spatial prepositions dans and sur includes some functional concepts belonging to "naive physics" (Hayes 1985).

A modular approach to the semantics of space in language

29

As we have seen before, support is essential in sur's semantics: an object hanging above a table, touching it, is not sur la table On the table'. As a matter of fact, support, along with contact, are the only features shared by the three cases of sur. Support makes reference to gravity: an object is supporting another one if it prevents it from falling. It is then clearly a physical concept; still it is a "naive" one, as few people are aware of how gravity is actually involved.2 As for dans 'in', we first note that in general, the inclusion does not link the spatial referents of the entities. In the sentence le livre estdans Varmoire 'the book is in the cupboard', the book and the cupboard do not share any portion of matter: their spatial referents do not even overlap. Here, the spatial referent of the book is included in the spatial referent of the inside of the cupboard. The inside of an entity is partly determined by the geometric function of convex hull, but there is more to it since any concavity does not generate an inside: the concavity must be a containing one (Herskovits 1986). Therefore, the notion of containment is important in the semantics of dans, even though the expression χ est dans b does not always imply that y contains χ (Voiseau est dans le del 'the bird is in the sky' and la main est dans le gant 'the hand is in the glove' do not involve containment). Containment can be described as the restriction of the contents' potential movements. Then, it also involves opposition to gravity, but what makes it different from support is the notion of restriction of lateral movements (compare sur un tabouret On a stool' and dans unfauteuil 'in an armchair'). Just as sur's semantics, the semantics of dans fall into three categories. The examples we have seen up tillnow illustrate the first two cases, where the spatial referent of the trajector is included in the spatial referent of the inside of the landmark. In thefirstcase - the prototypical one - this inclusion is total, as in le livre est dans Varmoire 'the book is in the cupboard'. In the second case, the inclusion is only partial (i.e., it is an overlap), as in la cuillere est dans la tasse 'the spoon is in the cup'. The third case is at first sight quite different: the spatial referent of the trajector is directly included in the spatial referent of the landmark, and there is a part-whole relation between the two entities. L'escalier est dans la maison 'the stairs are in the house' and I'homme est dans la foule 'the man is in the crowd' are examples of this case. As Vandeloise (1986) noted, not any part-whole relation can be described by a dans expression; for instance, le cerveau est dans la tete 'the brain is in the head' is acceptable while *le nez est dans la tete '*the nose is in the head' is not. This fact can be explained by a "contrast principle". A spatial

30

Michel Aurnague and Laure Vieu

expression describing a part-whole relation stresses the spatial position of the part within the whole: one considers the spatial relation as taking place between the part and, by "contrast", the whole without the part. As the nose is not included in any concavity of "the head without the nose", the last example is not acceptable. From this viewpoint, this third case is just derived from the other two.3 In fact, the distinction has to be refined further, as we use dans in different ways according to the type of the entities denoted by the trajector and the landmark. At the functional level, we distinguish three types of entities: the objects (all the examples above were concerned with them), the locations (countries, cities, gardens ...) and the space portions (as insides of objects, holes, cracks...). Various combinations are allowed; they are fully described in Vieu 199la. The formalization of the semantics of sur and dans is done at the functional level, on the basis of the observations we just made. A complete account of these formal definitions is given in Aurnague 1991; Vieu 199la; Aurnague and Vieu 1993a, 1993b. As we pointed out in the introduction, it is important to check whether the definitions we give in our system allow inferences in accordance with "natural" deductions. From le vase est sur le dessus de l'armoire 'the vase is on the top extremity of the cupboard', the system is able to infer le vase est sur le haut de l'armoire 'the vase is in the cupboard'. From le livre est dans l'armoire 'the book is in the cupboard' and 'the and l'armoire est dans le salon 'the cupboard is in the living room', the system deduces le livre est dans le salon 'the book is in the living room'. Still, dans is not always transitive, and the system reflects our intuitions in this respect too. From U y a un trou dans lefromage 'there is a hole in the cheese' and lefromage est dans le refrigerateur 'the cheese is in the fridge', the system does not infer U y a un trou dans le refrigerateur 'there is a hole in the fridge'; from la maison est dans Vile 'the house is in the island' and Vile est dans la mer 'the island is in the sea', the system does not conclude la maison est dans la mer 'the house is in the sea'. It may be noted that this semantics is then fairly more finegrained than what is usually proposed (in particular, dans is usually assumed to be fully transitive).

A modular approach to the semantics of space in language

31

5. The pragmatic level Some pragmatic rules act on the semantics obtained at the previous levels in a significant way. On top of functional knowledge, they use world knowledge (in particular, knowledge of typical situations), and information about context. The rules we will consider here may be seen as the instantiation of more general rules (such as Gricean cooperativity principles) in the spatial domain. First, pragmatic rules lead to deduce in some cases (often by implicature4), more information than is really present in the text and is represented at the first two levels. For instance, the sentence Marie est dans la voiture 'Mary is in the car' is generally understood as Mary is in the passenger cell, discarding at the same time the alternative Marie is in the boot. Second, they may rule out some expressions (for example, expressions inferred at the previous levels) because, even though their "crude" semantics is verified by the system and in the model, they cannot be uttered, since using the first process, these expressions would be regarded as conveying information contradictory with what is known. For instance, if we know that Marie est dans le coffre de la voiture 'Mary is in the car's boot' is true, then Marie est dans la voiture 'Mary is in the car' is not false, and yet, in general we cannot answer to where is Mary? with the latter sentence, for in most contexts, it is interpreted as Mary is in the passenger cell. A "fixation principle" underlies the examples cited above. This principle, first introduced in (Vandeloise 1986) expresses that the typical use of an object "fixes" some of its characteristics. For instance, the front and the back of a car are "fixed" by the usual - not the actual - direction of its motion; indeed, many intrinsic orientations are determined this way. Example (4-5) in the first section involved also this principle: the inside of a glove is "fixed". Several other principles may be found. We can mention the principle of "maximum trajector", an instantiation of the maxim of quantity: a spatial relation is generally stated for the biggest entity possible. For instance, if the car is in the garage, then the car's engine is also in the garage; but stating this last fact alone (the car's engine is in the garage) somehow implies that the engine is separated from the car. At the same time, there is also a principle of "minimum landmark", expressing that the smaller the landmark, the more precise the localizing relation. However, the application of such principles has to be controlled because some of them may lead to contradictory results. This part of the system is still under work, and at the moment we cannot give an algorithm for the application of the various rules.

32

Michel Aurnague and Laure Vieu

Third, we must mention the pragmatic phenomenon that enables us to loosen some conditions of the semantic definitions. For the preposition sur On' we may drop the condition of contact between the spatial referents of the trajector and the landmark. Doing this makes sur ι transitive; indeed, if we have a pile of books on a table, any of the books may be described as being on the table. However, the loosening is not possible in any situation. For instance, if instead of two books, we have a lid on a tea-pot, the tea-pot being on the table does not allow us to say that the lid is on the table. The loosening of some conditions must be controlled by Grice's maxim of relevance: if a relation is more relevant than another one, the first cannot be "forgotten" and the second used. Relevance is highly dependent on functionality; in the first example, the books have no special function that makes it relevant for a book to be on another one, whereas in the second, the lid being on the tea-pot fulfills its function: this fact cannot be "forgotten" by tansitivity. We must note also that relevance depends on context: if one wants to designate a particular book of the pile, le livre sur la table 'the book on the table' will only designate the book that is in contact with the table.

6.

Conclusion

A formal semantics of NL expressions referring to space has to integrate the different aspects which underlie their semantics. As we have shown with the analysis of dans 'in' and sur On', the mere geometrical data is not enough to express the semantics of such lexemes, and functional or pragmatic aspects must also be taken into account. These observations have led us to build up a three level system (geometrical, functional and pragmatic) by means of which we can represent the meaning of spatial expressions and draw some deductions. This modular construction permits thus to come closer to the natural reasoning expressed in the discourse and from this point of view constitutes a real cognitive approach. Before finishing we have to underline that this system also deals with some of the complex relations occurring between space and time, the individuals handled at the geometrical level being in fact spatio-temporal ones.

A modular approach to the semantics of space in language

33

Notes 1. These examples were introduced in Link 1983 about a golden ring, recently made out of some old Egyptian gold: "the ring" is new but "the gold making up the ring" is old; and about playing cards of several decks: counting the plural entities "the cards" and "the decks of card" gives different results. 2. For instance, how the forces reacting against gravity are distributed. 3. This principle accounts for some uses of sur as well: le bouton est sur le devant de la television 'the knob in on the front of the tv set', la table est sur ses pieds 'the table stands on its legs'. 4. Then we need a non monotonic logic at this level.

References Allen, James F. 1984 Towards a general theory of action and time. Artificial Intelligence 23.2: 123154. Aurnague, Michel 1989 Categorisation des objets dans le langage : les noms et adjectifs de localisation interne. Cahiers de Grammaire n° 14, Toulouse: Universito Toulouse-Le Mirail. 1-21. 1991

Contribution ä etude de la semantique formelle de l'espace et du raisonnement spatial : la localisation interne en fran^ais, semantique et structures inferentielles. Ph.D. dissertation, Universit6 Paul Sabatier, Toulouse.

Aurnague, Michel, Mario Borillo and Laure Vieu 1990 A cognitive approach to the semantics of space. COGNITIVA 90, Madrid. 321-328. Aurnague, Michel and Laure Vieu 1993a A three-level approach to the semantics of space. In C. Zelinsky-Wibbelt (ed.), The Semantics of Prepositions: from Mental Processing to Natural Language Processing. Berlin: Mouton de Gruyter. 395-439. 1993b

Toward a formal representation of space in language: a commonsense reasoning approach. In F. Anger, H.W. Guesgen and J. van Benthem (eds.), Proceedings of the IJCAI'93 Workshop on Spatial and Temporal Reasoning, Chambery.

Borillo, Andr6e 1988 Le lexique de l'espace : les noms et les adjectifs de localisation interne. In Cahiers de Grammaire n° 13, Toulouse: Universite Toulouse-Le Mirail. 122. 1992

Le lexique de l'espace : prepositions et locutions pripositionnelles de lieu en francais. In L. Tasmowski and A. Zrib-Hertz (eds.), Hommage a Nicolas Ruwet. Dpecial Issue of Communication and Cognition. 176-190.

34

Michel Aurnague and Laure Vieu

Clarke, Bowman L. 1981 A calculus of individuals based on "connection". Notre Dame Journal of Formal Logic, 22.3: 204-218. 1985 Individuals and points. Notre Dame Journal of Formal Logic, 26.1. Grice, H. Paul 1975 Logic and conversation. In C.P. Morgan (ed.), Syntax and Semantics. New York: Academic Press. 41-58. Hayes, Patrick J. 1985 The second naive physics manifesto. In J.R. Hobbs and R.C. Moore (eds.), Formal Theories of the Commonsense World. Norwood: Ablex Publishing Corporation. 1-36. Herskovits, Annette 1986 Language and Spatial Cognition: An Interdisciplinary Study of the Prepositions in English. Cambridge: Cambridge University Press. Leonard, Henry S. and Nelson Goodman 1940 The calculus of individuals and its uses. The Journal of Symbolic Logic, 5: 45-55. Link, Godehard 1983 The logical analysis of plurals and mass terms: A lattice approach. In R. Bäuerle, C. Schwarze and A. von Stechow (eds.), Meaning, Use and Interpretation of Language. Berlin: Mouton de Gruyter. 302-323. Talmy, Leonard 1983 How language structures space. In H. Pick and L. Acredolo (eds.), Spatial orientation theory, research and application. New York: Plenum Publishing Corporation. 225-282. Vandeloise, Claude 1986 L'espace enfrangais: semantique des prepositions spaliales. Paris: Seuil. Vieu, Laure 1991a Sdmantique des relations spatiales et inforences spatio-temporelles : vine contribution ä l'e"tude des structures formelles de 1'espace en langage naturel. Ph.D. dissertation, Universito Paul Sabatier, Toulouse. 1991b Quelques 616ments pour une somantique formelle de 1'espace. Actes des

joumees du UPN, Paris.

Beer and semantics Dirk Geeraerts 1. Belgium as Beer's Own Country1 If Cognitive Linguistics is about the relationship between language and culture, Cognitive Linguistics conferences should be concerned with learning about cultures as much as they are concerned with learning about language and languages. In approaching a culture, in fact, it seems only fair to tackle those areas of behaviour first that are experienced as being important by the members of the culture themselves. And it seems that there is indeed sufficient evidence for the claim that beer is part of the pride and joy of the inhabitants of that minuscule patch of land along the coast of the North Sea. First, some factual data may demonstrate the importance of beer in Belgium. To begin with, the consumption of beer per head of the population is high. The chart in figure 1 is based on data compiled by the Interbrew breweries in 1987. It specifies the consumption of beer in liters per head of the population in twelve European countries; it demonstrates the existence of a so-called European beer-belt, situated geographically between the spirit-belt in the north of Europe, and the v^ine-belt in the south. The figures, of course, do not show that Belgium is beer country number 1 as far as per capita consumption is concerned: it comes third after Germany and Czechoslovakia. The key concept regarding Belgium's status as a beer country, however, is not quantity, but quality. And the quality shows up specifically in the enormous variety and the abundant diversity of the types of beer produced in Belgium. In the data set that was used for this investigation (and which is taken from Peter Crombecq's annual overview of Belgian beers for the year 1992) no less than 1454 different brands are distinguished. In Crombecq's own classification of these beers according to taste and alcoholic strength, no less than 304 distinct types of taste appear. A comparison with Holland, Belgium's nearest neighbour, strengthens the impression of diversity. Against the 1454 different brands to be found in Belgium, there are only 354 Dutch ones. And while Crombecq distinguishes 304 different Belgian tastes, he only tastes 94 Dutch ones. In short, there is sufficient evidence for the variety of Belgian beers.

36

Dirk Geeraerts

Table 1. Per capita beer consumption in 12 European countries (in liters)

Germany Czechoslovakia Belgium Austria England The Netherlands Ireland

145 130 121 118 111 85 75

Sweden Norway

51 52

Spain Portugal Italy

65 40 25

At least as important as these factual data is the recognition that Belgians identify themselves with the country's production of high quality beers. If you ask people from Belgium what their country is good at, they will probably mention beer and chocolates. The recognition that beer, and the quality of Belgian beer, is an explicit topic in Belgian culture, may be further illustrated in two ways. One is the following slogan from a beer advertisement: Het bier van het land van het bier, i.e., 'the beer from the country of beer', or, more freely, 'the beer from beer's own country'. Belgium as beer's own country - this is apparently a live notion. The second illustration consists of the number of coffee table books that are published in Belgium on the topic of beer (although the name coffee table books is obviously somewhat paradoxical in this respect). Here is a list of some more or less recent ones; if nothing else, it at least shows that there is a live interest in beer, and that beer is considered a quality product worthy of luxurious publications of the type that is usually restricted to wine and art. M.Jackson 1977 W. Patroons 1984 J. Tulfer 1986 M.Jackson 1991 G. van Lierde 1992

Spectrum bieratlas (transl. of The world guide to beer). Utrecht/Antwerpen. Alles over Belgisch bier. Antwerpen/Weesp. Belgische biergids. Antwerpen. De grate Belgische bieren. Een volledige gids en een hulde aan een unieke cultuur (transl. of The great beers of Belgium). Anwerpen. Bier in Belgie. Gids voor bieren en brouwerijen. Roeselare.

Beer and semantics

37

2. Methodological considerations Given the importance of beer in Belgian culture, the cognitively interesting question is this: how is beer conceptualised in Belgium? What are the cognitively salient concepts associated with it? And more specifically, starting from a Cognitive Linguistic perspective: what does language reveal about the conceptualisation of beer? The two crucial aspects of this question that I would like to focus on are the following: - First, a general methodological question: how can salience effects in lexical fields be measured? - Second, a specific question about beer: what are the specific values associated with beer? Before I try to be more specific about these questions, I have to clarify a few points. To begin with, there are a number of restrictions on what I will be doing next that have to be mentioned explicitly. The vocabulary that I will be talking about consists basically of brand names, i.e., the names of types of beer. This means that various parts of the lexical beer field will pass unnoticed. For instance, I will not have the opportunity to mention the names of pubs, inns and cafes (in spite of such gems as The ship of fools, The last judgement, The kingdom of heaven, The ultimate hallucination). The restriction also implies that the vocabulary for talking about beer will not enter the picture. I will have nothing to say, that is, about predicates like earthy, fruity, refreshing, complex, seductive, and others that may be used to describe beers. (Surprisingly perhaps, this vocabulary is far less extended than the one that exists for wine and that has been analysed so elegantly by Lehrer 1983.) Another consequence of the restriction to brand names is that the everyday names of the beers are not considered systematically in what follows. The brand names, if you like, are full names, constituting a truly individual identification of the beers. The everyday names, on the other hand, are like Christian names: shorter but at the same time more general forms that are communicatively sufficient in most circumstances. In the technical terms of lexical semantics, the brand names are certainly not basic level terms: they are used much less frequently than the everyday names, and they are taxonomically much more specific. In most cases, the generic basic level denominations are based on overall types of beer. Whereas the full names are clearly proper names, the everyday names are common nouns. When you ask for a geuze in a cafe, it will mostly be of no interest to you whether you are served

38

Dirk Geeraerts

an Eylenbosch gueuze Iambic, a Drie Fonteinen geuze, a Mort Subite geuze or any of the other brands of geuze. But by a typical prototypicality effect, the everyday name is not necessarily originally a common name. When you ask for one of those typical white beers made on the basis of wheat, you would probably ask for a Hoegaarden, which is the most typical representative of this class - and you probably would not object if you were then served a Dentergems witbier, which is another example of the same class. It could be argued, no doubt, that the basic level terms are somewhat more interesting for lexical analyses than the brand names, precisely because they are cognitively more salient. Note, however, that systematic research into the actual use of the everyday names would require extensive observation in pubs and cafes, and I am afraid that the investigation would then be diverted very soon onto scientifically unfruitful although extremely pleasant side-roads. Moreover, precisely because the brand names are more specific than the everyday names, they may reveal more of the conceptualisation of beer. Again, it could be remarked that the conceptualisation they reflect is not the consumer's but the brewer's, but even so, the brewers are likely, for obvious commercial reasons, to choose names that have a broad appeal to the consumers. Even though we study names that are not actually given by the consumer, the names can be expected to fit the consumers' image of what beer should be. From a methodological point of view, the field of beer terms is interesting for two reasons - two complications, in fact. The first methodological point to note involves the fact that the names found in the data set subclassify into two fundamentally distinct subsets, which each require their own type of analysis. The majority of the brand names conforms to a general pattern: they consist of an identification of the firm producing the beer, and a specification of the type of beer involved. Here are some examples: Louwaege pils, Liefmans frambozenbier, Belle Vue geuze. There is considerable variation in this pattern, not least because neither the first nor the second element is absolutely necessary. In some cases, a single reference to the origin of the beer suffices. This is the case, for instance, with some of the beers produced in monasteries (I will come back to these further on). A name like Orval identifies the monastery together with the single kind of beer produced there: there is only one Orval kind of beer, and it is a trappist beer. Conversely, only the type of beer may be specified, without reference to the brewer, but then, the generic name will be subclassified by the addition of various specifications as in the following: Dentergems witbier, Aarschotse bruine, where a place-

Beer and semantics

39

name indicates that this is the type of white beer or brown beer or whatever as typically brewed in Dentergem, or Aarschot, or wherever. The subclassification may also occur in combination with an identification of the brewer, as in Liefmans geuze gefilterd, Liefmans geuze ongefilterd, where it is indicated that the geuze beer produced by the firm Lindemans comes in two varieties: a filtered one and an unfiltered one. In cases like the Orva/-type, where the element identifying the producer at the same time suggests the kind of beer involved, the subclassifying element occurs without an explicit mention of the generic type of beer. In the following set, the various additions identify the varying degrees of alcoholic strength of the beers. The name Westvleteren is again a name for a monastic brewery; it identifies the beer in question as a trappist beer: Westvleteren dubbel 4, Westvleteren special 6, Westvleteren extra 8, Westvleteren abt 12. All the names mentioned so far stand in dramatic contrast with a type of name illustrated by the following examples: Duvel, Lucifer, Judas, Verboden Vrucht ('Forbidden fruit'). The basic distinction between this type (which forms a sizeable minority of the total set) and the previous one resides in the fact that the type 1 names are literal names, whereas the type 2 names are figurative, or at least associative names. A name like Liefmans geuze gefilterd is made up of various informational elements that identify aspects of the beer: the fact that it is of the geuze type, the fact that it is produced by the Liefmans brewery, the fact that it has gone through a filtering process. Each of these informational elements is identified directly in the name of the beer, i.e. by means of words whose literal meaning it is to express the informational elements in question. The literal meaning of the type 2 names, on the other hand, does not as such identify the relevant informational element. The name Duvel literally means 'devil', but the information that is relevant for the beer name is only indirectly connected with this literal meaning: it is a concept that has something to do with the devil, and that for one reason or another is also appropriate to talk about this kind of beer. Not surprisingly, the identification of the relevant informational element is not always easy with type 2 names. The beer may have something devilish, but what exactly is that? Does it lead you to destruction? Is it a sin to drink it? Does it overpower you in the way the devil might? One could say, perhaps, that the type 2 names attribute somewhat vague qualities to the beers, whereas the type 1 names describe rather clearly defined features. In summary, the type 2 names are associative or, if one wishes, evocative names, because they function indirectly, and because they involve less clearly

40

Dirk Geeraerts

defined characteristics than the type 1 names. The consequence of the distinction is that the type 2 names have to be analysed on a different basis than the type 1 names. In the case of the type 1 names, one can base the analysis rather straightforwardly on the literal readings of the terms. In the case of type 2 names, one has to take into account the relationship between the original literal meaning and the evocatively associated meaning. This type of analysis, of course, is well-known within Cognitive Linguistics: it is analogous to what we all do when we analyze metaphorical patterns in Lakovian style, i.e. to identify source and target domains and pinpoint the precise motivational link that connects them. In the vocabulary subset that interests us here, the associative names are not all metaphorical, but the analytical procedure used for metaphors can be generalized. The second complication that I hinted at before involves the fact that one of the things that we are crucially interested in are salience effects in lexical fields. Given a set of categories that belong together in a conceptual field (like types of beer), what we would like to know is which categories are conceptually more important than others. This is not a question that is adequately answered in traditional lexical field theory, but I think we should recognize that even Cognitive Linguistics, with its outspoken interest in salience effects, has not yet developed a proper methodology for treating this kind of issue. The salience effects Cognitive Linguistics has been primarily interested in are salience effects within separate categories, an interest that is epitomized by prototype-theoretical research (cf. Geeraerts 1989). Apart perhaps from Berlin's basic level model (see Berlin 1976, 1978), not much systematic attention has been devoted to salience effects that involve not the various applications of a single category, but rather the relationship between various lexical categories as alternative names for the same kind of referent. There is, to be sure, a nuance to be added to this picture. As far as the associated values as expressed by type 2 names are concerned, we do have a clear-cut idea of how to measure salience: an associated value is more salient to the extent that it recurs throughout the set of evocative names. This is basically the measure of cognitive salience behind the search for generalized metaphors introduced by Lakoff and Johnson (1980). Note, however, that it involves only type 2 names, so that it remains important to try to find a measure of salience that specifically applies to type 1 names.

Beer and semantics 3.

41

The origins of the diversity

To see what is involved in tackling the methodological question raised a moment ago, we should first have a closer look at the structure of the referential field itself, i.e., at the variety that I hinted at before. In order to get a better grasp of it, it is important to know something about how beer is made. Basically, you put barley into water and let it ferment. However, an elaborate beer culture could not have developed if this process had not been subject to refinement. Schematically, what actually happens can be divided into three steps. First, the barley is let to soak and germinate, and the budding barley is then dried or sometimes roasted. The product of this first step is known as malt, which reveals, by the way, that whisky is the distilled counterpart of beer. In a second step, the malt is crushed and cooked, so that the natural sugars that it contains become free. During the cooking stage, additional substances for flavouring are added, such as hop or extra sugar. The product of this second step is called -wort. When yeast is added to the wort, it can then ferment, and at the end of this third stage, you get beer. At each of the three stages, alternatives with regard to the standard process exist. Instead of starting off with barley, for instance, you can begin with wheat, or rye, or a mixture. In the second stage, various other flavouring substances can be added. Typical for a number of Belgian beers is the addition of fruit flavours like cherry and raspberry, or certain herbs like ginger, licorice, coriander, or mint. And in the third stage, different forms of fermentation exist. This is an important point that needs some elaboration. Basically, there are three forms of fermentation: high fermentation, low fermentation, and spontaneous fermentation. In the first case, the fermentation process is relatively short and intense, producing strong ales. In the case of low fermentation, the process is a slow one, the ripening process takes a long time, and the resulting beer is less heavy than in the case of high fermentation. These are the beers of the regular pilsener or lager type. Spontaneous fermentation, on the other hand, is much less common than high or low fermentation, and is really extremely typical for one portion of the Belgian beers. In the surroundings of Brussels, the art of producing beers by spontaneous fermentation has been refined in ways not known anywhere else in the world, and some of the most typical names associated with Belgian beers have to do with spontaneous fermentation. Specifically, I first have to mention lambiek, which is the product of 100% spontaneous fermentation. As lambiek as such is rather sour, you will not find it very often in its pure form. Usually, sweet flavours

42

Dirk Geeraerts barley germination drying/roasting malt sugar etc.

hop cooking wort fermentation

yeast beer

Figure 1. Schematic representation of beer production

are added in the form of fruit extracts, which yields names like kriek lambiek 'cherry Iambic'. Another additional process that Iambic may go through is a second fermentation, which yields the well-known geuze, which is sweeter than Iambic, and which has a unique, champaign-like sparkling character. The variety does not stop there, however. At least two additional points have to be mentioned. For one thing, some Belgian beers are brewed only on a special occassion. Traditionally, this could be the harvesting season, or Christmas, or the annual festival of the local patron saint. Nowadays, these gelegenheidsbieren or 'occassional beers' also include beers made to commemmorate specific historical events, like the anniversary of the foundation of one or another city, or the birth of a historical figure. Another additional factor besides the temporal one is a geographical one, or at least, it involves the place where the beer is being brewed. This is particularly important for abbeys and monastries producing beer. In fact, refining the art of brewing is one salient aspect of the civilizing influence that monasteries and abbeys have exerted in the Low Countries. Traditionally, all monasteries brewed their own beer, like many farmers once did. Today, there are still five monasteries left where the monks brew their own beer. As they all belong to the Trappist order, their beer is commonly known as trappist.

Beer and semantics

43

The monasteries in question are those in Orval, Chimay, Rochefort, Westvleteren and Westmalle. These beers are all of the strong, high fermentation type, but there are many differences among their individual flavours. It is useful, by the way, to distinguish the trappist beers from the larger category of abdijbieren or 'abbey beers'. These are beers of basically the same type as the trappist beers, but they are being produced by commercial breweries who have taken over the original monasterial breweries, or who have merely paid for the license to use the abbey's name. Examples are Leffe, Grimbergen, Maredsous, Affligem, Corsendonck, and Tongerlo. On the basis of what I have discussed so far, we can now define, on a generic level, some of the most common types of Belgian beers. - pils: a light blond beer of the pilsener or lager type, produced through low fermentation (common brands: Stella, Jupiler) - witbier. 'white beer', a light beer on the basis of wheat rather than barley, with a blond, somewhat cloudy appearance (common brands: Hoegaarden, Dentergems) - geuze: a sparkling, spontaneously fermented beer with a reddish colour and a sour-sweet taste (common brands: Belle-Vue, Mort Subite) - kriek: a moderately sweet beer obtained by adding cherry juice to geuze or to lambiek (common brands: Belle-Vue, Mort Subite) - trappist: strong, mostly dark brown beer of the high fermentation type, produced in Trappist monasteries (common brands: Orval, Chimay, Rochefort, Westmalle). (Sometimes, the name trappist will also be used for the comparable commercial abbey beers).

4.

Salience effects for type 1 names

Although it would seem that the list at the end of the previous section gives us a nice and neat set of categories to start the onomasiological investigation with, things are less clear than they seem. Once we start looking at more specific categories, there appears to be a lot of overlap and cross-classification among the various groupings of beers. Consider the following example. As we have seen, we can distinguish between geuze, lambiek, and kriek in the category of spontaneously fermented beers. Among the beers with high fermentation, there are the whitebeers, based on wheat, but also a number of barley beers, such as the popular, very strong beer called Duvel ('devil'). Geuze and lambiek have in common with the whitebeers that wheat is used in

44

Dirk Geeraerts

the production process. Duvel and geuze, on the other hand, have in common that then· fermentation process continues when they are already bottled. And further, there are kriek beers that are not based on geuze or lambiek, like the Felix kriek oudbruin. The resulting picture is one of multiple overlapping, as shown in figure 3. Each beer type, in other words, can be considered a bundle of features, each of which can be the basis for the name of the beer (or part of the name of the beer). But how then can we go about determining the conceptual salience effects that we are interested in? If each beer type were to belong uniquely to a neatly separated, taxonomically well-behaved category, we would just have to investigate those categories. But if the categories themselves are extremely varied and exhibit multiple overlapping, what should one do ? I earlier referred to Berlin's basic level model as the only systematically elaborated model for salience within lexical fields. But note that Berlin's model deals with the salience of levels in a taxonomy, while we are interested here in salience phenomena among categories on fundamentally the same level of a taxonomy. So it seems that it may be worthwhile to suggest ways of dealing with intercategorial (rather than intracategorial) salience effects in a way that goes beyond the basic level model. (The ideas presented here are developed in more theoretical and empirical detail in Geeraerts, Grondelaers and Bakema 1994. The notion 'intercategorial salience' is briefly introduced in Geeraerts 1993; it is basically a quantitatively operationalized version of the concept of entrenchment defined by Langacker 1987: 59-60. A useful presentation of the distinction between intracategorial and intercategorial entrenchment may also be found in Kleiber 1991.) The obvious thing to do, I would suggest, is to start from the features themselves that constitute the bundles representing each individual type of beer. A practical methodology can then be formulated as follows: the more a particular feature is expressed in the names of beers, the more it is cognitively salient. When, for instance, the presence of fruit flavours never surfaces in the names of the beers that have such a flavour, it is unlikely that fruit flavour is a preponderant feature of beers. On a general methodological level, I would like to suggest the following operational definition of conceptual salience in lexical fields: a category in a lexical field is more salient to the extent that its members are more often identified by names that are typical for the category. Of course, we have to be more precise about what it means for a feature 'to be expressed in the names of beers'. In most cases, there is no problem: when you find a reference to kriek in a name, and when you know that kriek is the

Beer and semantics

j fermentation

Duvel

45

geuze

in bottle

witbier

lambiek

1.9 1

cherry flavour

oudbruin krlek

high fermentation

kriek lambiek

spontaneous fermentation

Figure 2. Multiple overlapping in the lexical field of beer names

name for a type of cherry, there is no difficulty in concluding that the addition of cherry juice is mentioned in the name of the beer. In other cases, the reference is more indirect. A striking example is the following: in the French speaking part of Belgium, a peach-flavoured beer is produced with the name La pecheresse, which translates as 'the female sinner, the sinning woman'. When you realise, however, that peche is the French for 'peach', it will be clear that the name contains a pun with an indirect reference to the presence of peach flavour. In general, a feature is expressed by a name (or part of a name) when that name (or the relevant part of it) only occurs in connection with that feature. Some more examples may further illustrate the point. The fact that a particular beer is low on alcohol may be expressed directly by expressions like alcoholarm 'low on alcohol' or alcoholvrij 'without alcohol'. Indirectly, however, the item tafelbier 'table beer' also signals the light character of beer,

46

Dirk Geeraerts

Table 2. Salience characteristics of sample referential features

Referential feature

based on wheat instead of rye very light, low alcohol occassional beers added herbs raspberry flavour cherry flavour trappist beer

Referential frequency in the corpus

Number and percentage of characteristic names

139

91

65,5

110

25

22,7

86 59 40 19 16

86

100 6,8

4 38 18 16

95,0 94,7

100

to the extent that it only occurs within the group of very light beers. Similarly, lambiek and geuze can be taken as expressions of the 'spontaneous fermentation' character of beers, to the extent that they only occur in names for beers with that characteristic. Kriek, however, does not express the feature of spontaneous fermentation, because (as we have seen in Figure 2) it also occurs in the name of beers of high fermentation. Following these guidelines, we can show that not all features of beers are equally important. There are clear differences in the extent to which specific characteristics are expressed in the names. Table 2 presents a sample to demonstrate the kind of differences in salience that occur. I should add immediately that it is practically impossible to systematically explore all possible features, mainly because their presence is often uncertain: even persistent tasting (to the extent that the researcher can muster it) does not always suffice to establish the objective characteristics of a particular brand. The table should be read as follows. The first column of figures indicates the number of beer brands in the data set that has the feature mentioned to the left. The rightmost column indicates how many of those beers actually carry a name that refers directly or indirectly to the feature in question. The relationship among the figures mentioned to the right gives an indication of which features are considered special enough to merit a name of their own, or at least, special enough to be referred to in the beer names. Trappist origin, for instance, seems to be highly valued, while the addition of herbs

Beer and semantics

47

is not a terribly individuating feature. It is tempting, of course, to speculate about the origin of these preferences. It is not unlikely that they reflect historical traditions within the culture, but this is, to be sure, something that cannot be established on the basis of this scanty set of examples alone - one would need a longitudinal diachronic analysis to get any certainty on this point. On the other hand, the examples at least allow for the falsification of one hypothesis. It might be suggested that the salience of a particular feature is inversely correlated with its referential frequency: an infrequent feature like the presence of cherry flavour is lexically expressed to a high degree, precisely because it is a marked, unexpected feature. This phenomenon probably does play a role in the lexical field as a whole: no one would expect a reference to water in the name of beers, precisely because it is obvious to anyone with the least encyclopedic knowledge of the domain that all beers contain water. However, the figures in the table clearly establish that such an inverse correlation with referential frequency could at most explain part of the data. There seem to be at least some real preferential phenomena at play.

5. Salience effects for type 2 names I would now like to turn to the analysis of what I have earlier called the associative values of the beer denominations. In Table 3, five salient associative motifs are identified. They are recurrent ones not just because they show up in various individual names, but also because they generalize over more than one source domain. It would lead us too far to discuss all the names separately, but let's look at a sample analysis of some illustrative cases. A first recurrent pattern represents beer as something prestigious. In its most direct form, the high status of beer is evoked by borrowing names from the lexical field of wine (the prestigious drink par excellence), as in Vlaamse Bourgogne 'Flemish burgundy'. The same value may be evoked more indirectly by referring to important people, like kings (Grand Monarque 'Great monarch') or members of the aristocracy (Ridder 'Knight'). A link with the religious motif that occurs a number of times in the set is constituted by names referring to hierarchically important people in religious organisations (like Monseigneur 'Monsignor' and Kapittel abt 'Chapter abbot'). Another set of prestigious figures involves historically important people, who may be political leaders like emperor Charles V (Charles Quint) or culturally prominent people like the painters Breugel, Rubens, and Rembrandt. In most cases,

48

Dirk Geeraerts

these figures have a direct historical link with the Low Countries, but names like Da Vinci show that this is not an absolute precondition. Next prestigious people, expensive materials and other valuables may be evoked, as in Safir; a cross-reference to one of the other subcategories already mentioned appears in a name like Gouden Carolus, which is originally the name of the gold coin minted by emperor Charles V. A second major motif represents beer as an adventure at the edge of the accepted norms. Up to a certain point, the notion of adventure (danger, physically extreme conditions, violence) and that of transgressive behaviour (flouting existing norms) may be separated, but because they so often occur together, they are here treated as one category. The norms that are violated may be basically religious or social. The first case yields names referring to sin (Verboden Vrucht, 'Forbidden fruit'), witchcraft (Sorcieres d'Ellezelles, 'Witches of Ellezelles'), and the devil (Lucifer, Satan, Duvel, 'Devil'). The second case yields names referring to pirates (Piraat, Boucanier}, smugglers (Smokkelaar), highwaymen (Bokkereyef) and related unorthodox professions. Also, names like Schavuit, Deugniet, and Stouterik 'Rascal' indicate moderated mischievousness; they are typically the kind of name to be used hypocoristically to a child. Brigand and Kerelsbier are particularly interesting cases. For instance, in the Belgian context, Brigand refers to the peasants and farmers who rebelled against the French occupation at the end of the 18th century (the Boerenkrijg 'Peasants' war'). Looking back at the historical figures mentioned above, it now becomes clear that they often (as in the case ofAmbiorix, Artevelde, Breydel and De Coninck) represent a struggle for independence against a foreign oppressor. (It will be suggested below that an irreverend or even rebellious attitude with regard to established authority may be a general characteristic of Belgian/Flemish culture.) Names like Stoeren Bonk 'Sturdy fellow, son-of-a-gun' evoke virility and force. Against the background of what was just said, it seems plausible that these names do not however evoke machismo pure and simple; rather, the image is at least in part one of strong and independent men who stand up for their rights. A final major source domain in this group points to the physical effects of overabundant consumption of beer, as in Delirium Tremens. The idea probably is that you have to be a strong person (of the type evoked in the previous group) to take such strong drinks. Note, in addition, that multiple associations may again be active: Houten Kop, 'Wooden head', may not only mean 'hangover', but also stubbornness, which is again related with the spirit of independence.

Beer and semantics

49

Names like Flierefluiter 'Layabout, idler', Lekkerbek 'Gourmet', and Pallieter (the name of a famous literary figure who has become symbolic for an epicurean attitude) are typical for a third major motif, in which beer is heralded as a source of pleasure and as part of an easy-going way of life. Characteristically, the religious motif appears once again when beer is so to speak presented as a foretaste of the pleasures that may await us in heaven. This is the case in a.o. Engeltjesbier 'Little angels' beer'. The religious source domain also contributes to the large group of names in which beer is represented as a link with the traditional culture of the past. Religion, to be sure, is such an aspect of the tradition, in the same way in which the guilds of old (as in Gildenbier 'Beer of the guilds') are. References to old methods of producing beer (Traditions des Moines The monks' traditions') serve a similar function. In A laferme On the farm', Verte campagne 'Green countryside', and Bosbier 'Beer of the woods' rural if not pastoral associations are activated: to the extent that beer is part of the legacy of the past, it is clearly the pre-industrial era that is at stake. Historical references, to be sure, are a recurrent phenomenon in the set presented here. Apart from the historical figures mentioned before and the references to local history mentioned below, there are straightforward references to the past in names like Vieux Temps Old times'. In Druide 'Druid', I'lguanodon 'The iguanodon', and Blonde du Menhir 'The blonde one of the menhir', the historical reference takes a remarkable prehistoric, pre-Christian shape: the historical reference is so to speak pushed to its extreme, but at the same time, the notions of primitivity and force that seem to accompany these references echo some of the vigorous, vitalistic values connected with the second motif discussed above. The final major motif represents beer as an aspect of local identity, by referring to people and things associated with a particular region. The relevant region can change in surface, to be sure. It may be Belgium or Flanders as a whole (as with historical figures like the painters Rubens and Breugel, or the cartographer Mercator), or it may be one of the main regions of Flanders (as when the historical figures Artevelde (Ghent), Breydel and De Coninck (Bruges) evoke the past of Flanders in particular, and when Elckerlyc and Pallieter refer to figures from Flemish literature); very often, it is a specific town or village (as in the many names that mention a particular placename, or the place's patron saint). The latter case is also represented by those names that make reference to aspects of local folklore, like famous local figures (Ketje), mysterious beings that appear in legends and traditional stories (Antigoon, Cnudde), or salient landmarks (Manneken Pis). Linguistically interesting, to

50

Dirk Geeraerts

round off, are those names where a local identity is evoked by dialect forms (which are often at the same time archaic). The relevant linguistic features include inflection (Stoeren Bonk would be Stoere Bonk in standard Dutch), pronunciation (Duvel instead of Duivel), or spelling (as the typically archaic sch in Oude Tongerschen Bruineri). These features may, of course, be combined with each other and with values discussed before. The result may be quite surprising: in particular, notice how the orthographical form Lowie Kators for the name of the French king Louis Quatorze (cf. the source domain of royalty and aristocracy) evokes a Dutch, dialect-like pronunciation, and as such ironically combines the lofty position of the foreign king with the down-to-earth status of the local dialect. There are three additional remarks to be made with regard to this overview of associated values. First, it is not always easy to get the etymological origin of the names right, or rather, the actual etymological origin is not necessarily the one that is relevant for the value that seems to be evoked now by the name. For instance, Mort Subite is 'Sudden death'. To the extent that it is interpretable at all, it could only be in terms of the physiological impact of strong beers. Originally, however, the name was that of a particular cafe plus brewery in Brussels; the name of the cafe itself apparently referred to a particular type of backgammon popular in the pub. Some methodological caution, in other words, is necessary when dealing with this kind of material. As a second remark, it will be noted that the list contains both French and Dutch names. In principle, it might have been possible to analyse both sets of names separately, on the assumption that we might then get an insight in the differences between the culture of the Dutch speaking part of Belgium (Flanders) and its French speaking part (the Walloon country). I will, however, proceed on the assumption that, as far as beer goes, Belgian culture is fairly uniform. This is supported by the recognition that there is no unambiguous link between the language of the brand name and the geographical position of the brewery. Some French names, for instance, derive from breweries in Flemish territory. Moreover, it appears from sociological surveys (most recently, Kerkhofs et al. 1992) that the opinions and behaviour that distinguish Flemings from Walloons are by far outweighed by what they have in common in contrast with the Dutch, Germans, or French (to name only the immediate neighbours). A third remark is simply that a specific name may be motivated by various evocative values at the same time, as has become apparent at various points

Beer and semantics

51

Table 3. Recurrent associative values in beer names Source domain

Examples

Beer is a prestigious entity Wine terminology Royalty & aristocracy; VJPs

Historical

figures

Wealth & riches

Vlaamse Bourgogne, Cuvee Chateau des Flandres, Caves, Rose de Cambrinus Ada Royal, Grand Monarque, King's Ale, Regal, Ridder, Royal pils, Marquise Monseigneur, Moeder Overste, Kapittel abt Ambiorix, Artevelde, Breydel en de Coninckbier, Hertog Jan pils, Charles Quint Breugelbier, Rubensbier, Rembrandt's bier, Vondel, Mercator, Rodenbach Lowie Kators, Napoleon, Da Vinci Fortuyn, Gold Pils, Gouden Carolus, Parel Pils, Silver Pils, Safir, Toison d'Or

Beer is an adventure at the edge of the accepted norms Transgression of religious norms Duivels bier, Duvel, Hopduvel, Judas, Lucifer, La Pecheresse, Satan, Verboden Vrucht, Sorcieres d'Ellezelles Transgression of social norms Boucanier, Piraat, Smokkelaar, Schavuit, Stouterik, Averechtsen, Bokkereyer, Deugniet, Bakelandt, Brigand, Kerelsbier, Voyou Virility and force Het geheim van Jan de Sterke, Kastaar, B ink, Brugse Straffe Hendrik, Stoeren Bonk, De Soldaat Strong bodily effects Delirium Tremens, Houten Kop, Katerke, Killer Pils, Mort Subite Beer is a source of pleasure Carpe diem

Heavenly pleasure

Bacchus, Cupido, Drongens plezierke, Flierefluiter, Joyeuse blonde, Lekkerbek, Pallieter, Plaisir, Relax, Cöte d'Azur speciale Engeltjesbier, Bons voeux, Sanctus, Onze vader z'n bier + various names of saints

Beer constitutes a link with the traditional culture of the pre-industrial era Traditional methods of producBrouwmeester, Gildenbier, Keurbier, Tradition, old trades tions des Moines, Oerbier, Retro oude methode, Natuurbier, Saison 1900 cont.

52

Dirk Geeraerts

Table 3. (continued) Source domain

Examples

Rural origins

A la ferme, Bosbier, Boskeun, Verte campagne, Ferme de Grand Pre, Pays vert, Peerdevisscher, Oelens boerke Antiek, Vieux Temps, Druide, Blonde du Menhir, 1'Iguanodon + see below ('Local history') and before ('Historical figures') as before: Monseigneur, Moeder Overste, Kapittel abt Engeltjesbier, Bons voeux, Sanctus, Onze vader z'n bier + various names of saints

History & prehistory

Religion

Beer is an aspect of local identity Dialect forms & archaic forms

National or local history

Local folkore

Duvel, Lowie Katars, Stoeren Bonk, Peerdevisscher, Drongens plezierke, Oude Tongerschen Bruinen Ambiorix, Artevelde, Bakelandt, Breydel- en de Coninckbier, Kerelsbier, Hertog Jan pils Breugelbier, Rubensbier, Rembrandt's bier, Vondel, Mercator Gouden Carolus, Toison d'Or Elckerlyc, Pallieter, Prutske bier, Hercule Poirot Antigoon, Cnudde kriek, Het geheim van Jan de Sterke, Ketje, Rosse Lei, Rooie Rietje Stekselbier, Tineke van Heule, Manneken Pis, Geitenbier

in the preceding discussion. A name like Ambiorix, for instance, refers to the leader of one of the Celtic tribes that rebelled against Julius Caesar's invasion of this part of Gaul. In this sense, the name Ambiorix combines the motifs of local identity, of prestigious leadership, of the remote historical past, and of resistance to an imposed authority. The fourth remark is the crucial one. If beer is as central to the Belgian selfconception as I claimed earlier, do these aspects of Belgium's conception of beer reveal something about Belgium's self-conception? Methodologically,

Beer and semantics

53

answering such a question requires going beyond linguistics and cooperating with sociologists, historians, and anthropologists. That is more than could be done in the space of this paper, but let us note that linguistics is not entirely unequipped to deal with the question. At least one thing we can do as linguists is to compare the Belgian situation with that in other countries. Taking The Netherlands as a point of comparison, some differences do appear that support the idea that the pattern that emerges for Belgium is a culturally specific one. For instance, while the prestige motif is as outspoken in The Netherlands as it in Belgium (which may be due to a universal commercial strategy of appraisal), the link with the pleasure motif is much weaker, and references to local identity (at a level below that of the country as a whole) are rather scarce. So are references to religion. Most tellingly, perhaps, references to anti-authoritarian transgressive behaviour are almost entirely absent. These differences can be meaningfully interpreted against the background of the different cultural and political history of both countries. (These differences can only be summarily evoked here. For more background information, see Verdoodt 1978, where a sketch of the cultural and political history of the Low Countries provides the backdrop for a sketch of the divergent linguistic evolution of Belgian Dutch and Netherlandic Dutch.) First, the protestant, calvinistic religion that was dominant in Holland from the 16th century on did not favour the enjoyment of earthly goods, while the Catholic religious atmosphere in Belgium was much less austere. This may explain why the link with the religious motif and with the pleasure motif is almost absent in the brand names found in The Netherlands. Second, Holland became a nation, a unified national state in the 17th century, while the Belgian provinces remained under foreign (Spanish, Austrian, French) influence until the 19th century. This correlates with the fact that Dutch people identify more readily with the national state, whereas the inhabitants of Belgium traditionally have a genuinely irreverend attitude with regard to the national political authorities at large. Apparently, then, this difference in mentality also shows up in the different patterns of associated values for beer names in both countries.

6.

Concluding remarks

Possible conclusions to be drawn from this field trip through lexical beer country include the following. First, a systematic investigation into salience phenomena in lexical fields should be high on the list of methodological pri-

54

Dirk Geeraerts

orities of Cognitive Linguistics, if it is to reach its aim of doing cultural analysis by linguistic means. Second, in analyzing such salience phenomena, a distinction has to be maintained between the salience of 'literal' referential features and the salience of associated values. In the foregoing pages, this distinction showed up in the distinction between 'type Γ and 'type 2' names. Third, proper names (and in particular, commercial brand names) are amenable to a cognitive analysis in terms of recurrent associative and metaphorical values. Fourth, the wealth and diversity of Belgian beers is reflected in the variety of beer names; moreover, there is at least some evidence that the semantic patterns that emerge from the linguistic analysis correspond with real culturally specific aspects of behaviour and mentality. And fifth, the lexicology of beer names can be fun - but whether it is more fun than the tasting of beer itself is for the reader to explore.

Notes 1. This paper is a revised version of the Opening Address presented on July 18, 1993 at the 3rd International Cognitive Linguistics Conference. The results presented in the article are to a considerable extent the outcome of a seminar that I conducted in the Spring term of 1993 with a group of 20 last year students. I explicitly want to acknowledge their contribution to this paper.

References Berlin, Brent 1976 1978

The concept of rank in ethnobiological classification: some evidence from Aguarana folk botany. American Ethnologist 3: 381-400. Ethnobiological classification. In E.R. Rosch and B.B. Lloyd (eds.), Cognition and Categorisation. Hillsdale, New Jersey: Lawrence Erlbaum Associates.

Geeraerts, Dirk 1989 Problems and prospects of prototype theory. Linguistics 27: 587-612. 1993

Generalised onomasiological salience. Belgian Journal of Linguistics 8: 4356.

Geeraerts, Dirk, Stef Grondelaers and Peter Bakema 1994 The Structure of Lexical Variation. Meaning, Naming, and Context. Berlin: Mouton de Gruyter. Kerkhofs, Jan et al. 1992 De versneide ommekeer: de waarden van Vlamingen, Walen en Brusselaars in dejaren '90. Tielt: Lannoo.

Beer and semantics

55

Kleiber, Georges 1991 Hierarchie lexicale: categorisation verticale et termes de base. Semiotiques 1: 35-57. Lakoff, George and Mark Johnson 1980 Metaphors we live by. Chicago: The University of Chicago Press. Langacker, Ronald W. 1987 Foundations of Cognitive Grammar, Vol. 1. Stanford: Stanford University Press. Lehrer, Adrienne 1983 Wine and conversation. Bloomington: Indiana University Press. Verdoodt, Albert (ed.) 1978 Belgium. The Hague: Mouton. Speicla Issue of (International Journal of the Sociology of Language 15).

Is everything black and white in conceptual oppositions? Gabor Györi and Iren Hegedus

1.

Introduction1

In Indo-European languages there are curious lexical correspondences that are safely treated as cognates, although their semantic contents, as reflected in the descendant languages, reveal antonyms and opposites. Such a seemingly contradictory semantic development within a cognate group can be discerned in an etymological survey.2 This phenomenon is current enough in the Indo-European stock (cf., e.g., York 1993)3 and also in other language families (e.g. Finno-Ugric) to give rise to the assumption that it reveals important facets of human cognition (cf. Key 1979). In this paper we seek to deduce implications of this diachronic phenomenon for cognitive linguistic studies and examine what cognitive mechanisms might be instantiated in such linguistic developments. On the other hand, we also try to show what cognitive semantics has to offer for diachronic studies, especially in explicating the semantic developments in question. The phenomenon of a lexeme acquiring an opposite meaning and at the same time retaining the original one is absolutely not uncommon. E.g. Latin dare 'give', Proto-Slavic *da- 'give', etc. and Hittite da- 'take' are cognates deriving from PIE (Proto-Indo-European) *do~ 'give'.4 In this case we can witness a semantic development from one pole to its opposite. We are not trying to explain this type of semantic change, but rather suggest that another type of process could also have led to the emergence of cognate sets involving opposites, especially in the case of basic oppositions, i.e. oppositions in basic perceptual domains. Thus, we will limit our investigations to basic oppositions and examine how their cognitively special status may have influenced their semantic development (see Tables 1-4 in the Appendix).

2. Problems in semantic reconstruction There is no doubt about the fact that the surveyed lexical items in each table can be phonologically traced back to a common protoform. The regular sound

58

Gabor Györi and Iren Hegedüs

correspondences that are required for the reconstruction of a common etymon are there. But what about the meanings of these forms? Though regularity of sound change is of course the more compelling evidence to reconstruct an etymon, this cannot be done without finding some semantic relationship between the postulated cognates, i.e. we have to be able to supply a plausible explanation for the semantic development from an earlier underlying meaning. In the above cases the semantic relationship is there: that of oppositeness, but this makes it very difficult to account for an earlier common meaning. In general we have to be circumspect when reconstructing meanings. At present we do not seem to have a better procedure than semantic reconstruction based on a feature analysis of meanings, which gives us some kind of a lowest common denominator that we posit as the original meaning. However, as Sweetser (1990: 24) has clearly shown, this does not seem to be realistic at all, since such a procedure yields a protovocabulary full of abstract meanings, which is contrary to our knowledge of semantic change running from concrete to abstract in the vast majority of the cases. In her opinion, a cognitive theory of meaning cannot subscribe to the idea that the basic mechanisms of semantic change can be reduced to loss, addition and recombination of semantic features. As Langacker (1987: 157) has pointed out, an encyclopedic view of meaning, contrary to an autonomous feature-based approach, is especially suited to explain semantic extension. Thus, the major controversy does not seem to lie with the idea of an original common meaning, but with the method of reconstructing one. In any case, no semantic reconstruction can be initiated unless we can find some semantic relationship between our tentatively cognate forms and supply a plausible explanation for the given development. Since it appears to be difficult, at first glance, to find a plausible explanation for semantic developments manifest in the examples (Tables 1-4 in the Appendix), theoretically we could postulate that the protoform that has been arrived at by phonological reconstruction had two meanings and what we have to do with are "binary homophones" or "homophonic antinomies" (York 1993: 238). However, an explanation as cases of homonymy does not seem to be really convincing because of two reasons. On the one hand there is a relatively large number of cases with this type of semantic correspondence, and on the other, this type of rather disturbing homonymic clash would probably have resulted in the lexical change of one of the lexemes of the antonym pair. It is well known that lexical change is a general solution for such cases

Is everything black and white in conceptual oppositions?

59

in the development of languages (cf. Palmer 1978: 331, Hock 1986: 298). Thus, if we had to do with homonymic clash in these instances, it would be rather puzzling why lexical change did not occur in their case. Let us examine the possibility of polysemy as well. Is it possible that these lexemes had always been polysemous? Our present knowledge of semantic change tells us that this is not possible because polysemy always implies some kind of a prior historical bifurcation process. Based on this, we could say that at the given stage of language development (Proto-Indo-European and Proto-Finno-Ugric in our examples) the bifurcation process had already taken place and what we have to do with are polysemous lexical items after all. However, this attempt at a plausible solution brings up another problem. In the case of polysemy there is always one meaning that can be considered basic, i.e. all other meanings are explicable as derived meanings. Such an explanation is a possibility only to the extent to which e.g. the concept WHITE can be assumed to derive from BLACK or vice versa. A plausible way out could be that the basic meaning was already extinct in Proto-Indo-European times and we have to do with two parallel extensions from it. The strange thing is, however, that these parallel developments are symmetrical in the sense that they are exact opposites, i.e. the two poles of a conceptual domain.

3. The cognitive background Having outlined the problems manifest in the above etymologies and considered them from the point of view of semantic change in some detail, we can now move on to investigating what sort of conceptualization might underlie such phenomena. Our starting point is the fact that any phenomenon can only be conceptualized on the basis of its relationship to other phenomena. The importance of this for semantics is that a predicate always has to be characterized relative to a domain (Langacker 1987: 147). The domains we have to do with here are all basic domains, since the oppositions dealt with are all perceptual oppositions, i.e. our knowledge about them derives from direct bodily interaction with our environment (see Table 5 in the Appendix). Since exclusively basic domains, and no abstract ones are involved in the characterizations of the predicates in question, it is crucial to have an exact understanding of what it means more specifically that our knowledge about these experiential domains derives from direct bodily interaction with our environment. The following consideration will reveal the importance of this.

60

Gabor Györi and Iren Hegedüs

When we say that a predicate is characterized relative to a domain, we mean that a domain used to characterize a predicate is always one level less abstract than the domain that the predicate itself defines. This holds for most predicates (cf. Langacker 1987: 150). Thus usually, if a predicate is characterized by a basic domain, the predicate itself already pertains to an abstract domain that is one level above basic level. However, the predicates treated here all pertain to basic domains directly. Langacker (1987: 149) writes: "By definition, basic domains occupy the lowest level in hierarchies of conceptual complexity: they furnish the primitive representational space necessary for the emergence of any specific conception." For our considerations this means that the predicates in question (viz. our etyma) can only be characterized by the domains to which they themselves belong, since these domains are not reducible any more to more fundamental ones. As such a situation exhibits a special case, we must also assume a special mechanism for characterizing such predicates. This is why we have to examine in more detail how knowledge about such domains is exactly derived. In our direct bodily interaction with the environment, i.e. in our physical experience gained through our sensory organs, we detect recurrent perceptual patterns (e.g., visual, kinesthetic, etc.). This gives rise to preconceptual configurations in our minds, which are then used to organize our experience. These preconceptual configurations are what Johnson (1987: 29) and Lakoff (1987: 278) call image schemata. We perceive our experience as organized because such image schemata structure the domains of our experience. Since relying on previous experience is a fundamental property of the mechanism of structuring all experience (Langacker 1987: 105, Seiler 1985: 117), image schemata also function in imposing structure on more abstract domains of experience. It is by projecting them onto such more abstract domains that we make these meaningful. Since the domains being considered here are all physical ones, they must be directly structured by image schemata. The image schemata that structure them are not projected on them from other domains because these are the domains in which these image schemata originally exist. In other words, basic domains are structured by their own image schemata. The next step we have to take in our investigation is to examine how image schemata structure domains, especially basic ones. As Johnson (1987: 41) has shown, they are capable of structuring domains because they themselves have an internal structure that they impose on the given domain. Thus, e.g.

Is everything black and white in conceptual oppositions?

61

the UP-DOWN image schema, which arises from visual and kinesthetic perception, imposes its internal structure on the domain in which it exists (or on a domain it is projected to). This is why we conceive of space as extending upwards and downwards among others. The internal structure of image schemata is basic level structure in the sense that it is analyzable but not decomposable, i.e. its elements are inseparable from each other. This is so because they pertain to basic domains, which are not characterizable relative to other domains, but only to themselves. Johnson defines the internal structure of image schemata as experiential gestalts. Image schemata have internal gestalt structure as they are coherent unified wholes within our experience. He writes: "Any given schema can, of course, be analyzed and broken down simply because it has parts. But any such reduction will destroy the integrity of the gestalt, that is, will destroy the meaningful unity that makes it the particular gestalt that it is" (Johnson 1987: 44). It is important to bear in mind that image schemata are characterizable as irreducible gestalts. Most gestalts can be broken down physically, which destroys their integrity, but what we get are new gestalts. In the case of image schemata, however, we have to do with basic gestalts, which can only be broken down theoretically, but not physically. What should be understood then by irreducible gestalts? For any phenomenon to have a structure, at least two contrasted elements are needed. Since every image schema possesses two such elements (see below for more explanation), it is obvious that they have an internal structure. However, it must also be obvious that these structures are irreducible basic gestalts. There is no OUT without IN, there is no DOWN without UP, etc., and vice versa. In other words, we cannot conceptualize one element of the structure without the other. Pertaining to the basicness of an image schema, there is also a seeming circularity in its conceptualization (or preconceptualization). Let us illustrate this with the example of the CONTAINMENT schema. We cannot have a sense of IN and OUT unless we have a sense of CONTAINMENT, and we cannot have a sense of CONTAINMENT unless we have a sense of IN and OUT. But this circularity should not disturb us; it is a result of irreducibility, i.e. a true sign of basicness. This irreducible structure is not only characteristic of image schemata. In general it can be stated that we cannot perceive something as something unless there is something else to contrast it to in the given perceptual domain, i.e. across one perceptual modality. As Langacker (1987: 101) has pointed

62

Gabor Györi and Iren Hegedüs

out, comparison is one of the most fundamental cognitive abilities relevant for semantic structure. Anything perceived will gain significance in our cognitive processing only in as much as it functions as a target that is compared to a standard. Because of this, for the perceived element of an opposition to be able to stand out as a figure the other element necessarily has to serve as the ground (cf. Langacker 1987: 120). Thus, it is the nature of oppositions, just like that of image schemata, that they can only be perceived as gestalts, i.e. as unified wholes made up of two poles of a domain. Above we assumed that the basic internal structure of image schemata always involves two contrasted elements; therefore we also think that image schemata naturally have an internal polarity. All the image schemata mentioned in Lakoff (1987) conform to this assumption. Thus only by way of example, the LINK schema involves an entity A as separate from an entity B to which it is connected, though of course the relationship is symmetrical (Lakoff 1987: 274), or there is the PATH schema, which by definition has to involve a source and a goal as separate points in space making up the basic structure of a path. Even schemata like LINEAR ORDER or FORCE conform to this "polar" structure. At least two separate entities are necessary to make up the most primitive linear order and in the case of force there always has to be one entity exercising force on another separate entity, since "there is no schema for force that does not involve interaction" (Johnson 1987: 43). Everything we perceive is perceived on the grounds that it contrasts with something else, i.e. we compare it to other experience or conceive of it as a figure that stands out against a contrasting ground (cf. Langacker 1987: 121). Because of this we postulate that the natural way for basic perceptual oppositions to appear preconceptually is in the form of image schemata (and maybe there is also some ground for the claim that image schemata are preconceptual forms of perceptual oppositions). The basis for such a claim lies in the observation that it is exactly perceptual oppositions that structure our sensations and make them meaningful, i.e. we make sense of our perceptions in terms of oppositions. This can well be observed in fundamental cognitive mechanisms as comparison or figure/ground alignment, where a target always stands in opposition to a standard or a figure to a ground respectively. These oppositions basically include the perception of features vs. non-perception of features on the one hand, and the perception of different degrees along a scale on the other hand.

Is everything black and white in conceptual oppositions? 4.

63

The lexicalization process of oppositions

Let us now examine our language material in the light of the above theoretical considerations. On the basis of these, we assume that just as it is impossible to conceptualize one pole of a conceptual opposition without conceptualizing the other, or conceptualize only one part of an image schema, it is impossible to have a lexical gap at one pole of an opposition. We assume that either the domain in question is not coded in any way in a language or both poles have to be coded. Michael York (1993: 238) writes that "[tjhese homophonic antonymies are not a universal feature of the proto-language, but one which occurs often enough to be indicative of the possible IE tendency toward polarized perception [...]". We think that this phenomenon should not be called homophony, because it is exactly "polarized perception" that requires both poles to be coded in parallel at any historical stage in a language and this presumes a simultaneous coding process. This can only be the case if the coding of the two poles has one common origin. This type of simultaneous coding, i.e. deriving the lexemes for the two poles from one common source, is not simply a convenient linguistic solution, but as we have seen, also has psychological reality. At least at the conceptual (or preconceptual) level the simultaneous recognition of the two poles is compelling. This would speak for the universality of this phenomenon, which of course might have already become obscure in many cases by the PIE stage. On this basis we try to give a new interpretation to the semantic developments within the cognate groups of the conceptual oppositions we dealt with above. We have seen that a conceptual opposition appears preconceptually in an image schema. Because of this we think that the etyma in question are cases in which a complete image schema became lexicalized, i.e., coded in the language. We suggest that the notion "archilexeme" can be usefully applied in the historical sense to this linguistic phenomenon. An archilexeme, as used in lexical field studies, is a lexeme which neutralizes two poles of a semantic dimension that are distinct in lexemes with otherwise identical semantic contents (cf. Kastovsky 1988: 197, Mettinger 1988: 151). Cruse (1986: 255) defines neutralization as the non-appearance of a semantic contrast. Thus by way of example, the lexeme child is an archilexeme in the sense that the gender difference readily apparent in its two hyponyms boy and girl is obliterated in it (cf. Mettinger 1988: 151). By utilizing the notion archilexeme in a historical frame we can consider our etyma to be archilexemes in the sense that the semantic contrast appar-

64

Gabor Györi and Iren Hegedüs

ent in their later developments can still be found in a neutralized state in the PIE etyma themselves. In cognitive linguistic terms we can say that the image schema structures were still coded (lexicalized) in a homogeneous way at the protolinguistic (e.g. PIE) stage. We assume that these archilexemes were at best only internally differentiated for poles, and expressed a preconceptual idea of 'verticality', 'temperature', and 'darkness-brightness'. However, the internal dynamism of the underlying schemata (cf. Johnson 1987: 29), due to their polarity, caused their splitting up into their poles, and a semantic development into opposite directions began. This splitting up took place of course only at the lexical level, because as we have seen, an image schema is not decomposable at the conceptual level. Thus, the etyma gave rise to parallel but contrasting lexical extensions in their own rights.5 This kind of semantic development can plausibly be postulated for basic oppositions since the conceptual inseparability of the two poles makes it likely that the development was not from one pole to the other, but it was a parallel one starting out from a common etymon denoting the complete notion of the opposition as such with only internally differentiated poles. In the case of non-basic oppositions, however, especially if they are non-binary, i.e., the two poles are conceptually separable, a semantic conversion of one pole into the other seems to be possible. An example for this is the already mentioned case of PIE *do- 'give', which retained its meaning in Latin (dare 'give') and in the Slavic languages (e.g. Russian dot' 'give') but changed its meaning to its opposite in Hittite (da- 'take'). The semantic reconstruction and thus the postulated semantic development are plausible because the two notions can be maintained independently: taking does not necessarily imply giving by the other party and vice versa. In spite of this, it would be mistaken to assume that such an opposition has no image schema structure. The opposition GIVE vs. TAKE is non-basic in the sense that several more basic domains form its matrix. Minimally it can be reduced to the basic oppositions contained in the LINK and SOURCE-PATH-GOAL schemata. In the following we examine the semantic developments of PIE *bhel- and take a look at the reconstructed meanings of its derivatives. The structure of the cognate group is depicted in Figure I.6 As can be seen from the figure, the semantic reconstructions within the cognate group postulate that the notions of 'black' and 'white' are extensions of 'burn'.7 Standard etymologies postulate this development as a kind of mixture of a plausible semantic explanation and feature analysis (yielding the

Is everything black and white in conceptual oppositions?

65

PCelt. *bslo- 'shining white' > V f . bal 'white blazed' Proto-Slavic *beit 'white' > Russian Jbelyj 'white' Greek phalarom 'white spot' ->

Armenian JbaJ 'palenese' Baltic: Lith. bSlaa 'white',

Latv. balm 'pale'

Albanian bal 'white haired; blazed animal' EXTENDED ROOT

— Proto- Germanic

*bla±k-Jan 'to make white' '— OK blaoan 'bleach' *blaik-ax 'shining, white' Norse bla±kr 'shining, white' b Old OE blac 'pale, white, bright' *blikkat-jan 'to flash, lighten' I— OHG blScchaMMon 'id. r > Germ, blitxon 'id.' *bla-a- 'shining, white' OE blsao 'torch' Middle Dutch bl«a 'white spot' » Old French ble(*)mir 'to make pale' *blend-/bland- 'to shine, dazzle, blind' OE blind 'blind' b OHG blontan 'to blind,

deceive'

*blonk~/blank- 'to shine, dazzle, blind'

OE blancan 'to deceive' b OHG blanc 'bright' 1

— » Old French Wane 'white'

— *blJ.sk- 'to shine, burn' I— OE blyecan 'to glow red' -> blueh Proto-Germanic "blSvax 'blue' Latin flSvua 'golden, reddish yellow' •iAleg-

Proto-Germanic *blaka* 'burned' I— OE blmc 'black' Latin /ulgfio 'to flash, shine·, flagrO 'to burn'

•LA,

Greek phlogoin 'to burn', phlox 'flame'

Tokharian pa'lfc- 'to shine, flash; to burn'

Figure 1. Cognate tree of PIE *bhel-

66

Gabor Györi and Iren Hegedüs

lowest common denominator of 'bright'/'dark'/'burn'). The semantic extensions can be said to be based on the metaphor BRIGHTNESS/DARKNESS IS BURNING. The metaphorical projection from the source domain BURNING to the target domain BRIGHTNESS/DARKNESS (cf. Lakoff 1990) can be explicated in the following way: Brightness/darkness is caused by burning in the sense that when something burns there is brightness but at the same time the thing that bums will become dark (i.e. burned) and because of this brightness/darkness is conceptualized as the state(s) caused by burning. However, based on the distribution of the different meanings in the daughter languages, Crepajac (1967) claims that PIE *bhel- could not have meant 'burnt', from which the two other meanings branched off, but rather denoted some light shades of colour. If we take a closer look at the cognate group, it in fact becomes apparent that terms denoting some shade of colour are more frequent than terms denoting 'burn'. This fact together with the psychological reality of simultaneous conceptualization (or preconceptualization) of oppositions seems to indicate that the metaphorical projection might have run in the other direction as well. Since both domains (source and target) are physical ones, this is not in the least inconceivable. A reversal of source and target domains in this case may be even more realistic on the grounds of the compelling simultaneity of the coding of the two poles. This would speak for the initial coding of a coherent image schema as a holistic phenomenon. On this view, the metaphor BURNING IS BRIGHTNESS/DARKNESS could be postulated as the basis of the extensions. Beside the above reasons for such a reversal, there is an even more fundamental motive. On the basis of Langacker's (1987: 149) distinction between basic and abstract domains, only the domain of BRIGHTNESS/DARKNESS can be considered to be a basic one, because it is grounded directly in sensory experience. Even though both domains (i.e. both BRIGHTNESS/DARKNESS and BURNING) pertain to physical reality, only the conception of BRIGHTNESS/DARKNESS relies directly on our visual sensation of light and is irreducible in the sense that it cannot be explicated in terms of more fundamental domains. BURNING on the other hand is clearly an abstract domain in a cognitive semantic approach, since it is readily characterizable in terms of, i.e. reducible to, basic domains like the sensation of light and temperature. It is on this basis that BRIGHTNESS/DARKNESS is more likely to function as source domain for the projection (and semantic extension) to the target domain BURNING. Since 'burning' is saliently characterized by 'brightness/darkness', 'burning' is best conceptualized as

Is everything black and white in conceptual oppositions?

67

the case when there is 'brightness/darkness' (i.e. light plus burned substance as the outcome). In the case of PIE *kel- 'burning, warm' and PFU *pal'a 'icecrust, frost; to freeze' (cf. Tables 2 and 4 in the Appendix) the semantic contents of the original etyma serve as an even less adequate starting point for a plausible explanation of semantic development, since the reconstructed meanings are reduced to only one pole of the opposition. However, the semantic reconstruction of PIE *upo 'under, up from under, over' (cf. Table 3 in the Appendix) seems to be more in line with our approach to the lexicalization of perceptual oppositions. We feel that our investigations of the semantic development of PIE *bhelalso have a bearing on the theory of the evolution of basic colour terms. Berlin and Kay (1969) consider the stage in the evolution of colour terms when only macro-white and macro-black are coded in a language to be stage one. On the basis of what has been said above, we postulate a stage prior to stage one, i.e. to the 'macro-white «->· macro-black' stage, which we would like to call stage zero, i.e. the stage of the coding of the undifferentiated image schema. We think that such a stage zero can be postulated for all the investigated oppositions. By this we do not wish to claim that people in those times did not have the perceptual and cognitive capacities to differentiate between the different poles of these oppositions. The question when such oppositions were realized belongs to the realm of pure speculation. However, through the methods of historical linguistics we are able to conduct investigations into lexicalization processes, from which we can make indirect inferences since they are inseparable from some kind of underlying conceptualization (Györi 1996). Note also that the term "protolinguistic" is not used to refer to a rudimentary linguistic stage, but to "the earliest form of a language established by means of the comparative method of reconstruction" (Jeffers and Lehiste 1982: 183), i.e. to a protolanguage, such as PIE, Proto-Uralic, etc. (cf. Hegedus 1997 for a taxonomy of stages and issues of time-depth). Even by keeping the above reservations in mind, several problematic questions immediately arise in connection with stage zero. The most urgent ones are the following: How real is stage zero? What could the real meaning of the archilexeme have been? Can we postulate a realistic meaning, i.e. a realistic archisememe? Below we would like to explicate the answers cognitive semantics has to offer to these questions. A totally neutralized form would be communicatively inefficient and un-

68

Gabor Györi and Iren Hegedüs

thinkable; it could serve no communicative purpose. However, as we have postulated, these archilexemes are lexicalized image schemata and because of this they should reflect the nature of image schemata. An image schema is a gestalt, a unified whole, but has internal structure given by its polarity. Thus, the archilexeme cannot be anything but a gestalt itself that displays internal polarity. The Lithuanian antonym pair Mtas 'warm' · saltas, 'cold' may serve as a model for illustrating this idea. In this word pair the consonant structure can be considered to give the gestalt (i.e. the perceptual dimension) and the vowels together with suprasegmental features (circumflex vs. acute accent) provide the internal polarity. Thus, in PIE the expression of the internal polarity could have been realized by different ablaut grades, or if PIE indeed had politonous accent (which may also be surviving in Lithuanian), then it could even have been realized by this, i.e. rising, falling, or rising falling accent, or as a third alternative, simultaneously by both ablaut grades and accent differences. Again Lithuanian may provide a good example: it has an accent that can be semantically distinctive. Many linguists consider Lithuanian to be archaic in this respect, and this may be a relic feature surviving from PIE times (cf. Szemerenyi 1972: 137).

5. Conclusion Our purpose in this paper was to attempt a clarification of the cognitive processes that could have underlain seemingly contradictory semantic developments within groups of etymologically related words. We think that the etyma investigated in our paper could very well be relics of an archaic process in which the two poles of an image schema gained lexical representation. The reflexes in the Indo-European daughter languages seem to have retained the former semantic unity of the archilexemes to different degrees. In fact the historical development of the cognate groups tends more and more to obscure this cognitively motivated semantic unity by eliminating (replacing) one member of the cognate word pair carrying the opposite meanings. Yet a few Indo-European languages still sporadically display the opposite semantic poles within one cognate set, as we have tried to demonstrate. Outside the Indo-European language family, such relic cognates are also attested in Finno-Ugric languages and probably in others as well, as we suspect. Furthermore, such a process is not necessarily confined to the protolinguistic stage but it may be inherent in language change. How widespread such pro-

Is everything black and white in conceptual oppositions?

69

cesses might be in different language families and whether such processes are still operative in language change of less remote periods should be the subject of further research, to which this preliminary study could serve as a theoretical basis.

Appendix Table 1. Pffi *bhel- 'to shine, flash, burn' χ = feasible but not surviving pair black

white

χ

Russian belyj 'white' < Proto-Slavic *bel$ 'white'

χ

Lithuanian b las 'white' Latvian b ls 'pale, palid'

χ

Armenian bal 'paleness'

Albanian blaze 'soot'

Albanian bal 'white haired; blazed animal'

χ

Welsh bal 'white blazed' Gaul, balio- 'white blazed' < PCelt. *bdo- 'shining, white'

OE bcel 'flame, funeral pile' ONorse b l 'fire' OE blcec 'black' Icelandic blakkr 'dark' Swedish black 'ink' OHG blah 'ink' OSwed. blakker 'black, dark' < PGmc. *blak- 'black'

OEng. blaec-an 'to bleach' ONorse bleikr 'shining, white' Swed. blek 'pale, fallow' Germ, bleich 'pale' OSwed. blakker 'pale, fallow' < PGmc. *blaik- 'white'

χ

Lal.fiilgeo 'to flash, shine'

χ

Greek phalos 'white'

χ

Tokharian polk- 'to shine'

70

Gabor Györi and Iren Hegedüs

Table 2. PEE *kel- 'burning, warm' = feasible but not surviving pair hot/warm

cold

Lithuanian siltas 'warm'

Lithuanian saltas 'cold'

Latvian silts 'warm'

Latvian salts 'cold' O.Ch. Slavic slana 'hoarfrost'

German lau 'tepid' ONorse hlyr 'warm' OE hleo(w) 'cover (from cold)' < PGmc. *hlewaz Latin calidus 'hot, warm' Italian caldo, French chaud Welsh clyd 'warm, warming* Avestan sarata-, Persian sard 'cold', Sanskrit s'isira- 'cool, cold' (reduplicated)

Table 3. PEE *upo 'under, up from under, over' = feasible but not surviving pair

up

under

PGmc. *upp- 'up, in' > OE up-, upp-, Goth. iup 'upward' PGmc. *ufana On, above' Gothic ufar Over' Welsh gorau 'best' /uppermost/ < PIE *uper-esu

Gothic uf 'under' PCelt. *vo- 'under' PCelt. *wasso- 'servant' /:one who stands under/' < PIE *upo-sto-

Greek hupsos 'height, top', hyper Over'

Greek hupo 'under'

Old Indie upama 'uppermost'

Skt. upa 'near to, under'

Latin subiciö 'to lift up'

Latin sub 'under' < PIE *(s)up Latin subiciö 'to place under'

Hittite up-zi 'rise' (of the sun)

Is everything black and white in conceptual oppositions?

71

Table 4. PFU *pal'a 'icecrust, frost; to freeze' χ = feasible but not surviving pair burn (hot)

freeze (cold) Finnish palele- 'feel cold, freeze' Lapp buol s- 'frost, frosty' Mordvin polo- 'freeze'

Finnish pala- 'bum, flame' Lapp buolle- 'bum' Mordvinpa/o- 'burn' Estonian pala- 'burn' χ

χ

Hung.fagy 'frost; freeze'

Table 5. Basic oppositions and their grounding in bodily experience opposition

perceptual modality

sensory domain

black : white hot : cold up : down

visual haptic visual, kinesthetic

colour temperature space

Abbreviations Gaul. Germ. Hung. Lat. Latv. Lith. O.Ch.Slavic OE OHG ONorse

Gaulish German Hungarian Latin Latvian Lithuanian Old Church Slavic Old English Old High German Old Norse

OSwed. PCelt. PFU PGmc. PIE PSlav. Skt. Swed. W.

Old Swedish Proto-Celtic Proto-Finno-Ugric Proto-Germanic Proto-Indo-European Proto-Slavic Sanskrit Swedish Welsh

Acknowledgements We are grateful to Raimo Anttila, Wolfgang U. Dressler, Lena Ekberg, Mary Ritchie Key and Joseph D. Salmons for their comments, suggestions and corrections on an earlier version of this paper. Naturally, all the remaining errors are ours.

72

Gabor Györi and Iren Hegedüs

Notes 1. The preparation of this paper was supported by a grant from the Hungarian National Fund for Scientific Research [project no. OTKA T006762]. 2. For the etymologies found in the text the following main sources were consulted: Pokorny (1959-69), Redei (ed.) (1986-91), Watkins (ed.) (1985). 3. We owe this reference to Joseph D. Salmons. 4. We are indebted to W. Dressler for this example. 5. It could be hypothesized here that due to this kind of development both poles of these opposites should be considered unmarked (and thus natural), at least in their initial states, if such a case is possible at all. However, this claim is very tentative and requires further investigation. 6. The figure shows the difference between words that are reflexes of the PIE root with or without a root extension. The function of root extensions is not yet perfectly clear in Indo-European linguistics: most often they convey grammatical information (e.g. determine the declination class) rather than modify the root semantically (cf. Lehmann 1993: 240-241, Szemerenyi 1990: 104-105). Thus we should not simply ascribe the particular semantic modifications to the presence or absence of these root extension morphemes. 7. It should not be forgotten here that on the basis of the Berlin-Kay investigations (cf. Berlin and Kay 1969) these notions actually have to be treated as macrowhite (bright) and macro-black (dark) in correspondence with the idea of different stages in the evolution of basic colour terms.

References Berlin, Brent and Paul Kay 1969 Basic Color Terms. Their Universality and Evolution. Berkeley and Los Angeles: University of California Press. Crepajac, Ljiljana 1967 Die idg. Wortsippe *gh/ghel- "hell, glänzend" im Griechischen und Slawischen. Zeitschrift für vergleichende Sprachforschung 81.3-4: 181-196. Cruse, D. Alan 1986 Lexical Semantics. Cambridge: Cambridge University Press. Györi, Gabor 1996 Historical aspects of categorization. In Eugene H. Casad (ed.), Cognitive Linguistics in the Redwoods. Berlin and New York: Mouton de Gruyter. 175206.

Hegedus, Iron 1997 Principles for paleolinguistic reconstruction. In Roger Blench and Matthew Spriggs (eds.), Archaeology and Language I. Theoretical and Methodological Orientations. London and New York: Routledge. 65-73.

Is everything black and white in conceptual oppositions?

73

Hock, Hans Henrich 1986 Principles of Historical Linguistics. Berlin and New York: Mouton de Gruyter. Johnson, Mark 1987 The Body in the Mind: The Bodily Basis of Meaning, Reason and Imagination. Chicago and London: University of Chicago Press. Kastovsky, Dieter 1988 Structural semantics or prototype semantics? The evidence of word formation. In Werner Hüllen and Rainer Schulze (eds.), Understanding the Lexicon. Meaning, Sense and World Knowledge in Lexical Semantics, 190-203. (Linguistische Arbeiten 210) Tübingen: Max Niemeyer Verlag. Key, Mary Ritchie 1979 Opposites - Once again. Interfaces. Linguistics and Psychoanalysis Newsletter. (University of Delaware, Newark, Delaware) No. 11: 2-5. Lakoff, George 1987 Women, Fire and Dangerous Things. What Categories Reveal about the Mind. Chicago and London: The University of Chicago Press. 1990 The Invariance Hypothesis: Is abstract reason based on image-schemas? Cognitive Linguistics 1: 39-74. Langacker, Ronald W. 1987 Foundations of Cognitive Grammar, Volume 1: Theoretical Prerequisites. Stanford, CA.: Stanford University Press. Lehmann, Winfred P. 1993 Theoretical Bases of Indo-European Linguistics. London and New York: Routledge. Mettinger, Arthur 1988 Pay Caesar what is due to Caesar...: Semantic features vindicated. In Werner Hüllen and Rainer Schulze (eds.), Understanding the Lexicon. Meaning, Sense and World Knowledge in Lexical Semantics, 148-156. (Linguistische Arbeiten 210) Tübingen: Max Niemeyer Verlag. Palmer, Leonard R. 1978 Descriptive and Comparative Linguistics. London: Faber & Faber. (First ed. 1972) Pokorny, Julius 1959-69 Indogermanisches etymologisches Wörterbuch. Bern: Francke. R6dei, Käroly (ed.) 1986-91 Uralisches etymologisches Wörterbuch. Band 1-3. Budapest: Akademiai Kiado. Seiler, Thomas B. 1985 Sind Begriffe Aggregate von Komponenten oder idiosynkratische Minitheorien? In Thomas B. Seiler and Wolfgang Wannenmacher (eds.), Begriffs- und Wortbedeutungsentwicklung, 105-131. Berlin: Springer. Sweetser, Eve 1990 From Etymology to Pragmatics: Metaphorical and Cultural Aspects of Semantic Structure. Cambridge: Cambridge University Press.

74

Gabor Györi and Iren Hegedüs

Szemere"nyi, Oswald 1972 Comparative linguistics. In Thomas A. Sebeok (ed.), Current Trends in Linguistics, 119-195. Vol. 9/1. The Hague: Mouton. 1990 Einführung in die vergleichende Sprachwissenschaft. (4th ed.) Dannstadt: Wissenschaftliche Buchgesellschaft. Watkins, Calvert (ed.) 1985 The American Heritage Dictionary of Indo-European Roots. Boston: Houghton Mifflin. York, Michael 1993 Toward a Proto-Indo-European vocabulary of the sacred. Word 44.2: 235254.

Categorization and analogical change: The case of athematic Isg -m in the Slavic languages Laura A. Janda

1.

Introduction

Every inflected language carries some seemingly useless morphological baggage bequeathed to it by previous generations in the form of irregular inflectional affixes. These relics of what were once productive and systematic linguistic distinctions often fade into oblivion, although they can persist for hundreds of years. However, it is not the case that all defunct morphology must wither and die. Languages are capable of recycling nearly extinct morphemes and using them either to restore distinctions that have eroded or even to build entirely new systems of distinctions. There are many examples of this type of analogical extension, yet it has received so little attention that it lacks a standard term. This type of analogical extension may be called "upstream" (i.e., against the "expected" flow of morphology which tends to eliminate irregular forms in favor of regular forms) or "exaptative", a term borrowed by Lass (1990) from biologist Stephen J. Gould. I will illustrate exaptative analogical extension with a chapter from the history of the Slavic languages: the spread of Isg -m from a handful of athematic verbs to much or all of the verbal lexicon in the West and South Slavic languages. "Exaptation" appears to have both a variety of sources and a variety of explanations. There are several different kinds of marginalized morphemes that can serve as source material for this process because a grammatical morpheme can become marginal in one of several ways: a) when the paradigm that uses the morpheme becomes limited to only a few lexical items (as in the case of the athematic Isg -m), b) when the paradigm is lost (e.g., the former «-stems of Slavic), and c) when the grammatical category that the morpheme represents is lost (e.g., the loss of the dual as a grammatical category in most of Slavic). The categorial status of marginal morphemes is thus compromised and they are available for recategorization. Furthermore, the pattern of recategorization is quite logical. Morphemes in situation a), where the paradigm is limited to a few lexical items, still represent active linguistic categories in live

76

Laura A. Janda

paradigms, and they will remain stable or decline unless phonological (and morphological) events "conspire" to change their status within their category. This is the case of the spread of athematic -m. In addition to the framework of cognitive linguistics, involving linguistic categorization based on prototypes and radial networks, I invoke Andersen's (1973) model of language change via abduction and subsequent deduction. The synchronic state of a language presents learners of the younger generation with data from which they abduce their grammar by making decisions about which linguistic categories are relevant and which characteristics are prototypical of those categories. There is inevitably some mismatch between the grammars abduced by successive generations, and this is the ultimate source of language change. Thus, in the analysis that follows, I will present the synchronic state of each relevant language just prior to the spread of Isg -m and discuss the abductive innovations that promoted -m from the status of a marginal morpheme to that of prototypicality, leading to the deduction that -m should be more widespread in the verbal lexicon.

2.

Preconditions common to West and South Slavic

Indo-European had two sets of verbal endings, one for thematic and another for athematic verbs. Athematic verbs further distinguished between primary endings for the present active indicative and secondary endings for all other categories. For the Isg, the primary ending was -m-i and the secondary was the same as the primary, except that it lacked the "hie et nunc" particle -i and was merely -m. Proto-Slavic inherited the primary ending -mi for its five athematic verbs and developed an ending consisting of -am for the Isg of the thematic verbs. The presence of the -i particle in the athematic forms vs. its absence in the thematic forms had serious consequences for the syllabification and subsequent fate of Isg forms. The law of rising sonority (in effect until the disintegration of Slavic unity) occasioned the simplification of consonant clusters and the monophthongization of diphthongs as described below: athematic verbs: Isg forms: CV.C (root) 4- m-i > simplification of consonant clusters > CV + .m-i > fall of jers > CV + m thematic verbs in *o/e:

Categorization and analogical change Isg forms:

77

CV.C (root) + om > monophthongization of diphthongs > CVC + Q

The period symbolizes the position of a syllable boundary. Where m was in syllable initial position it was retained, but in syllable final position m was monophthongized with the theme vowel ο to yield the nasal vowel Q. The close of the common Slavic period is associated with the elimination of short ι, Μ (called "the fall of the jers") which removed the final -i particle from the athematic verbs, thus opposing two very dissimilar expressions of Isg. Note that monophthongization not only removed the Isg marker -m but also altered the role of the theme vowel since it was no longer a separate entity in this form. A similar monophthongization took place in the thematic third plural forms with the same consequences for the theme vowel: thematic verbs in *o/e: 3pl forms: CV.C 4- ο + n.tu > monophthongization of diphthongs > CVC + qt > fall of jers > CVC + qt The development in thematic verbs in *i, though not identical, also effectively stripped the theme vowel of its independent role in non-past morphology: thematic verbs in *i: Isg forms: CV.C + ι + o-m > i becomes non-syllabic when not syllable peak > CV.Cj + om > deiotation1 and monophthongization of diphthongs > CVC + q 3pl forms: CV.C -I- / + nt > monophthongization of diphthongs > CVC + qt > fall of jers > CVC + qt Thus by the time of late Common Slavic the role of the theme vowel had been compromised and we can speak of an independent theme vowel only in the 2sg, 3sg, Ipl, and 2pl forms. The deiotation of the stem final consonant in Isg forms of thematic verbs in *i created a morphophonemic alternation that opposed the Isg to all other forms (e.g., Isg in pros- vs. all other forms in pros- for the verb 'beg'), and the palatalization of stem final velars before the theme vowel e in the 2sg2pl forms created another morphophonemic alternation opposing the Isg and 3pl to all other forms (e.g., Isg and 3pl in pek- vs. all other forms in pector the verb 'bake'). In most languages of West and South Slavic (i.e., everywhere but Polish) it was abduced that desinences, i.e. inflectional endings, not morphophonemic alternations, were the prototypical means for distinguishing person and number in the non-past. The spread of -m in these

78

Laura A. Janda

languages is associated with the elimination of morphophonemic alternations. The fate of the Isg markers cannot be discussed outside of the context of the verbal system to which they belonged, so it is necessary to outline the paradigms of this system here. At about the time of the fall of the jers, sequences of -VjV- contracted to yield -V-, and this included sequences in which they was the stem final consonant and the second vowel was the theme vowel, thus removing the original theme vowel and producing new conjugational paradigms. Contraction was not carried out uniformly in this territory, but certain generalizations can be made: - contraction was avoided wherever it would disfigure the shape of the root (e.g., forms of a verb like 'bark' in laj-ole- were never contracted) - contraction did not take place in the Isg and 3pl - Bulgarian and Macedonian contracted only sequences of -aje- in verbal forms, never -eje- Slovene and Serbo-Croatian merged the verbs of 'becoming' in original -eje- with the class II verbs (in -i-) prior to the time of the spread of -m and therefore do not show reflexes of contraction for these verbs - Polish avoided contracting any sequence of -aje- or -eje- that was identified with a suffix denoting imperfectivization or the deadjectival derivation of a verb meaning 'become X' (e.g., poznajesz 'you meet', leniwiejesz 'you become lazy'). Contraction created new long vowels. Length will be marked in the paradigms below only where it occurs in all languages except Macedonian and Bulgarian (which lost this distinction prior to the spread of -m). After contraction had taken place, the following system of non-past conjugations was present in West and South Slavic: Class I:

-o-l-eIsg -q 2sg -e-s 3sg -e

Ipl 2pl 3pl

-e-mV -e-te -q(t)

This class continues the Indo-European type of conjugation with the ablauting theme vowel -o-l-e-. The final vowel in the Ipl is a recent innovation, and varies according to language: in Polish it is y; in Czech, Slovak, Macedonian, and Bulgarian it is e; and in Slovene and Serbo-Croatian it is ο. The final t is present in the 3pl forms only in those languages that do not have phonemic length to distinguish the 3pl from the Isg.

Categorization and analogical change

Class II: i Isg 2sg 3sg

-q -i-s -i

Ipl 2pl 3pl

79

-i-mV -i-te

In this class the theme vowel is / and this vowel is long in all of West and South Slavic but Polish, Macedonian, and Bulgarian. Class III: -ej + -o-l-eIsg -ejq 2sg -S-s 3sg -£

Ipl 2pl 3pl

-S-mV -£-/e -ejQ(t)

This is a new conjugation class created by the contraction of original sequences of -eje- in the 2sg-2pl forms to produce a new theme vowel of -e-. This class does not exist in Macedonian and Bulgarian, and is limited to only three non-derived verbs in Polish. Verbs in these three languages with uncontracted -eje- are kept in class I. Class IV: -aj + Isg 2sg 3sg

-o-l-e-ajq -ä-s -ä

Ipl 2pl 3pl

-ä-mV -ä-te

Class IV is likewise a new conjugation class arising from the contraction of -aje- to produce the new theme vowel -a-. The Polish verbs in original -ajethat did not contract remained in class I. Class V: athematic2 vim, vSs, v&, vimV, vl(s)te, vedj(t) jSm, jgs, ji, jlm V, j$(s)te, jedj(t) (i)mäm, (i)mäs, (i)mä, (i)mamV, (i)mate, (i)maJQ(t) dam, das, da, dämV, dä(s)te, dadg(t)

'know' 'eat' 'have' 'give'

The parallels between the athematic verbs 'know' and 'eat' and class III (both types have the vowel preceding the person-number desinences) on the one hand and between the athematic verbs 'have' and 'give' and class IV (both types have the vowel ä preceding the person-number desinences) on the other hand were striking and were instrumental in the spread of Isg -m. Up until this point, distinctive athematic desinences had been eliminated in order to bring the athematic verbs into conformity with the thematic paradigms. Specific events in this trend include:

80

Laura A. Janda

- changing the 2sg athematic desinence from -si to -s - loss of final -t in the 3sg and (except in Macedonian and Bulgarian) 3pl, paralleling its loss in the thematic verbs - loss of -s- in the 2pl desinence, e.g., veste > v$te 'you know' (except in Slovene) The spread of the one remaining distinctive athematic desinence to thematic verbs was to mark a dramatic departure from this trend. One more piece of pertinent background information: the morphophoneme m was present in the following pronominal roots and verbal desinences, all of which signaled first person: - all of the oblique forms of the Isg pronoun - the nominative forms of the Ipl pronoun: my (everywhere but most Bulgarian and Macedonian dialects) - the Ipl non-past marker, also used in Ipl imperative formation: -mV - the Ipl marker for both the imperfect and the aorist in -m Thus the choice of m as a marker of first person had plenty of support in the grammar inherited by early West and South Slavic. To summarize, the following are preconditions shared by all languages that ultimately spread Isg -m: - 1 sg non-past forms end in -q or -m - m is used elsewhere in pronominal and verbal morphology to mark first person - independent theme vowels are present in 2sg-2pl only - morphophonemic stem alternations distinguish Isg and 3pl from other forms in class I, and Isg from other forms in class II - contraction creates new theme vowels -e- (most languages) and -a- (all languages) which parallel the predesinential vowels in the athematic verbs - up until this point the trend has been to level out distinctive athematic desinences in favor of thematic desinences. Although all of West and South Slavic extended the use of -m, this process occurred several centuries after the disintegration of Slavic. Sound changes that took place in the intervening time played a decisive role in determining what new categorial associations could be abduced in the grammars of individual languages. Both the ultimate extent of Isg -m and the route of cognitive leaps that brought about its extension varied greatly from language to language. There are three Slavic languages in which -m was universal-

Categorization and analogical change

81

ized: Slovak, Slovene, and Macedonian. This does not mean, however, that the course or even the outcome of this process was the same in these three languages, as we shall see. Although the discussion will focus primarily on these languages, parallels will be drawn to the remaining Slavic languages. 3.

The spread of -m in Slovak (and Czech, Polish, and SerboCroatian)

Here are the relevant phonological changes peculiar to Old Slovak (cf. Krajcovic 1988): g>h I, e > e, e q > ü; Q > u; > ; > ä/P_, elsewhere > a e > ie a>ialC'_ depalatalization of palatalized consonants

(circa fall of jers) (circa fall of jers) (10th c) (13th c) (15th c) (15th c)

Here is the inventory of Old Slovak verb classes at the time -m spread: Class I is divided into two major subclasses according to the length of the reflex of the original theme vowel in the 2sg-2pl forms, which will be referred to as Vj (since the theme vowel is no longer an independent morpheme throughout verbal paradigms). Class la: (sg) -u, -es, -e, (pi) -erne, -ete, -u3 This class includes verbs in - *-no/-ne following a C vydychnu, vydychnes, (vydychnu >) vydychnu 'exhale' - *-jo-/-je- following a C ukazu, ukazes, (ukazu >) ukazu 'show' The shortenings observed in the 3pl forms of these verbs are caused by the rhythmic law4 of Slovak and will be discussed below. Class Ib: (sg) -u, -ies, -ie, (pi) -ieme, -iete, -ü This class includes - verbs in *-no/-ne following a V or syllabic r or / minu, minies, minü 'pass' - non-suffixed stems nesu, nesies, nesü 'carry' peku (> pecu), pecies, pekü ( > pecü) 'bake'

82

Laura A. Janda tru, tries, trü 'rub' beru, beries, berü 'take' - verbs with stem in -Vjkryju, kryjies ( > kryjes), kryjü5 'cover' - verbs in -ova-l-ujkupuju, kupujies (> kupujes), kupujü 'buy'

The shortenings observed in the last two subtypes will also be discussed below. Class II: (sg) -u, -is, -f, (pi) -ime, -Ite, -ä This class includes - verbs in - prosu, prosis, pros'ä (> prosia) 'beg' Class III: (sg) -eju, -ies, -ie, (pi) -ieme, -iete, -ejü This class includes - non-derived verbs smeju, smies, smejü 'dare' - derived verbs, especially with the meaning 'becoming' beleju, belies, belejü 'become white' Class IV: (sg) -aju, -äs, -ä, (pl) -äme, -äte, -ajü This class includes - verbs in original *-ajehl'adaju, hl'adäs, hl'adajü 'look' - derived imperfectives in *-ajeklanaju sä, klanas sa(> klanias sä), klanajü sa 'bow' Class V: viem, vies, vie, vieme, viete, ved'ä (> vedia) 'know' jem, jes, je, jeme, jete, jed'ä (> jedict) 'eat' (i)mäm, (i)mäs, (i)mä, (i)mame, (i)mate, (i)maju 'have' dam, das, da, dame, date, dad'ä (> dadia, later > dajü) 'give' V j of 'eat' is long -e- (in all but the 3pl) but is diphthongized to -ie- and then the first part of the diphthong coalesces with the contiguous glide -j-, thus yieldingy'em, jes, je, jeme, jete, jedia. I will suggest that fourteenth century learners of Slovak faced with this verb system made the following abduction about their verbal morphology. The non-past forms of all verbs could be segmented as: stem - Vj - (C) - (V2)

Categorization and analogical change

83

where the stem ends in a consonant and has the minimal shape C(C), Vj is any vowel or diphthong, and the last two segments are optional. This abduction carries with it the assumption that the predesinential vowel of the athematic verbs is V 1? not a root vowel. All forms of the non-past paradigm have the same schematic shape, regardless of conjugation class, except the Isg and 3pl which are not uniform, giving: Isg

3pl

stem-Vj stem - Y! - C - V2 stem - Vj - C stem - Vj stem - Y! - C - V2

classes I, II classes III, IV class V classes I, II classes III, IV, V

This fact singles out the Isg as the form with the most variants. If we compare the paradigms of class Ib and the contracted classes (classes III and IV) with the athematic class (class V) in terms of both the schematic shapes of the forms and the quality and quantity of VI, we find that the following generalization can be made: - V j is long -ie- or -a- in all forms but the 3pl, with the exception of the - Isg of classes III, IV, and Ib The logical consequence of observing this generalization is that the Isg forms of classes III, IV, and Ib are exceptional, whereas the Isg forms in the athematic class are prototypical. Further, this generalization imposes a hierarchy of prototypicality judgements on the entire verbal system: generalization holds * generalization holds, but

class V the Isg is exceptional classes Ib, III, IV

*

generalization holds, but * generalization holds, but

the Isg is exceptional quality of V j varies class II the Isg is exceptional quality of V j varies quantity of V j varies class la

84

Laura A. Janda

The formerly marginalized -m has received a great boost in status, and is now licensed to spread to all classes, beginning with those with a Vj of -ie- or -a-, as we see in: Phase 1 Abductions: -m is a marker of first person - there is a category of verbs with (underlying) long Vj = or ie - class V is prototypical of this category Deduction: - classes Ib, III, IV use the prototypical athematic Isg in - V] + m Phase 1 (plus subsequent sound changes) deduces the modern Slovak forms: miniem Ί pass' neuem Ί carry' peciem Ί bake' triem Ί rub' beriem Ί take' (kryjiem >) kryjem Ί cover (kupujiem >) kupujem smiem Ί dare' Ί buy' beliem 'Hook' Ί become white' hi' adorn (klan m sa >) klaniam sa Ί bow' based on the prototypical pattern of viem Ί know' (jem >)jem Ί eat' ((i)mam >) mam Ί have' dam Ί give' The next step in the process of spreading Isg -m is the removal of the constraint on the quality of Vj. Now any verb with a long V^ is subject to the influence of the prototypical athematic class. This brings class II with its V j -ϊ- on board for: Phase 2 Abductions: - m is a marker of first person - there is a category of verbs with (underlying) long Vj - class V is prototypical of this category Deduction: - classes Ib, II, III, IV use the prototypical athematic Isg in V] + m and Phase 2 deduces prosim Ί beg.' The lifting of the quantity restriction on Vj permits the last group of verbs (class la) to come under the influence of class V. Three factors gradually eroded the distinctive value of length in V^ - the rhythmic law of Slovak - the presence of morphologized exceptions to the rhythmic law - the shortening of diphthongs after 7'

Categorization and analogical change

85

According to the rhythmic law of Slovak, wherever there are two consecutive long syllables (i.e., syllables containing a long vocalic segment or a diphthong), the second is shortened, as in class la: (vydychnü >) vydychnu 'they exhale,' (ukaiu >) ukazu 'they show.' A parallel example from class IV is (pomahäs >) pomahas 'you help.' Whenever the final syllable in the root of a verb was long, Vj was shortened, yielding forms schematically parallel to those in class la: Class la (with long stem syllable) other classes (with long stem syllable)

CVC - Vj CVC - V] -

The recognition of certain short vocalic segments as underlyingly long was compromised by morphologized exceptions to the rhythmic law and shortenings of long syllables attributable not to the rhythmic law, but to the absorption of the first part of diphthongs beginning in / by a preceding j. Particularly important here was the shortening of Vj ie > e in position after -j-, as in kupujes > kupujies > kupujes 'you buy' and je s > jies > jes 'you eat.' The underlying quantity of V j was no longer reliably recoverable, and it ceased to play a determining role in the spread of -m, as we see in: Phase 3 Abductions: - m is a marker of first person - length of V j is not distinctive in classifying verbs - there is one category of verbs (rhythmic law is not automatic; there are other shortenings; underlying length cannot be determined) Deduction: - all verbs use the prototypical athematic Isg in - Vj + m, where V] can be long or short which deduces: vydychnem exhale', ukazem show.' The spread of -m in Czech follows a similar path, except that length did not develop in class Ib and Czech lacked the rhythmic law and other shortenings of VI, so there was no change in the role of quantity in this process. Athematic Isg -m is thus spread only to classes II, III, and IV in Czech, by a series of abductive steps roughly equivalent to phases 1 and 2 (modified by the exception of class Ib). The spread of -m is even more circumscribed in Polish, where class II did not have a long theme vowel and the process is limited to the equivalent of phase 1, producing Isg forms in -m only in class III and class IV verbs (and recall that these classes were smaller because Polish did not carry out contraction as fully). Polish did not spread -m specifically

86

Laura A. Janda

to verbal classes (I and II) that had morphophonemic stem alternations; instead Polish enhanced these alternations to distinguish the Isg and 3pl from all other forms in both classes of verbs: class I bior$ Ί take', bierzesz 'y°u take', biorq. 'they take', where we see r alternating with rz and ο with e class II prosz$ Ί beg', prosisz 'you beg', (*prosig >)proszq 'they beg', where we see the alternation of sz with s. Early prosodic changes produced in thirteenth century Serbo-Croatian a system with long Vj in all conjugation classes, so quantity was never an issue. The spread of -m followed a two-phase development, parallel to the first two phases in Slovak, with the exception that original e did not merge with e in time for this change (or in some dialects ever), which means that -m spread to classes III and IV (constrained only by quality of Vj) in phase 1 and to classes I and II in phase 2. Isg -m now marks all verbs in Serbo-Croatian, with only two exceptions, the modal mogu Ί can' and the modal/auxiliary hocu Ί want/will'.

4. The spread of -m in Slovene The strategy in Slovene was to regeneralize the theme vowel, and the Isg and 3pl forms were perceived as: a) failing to conform to the otherwise universal non-past shape of stem + theme vowel + desinence, b) lacking unitary expression, c) lacking sufficient distinctiveness (since for many verbs the Isg and 3pl had the same form). The spread of -m helped to solve all of these problems, and was accompanied by innovations in the 3pl forms. Phonemic vowel length did not play any role in the spread of -m in Slovene. Note that Slovene has preserved the dual forms in its conjugations, as indicated in the inventory of classes below. The following relevant changes are particular to Old Slovene, and all of them took place in the 9th-10th cc. (cf. Lencek 1982): q, Q > ο; ξ, ξ > e l,e>e

Here is an inventory of verbal classes in Slovene just prior to the spread of Isg -m: Class I:

(sg) -o, -es, -e, (du) -eva, -eta, -eta, (pi) -emo, -ete, -o This class includes:

Categorization and analogical change

87

- verbs in *-no-l-neizdihno, izdihnes, izdihno 'exhale' mino, mines, mino 'pass' - non-suffixed stems neso, neses, neso 'carry' peko/peco, peces, peko/peco 'bake' tro, tres, tro 'rub' hero, beres, hero 'take' - verbs in *-jo-/-je- following a C pokazo, pokazes, pokazo 'show' - verbs with stem in -Vj krijo, krijes, krijo 'cover' - verbs in -ova-l-ujkupujo, kupujes, kupujo 'buy' Class II: (sg) -o, -is, -i, (du) -/να, -/ία, -/ία, (pi) -imo, -ite, -e This class includes - verbs in -/proso, prosis, prose 'beg' - the verbs of 'becoming' (beleju, belis, beleju) > belo, belis, bele 'become white' Class III: (sg) -ejo, -es, -e, (du) -eva, -eta, -eta, (pi) -emo, -ete, -ejo This class includes non-derived verbs smejo, smes, smejo 'dare' Class IV: (sg) -ajo, -as, -a, (du) -ανά, -ata, -ata, (pl) -amo, -ate, -ajo This class includes - verbs in original *-ajVgledajo, gledas, gledajo 'look' - all derived imperfectives (except those in -ova-l-uj-} klanjajo se, klanjas se, klanjajo se 'bow' Class V: vent, ves, ve, veva, vesta, vesta, vemo, veste, vede 'know' jem, jes, je, jeva, jesta, jesta, jemo, jeste, jede 'eat' imam, imas, ima, imava, imata, imata, imamo, imate, imajo 'have' dam, das, da, dava, dasta. dasta, damo, daste, dade 'give'

88

Laura A. Janda

If we make the generalization that the fourteenth century verb system presents a unified non-past conjugation, based on a pattern of stem + theme vowel + desinence, where there is one desinence per form in the paradigm, and each desinence consists of either a zero or a consonant that may or may not be followed by a vowel, we observe the following exceptions: - four verbs (from original class V) have -s- in their 2du, 3du, and 2pl desinences - the theme vowel is missing in the Isg and 3pl forms of classes I and II - the following are not uniform Isg, which has the exponents -o classes I & II -theme vowel-^'-o classes III & IV -theme vowel-w class V 3pl, which has the exponents -o class I -e class II -theme vowel-y-o classes III & IV & imajo -theme vowel-d-e dade, vede, jede In classes I, III, and IV, the Isg and 3pl forms are identical. The spread of -m combined with the generalization of the theme vowel restored distinctiveness and produced verbal paradigms unique in the Slavic world. Here is a formulation of the generalization and the hierarchichal organization it imposes on the verbal system: Generalization: the prototypical shape of a non-past verbal form is stem + -e-l-a- (theme vowel) + desinence generalization holds class V * * generalization holds - everywhere but Isg classes I, III, IV * generalization holds - everywhere but Isg - quality of VI varies class II

generalization holds - everywhere but 3pl classes I, II

Categorization and analogical change

89

Phase 1 Abductions: - the vowels -a- and -e- in class V are theme vowels equivalent to the theme vowels in classes I, III, and IV - the Isg forms in class V are prototypical because they conform to the pattern of stem + theme vowel + desinence Deduction: - Isg forms of verbs in classes I, III, and IV have the shape stem + theme vowel

+m Phase 1 (plus subsequent sound changes) deduces izdihnem Ί exhale' minem Ί pass' Ί carry' Ί bake' nesem pezem 'Irub' berem Ί take' tarem krijem Ί cover' pokazem Ί show' Ί dare' kupujem Ί buy' smem gledam Ί look' klanjam sa Ί bow' based on the pattern of vem Ί know' jem Ί eat' imam Ί have' dam Ί give.' Phase 2 Abductions: - all theme vowels have equal status, regardless of quality - the Isg forms in class V are prototypical because they conform to the pattern of stem + theme vowel + desinence Deduction: - Isg forms of all verbs have the shape stem + theme vowel +m Phase 2 deduces: prosim Ί beg', belim Ί become white.' In Slovene the underlying assumption was that all non-past verb forms have a theme vowel (rather than a stem enlargement, as is likely the case with Vj in West Slavic and Serbo-Croatian). This assertion is made on the basis of the changes that took place in phase 3 in Slovene, a phase that probably overlapped with phases 1 and 2. Phase 3 Abduction: - the 3pl form has the prototypical shape stem + theme vowel Deduction: - verbs in all classes have 3pl forms conforming to this shape Phase 3 deduces forms such as: izdihnejo 'they exhale', nesejo 'they carry', kupujejo 'they buy', prosijo 'they beg', vejo 'they kncw\jejo 'they eat', 'they give'.

90

Laura A. Janda

The existence of forms like izdihnejo 'they exhale' and kupujejo 'they buy' confirm that the extension of theme vowels was the driving force behind these changes. If Slovene were concerned only with preserving distinctiveness, there would have been no need to modify the original forms in *izdihno and *kupujo. Thus although the spread of -m in Slovene might appear at first glance to be part of an areal phenomenon shared with Slovak, Serbo-Croatian, and Macedonian, this result was obtained via a different set of abductions than in neighboring languages.

5.

The spread of -m in Macedonian (and Bulgarian)

In Macedonian, in contradistinction to Slovak and particularly to Slovene, the role of the theme vowel was not altered in the course of the spread of -m. Instead, sound changes created a situation in which all Isg non-past verb forms ended either in -a or in -am. It was abduced that -a and -am were allomorphs of a single morpheme, and the more distinctive of the two, namely -am, was selected as the prototypical allomoiph and extended accordingly. Here are the relevant changes that yielded the sound system of Macedonian at the time of the spread of Isg -m (cf. Koneski 1983, Janda and Friedman 1994): loss of distinctive length y>i ?>e;9>a (j)e > ja in initial position e > e elsewhere x>v/_C

(prior to 13th c.) (12-13thcc.) (12-13thcc.) (14-15th cc. or later, dep. on dialect) (16thc.)

Here is an inventory of Macedonian verb classes at the time of the spread of-m: Class I:

(sg) -a, -es, -e, (pi) -erne, -ete, -at This class includes - verbs in *-no/-ne izdivna, izdivnes, izdivnat 'exhale' mina, mines, minat 'pass' - non-suffixed stems nesa, neses, nesat 'carry' (peka >) peca, peces, (pekat >)pecat 'bake' tra, tres, trat (> trija, trijes, trijat) 'rub'

Categorization and analogical change

91

bera, beres, berat 'take' - verbs in *-jo-/-je- following a C ukaza, ukazes, ukazat 'show' - verbs with stem in -Vj krija, krijes, krijat 'cover' - verbs in *-ejesmeja, smejes, smejat 'dare' beleja, belejes, belejat 'become white' Class II: (sg) -a, -is, -i, (pi) -im(e), -ite, -at This class includes: - verbs in -iprosa, prosis, prosat 'beg' Class III: This class is not distinct in Macedonian, for contraction of original -eje- clusters did not take place, so all of its members remain in class I. Class IV: (sg) -α/α, -as, -a, (pi) -ame, -ate, -ajat This class includes - verbs in original *-aje- and derived imperfectives gledaja, gledas, gledajat 'look' klanjaja se, klanjas se, klanjajat se 'bow' - suffixed verbs in original *-ovakupuvaja, kupuvas, kupuvajat 'buy'

Class V: [*vedeti 'know' was lost early in the history of Macedonian] jam, jas, ja, jame, j te, jadetljadat 'eat' imam, imas, ima, imame, imate, imajat 'have' dam, das, da, dame, date, dadetldadat 'give' Note that the loss of *vedeti 'know,' coupled with the change je- > ja- in the verb for 'eat' meant that the predesinential vowel in all the athematic verbs was -a-. To summarize, the pattern of non-past verb forms in Macedonian at the time was the following:

92

Laura A. Janda

Isg

stem stem stem stem stem stem -

2sg-2pl 3pl

a classes I & II aj - a class IV am imam, dam, jam theme V - desinence all classes at classes I & II, dadat, jadat aj - at class IV, imajat

The verb *imati 'have' was identified with class IV, and -am was generalized to class IV verbs. At this point, all Isg forms were in either -a or -am, and this occasioned the spread of -am to all verbs. Subsequent to the generalization of -am, constraints on the shape of non-past verbal stems led to adjustments in the conjugation of the remaining two athematic verbs and non-syllabic verb types in class I. The generalization that could be drawn by language learners at this time and the hierarchies it entailed were: Generalization: Isg forms have the shape syllabic stem + am * * valid for most verbs valid for class V * * not valid for jam, dam, final segment -m and non-syllabic stems in is missing in class I classes I, II, III, IV The changes motivated by this generalization can be summarized as follows: Phase 1 Abductions: - there are three classes of verbs, based on the theme vowel present in the 2sg, 3sg, Ipl, and 2pl forms - ima 'have' is a prototypical class IV verb Deductions: - all verbs have a theme vowel of -e-, -i-, or -a- in the 2sg, 3sg, Ipl, and 2pl forms - class IV verbs have Isg in -am (i.e., their paradigm follows that of ima 'have') Phase 1 deduces gledam Ί look', kupuvam Ί buy', klanjam se Ί bow'. Phase 2 Abduction: - the Isg desinence is -a(m), and -am is prototypical Deduction: - all verbs have the Isg desinence -am Phase 2 (plus subsequent loss of j in sequences of -eja-) deduces:

Categorization and analogical change

izdivnam

nesam trijam ukazam smeam prosam

Ί exhale' Ί carry' 'Irub' Ί show' Ί dare' Ί beg'

minam pecam be ram

krijam beleam

93

Ί pass' Ί bake' Ί take' Ί cover' Ί become white'

Phase 3 Abductions: - non-past verb forms are minimally disyllabic, following the pattern: Isg (C n )VC n -am 2sg-2pl (Cn)VCn - theme V - desinence 3pl (Cn)VCn - (aj)at i.e., the verbal stem is syllabic - the 3pl forms dadat, jadat belong to class I (parallel to idat, vedat, kradat, kladat, etc.) and have the stems dad-,jadDeductions: - unprefixed resonant stems are all given syllabic shapes, e.g., tr- > trij- in trijam Ί rub' - the verbs 'give' and 'eat' have the following class I conjugation: dadam jadam dades dade dademe dadete dadat

jades jade jademe jadete jadat

Bulgarian underwent a phase 1 nearly identical to that of Macedonian, but because 9 > 9 rather than a, the spread of -m stopped with the class IV verbs, since there was no motivation for the type of abduction that led to phase 2 in Macedonian. 6.

Why -m did not spread in East Slavic

Throughout West and South Slavic, the contractions that produced classes III and especially IV were crucial in facilitating the abduction that these classes could be categorized with the athematic verbs. East Slavic never experienced contraction and this precluded the spread of -m. The trend was instead to regularize athematic verbs as members of class I (cf. Russian imet' 'have' and vedat' 'know' which are regular, but dot' 'give' and est' 'eat' which have retained the athematic pattern). Morphophonemic alternations involving

94

Laura A. Janda

stress and stem final consonants play an important role in verbal conjugations in East Slavic, contrary to what we observe in West and South Slavic.

7.

Conclusion

Analogy is a reinterpretation of categorial status. Items that are marginal in the morphology of one generation can be abduced to be prototypical by successive generations when sound changes have altered the perceptible relations among items in a category. This paper has traced the various routes by which an originally non-prototypical marker of Isg was reinterpreted as prototypical for some or all of the verbal lexicon of the West and South Slavic languages.

Acknowledgments This research was aided by a grant from the Joint Committee on Eastern Europe of the American Council of Learned Societies and the Social Science Research Council. I would like to thank Victor Friedman and Craig Melchert for comments and corrections. All errors in the present text can be attributed only to me.

Notes 1. The term "deiotation" is used here to refer to both the loss ofj and the palatalization of the preceding consonant. The symbol C is used to indicate the result of this process, which is either a palatal or a labial followed by palatal /. 2. The fifth athematic verb, byti 'be' has been eliminated from discussion in this paper because this verb is highly irregular and has an idiosyncratic development in each Slavic language. 3. The conjugations of verbs list segments following the stem of forms in the traditional order of Isg, 2sg, 3sg, Ipl, 2pl, 3pl, although in the case of Slovene Idu, 2du, 3du have been inserted between the singular and the plural. The examples show only the Isg, 2sg, 3pl, since the shape (excluding desinence) of the 3sg, Ipl, and 2pl (and in the case of Slovene, Idu, 2du, 3du) is the same as that of the 2sg. 4. Examples of morphological environments that do not show the expected shortenings of the rhythmic law (the first two cited by Kenstowicz and Rubach 1987): denominal (possessive) adjectives like kohuti 'rooster's', neuter collective nouns \ikeprutie 'twigs', deverbal nouns like prianie 'wish', class II 3pl verb forms

Categorization and analogical change

95

likepacia 'they please', verbs of 'becoming' like zmudriet' 'become wise', idiosyncratic lexemes like upiet' 'lament.' 5. Stanislav (1967) and Krajcovic (1988) do not posit inherited length in the theme vowel for the subtypes in -_/-, although they do posit length for the remaining subtypes. There are compelling reasons to include these two types here. One is structural: all other verbs with an original theme vowel -e- not preceded by an original consonant cluster belong to this class, (thus we seepecies (< *pek- e-si) 'you bake,' but not places (< *plak-je-si) 'you cry' in this group). The other reason is phonological: the initial portion of the diphthong -ie- was absorbed by the preceding -j- before any attestations are available, so it is impossible to prove that these two subtypes did not have a long Vj.

References Andersen, Henning 1973 Abductive and deductive change. Language 49: 765-793. Janda, Laura A. and Victor Friedman 1994 About the ja- in makedonskiot jazik: the fate of initial *6- and *e- in Macedonian. Journal of Slavic Linguistics 2: 282-286. Kenstowicz, Michael and Jerzy Rubach 1987 The phonology of syllabic nuclei in Slovak. Language 63: 463-497. Koneski, Blaze 1983 A historical phonology of the Macedonian language. Heidelberg: Winter. Krajöoviö, Rudolf 1988 Vyvin slovenskehojazyka a dialektologia [The evolution of the Slavic languae and dialectology]. Bratislava: Slovenske pedagogicko nakladatel'stvo. Lass, Roger 1990 How to do things with junk: exaptation in language evolution. Linguistics 26: 79-102. Lencek, Rado 1982 The Structure and History of the Slovene Language. Columbus: Slavica. Stanislav, Jan 1967 Dejiny slovenskeho jazyka. II· Tvaroslovie [The history of the Slovak language. II. Morphology]. Bratislava: Slovenskaa Akademia Vied.

Contrast and Schemas: Antonymous adjectives Arthur Mettinger

1.

Introduction: A new departure?

In this paper I want to briefly sketch the (image-)schematic properties of antonymous adjectives in English. The search for these characteristics is to be understood as one step in the investigation of 'contrastivity' as the cognitive substratum on which contrast phenomena in language and cognition are based. The ultimate aim of the research programme is to arrive at a characterisation of 'contrastivity' as a prototypically or/and schematically structured 'superconcept' conditioning the construal and interpretation of semantic contrast in a particular language.1 The investigation of semantic contrast has been given sufficient attention over the past decades with the study of 'antonymy' (in the broad sense of the term) as one type of sense- or meaning-relation most often performed within a basically structuralist framework against the background of the semantic systems of different natural languages, resulting either in typologies such as Lyons's (1977) subclassification of 'contrast' into antonymy (restricted to gradable opposites), complementarity, converseness etc., or in the classification of individual pairs of opposites elicited from a corpus as operating along semantic dimensions of various types and kinds (Mettinger 1994).2 These results are perfectly in accordance with a linguistic theory founded in the conviction that [...] because the meaning of a linguistic sign is determined by its position in the linguistic structures in which it takes part, the proper subject matter of semantics is [...] the semantic structure of the language that demarcates the meanings of individual words with regard to each other. (Geeraerts 1992: 259)

Since the advent of Cognitive Linguistics (CL), however, there has been a marked shift of emphasis from the description of langue-properties to the characterisation of language as "one of the basic cognitive tools of man" (Geeraerts 1988: 656f.). This new perspective necessitates a new approach to

98

Arthur Mettinger

phenomena that have already been investigated within non-CL frameworks (contrast being one of them) with the explicit aim of gaining new insights into the cognitive underpinnings of language and their translation into and codification in individual languages. Pionieering work on anatomy within a basically CL framework has been done by Cruse (1992) and Cruse and Togia (1995).

2.

Relevant axiomatic and methodological features of CL

2.1.

General

Above all, CL emphasises the experientially embodied nature of language. It is assumed that "lexical meaning is not [...] an autonomous phenomenon, but is [...] inextricably bound up with the individual, cultural, social, historical experience of the language user" (Geeraerts 1992: 266). Methodologically, this leads to giving up the structuralist language-immanent approach with the sharp distinction between linguistic and encyclopaedic knowledge and to replacing it by a basically hermeneutic approach that "consists of an interpretative attempt to recover the original experience behind the expressions" (Geeraerts 1992: 267). As it is claimed that "language is an integral part of human cognition" (Langacker 1987: 12), the task of linguistic semantics is conceived of as attempting the structural (not structuralist) analysis and explicit description of abstract entities like thoughts and concepts: Meaning is equated with conceptualization. [...] Because conceptualization resides in cognitive processing, our ultimate aim must be to characterize the types of cognitive events whose occurrence constitutes a given mental experience. [...] It is claimed [...] that semantic structures (which I call "predications") are characterized relative to "cognitive domains", where a domain can be any sort of conceptualization: a perceptual experience, a concept, a conceptual complex, an elaborate knowledge system, etc. The semantic description of an expression therefore takes for its starting point an integrated conception of arbitrary complexity and possibly encyclopedic scope. The basic observation supporting this position is that certain conceptions presuppose others for their characterization (Langacker 199la: 2f.)

Even a cursory glance at the literature soon reveals that CL exhibits "great diversity with respect to the analytical tools used, the points emphasized, and the perspectives adopted" (Rudzka-Ostyn 1993: 1). Still, I would like to

Contrast and Schemas: Antonymous adjectives

99

claim that there are four major 'strands' in CL research focusing on different aspects of linguistically manifest conceptual phenomena: the detection of image Schemas, the reinterpretation of 'grammar' and the exploration of scene construal, the disclosure of metaphorical processes and, finally, the discussion of categorization phenomena (including problems of category structure, conceptual hierarchies, and variation phenomena). The importance of the identification of these research areas for the characterization of 'contrastivity' lies first in their heuristic function as stepping stones in the process of investigation and secondly in their function as parameters for the cognitively oriented description, i.e., any theory of 'contrastivity' will gain in credibility the more it succeeds in simultaneously covering those four areas. In this contribution the focus will be on image Schemas; as we will look at adjectives, however, a brief discussion of their status in CL is called for.

2.2.

'Base', 'profile', and word-classes

In Langacker's (1987, 1991a, b) cognitive grammar framework every 'predication' (i.e., the meaning of a linguistic expression) imposes a 'profile' on a 'base' where "the base of a predication is its domain" (Langacker 1991a: 5) and the profile "is a substructure elevated to a special level of prominence within the base, namely that substructure which the expression 'designates'" (Langacker 1991a: 5). Moreover, a broad distinction is made between basic classes of predications depending on the nature of their profile: a noun is regarded as a symbolic structure that designates a 'thing', "where 'thing' is a technical term defined as a 'region in some domain'" (Langacker 199la: 20), whereas verbs, adverbs, adjectives and prepositions are regarded as 'relational' expressions profiling "the 'interconnections' among conceived entities" (Langacker 1991a: 20). They are thus conceptually dependent in that "one cannot conceptualize interconnections without also conceptualizing the entities that they interconnect" (Langacker 1987: 215). Relational expressions must therefore always be characterised in terms of two participants, viz. the Trajector' (Tr) and the 'Landmark' (Lm). "The Tr is the more salient participant in the relation. The less salient participant constitutes the Landmark [...], which serves as a kind of reference point for the specification of the Tr." (Taylor 1992: 10).3 As a consequence of the assumption of these two basic classes of predications we are bound to expect different manifestations of 'contrastivity': con-

100

Arthur Mettinger

trastive relations between concepts coded as nouns4 are expected to be of a different nature than those coded as relational expressions such as adjectives, verbs, or prepositions. The first step in the process of investigating 'contrastivity' thus involves the search for very general conceptual properties characterising opposites against the background of their respective word-classes. In this contribution the focus is on adjectival opposites, whose characterisation, as I will be claiming in the following, is best performed in terms of 'image Schemas'.

2.3.

Image Schemas

The postulation of and search for these mental 'images' is based on the assumption that "in order for us to have meaningful, connected experiences that we can comprehend and reason about, there must be pattern and order to our actions, perceptions, and conceptions" (Johnson 1987: 29). A schema in the sense of 'image schema' is thus defined in the following way: Λ schema is a recurrent pattern, shape, and regularity in, or of, these ongoing ordering activities. These patterns emerge as meaningful structures for us chiefly at the level of our bodily movements through space, our manipulation of objects, and our perceptual interactions. It is important to recognize the dynamic character of image schemata. I conceive of them as structures for organizing our experience and comprehension [...] They are dynamic in two important respects. (1) Schemata are structures of an activity by which we organize our experience in ways that we can comprehend. They are primary means by which we construct or constitute order and are not mere passive receptacles into which experience is poured. (2) Unlike templates, schemata are flexible in that they can take on any number of specific instantiations in varying contexts [...] (Johnson 1987: 29f.) Image Schemas, it seems, were first introduced in connection with the analysis of spatial prepositions: in contradistinction to cognitive categories (such as OBJECT5, ORGANISM, and ACTION categories) which exhibit a rich categorial structure and are best conceived of in terms of the prototype hypothesis of categorisation (with attribute lists, typicality gradients and gestalts), locative relations such as UP—DOWN, IN—OUT, FRONT—BACK, LEFT—RIGHT, OVER—UNDER etc. reflect "basic experiences" (Ungerer and Schmid 1996: 160) and are best treated in terms of image Schemas6.

Contrast and Schemas: Antonymous adjectives

101

This type of characterisation seems most suited to the cognitive nature of adjectives, too: Adjectives serve the purpose of linguistically coding property concepts which are usually predicated of members of OBJECT/ORGANISM categories (coded as nouns).7 When an adjective collocates with a noun, then, either one of the many properties the nominally coded concept possesses8 is highlighted, or the nominally coded concept becomes equipped with an 'extra' property it had not possessed as one of its 'regular' properties.9 The former case can be illustrated by collocations such as old man, tall man, wise man etc. (where MAN is characterized against the domains of AGE, SIZE, MENTAL ABILITIES, respectively), the latter case by dead man (which is a characterization with regard to the LIFE CYCLE domain).

3. Adjectival opposites and Schemas 3.1. Two relevant image Schemas In the following I am going to claim that the distinction between gradable (scalar)10 and non-gradable (digital) adjectival opposites11 is due to two different image Schemas, viz. the SCALE schema and the CONTAINER schema. I will proceed in the following way: in accordance with the CL assumptions outlined in section 2.2 the schematic properties of scalar adjectives will be sketched following Taylor (1992). In this paper Taylor discusses the three distinct senses of the English adjective old (OLD as in old box, OLD' as in old friend, and OLD" as in old regime) "as more specific instantiations, or 'elaborations', of a single, more abstract, or 'schematic' sense" (Taylor 1992: 20) vis-ä-vis the background of scalar adjectives in general. As Taylor was not interested in studying old with regard to its antonyms (young and new) I must go beyond his description in my search for Schemas capable of serving as a frame of reference for characterising the nature of adjectival opposites. Such Schemas must be able to account for pairs of adjectives, and they should ideally be able to offer a unified cognitive description of different kinds of contrast (both in the conceptual world of speakers as manifested in their linguistic expressions and in the various senses such linguistic expressions can have).

102

Arthur Mettinger

3.2.

Scalar opposites

3.2.1.

The semantic structure of scalar adjectives according to Taylor (1992)

Taylor assumes the following schematic representation of a scalar adjective:

Dom

[A scalar adjective] designates a relation between its Tr (a thing) and its Lm, a region on a scale. [...] The large box encloses the relevant cognitive domain of the adjective, e.g., 'length', 'height', 'speed', etc., symbolized by [Dom], while the horizontal line represents the dimension itself. The heavy portions of the diagram represent the profiled elements of semantic structure. The heavy circle represents the Tr of the adjective, the small box surrounding the circle symbolizing the cognitive domain of the Tr. The Tr is located within a profiled region of the dimension, represented by the heavy portion of the horizontal line. The profiled region of the dimension lies in excess of some norm, represented by the region surrounding the point n. The norm is represented as a region so as to capture the 'fuzziness' of scalar adjectives. There is, namely, no precise point on the dimension of, e.g., tallness, which clearly cuts off the class of 'tall' entities from the class of 'not tall' entities. (Taylor 1992: lOf.)

The semantic structure of the adjective old in attributive position as in old man, old box, etc., can be represented as an elaboration of the scalar adjective schema, cf.:

'···..,

Lm

r\ \^J

^-^ ^J

]

"

Tr

Θ

Dom

Essentially, old denotes that its Tr has been in existence for a period of time in excess of some norm. In Schema 2, the passage of time is represented

Contrast and Schemas: Antonymous adjectives

103

by the horizontal time-line, at the bottom of the diagram, while the double appearance of the Tr entity represents the continued existence of the Tr over the intervening period of time. The broken line linking the two instantiations of the Tr symbolizes the perceived identity of the Tr at the different times. R denotes the reference time, i.e. the time at which the Tr is characterized with respect to its oldness. (Taylor 1992: 11)

The combination of old with a nominal predication is possible if this nominal predication is able to elaborate the schematic Tr of the adjective. Moreover, the norm n associated with the adjectival predication receives its precise value only in this composite structure in that it depends on the kind of entity that serves as the adjective's Tr. (Taylor 1992: 12f.) Such schematic representations are meant to capture degrees of specificity: a schema (such as Schema 1) is more abstract and less specified than its elaborations (Schema 2) or instantiations, i.e., an instantiation is always fully compatible with the specifications of the schema it instantiates, but is characterised in finer detail. 3.2.2.

Schema versus image schema

It is vital, though, to be aware of the fact that 'schema' and 'schematic' as used in this section and 'image schema' as outlined in 2.3. are of a different nature: Taylor's use of 'schema' follows Langacker's understanding of 'schema' as "[a] semantic, phonological, or symbolic structure that, relative to another representation of the same entity, is characterized with lesser specificity and detail. A 'coarse-grained' (as opposed to a fine-grained) representation." (Langacker 1991b: 552) Schemas are abstractions capturing what is common to the members of a category (semantic, phonological, or symbolic), the process of abstraction taking usage events, i.e. "actual utterances in the full richness of their phonetic detail and contextual understanding" (Langacker 1991b: 2) as its starting point. Langacker's Schemas (as I see it) thus evolve as the result of bottom-up cognitive procedures and are primarily grounded in linguistic experience. Image Schemas, on the other hand, are abstractions grounded in non-linguistic experience: An image schema is a recurring, dynamic pattern of our perceptual interactions and motor programs that gives coherence and structure to our experience. The VERTICALITY schema, for instance, emerges from our tendency to

104

Arthur Mettinger

employ an UP-DOWN orientation in picking out meaningful structures of our experience. We grasp this structure of verticality repeatedly in thousands of perceptions and activities we experience every day, such as perceiving a tree, our felt sense of standing upright, the activity of climbing stairs, forming a mental image of a flagpole, measuring our children's heights, and experiencing the level of water rising in the bathtub. The VERTICALITY schema is the abstract structure of these VERTICALITY experiences, images, and perceptions [emphasis A.M.]. (Johnson 1987: XIV) For the characterisation of adjectival opposites both types of schema are called for: the SCALE schema (for scalar opposites) and the CONTAINER schema (for non-scalar opposites) account for the experientially grounded conceptual properties shared by all these adjectival opposites; on the other hand, as members of the category ADJECTIVE they share the properties discussed in 2.2. and can be represented in terms of the Schemas suggested by Langacker (and Taylor, as far as scalar adjectives are concerned). In the following I will, therefore, map the relevant (general) image schematic properties on the word-class specific schema. 3.2.3.

Scalar opposites and the SCALE schema

According to Mark Johnson "the SCALE schema is basic to both the quantitative and qualitative aspects of our experience" (Johnson 1987: 122) and exhibits the following properties: i) the SCALE schema has a more or less fixed directionality. [...] Normally, the further along the scale one moves, the greater the amount or intensity. [...] ii) Scales have a cumulative character of a special sort. If you are collecting money and have $15, then you also have $10. [...] iii) SCALES are typically given a normative character; [...] Having more or less of something may either be good or bad, desirable or undesirable. Having more heat in the winter can be desirable, while having more heat in the summer might be awful. In either case, however, norms are mapped on to the scale. iv) [Scales] can be either closed or open. [...] At any rate, SC ALARITY does seem to permeate the whole of human experience, even where no precise quantitative measurement is possible. Consequently, this experientially basic, value-laden structure of our grasp of both concrete and abstract entities is one of the most pervasive image-schematic structures in our understanding. The image schema which emerges in our

Contrast and Schemas: Antonymous adjectives

105

experience of concrete, physical entities is figuratively extended to cover abstract entities of every sort [...] (Johnson 1987: 122f.) If this image schema is accepted as basic to scalar adjectival opposites, we can characterise a pair of such opposites as in Schema SC: Lm

Don*

Schema SC/a

Don*

Schema SC/b

Schema SC/a captures the basic schematic properties of gradable adjectives denoting LESS of a scaled property (e.g., young, short, low, little etc. in combination with an appropriate nominal Trajector), Schema SC/b supplies the properties of the pair member denoting MORE (e.g., old, long, high, much etc. in such a combination). In both cases the Trajector is located within a profiled region on the scale that is situated either "below" or "above" some norm (n). In view of Johnson's characterisation of the SCALE schema it seems more plausible to adopt a vertical representation rather than a horizontal one as has been done by Taylor. Moreover, it must be noted that in each of the following schema representations the "a" and "b" versions of the respective figures highlight the difference between the members of an antonymous pair of adjectives; they are meant to be identical except for the profiled region on the scale and the position of the Trajector. The domains are identical as well (hence: Donij).12 The schema suggested above must, however, be elaborated, as it does not contain any information as to the directionality and boundedness of the SCALE, i.e., the profiled regions on the scale where a Trajector might be located must be specified in greater detail; this elaboration of the schema is necessary to account for three types of antonymous adjectives: a) Adjective pairs like long - short, high - low, much - little etc. presuppose a SCALE that is bounded at the lower end. Short, low, little profile the lower

106

Arthur Metünger

part of the scale, and the Trajector is then positioned anywhere in the region between the norm and the zero-point on the scale: the shorter, lower, less something is thought to be, the closer the Trajector is moved towards the zero value (Schema SC/a/1 below). On the other hand, long, high, much etc. profile the part of the scale above the norm, and in case of intensification the Trajector is positioned higher up on the scale which is unbounded at its upper end (Schema SC/b/1):

Dom

Schema SC/a/1

Dom;

Schema SC/b/1

b) In the case of beautiful - ugly, good - evil etc. the Trajector is again positioned either "above" or "below" the norm region. The SCALE is bidirectional and unbounded: it extends into the "positive" direction as well as into the "negative" one. When an expression such as ugly or evil is used the Trajector is always situated in the "negative" part of the scale; the more negative it is conceived of, the farther the Trajector is removed from the norm which represents some kind of "neutral state" (Schema SC/a/2). Intensification of beautiful, good etc. involves moving the Trajector towards infinity along the profiled "positive" part of the scale (Schema SC/b/2):

Lm

Dom,

Schema SC/a/2

Doni)

Schema SC/b/2

Contrast and Schemas: Antonymous adjectives

107

c) The third variant concerns cases like safe - dangerous, dry - wet, innocent - guilty, traditionally termed 'gradable complementaries', which are characterised by the fact that very often the scaled property is evaluated negatively: In the case of an UNDESIRABLE property [...] the most important question for the language user is whether the property is present or absent. The desirable status is zero value of the property, and this is signalled by one of the terms of the opposition; any positive value represents an unsatisfactory state, and this is signalled by the other term of the opposition. (Cruse 1980: 23) The scale is thus bounded at one end by the norm. In expressions such as a dangerous road, a wet towel, a guilty person the Trajector is placed in the "negative" region of the scale, and the more dangerous, wetter or guiltier the Trajector is thought to be the farther it is removed from the norm (Schema SC/a/3). The other member of the pair, on the other hand, coincides with the norm and thus is attributed the zero value (Schema SC/b/3):

Dom; Schema SC/a/3

3.3.

Schema SC/b/3

Non-scalar opposites and the CONTAINER-schema

I have claimed above that non-scalar adjectival opposites of the type right wrong, true -false, male -female, civilian - military, etc. should be regarded as linguistic manifestations of the CONTAINER schema - "a schema consisting of a boundary distinguishing an interior from an exterior. The CONTAINER schema defines the most basic distinction between IN and OUT" (Lakoff 1987: 271). Johnson (1987: 39) claims that this schema is also responsible for our understanding of negation: together with the metaphorical understanding of propositions as locations we assume that to hold a proposition is understood in terms of being located in a definite bounded space

108

Arthur Mettinger

(the space defined by the proposition), whereas to hold the negation of that proposition is understood as being located outside that bounded space. And yet, although non-scalar adjectival opposites are semantically analysable in terms of negation (each member of the pair can be characterised as the negation of the other member in such a way that 'male'='not female' and 'female'='not male'), the assumption of the CONTAINER-schema as defined above cannot do justice to the conceptual properties of non-scalar opposites as it does not specify the space outside the bounded region, i.e., the space that is occupied by the negation of a proposition.13 Therefore the schema has to be modified so as to account for the observation that certain domains are conceptualised as "bi-compartmental" and that the Trajector can be placed in either of the two compartments: if it is positioned in one compartment of the respective domain (Schema CONT/a) it is automatically excluded from the other compartment of the same domain and vice versa (Schema CONT/b):

Lm

Dom; Schema CONT/a

Lm

Donii Schema CONT/b

This schematic representation is basically compatible with Lyons's (1977: 27 If.) statement that "[upgradable opposites, when they are employed as predicative expressions, divide the universe-of-discourse [...] into two complementary subsets. It follows from this, not only that the predication of either one of the pair implies the predication of the negation of the other, but also that the predication of the negation of either implies the predication of the other".14 Moreover, this schema can also accommodate cases like almost true (the Trajector is moved into the direct vicinity of the Lm in Schema CONT/a) or half true (the Trajector is conceptualised as having in part entered the Lm) as well as slightly false (the Trajector is located partly inside and partly outside the Lm in Schema CONT/b).

Contrast and Schemas: Antonymous adjectives 4.

109

Conclusion

In this contribution I have tried to show how the notion 'schema' in both the Johnsonian and Langackerian understanding could be successfully implemented in the study of contrast in language. I am well aware of the fact that the picture I have presented is incomplete and lacking in both scope and detail. And yet I think it shows a major step in the right direction with regard to the study of CONTRASTIVITY: just as a polysemous lexical item can be shown to have a number of senses related network-wise via extensions of one sense to another on the one hand and as elaborations of a more abstract, schematic sense on the other hand, CONTRASTIVITY can be seen as a 'poly-schematic' network composed of a number of individual image Schemas. In order to substantiate this claim a lot more information will be needed, of course: On the schematic level, an inventory of the image Schemas responsible for nominal, verbal, and prepositional opposites must be established, on the basis of which research into the way these image Schemas arrange themselves with regard to CONTRASTIVITY should be made possible, the ultimate goal being a characterisation of the internal conceptual structure of CONTRASTIVITY. On the lexical level we will have to show how opposites instantiate these Schemas (including all sorts of metaphorisation processes) and how our interpretation of lexical items as opposites is guided by these Schemas. If this can be accomplished, the CL approach will have proved its superiority over all extant accounts of lexical opposition.

Acknowledgments This paper forms part of the research project "Contrastivity: cognition and lexical representation" subsidised by the Austrian Fonds zur Förderung der wissenschaftlichen Forschung (Erwin Schrödinger-Auslandsstipendium J0769-SPR). I wish to express my gratitude to Dirk Geeraerts, Stef Grondelaers, Maarten Lemmens, Leon de Stadier and all my colleagues at the Linguistics Department of the Catholic University of Leuven as well as to an anonymous reviewer, who have all helped me shape my ideas on the problems discussed here. All errors and misconceptions are, of course, my own.

110

Arthur Mettinger

Notes 1. Ultimately, the characterisation of 'contrastivity' as a possibly universal concept is desirable; at present, however, we are far from attaining this goal. 2. There is no point in discussing the structuralist approach in this paper. The reader is referred to Mettinger (1994: 4-7; 13-17) for an overview of studies in the structuralist framework and to Mettinger (1994: 94-147) for the corpus-based analysis. 3. "This use of trajector and landmark has been generalized in cognitive linguistics, so trajector stands for the [...] most prominent element in any relational structure (and is therefore indicated by very bold lines [in the following diagrammatic representations (A.M.)]), whereas landmark refers to the other salient entity in a relation [indicated by medium bold lines]." (Ungerer and Schmidt 1996: 161) The landmark is thus less prominent than the trajector, but it is salient compared with 'the background', which is now firmly established as part of the system and called the domain. This domain provides the context for an expression. 4. This statement is valid for primary, i.e. non-derived nouns only. 5. Concepts and categories will be typographically represented by SMALL CAPS, domains by ITALICISED SMALL CAPS, image Schemas by CAPITALS. 6. "The idea is that by experiencing for example many instances of things-overthings we have acquired some sort of cognitive pattern or schema of the -OVERrelationship which we can apply to other instances of this locative relation. The schema which has thus developed is obviously less concrete than the rich prototype categories of objects and organisms [...]. However, [...] an image schema is not just an abstract semantic principle, but should be understood as a mental picture which is more elementary than both concrete categories and abstract principles. This is something to keep in mind when we [...] look at pictorial representations used for locative image Schemas in cognitive linguistics." (Ungerer and Schmid 1996: 160) 7. Cf. Wierzbicka (1988: 463) on the semantic basis of the distinction between nouns and adjectives. 8. In technical terms, we would say that nominally coded concepts are usually characterisable relative to a multitude of cognitive domains (a so-called domain matrix; for details see Langacker 1987: 147ff.). 9. This is possible only if the domain of the adjectival concept does not clash with any of the domains relative to which the nominal concept is characterised. Within non-CL frameworks this phenomenon has been known as 'lexical solidarities' or 'selection restrictions' (cf. Likpa 1992: 160f.) 10. In a structuralist account of oppositeness of meaning 'gradability' is regarded as the syntactic manifestation of 'scalarity' which is taken to be a semantic property of the respective adjective. For a more detailed discussion see Mettinger (1994: 85).

Contrast and Schemas: Antonymous adjectives

111

11. This corresponds to the traditional distinction between antonymy and complementarity as established, for example, in Lyons (1977). 12. As far as the status of the schema representations is concerned I side with Van Hoek (1995: 315) who states that "[t]he use of iconic notation is intended only as a heuristic and does not entail any theoretical claim. In particular, it should not be construed as a claim that semantic representations consist solely of visual images." 13. Cf. in this context Steinthal's (1890: 36If.) statement that "[...] negation [...] and opposition are not the same thing: an opposite is as positive as the thing whose opposite it is [...]" [transl. A.M.] 14. Although Lyons discusses the predicative use of ungradable adjectival opposites I will take his statement to be basically valid for the attributive use of these opposites as well.

References Cruse, David Alan 1980 Antonyms and gradable complementaries. In: Kastovsky (ed.), 14-25. 1992 Antonomy revisited: Some thoughts on the relationship between words and concepts. In Lehrer and Feder Kittay (eds.), 289-306. Cruse, David Alan and Pagona Togia 1995 Towards a cognitive model of antonymy. Lexicology 1.1, 113-141. Geeraerts, Dirk 1988 Cognitive grammar and the history of lexical semantics. In: Rudzka-Ostyn, B. (ed.), 647-677. 1992 The return of hermeneutics to lexical semantics. In: Pütz, M. (ed.), 257-282. 1993a. Cognitive semantics and the history of philosophical epistemology. In: R. Geiger and B. Rudzka-Ostyn (eds.), 53-79. 1993b Vagueness's puzzles, polysemy's vagaries. Cognitive Linguistics 4.3, 223272. Geiger, Richard A. and Brygida Rudzka-Ostyn (eds.) 1993 Conceptualizations and Mental Processing in Language. Berlin and New York: Mouton de Gruyter. Johnson, Mark 1987 The Body in the Mind. The Bodily Basis of Meaning, Imagination, and Reason. Chicago and London: The University of Chicago Press. Kastovsky, Dieter (ed.) 1980 Perspektiven der lexikalischen Semantik: Beiträge zum Wuppertaler Semantikkolloquium vom2-3. Dezember 1977. Bonn: Bouvier. Lakoff, George 1987 Women, Fire, and Dangerous Things. What Categories Reveal about the Mind. Chicago and London: The University of Chicago Press.

112

A rthur Mettinger

Langacker, Ronald W. 1987 Foundations of Cognitive Grammar. Vol. 1: Theoretical Prerequisites. Stanford: Stanford University Press. 1988a An overview of cognitive grammar. In: B. Rudzka-Ostyn (ed.), 3-48. 1988b A view of linguistic semantics. In: B. Rudzka-Ostyn (ed.), 49-125. 199la Concept, Image, and Symbol. The Cognitive Basis of Grammar. Berlin and New York: Mouton de Gruyter. 1991b Foundations of Cognitive Grammar. Vol. II: Descriptive Application. Stanford: Stanford University Press. Lehrer, Adrienne and Eva Feder Kittay (eds.) 1992 Frames, Fields, and Contrasts. New Essays in Semantic and Lexical Organization. Hillsdale, NJ: Lawrence Erlbaum. Lipka, Leonhard 1992 An Outline of English Lexicology. Lexical Structure, Word Semantics, and Word-Formation. 2nd ed. Tübingen: Niemeyer. Lyons, John 1977 Semantics. Volume I. Cambridge: Cambridge University Press. Mettinger, Arthur 1994 Aspects of Semantic Opposition in English. Oxford: Clarendon Press. Pütz, Martin (ed.) 1992 Thirty Years of Linguistic Evolution. Studies in Honour of Rene Dirven on the Occasion of His 60th Birthday. Amsterdam and Philadelphia: John Benjamins. Radden, Günther 1992 The cognitive approach to natural language. In: M. Pütz (ed.), 513-541. Rudzka-Ostyn, Brygida (ed.) 1988 Topics in Cognitive Linguistics. Amsterdam and Philadelphia: John Benjamins. Steinthal, H. 1890 (1863) Geschichte der Sprachwissenschaft bei den Griechen und Römern mit besonderer Rücksicht auf die Logik. 2nd ed. Berlin: Dümmler. Taylor, John R. 1992 Old problems: Adjectives in cognitive grammar. Cognitive Linguistics 3.1, 1-35. 1995 Linguistic Categorization. Prototypes in Linguistic Theory. 2nd ed. Oxford: Clarendon Press. Ungerer, Friedrich and Hans-Jörg Schmid 1996 An Introduction to Cognitive Linguistics. London and New York: Longman. Van Hoek, K. 1995 Conceptual reference points: A cognitive grammar account of pronominal anaphora constraints. Language 71.2, 310-340. Wierzbicka, Anna 1988 The semantics of grammar. Amsterdam and Philadelphia: John Benjamins.

Figurative giving John Newman 1.

Introduction

Verbs which have the sense of 'give (something to someone by hand)', here designated as GIVE verbs, often have other senses which are understood as related to, and based on, this sense. The parallels in the extensions associated with GIVE across languages reveal common patterns and in what follows I explore the major patterns of extensions across languages and suggest motivations for these semantic extensions. I will refer to the sense of 'transfer possession of some tangible thing to someone (by hand)' as literal GIVE and any other sense of GIVE morphemes as figurative GIVE (whether or not speakers consider such senses to be poetic or novel). While this study is based on a cross-linguistic comparison of GIVE morphemes, much can be gained from an intensive study of the use of GIVE morphemes within a language too. One should mention in this connection the very thorough account of the figurative extensions of Brazilian Portuguese GIVE in Salomao (1990). In seeking explanations for these figurative extensions, one must acknowledge that some expressions to do with figurative giving may be culturally bound and a full explanation of such expressions would require a reference to culture-specific institutions, customs etc. In Jacaltec, for example, an expression which is literally 'give big stomach for me!' functions as an idiom with the meaning 'be patient with me!' (Craig 1977: 396). To understand how a word with the meaning of 'stomach' comes to be used in this idiom, one needs to appreciate the existence of a number of metaphors in Jacaltec based on 'stomach', which is associated with various emotions and mental states (Craig 1977: 274). So, for example, Ί would like to sing/play' is expressed as 'My stomach would like to sing/play etc' (Craig 1977: 315); Ί am sad that he is going' is expressed as 'My stomach is burning that he should go' (Craig 1977: 256). The 'give big stomach' construction is part of a larger, coherent system of metaphor which needs to be considered in a full account of the semantics of the construction. The emphasis here, however, lies not in analyzing aspects of GIVE extensions which are bound to specific cultural traditions or systems of metaphor. Rather this paper considers extensions which recur across languages or which show strong similarities across Ian-

114

John Newman

guages. This includes the comparison of grammatical, or semi-grammatical, functions of GIVE morphemes. The motivations I seek rely on an appreciation of the experiential reality of the giving act in its essentials and the semantic complexity inherent in the meaning of literal GIVE.

2.

Literal GIVE

It is useful to clarify the nature of literal GIVE before moving on to figurative exensions. Two aspects of literal GIVE in particular need to be mentioned: the basicness of the giving act in an experiential sense and the conceptual complexity of the meaning of literal GIVE. Both of these aspects are relevant to a full understanding of the figurative extensions of GIVE. In Newman (1996) I detail reasons for thinking of the giving act as a relatively basic act within the realm of ordinary, everyday human interaction. Firstly, general considerations about human social interaction lead one to see the giving of objects from one person to another as a commonly recurring feature of such interaction. The passing of objects from one person to another, whether or not the objects are intended as gifts, seems a fairly salient feature of interaction between humans. One might compare it with, say, the act of looking at someone or speaking with someone as some of the basic interpersonal acts. Secondly, language acquisition studies point to expressions like Give me X as being among the first expressions understood by children. A pertinent study is Chapman (1981), who investigated comprehension of twelve children in the age range of 13-15 months. Among other things, Chapman investigated the extent to which the children responded appropriately to requests requiring various actions on the part of the children. To some verbal requests of the form Kiss X, only one child responded appropriately. To requests of the form Give me X, on the other hand, all twelve children responded appropriately. The Give me X construction was shown, therefore, to be one which was comprehended relatively early in terms of first language acquisition. Thirdly, basic vocabularies typically include a verb meaning literal GIVE. A "basic vocabulary" is hardly a well-defined term in the linguistics literature, but I use it here to refer to a vocabulary of a very limited size which nevertheless functions to convey all the meanings which vocabularies of much larger size would normally convey. Sometimes, a basic vocabulary occurs naturally, as a kind of sub-language within a community. A taboo variety of the Australian language Dyirbal, no longer actively used,

Figurative giving

115

functioned like this, as described in Dixon (1971: 441). It contained an absolute minimum of verbs with each taboo verb corresponding to many verbs in the normal variety of the language. Ogden's Basic English, as described in Ogden (1968), would be an example of an artificially constructed basic vocabulary. It, too, has a minimum number of verbs (less than twenty) which are put to use in characterizing all the meanings which can be conveyed. Significantly, a verb meaning GIVE appears in each of these basic vocabularies. It seems attractive to humans, in other words, to conceptualize other acts/events in terms of giving, rather than reduce giving to other concepts. Fourthly, it is possible for the concept of giving to be realized without any overt wordform associated with it. The Papuan language Amele, described in Roberts (1987), is one such case. Unique amongst verb predicates, the sense of GIVE is conveyed as though there were a verb plus its affixes, except that the verb is a zero morpheme. Compare the 'show' sentence in (la) with the GIVE sentence in (Ib). Notice the similarities in the overall structures of the two sentences, except that there is no overt GIVE verb in (Ib): (1)

a.

b.

Jo eu ihac-i-od-ig-en. house that show-Pred-2P!Obj-lSgSubj-Future Ί will show that house to you (plural).' (Amele, Roberts 1987: 69) Naus Dege ho 0-ut-en. Naus Dege pig give-3SgObj-3SgSubj.Remote Past 'Naus gave Dege the pig.' (Amele, Roberts 1987: 34)

The fact that there is no overt form associated with GIVE suggests a relatively basic predicate, on a par with a verb 'to be' in some languages which may also appear as a null form. Finally, the richness of GIVE morphemes as sources for figurative extensions, as documented here, is another manifestation of this same conceptual bias. The basicness referred to in the preceding paragraph relates to the centrality of the giving act in terms of human interaction. When one reflects on all the components which make up the act of giving, however, it is clear that there is a relatively complex configuration of entities, interacting with each other. Since the various figurative extensions of GIVE are motivated by different components inherent in the semantics of literal GIVE, it is useful to unravel some of the complexity of literal GIVE, isolating some of the different layers of meaning of the predicate. I find it convenient to distinguish a number of different parameters for describing the meaning of GIVE. These different

116

John Newman

parameters help to break down the whole of the meaning into smaller parts. Each choice of parameter results in a particular view of the predicate which may be equated with the notion of a domain in Langacker (1987). The domains which I have found relevant to an understanding of figurative GIVE (as well as constructions involving literal GIVE) are as follows: (i) The spatio-temporal domain. This is the domain defined entirely by space-time co-ordinates, by means of which one may track the location and movement of tangible objects through space and time. Even limiting ourselves to this one domain, literal GIVE involves quite complex relationships. Two human participants are positioned in a relatively proximate and stable spatial configuration with respect to each other, while some object moves from one human to the other. A full description of this domain would require some discussion of the role of the hands of the human participants. These constitute what Langacker (1987) calls the active zone in the semantic characterization of a predicate, comparable to the role of the eye-lids in the characterization of the meaning of blink. Identifying the hands used in the giving act as the active zone of a GIVE predicate is consistent with the widespread use of the morpheme meaning 'hand' as the root for GIVE type predicates, as in English hand something to someone, as well as its numerous figurative extensions hand it to someone 'praise someone', hand over the meeting to someone etc. (ii) The control domain. The spatio-temporal domain does not allow us to say anything directly about the different ways in which the human participants referred to in the meaning of GIVE relate to the thing being passed. The control domain describes a predicate in terms of who exercises control over what at each point. With respect to GIVE, one needs to characterize the giver as the person initially exercising control over the object being passed and the recipient as the person who exercises this control at the end of the GIVE act. The object passes from one sphere of control to another. The sense of "control" which I appeal to here is the broad sense of having access to an object and having some physical control over the object, rather than any specific sense of, say, legal control of the object, that is, possession in the legal sense. The literal sense of GIVE which I am considering is neutral with respect to whether there is a transference of possession in a strictly legal sense. (iii) The force-dynamics domain. Another domain which one may distinguish is the force-dynamics domain. Within this domain one describes the ways in which the entities involved affect one another. In the case of GIVE, there is one human participant, the giver, who initiates the act and another

Figurative giving

117

human participant, the recipient, with whom the act of giving comes to an end. One may speak of the giver as the "energy source" and the recipient as the "energy sink" as a way of expressing this asymmetry in the involvement of the two human participants (cf. Langacker 1991: 292-293). There is a natural path to the energy flow inherent in GIVE: from the giver to the recipient. (iv) The domain of human interest. This domain characterizes the various ways in which the participants in an act or event are advantaged or disadvantaged. Recognizing a special domain of human interest in the characterization of literal GIVE is justified by the experiential reality. The extent to which humans are advantaged or disadvantaged is a significant and interesting aspect of the act of giving. In particular, the recipient is someone who typically stands to gain something by having been given some thing. Normally, we are given things in order that we might go on to do something with what we have been given. We can easily give things which harm the recipients or which are useless to the recipients, but there is a positive bias in the way we generally interact with others and this is reflected in the way we give things. While admitting the positive bias can be easily overridden, I believe nevertheless that, all else being equal, one would normally expect giving to be for the benefit of the recipient. Taken together, these domains constitute the meaning of literal GIVE which I take to be the source for the other meanings of GIVE morphemes documented in Section 3. The priority I assign to literal GIVE is consistent with the special status of literal GIVE as a relatively basic predicate, as discussed above. It is also consistent with what I know about the historical development of the meanings associated with GIVE morphemes. Where one can see a semantic development historically, then it is the meanings which I am calling figurative which develop from literal GIVE, rather than the other way around. Literal GIVE thus has a priority in my account which is based on my understanding of properties of GIVE predicates which I, as a linguist, have access to. It is not based on any psycholinguistic experiments which might establish a psychological reality to the priority of one meaning over another (testing subjects' response times to different meanings, for example, and identifying the meaning which is most quickly identified as "basic"). Informally questioning subjects about the "basic" or "most obvious" meaning of a GIVE morpheme can elicit widely variant responses. So, for example, one (educated) native speaker of English whom I interviewed regarded the meaning of give exemplified in The problem gave me a headache as a more

118

John Newman

typical use of give than the meaning exemplified in / gave her the book. The priority I assign to literal GIVE is linguistically informed and expedient as a way of explicating the full range of senses one finds, though the pyschological reality of such priority remains as yet unvalidated.

3.

Figurative extensions

It is convenient to distinguish subgroupings of the figurative extensions of literal GIVE across languages, as a way of appreciating some of the most important factors underlying such extensions. I have labelled these subgroupings interpersonal communication, emergence/manifestation, permission/causation, schematic interaction, recipient/benefactive, completedness/ perfectivity, and movement away. Taken together, they represent a broad spectrum of meanings which may seem sometimes to be quite unconnected to the sense of literal GIVE and even contradictory. For example, there is a use of reflexive GIVE in some languages which has the sense of 'come into being, present itself, arise', as in Spanish Si se da el caso, ... 'If the circumstance arises, ...'. This can be contrasted with the use of reflexive GIVE in the German sentence Das Fieber gibt sich 'The fever is going away', where reflexive GIVE carries the sense of 'disappear, die down'. The semantic contradiction in the meanings of reflexive GIVE in the two constructions here does not mean that one should give up the task of providing a rational account of the extensions in meaning. Rather, one needs to seek the basis for the different extensions in different components of the meaning of the source, i.e. literal GIVE. So, for example, the Spanish reflexive GIVE belongs to the emergence/manifestation group of extensions motivated by the emergence of a thing out of the sphere of control of a giver, while the German use relates to the movement away group of senses motivated by the physical movement of a thing away from the giver. While there is a very extensive array of figurative extensions of GIVE, it would be wrong to conclude that any verbal sense could be based upon an extension of literal GIVE. It is particularly instructive in this respect to compare figurative extensions of GIVE and 'take' verbs, although a full comparison of these verbs goes beyond the scope of the present paper. One may note, however, that the movement of a thing implied by both literal GIVE and 'take' is in opposite directions in each case: the thing given moves away from the subject referent in active GIVE clauses, but to the subject referent in active 'take' clauses. This difference is relevant to

Figurative giving

119

understanding the different ways in which the two predicates are extended to communicative acts where we find contrasts such as give a lecture but take notes, give advice to someone as well as take advice from someone. The extensions in these cases preserve the essentials of the force dynamics of the literal GIVE sense on which they are based. It must be emphasized that the overview of the extensions of GIVE which follows is not exhaustive. The number of specific meanings one can identify across languages is quite staggering and I have not attempted to list each and every one which I have encountered. Instead, I have attempted to characterize and motivate the main directions in which GIVE has been extended, illustrated in each case with a few examples. In this way I hope to convey a sense of the panorama of extensions which are found, without detailing every attested extension. I. Interpersonal communication (2)

give advice I a message I an opinion to someone etc.

(3)

/ give you my word 11 give you my blessing etc.

(4)

a. b.

give a concert I a recital I a lecture I an address I a speech etc. Kakvo davat po televizija-ta I v kino-to? (Bulgarian) what give.3Pl on television-the / at cinema-the 'What's on TV /what's on at the cinema?'

The examples in (2) and (3) have parallels in numerous languages and represent one of the most robust semantic extensions in terms of the number of languages which show evidence of this type of extension. The examples in (2) describe the communication of thoughts, information etc., while the examples in (3) illustrate various performative acts. Both types involve semantic components which can be easily mapped onto components of the domains of literal GIVE. Take as an example the expression to give an opinion of something to someone. In terms of the spatio-temporal entities involved, the speaker can be identified with the giver, the opinion with the thing passed, and the addressee with the receiver. The speaker has access to information comparable to the way the giver has control over an object, and the addressee comes to have access to information comparable to the way in which the receiver comes to have control over an object. The act follows a natural path from the speaker through the opinion expressed to the addressee. There are strong parallels, therefore, between the structure of communicating opinions

120

John Newman

to others and the act of giving and these parallels can be seen in a number of the domains (spatio-temporal, control, force-dynamics). This extension of GIVE to describe interpersonal communication is but one part of the larger "conduit metaphor" which underlies and shapes much of our discourse about communication (Reddy 1979). In (3), it is not information which is conveyed, but rather some act is performed, involving a speaker and some other person or person-like entity. A status or privilege is conveyed in a way comparable to how a thing is passed in the act of giving. Note that the performative use of give, like other performatives, uses the simple present tense of English which would not normally be possible with the literal GIVE sense of the verb to refer to an act taking place at the time of speaking. In so far as the performative act of promising, blessing etc. is accomplished in the instant of the utterance, the performative give may be compared with the completedness subgroup of extensions discussed below under VI. We may mention here that the giving act is typically brief, goal-directed, and has a clear endpoint and it is these features which may motivate the use of give as a performative verb, which at the time of utterance indicates an act performed. The expressions in (4) are not examples of communication in the strict sense, but seem similar enough to the other examples to warrant inclusion here. (4a) illustrates expressions pertaining to performances for an audience, whether or not the audience is overtly expressed. The performer may be equated with the giver, the performance itself with the thing passed, and the audience with the recipient. This suggests an equation like performing for an audience is giving something to someone. Alternatively, one could think of these uses of give as a special case of interpersonal communication. Note that there are other ways of talking about concerts, plays etc. which appeal to communication metaphors: the pianist communicated with the audience, I felt the pianist was speaking to me as she played the adagio, she expressed a lot of emotion through her playing, etc. On such a view, performing for an audience is more immediately related to communicating something to someone which in turn is like giving something to someone. Both of these seem possible ways to relate these uses of give to literal GIVE. There are also connections one might make with the 'emergence, manifestation' subgroup of extensions, where one might compare give a concert etc. with expressions such as Italian questo terreno da ottimo grano 'this land produces very good corn', where the GIVE verb dare takes on the meaning of 'produce, give rise to'. There are these similarities, to be sure. I have chosen to include the give a concert

Figu rative g iving

121

kind of example here, however, on the basis of the presence of two human participants (the performer and the audience) which seems a salient part of the sense of a performance, as it is with an act of giving. (4b) presumably belongs here too, even though the 'performer' (TV/radio company, the people arranging the programmes?) and 'audience' (viewing/listening public) are left unspecified. II. Emergence/manifestation (5)

a.

b.

(6)

Questo albero non da piu frutti. (Italian) this tree Neg gives little fruit This tree no longer bears fruit.' Zemja-ta tazi godina dade dobra rekol-ta. (Bulgarian) earth-the this year gave good crop-the 'The land yielded a good crop this year.'

a.

Este terreno da un buen cultivo de maiz. (Spanish) this land gives a good crop of corn. This land produces a good crop of corn.' b. El cultivo estaä se dando bien este ano. (Spanish) the crop is Reflex giving well this year. The crop is coming on well this year.' c. La planta no se da en el sur. (Spanish) the plant Neg Reflex gives in the south. The plant is not found in the south.' d. *Se dan buenas bibliotecas en Madrid. There are good libraries in Madrid.'

e. *Dios se da. 'God exists.' (7)

a.

b.

(8)

Es gibt ein-en Gott. (German) it gives a-Ace God There is one God.' Det gi-s I0sning paproblem-et. (Norwegian) it gives-Passive solution to problem-the There is a solution to the problem.'

khpom ruat tin sailai-rien ?aoy rahas (Cambodian) I run go school give quick run quickly to school.'

122

John Newman

The extension of literal GIVE to a meaning like 'yield, produce' is illustrated in (5). Unlike the senses dealt with under I, the 'yield, poduce' sense does not presuppose two human participants and in this respect it seems further removed from the spatio-temporal configuration which goes to make up literal GIVE. Nevertheless, one can see parallels between the 'yield, produce' sense and the dispensing of an object which forms part of giving. Thus, the tree in (5a) can be equated with the giver, and the fruit equated with the thing passed. The fruit comes out of the tree in the same way that the thing passed moves away from the giver. Similarly in (5b), the crop begins by being within the boundaries of the earth, gradually extending itself out of the bounds of the earth, comparable to what happens in the control domain of literal GIVE. The flow of action is understood most naturally as proceeding from the earth, in which reside minerals, nutrients etc, to the developing crop, matching the flow of action from giver to the thing passed. Although the characterization of 'yield, produce' sense does not necessitate any reference to a human participant, matching the recipient in a giving act, there may nevertheless be a weak implication of some human participant in the sense that the land produces a crop for people of a country/the world, in which case the people are beneficiaries of the event parallel to the way in which the recipient is the beneficiary of giving. The Spanish examples in (6) show further extensions on the idea of 'yield, produce'. (6a) illustrates the same 'yield, produce' sense as in (5). In (6b), a reflexive GIVE construction effectively backgrounds the role of the earth/land and, as is the case with other reflexive constructions, the crop is construed as both agent and patient in the growth of the crop. (6b) has something still of the dynamic nature inherent in literal GIVE and the 'yield, produce' sense in (6a). Consistent with this, the present participle construction esta dando is more appropriate than the simple present and the bien is naturally understood as a manner adverbial, describing the manner in which the growth is taking place. The sense is basically 'grow, emerge'. (6c) uses the same reflexive GIVE construction, but with more of a locative interpretation possible, at least compared with (6b). While there is still the dynamic sense of the plant growing, the overall meaning is largely equivalent to describing the location, or non-location as the case may be, of the plant. Even within the same basic construction, the reflexive GIVE construction, in the one language, Spanish, one can see a variability in interpretation, with an 'emerge' sense more predominant or with a 'location, existence' sense being the more predominant. This use needs to be contrasted with the unacceptable uses of reflexive GIVE

Figurative giving

123

in (6d) and (6e). Obviously, the reflexive GIVE does not extend to denoting simply location or existence. There are other syntactic devices, in addition to a reflexive construction, which turn a GIVE clause into a kind of presentative or existential construction. Impersonal and passive constructions can both be utilized to achieve such meanings. These constructions are illustrated in (7a) and (7b) respectively. In (7a), an impersonal subject es helps to blur the agent role, typically associated with the subject, with the result that the object referent is presented without the idea of anything having acted upon it. Rather than being about the emergence or growth of the object referent, the es gibt construction can be seen as the presentation of the object referent to the addressee/reader. Similar remarks apply to the Norwegian det gis construction, where the use of a passive voice makes it even more difficult to identify an agent role in the clause. Here, too, the effect is that of a presentative meaning. One could claim a dynamic property of all these presentatives, in the sense that a presentative is a kind of abstract motion of the object referent into the mental field of the interlocutors. But it is clearly much less dynamic in its meaning than literal GIVE or the 'yield, produce' sense. The use of GIVE in (8) is not an easy one to motivate, but I believe it can be dealt with as part of the same general tendencies exhibited in (5-7), particularly (7). In Cambodian, GIVE helps to form adverbial constructions, where GIVE relates an adverb-like quality (the 'quickness' in (8)) to a verbphrase ('run to school' in (8)). The manifestation of a quality evident in (8) is like the adjectival counterpart to a presentative construction in (7), which serves to introduce nouns. III. Permission/causation (9)

Ich gab ihm eine Tasse Tee (zu trinken). (German) I.Nom gave him.Dat a.Acc cup tea (to drink) gave him a cup of tea (to drink).'

(10)

a.

b.

Ich gab ihm zu verstehen, dass ... (German) I.Nom gave him.Dat to understand that gave him to understand that antaajonkun ymmärtää, että ... (Finnish) give person.Gen understand that 'give a person to understand t h a t . . . '

124 (11)

John Newman a.

b.

(12)

a.

b.

emu ne dali govorit' (Russian) him.Dat Neg gave.Pl speak 'They didn't let him speak.' Gef wo bäo-bao haizi. (Mandarin) give me hug child 'Let me hug the child.' cake x-'a'a-ni-ayoj ixim awal (Jacaltec) wind Aspect-give-suffix-fall Class cornfield The wind made the corn fall down.' (Craig 1977: 377) bele-wo (Guwinggu) clear-give 'make clean' (Gates 1964: 91)

All the uses illustrated here involve a sense of enabling or causing an event. (9) is a convenient example to begin with, since it seems to have the closest connection with literal GIVE. Both the German sentence and its English translation show how it is possible in these languages to elaborate on literal GIVE clauses with an additional verbal predicate, a verb meaning 'drink' in (9). The GIVE word can still be understood literally, but at the same time it expresses the means by which the further act of drinking can take place. Recall a basic feature of literal GIVE is the special advantaged status of the recipient. Within the domain of human interest, the recipient typically receives the thing passed so that the recipient may then proceed to do something with it. In adding the extra verbal predicate, an implicit part of the meaning of the literal GIVE is made explicit and specified. The verb GIVE has not changed its meaning in (9) from that of literal GIVE. However, the whole construction consisting of the literal GIVE clause integrated with a second verb includes the notion of giving leading to and enabling the act of drinking. The notion of enabling evident in (9) is also present in (10-11), as part of the meaning of 'enable someone to know something/do something'. In neither (10) nor (11), however, is the meaning of literal GIVE part of the overall meaning and from this point of view the meaning of the GIVE construction is further removed from literal GIVE than in (9). Nevertheless, there are parallels one may draw between the 'enable' sense and literal GIVE. With literal GIVE a recipient is passed an object, along with the control of the object, while in (10) a person comes to have access to information through some communication. In (11) a person comes to have rights to act through a granting of permission.

Figurative giving

125

Table 1. Literal GIVE and its 'enable' extensions

Source Domain

Target Domain 1

Target Domain 2

give something to someone

to be given to understand... cf. (10)

emu dali govorit' 'they let him speak' cf. (11)

Initially a giver has control over an object and a recipient does not have control over the object

Initially person A has some information and person B does not have this information

Initially person A controls the rights to some act and person B does not have rights to this act

The giver passes the object and control over the object to the recipient

Person A passes this information to person B

Person A bestows on person B the rights to the act

The recipient has control over the object and the giver no longer has control over the object

Person B has this information

Person B has rights to this act

The mappings between these conceptual spaces are extensive and are detailed in Table 1. There are gaps: a person does not cease to have information after sharing it with others. The nature of information sharing is such that information has no unique possessor and this conflicts with, and overrides, the idea of a unique controller of an object. Similarly, a person still retains authority as to who may perform an act even when permission has been given to someone to perform the act. In all three conceptual spaces, the force dynamics are such that there is a flow of action from one person to another, with the person at the "energy sink" end coming to have some kind of new control over something. The 'allow, permit' sense is an especially common extension of literal GIVE and there are many other extensions which are closely related to this, although I have not tried to document them all here. I would include, as related extensions, purposive and concessive senses, such as /'// give you

126

John Newman

that Ί will concede that point to you' and Russian ej nel'zja dot' bol'she dvadtsati let, lit. 'to her cannot give more than twenty years' meaning 'She does not look more than twenty years'. The examples in (12) extend this notion of enabling a little further by taking on the meaning of 'cause, make happen' without any stipulation that there be a human recipient of the transfer. Instead, it may simply be an event of some sort, rather than an act of some human agent, which appears as a result. This is the pure causative meaning as opposed to the 'enable, let' meaning of (ΙΟΙ 1). The force dynamics of literal GIVE motivate an extension to a causative sense: the giver is an intentional agent initiating an action involving another person and is appropriately described as a causer of an action. Consider also the following use of Mandarin gei: (13)

Jmyu gei m o chl-le. (Mandarin) goldfish give cat was eaten-Aspect 'The goldfish was eaten by the cat.'

The use of GIVE in (13) poses some challenges, but I believe one can make natural connections between these uses and literal GIVE. In Newman (1993), I detail a number of separate considerations which help to motivate the 'agentive' prepostional use of GIVE in (13). One of the relevant considerations is the fact that Mandarin gei is also used in the 'let, allow' sense, illustrated in (lib), and it is this sense which forms the bridge between the literal GIVE sense of gei and its use as an agentive preposition. Broken down into the conceptual steps involved, the chain of extension in the meaning of A gei Β . . . is that shown below: 1. A weakly permits Β to do something. 2. A is such that Β does something. 3. A is such that Β does something to A. (= the passive construction) In both the 'let, allow' sense and as a preposition-like word in the passive construction, gei functions as a relational predicate, introducing some other entity functioning as an agent. Furthermore, the transition from a verb-like word meaning 'let, allow' to a preposition-like word as the agentive marker is consistent with the extensive use of verbs as prepositions in Mandarin. While these few remarks fall far short of a full explanation of the use of gei in (13), they give some indication of where one might look for bridges to the literal GIVE sense.

Figu rative g iving

127

IV. Schematic interaction (14)

give the car a push (= push the car) give someone a shove (= shove someone) give the house a clean (= clean the house) give the floor a sweep (= sweep the floor) give the floor a scrub (= scrub the floor) give the rope a pull (= pull the rope) give it a try (= try it) give someone a scare ( = scare someone) give someone a hug/kiss (= hug/kiss someone) give the ball a kick (= kick the ball)

In the examples in (14) we find a ditransitive give construction functioning as a paraphrase of transitive verbs as shown in parentheses. I take the noun in the second object of the ditransitive construction (push, shove, clean etc.) to be a deverbal and will refer to it as such, although nothing crucial hinges on this assumption. The paraphrase, or near-paraphrase, relationship illustrated in (14) has attracted some attention in recent literature (see in particular the interesting discussion and references in Dixon 1991: 336-362). Grimshaw and Mester (1988: 229-230) suggest an analysis of the give paraphrase in terms of a process called Argument Transfer. Give is viewed as having an incomplete argument structure, like some other "light verbs" in English, e.g. make, put, take. The deverbal noun, on the other hand, is analyzed as having a complete argument structure which is transferred to give. One could proceed in a roughly comparable way in Cognitive Grammar, along the lines suggested by Langacker's analysis of periphrastic do (cf. Langacker 1991: 205-206). In either view give would have a relatively schematic semantic characterization which combines with the more richly elaborated deverbal predicate to form a new semantic whole. However one chooses to account for the semantics of the whole construction, it would be wrong to dismiss give as a semantically empty predicate in the construction. It is possible, as I show below, to isolate components of intentionality, energy flow, and punctual or perfective aspect in its use. Taken together these elements of meaning provide a schematic meaning of interaction between agent and patient-like entities. Before providing the contrasts which bring out the various components of this meaning of give, one should note that there appears to be a kind of formal constraint on the use of a deverbal in the give construction, to the effect that

128

John Newman

it exists as a deverbal independently of its use in the give construction, as discussed by Stein (1991: 7-8). So, for example, w'iih push one can say not only / gave it a push, but also Do you want a push ?, I got a big push etc. If we try to construct a sentence such as / gave the song a sing, its unacceptability might then be explained away as simply a result of the non-existence of a deverbal sing. If we want to tease out the semantic factors influencing the use of the give construction, then, we must use deverbals which have independent existence. Consider, then, spill which is used both as a verb and as a noun (presumably based on the verb, hence deverbal) independently of the give construction, as in There has been an oil spill off the coast. Notice that we do not find pairs such as those in (15): (15)

a. / spilled the milk. b. */ gave the milk a spill

A clear difference semantically between spill and the deverbals in (14) is the involuntary nature of spill. It is understood that spilling is strongly associated with something one does accidentally and this is presumably relevant to the impossibility of (15b). That is, the give construction is associated with an intentional component of meaning. Compare also (16a) and (16b): (16)

a. b.

Lin missed the show. Lin gave the show a miss.

Both miss and give (something) a miss have the sense of 'avoid (something)', but the give construction must be interpreted as an intentional act of avoidance, whereas the plain verb could be intentional or non-intentional. Another component of meaning present in the give construction is a momentary, punctual, or perfective sense, consistent with the extensions of meaning of GIVE noted in VI below. In (17a), kick can easily be used to describe a single instance of kicking as well as repeated instances, whereas give a kick is appropriate for only the one instance: (17)

a. Tom \vas kicking the ball the whole afternoon. b. *Tom was giving the ball a kick the whole afternoon.

Compare also the possibility of give the ball a kick with the impossibility of give the ball a hold (in spite of the independent existence of a nominal use of hold as in He had a good hold on the ball).

Figurative giving

129

Furthermore, where there is a transference of energy from one entity to another, the transference of energy has to be from the subject referent to the object referent. Compare (18a) and (18b): (18)

a. / gave the ball a good throw. b. */ gave the ball a good catch.

In the case of throwing an object, the thrower initiates the energy flow and the object thrown is an energy sink, so the energy flow is outwards from the subject referent. In (18b), on the other hand, the subject referent functions most obviously as an energy sink. The ball comes to rest in the hands of the catcher and energy flows from the ball to the catcher. This is not the entire story, of course. Catching a ball is not the same as a ball landing in someone's hands. To catch a ball, one must also initiate some activity affecting the ball. However, the role of the catcher as a kind of endpoint in the energy flow seems to make a stronger impression than the role of the catcher as the initiator of the catch. This orientation in the energy flow of catching seems relevant to understanding why (18b) is unacceptable. These nuances of the give construction, compared with the corresponding plain verb construction, are familiar as features associated with higher transitivity as discussed in Hopper and Thompson (1980) and Rice (1987). The relatively high transitivity inherent in the give construction does not reduce to any simple presence or absence of a feature, as can be seen in the variety in the give constructions reviewed here. It is helpful to consider the features of the give constructions with respect to the different domains which are relevant, as proposed by Rice (1987: 78-88). So, for example, when physical entities are acting upon each other, as in give the ball a kick, the flow of energy involved results in a change in the location of the affected entity (the ball). In the case of give the show a miss, on the other hand, we are dealing more with a state of mind of a person with respect to some event and there is no obvious effect on the object referent (the show). These different properties should be understood as related to the different effects associated with transitivity in different domains. These properties of the give construction are consistent with and motivated by properties of literal GIVE. Intentionality is very much present in the mental world of the giver in the characterization of literal GIVE, as well as in many of its extensions. It is present in the interpersonal communication extensions and the extensions relating to permitting, enabling etc. Extensions of GIVE to benefactive marking, as discussed below, also involve typically an

130

John Newman

intention that someone be the beneficiary. The momentary, perfective element inherent in the give construction is a further example of the 'completedness' group of extensions discussed under VI, motivated by an inherent perfectivity in literal GIVE. The flow of energy which is present (though understood differently in the different domains) has its source in the clear flow of energy (outwards) from the giver (through the thing) to the recipient. It is appropriate that this particular give construction makes use of the double object structure in English, rather than the Object NP + to phrase alternative (*give a wash to the car, *give a push to the car, *give a shove to the man etc.). In the double object construction the primary object is the thing affected, as is the object of the corresponding monotransitive verb. So, for example, the car is the (primary) object and the thing affected in both give the car a wash and wash the car. Both treat the main interaction as one between the agent washing the car and the car. The high transitive flavour of the GIVE construction noted above is consistent, too, with the agent and thing affected being expressed as a grammatical subject and grammatical object rather than having the thing affected integrated into the clausal structure as an oblique phrase. V. Recipient/benefactive (19)

a.

b.

(20)

a.

b.

(21)

a.

0 fa a fun mi. (Yoruba) he/she sold it give me 'He/she sold it to me.' (Pulleyblank 1987: 989) Wöß-le yi feng gei fa. (Mandarin) I mail-Aspect one Class letter give him/her mailed a letter to him/her.' Tä gei wo zao-Je yi döng fängzi. (Mandarin) he/she give me build-Aspect one Class house 'He/she built a house for me.' afsa' kiboz kuki Kofi kuki me (Siya, Ghana) he-paid money give Kofi give me 'He paid the money to Kofi for me.' (Ford 1988: 144) wale paa-ny-a'n-ya 'aayafo-na knife 3Subj.3Obj-make-give-Past woman-Obj (Nez Perce) 'He made the woman a knife.' (Rude 1991: 186)

Figurative giving

b.

131

maajhaa saaThi he kaam kar-un de-Sil kaa my.Obi sake this job do-Suffix give-Fut question (Marathi) 'Will you do this job for me?' (Hook 1991: 66)

The examples in (19) illustrate a use of GIVE predicates often found in languages having serial verb constructions. Here, GIVE functions as the second of two verbal predicates. GIVE may still have the sense of passing something by hand, but it may just as well function with a more prepositional meaning (e.g. 'direction towards'). It is not always clear, in fact, whether one should call the GIVE predicates in such constructions verbs or prepositions. Semantically, at least, the GIVE predicate functions to integrate a literally or figuratively understood recipient into the larger clause. As can be seen in the Mandarin example (19b), the construction is used with/7 'mail, send'. As such, the person doing the mailing does not actually pass over anything by hand to the recipient, there being one or more intermediaries. The receiver of the letter in (19b) is a recipient, but in a much more indirect sense than in the case of literal GIVE. In languages which allow a GIVE predicate to function in a more preposition-like way as in (19), it is not uncommon for the GIVE predicate to function as a kind of benefactive marker. It was noted above that the typical giving act advantages the recipient in some ways, leading us to recognize the relevance of a domain of human interest in describing the meaning of literal GIVE. This component of the meaning of literal GIVE is a source for the use of GIVE predicates as benefactive markers in languages. Literal GIVE (weakly) implies a benefactive effect on the recipient, just as a benefactive role (weakly) implies that the beneficiary is most likely a recipient of something. Because of these implications, it is not surprising that it is often difficult in practice to identify a purely benefactive role as opposed to a purely recipient role. In (20a), for example, the beneficiary is to 'receive', eventually, a completed house. This example is typical of the blurriness often found with recipient and benefactive meanings. (20b) illustrates an unambiguously benefactive use of GIVE. The examples in (21) illustrate a further variation on this type of semantic extension, whereby the GIVE morpheme is integrated into a benefactive construction, though not in a prepositional way. In Nez Perce, the GIVE morpheme a'n is part of the larger verbal word and its presence in the verb adds the notion 'for the benefit of the object referent'. In Marathi, the GIVE

132

John Newman

morpheme forms part of what is traditionally called a 'compound verb' and signals a benefactive interpretation of the Oblique phrase. (22)

Cantonese a. chit ping-gwo bei ngoh cut apple give me 'Cut an apple for me.' b. bäh-bäjaahn chin bei ngoh dad earns money give me 'Dad earns money for me.' c. ?/* cheung gö bei ngoh OK cheung gö bei ngoh teng sing song give me hear 'Sing a song for me.' d. *ja go ping-gwo bei ngoh hold Class apple give me 'Hold the apple for me.' e. *John heui Wellington bei ngoh John go Wellington give me 'John went to Wellington for me.'

(23)

Hokkien a.

b.

c.

d.

e.

chiat pin-ko ho goa cut apple give me 'Cut an apple for me.' pa-pa than lui h5 goa dad earns money give me 'Dad earns money for me.' chhlu koa h5 goa sing song give me 'Sing a song for me.' gim pin-ko h5 goa hold apple give me 'Hold the apple for me.' John khi Wellington h5 goa John go Wellington give me 'John went to Wellington for me.'

Figurative giving

133

A comparison of two Southern Chinese dialects, Cantonese and Hokkien, shows a possible path of semantic extension of GIVE to benefactive marking. Both dialects extend GIVE (Cantonese bei, Hokkien h$) to benefactive marking, but to different degrees as shown in (21) and (22). Hokkien GIVE has extended to a benefactive marker in a very complete sense, as illustrated in (23e), where there is no implication that anything will be handed to the beneficiary in this sentence. Cantonese GIVE, on the other hand, can function as a benefactive marker only when there is an easy way to understand the beneficiary as a recipient of some act of giving. (22d-e) are unacceptable since there is no clear implication that there is an act of giving involved. It is tempting to think of the Cantonese state of affairs as a stage on the way towards a full-blown benefactive marker as we have in Hokkien. VI. Completedness/perfectivity (24)

a.

b.

mAl ne uske hoThO ko ... TaTol-naa ... Suruu kar diyaa I Erg her lips Dat feel-Inf start do gave Ί began to feel her lips.' (Hindi-Urdu) (Hook 1991: 6, cited from Vaid 1970: 13) Geli-ver! (Turkish) come-give 'Come quickly!'

There is a cluster of grammaticalized extensions of literal GIVE which are motivated by what one might call aspectual properties of literal GIVE. For one thing, literal GIVE is strongly telic in its meaning. There is a definite goal towards which the act of giving is directed, namely the recipient coming to have the thing being passed. In the force-dynamics domain of literal GIVE the recipient constitutes the endpoint or energy sink of the action. Furthermore, the completion or accomplishment of an act of giving is more or less instantaneous in its most typical manifestations. These properties of literal GIVE account for a kind of inherent perfectivity or completedness to the verb. A reflection of this perfective bias within the meaning of literal GIVE can be seen in the verbal morphology associated with Russian dat' GIVE. In Russian, most verbs exist in both an imperfective and perfective form. In the overwhelming majority of cases, a simple stem will be imperfective and the perfective will be formed from that stem through the addition of a prefix, e.g., citat' 'write (Imperfective)' andprocitat' 'write (Perfective)'. There is, however, a small number of verbs where the simple stem itself functions as

134

John Newman

a perfective, including: past' 'fall', lee' 'to lie down', sest' 'sit down', det' 'put', and dal' GIVE (Vilgelminina 1963: 21). These may be considered to be inherently perfective in meaning and, significantly, the verb GIVE is one of these. Although I have no examples where it is the morpheme GIVE which appears as the perfective morpheme, transforming the aspectual nature of the accompanying verb, there are examples where a grammaticized GIVE cooccurs with perfective type verbs. One example of this is found in HindiUrdu, where GIVE functions as a so-called "vector", i.e. an auxiliary-like component, in the compound verb construction, with a meaning sometimes described as perfective. The discussion by Hook (1991: 66-74) suggests, however, that the usual function of GIVE in this construction in Hindi-Urdu is not really to perfectivize the main verb since the main verb is typically one which already contains the idea of completedness. It would appear more correct to speak of GIVE in this construction as indicating the presence of a perfective element of meaning in the verb phrase, rather than creating it. An example of this use of GIVE is (24a). Here the notion of starting an action gives the verbal meaning a momentary and perfective quality, with GIVE as an appropriate accompanying auxiliary. As discussed in Hook (1991: 6263), the same auxiliary GIVE in Hindi-Urdu may also function to indicate the prior action of the main verb it co-occurs with. A related development is the use of GIVE in Turkish, illustrated in (24b). The verb vermek GIVE functions as a suffix/auxiliary verb which attaches to a gerundive form of the main verb with the approximate meaning 'to do the action of the main verb quickly'. To speak of the meaning of GIVE in this construction as adding the sense of 'quickness' to the action is somewhat simplified. It is difficult, however, to pin down a very precise semantic effect of GIVE in this construction. Tarring (1886:107) characterizes the semantics of the auxiliary GIVE in terms of 'promptitude, readiness, facility, offhandedness' and 'cheapening ... or making light of an action'. The example in (24b) is translated by Tarring as 'Just come!'.

VII. Movement away (25)

a.

O navio deu no rochedo. (Brazilian Portuguese) the ship gave in.the rock The ship hit the rock.' (Salomao 1990: 85)

Figurative giving

(26)

b.

Höatu, me waiho maua i konei (Maori) give(away) Particle leave us here 'Go ahead, leave us here.' (Williams 1957: 55)

a.

Das Fieber gibt sich. (German) the fever gives Reflex The fever is going down/away.' Ikkunat antavat pohjoiseen. (Finnish) windows give north-Illative 'The windows face to the north.'

b.

135

A salient part of the meaning of literal GIVE is the movement of the thing being passed from the giver to the recipient. This feature of literal GIVE is the motivation for a variety of extensions of GIVE which relate to concrete or abstract motion. A sense of motion is present, in fact, in a number of the extensions already looked at. So, for example, the extensions having to do with emergence/manifestation all involve a component of motion. In (25a) the GIVE verb is being used as a motion verb, though without any volition being attributed to the thing moving. This is consistent with the absence of any volition on the part of the typically inanimate thing which is passed from the giver to the recipient in the the act of giving. The idea of unexpectedly encountering something also finds expression in Brazilian Portuguese in the dar com construction, literally 'give with'. Thus, dar com o livro means 'to unexpectedly find the book'. In (25b), the Maori verb is the 'give away' form, here used with the meaning of 'go on'. One should mention that the example cited is taken from a nineteenth century text and is not a usage familiar to all Maori speakers to whose attention I have brought the example. It is possible for GIVE morphemes to have the sense of 'movement away', even without a specific 'away' morpheme being present. This relates to the movement of the thing passed in the act of giving being seen from the perspective of the giver, as opposed to the recipient. The use of GIVE in (26a) is perhaps best understood by relating it to this perspective of literal GIVE. In (26a), the fever is construed as moving away, out of sight, as it were, until it is no longer evident. The example in (26b) represents a slightly more abstract version of this 'movement away' sense. Although there is no physical motion of any kind involved in (26b), the spatial orientation of an entity (here, the windows) is construed as a line which moves out from the entity to a compass

136

John Newman

point. In (26b) the compass point functions semantically like a spatial goal of motion. Give way, as in The floor gave way under our feet, may be included in this category as well, since there is a sense of movement (of the floor) away from the speaker. At the same time, give way might also be thought of as primarily a kind of losing, yielding, or conceding, of power or control. Thus, the floor giving way is like the floor yielding to the feet and thereby losing its stability. Understood in this way, give way would be similar to expressions such as give ground and give the right to someone, which in turn are close to the permisssion class of extensions discussed above.

4.

Competing motivations

In the overview just presented, I have necessarily made simplifications for the sake of giving some order to the discussion. Subgrouping the various extensions into seven categories as I have done, for example, is one kind of simplification. These categories help to give some order to the discussion, but there is a certain degree of arbitrariness in categorizing the meanings in this way and it is possible to expand or reduce the number of such categories. The 'transitivity' category of extensions dealt with under IV, for example, has features in common with a number of other categories, such as 'completedness'. It is important to observe that there are alternative paths which could be followed in tracing the connections between the meanings. This is a familiar situation in any synchronic account of polysemy which relies on the intuitions of the linguist to establish the connections between meanings. We have had occasion to point out some of the alternative paths to motivate certain extensions. In the case of the extension of literal GIVE to the use in give a concert, for example, one might connect this use directly to literal GIVE, with a metaphorical giver (the performer), thing (the music), and an implied metaphorical recipient (the audience). However, as discussed above, there is also an extensive metaphorical mapping available which understands the playing of music as a form of communication, whereby a metaphorical speaker (the performer) says something (the music) to an addressee (the audience). This would be an alternative way to understand the path of extension. A third way to motivate the 'perform' sense is to relate it to the 'produce, create' sense found in uses such as give warmth, give off heat, whereby giv-

Figurative giving

137

ing a concert is understood as a kind of external thing produced by the performer, proceeding as it were from out of the performer. This multiplicity of motivations is a common feature of polysemy and adds considerably to the complexity and ambiguity of the networking of the various meanings. An historical account which documents the order in which various extensions develop can narrow down the possbilities. So, for example, we would not wish to motivate the development of sub-meaning X on the basis of another sub-meaning Υ if we know that Υ did not appear until after X had become established. But even a careful documenting of the stages by which a set of related meanings develops does not provide any direct evidence for connecting a new sub-meaning with any one particular pre-existing sub-meaning. If a form has meanings X and Υ at one stage and in the next stage a related meaning Ζ develops, one can not know if it was X or Υ or both which motivated the new extension. So, even when one deals with historical evidence, there still remain many unknowns concerning the details of the semantic connections between related meanings. Furthermore, an historical account of the evolution of polysemy can not be equated with a synchronic account of the relatedness of the meanings. Even when one considers just a single connection between two senses, there are alternative motivations which can be offered. It would be simplistic to think that one can always isolate one component of the meaning of literal GIVE which motivates the extension. The meaning of literal GIVE itself is so complex in its semantic structure that it is normally not possible to draw the boundaries around just one component of meaning of literal GIVE. The recipient, for example, functions simultaneously in various ways, depending on the domain: the recipient represents a kind of spatial goal in the spatiotemporal domain, the centre of a new sphere of control over the thing passed in the control domain, an energy sink in the force-dynamics domain, and the person typically benefiting from the act of giving in the domain of human interest. Distinguishing the various domains, as I did in Section 2, helps in our description of meanings, but at the same time it must be recognized that the domains are closely inter-related and one can not entirely ignore any domain in the description of part of the meaning of GIVE. If I appear to have done this in my discussion of the figurative extensions of GIVE, it is because I am directing attention to what I consider to be the most relevant component of the meaning of literal GIVE at that point in the discussion, rather than the only component of meaning which is relevant. In general, one must always be prepared for multiple motivations in the extension of meaning. The more

138

John Newman

conceptual complexity in the meaning of the source morpheme, the more motivations one is likely to find, whether one is thinking about synchronic relatedness of meanings, i.e. semantic networks, or diachronic paths of evolution of these meanings. Finally, there is a kind of motivation for extensions in the use of GIVE morphemes which relates to the forms of constructional types in a language. Throughout this discussion I have been assuming the existence of 'constructions' which pair constructional meanings and forms as real symbolic units, on a par with symbolic units at the lexical level. I have focused on semantic motivations for figurative extensions, but one must also consider aspects of the form of the figurative extensions. In many cases, the form of an extended sense of GIVE is a form which has independent existence in the language and this in turn provides some additional motivation for the new use of the GIVE morpheme. For example, German has extensions of literal GIVE which involve impersonal, reflexive, and impersonal reflexive constructions as illustrated below: (27)

a.

b.

c.

Es gibt sieben Universitäten in Neuseeland. it gives seven universities in New Zealand (Impersonal) 'There are seven universities in New Zealand.' Das Fieber gibt sich. (Reflexive) the fever gives Reflex The fever is going away.' Es be-gab sich, dass ... (Impersonal reflexive) it Prefix-gave Reflex that 'It happened that...'

Each of these constructions exists in German quite independently of these figurative GIVE uses. That is, one must recognize for German these three constructional types each with its own characteristic form and schematic meaning. In accounting for these figurative GIVE uses, then, one can appeal to already existing morphological/syntactic devices as part of the explanation of why the new uses are possible. The extension of literal GIVE in (27) elaborates the polysemy of existing construction types rather than creating brand new construction types. Thus, it is no coincidence that languages such as German, Italian, Spanish are rich in figurative extensions of GIVE involving grammaticized Reflexive and Impersonal constructions but lack extensions of GIVE to preposition-like uses meaning Benefactive. These languages have

Figurative giving

139

well established Reflexive and Impersonal construction types but do not have well established serial verb type constructions in which morphemes are ambiguous between verbs and prepositions. Certain African languages and Chinese dialects, on the other hand, behave in exactly the opposite way: they have well-developed preposition-like uses of GIVE morphemes, but lack Reflexive and Impersonal types of extensions. This reflects the availability of different conventional devices in the languages.

References Chapman, Robin S. 1981 Cognitive development and language comprehension in 10-21-months-olds. In Rachel E. Stark (ed.), Language Behaviour in Infancy and Early Childhood, 359-391. New York: Elsevier North-Holland. Craig, Collette G. 1977 The Structure ofJacaltec. Austin: University of Texas Press. Dixon, Robert M. W. 1971 A method of semantic description. In Danny D. Steinberg and Leon A. Jakobovits (eds.), Semantics: An Interdisciplinary Reader in Philosophy, Linguistics and Psychology, 436-471. Cambridge: Cambridge University Press. 1991 A New Approach to English Grammar, on Semantic Principles. Oxford: Clarendon Press. Ford, Kevin C. 1988 The Central-Togo languages. In M. E. Kropp Dakubu (ed.), The Languages of Ghana, 119-154. London: Kegan Paul International. Grimshaw, Jane and Armin Mester 1988 Light verbs and θ-Marking. Linguistic Inquiry 19.2: 205-232. Hook, Peter Edwin 1991 The emergence of perfective aspect in Indo-Aryan languages. In Elizabeth Closs Traugott and Bernd Heine (eds.), Approaches to Grammaticalization. Vol. II, 59-89. Amsterdam: John Benjamins. Hopper, Paul and Sandra A. Thompson 1980 Transitivity in grammar and discourse. Language 56: 251-299. Langacker, Ronald W. 1987 Foundations of Cognitive Grammar, Vol. 1. Stanford: Stanford University Press. 1991 Foundations of Cognitive Grammar, Vol. II. Stanford: Stanford University Press. Newman, John 1993 The semantics of giving in Mandarin. In Richard A. Geiger and Brygida Rudzka-Ostyn (eds.), Conceptualizations and Mental Processing of Language, 433-485. Berlin/New York: Mouton de Gruyter. 1996 Give: A Cognitive Linguistic Study. Berlin/New York: Mouton de Gruyter.

140

John Newman

Gates, Lynette Frances 1964 A Tentative Description of the Gunwinggu Language. Oceanic Linguistic Monographs 10. Sydney: University of Sydney. Ogden, Charles K. 1968 Basic English: International Second Language. New York: Harcourt, Brace & World. Pulleyblank, Douglas 1987 Yoruba. In Bernard Comrie (ed.), The World's Major Languages, 971-990. New York: Oxford University Press. Reddy, Michael J. 1979 The conduit metaphor - a case of frame conflict in our language about language. In Andrew Ortony (ed.), Metaphor and Thought, 284-324. Cambridge: Cambridge University Press. Rice, Sally Ann 1987 Towards a cognitive model of transitivity. Unpublished Ph.D. dissertation, U.C.S.D. Roberts, John 1987 Amele. Beckenham, Kent: CroomHelm. Rude, Noel 1991 Verbs to promotional suffixes in Sahaptian and Klamath. In Elizabeth Closs Traugott and Bernd Heine (eds.), Approaches to Grammaticalization, Vol. II, 185-199. Amsterdam: John Benjamins. Salomao, Maria-Margarida 1990 Polysemy, aspect and modality in Brazilian Portuguese: The case for a cognitive explanation of grammar. Unpublished Ph.D. dissertation, University of California, Berkeley. Stein, Gabriele 1991 The phrasal verb type 'to have a look' in Modem English. International Review of Applied Linguistics XXIX: 1-29. Tarring, Charles James 1886 A Practical Elementary Turkish Grammar. London: Kegan Paul, Trench & Co. Vaid, Krishna Baldev 1970 duusre kinaare se... (From the other shore ...). Delhi: Radhakrishna Prakashan. Vilgelminina, A. A. 1963 The Russian Verb. Aspect and Voice. Moscow: Foreign Languages Publishing House. Williams, Herbert W. 1957 A Dictionary of the Maori Language. Wellington: R. E. Owen, Government Printer.

Nominalizations, metonymy and lexicographic practice Kiki Nikiforidou

1.

Introduction

This paper examines the semantics of nominalizations in English and Modern Greek. I first suggest that the different meanings of nominalizations are systematic enough to allow description through general principles. These principles can be analyzed as general metonymies which link the various meanings. Secondly, I propose that these meanings should be taken into account when composing dictionary entries in a way that reflects their regularity and systematicity. The object of this study are nominalizations which derive from verbal and adjectival sources. More specifically, I will examine the nominalizations which, in Vendler's early treatise of verbal semantics (Vendler 1967), are described as "a means of packing a sentence into a bundle that fits into other sentences" and are usually referred to as "action" or "state" nominals. Thus, nominals such as English purchase, knowledge, predictability and refusal are all within our scope, while farmer, diner or ancestry are not. It should be also made clear that the results in this paper concern only "pure" nominalizations of the type just mentioned, and not the -ing complements, an example of which is found in (1): (1)

Their refusing our request was rather annoying.

There are several differences between the -ing derivational complements and pure nominalizations. First, the -ing derivation, unlike pure nominalizations, is completely regular. Secondly, there are good reasons to assume that refusing in (1) is immediately dominated by a verb rather than by a noun (McCawley 1988, Vol. 1 contains a long list of arguments to this effect). Third, the range of meanings of refusing is different from that of refusal.1 Langacker (1991: 32-33) characterizes the difference between the -ing forms and pure nominalizations in semantic terms, as a difference deriving from the "process type" character of an action nominalization vs. the "ungrounded instance of the type" that we have in the -ing forms. In Greek, there is not a regular

142

Kiki Wkiforidou

derivational pattern corresponding to the -ing complements and thus Greek has only two options: a finite clause in subject, object etc. position and pure nominalizations. A big issue in the literature on nominalizations has been the issue of predictability both on the morphological and the semantic levels. Earlier discussions of nominalization centered on whether it should be considered a matter of the lexicon or a matter of syntax. On the basis of syntactic and semantic idiosyncrasies, Chomsky (1970) opted for the former, and the conclusion that nominalization was rather a lexical phenomenon was also reached in other studies (Jackendoff 1975, Aronoff 1976, Levi 1978, Selkirk 1982 and others). Indeed, it has not seemed possible to predict whether a given nominal will conform to a given derivational pattern, nor has it been always possible to predict the syntactic and semantic properties of the resulting form. An alternative analysis, which relates nominals to sentential sources by transformation, is proposed in Lakoff (1970) and more recently in McCawley (1988) with an intermediate derivational stage. Cross-linguistically, languages show varying predictability with respect to their noun formation processes. In Hebrew, for example (cf. Comrie and Thompson 1987: 357), there is no way of predicting the form of the action nominal from the form of the root. The same is true of Modern Greek, where the form of the verb or the adjective does not determine the form of the nominal. Thus, aSikia 'injustice' is the derived nominal from aoiko 'to wrong someone', but ayanaktisi is the nominal from ayanakto; although the source verbs belong to the same conjugational paradigm, the nominal endings belong to different morphological classes. English is slightly more predictable in this respect; as observed by Comrie and Thompson (1987: 357) and others, almost any polysyllabic verb will form its action noun in -tion and most adjectives in -able/-ible will form nouns in -ity. On the other hand, there is no way of predicting that refuse will form the nominal refusal while accuse will give accusation. On the semantic level, we can safely say that it is not uncommon to find a deverbal noun taking on special and unpredictable meanings. English deed, for example, does not simply mean 'the action of doing' and saying is not only 'what someone says' (I will discuss such cases in section 4). However, nominalizations have also meanings such as the action or state, the product of the action, the manner in which the action was done and certain others, which are recurrent and common enough to suggest regularity. Such meanings are abstract and cross-lexemic, since they characterize nominalizations

Nominalizations, metonymy and lexicographic practice

143

as a class, and at the lexical level they flesh out by the specific meaning of each nominal. The rest of this paper is thus concerned with identifying and describing such regular patterns in the semantics of nominalizations and with proposing possible applications in lexicography.

2. The data Consider examples (2-4): (2)

The final delivery lasted an hour.

(3)

The final delivery was in huge paper boxes.

(4)

The final delivery was careless and sloppy.

In (2), the nominalization delivery refers to the action of delivering and the predicate specifies the duration of this action. In (3), delivery denotes the concrete product of the act of delivering, while in (4) the adjectives careless and sloppy refer to and specify the manner in which the act of delivering was performed. A first observation is that this kind of polysemy2 of action, product and manner meanings is very common both in English and in Greek. Examples (5-7) are corresponding examples from Greek with the nominal apotamiefsi 'saving money' in action, product and manner meanings: (5)

/ apotamiefsi ine aparetiti the saving-up be-3SG-PRES necessary 'Saving up is necessary.'

(6)

pire us apotamiefsis tu ke pije sto eksoteriko take-3SG-P the savings his and go-3SG-P abroad 'He took his savings and went abroad.'

(7)

ekanan tin apotamiefsi ipomonetika ke xoris piesi do-3PL-P the saving patiently and without pressure They did their saving patiently and without pressure.'

Criticism, painting, work, writing, recording, description, applause, performance are all straightforward examples of the action, product, manner polysemy. A description may, for example, 'run for an hour', 'appear on the front page' or 'be thorough'. The product in each case may be more or less abstract and more or less distinct or separate from the action, depending on

144

Kite Nikiforidou

the corresponding verbal semantics; when you paint a painting, the product of the action is identified with the patient member in the verb frame, but when you criticize a work, the work is not the product of your criticism. Something that emerges clearly from this group of nominalizations is that this general semantics of action, product and manner characterizes nominalizations independently of their derivational class.3 Action, product and manner are common meanings of nominalizations which are derived from verbs that denote activities. Nominals derived from stative verbs and adjectives show a different kind of polysemy exemplified in (8-11): (8)

John's knowledge of the secret dates back to 1960.

(9)

John's knowledge of the subject is practically unlimited.

(10)

John's resentment of his stepmother has been going on for years.

(11)

John's resentment of his stepmother exceeds his common sense.

In (8), the derived nominal refers to the state of John's knowing and the sentence affirms that this state begins to hold at a certain point in time. In (9), however, what is actually predicated of is the extent or degree to which the state holds or is true. Similarly in (10), resentment denotes the state expressed by the corresponding verb and the sentence is an assertion of the state's lasting for a long time. In (11) though, we have again the extent meaning, the verb "exceed", requiring a subject with a degree meaning-component. Sentences (12) and (13) are corresponding examples from Modern Greek with the nominal amixania 'embarrassment' derived from the adjective amixanos 'awkward, embarrassed': (12)

i amixania tu krause Ιΐγαίβρία the embarrassment his last-3SG-P few minutes 'His embarrassment lasted a few minutes.'

(13)

telefiea δίχηί megali amixania brostase kosmo recently show-3SG-PR great embarrassment in-front-of people 'Recently, he shows great embarrassment in front of other people.'

Love, hate, resemblance, similarity, indignation, purity, sympathy, clumsiness, stability are other examples of the state-degree polysemy. Hate can, for example, "last for years" or can be described as "great" or "excessive", and

Nominalizations, metonymy and lexicographic practice

145

so on. Again, what is true of both English and Greek is that this polysemy characterizes nominalizations of different derivational classes. A general observation concerning state nominalizations is that nominals derived from Stative adjectives often express qualities. Nominals such as stability and clumsiness mentioned above, innocence, intelligence and several others, besides their stative meaning, may have a quality meaning as well (e.g., "John's intelligence is his most remarkable quality"). Here, I will just take for granted what has been pointed out very early by Vendler (1967), namely that from the point of view of a time schema, qualities behave exactly like states; in Vendler's words, qualities are "inherent states". The similarity between states and qualities is also captured by Langacker's (1987) treatment of both as atemporal relations. It should not therefore be surprising that nominalizations deriving from stative adjectives have both state and quality meanings. Whether the state or the quality meaning is the basic one in such nominals is left as an open question, amenable to considerations of centrality and basicness that apply to all cases of polysemy. Consider now (14-18): (14)

John's love of antiques is well-known.

(15)

Mary's instability in her relationships is not surprising.

(16)

Their survival was pure luck.

(17)

The government's decision to sell public companies raised havoc.

(18)

The writing of that particular novel by John was rather surprising.

These examples illustrate yet another meaning common in both action and state nominalizations, the "fact that verbs" meaning. It is the fact that he loves antiques which is well-known, the fact that the government decided to sell which raised havoc, etc. The "fact" meaning characterizes Greek nominalizations as well, as for example in (19): (19)

i apofasi tu na xorisi Sen me ekseplikse the decision his subj. divorce NEG me surprise-3SG-P 'His decision to get a divorce did not surprise me.'

The final meaning of nominalizations which is common enough to deserve mention is exemplified in (20-24): (20)

The explosion was costly.

146

Kiki Nikiforidou

(21)

The first experiment encouraged us to go on.

(22)

The examination was disastrous for most of the students.

(23)

The damage was permanent.

(24)

We need to clean up a leak in the basement.

The nominalizations in (20-24) are action or event nominalizations (Langacker 1987) defines an event as "a cognitive occurrence of whatever degree of complexity"). Yet, in these examples the nominalizations do not denote the action or the event but rather the results or the effects of it. It is the results of the explosion which are costly (at least under one interpretation), the results of the experiment that encouraged us to go on, the result of the damage which is permanent and the results of the leak that we really have to clean. A corresponding example from Greek is given in (25): (25)

i ektimisi tis periusias θα Simosiefti sindoma the evaluation the property-GEN PUT appear-3SG soon 'The evaluation of the property will be published soon.'

where ektimisi 'evaluation' is referring to the concrete results of the evaluation that will be published.4

3. Relating meanings by metonymy An action, the product or the result of this action and the manner in which it is performed are of course intuitively related. To make this intuition more concrete, we may say that they are related by means of the scene and the corresponding linguistic frame (terms used by Fillmore 1971, 1985 and elsewhere, and by many others) evoked by the source verb. The frame of the verb record, for example, includes at the very least an action and a product component. The type of the product is determined by the specific semantics of the verb record. By virtue of its action component, the frame also includes a manner component since every action can be qualified with respect to the manner in which it was performed. A speaker of English knows that the product of an act of recording can only be one of a number of things (tape, videotape etc.). He/she also knows that the action of recording, by virtue of being an action, can be modified as to the manner in which it was done (good, bad, clear etc. recording). The same is true of the state and degree meanings which are both components of the relevant frame; it is part of our

Nominalizations, metonymy and lexicographic practice

147

knowledge about states that they continue to hold unless actively stopped and that they can characterize someone or something to a certain extent. Given this common experiential background, we may now look at the nature of the links between these meanings. The suggestion forwarded here is that these meanings are related metonymically, through general metonymies operating at the cross-lexemic level. The relevant metonymic mappings are stated in (26): (26)

a. b. c. d. e.

ACTION (stands) for the PRODUCT of the action ACTION for the MANNER in which the action is performed STATE for the DEGREE to which the state holds ACTION/EVENT for the RESULT(S) of the action/event ACTION/STATE/EVENT for the FACT that action/state/event occurred

The statement in (a) relates the product to the action meaning, (b) relates manner to the action, (c) links the degree to the state, (d) the event to its results and (e) relates the fact meaning to the action/state/event ones. The assumption in all (a-e) is that the state, action or event meanings in nominalizations are the basic ones. The answer to "What is metonymy?" differs among linguists. As pointed out by Emanatian 1991, the term, as used in the literature, refers to at least three things: 1) the link from one meaning to another, 2) the motivation for such a link between meanings and 3) a process of semantic extension, especially in the context of historical semantics. Metonymy and metonymic mapping, as used in this paper, come closer to the first of these uses, that is they refer to the type of link between the different meanings of nominalizations which is motivated by the common experiential background (the frame). In section 4,1 will attempt to qualify this further, relating metonymy to meanings of varying degrees of conventionality. The idea that an action may metonymically stand for its product or result is found in Ullmann (1979: 220), who discusses examples such as binding, performance, addition etc. in action and product meanings. It is also found in Breal's (1900: 134ff.) notion of "concretion of meaning", which refers to examples such as Latin fructus 'the action of enjoyment' eventually coming to mean 'fruit of the earth'. The idea that concretion is involved in this process is also found in Ullmann, who notes that, in this context, metonymy tends to give abstract words a concrete meaning. The metonymies in (26) differ, however, from both Breal's and Ullmann's accounts in two significant

148

Kiki Nikiforidou

respects. First, they cannot always be identified with concretion of meaning. Concretion may perhaps characterize the action-product polysemy in some cases, but it is in no way involved in the action-manner or the action-fact extensions. Secondly, the metonymic mappings in (26) are directly related to nominalizations as a morphological class, and it is only in the context of nominal semantics that they can be seen as productive and systematic, and not random or sporadic meaning extensions. As noted above, the product or the result is only one component of the verbal frame; manner or degree are different aspects of meaning which may also be systematically related to the basic action/state component. The clearest cases of metonymy are these in which one entity is used to refer to another which is somehow related to it. Example (27), originally due to Geoffrey Nunberg, was discussed by Lakoff and Johnson (1980: 35) in order to illustrate a different cognitive mechanism from metaphor: (27)

The ham sandwich is waiting for his check.

As evidenced by (27), in order for metonymy to work there has to be a common conceptual structure which contains both the target concept, in this case the actual person referred to, and the source concept which is the one actually used. The common structure in this case is the real-world scene which includes the customer and his order. What makes this sentence a good example of metonymy is that it satisfies the requirement (discussed by Lakoff 1987 and Fauconnier 1985) that the choice of the source concept, in this case "the ham sandwich", should uniquely determine the target concept, as indeed happens in (27). The target and the source concepts have thus to be somehow related within a common conceptual structure. One type of such relationship is a part-for-part relationship, where one part of the conceptual structure metonymically stands for another. Part-for-whole and whole-for-part are other possibilities discussed in the literature through many examples (cf. selectively Ullmann 1979, Nikiforidou and Sweetser 1989, Wilkins 1981). For the "ham sandwich" example, we may therefore say that it is a part-for-part metonymy, with one part of the scene used to refer to another. At a different level of abstraction, the metonymies in (26), which motivate the meanings of nominalizations, are also part-for-part metonymies, where one aspect or component of the relevant frame, the action or the state or the event, is used to refer to another. The product and result meanings, linked metonymically to actions and events, are components of the frame or, in Lakoff's (1987) terms, of the Idealized Cognitive Model associated with the

Nominalizations, metonymy and lexicographic practice

149

source verb or adjective, and they are more or less prominent participants in the frame. Similarly, the manner and degree meanings, as noted above, are related to the frame through the general pragmatic knowledge of a speaker about actions and states and in this sense they are also parts of the frame. The fact meaning is more general and not directly linked to the verbal frame (and can therefore be triggered globally by actions, states and events). It is the salience of the action or state parts of the frame that enables them to stand for the other components, in the same way that in Langacker's description of Active Zone phenomena (1987 and elsewhere), it is a prominent substructure within a profile which is used to refer to some other aspect or substructure. Positing a salient action or state component in the semantic frame of nominalizations is intuitively appealing, since it can successfully distinguish between a simple noun, e.g. chair, and a nominalization such as delivery. Relating nominals to their source verbs or adjectives derivationally may not be necessarily desirable or obtainable given the morphological and semantic irregularities we noted before (but cf. McCawley 1988 for a different view). The fact remains, however, that an action or state meaning component is really what distinguishes a nominalization from a simple underived noun.

4.

Conventionality, polysemy and predictability

I have so far identified action, product, manner, state, degree, event, result/effect and fact meanings for derived nominals. To the extent that a given nominal manifests two or more of these meanings, I have suggested that they are related by metonymy. In this section, I will address questions which concern the cognitive and the analytical status of these meanings referring specifically to (a) their degree of entrenchment and conventionality, (b) their status as semantic or pragmatic extensions and (c) their degree of predictability given the meaning of the source verb or adjective. The suggestion I forward here is that the answer to these questions cannot be general and unified for all cases, but has to be examined separately for each nominal and for each category of meaning. Entrenchment, as defined by Langacker (1987: 59-63), is basically a matter of repeated use. Entrenchment, together with "sharedness" (by speakers of a language) of a given word or meaning, determines, according to Langacker, the degree of conventionality of the word or meaning. Both entrenchment and "sharedness" (and therefore conventionality) are a matter of degree and

150

Kiki Nikiforidou

there is no significant cut-off point beyond which we could say with safety that something is entrenched or conventional. It follows that any judgements of entrenchment or conventionality (in this sense) are to a large extent subjective. Yet, with respect to nominalizations, it is useful and illuminating to recognize a gradation of entrenchment if only to describe a speaker's intuition that certain of the meanings I discussed above are more "there" for certain nominals than for others. For nominalizations such as painting, recording, writing, criticism or description, we may safely say that the product meaning has been sufficiently entrenched to represent a conventional meaning of these words. The product sense in this case has entered the lexicon as an established unit which refers to a particular kind of object. By contrast, for nominals such as evaluation, accumulation, reading or investigation, the product meaning is not entrenched to the same degree, although it is readily available in context, (e.g., "The evaluation of the king's property will appear in tomorrow's papers", where the nominal evaluation refers to the product or result of the act of evaluating). We would therefore like to say that there is a difference in the degree of entrenchment between the product meaning of painting on the one hand and the product meaning of evaluation on the other. Finally, the product or result meaning of explosion (cf. example (20)) appears to be even less entrenched, requiring strong contextual support. In the same line, we can say that the manner meaning of, e.g., description or performance is much more entrenched and shared (e.g. a thorough description, a good performance etc.) than the manner meaning of allocation if only because the former are much more often contextually relevant. Given the tenet of Cognitive Grammar that conventionality is a matter of degree, the conclusion reached here is rather natural and expected. It points, however, to a need of determining the status of these meanings on an individual basis for each nominal, since the degree of entrenchment (and in this sense of conventionality) appears to vary considerably for different nominals. The distinction between polysemy and vagueness (or, in Horn's (1985) terms, pragmatic ambiguity) is the second issue that enters into nominal semantics. This distinction has always been a problematic one and standard analyses of lexical semantics (cf. Lyons 1977/1983, Cruse 1986) treat the subject with reserve, pointing out the complexity of the issue. Geeraerts (1993) discusses in detail the instability of the vagueness-polysemy distinction showing that the proposed tests and criteria often lead to contradictions. Indeed, the linguistic tests that have been proposed, e.g., the "so" deletion test or the deletion of a constituent only if it has the same meaning with a

Nominalizations, metonymy and lexicographic practice

151

previous occurrence, are not always applicable nor do they always yield consistent results. In the case of nominalizations, linguistic tests are also rather inconclusive; thus, a sentence like "The painting, which took him an hour, is hanging on the wall" is acceptable, while "The painting is hanging on the wall and took him an hour" is decidedly worse, leading to different conclusions as to the distinctness or not of the product meaning and its status as a contextual variant. What we can safely say for the product meaning is that the particular type of product involved naturally does play a role in its status as a pragmatic or semantic extension. The fact, for example, that recording or painting refer to very specific kinds of objects (although not necessarily unpredictable from the meaning of the source verbs) associated with salient properties, stable and constant across different contexts, enhances their lexical meaning status. And of course, specialization of meaning in the product sense also strengthens semantic status (e.g., saying, deed, admission (=money)). We could predict, therefore, that both the specificity and the specialization of the product meaning should correlate with distinctness and conventionalization (cf. also the discussion on predictability below).5 The possibility of finding a single definition, which has also been proposed as a criterion of vagueness or abstraction (if we can find a definition that encompasses all possible meanings of a word, then these meanings should be treated as pragmatic variations and the word as vague rather than polysemous), is highly subjective and often relies on the ingenuity of the researcher and his ability to come up with a definition general enough to cover all meanings. For nominalizations, the definitional test is hardly any more conclusive since it does not seem any harder to come up with a definition encompassing action and product than to imagine one which would include action and manner; and yet, manner appears to be a completely general interpretation, independent of the specific lexical semantics. Another criterion, proposed by Horn (1985), Sweetser (1986) and others, suggests that the existence of distinct forms in some languages argues for the postulation of corresponding distinct meanings in languages that do not make the same formal distinctions. While both in English and in Greek all these meanings are carried by the same form, that is by the action or state nominalization, there are languages which have distinct morphological forms for their product and manner nominalizations. As reported by Comrie and Tompson (1987), Turkish has a special derivational pattern for forming nouns which mean "way of verbing" and so do Hebrew and Zulu. Similarly, many Bantu languages, as reported by Givon (1970), have special morphological

152

Kiki Nikiforidou

forms which designate the result or the cognate object of an action, and the same is true in some Indian languages. This would seem to favor the existence of distinct meanings in English and Greek nominalizations as well. It does not help us, however, with the distinction between manner, fact and degree on the one hand, which for English and Greek are completely general, and product or result on the other, which are more restricted and lexically dependent. Indeed, the fact and manner meanings appear to be completely general, in the sense that they are available in context for all action/state and for all action nominalizations, respectively. I have not found a single case where the lexical meaning of a given nominal effects the manner and fact interpretations, barring them from arising. Every time that we have an action nominalization, we can always, in context, refer to the manner in which the action was performed (cf. (4) and (7)) and every state or action nominalization can be used to refer to the "fact-of-verbing" (cf. (14-18)). This seems to suggest that manner and fact are contextually triggered interpretations, not necessarily part of the semantic content of nominalizations. In the same line, the degree interpretation of state nominalizations, by the same criterion of generality, seems also to fall into the category of pragmatic interpretations. Given a state nominal, we can always refer to and modify the extent to which the state holds (cf. (9, 11, 13)) and apparent exceptions can be readily explained by the lexical semantics of the specific nominal. Thus, in a sentence such as (28), (28)

His availability is in doubt.

the nominal availability appears to have a "whether or not" reading rather than, or in addition to, a "degree to which" one. We can, however, account for this simply by saying that the semantics of available is such that it allows reference to only two degrees, hence the "whether or not" interpretation. Still in the context of generality and predictability, let us look at the nominal knowledge mentioned in previous examples as an instance of the statedegree polysemy. The frame of know includes at least two basic senses, that of knowing a fact (e.g., "I know that he is a fool") and that of knowing a field (e.g., "I know French"). Compare, for example, (29) to (30): (29) (30)

I finally begin to know linguistics. ?I finally begin to know that Sam is a fool.

Nominalizations, metonymy and lexicographic practice

153

Both senses are stative but as shown by (31), it is only the "knowledge of a field" sense of the nominalization which allows a degree reading: (31)

His knowledge of the subject/*the fact is fairly extensive.

In the case of knowledge, I suggest that we may again speak of regular or predictable determination of the meaning of the nominal by the meaning of the verb (cf. also the discussion on availability above), since it is only know in the "field" sense which is acquired in steps and thus allows a degree interpretation. The metonymy "state-for-degree" can only be triggered by one of the meanings of the verb but the trigger need not be stated explicitly since it is predictable from the semantic analysis of the verbal frame. In this sense, knowledge does not represent an exception to the generality of the state-degree metonymy which, like manner and fact, appears to be completely regular. Going back to the original question of whether the meanings associated with nominalizations are a conventional part of their meaning or simple contextual interpretations, we may conclude that the answer cannot be unified and general but has to take into account the kind of general cross-lexemic meaning involved. Moreover, for the product and result meanings the answer may reflect different degrees of conventionalization for different nominals. This suggests that semantics and pragmatics should be seen as points on a continuum rather than as a strict dichotomy. This is a conclusion reached by many researchers with respect to different phenomena and kinds of data (cf. selectively Fillmore and Kay 1987, Fillmore, Kay and O'Connor 1988, Lakoff 1987, Langacker 1987), in accordance with the widely held view in Cognitive Linguistics that absolute distinctions are not characteristic of natural language in any domain. Metonymy is, therefore, taken as linking both semantic extensions and pragmatic (and in this sense non-conventional) interpretations to the basic nominal meanings of action or state. I thus suggest that the same cognitive mechanism, metonymy, allows the source semantics of nominalizations to stand for and refer to the other meanings of the nominal, regardless of their conventional or non-conventional status (which of course presupposes that both semantic and pragmatic extensions are distinct enough and identifiable). Metonymy is the basic motivation for the polysemy of nominalizations and underlies extensions of all kinds. This suggestion is similar to that made by Sweetser (1990), namely that it is possible for both polysemy and pragmatic ambiguity to have the same metaphorical motivation. The same general

154

Kiki Mkiforidou

metaphor "internal world as external world", according to Sweetser, motivates both the different (conventional) meanings of the modal verbs and the different pragmatic interpretations of various conjunctions. In the same way, a single cognitive mechanism, metonymy, realized as different metonymic mappings, has been argued to motivate both semantic and pragmatic extensions of the source meanings of nominalizations. The final point I will examine in this section is the issue of predictability or regularity of the metonymic mappings which, as I suggested, motivate the polysemy of nominalizations. My main concern is to show that, just as with entrenchment and conventionalization, the predictability of the extensions with respect to the meaning of the verbal or adjectival source is again a matter of degree and can only be determined on an individual basis. The verb write, for example, (discussed by Fillmore in a frame semantics context) is linked to a frame and an associated scene part of which is the knowledge of a speaker of English that the product of the act of writing can only be something linguistic; it cannot be, for instance, a picture or a smear. In this case, the meaning of the nominalization writing in its product sense may be said to be completely predictable from the corresponding verbal frame. Similarly, draw includes in its frame the knowledge that the product is going to be something representational and this is indeed the meaning of drawing. For a nominal such as deed, however, the story is more complex. The semantic oddity of (32), (32)

*One of John's deeds this morning was to brush his teeth.

was used by Chomsky (1970) in order to argue that the semantic relationship between nominals and their source verbs is not always predictable; deed is not simply what you do, but rather a notable or heroic achievement. While deed does also have the regular meaning of "something that is done", the fact remains that its meaning includes a specialized extension that has to be specified separately and explicitly. An extension towards specialization of meaning also characterizes the meaning of saying which is not what someone says but has acquired the meaning of a "maxim" or "proverb". This information is not redundant given the verbal frame of say and the metonymic mapping "action-for-product", but it is something that has to be stated explicitly in the definition of the nominal. Similarly, the Greek derived noun kataptosi comes from the verb katapefto which means 'to get exhausted, weakened'. The nominal, however, has broadened its meaning to include both 'physical exhaustion, weakening' and 'decadence' in general. In this case, we have

Nominalizations, metonymy and lexicographic practice

155

a generalization of the original meaning which again is not completely predictable from the statement of the relevant metonymy. We may thus conclude that predictability in the semantics of nominalizations is a matter of degree ranging from the completely predictable (e.g., drawing) to the partially predictable (e.g., kataptosi) and to the completely unpredictable (e.g., saying).6 In the following section, I will suggest that the number of regular cases is very high both with respect to the availability of the product/result meaning and with respect to its regular semantic determination by the corresponding verb or adjective frame. In fact, it is so high that it deserves systematic treatment in lexicography. Here, I will simply note that the existence of unpredictable (generalized or specialized) semantic extensions, as for example with deed, or even the total lack of a product/result meaning in some nominals, do not invalidate the operation of general metonymies such as (26a) or (26d). Rules or patterns in a cognitive framework need not be completely regular or productive (cf. Langacker 1987 and others) and the existence of exceptions should not stop us from recognizing the existence of regularities.

5. Nominalizations in lexicography The final point I will examine is the possible application of such metonymies in lexicography. The problem of how exactly to treat polysemy, how to distinguish it from homonymy and how to organize polysemous entries is well recognized in lexicographic practice (cf. Landau 1989 and Burchfield 1987 for detailed discussions of the treatment of polysemy in different dictionaries and the advantages and disadvantages of different approaches). Maximize or minimize polysemy and at the same time reconcile this decision with limitations of space is an issue that every lexicographer faces in the planning stage. In practice, metonymy figures surprisingly little in dictionary entries, especially when compared to metaphor. There are seven major monolingual dictionaries of Modern Greek - two still incomplete. Of these, five mark a given sense as metaphorical but only three mark explicitly metonymy. Moreover, of all the types of metonymy only the "part-for-whole" metonymy, known as synechdoche, is marked and even that quite randomly. I do not have a general picture of the use of metonymy in English lexicography - there are

156

Kiki Nikiforidou

many more dictionaries of English. I have, however, looked more closely at the Longman Dictionary of Contemporary English (in the 1987 edition) and at the Cobuild English Language Dictionary (1993), neither of which uses metonymy in characterizing semantic relations. In fact, one of Cobuild's definitions for the word no, "A no is a person who has answered no to a question or who has voted against something", caused Fillmore (1989) to note that "this definition makes it seem that everybody on earth must be a no" and that "definitions like this make it clear that dictionaries need some standard way of speaking of conventional metonymy". It is true that while metonymy has its place in lexicographic theory (Geeraerts 1990, for example, includes it in the organization of dictionary entries), in practice it has not been exploited. Metonymy has certainly not been exploited in nominalizations, the polysemy of which has been also treated quite randomly. In all the Greek and English dictionaries that I looked at, the different meanings of derived nominals have been in general recognized and noted, but in a rather unsystematic and random fashion. To take a few examples, the Patakis Contemporary Dictionary of the Greek Language (1991) notes that the nominal apangelia 'recital' can refer to the action, the manner or the product but for apotamiefsi 'saving money' only the action meaning is mentioned; for abalarisma 'wrapping' it only mentions the product and manner meanings but not the action one. Cobuild includes the action and product meanings of acknowledgement, but leaves out manner, while for adherence, the state and fact meanings are mentioned but not the degree. In both Greek and English dictionaries, the fact meaning which, as noted, is completely general, is not mentioned in most entries for nominalizations with, however, sporadic exceptions. Finally, another aspect of the non-systematicity manifests itself in redundant specifications of polysemy: The different meanings of accept are, for example, all repeated in the entry for acceptance. To some extent, this inconsistency may reflect the variation in the degree of conventionality, conventionalization and predictability of such meanings and the lexicographer's decision to include or not include them accordingly. However, nominals such as apangelia and apotamiefsi do not seem to differ considerably in the extent to which they have conventional product meanings, while the omission of an action meaning for abalarisma when the manner interpretation is included is clearly inconsistent. Finally, the generality of the fact, manner and extent interpretations in nominalizations has gone rather unnoticed since they are neither all included nor all excluded as predictable; instead they appear randomly in some entries while lacking from others.

Nominalizations, metonymy and lexicographic practice

157

To substantiate the claim that the polysemy of nominalizations is fairly systematic and regular, I went through the letter A of the Patakis Dictionary and of the Cobuild English Language Dictionary. There were 510 action/state nominalizations in Greek and 187 in English whose morphological sources are synchronically present in the two languages. These numbers do not include agent and instrument nominalizations such as accuser or adaptor and place nominalizations such as Greek anapqftirio 'place of relaxing'. Of the 510 action/state nominalizations in Greek and the 187 in English all were completely regular with respect to the fact, manner and extent meanings, as predicted by the analysis given above. If manner, extent and fact are pragmatically triggered interpretations, we should expect them to be generally available and indeed they are. Every time that an action meaning was present, a manner interpretation was also available, and state nominalizations always had a degree interpretation as well. Fact, also expectedly, was completely regular as an available interpretation of both action and state nominals in the letter A. In the Greek letter A, 260 (out of the 510) were action or event nominalizations. Of these, 212 were also regular with respect to the product or result meanings. These were nominalizations which had product or result meanings and in which the lexical meaning of the product or the result was completely predictable from the meaning of the source verb, e.g., anakalipsi 'discovery', anadiorganosi 'reorganization', abalarisma 'wrapping' etc.. In Cobuild's A, there were 92 action or event nominals of which 67 were again regular in their product/result extensions. In these, the meaning of the product was predictable from the meaning of the corresponding verb frame, e.g., acceptance, acknowledgement, adaptation etc. This leaves 48 cases of irregularity for Greek and 25 for English. These include cases where 1) the meaning of the product is not predicted by the corresponding verb frame, for example English audition, 2) the action or state expressed by the nominal is not coextensive with that of the source verb or adjective, for example the Greek adjective asQenis means both 'weak' and 'sick' but the nominal asQenia means only 'sickness', 3) the predicted product meaning is present but there are also specialized, unpredictable extensions, for example English admission with the meaning 'amount of money you pay to go into a place' and finally, 4) either the action or the product meaning is missing, for example Greek etima which means 'demand' but not the act of demanding. These are often cases of suppletion, with another, mostly related, word taking on the missing function; for example, Greek anaynosi refers to

158

Kiki Nikiforidou

the act of reading while the related form anaynosma means 'that which you read'. One would have to look through the entire lexicon before final judgements. However, these preliminary numbers suggest the existence of highly regular recurring semantic patterns. Stating explicitly general metonymies such as those in (26) helps the lexicographer recognize both the range of meanings in nominalizations and their systematicity. For meanings such as manner, extent and fact, which are completely general and should be viewed as pragmatic interpretations, the decision of whether to include or omit them rests with the lexicographer. Whichever treatment is chosen should, however, reflect the fact that they are completely general; if they are omitted, they should be omitted consistently, if they are included they should be included in every entry. For the product and result meanings, the decision of whether they represent part of the conventionalized semantic content of the nominalization has to be made, as I suggested in section 4, on a case-by-case basis, taking into account factors such as the specificity and/or specialization (i.e. unpredictability) of meaning. The results I reported above from the letter A, give us an idea about the predictability of the product/result meanings but, except in clear cases of specialization or generalization, do not always help with deciding the extent to which such meanings are conventionalized. The fact remains that whether or not metonymy is explicitly mentioned in describing the possible meanings of nominalizations, the metonymic mappings stated in (26) provide a framework for a more consistent and systematic treatment of the polysemy of nominalizations.

6.

Conclusions

There are several features of nominalizations which do not follow from any general rule: First and foremost their form which, as noted in the beginning, is not always predictable. Secondly, their valency peculiarities with respect to the source verb or adjective (e.g., 'fond of something' but 'fondness for something'). Third, the control properties of the sentence which contains a given nomimalization. For example, in (33), with the nominalization realization, 'George' is obligatorily the one doing the realizing, but in (34) with discovery 'George' is not necessarily the discoverer. (33)

George enjoyed the realization that Mary was an Elvis fan.

(34)

George enjoyed the discovery that Mary was an Elvis fan.7

Nominalizations, metonymy and lexicographic practice

159

All of these are of course related to the semantics of nominalizations, but like many things about the semantics itself, do not seem to fall out of general principles. And yet, as Langacker (1991: 44) observes, "the picture offered by nominalization is not one of total chaos and idiosyncrasy, there are patterns to be discerned and characterized". The present paper has been an attempt in this direction, aiming to identify patterns and relate meanings in the semantic structure of nominalizations. Other questions have still to be answered before we have a complete picture of the nominalization semantics. We may, for one, investigate other areas of the vocabulary where we may expect to find similar polysemies. Deverbal Stative adjectives, for instance, not surprisingly appear to have also degree readings (e.g., "He has been repeatedly affected by the fluctuations in the stock market" (stative reading) vs. "He has been deeply affected by the loss of his wife" (degree reading)). Such filling-in of the picture would place the polysemy of nominalizations in a broader context rendering it more systematic and motivated. We may also ask why we find this particular grouping of meanings for action/state nominals as opposed to any other. Why is, for example, the locative or source meaning of nominalizations (as reported by Dressier 1986) cross-linguistically grouped with the agent form (so, we have diner like gardener), while manner, as reported here, is expressed by the action form? A partial answer has of course been given in section 3, where it was noted that all of the meanings associated with action/state nominalizations (except for fact for which we might predict a different expression in some other language) are one way or the other directly related to the source verb or adjective in the sense that they form part of the verbal (or adjectival) frame, filling in various aspects of it. In fact, these are the semantic participants which in a syntactic representation would either be sisters of the verb (when the product happens also to be the patient in the verbal frame) or, for any manner or degree modifiers, constituents of the VP. This is not so for the agent role (if expressed by the subject) or for any location modifiers which would be daughters of the sentence. A complete semantic analysis of nominalizations would have to examine these and possibly other questions. What I have tried to do here is to identify some of the factors that a semantic analysis of nominalizations would have to take into account, such as the general cross-lexemic meanings associated with action/state nominals and cognitive mechanisms like metonymy.

160

Kiki Nikiforidou

Acknowledgements An original version of this paper has profited from discussions with George Lakoff. I am also indebted to the two reviewers and to Claudia Brugman, Michele Emanatian and Eve Sweetser for their comments at the conference. All errors are solely my responsibility.

Notes 1. This of course does not mean that pure nominalizations never have the -ing form. Writing or recording, for example, may function both as -ing complements (e.g., "Writing letters always bores me") and as real nominalizations (e.g., "The writing of the letter took me an hour"). 2. At this point, I use the term "polysemy" rather loosely to indicate the existence of distinct, identifiable interpretations. In section 4,1 will explicitly discuss the extent to which such interpretations are conventionalized and can, therefore, be considered part of the semantics of the nominal. 3. This non-correlation between the semantics and the morphology led Jackendoff (1975) to postulate separate semantic and morphological rules which may, however, cross-cut in different nominals; thus discussion and congregation are both the output of the same morphological rule but the former belongs to the "act of verbing" semantic rule while the latter to the "group that verbs" rule. 4. In some respect, the result/effect meaning is related to the product meaning exemplified above in (3) and (6). Result, however, is a more general notion pertaining to both actions and events. At the same time, it is distinguishable from the product meaning for a given nominal. Consider, for example, (i), which has a product interpretation, alongside (ii), which expresses result or effect: (i)

Hand in your examinations to the supervisor.

(ii)

The examination was disastrous for most students.

5. Conventionality, as defined by Langacker, should not be confused with conventionalization as it pertains to the polysemy-vagueness distinction (and for this reason I have avoided the term "conventionalization" as much as possible in the polysemy-vagueness discussion). Conventionality for Langacker implies "sharedness" and repeated use, whereas different kinds of considerations (e.g., specificity, specialization and in general distinctness of meaning) are relevant to the semantic-pragmatic distinction. We may, for example, want to say that the product meaning of description is highly entrenched and shared (and in this sense conventional), but not necessarily part of the semantic specification of the nominal since it is completely predictable and not so (out of context) specific. 6. The question of predictability arises thus only with respect to the product/result

Nominalizations, metonymy and lexicographic practice

161

meanings, where we have cases of nominalizations with no product sense (although the source meaning would allow for one) or cases with unpredictable extensions (e.g., deed). Manner and fact, as noted before, are completely regular, while apparent exceptions to the degree interpretation can be explained by reference to the source semantics (cf. knowledge and availability). 7. The control properties of nominalizations are perhaps not unrelated to the assignment of a subjective or an objective role to the accompanying genitive, in which case a principled account of the former relies on a systematic account of the latter (for a cognitive account of the assignment properties of nominalizations see Taylor 1994).

References Aronoff, Mark 1976 Word Formation in Generative Grammar. Cambridge, Mass.: MIT Press. Broal, Michel 1900 Semantics. Translated by N. Cust. London: Heinemann. Burchfield, Robert (ed.) 1987 Studies in Lexicography. Oxford: Clarendon Press. Chomsky, Noam 1970 Remarks on nominalization. In R.A. Jakobs and P.S. Rosenbaum (eds.), Readings in English Tranformational Grammar, 184-221. Waltham, Mass.: Ginn. Collins Cobuild English Language Dictionary 1993 Editor-in-Chief John Sinclair. London: HarperCollins. Comrie, Bernard and Sandra Thompson 1987 Lexical nominalization. In T. Shopen (ed.), Language Typology and Syntactic Description, 349-98. Cambridge: Cambridge University Press. Cruse, D.A. 1986 Lexical Semantics. Cambridge Textbooks in Linguistics. Cambridge: Cambridge University Press. Dressier, Wolfgang U. 1986 Explanations in natural morphology, illustrated with comparatives and agentnoun formation. Linguistics 24, 519-48. Emanatian, Michele 1991 Grammatical Polysemy: The Systematicity of Multiple Meanings in Grammar. Ph.D. Dissertation. University of California, Berkeley. Fauconnier, Gilles 1985 Mental Spaces. A Bradford Book. Cambridge, Mass.: MIT Press. Fillmore, Charles 1971 Verbs of judging: an exercise in semantic description. In C.J. Fillmore and D.T. Langendoen (eds.), Studies in Linguistic Semantics, 273-89. New York: Holt, Rinehart and Winston. 1985 Frames and the semantics of understanding. Quaderni di Semantica V1.2, 222-54.

162

Kiki Nikiforidou

1989 Two dictionaries. International Journal of Lexicography 2.1, 57-83. Fillmore, Charles and Paul Kay 1987 The goals of construction grammar. Technical Report 50. Cognitive Science Program. University of California, Berkeley. Fillmore, Charles, Paul Kay and Mary Catherine O'Connor 1988 Regularity and idiomaticity in grammatical constructions: the case of let alone. Language 64.2, 501-38. Geeraerts, Dirk 1990 The lexicographical treatment of prototypical polysemy. In S.L. Tsohatzidis (ed.), Meanings and Prototypes: Studies in Linguistic Categorization, 195210. London and New York: Routledge. 1993 Vagueness's puzzles, polysemy's vagaries. Cognitive Linguistics 4.3, 223-72. Givon, Talmy 1970 The Si-Luyana Language: A Preliminary Description. Institute for Social Research, Communication 6. Lusaka, University of Zambia. Horn, Lawrence 1985 Metalinguistic negation and pragmatic ambiguity. Language 61.1, 121-174. Jackendoff, Ray 1975 Morphological and semantic regularities in the Lexicon. Language 51, 63971. Lakoff, George 1970 Irregularity in Syntax. New York: Holt. 1987 Women, Fire, and Dangerous Things: What Categories Reveal about the Mind. Chicago and London: The University of Chicago Press. Lakoff, George and Mark Johnson 1980 Metaphors we Live by. Chicago and London: The University of Chicago Press. Landau, Stephen 1989 Dictionaries: The Art and Craft of Lexicography. Cambridge: Cambridge University Press. Langacker, Ronald 1987 Foundations of Cognitive Grammar. Vol.1: Theoretical Prerequisites. Stanford, CA: Stanford University Press. 1991 Foundations of Cognitive Grammar. Vol. 2: Descriptive Application. Stanford, CA: Stanford University Press. Levi, Judith 1978 The Syntax and Semantics of Complex Nominals. New York: Holt. Longman Dictionary of Contemporary English (new edition) 1987 Editorial director Delia Summers. Lyons, John 1977/1983 Semantics. Cambridge: Cambridge University Press. McCawley, James 1988 The Syntactic Phenomena of English. Vol. 2. Chicago and London: The University of Chicago Press. Nikiforidou, Kiki and Eve Sweetser 1989 Diachronie regularity and irregularity: structural parallels between semantic

Nominalizations, metonymy and lexicographic practice

163

and phonological change. Berkeley Cognitive Science Report Series No 60. Institute of Cognitive Studies, University of California, Berkeley. Patakis Contemporary Dictionary of the Greek Language 1991 Athens: Patakis Publications. Selkirk, Elisabeth 1982 The Syntax of Words. Cambridge, Mass.: MIT Press. Sweetser, Eve 1986 Polysemy vs. abstraction: mutually exclusive or complementary? In K. Nikiforidou and P. VanClay (eds.), Proceedings of the 12th Annual Meeting of the Berkeley Linguistics Society, 528-38. Berkeley Linguistics Society: University of California, Berkeley. 1990 From Etymology to Pragmatics: Metaphorical and Cultural Aspects of Semantic Structure. Cambridge: Cambridge University Press. Taylor, John 1994 "Subjective" and "objective" readings of possessor nominals. Cognitive Linguistics 5.3, 201-42. Ullmann, Stephen 1979 Semantics: An Introduction to the Science of Meaning. New York: Barnes and Noble. Vendler, Zeno 1967 Linguistics in Philosophy. Ithaca, NY: Cornell University Press. Wilkins, David 1981 Towards a Theory of Semantic Change. Doctoral Dissertation. Australian National University.

Langacker semantics for three Coeur d'Alene prefixes glossed as On' Roy H. Ogawa and Gary B. Palmer

1.

Introduction

The Coeur d'Alene, traditionally known to themselves as the Sn'chitsu'umsh, the 'discovered people', are a North American Amerind1 group who derive from an area in western Washington state and northern Idaho. They are speakers of a language from the interior subgroup of languages from the Salish family. Johnson estimates that there might have been less than 10 proficient speakers of the language in 1975, all over the age of 50 and all bilingual (1975: 1). This paper is based largely on published sources.

Coeur d'Alene orthography Before discussing the focus of the paper, we present some information about the orthography and the morphology of Coeur d'Alene. The orthography for Coeur d'Alene is slightly problematic because Reichard (1938), Johnson (1975), and Nicodemus (1975) each use a different system. Reichard's orthography and phonology was analyzed by Sloat (1968, 1980), who suggested modifications and these modifications were accepted by Johnson (1975) except for the substitution of Ό' for the V of Reichard and of Sloat. Nicodemus, whose goal was to use only symbols available on the standard typewriter, uses a "practical" orthography (vol. II, 1975: 1). The orthographies and their relationships are given in Appendix 1. In the examples cited below, transcriptions given in parentheses are from the original source, which is indexed below the example.

Coeur d'Alene morphology Coeur d'Alene words are usually analyzable into prefixes, a root or stem, and suffixes. For example2

166 (1)

Roy H. Ogawa and Gary B. Palmer cfaq'walqw (chiaq'walqw) Ic - \lleq'w - ilqw/3 on - remove.skin - stick-like.object vt. 'he stripped bark off a tree'. NI:47

has a prefix Ic-l followed by a root /leq'w/, conventionally indicated with a prefix '\T in Salishan linguistic studies, and a suffix /-ilqw/. Sometimes the root is reduplicated to form a stem indicating intensification, plurality, diminution, or spatial distribution, as in (2)

cyeryel'xwal'qsQn Ic - REDUP Nyil'x w -flqw- san/ on - REDUP cover, with.fabric - stick-like.object - leg;foot 'leggings' J: 241, R: §498

In some cases what appears to be a suffix is treated as a stem. For example, the suffix /-ini?/ functions as a stem in

(3)

fine? (t'inä'ä) /t-Mini?/ on - ear Outer ear' R: §473, PN: 356

Finally, there are a few words which contain two stems, as in (4)

cte?elwlwlimn (chte'eiwlwlimn) Ic - REDUP Nti? - al - \lwlwlim - an/ on - REDUP hit,pound - CONJ - metal - NOM n. 'anvil'

NI:54 and these are joined by a conjunctive morpheme /-at/. Coeur d'Alene roots are most often verbal, nominal, and Stative (Reichard, 1939). Kinkade (1983) suggests that the stems should be treated as predicative and that the lexical categories of noun and of verb do not exist. A prefix, /s-/, often nominalizes a verbal, while a suffix, /-an/, usually refers to a nominal which is involved with the verbal process described by the stem. Examples are:

Langacker semantics for three Coeur d'Alene prefixes

(5)

stgwepgwepicen'us (s-t-gwcpgwep-icen'-us) /s -1 - REDUP gwep - icen' - us/ NOM - on - REDUP hairy - back; ridge - eye;face 'eyebrows' PN:354

(6)

ctel'qmindn (c-tel'q-min-en) Ic - tel'q - man - 9n/ on - touch.with.foot - used.for - NOM 'stirrups' R:§452

167

Verb inflections relate to case, person, number, tense, aspect, mood, and modality. They are tabularized as paradigms by Reichard (1938: 686-694). One particular case deserves mention here: the S.s.subj.INTRANS.PAST case is often unmarked. Verb inflections are not well understood in Coeur d'Alene. Similar inflections exist in Thompson (Thompson and Thompson 1992). Morphological analysis requires application of morpho-phonological rules as well as analysis of verb inflections. Morpho-phonological rules are given in Johnson (1975), which further develops the work of Reichard (1938: 543568). One such rule is a rule of assimilation of/t/: If a /t/ or Id stands before a bilabial, a palatal (except for /y/ or /y'/), a velar or a faucal, it retains its identity [...] If, however, /t/ or Id stands before a dental (/d/, /t/, /t'/, /n/, /nY), a sibilant (/s/, Id, /c'/, /s/, /}/, /c/, /c'/, /y/), or a lateral (/I/, /!/, /1V), it becomes assimilated to that sound, and if the accent is on the syllable immediately following, the assimilation is evident in the echo vowel [...] If, on the other hand, the accent is on a syllable farther removed, there is no vocalic indication of the assimilation. (Reichard 1938: 547)4

This rule applies to the final t in /cit-/ but it does not apply to the prefix /t-/ (Reichard 1938: 548). Thus /cit-t-/ > /ci:t-/ if the accent is on the next syllable and /cit-t-/ > /cit-/ if not. The Family of Spatio-Relational Prefixes Some prefixes indicate a spatial relation between the concept indicated by the stem and a thing or place indicated by some combination of nominal suffixes. The latter often indicates a human anatomical part. For example:

168

Roy H. Ogawa and Gary B. Palmer

(7)

ciaq'taq Yne? (chiaq 'iaq 'ine') /c-REDUP\laq'-ini?/ on - REDUP broad - ear;over vt. 'he has big ears' NI:47

The set of spatial relational consists of: Ic-l On', /t-/ On, part of, Than-/ 'in, affect', /cs-/ 'behind', /i-/ 'a general preposition [which usually has a spatiorelational meaning]', /u-/ 'part of; exactly', /ein-/ 'under', /mil'-/ 'near', /ni?-/ 'amidst', /etc-/ 'hither', /cit-/ On', /tu?-/ Over [...] in order to', /citus-/ 'so far hither'. Our group has begun a study of these spatio-relationals, starting with /han-/ (Palmer 1990), /ni?-/ (Occhi, Palmer and Ogawa 1993), and /ein-/ (Kendrick-Murdock in preparation). This paper reports on the part of the set involving the concept of contact: /cit-/, /t-/, and Ic-l, all of which are glossed by workers on the language with some usage of the English word On'. These are analyzed as a group because their semantics needs to be delineated clearly enough to determine the reasons for the selection of one over another in particular usages and because they also operate in combinations, in /cit-t-/ and /t-c-/. For /cit-/ Reichard gives "on a surface or something broader than subject, above, over" (1938: 595) while Johnson, whose main focus was phonology, just glosses it as "on" (1975: 34). For example: "/cetpu?ui-itkwe?/ 'bubbling on surface of water' (/put/ 'foam, bubble'; /-itkwe?/ 'water')" (Reichard 1938: 595). For /t-/ Reichard gives "on and a part of" (1938: 594) while Johnson gives "on, part of" (1975: 34). For example: "/t-gwcxw-9nts/ 'he hung them up on it', where /gwexw/ is a root meaning 'plural objects hang'" (Reichard 1938: 594). Of/t-/, Reichard says in addition, These prefixes often mean the opposite of the literal idea they indicate. Instead of meaning 'it is on', the combination of prefix, stem and suffix may mean, e.g., Off of or Off from': /citus-t-kwin-9w'es-3nts/ 'he went and took it off of it' (/citus-/ 'so far hither'; /kwin/ root 'take hold of one'; /-iw'es/ 'together, where it made contact'; /-an/ tr.; Ms/ 3-3 compl.) (1938: 595).

For Ic-l Reichard gives "on, attached to but not part of, at a point" (1938: 595) while Johnson gives "on" (1975: 34) and Nicodemus gives "adv. on; on the surface" (vol. I, 1975: 33). For example: "/ε-δ-§έΙ'-θ9η/ 'it projects out from mouth (like a tooth)', (ec-c- > εδ-); cc- 'made so'; /§et'/ One projects; /-COT/ 'mouth'." (Reichard 1938: 595).

Langacker semantics for three Coeur d 'Alene prefixes

169

These prefixes may combine. Reichard commented on one such combination, /tc-/, /tc-/ (/t-c-/ ?) This prefix is used with numerals in counting persons; it means "astraddle" or "straddling object"; or that the object on top is larger than the one it is on. /tc'-9m-£s-alqw/ 'he sat down astride a log' (/tc/ - vowel stem initial > /tc'/; /-am/ weak form of/-em/, sit; /-ε§/ weak form of /-is"/ be in the act of; /-alqw/ long object). /tc'-upan-tct-e?st/ 10 round objects which have covering (orange, berry), (/upan/ 10; /-ct/ finger; /-i?st/ surface of sphere.) (1938: 595).

The glosses given above by Reichard, Johnson, and Nicodemus do not describe all instances of usage correctly. For example, the ί in tgwexwdnts does not seem to be an example of On and a part of. (8)

tgwexw9nts /t-g w ex w -9nts/ on - plural objects hang - 3.s.subj.3.s.obj.TRANS.PAST 'he hung them up on it' R: §393

Similarly, the c in chdlamalqw is not an example of On, attached to, but not part of, at a point'. (9)

cdlamalqw (chdlamalqw) /c - delim - ilqw/ on - galloped - stick-like object n. 'train' (lit. he galloped on the log) NI:36.

And, the cit in Nicodemus' chat(e'line'n is not what we would normally mean by On a surface or something broader than subject, above, over'. (10)

catferine?n (chat(e'line'n) /cit-NTel'-ini?-9n on - cover - ear;over - NOM n. 'cover, something that conceals or disguises' NT. 34

170

Roy H. Ogawa and Gary B. Palmer

Our Goal Our main goal is to analyze the semantics of the prefixes using the concept of complex category as described and developed by Langacker (1991). In this framework a category of meaning is represented as a network of Schemas. Each schema represents a subcategory of usage, some of which have multiple members, and some of which have only one. We have devised a methodology for determining the semantics of Coeur d'Alene morphemes. An entirely secondary effect of the work is that we are able to give better glosses for the morphemes. However we must beware of thinking that these glosses define the morphemes because we do not arrive at intensional definitions by descriptive gloss but at extensional5 definitions presented as particular networks of Schemas. Langacker warns, speaking of the Cora u/a contrast, that it is most improbable that a single abstract meaning can be found that would be schematic for all of the specific values attested for u and a, and even if one were found it would also be schematic for indefinitely many values that M and a happen not to have. A linguistic description that limited itself to stating a single meaning for u and a - whether an all-subsuming schema or the prototypical sense - would therefore fail to provide an explicit account of the facts of the language, in particular the range of conventionally established senses and usages characteristic of these morphemes (1991: 55).

Section 2 includes a discussion of our methodology; section 3 discusses the analysis of the prefixes. In section 4 we summarize our findings. In appendices we include information about Coeur d'Alene orthography and a list of the words that we used as data for our analysis.

2.

Method

Our problem is to understand spatial prefixes. Our method is based on concepts developed by Langacker (1991: 9, 266) who defines semantics of lexical or grammatical items as complex categories, which are networks of Schemas. In the case of a spatio-relational morpheme each node of the network consists of a subcategory of words used in similar ways and thus corresponds to a sense of the morpheme. The node is identified and symbolized by a schema of relationship which profiles characteristics of the relation itself and the two participants in the relation: the trajector and the landmark. The senses of the morpheme are related in the network through extension and elaboration to build meaning-complexes.

Langacker semantics for three Coeur d'Alene prefixes

171

Our data consists of words with glosses given by bilingual native speakers. These came from Johnson (1975), Nicodemus (2 vols. 1975), Palmer and Nicodemus (1985), and Reichard (1938). Each word has two glosses while some have three. One is the gloss for the word as a whole as provided by a bilingual speaker and one is a gloss constructed from glosses of the morphemes that constitute the word. Some of the words also have 'literal' glosses given by the native speakers. What we mean by this is illustrated by the entry "chtelqhwa'lqw n. music, flute, cembalo, harpischord (lit. Holes bored on a hollow stick, log, etc.)" from Nicodemus (NI: 1975). The parenthesized phrase contains the elements of a morphemic analysis, which we find to be: /c- REDUP Nlex w - ilqw/ On REDUP perforate - stick-like.object'. If the word is given such a so-called literal gloss by Nicodemus then that is the third gloss. The whole-word gloss may sometimes be the same as the constructed gloss but at other times they differ. For an example where the whole-word gloss is the same as the constructed gloss, consider the word cliji?snts 'He stabbed a spherical object'. The morphemic analysis of the word is /c - \llej - i?st - ants/ on- stab - spherical.object - 3.s.sub.3.s.obj.TRANS.PAST. The constructed gloss is 'he/she/it stabbed onto a.spherical.object he/she/it'. For an example in which the whole-word gloss and the constructed gloss differ, look at the word ct'lt'lqicn' 'bike, bicycle', analyzable as /c - REDUP Nt'elq - ican'/ On - touch.with.foot,both - back;ridge'. The constructed gloss is therefore 'touch.with.foot,both, covering.on backjridge'. This might describe the operation of the bicycle, with the foot on the pedal (back; ridge), but could only refer metonymically to the bicycle itself. These glosses are links to the semantics of the morphemes and words but since they are only glosses and not full meanings acquired by a person in the natural way (by normal acquisition and usage), we have to treat them as somewhat schematic for the semantics. The constructed gloss is therefore also schematic because the parts of the word name abstracted entities. The schema, built up from these parts, constitutes the meaning of the constructed gloss. The constructed-gloss schema, or constructed schema for short, is related in some way to the whole-word meaning. In the case of '3.s.subject stabbed a.spherical.object 3.s.object' the constructed gloss is equivalent to the whole-word meaning, while with 'bicycle' above, the bicycle is metonymically connected to the situation described by the constructed schema. With the word cdlamalqw 'train' which morphologizes as /c - Ndelim - ilqw/ On - gallops - stick-like.object', the train is metaphorically galloping across the railroad ties.

172

Roy H. Ogawa and Gary B. Palmer

The schema for a spatial relation contains three parts: a focal or foreground entity called the trajector, a contextual or background entity called the landmark, and the relation itself between the trajector and the landmark. (Langacker 1991: 9) For example, in the word for bicycle discussed above, we postulate that the trajector is the reduplicated process of touching the pedal with the foot. The landmark is the pedal. The relation is the contact of the foot covering over the pedal. We postulate in this paper that the trajector in a Coeur d'Alene word is described by a constituent containing the stem. Since Coeur d'Alene word stems are verbal (processual), nominal, or stative, these are the major classes of trajectors. The landmark is either explicitly referred to by a suffix constituent or is tacit and can be determined by examining the whole-word gloss. The relation between the trajector and the landmark is described by the prefix. In the word for train, the trajector is the process 'galloped'; the landmark consists of the railroad ties under the tracks; and the relation, we postulate to be contact or near contact with the ties since the tracks are between the train wheels and the ties. In the word chlek'wnts 'he broiled it (salmon)' which morphologizes as /c - lek'w - ants/ On - pierce - 3.s.sub.3.s.obj.TRANS.PAST', we have the processual 'pierce' as trajector, the relation of/c-/ On', with the tacit landmark being a surface which pertains to the salmon. After collecting schematic definitions for each word containing the prefix being studied, we categorize them. (Here is where our English speaking cultural intuitions are used and may lead us astray.) Schemas are categorized for similarity according to our intuitions about spatial contact, each category having a representative schema. These categories and Schemas are easily seen to be related by extension and elaboration. The relation of extension "implies some conflict in specifications between the basic and extended values" (Langacker 1991: 266f). The relation of elaboration "amounts to one of specialization [...] characterized with finer precision and detail." (ibid.: 267). After the relations are determined, we have a network of Schemas which makes up the semantics of the morpheme. At this point, if we wish, we may write descriptions of the Schemas in the networks to serve as some sort of descriptive summary of the network, but these descriptions are not intended as glosses. To summarize the method: we collected a group of words that include a common spatial prefix and morphologized them. The constructed gloss was then interpreted in the whole-word gloss of each word. We then interpreted the relation of contact in each word and identified the trajector and landmark in the interpretation. (Because the semantics of the relational prefix is be-

Langacker semantics for three Coeur d'Alene prefixes

173

ing studied, it's nominal gloss from past investigators was withheld while the glosses of the other word parts were taken at face value.) The types of contact were then subcategorized, with Schemas constructed for each subcategory, and then the relations between these Schemas were delineated. The relations are those of extension and elaboration. This network of senses in the categorial relationships make up the semantics of the morpheme. We might be criticized at this point for lacking native speaker knowledge of the semantics but we proceed anyway, knowing that native speakers of any language normally do lack conscious understanding of how they think. Of course, there is no substitute for a linguist who is also a native speaker. But our results do show that there is value in using this method.

3. Analysis This section presents and explains the small set of Schemas that describes each of the prefixes. Words which include combinations of the prefixes, namely those including /t-c-/ and those including /cit-t-/, are discussed separately. The English word On', used in all of the glosses for the prefixes, involves, in its most general meaning, some sort of contact, with physical contact being the prototype. We do find an interpretation of contact for all words whose morphology we understand, though in some instances the contact is metaphorical. We discuss the prefixes in the order: /cit-/, /t-/, /cit-t-/, /c-/, /t-c-/. 3.1. /cit-/. We found 51 words with this prefix. They may be divided into two categories: one involving contact with a broad surface and one involving a kind of contact 'all over', even if the surface is not broad. 3.1.1. The first category of words involves contact with explicit broad surfaces or tacit broad surfaces. A typical word of this category involves contact with the surface of a table. (11)

catq 'ey 'minan (chatq 'e 'yminn) /cit - Nq'iy' - man - an/ on - writing;design;graphics - that.which.is.used.for - NOM n. 'writing desk' NI:34

174

Roy H. Ogawa and Gary B. Palmer

Since all the words involve contact with broad surfaces, there is only one category and one schema for this subcategory. The schema consists of an abstract trajector in contact with a broad surface landmark. We see no evidence of clear subcategorization of the trajectories. We can understand what the Coeur d'Alene mean by a broad surface by subclassifying the landmarks of /cit-/. We must be careful in using the English phrase 'broad surface'. We do not mean that any thing recognized by an English speaker to be a 'broad surface' may be used as a landmark of contact for /cit-/. We mean that the phrase summarizes the list of those things which may be landmarks of contact for /cit-/. The first subcategory subsumes partial surfaces of the human body. The second subcategory consists of supporting surfaces. The third is land; the fourth is the surface of water. The fifth is clothing on the body and the sixth is a general unspecified surface. We give examples of the subcategories below. 3.1.1.1. ample, (12)

This subcategory is of partial surfaces of the human body. For exC9ncac'a?ril'kwe?us /can - cit - Nc'isr - il'kwi? - us - is/ 1.s - on - hurt;ill;cold - forehead - eye;face - POSS 'my forehead hurts' J: 234, F: 57, R: §484

The constructed gloss is 'my hurt; ill; cold on forehead of face'. The profiled trajector is the process of 'hurt; ill; cold'; the profiled landmark is 'forehead of face'. In this term, contact is metaphorical. 3.1.1.2. The second subcategory of words have covering surfaces for the landmark. In several words for example, the landmark surface is a cover such as a roof: (13)

catxepilxw9n (chatqhepilkhwn) /cit - Nxep - ilxw - an/ on - pile.things.flat - house - NOM n. 'shingle, roofing, thatch' NI:34

Here the constructed gloss is 'something which is piled flat on the house'; the profiled trajector is the process of piling things flat; the profiled landmark is

Langacker semantics for three Coeur d'Alene prefixes

175

the (roof of the) house. The house is metonymically used for its roof.

3.1.1.3. Other words form a subcategory and have landmarks which are supporting surfaces of a table, a chair, or a bed. (14)

catq'ey'min9n (chatq'e'yminn) /cit - Nq'iy' - man - 9n/ on - writing;design;graphics - that.which.is.used.for - NOM

n. 'writing desk' NL34 The profiled trajector is the verb 'to write'; the profiled landmark is specified only through reversal as 'that which is used for writing'. 3.1.1.4.

(15)

Another subcategory of landmark is the land.

c 'dncetcigwiS (ch 'nchetchigwish) [sic] /can - cit - \lcigw - as/ l.s.subj - on - extend.across - VB vi. Ί traversed the plain' NL59

The profiled trajector is the verb 'to extend across' and the profiled landmark is tacitly the plain. 3.1.1.5. A further subcategory of landmark is the surface of water. (16)

catpo?satkwe? (chatpo'satkwe') /cit - Npu?us - itkwe?/ on - bubble - water 'to bubble (lit. it (water) gave off bubbles)' NI:34

The profiled trajector is the verb 'to bubble' and the profiled landmark is the water. Notice that we must treat the surface of the water as tacitly understood; but also notice there is no physical reason why the water as a whole isn't treated as giving off bubbles rather than the surface of the water giving off the bubbles. It is the prefix /cit-/ that evokes the schema of a surface. 3.1.1.6. is:

A one-word subcategory in which the landmark consists of clothing

176

(17)

Roy H. Ogawa and Gary B. Palmer

catqältcal'qs (chatqaltcha'lqs) /cit - Nqiltc - il'qs/ on - flesh;meat - clothes n. 'coat, cloak, overcoat (lit. outer garment)' MI: 34

The trajector is 'flesh; meat' and the landmark is the 'coat.clothing'. 3.1.1.7. The last subcategory consists of words that have tacit surfaces. A revealing example of such a word is: (18)

caiyaq'ants (caiyaq'ants) /cit - Nyeq' - ants/ on - sharpen;abrade - 3.s.subj.3.s.obj.TRANS.PAST 'he filed the surface' J: 241, F: 95

Here the constructed gloss is '3.s abraded an unspecified surface'. The profiled trajector is the verb 'to abrade' and the profiled landmark is a nominalized thing being filed. That it is a surface and not a point or an edge can only be lexicalized within the prefix /cit-/. This shows that the sense SURFACE is already part of/cit-/ and the gloss On a surface', where the surface is either specified by a suffix constituent or is tacit, is correct. This is Reichard's gloss for cit. 3.1.2. The second category of words which include /cit-/ involves the configuration /cit - STEM - ini?/ indicating that the stem is in a kind of contact 'all over'. The suffix /-ini?/ is always the landmark in this complex. A typical example is the word:

(19)

cetgwelpine? /cit - Ngwelp - ini?/ on - uncontrolled.burn - ear;over 'he burned all over' R: §541

The profile for the spatial relation consists of the trajector process of burning, the landmark Over', and contact between the burning and the Over': The frame [cit- STEM -ini?] is extended to mean something like the English 'all

Langacker semantics for three Coeur d'Alene prefixes | L Contact with Broad Surface |

111

12. Contact 'all over' |

Figure 1. Schemes of prefix /cit-/.

over'. Kinkade has mentioned a suggestion by Larry Thompson that the English 'up to his ears in' might suggest the connection between 'ear' and 'all over' (Kinkade 1994). Another word is: (20)

cceiie]'ene?ncui (chchet(e'lene'ntsut) /c - cit - M Tel' - ini? - an - cut/ we - on - obstruct,block,fence,curtain - ear;over - TRANS.PAST REFL vt. 'we covered ourselves'. MI: 35

The profiled relation has a trajector of 'to cover'. Another example is: (21)

ecetc'lxwine? (echetch'lkhwine') /e-cit-Nc'elx w -ini?/ PREP - on - concave - ear;over n. 'covered wagon (lit. a wagon with an arched cover, tent, over it)' MI: 65

The profiled trajector is the concavity; the profiled landmark is the cover.6 We interpret it here as "cover over", meaning 'covering over the body of the wagon'. 3.1.3. The complete category of Schemas and their relationships may be visualized in Figure 1. The fact that the semantics for this prefix involves two different senses of meaning is visible as the two separate networks. 3.2. /t-/ We found 94 words which used this prefix. Many words in this class refer to parts of the body. We will discuss the body part names after discussing the others since they fall naturally into the categories with the others. The other words fall into four groups of related meanings: an attached contact, metaphorical attached contact, specification of a state of beipg or quality, and trajectories onto or off from. The specification of state of being

178

Roy H. Ogawa and Gary B. Palmer

or quality is nothing more than a metaphorical attachment of the state or quality to something. This sense of meaning of being a state or quality of something would appear to have very common use and this is confirmed in that the number of /t-/ words is very high compared to all others except for the number of Than-/ 'in' words. 3.2.1. A prototypical kind of attached contact is that of hair on the head as exemplified by: (22)

tgwapqdn (tgwapqn) li - \lgwep - qan/ on - hairy - head;top v. 'H/s has thick hair.' NL259

Here the trajector hair is attached to the landmark head. Another kind of attachment is shown by (23)

tcoqcequs (ttsoqtsequs) /t - REDUP Nceqw - us/ on - REDUP pink - eye;face adj. 'blood-shot, he has bloodshot eyes.' NL277

In this case, the pink blood vessels are attached to the eyes. Another kind of attachment is more a 'growing out of as shown by the two words: (24)

(25)

tSt'iw'es (tsht'iw'es) li - NSet' - iw'es/ on - to.extend,to.project - between;waist n. 'kettle (lit. a vessel with a projection from its body.)' NL277 tset'ect (tshet'echt) /t - \lset' - ict/ on - to.extend.project - whole hand n. 'branch (lit. projection from a tree.)' NL277

3.2.2. A group of words show a metaphorical attachment to the landmark. For example:

Langacker semantics for three Coeur d'Alene prefixes

(26)

179

teen'can'am (tche'nch'nm) /t - REDUP Ncen' - 9m/ on - REDUP hold.one.object - MID v. 'cling (lit. he held fast to something physically or emotionally.)' NL255

Here emotion is metaphorically being held - attached to. The next word reveals, if we are correct about /t-/ words involving attachment, that the Coeur d'Alene think of taking care of children as being attached to them. (27)

tst'ilt (tsht'ilt) /t - N set' - ilt/ on - take.care.of - offspring vi. 'to babysit (lit. he cared for his own or others' children.)' NL277

A group of words involve an interesting type of attachment to a 'handle' as shown by: (28)

tce?wple?nts (tche(wple'nts) /t - Ne'e?*-ipli?-ants/ on - pray - handle;attachment - 3.s.subj.3.s.obj.TRANS.PAST v. 'bless, pray over' ΝΓ255

The concept seems to be that prayer involves making an attachment to a special place or thing. This shows up in a group of other words including 'to sing hymns' and 'to pronounce a judgement'. This group suggests the study of the suffix ipli? which appears to indicate a magical place or thing but also has mundane meanings such as where a fish line is attached. Among this group is an interesting pair, the words for 'to attract' and 'to detract'. They use the same root and the same schema, but indicate the different meanings through inflection. (29)

tcekwnts (ttsekwnts) /t - Ncekw - ants/ on - pull - 3.s.subj.3.s.obj.TRANS.PAST vt. 'detract (lit. He drew it off.)' NL277

180

(30)

Roy H. Ogawa and Gary B. Palmer

tcekwukwum (ttsekukum) /t - REDUP Mcekw - am/ on - REDUP pull - 3.s.subj.PASS.PAST vi. 'attract (lit. He was drawn.)' Nl:277

3.2.3. The next group of words are closely related to the previous group. They involve the landmark having a quality or state of being which is metaphorically attached to the object. For example: (31)

iiuxoxw/: (ttuqhoqhwiii) k -\texw REDUP - iii/ on - one.dies.stops REDUP - INTENS vi. 'He was found dead.' NL277

In the next word, the quality of being high is attached to the nose. (32)

tgwesi?qs (tgwesi'qs) /t - Ng w is - 9?qs/ on - ascend;high - nose;beak v. 'H/s has a high nose' NI:259

An interesting group of words each involve a group of like objects. A numerical count of the group is a property of the group and is metaphorically attached to the group. (33)

tce?Hselps (ic/ie'fjseJps) /t - \lci?lis - ilps/ on - three - throat,mane.(counting.classifier) n. 'three pigs, three grizzly bears' NL255

(34)

the?in'9m (the'i'nm) /t - Nhe?in'm/ on - eight n. 'eight persons' NI:259

Langacker semantics for three Coeur d'Alene prefixes

181

Later we will see that this includes counts of ten but they also involve the prefix Ic-l and are discussed in the group of /t-c-/ words. 3.2.4. Finally, we have the group of words which involve trajectories to or from something. Here are two words which indicate a trajectory on to something. (35)

tpoxwqents (tpoxwqents) /t - Npux w - qan - ants/ on - blow - head;top - 3.s.subj.3.s.obj.TRANS.PAST 'he blew on her head' J: 237, R: §208

(36)

tco?qintem (tcoYqintam) /t - \lcuw' - qan - antam/ on - punch - head;top - 3.s.subj.PASS.PAST 'he was hit on the head' J: 233, S: 56

For a trajectory off of something, we have: (37)

tccminpixwnt9m (tchchminpikhwntm) /t - Ncec - man - p - ilxw - antam/ on - throw - that.which.is.used.for - unwillingly - house 3.s.subj.PASS.PAST v. 'degrade, eject (lit. he was thrown out of the house, he was deprived of dignity, he was ousted.)' NI: 255

3.2.5. Body Parts. These words use several of the spatial-relational prefixes, especially when locating a body part relative to others. The prefix /t-/ is most common. In many of them the trajector stem has been glossed as either 'right' or 'left'. This sense of /t-/ is the meaning that a body part is attached on the right side or on the left side. In others the body part is attached to another body part. (38)

stc'ihict /s-t-Nc'ih-ict/ NOM - on - right, side - whole hand 'right arm' PN: 357

182

Roy H. Ogawa and Gary B. Palmer

1. Attached contact

2. Metaphorically attached

4. Movement onto or off from

1

3. Property or quality Figure 2. Schemas of prefix ft-/.

(39)

stc'ikwi?ini? /s -1 - \lc'ikwi? - ini?/ NOM - on - left.side - ear;over 'left ear' PN: 357

Some body part words do not denote a direction; they denote a location and an attachment.

(40)

fine? (t'inä'ä) /t - Nini?/ on - ear;over Outer ear' R: §473, PN: 356

Here the trajector is the outer ear (normally a suffix but here a stem) and the landmark is tacitly the head. The inner ear is denoted by Than - Nini?/ 'in ear.' A common trajector is /c'em/ glossed as 'surface', /t - c'em -/ describes 'the surface attached to':

(41)

stc'emgwdl /s -1 - Mc'em - gw9l/ NOM - on - surface - abdomen,stomach 'abdomen' PN: 356

3.2.6. The complete category of Schemas and their relationships may be visualized in Figure 2. Lines going across indicate a relation of extension; lines going down indicate a relation of elaboration. The fact that the semantics for this prefix

Langacker semantics for three Coeur d'Alene prefixes

183

involves two different senses of meaning is visible as the two separate networks, one of which has only a single node. 3.3. /cit-t-/ This combination of prefixes was found in four words. Morphophonological rules shifted the prefixes to /ci:t-/ (cf. Introduction). The first word is: (42)

ceitc'uw /cit -1 - Nc'uw/ on - on - gone 'it is gone from the table' J: 235, R: §133

The use of each prefix here is coherent with what we found for /cit-/ and /t-/ separately. The action is to be gone via a trajectory Off from' and the action takes place on a broad surface, the table. In the following case the 'attached' sense of /t-/ applies. (43)

ceitjements /cit -1 - Njem - ants/ on - on - pin - 3.s.subject.3.s.object.TRANS.PAST 'she pinned it to it' J:236,R: §133

The action is to pin something to a surface and the action takes place on a tacit unspecified surface. In the following case, the action of eating takes food off from the table and it is done from on a surface. (44)

ceit'iinan (cheet'iinn) /cit -1 - Niten - an/ on - on - eat - NOM n. 'table, buffet, sideboard (lit. that on which one eats)' NI:36

In the following word, the /cit-/ type of contact is clearly with the surface of the table but the /t-/ type of contact is harder to identify. It appears that a hand or something else is attached to the cloth. (45)

ceit'ip'eman (cheet'ip'emn) /cit -1 - Nip' - 9m - an/

184

Roy H. Ogawa and Gary B. Palmer

on - on - wipe - MID - NOM n. 'mop, wiping cloth for top (of floor, table, etc.)' NI:36 3.4. Ic-l We found 42 words which include Ic-l. They all involve contact of some sort, some metaphorical, but are not all classifiable as On, attached to, but not a part of as suggested by Reichard's gloss. 3.4.1. The first subcategory of words which include Ic-l involve a repetitive or scattered contact. This involves the notion that the contact may occur distributed over time or over space-time. The suffixes /-ilqw/ 'stick-like object', /-ic'e?/ 'blanket, skin, hide', or /-icanY 'back; ridge' appear frequently. For example: (46)

cddexdextal 'qw (chddeqhdeqhta 'Iqw) Ic - REDUP Ndext -ilqw/ on - REDUP walk - stick-like.object n. 'bum (lit. those who walk on log (RR))' NL35

3.4.2. The second subcategory is of smeared or spread-out contact. The landmark is often /-ilqw/ 'stick-like object', /-ic'e?/ 'blanket, skin, hide', or /-ican'/ 'back; ridge'. (47)

cyanp Yc 'e? (chynp 'its 'e') /c - \lyenp' - ic'e?/ on - to.clamp - blanket,skin,hide n. 'barrel (lit. a wooden container held together by hoops)' NL55

3.4.3. A third subcategory consists of words involving a distributed contact with generalized or tacitly given surfaces. A word in this category is: (48)

ctemantam (chtemntm) /c - \llem - 9nt9m/ on - dew - 3.s.subj.PASS.PAST vt. 'bedew (lit. it became wet with dew)' NI:47

This subcategory may be considered to be an abstraction of the first and second subcategories.

Langacker semantics for three Coeur d'Alene prefixes

185

3.4.4. A fourth subcategory is of words whose sense of meaning is metaphorical contact with a generalized surface. For example: (49)

cl6kwut (chlekut) Ic - Nlekw -1/ on - distant - inherently (or, VB) adv. 'aloof NI:47

3.4.5. A fifth subcategory consists of words involving a long, slender thing penetrating a surface. This consists of few words. (50)

ctek'wnts (chlek'wnts) Ic - Nlek' w - ants/ on - pierce.with.pointed.stick.(barb;spike;needle;fork) 3.s.subj.3.s.obj.TRANS.PAST vt. 'he broiled it (salmon)' NL47

(51)

cliji?snts Ic - Nlej - i?st - ants/ on - stab - round.object - S.s.subj.S.s.obj.TRANS.PAST 'he stabbed a spherical object' J: 236, R: §513

This category seems to be independent of the previous categories and represents a distinct sense of meaning for Ic-l. 3.4.6. The complete category of Schemas and their relationships may be visualized in Figure 3. The fact that the semantics for this prefix involves two different senses of meaning is visible as the two separate networks, one of which has only a single node. Each connected component of this network is a meaningcomplex, a set of related meanings. 3.5. /t-c-/ These words include the combined prefix /t-c-/ and involve an attached /t-/-like contact onto something in a /c-/-like manner, that is, they involve attachment, often holding onto a stick-like object or spread over a surface. This combination of prefixes was found in 13 words. Reichard considers

186

Roy H. Ogawa and Gary B. Palmer 4. Metaphorical distributed contact

3. Distributed contact

1. Scattered and/or repetitive contact

5. Piercing contact

2. Smeared or spread out contact

Figure 3. Schemas of prefix /c-/.

/tc-/ as a single prefix, glossed as 'sit up on, astraddle; a counting prefix.' (Reichard 1938: 595). Here we consider them as two prefixes to see whether the interpretations of the meanings of /t-/ and /c-/ are coherent with the gloss for/to-/. For example: (52)

tc 'okwtalqw (ich 'okwtalqw) /t-c-Muk w t-ilq w / on - on - crawl - stick-like object vi. 'H/s crawled on a log, stick.' NL255

The trajector is to crawl; the landmark a stick-like object; and the contact is of attachment (you have to hold on to a log) and distributed contact in time with a stick-like object.

(53)

?etc'emutw'es /?ic -1 - c- N?em - ut - iw'es/ CUST - on - on - sit - position - waist,middle 'he is on horseback' J: 233, F: 136, R: §396

The trajector is to sit; the landmark is the waist of the horse; and the contact is both an attachment with the legs hanging on astraddle and with legs spread over a cylindrical object. In the following case the numbness (sleep) is a state of the leg and it is distributed over the surface of the leg.

Langacker semantics for three Coeur d'Alene prefixes

(54)

187

tc 'it&n (tch 'itshn) /t - c - Nit - ssn/ on - on - sleep - leg;foot n. 'bedtime, numb (lit. his leg went to sleep)' NL255

For the count of 10 below, the /t-/ prefix means that the count is a quality of the group, but the meaning of the /c-/ prefix is not as clear. Perhaps all the fingers is a collection of multiple objects, distributed over the hands to form a count of 10 and this 10 is a quality of the group of grizzlies or hogs for which /-ilps/ is a noun classifier. (55)

tc'upenctelps (tch'upenchtelps) /t - c - Nupan - ict - ilps/ on - on - ten - whole.hand;finger - throatjmane v. 'There are ten hogs or grizzlies.' NL256

We do find, then that their senses of meaning are coherent with the combined meanings of the two prefixes. It thus appears that Reichard's interpretation of this as a single prefix is not necessary.

4.

Summary

We developed a methodology for understanding the semantics of three morphemes of contact in Coeur d'Alene using the concept of network of related Schemas of Langacker. These are the word prefixes /cit-/, /t-/, and /c-/. Using these networks, we studied the prefix combinations /cit-t-/ and /t-c-/. We found that the senses of the combinations are understandable as combined applications of the single prefixes and do not require new, independent analyses. The /cit-/ words can be subcategorized into two classes, one in which the trajector-landmark contact may be described as that between a trajector with a broad surface landmark. The other class may be described as situations involving contact 'all over' the landmark. A good description for /cit-/ is: 1. contact with a broad surface, 2. covering all over. A prototypical configuration for the first description is /cit- STEM - ic'e?/ On a hide, mat' and a prototypical configuration for the second is /cit - STEM - ini?/ 'all over'.

188

Roy H. Ogawa and Gary B. Palmer

These descriptions should not be treated as glosses but as summary descriptions of the categories. The /t-/ words may be subcategorized into four classes, three related to each other. The first class may be described as words where the trajectorlandmark relation is that of attachment. The second class is an extension of the first by metaphor. The third class may be described as words where the trajector is a quality which is possessed by the landmark; this is a metaphorical attachment of the trajector onto the landmark. The fourth class may be described as words where the trajector involves motion either onto or off of the landmark. Thus the /t-/ words have two distinct senses of meaning which may be described as: 1. trajector attached to landmark, or metaphorically attached to landmark, or as a quality of the landmark, 2. trajector involves motion onto or off of the landmark. The /cit-t-/ words may be understood as a combination of applications of /cit-/ and /t-/ separately. The /c-/ words may be subcategorized into four related classes of meaning plus an unrelated class of meaning. The first class involves contact which is repetitive or scattered. Landmarks include stick-like objects, hides or blankets, or backs. The second involves smeared or spread-out contact. Landmarks again include stick-like objects, hides or blankets, or backs. The third class involves smeared or spread-out contact with general outer surfaces. The fourth class consists of meanings which are metaphorical extensions of those of the third class. The fifth class involves a piercing contact with a surface. The first four classes can be seen as related through elaboration and extension. They may be described as involving a distributed contact in space, a smeared or spread-out contact, or in space-time, a repetitive or scattered contact. The fifth class stands by itself so we can describe the meaning of /c-/ as 1. distributed contact, or 2. piercing contact. The /t-c-/ words may be understood as a combination of applications of/t-/ and /c-/ separately. In conclusion, using the concept of network of Schemas for semantics developed by Langacker, we have elucidated senses of meaning of three spatiorelational prefixes of Coeur d'Alene. The senses of meaning that we found differs from those described in the glosses given by previous workers. This should help in our understanding of the kinds of spatial-relations used by the Coeur d'Alene; this work also serves to validate the concepts of linguistic usage developed by Langacker.

Langacker semantics for three Coeur d'Alene prefixes

189

Appendix 1: Orthography Table 1. Cour d'Alene vowels according to Reichard (1939: 528). I

II

III

IV

α α Ε i i u 3 u

αα αα

α'α α'α

α'α α'α

i*

i'1

i'i

uu D3

u'u D' D

u'u D'D

α

Nicodemus' (1975), Sloat's (1968), and Johnson's (1975) vowels correspond with Reichards' as: Table 2. Correspondence of vowels in orthographies of Reichard (1938), Nicodemus' (1975). Sloat's (1968), and Johnson's (1975). Reichard

Nicodemus

0 α d e 1 i E u u

o a e i i u

Sloat and Johnson D (ο) a ε e ι i 3 u

o

Reichard's glottalized or lengthened vowels are indicated by Sloat and Johnson with a glottal stop and by a ':'. Nicodemus does not show schwas and indicates lengthening by repetition. The consonants for Reichard are shown in Table 3. Reichard shows glottalized vowels with an apostrophe, for example a'a or i'i. The consonants for Sloat and Johnson are shown in Table 4, and the consonants for Nicodemus are shown in Table 5.

190

Roy H. Ogawa and Gary B. Palmer

M

A p

Ο

c "o H-s Ό

T3

e

u. F

>

oo

13

(T "er

x·

er "er

x»

> *- >f

c^

^4 ^i

Tj

C cd

«"«

—'

ce

XI

^

cd

o

/

scitc'emicen'ict /s - cit - Nc'em - icen' - ict/ NOM - on - surface - back;ridge - whole.hand 'surface on the back of the hand' PN: 355 saic 'emicen ' /s - cit - Nc'em - icen' - SOT/ NOM - on - surface - back;ridge - leg;foot 'instep, front of the foot' PN: 355 m (chat(atsilkhwn) /cit - Nfac - Hxw - an/" on - tie - hide;mat;covering - VB v. 'bind, tie' NI:34 ctmelxwncut (chtmelkhwntsut) /cit - Nmilx w - an - cut/ on - naked, bare - 3.s.subj.TRANS.PAST - REFL vi. 'divest (lit. he/she made his/her self bare)' NI:54 Subcategory 2: Covering surfaces cac'p'q'ilxwn (chats'p'q'ilkhwn) /cit - Nc'ap'q' - ilxw - an/ on - adhere - hide;mat;covering - NOM n. 'asphalt, glue for the roof NI:34

Langacker semantics for three Coeur d 'Alene prefixes catxepiixwn (chatqhepilkhwn) /Hi- Nxep - ilxw - an/ on - pile.things.flat - house - NOM n. 'shingle, roofing, thatch' NI:34 cetc'mciix™ (chetch'mtsiikhw) /cit - NJc'em - can - ilxw/ on - surface - mouth,shore,edge - house n. 'porch, eaves, awning' NI:41 Subcategory 3: Supporting surfaces catq 'ey 'm/nan (chatq 'e 'yminn) /cit - Mq'iy' - man - an/ on - writing;design;graphics - that.which.is.used.for - NOM n. 'writing desk' NI:34 /cit - \lcis - 1/ on - long;heated - inherently 'table' J: 233, R: §133

cet'ifnan (chet'iinn) /cit - Milan - an/ on - eat - NOM n. 'table (lit. place on which to eat)' NI.-42 'si'am (echet'we'Iw'lch"mst'm) Ice · cit - REDUP Nwelc' - am - stam/ CUST - on - REDUP roll.solid.object - MID - 3.s.subj.PASS.PAST n. 'bowling ball (lit. that w/c is rolled in bowling)' NI:65 hancafxo/ 'xol 'q '^/can (hnchatqho 'Iqho 'Iq 'wichn) /han - 5it - REDUP Nxol'q' w - ican/ in - on - REDUP wrap. with. string - back;ridge n. 'pool table, billiards (lit. place for rolling things)' NI:88

195

196

Roy H. Ogawa and Gary B. Palmer cet'dmiSdnts /cit - N?am - e§ - ants/ on - sit - act.of - 3.s.sub.3.s.obj.TRANS.PAST 'he sat on it' J: 232, R: §398, F: 55 (s-tcat-Em-ut-En) /s - cit - N?am - ut - an/ NOM - on - sit - state.of;be.in.position - NOM n. 'chair' R: §571 sect' ek' uten /s - cit - Nt'ek' w - ut - an/ NOM - on - one.lies.down - state.of;be.in.position - NOM 'bed' J: 240, F: 56, R: §447 cairn Oq 'wm Oq 'wmindn (chat'moq 'w'moq'w'minn) /cit - REDUP \lmoq'w - min - an/ on - REDUP pl.obj.lie;pile - MID - NOM n. 'boxcar (lit. that on which things are loaded repeatedly)' NI:34 catp'ert (chafp'ert) /cit - \Jp'er - 1/ on - flood - inherently vt. 'the floor was flooded' NI:34 catpexsm (chatpeqhm) /cit - Npex w - am/ on - spit - MID vi. 'h/s spat on floor' NI: 34 catxelflupan (chatqhelilupn) /cit -\xel- flup - an/ on - lay.side.by.side;lay.evenly.(as.lumber) - floor - NOM n. 'floor' NI:34

Langacker semantics for three Coeur d'Alene prefixes

Subcategory 4: The land caianna" /cit - ban - Nnäs/ on - in - wet n. On the Meadow; Wet Meadow (Clarkia, Idaho)' P:276 cafxesnv'es (chaiqnesi'wes) /cit - NJxes - iw'is/ on - good - middle n. 'plain (lit. good middle of surface)' NI:34 cancetcig"is (ch 'nchetchigwish) /en - cit - \cigw - 3sV l.s.subj - on - extend.across.plain - act.of vi. traversed the plain' NI:59

Subcategory 5: The surface of water catpo?satkwe? (chatpo'satkwe') /cit - \Jpu?us - itkwe?/ on - bubble - water 'to bubble (lit. it (water) gave off bubbles)' NI:34 catq 'ele? (chatq 'ele') /cit - Nq'ele?/ on - lake n. 'lake' NI:34 cetp'ut'mkwe? (chetp'ut'mkwe') /cit - Np'ut' - am - itkwe?/ on - come.to.end.(as.river.road.woods) - MID - water n. 'coast, beach (lit. end of water)' NI:42 cetpihih (chetpihih) /cit - REDUP \pih/ on - .REDUP.rise.to.surface.and.separate.as.with.cream vt. 'it rose to the surface and spread out like cream' NI:42

197

198

Roy H. Ogawa and Gary B. Palmer

cetpuh'tkwe? (chetpu'titkwe') /cit - \Jpul - itkwe?/ on - foam - water n. 'rough water (lit. the surface of the water was in turmoil, as near a falls)' NI:42 ?ecet'3k'witkwe? /?i - cit - Nt'ek' w - itkwe?/ 3.s - on - one.lies.down - water 'he floats' J:239,F: 17, R: §519

Subcategory 6: Clothing on the body The trajector of the following word is the verbal 'flesh;meat' while the landmark is 'clothes'. The trajector-landmark relationship appears to indicate a person who is inside of clothing which is the overcoat. catqältcal'qs (chatqaltcha'lqs) /cit - Nqfltc - il'qs/ on - flesh ;meat - clothes n. 'coat, cloak, overcoat (lit. outer garment)' NI:34

Subcategory 7: A broad surface not specified ?ece:t'ecdn' /Tic - cit - N?ecan - an/ GUST - on - do.it;put - NOM 'it exists put on a flat object' J:232,R:§137 caiyaq 'ants /cit - \Jyeq' - ants/ on - sharpen;abrade - 3.s.subj.3.s.obj.TRANS.PAST 'he filed the surface'

J: 241, F: 95

ceic'uw fcit - Nc'uw/ on - gone 'it is gone from the surface' J: 235, R: §133

Langacker semantics for three Coeur d'Alene prefixes

199

ceijomants /cit - \Jjem - ants/ on - pin - 3.s.subj.3.s.obj.TRANS.PAST 'she pinned it to surface' J:236,R: 133

catxwar'i&nten /cit - \lxwer' - & - anton/ on - step.over - act.of - 3.p.subj.3.s.obj.TRANS.PAST 'they stepped over it' J: 240, F: 95 hancetmuln (hnchetmuln) /han - cit - Nmul - an/ in - on - dip - NOM n. 'creamery (lit. place where butter, etc., are made)' NI:88

Category 2: /cit-STEM-ini?/: 'All over' catp't't'äne? (chatp't't'ane') /cit - REDUP Np'at' - ini?/ on - REDUP pour.mushy.stuff - ear;over n. 'covering by lava, magma' NI:34 catYel'ine?n (chat(e'lme'n) /cit - NTel' - ini? - an/ on - cover - ear;over - NOM n. 'cover, something that conceals or disguises' NI:34 cat fei 'ene?ncut (chat(el 'ene 'nfsut) /cit-\l?el'-ini/-9n-cut/ on - cover - ear;over - 3.s.subj.TRANS.PAST - REFL imper. 'h/s covered himself/herself NI:34 ccetfel 'ene?ncut (chchetfe 'lene ' /c - cit - N Tel' -ini? -an -cut/ we - on - obstruct,block,fence,curtain - ear;over - 1. pi. TRANS. PAST REFL vt. 'we covered ourselves'. NI:35

200

Roy H. Ogawa and Gary B. Palmer cet'uxwil$ene?ents(tcä-t'uxw-il§-äne'-än-ts) /cit - Mt'ux w - il§ - ini? - ants/ on - wade;jump.off - arched.motion - ear;over 3.s.sub.3.s.obj.TRANS.PAST 'he jumped on him' R:§541 C9tgwelpine? (tcät-gwäl-p-inä') /cit - Ng w il - p - ini?/ on - burn - involuntarily - ear;over 'he burned all over' R: §541 catel'qini?ntem (tca-tal'q-ini?-ntem) /cit - Mt'el'q - ini? - antom/ on - step.on - ear;over - S.s.subj.PASS.PAST 'he was stamped on' R: §541

e cetc'lxwin'e? (echetch'lkhwi'ne') l\ cit - Nc'elx w - ini?/ PREP on - concave - ear;over n. 'covered wagon (lit. a wagon with an arched cover, tent, over it)' NI:65 ecetce?ine?stus (echetche'ine 'stus) lie - cit - Nci? - ini? - stus/ CONT - on - uncover - ear.over - 3.s.sub.3.s.obj.TRANS.CONT. vt. 'divulge, unveil (lit. he uncovers it)' NI:65

Appendix 3a: /t-/ The /t-/ words are divided into four categories: 1. something is physically attached to something. 2. something is metaphorically attached to something. 3. something is a state of something, and 4. something is taken from or put on something. The second category includes the third. The second category is an extension of the first category. The fourth category appears to be unrelated to the others and therefore gives a variant sense of meaning for /t-/.

Langacker semantics for three Coeur d'Alene prefixes

201

Category 1: Physically attached to: tcoqcequs (ttsoqtsequs) Λ - REDUP Nceqw~- us/ on - REDUP pink - eyejface adj. 'blood-shot, he has bloodshot eyes.' NI: 277 f fa faxe 'e? (t(a(aqhts'e') /t - REDUP \IYexο - ic'e?/ on - REDUP wrap evenly - blanket;hide;skin n. 'cantalope (lit. a melon ribbed around.)' NL279 tgwalpalqw (tgwalpalqw) /t - Ng w el - ρ - ilqw/ on - blaze - unwillingly - stick-like.object vt. 'the log burned' NL259 tcoSceSqan (ttseshtseshqn) /t - REDUP N|ci§ - qan/ on - REDUP long;heated - head n. 'comet (lit. it (star) is long-hair.)' NI:277 tc 'amasq 'it (tch 'masq 'it) /t - Nc'em - sq'it/ on - surface - day,sky,atmosphere n. 'sky, heaven, firmament (lit. vault of the sky.)' NI:256 tc'emipixw (tch'miplkhw) /t - Nc'em - ip - itx w / on - surface - bottom;behind - house n. 'ecology, surface of the earth (lit. outside of a house, surroundings of a house.)' NI: 256 tccen'gul (tch'e'ngul) /t - Nc'en' - gwai/ on - lay.round.object - stomach n. 'person with bulging stomach' NL255

202

Roy H. Ogawa and Gary B. Palmer tc'dlxwic'e?is (tch'lkhwits'e'is) /t - Nc'elxw - ic'e? - is/ on - be.receptacle.with.concave.surface.down - blanket;skin;hide 3.S.POSS n. 'crust (lit. covering (crust) of turtle.)' NI: 255 tc'dlxwic 'e?n (tch 'Ikhwits 'e 'n) /t - Nc'elx w - ic'e? - an/ on - be.receptacle.with.concave.surface.down - blanket;skin;hide - NOM n. 'eggshell (lit. the hard (but brittle) outside covering as on an egg.)' NL255 tc'ew'ect (tch'e'wecht) /t - Nc'ew' - ict7 on - widening - whole.hand n. 'bough, limb (lit. a large arm (hand).)' NL255 tSt'iw'es (tsht'iw'es) /t - Nset' - iw'es/ on - to.extend,to.project - between;waist n. 'kettle (lit. a vessel with a projection from its body.)' NI: 277 tSeVect (tshet'echt) ft - N§et' - ict/ on - to.extend,project - whole hand n. 'branch (lit. projection from a tree.)' NI:277 tgw6xwn (igwekhwn) /t - \lgwexw - an/ on - plural.objects.hang - l.s.subp.p.obj.TRANS.PAST ν. Ί hung them up' NI:259 tgwcxwnt (tgwekhwnt) /t - Ng w ex w - ant/ on - plural.objects.hang - 2.s.subj.3.p.obj.IMPER v. 'You (sing.) hang them up (imper.)' Nl:259 tfeciw'es (t(etsj'wes) /t - \ITec - iw'es/

Langacker semantics for three Coeur d'Alene prefixes

203

on - tie - between;mutually n. 'Sunday (lit. flag day)' NL280 tgwepelxw ha sxwet'i? (tgwepelkhwha sqhwet'i') li - Ng w ep - ilx* ha sxwet'iV on - hairy - hide;mat;covering this goat n. 'Angora (lit. He (the goat) has long hair.)' NL259 tgwapqan (tgwapqn) li - Ng w ep - qan/ on - hairy - head v. 'H/s has thick hair.' NI: 259 tgwepgup$9n (tgwepgupshn) li - REDUP Ng w ep - §an/ on - REDUP hairy - foot;leg v. The horse has hairy feet (ankles).' NI: 259 tc'9lxwip}xw (tch'lkhwiplkhw) li - \lc'elxw - ip - ilx w / on - be.receptacle.with.concave.surface.down- bottom;behind;after - house n. 'annex (lit. a building added on to another building.)' NI:255 tc'ap'q' (tts'ap'q') li - Nc'ep'qV on - adhere n. 'adhere, stick to. (lit. it stuck fast as if glued to something; he became firmly attached.)' NI: 277 tc'ap'q'an (tts'ap'q'n) li - Nc'ep'q* - an/ on - adhere - l.s.subj.3.s.obj.TRANS.PAST v. 'annex (lit. I glued it to something else.)' NI: 278 tc'p'xwumin3n (tch'p'khuminn) li - Nc'p'x w - man - an/ on - clip - that.which.is.used.for - NOM

204

Roy H. Ogawa and Gary B. Palmer n. 'clothespin (lit. a clip for fastening clothes on a line.)' NI: 256 t$an'alqw (tsha'nalqw) Λ - Nsen' - ilqw/ on - to.lie.(flat.object) - stick-like.object n. 'billboard' NI:276 'alqwdn (tsh'nalqwn) Λ - Nsen' - flqw - an/ on - to.lie.(flat.object) - stick-like.object - NOM n. 'cardboard (lit. flat object that is put on a post.)' NI:277 tcukw§niw'esnts (ttsukwshni 'wcsnts) /t - Ncekw - San - iw'es - ants/ on - pull - foot;leg - between;mutually - 3.s.subj.3.s.obj.TRANS.PAST v. 'to carry off, drag away (lit. he carried h/h off.)' NI: 277 (t-shepm) /t - Nsep - am/ on - chew,eat.off.bone - 3.s.subj.3.s.obj.TRANS.PAST vt. ' (lit. h/s chewed meat from bone.)' NI:276 tgw9$uxwnts (tgwshilkhwnts) it - Ngwe§ - flxw - ants/ on - comb - hide;mat;covering - 3.s.subj.3.s.obj.TRANS.PAST v. 'H/s combed the horse's hide (hair)' NL259 (tch'mtsinm) /t - Ncem' - con - am/ on - take.hold.of.many - mouth - MID v. 'he took them (berries) off bush to eat.' NI:256 tcom'am (tche'mm) /t - Ncem' - am/ on - take.hold.of.many - MID v. 'detract (lit. he plucked (berries, etc.))' NI: 255

Langacker semantics for three Coeur d 'Alene prefixes

205

Category 2: Metaphorically attached tc6kwukwum (ttsekukum) ft - REDUP Ncekw - am/ on - REDUP pull - S.s.subj.PASS.PAST vi. 'attract (lit. He was drawn.)' NL277 tcekwdnts (ttsekwnts) Λ - Ncekw - ants/ on - pull - S.s.subj.S.s.obj.TRANS.PAST vt. 'detract (lit. He drew it off.)' NL277

tcukwukwutom (ttsukukutm) ft - REDUP Ncekw - ut - am/ on - REDUP pull - state.of;possibility - MID adj. 'attractive (lit. capable of drawing (something).)' NL278 ' (tse'se'q"ms'n) /t - REDUP Nse?q' - am - san/ on - REDUP split;gaping - MID - l.s.subp.s.subj.TRANS.PAST vi. Ί punned on it. (lit. I took it (word) off the beaten path repeatedly; I played with words.)' NI:270 tcukw§9niw 'esantam (ttsukwshni 'wesntm) /t - Ncekw - san - iw'es - antom/ on - pull - foot;leg - between;mutually - S.s.subj.PASS.PAST vt. 'banish, debunk (lit. he was snatched away by the feet.)' NI:278 icen 'can 'am ( tche 'nch 'nm ) /t - REDUP Ncen' - am/ on - REDUP hold.one.object - MID v. 'cling (lit. he held fast to something physically or emotionally.)' NI:255 tselcan (tshelchn) /t - Νδέΐδ - an/ on - circle - l.s.subp.s.obj.TRANS.PAST vt. '[en]compass (lit. I circled it;I went around it.)' NI: 276

206

Roy H. Ogawa and Gary B. Palmer tSelcants (tshelchnts) /t - Νδέΐδ - ants/ on - circle - 3.s.subj.3.s.obj.TRANS.PAST vt. 'circle (lit. He circled it;h/s went around h/h/i.)' NI:276 tSelcanten (tshelchntm) /t - NSelc - antom/ on - circle - S.s.subj.PASS.PAST v. 'around, beset, he was surrounded, beleaguer (lit. he was encircled, crowded around.)' NI: 276 tce?wple?nts (tche(wple'nts) /t-Nc'e? w -ipli?-3nts/ on - pray - handle;attachment - 3.s.subj.3.s.obj.TRANS.PAST v. 'bless, pray over' NI: 255 tc'uxwipele? (tts'uqhwipele') /t - N°c'exw - ipli?/ on - promise - handle;attachment v. 'decide (lit. he judged.)' NI: 278 t§im'9nts (t-shi'mnts) /t - Ν sim' - ants/ on - to.benefit - 3.s.subj.3.s.obj.TRANS.PAST v. 'H/s benefited from h/h' NI:277 (tsht'ntsutn) ft - Μ set' - an - cut - an/ on - take.care.of - 3.s.subj.TRANS.PAST - REFL - NOM n. 'amulet (lit. means of taking care of oneself.)' NI:277 tSt'ilt (tsht'ilt) k - Nset' - ill/ on - take.care.of - offspring vi. 'to babysit (lit. he cared for his own or others' children.)' NI: 277 tgw6nust (tgwentust) /t - \lgwen - 1 - ust/

Langacker semantics for three Coeur d'Alene prefixes on - low - inherently - movement.along vi. 'He/s went by a low road, path.' NI: 259 tcehican' (ttsehich'n) Λ - Ncih - ί&»ηΥ on - be.next.to.move - back;ridge v. 'He went along the ridge.' NL277 Category 3: State as a metaphorical attachment of a property ttuxwoxwiii (ttuqhoqhwin) w

/t -'REDUP Ntex - my

on - REDUP one.dies,stops - INTENS vi. 'He was found dead.' NL277 tgwesi?qs (tgwesi'qs) /t - Ng w is - 3?qs/ on - ascend;high - nose;beak v. 'H/s has a high nose' NL259 ?atkosqen /?ic - t - N k u s - qan/ CUST - on - curl - head 'his hair is curled' J: 236, R: 208 tc'it'elx™ (tch'it'elkhw) /t - \c'it'- ilxw7 on - brown - hide;mat;covering adj. 'brown (lit. a horse with brown fur (hide).)' NI: 256 thanqan (thanqn) /t - Nhdn - qan/ on - be.grayish.in.color - head n. 'brunette' NI: 259 hiitc 'axws /h3c-t-Mc'axw-is/

207

208

Roy H. Ogawa and Gary B. Palmer 3.S.POSS - on - guilt - Ss.POSS 'his fault' J: 234, F: 77, R: §375 tc'esesmilgwes (t-tcJes-es-m-ilgwes) fi - REDUP Nc'es - am - ilgwes/ on - REDUP be.bad - MID - stomach;heart 'he got indigestion' R: §490 tcaypcan (tchayptsn) ft - Ncäy - p - can/ on - hard - unwillingly - mouth adj. 'agape (lit. his mouth became hard)' NI: 255 tcc9m'e?us (ttsts'me'us) /t - REDUP Ncem' - i? - us/ on - REDUP small - exaggeration - eye;face n. 'beads (lit. little eyes.)' NL277 tcal'el'ine? (tchTeline'} /t - REDUP \lcel - ini?/ on - REDUP await.eagerly;anxiously - ear;over v. 'he has sharp hearing (lit. he is alert as to the ears.)' NI: 255 tcaqene?utem (ttsaqene'utm) /t - \caq - ini? - ut - am/ on - receptacle.stands.with.concave.side.up - ear;over - state.of;position MID n. 'audible (lit. capable of being heard.)' NI:277 tcaqineYmants (ttsaqine'mnts) /t - Ncaq - ini?- am - ants/ on - receptacle.stands.with.concave.side.up - ear;over - MID 3.s.subj.3.s.obj.TRANS.PAST vt. 'he heard him, he obeyed him, he heeded him.' NI: 277 tc'äcalqs (tts'achalqs) li - Nc'ec - alqs/

Langacker semantics for three Coeur d'Alene prefixes

on - count - end;road n. 'Tekoa mountain (lit. numbered point.)' NL277 ic/fes (tchifes parson) Λ - Nci?lis parson/ on - three person n. 'Blessed Trinity (lit. three persons)' NI: 255 tce?Hselps (tche'iiselps) Λ - Nci?h's - ilps/ on - three - throat,mane.(counting.classifier) n. 'three pigs, three grizzly bears' NI: 255

the?in'9m (the'i'nm) Λ - \lhe?in' - am/ on - eight n. 'eight persons' NI: 259 Category 4: Trajectory to or from tm'm'exw9n (t'm'mekhw'n) /t - REDUP \lmexw - an/ on - REDUP laugh - NOM n. 'butt (lit. a laughing stock.)' NL287 tm'uxwm'uxwdn'mints (t'mukhw'mukhw'nmints) /t - REDUP \lmexw - an - am - ants/ on - REDUP laugh - NOM - MID - 3.s.subj.3.s.obj.TRANS.PAST v. 'deride (lit. he laughed at him, he mocked him.)' NI: 287 tpdr'kwqents /t - Npar'k w - qan - ants/ on - nail - head - 3.s.subj.3.s.obj.TRANS.PAST 'he poked him on the head' J: 237, F: 98 tc'iteYmants (tch'ite'mnts) /t - Nc'ite? - am - ants/

209

210

Roy H. Ogawa and Gary B. Palmer on - come.near - MID - 3.s.subj.3.s.obj.TRANS.PAST v. 'to approach (lit. he came close to him.)' NI:255 tpo\wqents /t - Npux w - qan- ants/ on - blow - head - 3.s.subj.3.s.obj.TRANS.PAST 'he blew on her head' J: 237, R: §208 tco?qintdm li - Ncuw' - qan - antam/ on - punch - head - 3.s.subj.PASS.PAST 'he was hit on the head' J: 233, S: 56 tcow'aqsantem /t - Ncuw' - iqs - antam/ on - punch - chest - S.s.subj.PASS.PAST 'he was hit on the chest' J: 233, S: 56 tp'at'a?stqents /t - Np'at' - i?st - qan - ts/ on - pour.mush - round.object - head - 3.s.subj.3.s.obj.TRANS.PAST 'she poured mush on his head' J: 237, R: §513 tsec'alq™ (tshech'alqw) li - NSec' - ilqw/ on - ?? - stick-like.object vt. 'derail (lit. it (train) ran off the rails.)' NI:276 tc3cminplxwdntem (tchchminpikhwntm) li - Ncic - man - p - itx w - antom/ on - throw.one.object - that.which.is.used.for - unwillingly - house 3.s.subj.PASS.PAST v. 'degrade, eject (lit. he was thrown out of the house, he was deprived of dignity, he was ousted.)' NI: 255 t?ecinitkupen li - \?ecan - it - kwup - an/

Langacker semantics for three Coeur d'Alene prefixes

211

on - to.do.with - for.use - fire;fuel - NOM 'fire poker' J: 232, R: §451 Appendix 3b: Body parts with ft-/ This category of words have /t-/ as prefix. They are all words which describe body parts and for no other reason we separate them. They all include /t-/ as a prefix because the body parts are attached to the body. Many attach to either the right side or to the left side. The first two groups involve neither right nor left. fine? (t'inää) /t - Nini?/ on - ear;over Outer ear' R: §473, PN: 356 stccintus /s -1 - REDUP ^scint - us/ NOM - on - person - eye;face 'pupil of the eye (lit. little Indian person on the eye)' PN: 354 stgwepiic'i?us /s -1 - Ng w ep - ilc'i? - us/ NOM - on - hairy - edge - eye 'eyelash' PN: 354 stgwepgwepican'us /s -1 - REDUP \lgwep - icon' - us/ NOM - on - REDUP hairy - back;ridge - eye;face 'eyebrows' PN: 354 The following category of words all involve /t-c'em/ 'surface attached/part of. stc'emcan /s -1 - Nc'em - can/ NOM - on - surface - mouth ;edge 'mouth' PN: 356

212

Roy H. Ogawa and Gary B. Palmer stc'emgwdl /s -1 - Nc'em - gw9l/ NOM - on - surface - abdomen,stomach 'abdomen' PN: 356 stc'emilqws9n Is -1 - Nc'em - ilqw - San/ NOM - on - surface - stick-like.object - leg 'the whole surface of the leg' PN: 355 stc'emusixan « /s -1 - Nc'em - us - ixan/ NOM - on - surface - eye;face - arm, wing 'shoulder' PN: 356

The following words include /t-c'ih-am/ meaning 'attached to the right.side' stc'ihamilq&n /s -1 - Nc'ih - am - ilqw - son/ NOM - on - right.side - MID - stick-like.object - foot;leg 'right leg' PN: 357 stc'ihamusixan /s -1 - Nc'ih - am - us - ixan/ « NOM - on - right.side - MID - eye;face - arm;wing 'right shoulder' PN: 357 stc'ihict /s -1 - Nc'ih - ict/ NOM - on - right.side - whole hand 'right arm' PN: 357 stc'ihini? /s -1 - Nc'ih - ini?/ NOM - on - right.side - ear;over 'right ear' PN: 357

Langacker semantics for three Coeur d'Alene prefixes stc'ihsan /s -1 - Nc'ih - San/ NOM - on - right.side - foot;leg 'right foot' PN: 357 sic'ihus /s -1 - Nc'ih - us/ NOM - on - right.side - eye;face 'right eye' PN: 357 The words below have /t-c'ikwi?/, meaning 'surface on left'. stc'ikwi?ini? /s -1 - Nc'ik w i? - ini?/ NOM - on - surface.on.left - ear;over 'left ear' PN: 357 stc'ikwi?husixdn β /s -1 - Nc'ik w i? - us - ixan/ NOM - on - surface.on.left - eye;face - arm, wing 'left shoulder' PN: 357 stc'ikwi?ilqwsdn /s -1 - Nc'iVi? - ilqw - SOT/ NOM - on - surface.on.left - stick-like.object - foot;leg 'left leg' PN: 357

stc'ikwi?$9n

/s -1 - \lc'ikwi? - son/ NOM - on - surface.on.left - foot;leg 'left foot' PN: 357 stc'ikwi?us /s -1 - Nc'iVi? - us/ NOM - on - surface.on.left - eye;face 'left eye' PN: 357 stc'ikwi? /s -1 - \lc'ikwi?/

213

214

Roy H. Ogawa and Gary B. Palmer NOM - on - surface.on.left 'left hand, left arm' PN: 357

Appendix 4: /cit-t-/ These combinations consist of a /t-/, 'attachment or trajectory', and /cit-/, On a surface' . Their senses of meaning are coherent with those of the two prefixes combined.

Trajectory off (/t-/) from surface (/cit-/) ceitc'uw /cit -1 - Nc'uw/ on - on - gone 'it is gone from the table' J: 235, R: §133

Attached (/t-/) to surface (/cit-/) ce.'tjemenis /cit -1 - N jem - ants/ on - on - pin - 3.s.subject.3.s.object.TRANS.PAST 'she pinned it to it' J: 236, R: §133

Trajectory off from (/t-/) surface (/cit-/) ceit'iindm (cheet'ilnri) /cit -1 - Milan - an/" on - on - eat - NOM n. 'table, buffet, sideboard (lit. that on which one eats)' NI:36

Attached to cloth (/t-/) at surface (/cit-/) ce:t'ip'eman (c/ieei'jp'emn) /cit -1 - Nip' - am - an/ on - on - wipe - MID - NOM n. 'mop, wiping cloth for top (of floor, table, etc.)' NI:36

Langacker semantics for three Coeur d 'Alene prefixes

215

Appendix 5: /c-/ The words in this group subcategorize into 5 senses of meaning. One sense, Category 4, is metaphor from another, Category 3. Subcategory 1 : Repetitive or scattered contact c?actartal'qalqw /c - ?ic - c - REDUP \ltel'q - ilqw/ we - CUST - on - REDUP touch.with.foot - stick-like.object 'we customarily step on a tree' J: 239, F: 134 cddexdextal 'qw (chddeqhdeqhta 7qw) Ic - REDUP \ldext -ilqw/ on - REDUP walk - stick-like.object n. 'bum (lit. those who walk on log (RR))' NI:35 cleixwal'qw (chtelqhwa'lqw) Ic - REDUP \ltexw - ilqw/ on - REDUP pierce.with.pointed.object - stick-like.object n. 'music, flute, cembalo, harpischord (lit. holes bored on a hollow stick, log, etc.)' NI:47 cdlamalqw (chdlamalqw) Ic - ^delim - ilqw/ on - galloped - stick-like.object n. 'train (lit. he galloped on the log)' NI:36 ' (cht'lt'lqich'n) Ic - REDUP Ntel'q - ican'/ on - REDUP touch.with.foot - back; ridge vn. 'bike, bicycle' Nl:55 (cht'ltl'qminn) Ic - REDUP Ntel'q - man - an/ on - REDUP touch.with.foot - INSTR - NOM n. 'steps' NI:55 In the following case nests are scattered in the water.

216

Roy H. Ogawa and Gary B. Palmer hncdmine?kwe? /han - c - Ndmine? - itkwe?/ in - on - nest - water 'Nests on the Water' P:275 ctuxwcindn (chtukhwtsinn) Ic - ^taxw - can - an/ on - collect - mouth - NOM n. 'relishes, condiment' (lit. something added to food, eaten). NI:55 cte/ef waJwa/iman (chte 'ef wlwlimn) Ic - REDUP Nti/ - at - Nwlwlim - an/ on - REDUP hit,pound - CONJ - metal - NOM n. 'anvil' NI:54

Subcategory 2: 2-dimensionally smeared or spread-out contact cyel 'yel 'x^ Ic - REDUP \lyil'x w - flqw - San/ on - REDUP cover. with.fabric - stick-like.object - leg;foot 'leggings' J: 241, R: §498 cyoryerk 'walq&n (chyeryerk ' walqshn ) Ic - REDUP \lyerk'w - ilqw - San/ on - REDUP curved - stick-like.object - leg adj. 'bow legged (lit. legs that curve outward at the knees)' NI:55 ciaq'walqw (chiaq'walqw) Ic - \lleq'w - ilqw/ on - remove.skin - stick-like.object vt. 'he stripped bark off a tree'. NI:47 In the following two cases, the landmark is tacitly the axle. cyarpan (chyarpn) Ic - Myir - p - an/ on - revolve - INCHOATIVE - NOM n. 'axis, axle (lit. something on which something turns (wheels))' NI:55

Langacker semantics for three Coeur d 'Alene prefixes cyarpSan (chyarpshn) Ic - \lyfr - p - San/ on - revolve - INCHOATIVE - foot;leg n. 'cartwheel'. NI:55 ci'dic'e?ntom (chi'ofis'e'ntm) Ic - Nt'ol - ic'e? - antom/ on - lumpy,sticky - blanket,skin,hide - S.s.subj.PASS.PAST vi. 'it was besmeared' NI:55 cyanp 'ic 'e ? ( chynp 'its'e') Ic - Myenp' - ic'e?/ " on - to.clamp - blanket,skin,hide n. 'barrel (lit. a wooden container held together by hoops)' NI:55

Ic - REDUP Ntam - icanV on - REDUP scorch - back;ridge 'Scorched Mountain' J: 239, R: §231 ciaq'wdnts (chlaq'wnts) Ic - Nleq' w - ants/ on - remove.skin - 3.s.subj.3.s.obj.TRANS.PAST vt. 'he peeled it off' NI:47 cyal 'yal 'xwosaxdn (chy 'ly 'Ikhosaqhn ) Ic - REDUP \lyil'x w - us - ixan/ on - REDUP cover. with. fabric - face;eye - arm n. 'cape (lit. covering on the shoulders)' NI:55 cnd?c 'emiw'isqan Ic - no? - \c'em - iw'es - qan/ on - amidst - surface - between - head;top 'top of the head' PN: 355 cyac 'ydc 'amactem Ic - REDUP \lyac' - am - ict - am/

217

218

Roy H. Ogawa and Gary B. Palmer on - REDUP tighten - MID - whole.hand - MID 'hold on tight' J:241,R:§240 ci'ulc>wlc'e?ip (cht'uk'wlts 'e 'ip) Ic - Nt'uk' w -1 - ic'e? - ip/ on - bumpy.uneven - DITRAN - blanket,skin,hide - bottom n. 'beaded belt purse on the hip.' NI:55 scfusman /s - c - \his - man/ NOM - on - eye;face - that.which.is.used.for 'eye' PN: 354

Subcategory 3: Contact with a generalized tacit outer surface ciemantem (chiemntm) Ic - Niem - antom/ on - dew - 3.s.subj.PASS.PAST vt. 'bedew (lit. it became wet with dew)' NI:47 ?uw'esc}emt /?u-?i-s-c-Nlem-t/ just - PREP - NOM - on - dew - inherently 'just in the fog' J:236,F:4,R:§127 ct'ek'w9nts (cht'ek'wnts) Ic - Nt'ek' w - ants/ on - one.lies.down - 3.s.subj.3.s.obj.TRANS.PAST vt. 'to apply (lit. he put it on something)' NI:55 cte?emfa'caw'eS (chte'emi'nts'wesh) Ic - \lti? - am - ants - iwes - a§/ on - hit,pound - MID - 3.s.subj.3.s.obj.TRANS.PAST - between 2.pl.subj.INTRANS vi. 'clash (lit. they collided)' NI:54 ct'al't'al'fw'es (cht'Wli'wcs) Ic - REDUP Mt'el' - iw'es/

Langacker semantics for three Coeur d'Alene prefixes

219

on - REDUP tear,rip - between vi. 'it burst (lit. it was torn apart from internal pressure)' MI: 5, R: §395. ?a:'axdmfon /?ic -°c - Mex - am - r - an/ CUST - on - move - MID - on.behalf.of - l.s.subp.s.obj.TRANS.PAST Ί customarily move it for him' J: 233, F: 21 Compare ci'apfnew'ancex"' below with the English 'you cover my flank'. ct'apinaw'9ncexw /c - Nt'ap - ilniw' - an - cixw/ on - shoot - alongside - VB - willingly 'you (sg.) shoot alongside me' J: 239, R: §222 Subcategory 4: The following words involve metaphorical usage of Category 3 ciaq'iaq'ine? (chfaq'laq'ine') /c-REDUPNlaq'-ini?/ " on - REDUP broad - ear;over vt. 'he has big ears' NT: 47 ciemiemnul (cniemfemny]) /c - REDUP Ntam - an - ul/ on - REDUP scorched - VB - habitually vi. 'he is captious (lit. he is marked by a disposition to find fault)' NL54 ciexwiuxw9n'ts (chtekhwtukhw'nts)

/c - REDUP Ntexw - ants/ on - REDUP collect - 3.s.subj.3.s.obj.TRANS.PAST vt. 'he/she exaggerated it (lit. he/she colored it highly)' NI:54 ctexw9nts (chtekhwnts)

/c - \ltexw - ants/ on - collect - 3.s.subj.3.s.obj.TRANS.PAST vt. 'to add, aggrandize, he increased it (lit. he added it to something else)' NI:54

220

Roy H. Ogawa and Gary B. Palmer clekwut (cMekut) Ic - \lek w -1/ on - distant - inherently (or, VB) adv. 'aloof NI:47

Compare cdoln with the English 'I'll cover you'. cdehn (chdeln) Ic - Ndel - an/ on - sit.up.like.animal - l.s.subj.3.s.obj.TRANS.PAST vt. 'defend, guard, shield, protect (lit. I defended him/her).' NI:35 cyac'p (chyafs'p) Ic - \lyac' - p/ on - tighten - unwillingly adj. 'fond (lit. he/she became very attached to someone, a home, an occupation, etc.)' NI:55 cy'oqwdtits (ch'yoqw'nts) Ic - Vuq w - ants/ on - lie - 3.s.subj.3.s.obj.TRANS.PAST vt. 'calumniate, slander, lie' (lit. he spoke falsely and maliciously of him, he told lies to him, he slandered him). NI:62 cy'oqwp9n'tdm' (ch'yoqwp'nt'm) /c - Ny'uq w - p - antom/ on - lie.(untruth) - unwillingly - 3.s.subj.PASS.PAST vt. 'defame, slander (lit. false things were said about him/her)' NI:62

Subcategory 5: This group of words involve a long slender thing penetrating the landmark ciek'wznts (chiek'wnts) /c-Nlek' w - 9 nts/ on - pierce.with.pointed.stick.(barb;spike;needle;fork) 3.s.subj.3.s.obj.TRANS.PAST vt. 'he broiled it (salmon)' NI:47

Langacker semantics for three Coeur d'Alene prefixes ?ec$6t'cdn /?i - c - M§6t' - can/ PREP - on - one.stands.up - mouth;shore;edge 'it sticks out of mouth' J:239,F: 135 cli]i?snts /c - Nie] - i?st - ants/ on - stab - round.object - 3.s.subj.3.s.obj.TRANS.PAST 'he stabbed a spherical object' J: 236, R: §513

Appendix 6: /t-c-/ tcokwtalqw (tch'okwtalqw) Λ - c- Nuk w t - ilqw/ on - on - crawl - stick-like object vi. 'H/s crawled on a log, stick.' NT: 255 tc'eme$alqw / t - c - N?am - as - ilqw/ on - on - sit - act.of - stick-like.object 'he sat on a log' J: 235, F: 135 tc'igwul (tch'igul) / t - c - Nigwal/ on - on - climb v. 'clamber (lit. he climbed.)' NI: 255 tc'igwubn (tch'iguln) / t - c - Nigw3l - an/ on - on - climb - NOM n. 'step ladder' NI: 255 ?atc'dm6$a?stqsn /?ic - t - c - N?em - as - i?st - qan/ CUST - on - on - sit - act.of - round.object;surface - head;top 'it customarily perches on his head'. J: 234, J: 235, R: §209

221

222

Roy H. Ogawa and Gary B. Palmer tc'emiS Λ - c - N?em - aS/ on - on - sit - act.of 'it perched' J:232,F: 131 Fetc'emutw'es /?ic -1 - c- N?em - ut - iw'es/ CUST - on - on - sit - position - waist;between 'he is on horseback' J:233,F: 136, R: §396 tc'amutewes (tch'mute'wes) /t - c - N?em - ut - iw'es/ on - on - sit - position - waist;between adv. On horseback' NI: 256 tc'itSdn (tch'itshn) /t - c - Nit - san/ on - on - sleep - leg;foot n. 'bedtime, numb (lit. his leg went to sleep)' NI: 255 tc'upenctelps (tch'upenchtelps) /t - c - Nupan - ict - ilps/ on - on - ten - whole.hand;finger - throat;mane v. There are ten hogs or grizzlies.' NI: 256 fc'upen (tch'upen) /t - c - Nupan/ on - on - ten n. 'ten persons' NI: 256

Uncategorized usages The use of /c/ in the following is puzzling. tcnkwinpele? (tchnkwinpele') Λ - C - ban - \k w in - ipli?/ on - on - in - sing - handle;attachment v. 'carol (lit. he celebrated or praised in song.)' NI: 255

Langacker semantics for three Coeur d'Alene prefixes

223

The following is also puzzling. Perhaps the person is sitting on a log by the water. tc'mutpkwe? (tch'mutpkwe') /t - δ - N?em - ut - p - itkwe?/ on - on - sit - position - unwillingly - water n. 'Mt. Baldy in Idaho, near the St. Joe River (lit. one who sits by the water.)' NI:256

Notes 1. We use the reference 'Amerind' instead of 'Native American' or 'American Indian'. 2. We refer to the sources of examples by initials and page number for Johnson (1975) [we use an 'F' to refer to his field notebooks], Nicodemus (1975, vol. I and vol. Π), Palmer and Nicodemus (1985), and Palmer (1990). For Reichard (1938) we use initial and section number. 3. In our morphology, we use the forms selected by Johnson (1975) as underlying forms, in particular his choice of vowels. Perhaps a better choice would be to use the results on Proto Interior Salish as a basis for selecting the vowels. In that case, for example, /- flqw/ would be /- alqw/. (Kinkade, 1994) 4. This quote may be succinctly summarized as: If a /t/ or Id stands before a coronal it becomes assimilated to that sound. If the accent is on the immediately following syllable the assimilation is evident in the echo vowel. (Kinkade 1994) 5. Extensional: definition of a term by listing its referents. Intensional: definition of a term by listing properties satisfied by its referents (see Crystal 1991). We view our network of Schemas as giving an extensional definition of a prefix. A gloss for a prefix is a predicate, a property which the referents of the prefix must satisfy. 6. The prefix /ε-/ is glossed as "a general preposition used with nominal forms. It means many things a preposition may mean in English; locative ideas, on, in, into, at, to; instrumental, with, by means of; agent, by." (Reichard 1938: 591) . Another possible analysis interprets the /ε-/ as Itc-l > /ε-/ before /cit/ and therefore representing the continuative.

References Card, Orson Scott 1991 Xenocide. New York: Tom Doherty Associates, Inc. Crystal, David 1991 A Dictionary of Linguistics and Phonetics, 3rd ed. Oxford: Basil Blackwell.

224

Roy H. Ogawa and Gary B. Palmer

Johnson, Robert 1975 The Role of Phonetic Detail in Coeur d'Alene Phonology. Washington State University, Ph.D. Kendrick-Murdock, Debra 1994 Semantic Analysis of the Coeur d'Alene Prefix cen-. Las Vegas: (in preparation). Kinkade, Dale 1983 Salish evidence against the universality of 'noun' and 'verb'. Lingua 60: 2539. 1994 Personal communication. Langacker, Ronald 1991 Concept, Image, and Symbol, the Cognitive Basis of Grammar. Berlin: Mouton de Gruyter. Nicodemus, Lawrence 1975 Snchitsu'umshtsn, The Coeur d'Alene Language, vol. l Coeur d'Alene-English, vol. 2 English-Coeur d'Alene. Spokane: University Press. Occhi, Debra, Gary B. Palmer, and Roy H. Ogawa 1993 Like Hair or Trees: Semantic Analysis of the Coeur d'Alene Prefix ne' 'amidst'. Proceedings of 1993 Summer Conference of the Society for the Study of Indigeneous Languages of the Americas, Berkeley Linguistics Department. Palmer, Gary B. 1990 'Where there are muskrats': the semantic structure of Coeur d'Alene place names. Anthropological Linguistics 32, 263-294. Palmer, Gary B., and Lawrence Nicodemus 1985 Coeur d'Alene exceptions to proposed universals of anatomical nomenclature. American Ethnologist 12, 341-359. Reichard, Gladys A. 1938 Coeur d'Alene, Handbook of American Indian Languages. New York: J. J. Augustin. 1939 Stem-list of the Coeur d'Alene language. International Journal of American Linguistics X, 92-108. Sloat, Clarence 1968 A skeleton key to Reichard's Coeur d'Alene transcriptions. Anthropological Linguistics 10,8-11. 1980 Vowel alternations in Coeur d'Alene. International Journal of American Linguistics 46, 14-20. Thompson, Laurence C. and M. Terry Thompson 1992 The Thompson Language, University of Montana Occasional Publications in Linguistics 8. Missoula: University of Montana.

Aspects of prepositions and prepositional aspect Sally A. Rice 1.

Introduction

This paper is a small part of a large-scale empirical investigation into the syntax and semantics of the English prepositions, the theoretical implications that their multivariate behavior has on models of lexical representation, and whether the prepositions form a homogeneous or heterogeneous word class.1 I will focus here on aspectual properties of prepositions using concepts and notation from Cognitive Grammar (Langacker 1987, 1990, 1991) and drawing on insights by Talmy 1985 and 1991, Rauh 1991, and others about the wide-ranging behavior of prepositions in English.2 This research fits squarely in the realm of cognitive linguistic analyses which explore the historical and logical order of senses or "grammaticalization channels" or "paths of metaphorical extension" of individual lexemes. As a word class, the English prepositions are known to have quite pronounced grammaticalized functions in addition to their regular prepositional duties (cf. Talmy 1985, Rauh 1991, Langacker 1992a and b). For example, many prepositions appear in constructions in which their canonical lexical usage (usually coded as the head of a prepositional phrase which marks the location of an entity or an event) is, by and large, irrelevant. Rather, in their guise as grammatical particles, they can mark quite diverse functions (e.g., case, causation, subordination, and periphrasis).3 Consider the examples in (1-5), which pair a lexical and grammatical usage of a preposition: (1) (2) (3) (4) (5)

a. b.

Mary flew ίο London. Mary wants to leave.

a. b.

John drove by the store. John was sued by the store.

a. b.

Audrey cooked over a hot stove. Audrey overcooked the pasta.

a. b.

He climbed up the ladder. He used up all the toilet paper.

a.

I killed him with a hammer.

INFINITIVE MARKER AGENT MARKER COMPARATIVE PREFIX COMPLETIVE PARTICLE

226

Sally A. Rice b.

I kissed him with abandon.

MANNER MARKER

Taken collectively, we find propositions used in construction types like the following (which by no means constitutes an exhaustive list): (6)

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

HEAD OF PREPOSITIONAL PHRASE in the box, on the table HEAD (?) OF ADVERBIAL PHRASE at risk, on fire COMPONENT OF PHRASAL VERB yell at, depend on COMPONENT OF VERB-PARTICLE CONSTRUCTION call someone up, turn it off, run someone over VERBAL PREFIX underestimate, overeat, outrun NOMINAL PREFIX o/terthought, owigassing, w-crowd PREDICATE I'm down today; They outed that politician ADVERB/ADJUNCT He stayed in; He continued on

When considering the full array of prepositional behavior, one cannot help but wonder whether a preposition is always a preposition in its meaning and function. Most researchers simply say, "No, it's not", and admit that, Januslike, they play both lexical and grammatical roles - a conclusion that's neither adequate given the full range of usage types nor helpful in explaining why there is so much semantic and categorial diversity afflicting this word class. Indeed, the lexical items which we call prepositions show up in such a wide variety of constructional frames in English that it is tempting to adopt Talmy's (1985) practice of referring to some of their instantiations as part of a broader class of verbal satellites and despair of ever linking their behavior to inherently prepositional properties. Despite the temptation to distinguish between use types of these lexemes on purely formal grounds and leave it at that,4 there remain a number of very important theoretical questions concerning the so-called lexical/grammatical alternations exhibited by adpositions in many languages and clearly manifested in the English prepositional system: (i) Are these target lexemes always prepositions? (ii) Are they equally meaningful in all applications? (iii)

Aspects of prepositions and prepositional aspect

227

Are the various usages or functional alternations of a preposition related lexically? Rather than accept unconditionally the descriptively convenient but explanatorily unsatisfying position that such alternations really do indicate a lexical/grammatical split between true prepositions and quasi-homophonous satellites or particles, let us assume that the meanings and usages of a given prepositional form are multidimensional in some vague, as yet undetermined way. For now, let us simply refer to the usages in the (a) sentences in (1-5) as basic and primarily spatial, while treating the (b) usages as extended and primarily non-spatial, bearing in mind that this distinction is more continuous than categorical. In addition, I will refer to these target lexemes as prepositions in the remainder of this paper, no matter what may be their particular lexico-syntactic manifestation in a given context. Persson 1988 and Lichtenberk 1991 call instances like this of purported grammaticalization of a lexical item by the not altogether transparent term, heterosemy, invoking it where "two or more meanings or functions that are historically related, in the sense of deriving from [perhaps] the same ultimate source, are borne by reflexes of the common source element that belong in different morphosyntactic categories" (Lichtenberk 1988: 476). Indeed, many cognitive linguists have long argued for recognition of prepositional polysemy, for one can reliably posit multiple, yet related meanings for individual prepositions even in their purely locational, that is, inarguably lexical usages (cf. Bennett 1975, Brugman 1981, Hawkins 1984, Herskovits 1986, Vandeloise 1991, Rice 1992, among others). Moreover, few dispute the fact that most prepositions have grammaticalized into functional particles or grams, to use Joan Bybee's (1988) term. I would, therefore, like to add another variable to the preposition/non-preposition debate and argue for both prepositional polysemy as well as what I will call prepositional polygamy or poly functionality, to adopt Janssen's (1993) term. These terms are more or less equivalent notions to heterosemy, but neither the underlying concept of related meaning despite cross-categorial function nor an agreed-upon descriptive term has attracted much theoretical attention or terminological consensus as of yet. Semantic extension from spatial meaning into the temporal domain is fairly ubiquitous among the English prepositions. What is of concern in this paper is an even less spatial, partially temporal, and clearly grammaticalized function of certain prepositions in English. In the remainder of this paper, I will address one such polygramous, poly functional, or heterosemous usage, that of aspectual particle, arguing that in addition to their myriad other usages and

228

Sally A. Rice

senses, many prepositions play a discernible aspectual role or at least convey aspectual meaning as well. By no means do I assume that the preposition is a fully grammaticalized aspectual marker in English or that it is unequivocably functioning to signal aspect. But indicating some sort of aspectual meaning is part of the preposition's function in the sentences under examination here. In some cases, the preposition is solely responsible for the aspectual interpretation that the sentence receives. Signalling aspect isn't such a strange or unexpected function for prepositions to have, considering that their aspectual meanings seem to be natural extensions of certain spatial meanings. Just as they can modulate the location of an entity in space or highlight the relevant contours or topography of a landmark object, so too can they modulate or reshape the contours of an event. Prepositions have long figured prominently in philosophical and linguistic work on verbal aspect since Vendler 1957 invoked them as a test for determining the inherent aspectual categories of verbs and sentences. Specifically, temporal expressions with in and for have usually been compared. Accomplishments or verbs and their complements denoting completed or culminated events readily combine with temporal PPs headed by in, whereas verbs and their complements denoting activities (that can, in principle, continue without limit) tend to combine with phrases headed by for. Some contrasting pairs are given in (7) and (8): (7)

a. b.

Mozart wrote the opera in I *for 6 weeks. Mozart wrote operas for/*in 26 years.

(8)

a.

Ever the insomniac, I (finally) fell asleep in 6 days. ACCOMPLISHMENT Ever the narcoleptic, I (?finally) fell asleep for 6 days. ACTIVITY

b.

ACCOMPLISHMENT ACTIVITY

Moreover, the traditionally labelled verb particles, up and out, are regularly associated with perfective or completive aspect. However, although they both carry roughly the same aspectual meaning, their use is heavily contingent on particular verbs and complements: (9)

a. b.

I filled up I *out the drawer. I emptied out I *up the drawer.

(10)

a. b.

She straightened up I *out the room, She straightened out/ *up the blanket.

Aspects of prepositions and prepositional aspect

(11)

c.

She straightened up I *out her husband.

a. b.

I tore up the letter. I destroyed (*up) the letter.

229

We find the same holds true for the prepositions under examination in this paper. I will ignore aspectual-temporal phrases like those in (7) and (8) containing in and for as they have received extensive treatment elsewhere (although not necessarily from a lexical semantic point of view).5 I likewise refer the reader to Lindner 1981 for an exhaustive study of the perfective verb particles up and out in English like those given in (9-11). In this paper, I will concentrate instead on the semantics of a number of other prepositions such as after, on, away, and others. Assuming that prepositions do indeed form heterosemous categories, I would like to build on existing work that has investigated one path of grammaticalization or metaphorical extension from concrete to abstract meaning and function, namely, the path from spatial motion to temporally situated action. I will suggest that aspectual usages represent a small detour or extension along this path. Next, I will look at a handful of prepositions and discuss the type of aspectual categories they seem to correlate with as evidenced by the types of verbs, inflectional aspects, and complement types that they co-occur with. Finally, I will speculate why certain prepositions have come to take on an aspectual function or why they seem to resonate aspectual properties present in some verbs. I propose that since prepositions are already so inherently flexible semantically and syntactically, they represent a fairly cheap and readily available resource for indicating minor aspectual categories for which English otherwise does not have a verbal inflection. In most cases, the prepositions under examination simply complement or reveal the inherent aspectual character of the underlying event. In present-day English, they are still restricted to certain predicates, argument structures, or inflectional aspects, but they seem to be gaining in productivity. Nevertheless, when taking both diachronic and synchronic evidence into account, certain prepositions are clearly undergoing specialized grammaticalization towards marking aspect.

2. Diachronic evidence for semantic shift and grammaticalization In the past decade, much research in cognitive linguistics has set about demonstrating the wide-ranging syntactic and semantic behavior of prepositions

230

Sally A. Rice

in English, French, German, and other languages, both synchronically (cf. Lindner 1981; Talmy 1985; Radden 1985; Rice 1993; and Langacker 1992a, b) and diachronically (Traugott 1978, 1982; Sweetser 1986; Genetti 1986). Researchers have invoked conceptual operations such as construal, schematicization, image-schema transformations, metaphoric extension, contextualization, etc., to account for the extension of these lexical items from purportedly basic uses in expressions of location, orientation, or direction, to extended uses in expressions of time, manner, comparison and various abstract functions. There are fairly compelling historical reasons for why the extensions happened in the way and direction that they did from concrete to abstract and objective to subjective/expressive usages. It has been shown, both historically and cross-linguistically, that distribution in time is readily conceived of in terms of distribution in space. In addition, few cognitive linguists would deny that there is a relationship between lexical and grammatical meaning, or in George Lakoff's (1990) terminology, that there is some preservation of image-schematic structure as the constructs associated with one domain are metaphorically mapped into another. Even Talmy (1985: 103) concedes that prepositional satellites in English "are mostly involved in the expression of Path", thus, perhaps, evoking vestiges of their basic spatial meaning. However, although certain kinds of inferential structures are preserved across meaning shifts like this, there is usually only a partial mapping. In fact, during the mapping, some new meanings may be gained. In the interest of space, I will be necessarily schematic about the exact nature of the mapping from the domain of space to the domain of action or the grammatical extension from spatial to aspectual usage for the prepositions being examined. The argument that prepositions in English behave aspectually and that their aspectual function is a natural and motivated extension of their basic lexical function is only convincing to the extent that I can demonstrate general similarities holding across the various prepositions and elucidate the regularity of the correspondence between location or motion in space and location or motion in time. Many historical linguists have argued that grammaticalization processes involve fairly regular semantic shifts, usually brought about through metaphorical extension (cf. Heine et al. 1991 for an extensive overview). One of the most common shifts cited has been that linking some property of the source domain of space (here, spatial location or motion along a path) with some property of the target domain of time (here, a state or the unfolding of intentional action) as schematized in Figure 1.

Aspects of prepositions and prepositional aspect

231

metaphorical mapping DOMAIN l SPATIAL LOCATION or ΜΟΉΟΝ ALONG PATH

DOMAIN 2 STATE or INTENTIONAL ACTION

LEXICAL GRAMMATICAL >=>> USE USE Figure 1. A common path of semantic extension and grammaticalization

In this vein, Traugott (1978, 1982), Bybee (1988), and Sweetser (1988) have demonstrated how it is possible to get aspectual meaning out of spatial terms via temporal usages. Furthermore, they believe that synchronic associations often recapitulate historical developments. I shall not summarize their arguments here. They have looked at some of the more obvious cases in English in which the preposition has actually lost most of its spatial thrust, as in the case of before or after or cases in which phonological or orthographic changes have precipitated or accompanied the addition of temporal or aspectual meaning, as in the case of against > again or through > throughout > thoroughly. What I would like to do is demonstrate the same phenomenon for less obvious cases, cases that have not undergone morphological mutation, or cases in which the propositions are still overwhelmingly spatial in their canonical usages.

3. 3.1.

The data After and RETROSPECTIVE aspect

I begin by looking at an unfamiliar case, but one which I hope will illustrate the aspectual potential of certain prepositions. In dialects of English spoken in Newfoundland and Ireland, one finds idiomatic, yet productive expressions with after like those given in (12): (12)

a. b.

I'm after breaking my leg / *I've after broken my leg. They're after closing down the fish plant and now everybody's out of work.

232

Sally A. Rice c. d.

e. f.

You're after spilling it on the floor and now you want me to clean it up?! The Provincial Government's after setting up this offshore petroleum impact committee. [St. John's Evening Telegram; 17 March 1980] Look what he's q/ter doing now! How many times am I after tellin' you?

Roughly, these constructions take the form [be after V-ing] and involve progressive inflected aspect with after inserted between the auxiliary be and the present participle. There are a number of semantic properties associated with this construction. First, the construction emphasizes that the event occurred in the immediate past and that it has current implications. As such, after can take a specifier like right or just, but modifying expressions signifying the distant past make little sense, as shown in (13): (13)

a. b.

I'm just/right after breaking my leg. I'm after hurting my back (*last year)

Thus, we could call after an indicator of retrospective aspect. Secondly, the construction conveys a notion of unexpectedness or that the event has or will have negative consequences. The source of this aspectual usage probably stems from the sense of sequentiality that is ever-present in both its spatial and temporal usages. Consider the purely spatial and temporal usages of after in (14) and (15), respectively: (14)

a. b.

He came after Tom in line. My house is just after the turn in the road.

(15)

a. b.

I'm usually home after 8 pm. They went out for coffee after the concert.

We could schematically represent the extension of after from subsequence in space to subsequence in time (where the TR in the target domain of action is understood to be some state or process) as in Figure 2. There are, in addition, constructions in standard English that more readily illustrate a close association between after and perfective meaning with progressive inflection, as shown in (16), so we should not be too surprised to find this same association in dialectal usages like those given above in (12).

Aspects of prepositions and prepositional aspect

233

metaphorical mapping

SPACE

ACTION

ubsequent location along path

tubscqueot location or occurencc In time

TR

~LM

TR

Spatial AFTER

LM

Temporal AFTER

Figure 2. Metaphorical origins of aspectual AFTER

(16)

a. b. c.

After lunch, he took a nap. After eating lunch, he took a nap. After having eaten lunch, he took a nap.

After even shows up as a sort of perfective conjunction in afterwards: (17)

Afterwards, he took a nap.

The point that I want to emphasize, though, is that in the after construction in Irish and Newfoundland dialects, the perfective is being signaled exclusively by after. The preposition undeniably has an aspectual function since its overall effect is to contradict the inflected aspectual morphology of the verb. In most of the other cases discussed below, the preposition more closely resonates aspectual properties present in the verbs, either by complementing them or by revealing the inherent aspectual character of the underlying event.

3.2. Away and CONTINUOUS aspect As a prepositional particle, away also has a clearly aspectual function, marking the continuation of an already ongoing activity, as shown in the examples in (18): (18)

a. b.

I've been working away on a new novel, They'll be partying away tonight as usual.

One can make a few preliminary observations about its distribution. First, it shows up most frequently with verbs inflected for progressive aspect, thus it

234

Sally A. Rice

nicely resonates semantically with overt verbal morphology. Compare (18a) with the sentences in (19): (19)

a. ?I worked away on a new novel all of last year. b. *?I have worked away on a new novel, but haven't made much progress. c. ?I will work away diligently all evening.

Secondly, it is confined to intransitive verbs (or transitive verbs with an omitted object): (20)

a. He's really snoring away in there. b. She's eating away like there's no tomorrow. c. *I'm still writing away a new novel. d. I'm still writing away on a new novel.

Moreover, it tends to co-occur with verbs denoting some sort of canonical, overt action: (21)

a. *?He's learning/teaching away like crazy this semester. b. *Don't bother him - let him think away in peace and quiet. c. ?Just look at him dreaming away in there.

Finally, it strongly disfavors verbs of motion with which it could take on an exclusively directional interpretation: (22)

a. ?You want to drive - drive awayl b. *He's such a maniac behind the wheel - he just speeds away no matter what the speed limit. c. *He just went away reading his book.

This last restriction is probably related to its spatial, source-oriented origins, as schematized in Figure 3. What might motivate an aspectual usage of away which highlights the continuation of some action? Away is used both prepositionally and adverbially to denote movement away from a source, whether that source location is specified or not. Some examples are given in (23): (23)

a. b. c.

He moved away (from the farm). She took their plates away (from the table). He's finally going away (from here).

Aspects of prepositions and prepositional aspect

235

metaphorical mapping

SPACE fource-based directed path

LM

TR

Spatial AWAY

ACTION continuing action

tune

Aspectual AWAY

Figure 3. Metaphorical origins of aspectual AWAY

d. e.

Away with him! Take it away, boys! [in show business sense]

These source-oriented, directional usages might help explain why aspectual away cannot co-occur with transitive verbs, which would otherwise invite a goal-oriented, perfective interpretation. Nevertheless, away figures in many transitive constructions of the form, [to V NP away] or [to Vaway at NP], as shown in (24), in which the verb usually expresses some sort of disintegrating activity. Away contributes something more to the overall meaning of these constructions than its aspectual sense, although it is present and salient. (24)

a. b. c.

Termites are eating my woodpile away I Termites are eating away at my woodpile. He kept on hammering away at the stack of bills. Government officials beavered away at the new legislation all winter.

These somewhat transitive constructions may sanction, or indeed contaminate, certain imperative transitive usages like You want to use my computer? Use it away!, but the majority of usages with aspectual away are intransitive and imperfective, as further exemplified in (25). These represent some of the more common usages of aspectual away. (25)

a. b. c. d.

He's saving away for a new house. She's planning away for her retirement. He's still smoking away, despite the doctor's warnings. Do you hear those two cats? They're really screwing away like crazy out there!

236

Sally A. Rice e.

You need to keep typing your report on my computer? Go ahead, type away I *type it away I *type on it away!

In general, the emphasis is on an activity that is or was in progress and far from over. Except in a few imperative forms, it doesn't really convey incipience, rather more of a sense that some event is ongoing and continuing along as planned or expected, but without any real endpoint or goal.

3.3.

On and RESUMPTIVE aspect

Turning now to a different case, that of on, we can make a few general observations. First, aspectual on signals a slightly different type of imperfectivity than does aspectual away. Whereas away signals what I might call "unfettered" continuation of an activity - a sort of default progressivity, on tends to signal a perseverance in the face of disturbance or an expectation of stopping, thus conveying a more effortful or purposeful resumptive progressivity. The difference is subtle, but fairly perceptible. Compare the pairs in (26): (26)

a. b. c.

She studied away I on all night. We were forced to toil laway I on despite the many hardships. We staggered 7away I on in the dark in search of shelter.

As described in the previous section, away conveys a slightly more effortless or purposeless continuation than does on, which suggests that, as an aspectual particle, on selects more force-dynamic predicates. This may be more evident in the examples in (26b and c). A second observation has to do with verbal aspect. Curiously, on slightly disfavors progressive verb forms, preferring simple tense inflection instead: (27)

a. He read on (despite the noise). b. *He has read on. c. ?He is reading on.

Third, unlike the case of away, it is perfectly acceptable with verbs of motion like go, come, or move: (28)

a. b. c. d.

He went on reading. Come on\ We'll be late. He's finally moving on. She drove on while looking for an exit.

Aspects of prepositions and prepositional aspect

237

Four, because the emphasis is on the continuation of a process itself, on strongly disfavors overt direct objects, although there may still be a sense of an implicit object or goal: (29)

a. He sang on (to the end of his song), despite their criticism. b. *He sang his song on, despite their criticism. c. *He sang on his song, despite then" criticism.

(30)

a. ?She drove the car on, while searching for an exit, b. *She drove on the car, while searching for an exit.

(31)

a. *He cooked on with his soup. b. ?He cooked on, despite our nasty comments about his culinary talent.

We should not be at all surprised, then, that aspectual on also tends to disfavor verbs which are inherently telic (the way phrasal verbs are). It is not used with verb-particle constructions or with phrasal verbs, whether they contain on or not: (32)

a. *He spent the morning trying on shoes on. b. *He depended on her on, even though they no longer lived together. c. *She's still yelling at him on, despite the complaints from the neighbours.

Aspectual on may have originated with spatial usages tied to continued motion, as indicated in Figure 4 and discussed below. Nevertheless, the aspectual meaning conveys a sense that the action continues despite a possible interruption or expectations of cessation, as indicated by the broken line in front of the processual TR. The link between continued motion and continued actions seems especially strong in the case of on. First of all, there are many expressions in English with prepositional on used to indicate an ongoing journey through space: (33)

a. b. c. d.

on one's way on a trip / an errand / a mission an ongoing journey Onwards!

238

Sally A. Rice metaphorical mapping

SPACE

ACTION

continued motion along path

continued or resumed (purposeful) action

TR LM=palh

Spatial ON

Aspectual ON

Figure 4. Metaphorical origins of aspectual ON

More relevant, however, may be the fact that in earlier stages of English, one form of the progressive was overtly locative, formed with a verbal noun preceded by a locative preposition, most often on, but at and in also. In (34), I have given an example from the OED from the 14th century: (34)

While pe masse is on syngynge; [...] while be gospel was on redynge. (1387)

As Hatcher 1952, Bybee 1988, and others have noted, vestiges of this usage live on in colloquialisms like those in (35): (35)

a. b. c.

Α-hunting we will go. I'm α-hoping to see him before he leaves. He's a-fixin' to get a spanking.

or paraphrases like those in (36): (36)

a. b.

He is living / He is alive. He is praying /He is at prayer.

The point is, on, a locative term, has a long association with present participles, if not progressive aspect. Let me add that aspectual on may also have its origins in a host of idiomatic phrasal verbs with on like keep on, carry on, come on, go on, hold on, take on, continue on which denote continuity and which sometimes take gerundive complements. Many of these phrasal verbs with on can be paraphrased by a verbal form of the gerundive complement and aspectual on. (37)

a.

Keep on smiling (« Smile on).

Aspects of prepositions and prepositional aspect

239

metaphorical mapping SPACE pith traversal

ACTION action retrareral

TR

-TRLM=path Sine Spatial OVER

Aspectual OVER

Figure 5. Metaphorical origins of aspectual OVER

b. c.

3.4.

Carry on (working), soldier (« Work on, soldier). Continue on (with what you were doing).

Over and SEMELITERATIVE ("CORRECTIVE") aspect

Let us examine one more locative preposition that definitely carries aspectual meaning, the preposition, over. In general, aspectual over conveys a sense of iteration, originating no doubt from a spatial sense of traversing a path or, more exactly, retraversing an established path. This spatial sense seems to be metaphorically extended to the domain of intentional action to convey a sense of action retraversal, as shown in Figure 5. But unlike inflectional aspects or morphemes which signal iterative or habitual action, over tends to indicate a single, second occurrence of an action, i.e., semeliterative aspect. There is, however, something more to the meaning of over in its aspectual usage. It conveys a sense that the activity is being repeated for corrective purposes and that it is or was being completed in a slightly different way the second time around: (38)

a. b.

He did it over (« He redid it in a different way). I'm making the bed over as the blanket has come undone.

Let me first note some co-occurrence restrictions. First, the verbs over cooccurs with must be transitive and only semi-perfective, as shown in (39): (39)

a. b.

I made it over I *I made over. I wrote my paper over I *I wrote over.

240

Sally A. Rice

By semi-perfective, I mean that the emphasis must be to a lesser degree on the final stages of the event and more on the process itself, since over entails a correction. But, unlike the case with aspectual on or away, the process must be construed as bounded or able to reach some sort of conclusion so that it can be judged as successful or not, that is, as having necessitated repeating in the first place. Secondly, aspectual over does not co-occur with ditransitives or phrasal verbs, which tend to be inherently telic and, moreover, semelfactive (occurring only once and consisting of a single act): (40)

a. b. c. d.

*I gave John a haircut over. ?She is crocheting the sweater for me over. *We talked about the proposal over. *I cut up the chicken over.

Finally, being semeliterative (occurring only twice), over cannot occur with verbs prefixed by the iterative, re-: (41)

a. *I returned the book over. b. "They re-printed the book over. c. *I had the book re-covered over.

Perhaps it is easier to compare aspectual over with the aspectual meaning of again, a closely related, but different aspectual particle, than it is to characterize it directly. Both forms presuppose an original traversal of some path, that is, a completed turn, if you will. Unlike again, though, over is restricted to a fairly small class of verbs with particular properties. Over is very sensitive to a small "degrees of realization" window and only a single, second occurrence of the activity. Moreover, it requires volitionality on the part of the agent and a discernibly different outcome. Again, on the other hand, is fairly insensitive to different "degrees of realization", it is infinitely iterative, no volitionality is assumed of the agent, and it does not require a different outcome after each iteration. Compare the paired examples in (42): (42)

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

He did it over I again. She fixed it lover I again. She took her turn over I again. He played the cassette lover I again. He ruined it / wants it *over I again. She failed / fell asleep / finished *over I again. You seem to be resembling him *over I again.

Aspects of prepositions and prepositional aspect h.

241

I repeated myself *over I again.

Next, we turn to some conjoined prepositions having aspectual import.

3.5.

Conjoined prepositions

Some aspectually inclined prepositions or particles like on, off, over, and again, also surface in coordinated pairs which have their own aspectual force. Space does not permit me to go through these examples in detail, but I would like to note that many of these conjoined prepositions carry different specialized meanings than their singleton counterparts and that they are often used in contexts unavailable to the single form. Moreover, they signal very particular though, again, minor aspectual categories: (43)

On and on and EXTENDED DURATIVE aspect a. She droned on and on. b. He worked on and on through the night. c. She drove on and on looking for a motel. d. He talked on and on (on many subjects). [Compare: He talked on (*on many subjects).] e. He cooked on and on. [Compare: ?He cooked on.]

(44)

On and off and INTERMITTENT aspect a. They're together on and off I *on I *off I *on and on. b. He worked away at it on and off. c. He cooks dinner on and off. d. ?She likes him on and off.

(45)

Over and over and ITERATIVE aspect and again and again and HABITUAL aspect a. He tried to fix it (over and over I again and again I over and over again I *over I again}. b. I repeated myself {over and over I again and again I *over I again}. c. They saw each other {over and over I again and again I *over I again}. d. He failed the exam {over and over I again and again I *over I again}. e. He fell asleep {over and over I again and again I *over I again}.

242

Sally A. Rice f. g.

She burst into tears [over and over I again and again I *over I again}. He took his turn {over and over I again and again I over I again}.

Unlike singleton over and again, the differences between over and over and again and again are not very significant. Indeed, they often combine to form the sequence over and over again, as exemplified in (45a), which means roughly the same thing and has the same distribution.

3.6.

Prepositional conjunctions and INCREMENTAL aspect

Finally, I note in passing the existence of some conjunctive uses of the prepositions, after, upon, and by, which seem to carry an aspectual quality. In most cases, the conjoined elements are not restricted to time nominals, as shown in (46a-b), (47a), and (48a), for these prepositions can conjoin any nominal which figures as a salient component of some generic event or habitual activity, as shown in (46c-f), (47b) and (48b), or a nominal which stands metonymically for time, as in (48c-d). In general, these conjoined expressions convey a sense of incremental progression through and activity or even incremental accomplishment, depending on the meaning of the overall predication. (46)

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

time after time / day after day (after day) / week after week (week *before I *at I *on I *into week) one after another / step after step She had dream after dream about her ex-husband. He watched video after X-rated video. Woman after woman gave birth to a stillborn baby in the contaminated area. Baby after baby was born dead in the months following Chernobyl.

(47)

a. b.

For week upon week, nothing happened, He read book upon book about his hero.

(48)

a. b. c. d.

After the operation, she felt herself getting stronger day by day. One by one, the mourners filed past. We'll succeed if we take things step by step (« a step at a time). Little by little, we reached our destination.

Aspects of prepositions and prepositional aspect

4.

243

Conclusion and prospectus

Perhaps we can best take stock of the fairly pronounced interaction between a preposition behaving aspectually, morphological aspect, inherent aspect, and verb transitivity, by looking at just a few paradigmatic cases, as given in (49). The point to emphasize is that the acceptability of each sentence in (49) is different depending on the particular preposition or coordinated prepositions used. Each of these prepositions or prepositional collocations, in its aspectual guise, has more or less found a unique ecological niche. Just as the prepositions carve up space differently, so too do they differentially partition the aspectual territory.

(49)

a.

b.

c.

d.

e.

f.

g.

h.

She fell asleep {*away / *over / *on / again / on and on / on and off / over and over / again and again}. (fell asleep = INTRANSITIVE; INCIPIENT) He finished {*away / *over / *on / again / *on and on / *on and off / over and over / again and again). (finished = INTRANSITIVE; PERFECTIVE) She's continuing (?away / *over / on / again / on and on / *on and off / over and over / again and again}. (is continuing = INTRANSITIVE; PROGRESSIVE) He blinked {away / ?over / on / again / on and on / on and off / over and over / again and again}. (blinked = INTRANSITIVE; ITERATIVE) She ran {*away / ?over / on / again / on and on / ?on and off / over and over / again and again}. (ran = INTRANSITIVE; IMPERFECTIVE; DIRECTIONAL) He upset her {*away / ?over / *on / again / on and on / *on and off / over and over / again and again}. (upset someone = TRANSITIVE; PERFECTIVE) She killed him {*away / *over / *on / *again / *on and on / *on and off/ *over and over / * again and again}. (kill someone = TRANSITIVE; SEMELFACTIVE, TELIC) Since the police were unable to catch the serial killer, he murdered {away / ?over / on / again / on and on / on and off / over and over / again and again}. (murder = INTRANSITIVE; IMPERFECTIVE)

244

Sally A. Rice

This paper has not been about grammaticalization of prepositions per se, since the prepositions are already highly grammaticalized. My aim has been to highlight one such grammaticalized role that is often overlooked in catalogues of prepositional function: that of aspectual particle. This function is evidenced by a number of the English prepositions. Such prepositional heterosemy or polyfunctionality allows us to extend the range of aspectual categories that we can code in English and to make several fine-grained aspectual distinctions that verbal inflection is not always capable of making. The data I have presented here do not exhaust the entire inventory of prepositions being used aspectually. For example, along and back are two prepositions that must also be examined for aspectual content. I note, too, that both about and to as in She 's about to have a baby carry a definite prospective or impending quality. The point is that prepositions, as a class, have taken on the functional role in English of signalling aspect, especially for minor aspectual categories, and have done so rather seamlessly as extensions from their canonical spatial senses. While there is no denying that, when used aspectually, the semantic content of the prepositions is fairly schematic, such meaning is, however, present and discernible. Though semantically schematic and overwhelmingly grammatical in function, these aspectual particles are no less important to the overall meaning of the predications that they occur in than are their full-fledged lexical counterparts. And they do seem to be related to them semantically, although it remains to be seen empirically or experimentally how close or far native speakers assess the semantic relationships. While most cognitive linguists would probably never dismiss prepositions, in whatever guise, as empty particles, I urge that, among other use types, aspectual usages of prepositions be recognized and studied and integrated into the full panoply of prepositional meaning and function. It is only with a complete catalogue and understanding of all usage types that we can start to address the theoretical implications that the multivariate behavior of prepositions have on models of lexical representation and debates about prepositional monosemy, homonymy, polysemy, or far more probably, heterosemy.

Notes 1. The research project of which this paper constitutes a part is supported by grant #410-930205 from the Social Sciences and Humanities Research Council of Canada.

Aspects of prepositions and prepositional aspect

245

2. This paper benefitted greatly from helpful discussion following its presentation at the third conference of the ICLA, in Leuven, Belgium. Many discussants suggested that I use the term aktionsart (to reflect what they felt was a clearly lexicalized coding of inherent aspect) rather than aspect (which is usually reserved for grammaticalized expression of inherent aspect) in my title and throughout my talk when referring to the aspectual contribution of the prepositions. I have not chosen to follow their suggestions, intending instead, in characteristically nonEuropean fashion (cf. Comrie 1976: 6-7 or Frawley 1992, Chapter 7), that my use of aspect throughout this paper be interpreted neutrally to mean the way in which an event is construed as distributed through time (cf. Talmy 1985) rather than the more narrow sense of inflectional aspect marked on the verb. 3. Poustma 1926 used the term particle, which was defined as "inflectionally inert", to cover, among other things, prepositions being used grammatically. 4. Formal properties that have been used to distinguish satellites from prepositions, with which they share phonetic content, are closer placement vis-ä-vis the verb, their receiving heavy stress vs. the light stress prepositions receive, their ability to not take an overt complement, and the possibility of their being "topicalized" or otherwise "dislocated". Cf. Talmy 1985 or Rauh 1991 for further discussion. 5. A number of arguments, such as the felicity of the preposed PPs, are usually advanced for treating the in-type phrases as verbal modifiers and the far-type phrases as sentential ones. Although such structurally-based distinctions hardly constitute a thorough semantic analysis, I will not be treating in and for phrases in this paper.

References Bennett, David C. 1975 Spatial and Temporal Uses of English Prepositions. London: Longman. Brugman, Claudia 1981 The story of OVER. MA thesis at the University of California at Berkeley. Bybee, Joan 1988

Semantic substance vs. contrast in the development of grammatical meaning. Berkeley Linguistics Society 14: 247-264.

Comrie, Bernard 1976 Aspect. Cambridge: Cambridge University Press. Frawley, William 1992 Linguistic Semantics. Hillsdale, NJ: Lawrence Erlbaum. Genetti, Carol 1986 The development of subordinators from postpositions in Bodic languages. Berkeley Linguistics Society 12: 387-400. Hatcher, Anna 1952 The use of the progressive form in English. Language 29: 254-280.

246

Sally A. Rice

Hawkins, Bruce 1984 The semantics of English spatial prepositions. [Unpublished PhD diss., University of California at San Diego.] Heine, Bernd, Ulrike Claudi and Friederike Hünnemeyer 1991 Grammaticalization: A Conceptual Framework. Chicago: University of Chicago Press. Herskovits, Annette 1986 Language and Spatial Cognition: An Interdisciplinary Study of the Prepositions in English. Cambridge: Cambridge University Press. Janssen, Theo 1993 Heterosemy or polyfunctionality of Dutch maar 'but, only, just'? Berkeley Dutch Linguistics Conference. Lakoff, George 1990 The invariance hypothesis: Is abstract reason based on image-schemas? Cognitive Linguistics 1: 39-74. Langacker, Ronald 1987/1991 Foundations of Cognitive Grammar, Vol. I-II. Stanford: Stanford University Press. 1990 Concept, Image, and Symbol. Berlin/New York: Mouton de Gruyter. 1992a Prepositions as grammatical(izing) elements. Leuvense Bijdragen 81: 287309. 1992b The symbolic nature of cognitive grammar: The meaning of of and of ofperiphrasis. In Martin Pütz (ed.), Thirty Years of Linguistic Evolution, 483502. Amsterdam: John Benjamins. Lichtenberk, Frantisek 1991 Semantic change and heterosemy in grammaticalization. Language 67: 475509. Lindner, Susan 1981 A lexico-semantic analysis of English VPCs with UP and OUT. Unpublished doctoral dissertation, University of California at San Diego. Persson, Gunnar 1988 Homonymy, polysemy, and heterosemy: The types of lexical ambiguity in English. In K. Hyldgaard-Jensen and A. Zettersten (eds.), Proceedings of the Third International Symposium on Lexicography, 269-280. Tübingen: Niemeyer. Poutsma, Hendrik 1926 A Grammar of Late Modem English, vol. 5, The Verb and Its Particles. Groningen: P. Noordhoff. Radden, Gunter 1985 Spatial metaphors underlying prepositions of causality. In Wolf Paprottd and Reno Dirven (eds.) The Ubiquity of Metaphor, 177-207, Amsterdam: John Benjamins. Rauh, Gisa 1991 Prepositional forms in the lexicon. In Rauh, Gisa (ed.), Approaches to Prepositions, 169-223, Tübingen: Gunter Narr Verlag.

Aspects of prepositions and prepositional aspect Rice, Sally 1992

1993

247

Polysemy and lexical representation: The case of three English prepositions. Proceedings of the Fourteenth Annual Conference of the Cognitive Science Society, 89-94. Hillsdale, NJ: Lawrence Erlbaum. Far afield in lexical fields: The English prepositions. ESCOL '92 Proceedings, 206-217. Ithaca: Cornell University Press.

Sweetser, Eve 1988 Grammaticalization and semantic bleaching. Berkeley Linguistics Society 14: 389^105. Talmy, Leonard 1985 Lexicalization patterns: Semantic structure in lexical forms. In Shopen, Timothy (ed.), Language Typology and Syntactic Description, Vol. Ill, 57-149. 1991 Path to realization: A typology of event conflation. Berkeley Linguistics Society 17: 480-519. Traugott, Elizabeth 1978 On the expression of spatio-temporal relations. In J. Greenberg (ed.), Universals of Human Language, vol. 3, 370-400. Stanford, CA: Stanford University Press. 1982 From prepositional to textual and expressive meanings: Some semantic-pragmatic aspects of grammaticalization. In W.P. Lehmann and Y. Malkiel (eds.), Perspectives on Historical Linguistics, 245-271. Amsterdam: John Benjamins. Vandeloise, Claude 1991 Spatial Prepositions: A Case Study from French, Chicago, University of Chicago Press. Vendler, Zeno 1957. Verbs and times. Philosophical Review 66: 143-160. Reprinted in Vendler, Linguistics in Philosophy (Ithaca: Cornell University Press), 97-121.

Some aspects of relational nouns Yoshiko Tagashira

1.

Introduction

Relational nouns (hereafter RNs) exhibit both morphological and syntactic properties of nouns, although the degree of nominality varies among languages and also among the individual items in the lexicon of a given language. The nominal properties exhibited by at least some RNs include: taking demonstratives, taking adjectival qualifiers, permitting relativization, and forming sentential subjects (Heine et al. 1991: 133-4). At the same time RNs are distinguished from regular nouns in that each one takes a certain relationship as its background concept and defines some entity which exists within this relationship in reference to another entity also within this relationship. Thus, the English RN pet, for example, is defined in terms of the relationship of ownership and in reference to the owner who also participates in the relationship (Langacker 1990: 179). Another example is a set of locational nouns in many African languages (Heine et al. 1991); ta'-me On' of Ewe, for example, is balanced by gsme 'under', and both are conceptualized in the relationship of vertically oriented space.1 The two basic properties of RNs mentioned above, i.e., their nominality and relationality, have been pointed out by various scholars including Langacker (1987, 1990, 1991) and Croft (1991). Beyond the recognition of these properties, however, the most important work on RNs appears limited to the detailed study of locational RNs in African languages by Heine et al. (1991). In the course of their study of the phenomenon of grammaticalization, the authors examine the wide range of meanings expressed by the locational RNs and explain them as the result of grammaticalization processes that these RNs have undergone. The Japanese language also employs constructions involving RNs to express location. Furthermore, the phenomenon of grammaticalization discussed by Heine et al. (1991) in relation to African languages seems to be closely paralleled by Japanese. At the same time, however, the Japanese RNs examined in the present paper exhibit features which are different from those observed among the RNs of African languages, and present us with some

250

Yoshiko Tagashira

questions of their own that must be discussed if we are to understand the nature of the linguistic category of RN. In this paper, we propose, in particular, to (1) discuss the peculiarities of the nouns which serve as ground in the locational expressions, (2) examine the etymologies of the RNs and discuss their nominality, (3) investigate the evidence of grammaticalization as exhibited by the RNs, and finally (4) in order to clarify further the properties of the RNs, discuss the difference between the construction involving the RN and the one without it.

2. Data We will begin by presenting some basic data of the RNs of Japanese.

2.1. Locational expressions The locational expressions of Japanese take the following two forms: (1)

'Noun + Particle, no + RN + Locative Postposition (e.g., yuka 'floor' + Particle, no Of -f RN, ue On' + Locative Postposition, ni 'at' = On the floor')2

(2)

'Noun + Locative Postposition' (e.g., yuka + ni = On the floor')

2.2.

The nominality of the RNs

We will examine the degrees of nominality of the RNs by checking each RN against the parameters which were used by Heine et al. (1991) for determining the nominality of various postpositions of Ewe (1991: 133). The results are presented in the Nominality Chart below, which follows the same format as their Table 5.5. (1991: 134). The nature of the RNs will be examined in detail later in Section 4, when the data given in the chart above will be discussed. For the present, we will simply point out the three outstanding features shown in the chart and briefly comment on the parameters where all the RNs have 'minus'-values: 1. RNs are rather uniform as regards their nominality, i.e., in most cases they have the same values ([+] or [-]) within a particular parameter, and discrepancies can be observed only in four cases.

Some aspects of relational nouns

251

Table 1. Nominality chart RN§s Cognitive concept English translations

sita

mae FRONT front

on, above

UNDER under, bottom, below

-

—

_

ue ON top,

uti IN

usiro BACK back, behind

naka

IN in,

inside,

inside

in

_

_

soto OUT out, outside

Parameters ( a ) numerical distinction (b) take demonstratives (c) take adjectival quantifiers (d) permit relativization (e) form the sentence subject (f) take first-or second-person possessive pronouns (g)genitival morphology (h) permit relativization when qualified by a genitive noun phrase (i) form the sentence subject as the head of a genitive noun phrase (j) take third-person possessive pronoun as modifier

_

_

+ -

4—

+ _

+ _

+ _

+ +

_

+

+

+

+

+

+

+

—

—

-

—

+

+

+

-

-

-

-

+

-I-

+

4-

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

—

—

_

_

_

_

)

.

_

2. With most of the parameters, most RNs have [+] values, indicating their high nominality. 3. The inability 'to express a morphological number distinction' is readily explained. It is a feature which applies not only to the RNs but to all Japanese nouns.

252 2.3.

Yoshiko Tagashira Etymology of RNs

The RNs are each perceived and used as one word in present-day Japanese, but etymologically all except one must be analyzed into constituent morphemes. Etymologies are indicated to the right of the colon. ue: sita: mae: usiro: naka: uti: soto:

u 'above' + he 'side'3 si 'under' + ta 'direction' ma 'eye' + he 'side' mu 'body' + siri 'buttock', which was commonly pronounced usiro; cf. strike: siri 'buttock' + he 'side' na 'inside' + ka 'place' «— utu 'sky' shortened form of soto 'back (of the house)' + mo 'side'

The important thing to note about these RNs is that, with the exception of uti and usiro, they denote 'an area' or 'a direction', rather than concrete objects; mae, for example, does not represent the eyes themselves but a direction or an area to which the eyes are directed. One of the 'exceptions', namely uti 'sky', can also be understood as denoting an area; cognitively, an unbounded area of 'sky' above us is of the same kind as an unbounded area extending far into the space in front of us denoted by mae. As for usiro, it was used to designate a part of the body, while sirihe was the locational RN representing an area. Later, from the 13th century on, sirihe disappeared and its meaning was taken over by usiro (Miyaji 1978: 105-7). Thus, it appears that in usiro, we find the case where a lexeme representing a part of the body has come to be used for locational meaning. However, even in this case, it is what Svorou calls 'relational object part': it is an inherent part of an object and "may not be individuated and separated from the object itself." (Svorou 1993: 83) Heine et al. (1991) propose a hypothesis regarding the likelihood of locational words being derived either from landmark models or body-part models. The hypothesis states that, when the SPACE concepts are "arranged on an implicational scale o f . . . UNDER > ON, IN > FRONT > BACK, if any of these spatial concepts is derived from body-part models, then none of the concepts to its right may be derived from landmark models." (1991: 12930) Our data above support their hypothesis; the RNs representing FRONT

Some aspects of relational nouns

253

and BACK are derived from body-parts words while those to the left of FRONT in their scale (i.e., IN, ON, UNDER) are all related to landmark features.

3. PLACE as Ground As noted in 2.1., one kind of locational construction involving an RN takes the form of 'Noun + Particle, no + RN + Locative Postposition': e.g., yuka no ue ni '(floor + of + on + at -») on the floor'. In this section, we will examine the properties of the nouns which occur as the first 'Noun' in the above formula. I will demonstrate that most of these nouns denote PLACE rather than OBJECT, and will refer to them as PLACE nouns. Before examining the nature of PLACE nouns, we should determine the semantic composition of the above structure involving a PLACE noun and an RN. The PLACE noun is attached to the particle no, which is used for the expression of a wide range of meanings. When no is followed by a word representing a relational concept (like RN), it serves to show that the preceding PLACE noun is the entity to which this relational concept is applied; thus, in yuka no ue, the PLACE noun, yuka 'floor' is the entity in reference to which the concept of ue On' is defined. In other words, the RNs profile a SPACE and the conceptualization of each of these RNs is defined relative to the entity represented by the PLACE noun. In connecting a PLACE noun to an RN, the function of no is to provide an entity to which the understanding of the following RN is conceptually dependent.4 The nouns which precede no and RNs are numerous and varied in kind, ranging from landscape entities (e.g., yama 'mountain', unkai 'sea of cloud') to parts of the body (e.g., atama 'head', hana 'nose') to artificial objects (e.g., sen 'line', tukue 'desk'). I would claim that these nouns designate PLACE rather than OBJECT, and that the aforementioned locational expression (yuka no ue), which contains a PLACE noun and an RN, identifies the type of PLACE and a particular facet of this PLACE to which a certain located object (i.e. Figure; see below) relates. In the following, we will examine how this concept of PLACE is distinguished from that of OBJECT. First of all, let us consider OBJECT. An OBJECT is a three-dimensional bound entity, the volume of which can be determined. Examples are plentiful, of course, but let us consider just a few. A carpet hung up in a store as an article for sale: it is a clearly bound OBJECT with

254

Yoshiko Tagashira

a definite length, width and thickness. Another example is a pencil the diameter and length of which can of course be precisely determined. The third example comes from a game and is a three-dimensional mountain-shaped human nose made of paper, which the blind-folded players are supposed to put in the center of a picture of a human face. Being three-dimensional, bound and having measurable volumes, each of these OBJECTS can be described by all or some of the following adjectives, which refer to their sizes: e.g., atui zyuutan 'thick carpet'; mizikai enpitu 'short pencil'; ookii hana 'big nose'. Now, let us consider PLACES.5 First, a PLACE provides a location where human activities are carried out. If we buy the aforementioned carpet and bring it home, we can place it on the floor, sit on it and play games there. In this situation, a carpet is a two-dimensional expanse which is used as a PLACE for human activity; its thickness is ignored and also its boundaries (both lengthwise and widthwise) are also unspecified. The important thing is that people can be 'within' the area defined by the carpet and carry on an activity there. Another example is an automobile, which is a three-dimensional OBJECT when placed in a showroom, but when people sit inside it and move from one place to another, it provides them with a PLACE for this activity. What we saw above is that the same entity can be an OBJECT or a PLACE, depending on whether it is viewed as something existing independent of human activity or whether it is viewed within the framework of human activity.6 OBJECT and PLACE do not exist as two distinct, objective entities in the metaphysical world; they exist in our perceived world and identified as OBJECT or PLACE on the basis of people's knowledge and experience.7 There are PLACES which are defined by the feature of being a part of the 'main body' and not separated from it. The most basic main body is the ground on which we exist. Now, if we consider such natural entities as 'lakes', 'fields' and 'mountains', they are a part of the ground, since (i) they cannot be moved and (ii) they are extensions of the ground and their borders are not marked.8 Another group of examples is constituted by parts of the body, the main body of which is the human body. Compare two kinds of noses: the one mentioned above (which is made of paper and used in games) and the one on a human face. The latter is an extension of the surface of the face and its boundary is not specified; it is merged with its surroundings. Unbound objects such as zyuusu 'juice' can be mentioned here also. In Zyuusu no naka ni doku ga mazete aru (juice + of + inside + at + poison + Nom -f mixed + be —>) The poison is mixed into the juice.' The poison

Some aspects of relational nouns

255

is not in one spot in the liquid; it is mixed with the juice and therefore is dispersed throughout it. The conceptualization of 'juice' in such a sentence is a PLACE where the mixing of the poison took place. The entities in the third group are PLACES by virtue of the fact that people can put things on or in them: e.g., tables, chairs, shelves, slippers, buckets and vases. The function of these PLACES is to keep the things at a constant distance from the ground or from the base point (whatever that may be). Sitting on a chair, for instance, one is at a certain distance from the ground, but this distance is constant. With regard to buckets, the water is kept at a constant distance measured from the rim. Thus, we should distinguish between OBJECT and PLACE. This conclusion is supported by the existence of two kinds of adjectives. An OBJECT, but not a PLACE, can be ookii 'big' or tiisai 'small'. PLACES must be qualified only by the following adjectives: hiroi 'spatious' or setnai 'narrow', or hukai 'deep' or asai 'shallow'. In the linguistic description of the relative location between two objects, the position of the located object is identified in terms of the reference object. The two objects are called, respectively, the Figure and the Ground. Thus, in the sentence Ki no süa ni goza o siita. '(tree -f of + under + at + sheet + Ace + spread ->) We spread the sheet under the tree.', 'the sheet' is the Figure and its location is identified with reference to 'the tree', which is the Ground. In the locational expressions involving RNs (in bold face in the last example), the Figure, goza is found outside the locational expression, but the Ground is within the expression and is represented by the PLACE noun ki. The purpose of this section was to demonstrate that the noun which precedes no + RN is represented by the nouns which identify PLACE rather than OBJECT. In the light of what has been noted above, we can now state that the Ground of a locational expression designates a PLACE, and the RN specifies a particular dimension or facet within that PLACE. We concluded above that the no + RN is preceded by a noun which represents PLACE and that this PLACE functions as Ground. We must now ask whether the nouns which precede no + RN can denote an OBJECT and provide a Ground. An OBJECT is three-dimensional, movable, and not a part of the main body. The bold face nouns in the following examples appear to represent an OBJECT: Kodomotati wa sensei no usiro ni kakurete simatta. '(children + Top + teacher + of 4- behind + hide + finished ->) The children hid behind the teacher.'; Karasu no go-meetoru ue o tobi ga tonde iru. '(crow + of + 5 meters + above + kite -f Nom + fly 4- be —>·)

256

Yoshiko Tagashira

The kite is flying five meters above the crow.' Given the evidence of these sentences, we cannot justify our claim that Ground is not represented by OBJECT.

4.

Nominality of RNs

In 2.2. above, we indicated the degrees of nominality of the Japanese RNs in the Nominality Chart by adopting the nominality-parameters proposed by Heine et al. (1991: 133) At the end of the Chart, we made the following three general observations about them: (i) their overall high nominality; (ii) the relative uniformity among them; (iii) their inability to make a morphological numerical distinction, which is a general restriction applicable to all Japanese nouns. Now, let us look more closely at the Chart and examine some of the facts, which may shed light on the nature of the RNs. A. Parameter (b), i.e., RNs can take demonstratives. RNs in Japanese are preceded by the demonstratives kono 'this', ano 'that (over there)', sono 'that' and dono 'which'. (These demonstratives should be distinguished from the demonstrative pronouns kore, sore, are and dore.) Until about the end of the 10th century, these words were each 'pronoun ko (or a or so or do} + no Of ', although in present-day Japanese, they are used exclusively as demonstratives. That is, in earlier times, sono mae, for example, meant so no mae '(that thing + of + front —>) in front of that thing' rather than 'that (= demonstrative) front'. We should note, however, that when the noun which follows the demonstrative is an RN, there is no conceptual difference between 'a demonstrative + RN' and 'a demonstrative pronoun + of + RN': the only possible conceptualization of the phrase, sono mae '(demonstrative, that + front —») that front' is 'in front of that thing', which is identical with the conceptualization we get from sore no mae '(pronoun, this + of + front —>·) in front of this thing'. It is not surprising, then, that the earlier construction, so no RN 'that thing + of + RN' came to be used as 'demonstrative, sono + RN'. A demonstrative can not cooccur with soto Outside'. The conceptualization of INSIDE/OUTSIDE is as follows: an egocentric territory is INSIDE, which is surrounded by OUTSIDE. It seems that the territory defined as INSIDE can not be referred to by a pronoun, ko (or so or do) and the construction 'demonstrative -I- RN' (i.e., kono soto, sono soto, ano soto, dono soto) cannot be conceptualized.

Some aspects of relational nouns

257

B. Parameter (c), i.e., RNs cannot take adjectival quantifiers. It has already been pointed out that RNs are nouns which denote 'a direction' or 'an unbounded area', rather than concrete objects. Moreover, we have discussed in Section 3 above that their function is to bring into focus the shape or dimensionality of the ground. As such, the RNs can not be quantified by adjectives.9 Adverbs, on the other hand, can modify RNs, since, instead of quantifying an area, they specify where this focused area is located. For example, in Tukue no sugu sita ni otita, '(desk + of + adverb, immediately + under + at + fell —») It fell right below the desk', the adverb sugu indicates the proximity between the area designated by sita and the desk (in relation to which the area of sita is defined). Another example: Watasi no naname mae ni suwatta, '(I + of + adverb, diagonal -I- front -f at + sat —>·) He sat diagonally in front of me.' C. Parameters (d) and (h), i.e., RNs permit relativization, with or without qualification by a genitive noun phrase. The examples where an areaindicating RN is relativized are rather few and all of them appear before the eleventh century. Only one example will be cited: Tada itaru ue yon, tori no takaku nakite iku, where Tada itaru is the relative clause modifying ue, '(simply + lying down + above 4- from + birds -I- Nom + high -I- shrieking + go —>·) Above the place I was just lying on, the birds flew away shrieking with high voices.' (from Makura no Soosi, which is a collection of essays completed around 995).10 As for the parameter (h), i.e., ability to be relativized when the RNs are preceded by a genitive noun phrase, the value is [+] for all words. This is to be expected, when we recall that a demonstrative, sono 'that', for example, has derived from sore no '(that thing + of —>·) of that' and includes the meaning of the pronoun sore 'that thing'. That being the case, in the construction 'relative clause -f sono + ue\ the relative clause is actually modifying the pronoun so which has been absorbed into the demonstrative sono. In other words, a relative clause is modifying the pronoun so, and there is no reason that an RN cannot be preceded by a relative clause, so long as there is this sono, the pronoun-including demonstrative between them. One example is cited: Kare no iru gyooretu no sono soto ο arukinasai, '(he + of + exist + procession + of + that + outside -I- Ace + walk —>·) You'd better walk outside the procession which he is in.' As we recall, the RN soto Outside' cannot be preceded by a demonstrative. However, as we can see in the last example, when the relative clause is followed by the demonstrative, sono 'that', it modifies this demonstrative (more correctly, the pronoun, so of sono) and

258

Yoshiko Tagashira

soto can be preceded by the demonstrative. We can conceptualize a definite area perceived as INSIDE (gyooretu 'procession' in the present example), and OUTSIDE must be an expanse surrounding this INSIDE area. As we will see in Section 5, RNs are used for meanings which belong to the category of TIME. When used for this meaning, RNs are more freely preceded by relative clauses. Several examples will be cited: /"...] ko aru naka nan kereba [...] 'child 4- exist 4- inside + be 4- and so -4) [...] since their relationship is such that they have borne a child [...]' (from Isemonogatari, the collection of poems; 10c.); Yo o sarinan usiro no koto sirubeki koto ni arane do [...] '(world + Ace 4- leave + back + of 4- thing + could know + thing 4- at + not so + although —») although I cannot know about the world after I die [...]'; Ikite iru uti ni ome-ni-kakari tai, '(alive + be -I- inside 4- at 4- see 4- want —>) I want to see you before I die.'; Yoku kangaeta ue de syootisita, 'well + thought + on + be + agreed ->) After I well thought about it, I agreed.' D. Parameters (e) and (i), i.e., an RN can be a sentence subject, either by itself or as the head of a genitive noun phrase. We propose to consider the second parameter first. As the head of a genitive noun phrase, an RN is preceded by 'PLACE noun + no\ and, as illustrated in the following example, in this context, it can be the subject of a sentence. In the example Sen no soto ga kikentiiki desu '(line + of + outside 4- Nom + danger zone 4- be —») Outside the line is the danger zone,' the subject is the noun phrase sen no soto and the RN soto is the head of this phrase. This property is a feature of all RNs. As regards the ability to be a subject by itself, this is limited to the three RNs soto, uti and naka. Furthermore, each of them is used with one definite meaning; soto, for instance, can only mean Outdoor' and no other plausible meanings such as Outside world' or 'exposed surface of an object' are admissible. Such is the case also with uti and naka, which have already been discussed. Thus, these three words must be used in an appropriate context in Japanese. Other RNs which occur by themselves as subjects should be viewed as the result of the deletion of some words. We cite a few examples: (i) deletion of a pronoun: (Beddo wa, ue no ga ii —>) Beddo wa, ue ga ii. '(bed + Top 4- up + one 4- Nom + good —>·) As for beds, I prefer the upper bunk.' (ii) a part of a PLACE is designated but the PLACE is only implied: e.g., (Anata no karada wa mae ga nurete im —>) Mae ga nurete iru. '(you 4- of +

Some aspects of relational nouns

259

body + Top 4- front + Nom + wet + be —>) The front part (of your body) is wet.' E. Parameter (g), i.e., existence of genitival morphology. The [+]value recorded for this parameter means that an RN can occur without a preceding PLACE noun. The fact that this is possible is explained by the situational context which supplies the PLACE: usiro no hon '(back + of + book ->) the book which is behind' is understood to mean, 'the book behind the PLACE, which is most likely to be the listener.' F. Remaining parameters (f) and (j), i.e., ability to take first-, second-, thirdperson possessive pronouns. There is only one RN which can denote entities that can be 'possessed' by people. It is uti 'house', which we find in examples such as Watasi no uti dake yaketa. '(I + possessive particle + house + only 4- burnt down —») Only my house was burnt down'. In the locational phrase introduced at the beginning of the paper, RNs are preceded by no, which is homogeneous with the possessive particle. As mentioned earlier, however, the pre-RN no has a different function, which is, by connecting the noun which precedes it to an RN, to provide an entity on which the understanding of the following RN is conceptually dependent.

5.

Grammaticalization of RNs

In the case of the RNs found in many African languages, grammaticalization is a process in which words belonging to the domain of OBJECT come to represent entities in other domains including LOCATION, TIME, QUALITY and CAUSE (Heine et al. 1991). Can we also observe this kind of grammaticalization with Japanese RNs? We saw in Section 2 that all RNs except one are compound words, each consisting of a morpheme denoting a part of the body or a part of spatial orientation, and a suffix meaning 'side', 'part', 'place' or 'area'. This means that these RNs are already locational expressions. The RN uti 'inside' is an apparent exception. It has derived from uti 'sky'. This too can be understood as denoting an area. Cognitively, an unbounded area of 'sky' above us is of the same kind as an unbounded area extending far into the space denoted by mae 'front'. The result is that none of our RNs has been clearly derived from OBJECT. Related to this fact is what I call the generic usage of RNs. In this usage, the RNs are used by themselves without a preceding PLACE noun: cf.

260

Yoshiko Tagashira

ue e mairimasu '(This elevator is) going up'; usiro o hurimuita 'He turned back'; (momo wa) naka ni tone ga aru '(Speaking of peaches,) there is a stone inside.' In these examples the RNs denote a direction or an area in the basic parameters of human conception: (i) we recognize the direction of the pull of gravity and identify the UP/DOWN axis and then the areas of ue and sita; (ii) recognizing the cardinal movement of human beings and their direction of sight, the opposition of FRONT/BACK is established; (iii) the area bound by UP/DOWN and FRONT/BACK is naka; (iv) an egocentric territory is INSIDE, which is surrounded by OUTSIDE. This generic usage of RNs indicates the basic nature of RNs, namely that they are in themselves locational expressions. RNs represent entities of various categories, some of which are less concrete than others. Thus, grammaticalization on a synchronic basis is indeed observed with RNs. At the same time, some semantic differences are such that they do not belong to separate categories but instead consists of focused meanings within particular categories. In the following, we list the various focal senses expressed by RNs, starting from those of least degree of grammaticalization and moving on to those of advanced grammaticalization. The identification of each focal meaning is followed by a brief definition or explanation of it and then the example sentences containing the RNs used for this meaning. 1. Generic meaning. People recognize the various spatial orientations in the world in which they exist and refer to the directions or areas defined by them. In this usage, the RNs are never preceded by a 'PLACE noun + no\ (3)

Moo-sukosi ue o muite kudasai. a little more on Ace look to please 'Please look up a little more.'11

(4)

Rooraakoosutaa kam sita o miru to kowai. roller coaster from under Ace look if frightful 'It is frightful to look down from a roller coaster.'

(5)

Mae ο mite aruki nasai. front Ace look walk Command 'Look ahead of you when you walk.'

(6)

Dareka ga usiro kara tikayotte kuru ki-ga-sita. someone Nom behind from approach come felt Ί felt someone was approaching me from behind.'

Some aspects of relational nouns

(7)

261

Naka o kurinuite kawa dake nokosimasu. inside Ace scoop out and skin only retain 'You will scoop out the inside and make it hollow.1

(No examples were found with uti and soto.12) 2. Object. The only RN which can represent a three-dimensional object and thus be modified by such adjectives as ookii 'big' and tiisai 'small' is uti. (8)

Taterareru uti no ookisa wa hooritu de kimatte im. buildable house of size Top law by decided be The size of a house you can build is regulated by law.'

(9)

Soto e dete asobinasai. outdoor to go out play 'Go outside and play.'13

3. Part of an object. The RN designates some part of an object, which is represented as PLACE. The boundary of this part is unspecified. (10)

Hon nara tukue no ue ni arimasita yo. book if be desk of on at existed Sentential Pel. 'If you are looking for the book, it was on the desk, I tell you.'

(11)

Saka no sita de hitoyasumi siyoo. hill of under at a rest let's do 'Let us rest for a while at the foot of the hill.'

(12)

Rikisi wa atamanimage ο yui mae wa aoku sumo wrestlers Top head at topknot Ace make front Top blue sotteiru. is shaved 'Sumo wrestlers have the topknot on their heads and their fronts (=foreheads) are shaved blue (=so close that they look blue).'

(13)

Kuruma ni wa usiro yori mesare soorae. carriage to Top back from get on Command 'When getting into a carriage, get into from its back.' (from Heike Monogatari 'Tale of the Heike' of early 13c.)

(14)

Atumatta hito nonaka nikarega ita. gathered people of inside at he Nom was 'He was among the people who gathered there.'

262

(15)

Yoshiko Tagashira

Senro noutio arukuno wa kiken da. railway track of in Ace walk thing Top dangerous be 'It is dangerous to walk between the rails.'

(No example was found with soto Out'.) 4. Area in the immediate vicinity of ah object. (16)

(17)

(18)

(19)

(20)

Kami-hikooki wa watasi no atama no sugu ue o kasumete paper-plane Top I of head of right on Ace scoop tootta. passed 'His paper aeroplane flew past right above my head.' Taipuraitaa no sita ni taoru o siite oto o yawarageta. typewriter of under at towel Ace spread sound Ace soften Ί placed a towel under the typewriter and thus softened the sound (of typing).' Fendaa ni sawatte, kuruma no mae de asondeita kodomo ga fender to touch car of front at was playing child Nom kega-o-sita. was hurt 'The child who was playing in front of the car touching its fenders was hurt.' Kuruma no usiro wa musubitukerareta takusan no car of back Top be tied to many of kekkon-sitabakari no kan ya omotya de, yakamasii. marriage-just done of can or toys by, noisy 'The back of the car was noisy because of the cans and toys which were painted 'Just Married' and were tied to it.' Kono kusuri wa doku ο karada no soto ni dasu. this drug Top poison Ace body of out to take out 'This drug brings out the poison to the surface of the body.'

(No examples are found with uti 'in' and naka 'in'.) 5. Area at a distance from an object. A placed object need not be in the vicinity of the PLACE relative to which the location of the former is identified by RN. Despite a certain amount of distance between PLACE and the placed object, the two constitute a perceptual unity.

Some aspects of relational nouns

263

(21)

Herikoputaa ga soonangenba no ue o senkaisiteiru. helicopter Nom accident-site of above Ace be going around Ά helicopter has been flying around above the site of the accident.'

(22)

Sakura no ki no sita de ohanami o sita. cherry blossom of tree of under at flower-looking Ace did 'We admired the blossoms underneath the cherry trees.'

(23)

Kare nara, watasi no haruka mae o hasitte iru. he if be I of far away front Ace running be 'If you are talking about him, he is running far ahead of me.'

(24)

'Sanzyaku sagatte si no kage sae humazu' Ί meter withdraw teacher of shadow even not step on' yo. Hanarete sensei no zutto usiro o aruki nasai. you know, be away teacher of clearly back Ace walk Command 'You know they say 'You should walk some way behind the teacher so as not to tread on his shadow. Walk far behind the teacher.'

(No examples were found with RNs referring to IN/INSIDE/OUTSIDE.) 6. An individual's perception comes into play. The central meaning intended here is that usage depends not only on physical location but whether something is visible or perceptible by people. (25)

Hito no miru ue wa musubite, ... people Nom see on Top tie 'He ties the sash on the upper side which people see.' (from Manyoosyuu, a collection of poems before 12th century; in present-day Japanese, this meaning is found only in compound nouns containing ue.)

(26)

Tenugui ο boosi no sita ni siite, atusa o sinoida towel Ace hat of under at spread heat Ace endured Ί put a towel under the hat (=between (the top of) the hat and my head) and lessened the heat.'

(27)

Sakusya ga zinbutu no usiro ni atte, sibasiba ito o hiita. author Nom actor of back at exist often thread Ace pulled The author often stood behind the actor and guided him.'

(28)

Watasi no naka no onna wa sinde simatta. I of inside of woman Top die finished The woman in me has died.'

264

Yoshiko Tagashira

(29)

Uti ni toosi o himete iru. inside in fighting spirit Ace hide be 'He hides deep inside himself a strong fighting spirit.'

(30)

Kare wa ikari o soto ni arawasita. he Top anger Ace out to exposed 'He showed his anger on his face.'

(No example was found for mae.) 7. Abstract place. What is described in a sentence does not refer to concrete objects, but to abstract entities such as social structure and situation. (31)

Nedan ga takai sono ue ni situ ga yokunai. price Nom high that on to quality Nom not good 'It is expensive and, on top of that, the quality is bad.'

(32)

Kare no sita de wa syusse wa muri da. he of under in Top promotion Top difficult be 'Being under him, getting promoted will be difficult.'

(33)

Kuike no mae ni wa seisin ga dakyooteki ni naru. appetite of front at Top spirit Nom compromising to become 'Faced with this desire to eat, my spirit has to compromise.'

(34)

Teki to kessen-nasi, syukun no usiro o yasuku sen. enemy with fight, master of behind Ace easy let's do 'By defeating the enemy, let us make our master's home territory safe and calm.'

(35)

Hutari no naka o torinasite kureru hito ga inai. two men of inside Ace mediate favor person Nom not exist There is nobody who is kind enough to mediate between those two people.'

8. Abstract generic use. The RNs no longer refer to physical regions but to abstract social areas. In this case too, an RN occurs without a preceding 'PLACE noun+ «£>'.

(36)

Ue ni wa ue ga aru. on at Top on Nom exist There is always one still further up (in the hierarchy).'

Some aspects of relational nouns

(37)

(38)

265

Uti no syatyoo wa nakanakano yarite da. in of president Top quite go-getter is The president of our company is quite a go-getter.' Uti de wa sinmyoo da ga, soto de wa nani home at Top of good behavior be but out of the home in Top what o siteiru wakaranai. Ace is doing do not know 'He behaves well at home but you never know what he's doing when he's out.'

9. Arrangement. An entity whose location relative to others within an object or area is profiled and PLACE is usually unprofiled. (39)

(40)

(41)

(42)

(43)

(44)

Ue hanbun wa kuuhaku ni site oite kudasai. on half Top blank to do leave please 'Please leave the upper half blank.' Itiban sita no tokoro ni kaite aru. first under of place at written be 'It is written at the very bottom (of the page).' Kare wa hutatu mae no seki ni ututta. he Top two front of seat to moved 'He moved to a seat, two seats in front of it.' Usiro no seki kara wa mienai. Back of seat from Top cannot see 'You can't see it from the back seat.' Kore ga naka no ani desu. this Nom inside of elder brother be 'This is the middle one of my three elder brothers.' Nizi wa itiban soto ga aka da. rainbow Top most outside Nom red is Ά rainbow's outermost colour is red.'

(An area cannot consist of a plural number of wff s.) 10. Time. We will examine how each RN is used for some meaning in the category of TIME. (i) ue. The function of ue may be said to be that of a conjunction. First, it attracts attention to the situation defined by what precedes ue; then it serves to indicate that what follows ue is what may happen on the basis of this situation.

266

(45)

Yoshiko Tagashlra

Uso amai mono o kuroota ue naraba nanika wa fake sweet thing Ace ate on if anything Top yokaroo. may be good 'After having eaten things which had a faked sweetness about them, anything may taste good.' (from the Japanese translation of Aesop Stories; 1593)

(ii) sita. Given that an activity (of speech, for instance) proceeds through ΉΜΕ, or, in terms of SPACE, from the top of the page to the bottom, then, if a certain action takes place, the time immediately after the action will be sita of the action. (46)

Kodoku o kakotu kuti no sita kam tomo no to be alone Ace complain mouth of under from friend of muzyoo ο uramu ... heartlessness Ace resent ... 'She complains about her loneliness and immediately afterwards resents her friends' heartlessness.'

(iii) mae. The duration of time before a certain event takes place or a goal is reached is referred to by mae. (47)

Daigaku ni hairu mae wa Koobe ni sundeita. college to enter before Top Kobe at was living 'Before entering college, I was living in Koobe.'

(iv) naka and uti. The area bounded by three dimensions (=naka) and the spatial territory (=uti) can be understood as delimited amounts of TIME. (48)

(49)

Isogasiinaka o yoku irasitekudasaimasita. busy inside Ace kindly came 'Thank you for coming over during at a time when you are so busy.' Kuraku naranai uti ni kaeru. dark not become in at come back home Ί will come back before dark.'

(No examples were found for usiro and soto.) 11. Quality. It is to be expected that RNs which represent orientational opposites will come to be used to express opposing qualities.

Some aspects of relational nouns (50)

Kaku ga ue da. rank Nom on be 'He is higher in rank.'

(51)

Sita no mono ga toyakaku iu koto dewa nai. under of people Nom comment say thing be Top not This is not something which subordinates should comment on.'

267

(No examples were found for the remaining RNs.) 12. Cause. The situation of being at a certain place can easily be the reason for what happens next. (52)

Koo natta ue wa moo yaru yori sikata nai. this way became on Top already do than way not exist 'The situation having turned out like this, there is nothing else for us to do but act.'

(53)

Kono mujun no mae ni kaiketuhoo wa ariyoo this contradiction of front at solution Top possibility mo nakatta. Emphasis not existed 'Faced with this contradiction, no solution could possibly be found.'

(No examples were found for the remaining RNs.) Having discussed most of the focal senses represented by Japanese RNs, we can see that the concepts and categories identified by Heine et al. (1991) in their study of RNs are also found among our data. Concepts such as part-ofan-object and area in the vicinity of an object are indeed also expressed by Japanese RNs, and, at the same time, there are senses which belong to the categories of QUALITY, TIME or CAUSE. One important category which is peculiar to Japanese is that of GENERICS: here, the senses refer to the four basic conceptualizations of the environment; in particular, vertical spatial orientation based on the pull of gravity of the earth, horizontal directional orientation based on the cardinal movement of people, bounded space, and egocentric territory. The most important conclusion presented by Heine et al. (1991) is that grammaticalization is the result of two kinds of human activity: pragmatics and conceptualization. In their actual use of language, people practice metonymy, i.e., they extend the sense of a certain word in order to represent

268

Yoshiko Tagashira

additional concepts and thus create the additional senses for the word. At the same time, they expand the meaning of a word by transferring its use from one category to another. This is a metaphorical transfer, which is another major conceptualization pattern of human beings. The various senses of Japanese RNs we have examined indeed illustrate these two types of activity. Words which designate part-of-an-object come to be used for vicinity of the object (in terms of the preceding list of senses and example sentences above, from the senses under 3 till those under 4). This usage has in turn led to the designation of area distanced from an object (sense under 5). As for the use of metaphor, several examples can be noted: (a) the conceptualization of the environment is transferred to an arrangement within an object (from 1 to 9); (b) physical area to abstract place (5 to 7); (c) abstract generic use to QUALITY (1 to 11); (d) a bounded portion on the horizontal area to a bounded portion on the path of TIME (1 to 10); (e) TIME to CAUSE (10 to 12). Of these five, the last two are found in the RNs discussed by Heine et al. (1991) and, perhaps, in many other languages as well. The first three, on the other hand, are yet to be supported by data from other languages other than Japanese.

6. Absence of RN Now, for a deeper understanding of the nature of RNs, let us consider the two alternative constructions used to express locations mentioned at the beginning under (1-2); one of them involves an RN and another does not. If we consider the two RNs, ue On' and naka 'in', we find that they are interchangeable in some contexts, while in other contexts, only one and not the other is allowed. Some examples which illustrate this are given in (5459): (54)

a.

b.

(55)

a.

Niguruma no ue ni nosete hakobu. cart of on to14 put on carry 'We put them on a cart and carry them.' Niguruma ni nosete hakobu. cart to put on carry 'We put them on a cart and carry there.' Kodomo ga beddo no ue o haimawatteiru child Nom bed of on Route crawl

Some aspects of relational nouns

269

'The child is crawling on the bed.' b. *Kodomo ga beddo ο haimawatteiru. child Nom bed Route crawl (56)

a. *Kizuguti no ue ni kusuri o nuru. wound of on to medicine Ace. apply b. Kizuguti ni kusuri o nuru. wound to medicine Ace apply 'Apply this ointment to the wound.'

(57)

a. b.

/ no naka de syooka dekizu, tyoo e hakobareta. stomach of inside in digest cannot bowel to be carried / de syooka dekizu, tyoo e hakobareta. stomach in digest cannot bowel to be carried 'It could not be digested in the stomach and was carried to the bowels.'

(58)

Tansu no naka de15 nemutta mamada. drawers of in in sleep stay They still lie unused in the drawer.' b. *Tansu de nemutta mamada. drawers in sleep stay

(59)

a. *Kabin no naka ni mizu o ireru. vase of in at water Ace put in b. Kabin ni mizu ο ireru. vase to water Ace pour in 'Put some water in the vase.'

a.

With other RNs, the deletion of RNs always results in unacceptable sentences, as shown in (60-64). (60)

a.

Booru ga sofaa no sita ni haitte simatta. ball Nom sofa of under to enter finished 'Too bad. The ball rolled under the sofa.' b. *Sofaa ni haitte simatta. sofa to enter finished

(61)

a.

Zinzya no mae de syasin o totta. shrine of front in photo Ace took 'We took the pictures in front of the shrine.'

270

Yoshiko Tagashira b. *Zinzya de syasin o totta. shrine in photo Ace took

(62)

a.

Isu no usiro ni tatte kudasai. chair of back at stand please 'Please stand behind the chair.' b. *Isu ni tatte kudasai. chair at stand please

(63)

a.

Kuruma o man no uchi ni ireta. car Ace gate of inside to put in 'He put the car inside the gate.' b. *Kuruma ο mon ni ireta car Ace gate to put in

(64)

a.

Mon no soto de matte iru. gate of outside in wait be 'He is waiting outside the gate.' b. *Mon de matte iru. gate in wait be

Given the data, three important questions must be answered. First of all, how do we explain the fact that only ue and naka, and not the other five RNs, can have the two kinds of locational expressions (i.e., a. and b. sentences), while with the other five RNs their use is obligatory? The answer may be that some PLACES have their own canonically salient facets, but these are limited to the notions of ON and IN. Let us consider the case of a cart, which is a PLACE by virtue of the fact that it is a vehicle for the activity (executed by people) of loading the objects onto it and transporting them. The loading is done by putting the objects on its platform or bed. This is the most common, unmarked manner of using the cart and it makes ON the salient facet of PLACE. Consequently, we do not need to mention ue On' overtly; to use a cart in a usual manner means to put the objects ONTO its bed.16 If one is, for one reason or another, to put objects UNDER a bed, the RN sita must be overtly mentioned, since UNDER is not a salient facet of the PLACE identified as bed. Similarly in the sentence (57b): stomach is the PLACE where food is digested, and digestion always takes place INSIDE the stomach, making INSIDE the salient facet of this PLACE. The second question to be asked is why the salient facets are limited only to ON and IN. I do not have an answer to this question at the moment. I feel

Some aspects of relational nouns

271

that further clarification of the concept of PLACE and the investigation of more RNs are required. The third question is when can we omit the RNs ue and naka, and when must they be used? For some PLACES (such as a cart and a vase) the RNs represent their salient facets and therefore need not be overtly mentioned. However, their occurrences are obligatory in some cases (cf. 55) and prohibited in others (cf. 56 and 59). What is responsible for this difference of usage? The answer is that the use of the RNs brings a particular area of a place into profile and the context determines whether this focusing is warranted. Thus, in (58), when lamenting the fact that the recently bought clothes have been kept unused and put away in a drawer, their being deep inside the drawer is emphasized and the overt use of naka is warranted. In contrast, in (56), since it is the wound which is being looked after and not its upper surface, and the profiling of the latter is inappropriate. Likewise, in (59), given both the act of putting in the water and the vase as the recipient of this act, where else but the inside of the vase do we pour water in? To specify that the act of pouring is directed at the inside is superfluous and incongruous with our understanding of the function of a vase.17 These observations confirm our conclusion stated earlier: viz. the function of RNs is to bring into focus some particular facet of the ground.

7.

Conclusion

The RNs of Japanese are themselves locational expressions denoting regions in our basic conceptualization of the environment: UP and DOWN-regions in vertical orientation, FRONT and BACK-regions in the line of movement, a region bound by six sides, UP/DOWN, FRONT/BACK and RIGHT/LEFT, and an egocentric INSIDE surrounded by OUTSIDE. This basic fact has resulted in my identifying the 'structural use' of RNs, where our attention is directed towards determining whether a particular relation refers to (a) a vertical axis or a horizontal axis or an area with IN/OUT distinction, and (b) the particular positions two (or more) entities occupy on these axes or in the area (as defined). The RNs exhibiting this 'structural use' are found in the example sentences cited under the following focal senses of Section 5: Generic meaning, Abstract generic meaning, Arrangement, Time, Quality, and Cause. Let us consider three examples. First, Doi-kun wa Inoue-kun no sita de wa syusse dekinai '(D. + Top + I. + of + under + in + Top + get promoted

272

Yoshiko Tagashira

+ cannot -») Being under Mr. Inoue, Mr. Doi cannot get promoted.' Here, our conceptualization is a matter of our recognizing a vertical axis in human society and two regions on this axis, viz. regions occupied by Doi and Inoue. In the second example Maeda-kun wa Atomiya-kun no zutto mae o hasitte ita '(M. 4- Top + A. + of + plenty + front + Ace 4- run -f was —>) Mr. Maeda was running far ahead of Mr. Atomiya', the runners are perceived as being on a horizontal line and the relative positions of the two runners on this axis is the object of our interest. In the third example Huri-dasa nai uti ni kaerimasyoo '(rain-start + not 4- inside + at + let us go home —>·) Let us go home (within the time when it doesn't start raining —>·) before it starts raining', the focal point of our interest refers to the opposition of 'the m-areas vs. the οΜί-area' and the situation in the 'm-area and the one in the owf-area' are compared. The other use of RNs will be referred to as 'focused PLACE'. Here, our image consists of an OBJECT and some PLACE and, most importantly, the focused facet of this PLACE. Typically, PLACE is some unbounded, twodimensional expanse but some facet of it (it may be the surface, the vicinity, or the inside) is focused. In constructing the image of yama no ue, for instance, our attention is not directed to the shape or to any other features of the mountain other than one of its facets, viz. its summit. This observation is supported by the grammatical structure as well: in the noun phrase, 'PLACE + no + RN', the RN is the head noun modified by the preceding two words, just as in the phrase watasi no hon '(I + of + book —>) my book', hon is the head of this phrase. Based on the preceding discussion, I would like to suggest that 'structural use' is the result of metaphor, where the concepts and structure within one category are transferred to another category. In contrast, in the 'focused PLACE' use, pragmatics are at play: there are many varied facets which can be focused and a limited number of RNs come to be used to express these varied facets.

Notes 1. Eight postpositions of Ewe - te, me, dzi, ijgo, megbe, ta '-me, go-me, dz-me are listed by Heine et al. in Table 5.5 (1991: 134). The authors recognize three different groups among these lexemes: (i) the first three which show three or four nominal features and "have been grammaticalized to the extent that they have lost many of their former nominal characteristics" (1991: 136); (ii) the last three

Some aspects of relational nouns

273

"are nominal with respect to all ten parameters [...] (and) are virtually indistinguishable from nouns" (1991: 135). They constitute a new set of lexemes that the language has developed "to make up for the loss" of nominality of the first three lexemes; (iii) ng3 and megbe are also fully nominal. The term relational noun is used by the authors only when discussing the Kabiye language (1991: 148) or when citing the study of Ewe by Lehmann (1991: 19). In their discussion of Ewe, they seem to stick to the term N-adposition. However, since they emphasize the nominality of the lexemes of (ii) and (iii) and since each of the lexemes represents an entity which must be defined in terms of the relation that holds between it and some other entity, they can justifiably be called relational nouns. 2. In Japanese there is a set of bound morphemes which are variably called 'particles', 'case-particles' or 'postpositions'. In the present paper, we use, without any strong justification, the following groups of forms: (i) Case-particles: those morphemes used for indicating 'agent', Object' and 'indirect object'. These meanings are expressed by the Nominative-case particle, ga, the Accusative-case particle, o, and the dative-case particle, ni. Most examples to be cited in the paper will have a 'morpheme gloss line', where the above three morphemes are given the abbreviated Case-identifications of Nom, Ace, and Dat, respectively. (ii) Topic particle, wa. In a morpheme gloss line, it will be Top. (iii) Locational postpositions: morphemes which refer to locations are particles such as ni 'at, to', e 'to', de 'in'. In a morpheme gloss line, they are allocated equivalent English prepositions. (iv) The particle no expresses a very wide range of meanings, which will be discussed later. Since in most cases, no can be translated by English Of, we give this as its gloss. 3. The etymologies of the second constituents of these words (i.e., he, ta, ka and mo) are well attested. However, the etymologies of the first constituents, i.e., u, si, no) are far from agreed upon. What is cited as supporting evidence for the etymologies given here is the fact that phonologically as well as semantically the same or similar morphemes are found in other languages that are reputedly related to Japanese, including Korean, Manchurian, Mongolian and Tungus. I have not succeeded in finding out when the pronunciation of [he] was changed to [e]; the orthography has remained the same, and in present-day Japanese the latter is read as [he] only when it occurs as a free form. 4. One popular dictionary, Kokugodaiziten [Comprehensive Dictionary of Japanese] (Gakusyuu Kenkyuusya; 1990) lists nine usages of no (including the one mentioned in the main text): 'possession' (e.g., Taroo no tukue 'Taroo's desk'); 'location' (e.g., Nihon no saru 'monkeys living in Japan'); 'author' (e.g., Ooe-si no syoosetu 'novels of Mr. Ooe'); 'content' (e.g., ryoori no hon 'book on cooking'); 'naming' (e.g., Kobe no mati 'City of Kobe'); 'apposition' (e.g., tomodati

274

Yoshiko Tagashira

no Taroo 'my friend, Taroo'); 'renders substance to the following abstract noun' like tame 'sake' (e.g., anata no tame ni '(you + no + sake + to —>·) for your sake'; 'agent' (e.g., tuma no sentakusita mono (wife + no + selected + thing -4) 'the thing my wife selected'). 5. Corresponding to the two types of noun, nouns denoting OBJECT and nouns denoting PLACE, are two distinct types of adjective. The existence of these adjectives has long been noted by various scholars. They have, however, only recently been examined in detail, and references are made to the nominal distinction of OBJECT and PLACE in Kushima (1993). The discussion in this section owes much to the insights found there. 6. Herskovits states that in the sentence, "There is a rabbit under the bush, the rabbit is not under the whole bush, which would be under the roots". Based on this and other examples, she suggests the following principle: "One can use a noun which denotes a whole object to refer to a part of it that is typically salient." (1988: 285) I would like to claim that the saliency is determined by the degree of human interest, and the PLACE of human activities, naturally, attracts a high degree of interest on the part of humans. 7. This has been pointed out by Rosch (1978: 29), as cited by Lakoff (1987: 50). 8. In representing them on a map, we must represent, following pre-determined rules, their 'boundaries'. In the actual world, however, they are extensions of the ground and have no clear lines of demarcation. These two different situations are reflected in the adjectives used: 'Lake Biwa on the map' is described as ookii 'big'; but 'Lake Biwa in which I swam yesterday' is described as hiroi 'spatious'. 9. As already mentioned, the RN uti represents an egocentric territory surrounded by OUTSIDE. However, this territory can be One's residence', a concrete object. With this meaning, uti can be quantified or modified: e.g., Hidoku syareta ookina uti ο tateta mono da na! '(terribly + modern + big + house + Ace + built + fact -f be + Exclamatory Sentential Particle —)·) I am amazed you built such a big modern house.' 10. As for the examples quoted in this article, we will indicate their sources, only when they are from pre-20th century literary works. Examples from works published after 1900, examples from present-day works, and those the present author composed will be cited with no additional information. 11. The examples are given in the order of the RNs listed in the chart in Section 2, except when a sentence contains two RNs with contrasting meanings. 12. The data which were excluded from the present paper are compound nouns which contain RNs as their components. The data are very large and constitute a very important portion of the grammar of RN. When it is said, as it is here and in other places later on, that the particular meaning being discussed is not represented by some of the RNs, such a meaning may be found in the RNs which occur as components of compound nouns. The examination of such compound nouns is definitely a task that must be taken up.

Some aspects of relational nouns

275

13. Although Outdoor' is not a three-dimensional object, it means 'an area outside a building', and its meaning must refer to some three-dimensional building. Accordingly, this example is classified under the sense of Object'. 14. The particle ni has several meanings, depending on the meanings of the cooccurring verbs. We will mention only the two meanings which are relevant here. When ni cooccurs with the 'verb of existence' aru 'exist', it designates a 'spot where it exists' and has a close equivalent in the English preposition 'at'. With verbs of motion such as noseru 'put on', it designates an end-point of the movement, corresponding to the English 'to' or Onto'. 15. The particle de is distinguished from ni discussed in the preceding footnote in that it indicates a place where a certain activity takes place. Compare the following sentences: (i) Taroo is doing some work. Taroo wa Tookyoo de hataraite iru '(Taroo -f Topic + Tookyoo + in + work + be —>) Taroo is working in Tookyoo.'; (ii) Taroo merely exists. Taroo wa Tookyoo ni sunde iru. '(Taroo + Topic + Tookyoo + at + live -f be —») Taroo lives in Tookyo.' 16. A parallel situation is observed with denominal verbs like 'knife'. Hearing the sentence 'He knifed the meat skillfully', we understand that he cut the meat with a knife rather than, for example, he decorated the meat with knives. So long as knives are employed for their usual, unmarked objectives of cutting things, one need not overtly mention how they were used; we do not need to say 'He cut the meat skillfully with a knife.' 17. It may be argued that the act of pouring water is directed at the open mouth of the vase, and that the INSIDE may not be the focused facet. We reject this suggestion in the light of the following evidence. We noted earlier that two sets of adjectives must be distinguished: takai 'tall' and hikui 'low', on the one hand, and hukai 'deep' and asai 'shallow', on the other. The first set of adjectives is used to refer to the PLACE on which people put things and the 'length' which is measured from the ground (e.g., takai isu 'tall chair'). A vase is not such a PLACE, and in support of this, a vase cannot be described as takai 'tall' or hikui 'low' in Japanese. In contrast, the second set of adjectives is used for PLACES in which people put things and 'length' which is measured 'downwards' from the opening. A vase must be identified as such a PLACE, since it is qualified in Japanese by the second set of adjectives, namely, hukai 'deep' and asai 'shallow'. Therefore, its salient facet must be the INSIDE, and not the highest point from the ground.

References Croft, William 1991 Syntactic Categories and Grammatical Relations. Chicago: The University of Chicago Press. Heine, Bernd et al. 1991 Grammaticalization. Chicago: The University of Chicago Press.

276

Yoshiko Tagashira

Herskovitz, Annette 1988 Spatial expressions and the plasticity of meaning. In Brigida Rudzka-Ostyn (ed.), Topics in Cognitive Linguistics. Amsterdam and Philadelphia: John Benjamins. Kushima, Shigeru 1993 Nihongo no ryoo ο arawasu keiyooshi no imitaikei to ryoo-kategorii no huhensei (The semantic system of Japanese dimension adjectives and universale of dimension categories. Gengokenkyuu [Study of Language] 104: 49-91. Lakoff, George 1989 Women, Fire, and Dangerous Things. Chicago: The University of Chicago Press. Langacker, Ronald 1987 Foundations of Cognitive Grammar, Vol. I . Stanford: Stanford University Press. 1990 Concept, Image, and Symbol. Berlin/ New York: Mouton de Gruyter. 1991 Foundations of Cognitive Grammar, Vol. 2. Stanford: Stanford University Press. Miyaji Atsuko 1979 Shinshin Goi no Shiteki Kenkyuu [Historical Study of Body-Mind Vocabulary]. Tokyo: Meijishoin. Morita, Yoshiyuki 1989 Kiso Nihongo Jiten [Dictionary of Basic Japanese]. Tokyo: Kadokawa Shoten. Rosch, Eleanor 1978 Principles of Categorization. In Eleanor Rosch and B.B. Loyd (eds.), Cognition and Categorization. Hillsdale, N.J.: Lawrence Erlbaum Associates. Svorou, Soteria 1993 The Grammar of Space. Amsterdam and Philadelphia: John Benjamins. Yamanaka, Joota 1976 Kokugo Gogen Jiten [Etymological Dictionary of Japanese]. Tokyo: Azekura Shoboo.

Prototype marker or reflexive marker: Russian -sja and categorical change Adger Williams

1.

Introduction

The purpose of this paper is to present an analysis of the Russian "reflexive" marker -sja that seems to have interesting implications for the analysis of reflexives, and possibly for grammatical theory within the cognitive framework as well. The paper can also serve as an illustration of a path of extreme grammaticalization that a reflexive marker might follow.

1.1.

Outline of the paper

Speakers of Russian can use the marker, (a verbal affix, -sja /-s') to indicate that the situation presented by an utterance, containing a verb bearing the affix, is a situation prototypical of the set of situations presentable by that verb when it does not bear the -sja affix. I refer to this function as PrototypeAccessing. The second part of this paper is devoted to a justification of the claim that -sja has such a function. Given that only some senses of a verb trigger the use of the affix (in its Prototype-Accessing function), a schematic network of the category symbolized by the verb should contain differentiation between triggering senses, and non-triggering senses. A tentative proposal is suggested in the third part of the paper. Not all s/'cz-predicates, however, are used for Prototype-Accessing. This function is only one of the many functions of -sja. Placing it within the schematic network of other uses of the affix suggests a re-evaluation of the network. The fourth part concludes with a brief introduction to and sketchy discussion of some of the uses of -sja.

1.2.

Background

-Sja is attached to verbs with syntactic consequences that differ depending on the meaning of the verb to which it is affixed.1 It is used therefore, for a number of different purposes.2

278

Adger Williams

The most common use of -sja is its addition to Patient-describing (or "accusative"3) transitive verbs to make imperfective passive predicates, in which a Patient appears as subject, and the Agent appears (if at all) in the Instrumental case. (1)

a.

b.

Okno moetsja (rabocimi). window-N wash-pr-3s-sja (workers-I). 'The window is washed (by the workers).' Rabocie mojut okno. workers-N wash-pr-3p window-A. 'The workers wash the window.'

As a special instance of this pattern, the Patient may be construed to be identical to the Agent, resulting in a predicate's interpretation as reflexive. (2)

a.

b.

Mai'cik moetsja. boy-N wash-pr-3s-sja. The boy is washing (himself).' Mal'cikmoet sebja. boy-N wash-pr-3s refl-pro-A. 'The boy washes himself.'

The difference between the two strategies for presenting reflexive situations displayed in example (2) will be discussed at a later point. Suffice it to say, for now, that examples (2a) and (2b) are not exactly synonymous.4 -Sja can also be added to an Agent-describing ("ergative") transitive verb, resulting in a predicate in which the Agent is the subject, and the Patient does not appear. Such predicates are most often interpreted as generic statements about properties of the subject. A generic Patient (everyone, anyone who gets close enough, etc.) is inferred, and the predicate is interpreted as referring to a generic state of affairs. (3)

a.

b.

Sobaka kusaetsja. Dog-N bite-pr-3s-sja 'The dog bites (generally)' Sobaka kusaet vsex. dog-N bite-pr-3s everyone-A 'The dog bites children.'

As a special instance of this pattern, these are verbs whose Patients are predictable from the meaning of the verb; e.g. tarascit' 'to bug out [the eyes]'

Prototype marker or reflexive marker

279

and nesti 'to lay [eggs]'.5 When a verb of this kind appears in a 5/a-predicate, the predicate need not be interpreted as generic, because the Patient for any instance is easily inferable. (4)

a.

b.

Ivan scurilsja pod solncem. Ivan-N squint-pa-m-sja under sun-I 'Ivan squinted in the sun.' Ivan scuril glaza pod solncem. Ivan-N squint-pa-m eyes-Α under sun-I 'Ivan squinted his eyes in the sun.'

In (4a), Ivan could squint only his eyes, so the addressee assumes that that is what Ivan squinted. Because this interpretation does not involve a generalized Patient (everyone, anything that gets close enough, etc.), as in (3a), it can be interpreted as referring to a specific occurrence. Verbs with inferable Patients provided the inspiration for the experiment described in part two, and may have provided language-users with the inspiration to make predicates like the next type: made from intransitive verbs that combine with -sja. (5)

a.

b.

On dolgo stucalsja v dver', no emu ne He-N long-time knock-pa-m-sja at door-Α, but he-D not otkryli. open-pa-3p. 'He knocked at the door for a long time, but they did not let him in.' On dolgo stucal v dver'. He-N long-time knock-pa-m at door-A. 'He knocked at the door for a long time.'

Note that there is no difference in the argument structure of the predicates in (5a) and (5b). Yet neither this pair of sentences, nor the pair in example (4) are completely interchangeable. Nor is the difference purely stylistic as is sometimes stated.6 Informants have definite preferences for one predicatetype over another in many contexts, as I document in part two. One must, therefore, consider how the presence of -sja effects this difference in preference. Other types of s/a-predicates exist, differing more in the inferences drawn by the addressee than in the meaning of the affix.

280

2. 2.1.

Adger Williams

Explanation of the experiment Outline of the experiment

To provide some validation of the hypothesis that the affix -sja serves as a marker that the speaker considers the situation presented by the utterance containing -sja to be prototypical of the set of situations that the verb in the utterance could present without -sja, I devised the following experiment. I compiled a list (from an authoritative dictionary7) of the verbs that could make predicates like (4a). From this group, I chose to investigate six verbs of facial expression, because of their relative commonness and usefulness:8 zmurit' 'to clench shut [eyes]', scurit' 'to squint [eyes]', tarascit' 'to bug out [eyes]', xmurit' 'to draw together [forehead, brows], 'frown', supit' 'to draw together [brows]', sulk', morscit' 'to wrinkle [e.g. the forehead]'. Only the last of the six can be used in combination with a Patient other than a bodypart. I then asked 11 informants to define the verbs, give their preferred direct object, and a couple of good examples. From the overlap of definitions and examples and from the examples cited in the dictionary, I hoped to deduce what the prototypical situations associated with each verb were. Armed with these deductions, I compiled a questionnaire composed of sentences containing these six verbs in context. The verb and the immediate actants were left out with any of the three possibilities provided for informants to complete the sentences; e.g. (6)

Sergej prisel i rasskazal o torn, do Sergej-N arrive-pa-m and recount-pa-m about that-P that slucilos'. ot ploxix novostej. Sobytija razvivalis' ne happen-pa-ne from bad-G news-G events-N develop-pa-p-sja not kakej by xotelos'. as she-D ptc want-pa-ne-sja a. Anja naxmurila lob Ann-N frown-pa-f forehead-A b. Lob Ani naxmurilsja forehead-N name-G frown-pa-m-sja c. Anja naxmurilas' Ann-N frown-pa-f-sja 'Sergej came and recounted what had happened. Anja frowned at the bad news; events had not developed as she might have wished.'

Prototype marker or reflexive marker

281

Choices were always presented in the same order; i.e. the option for a transitive verb with both Agent and Patient overtly expressed was option (a), a j/a-verb with Patient as subject was option (b), and a s/a-verb with the Agent as subject was option (c). Sentences with the same verb did not typically occur one after the other, so as to minimize contamination. There were three types of sentence for each verb: I. those presenting a situation prototypical of the verb, II. those presenting a situation as far from prototypical of the verb as feasible, and ΠΙ. sentences that were indeterminate, based on the information gathered in the definition interview. The sentences in the questionnaire were also scrambled on this basis, to avoid contamination. For the first type of sentences (prototypical situations), informants responded by selecting the (c) option most often; for the second group of sentences (decidedly un-prototypical situations), informants responded with a preponderance of (a) options. Responses to the third group of sentences were more mixed.9 Thus, for utterances presenting a situation prototypical of the set of situations that the verb can describe, a 5/a-predicate like (4a) was preferred over a transitive predicate or a 5/a-predicate with the Patient (Active Zone) as subject, and for utterances presenting un-prototypical situations, the transitive configuration of the predicate was preferred. This distribution of forms is precisely in accord with the hypothesis tested; use of -sja depends on the prototypicality of the situation being presented. The remainder of this section is devoted to an in-depth examination of one verb as an example. Further information about the other verbs or general information can be obtained from Williams (1993).

2.2.

Detailed description of one verb

I have chosen the verb scurit' 'to squint' for exemplification. According to the dictionary, the verb without -sja means "bringing the eyelids closer together, to make the eyes be more closed." scurit'sja 'to squint-sja' is defined as 1. "to squint the eyes" and 2. "to make more closed by pushing together the eyelids (of eyes)." The second definition is the Patient-prominent configuration (not being studied here), and the first is the Agent-prominent configuration, which is defined by using the verb in its transitive configuration with a specified object and no other differentiating factors.

282

Adger Williams

2.2.1. First interview In their first interview, six informants specifically gave a definition something like "narrow the space between the eyelids", while the other five simply referred to the eyes as the object. Of these all but one used examples that showed that the eyes were not closed, because the owner of the eyes could still see. As examples, eight mentioned a person wanting to see better, some with no further specification. Seven mentioned wanting to see better in bright sun. Five mentioned nearsightedness as the reason for squinting. One mentioned looking at something small. Four informants mentioned looking at someone or something not entirely understood, or of which one was suspicious. One informant mentioned doing this when one is being sly oneself. From these responses, it would seem that the best example of use of the verb scurit' 'to squint' involves a situation, where the Agent wishes to see better, and it is likely to be associated with sun or bright light, although nearsightedness (and, possibly, a small object) seem to be other reasons for engaging in the activity. 2.2.2.

Questionnaire

The following situations were presented to the informants. Sentences (7) and (8) served as prototypical situations of reaction to bright light. (7)

Liza squinted as she looked around her. Yes, the sun was shining brightly.

(8)

Alia pressed the switch and the room was filled with light. Unaccustomed to it, we squinted for a little while.

Opportunities to respond to stimuli involving nearsightedness and small objects were provided by sentences (9) and (10) respectively. (9)

How could you tease her? She always squints, she is nearsighted.

(10)

She squinted at the coins on the table. They were old, and the engravings were worn down; reading them was difficult.

The next two sentences tested informants' reactions to situations dissimilar from those they had mentioned in the first interview.

Prototype marker or reflexive marker

283

(11)

Then she played the role of their grandmother. So as to seem nearsighted, she squinted fiercely. Her brothers nearly fell over with laughter.

(12)

She squinted her eyes to powder her eyelids.

Various of my female informants pointed out that this particular way of applying powder (with the eyes squinted, but not closed) was quite silly, but they were happy to suggest that painting the eyelids with the eyes squinted was appropriate. As situations that might be close to the prototype (or might not), I tested a few others, listed below. (13)

For ten minutes, the surgeon washed his hands. During that time, he squinted at the information about the wounded soldier.

(14)

Liza squinted her eyes and looked into the distance.

(15)

"Where are they, the damned dogs", the captain asked. The scouts squinted their eyes under the rain. Silence.

(16)

The sunset was wonderful, but Sergej did not notice it. He was sitting under a tree and reading, passionately and avidly reading. Night arrived, just as unnoticed. He squinted more and more in the twilight.

Example (13) differs from the schema of small-item-looking in that, while the surgeon is looking at something small (printed information), its smallness is not referred to. Example (14) offers the possibility of distance as a cause for squinting, and (15) offers the possibility of squinting to keep rain out of the eyes. Example (16) offers dim light as a possible motivation for squinting. All these situations differ from the examples suggested by informants in small, but potentially significant ways. Yet they are not completely divorced from the general thrust of the examples provided as are sentences (11) and (12). 2.2.3.

Results of the questionnaire

According to the hypothesis being tested, informants should have preferred option (c) (j/a-verb with Agent as subject) for the sentences presenting good examples of the verb. The sentences cited in examples (7) and (8) should have the highest concentration of option (c) responses, since they are the

284

Adger Williams

sentences that present situations cited as examples of the verb by most informants. The sentences cited in (9) should come next, because, although several informants mentioned situations very much like it, fewer than half did so. The situation in (10) was only specifically mentioned by one informant but was consistent with what others said, so our predictions concerning its results must be kept: there might be considerable use of option (c). The sentences presented in (11) and (12) should have the lowest concentration of option (c) responses, since they are as different from the examples mentioned by the informants as I could make them, and still have informants willing to use the verb. The sentences presented in (13-16) might have a lower density of option (c) responses than examples (7-9), since they present situations similar to those mentioned by informants, but not actually mentioned by any person, but more than the anti-prototypical examples in (11) and (12). In fact, these are approximately the results that are observed.

2.3. Summary For sentence (7) and (8), the percentage of use of the option (c) was 67. For (9), it was 71. For (10), it was 45. For sentences (13-16), it was 59. And for sentences (11) and (12), it was 24. The difference between (7) and (8) as opposed to (9) is not statistically significant, but the difference between (7-9) (situations explicitly mentioned by more than one informant), and (11) and (12) (situations that are not related at all) is statistically significant. Situations that have nothing to do with good examples of the verb are preferably presented by a verb bare of the -sja affix. This pattern holds true for the other five verbs tested.10 Therefore, it is legitimate to say that one function of the affix, -sja is the marking of prototype-accessing. Previous researchers (e.g. Schenker among others) have explained the difference between sja- and non-sy'a-predicates by stating that predicates like (4a) or (5a) are similar to reflexive predicates, in that the subject is more involved in the activity presented by the verb when the verb bears the affix than when it does not. In reflexive predicates the subject of the predicate fills both Agent and Patient roles, as opposed to only one case-role, when the verb is bare. In predicates like (4a) and (5a), the increased involvement of the subject is evident in that the verbal action "never leaves the sphere of the subject." (Schenker, p. 32, quoting Vinogradov) But the activity does not leave the sphere of the subject in an intransitive predicate without the affix (or in

Prototype marker or reflexive marker

285

a predicate made with a verb like scurit' 'to squint' that presents movement of a part of the Agent's body). What then is the difference between (5a) and (5b), or (4a) and (4b)? Unsatisfied with this formulation, I have endeavored to add some specificity to the idea of the subject's increased involvement in the activity. I propose instead that reflexive predicates are similar to predicates like (4a) and (5 a) in that all these different kinds of predicates share the assertion that the situation presented by the ί/'α-predicate is prototypical of the set of situations presentable by the verb to which -sja is affixed. Thus reflexive predicates are only coincidentally reflexive; they are asserted by a speaker to be prototypical situations, and situations prototypical of certain verbs (myf 'to wash', brit' 'to shave', etc.) are reflexive.

3.

Implications for the meanings of these verbs

The existence of the property of prototype-accessing for verbs affixed with -sja has implications for analysis of the treatment of verbal meaning by Russian. If we follow Langacker (1987) in creating a schematic diagram to provide a representation of the category symbolized by a word, then we are obliged to posit two different kinds of nodes in the network. There are those nodes that, on being appropriate to a particular speech event, do not trigger the use of the affix -sja, and there are nodes that, on being appropriate to a particular speech event, do trigger the use of the affix. What we know about a node (sense of a word) that triggers the use of the affix is that it is cognitively the most salient in the informants' minds. We know this because, when asked to give an example of the use of a word, informants responded with a situation for which that sense was appropriate. It is a prototype of the category symbolized by the word. For the example word described above, scurit' 'to squint', one could posit the following schematic network shown in Figure 1. Within the network, solid lines represent relationships of specialization, and dotted lines represent relationships of elaboration. Curved lines lead from nodes of the schematic network to the senses they represent. The exact relationship between squintj and squint2 is not entirely clear. They may be joint prototypes, neither of which is an elaboration of the other, both of which are basic somatic experiences with the same results, though triggered by different causes. Perhaps, we should state that the likelihood of a verb's bearing the affix is dependent on the predicate's presenting a situation for which the ap-

286

Adger Williams

narrow the eyes due to near-sightedness

Figure 1. Schematic network for scurit'

propriate sense of the verb is depicted in the bottom layer of the schematic network. This suggestion seems adequate until more investigation reveals whatever shortcomings it may have. What is important is that some part of the schematic network is differentiated from the rest with respect to the use of the affix. However we represent schematic networks of categories symbolized by words, we must have some way to delineate nodes symbolizing prototypical senses from other nodes.

4.

Implications for the category represented by -sja

4.1.

Previous work within the cognitive framework

The place of the prototype-accessing function within the network of the uses of -sja must now be established. Interestingly enough, the currently available understanding of -sja and reflexives/middles in general does not provide much obvious place for the prototype-accessing function. 4.1.1.

Kemmer

Kemmer's (1988) comparative study of the middle voice mentions the Russian marker -sja as an extreme example with some odd uses that do not fit

Prototype marker or reflexive marker

287

ACTIVE

.gophoric Middle Logophoric Reflexive Passive Middle V»

Non-Iran slat kma] Motion

Cognitive Middle Spontaneous Action or Process

One-participant — Eyfipf« ACTIVE

Figure 2. Kemmer's network of situations

well into her basic network, although most of its senses are firmly embedded in it. Here reproduced is Kemmer's network of situations in which appropriate areas for use of various middle markers can be charted as shown in Figure 2. The marker -sja can be affixed to verbs to depict any of the above situations except Active, two-participant events11 and any kind of Logophoric situation. As Kemmer points out, the display of situational types does not include situations like that presented in (3a), where a generic property is attributed to the subject of a s/'a-predicate, unless this falls under the rubric of Active One-participant events, where (4a) and (5a) seem also to belong.

288

Adger Williams

Kemmer suggests by her exclusion of the Russian marker in her study of middle markers that perhaps it is not exactly a middle marker, an idea I will pursue later. Furthermore, there is no place in a description of situational types for including a property like prototype-accessing. Prototype-accessing is not dependent on situational type. It can, at least in theory, occur in any situational type. Whether it does or not is a matter for empirical investigation. We see, therefore, that the framework for understanding middle markers is not entirely satisfactory for the understanding of -sja. 4.1.2.

Janda

Janda's work, a comparison of the Czech marker se and the Russian -sja provides a different network of situational types, designed specifically for the study of sja (see Figure 3). The two senses "Intensive Action by-the-Self" and "Aggressive Action Bythe-Self" are proposed to account for the meaning differences between pairs of sentences like examples (5a) and (5b), or possibly (4a) and (4b), though the verb scurit' 'to squint' is not mentioned in her treatment, nor are others with predictable Patients. Intensivity and/or Aggressiveness are far from the whole addition in meaning that -sja provides in sentences like these. The difference between (9) and (11) is hardly explicable in terms of intensiveness or aggressiveness. According to Janda, the situation in (9) should occasion less use of -sja than the situation in (11) where there is specific lexical support for intense squinting. However, just the opposite is observed. There was more use of -sja in response to sentence (9) (8.5 out of 11) than to sentence (11) (2.8 out of II). 12 Similarly, while it might be argued that the difference between (5a) and (5b) is persistence or aggressiveness in knocking, no amount of persistence or aggressiveness will render the following example anything but humorous. (17)

V apparate cto-to stucitsja. In machine-P something-N knock-pr-3s-sja There is something knocking inside the machine.'

Russians confronted with sentence (17) unanimously claim that it means that there is an individual inside the machine, asking to be let out. Knocking to be admitted through a door is prototypical knocking; the accessing of this prototype justifies the meaning forced on the above sentence and the acceptability of example (5a). There does not seem to be any general unifying

Prototype marker or reflexive marker

289

On-Self By-Another

Controlled-by-Se

On-Selfi By-Setf2 On-Self: By-Self,

Generic Action By-the-Setf

Intensive Action By-the-Self

Agressive Action By-the-Self

Figure 3. Janda's network of situational types

principle that unites the prototypes of the six verbs I investigated, like Intensivity or Aggression. (One of them is used prototypically only when the speaker thinks the performer of the action is stupid.) Explaining the use of -sja with these verbs on the basis of some common meaning provided by -sja is a hopeless endeavor. The differences in meaning between sja-forms and non-5/a-forms vary from verb to verb in accordance with the meaning of the verb. Nowhere in Janda's description of the meanings of -sja is there a place to add the property of prototype-accessing, because her network is a list of meanings of the affix, not interactions with the meaning of the verb to which it is affixed. Therefore, this approach too is in need of some re-evaluation in light of the existence of prototype-accessing.

290

Adger Williams

4.2.

Focus-based analysis

I propose instead an analysis that takes -sja to be primarily a grammeme associated with focus. A speaker uses the affix to focus the attention of the addressee on one participant in the situation presented by the verb. As a result, other participants are backgrounded, or not mentioned at all; see examples (1), (2), and (3) for instance. This is the prototypical use of -sja. There seem to be two ramifications. One is that information about the focused participant is provided in greater abundance than usual; i.e. the participant's prototypical involvement in the activity is presented. This is the prototype-accessing property that we have discussed so far. I have, as yet, observed no verb whose prototype did not involve specification of the involvement of one or more main participants. Whether a verb like hammer with a prototype that specifies the nature of the involvement of a peripheral actant (use of a hammer, as opposed to a shoe) exists in Russian, or take a walk prototypically while the sun is shining, remains to be seen. I believe that no such verb will be found to form prototype-accessing s/a-predicates. Since -sja serves to focus on participants, only the part of the information about prototypical activity relating to participants is presented by a s/'a-verb. The other ramification is that information about the defocused participant is necessary in the discourse in order to describe the focused participant. In such cases, the defocused participant is included but not in a direct case. Its defocused status is thus reflected by the case system. (18)

a.

b.

Belyj pesok otrazal solnce. white-N sand-N reflect-pa-m sun-A The white sand reflected the sun.' Solnce otrazalos' ot belogo peska. sun-N reflect-pa-n-sja from white-G sand-G 'The sun (was) reflected off the white sand.' h

In 18b, the sun is the focused participant. It appears as the subject. The sand is defocused, appearing in a non-direct case (here the genitive, but strategies vary); it appears in the predicate, because understanding that the sun is reflected from the sand provides information about the sun. Agents in passive predicates appear in the Instrumental, as in example (1) in order to provide information about the nature of the Patient-subjects.13 Note how the prototypical use of -sja is related to passing information about participants in the verbal activity as are both of the ramifications. These

Prototype marker or reflexive marker

291

three uses form a more cohesive network than the network that would result from adding the prototype-accessing function to the networks of situation types presented above. Another reason to believe that -sja is not primarily a reflexive marker can be found in a comparison of the two different strategies for making reflexive predicates presented in example (2). Use of the pronoun sebja is unambiguously reflexive, while use of the affix -sja is usually ambiguous between reflexive and passive reading, and the identity of the Agent is inferred from context. (19)

a.

b.

Over' otkryvala sebja. door-N open-pa-f refl.-pro-A 'The door opened itself.' Over' otkryvalas'. door-N open-pa-f-sja The door was opened/opened itself.'

Further, for sja-verbs that are interpreted out of context as reflexive (prototypically reflexive), there is a difference in meaning between the predicates with the pronoun and the affix, as Haiman discussed in his 1983 article on iconic motivation. (20)

a.

b.

Ivan myl sebja. John-N wash-pa-m refl.-pro-A 'John washed himself (overtones of difficulty: the parts of the Agent's body do not cooperate; the Agent is weak or sick and not able to perform the action normally), Ivan mylsja. John-N wash-pa-m-sja 'John washed.' (himself, in the normal way)

In a prototypical washing situation, the Agent and Patient are not viewed as very distinct from each other. If some circumstance causes the Agent and Patient to be viewed as distinct (but still coreferential), that situation will contain elements that render it non-prototypical. The pronoun is used to present situations that are reflexive and not prototypical of the verb by virtue of reasons not related to reflexivity, while the affix is used to present situations that are prototypical of the verb in terms of reflexivity and anything else. Postulating that -sja functions to mark prototype-accessing explains handily both

292

Adger Williams

the ambiguity for some verbs (ones that are not prototypically reflexive), and the differences between the use of the affix and the pronoun.

4.3.

The change in meaning of -sja

If we suppose that -sja now symbolizes a category, centered on the focusing of attention on one participant, we are led to hypothesize about its development to this point. Historically, it was a reflexive marker. It was a clitic that appeared in Wachernagel's position, and then, in Russian, became gradually more and more bound to the verb, while the non-clitic reflexive pronoun, sebja remained relatively unchanged. Exact diachronic cataloguing of the functions of -sja as opposed to sebja has not been accomplished, but some theoretical speculation may be made, based on cross-linguistic studies. Proceeding from the hypothesis that Russian, having two reflexive markers (clitic and non-clitic) might be similar to other languages with two reflexive markers, we may produce useful insight. Pederson's 1989 crosslinguistic study of two marker languages reveals that such languages often have a phonologically less notable form (like -sja) with a wide range of senses associated with it, and a heavier form (like sebja) with a smaller range of associated senses. Typically, the senses of the heavy marker are those closest to the prototypical reflexive, while the lighter marker marks senses more peripheral in a network like Kemmer's. The newer heavier marker evolves "at the expense" of the older marker. Pederson cites Icelandic sig vs. -st as an example. Sig is used for "semantic reflexive, partitive reflexives, body-movement, and self-benefactive" situations, while -st is used for "body-movement, reciprocal, inchoative, converse, passive, and impersonal" situations (Pederson, pp. 51-55). Pederson points out that the old light form of the reflexive marker comes to symbolize a category whose internal structure is rather peculiar. A diagram of the network displaying the structure of the category symbolized by Icelandic -st in the example above would look approximately as shown in Figure 4. As is evident from the diagram, there is no central prototype for the category to radiate from. Pederson suggests cautiously that our knowledge of categories may be in need of further development, given the existence of such categories. For Russian, as I have argued above, such a diagram of the category represented by -sja is not appropriate now for several reasons, but perhaps at one time it may have been. If so, the instability of the disjointed category may have forced a re-analysis along current lines.

Prototype marker or reflexive marker Natural Reciprocal Events

/

. ^\^

293

—— Passive Middle Emotion / Middle Cognitive Middle Spontaneous Action or Process

Change in Posture Grooming

Figure 4. Icelandic -st

5. Summary I have outlined here an experiment that I performed to verify the hypothesis that the Russian marker -sja is, in certain cases, used to mark a predicate that presents a situation prototypical of the verb that the marker is affixed to. I have argued that this implies that schematic networks of the categories symbolized by verbs in Russian must have two different kinds of nodes, and that the grammar recognizes this distinction by affixation of -sja to a verb that is used to present a situation symbolized by a marked node. I have argued further that this development represents another instance of grammaticalization of a former reflexive marker.

Notes 1. For extensive discussion, see Williams 1992. 2. This statement is not intended to present the idea that there is no overarching schema that encompasses all the meanings of -sja, merely that it is convenient to discuss its uses at a different level at present. 3. See Davidse (1992) for terminology. 4. For extensive discussion see Haiman (1983) and Janko-Trinickaja (1962) to name only a few. 5. This is a homonym of the verb 'to carry'. 6. Brecht and Levine( 1984). 7. The four-volume Academy (1980-94). 8. Verbs like xoxlit' 'to ruffle [feathers]' and nastorozit' 'to prick up [ears]' were excluded. For a full list of such verbs, see Williams (1993).

294

Adger Williams

9. Use of the (b) option seemed to be governed by mention of contrasting activities by different parts of the Agent's body, though even then, it is not a preferred choice often. 10. I have also compared the dictionary definitions of -sja forms of verbs and non-sja forms to make sure that it is reasonable to suppose that the pattern holds outside of the realm of verbs of facial expression. For a discussion, see Chapter 4 of Williams (1993). 11. Actually, there are some Active two-participant events that can be presented using j/a-verbs, but they are limited to events that can be presented with converses or three-participant predicates. 12. Fractions represent choices of more than one option by an individual informant. In such a case, that informant's "vote" was split equally between the options he or she sanctioned. 13. See Gerritsen (1990) and Roughier-Willoughby (1993) for extensive discussion of the characterization of the subjects of sja-predicates by other elements in the predicate.

References Brecht, Richard and James Levine 1984 Conditions on Voice Marking in Russian. In M. Flier and R. Brecht (eds.), Issues in Russian Morphosyntax. Columbus: Slavica. Davidse, Kristin

????

Transitivity/Ergativity: The Janus-headed grammar of actions and events. In

Martin Davies and Louise Ravelli (eds.), Advances in Systemic Linguistics Recent Theory and Practice. London: Pinter. Evgen'eva, A. P. (ed.) 1981-84 Slvoar' russkogo jazyka. Moscow: Russkij Jazyk. Gerritsen, Nelleke 1990 Russian Reflexive Verbs: In Search of Unity in Diversity. Amsterdam and Atlanta: Rodopi. Kaiman, John 1983 Iconic and Economic Motivation. Language 59.4: 781-819. Janda, Laura 1992 Cognitive Linguistics as a continuation of the Jakobsonian Tradition: the semantics of Russian and Czech reflexives. Unpublished MS. Janko-Trinickaja 1962 Vozvralnye glagoly v sovremennom russkomjazyke. Moscow: Nauka. Kemmer, Suzanne 1988 Middle voice systems in typological and diachromc perspective. Stanford University: unpublished PhD dissertation. Langacker, Ronald 1987 Foundations of Cognitive Grammar, Vol. 1: Theoretical Prerequisites. Stanford: Stanford University Press.

Prototype marker or reflexive marker

295

Pederson, Eric 1989 The Typology of Reflexive Voice. Unpublished MS. Schenker, Alexander 1986 On the Reflexive Verbs in Russian. International Journal of Slavic Linguistics and Poetics XXXIII: 27-42. Vinogradov, V.V. 1972 [1945] Russkijjazyk (grammaticeskoe ucenie ο slave). Moscow: Vyssaja skola. Williams, Adger 1992 The Argument Structure of sja-Predicates, in Journal of Slavic Linguistics. Bloomington: Indiana University Linguistics Club. 1993 The so-called Reflexive Marker -Sja in Russian: semantically motivated syntax and cognitively motivated semantics. UCLA: unpublished PhD dissertation.

Metaphor

The strengths and weaknesses of the left/right polarity in Russian: diachronic and synchronic semantic analyses Alan Cienki

1. 1.1.

Introduction Strong polarity

In many languages, there is a semantic asymmetry between the words built on roots which refer to left and right spatially. The positive and negative connotations of the terms for right and left, respectively, has been studied in some detail before; see, for example, Fritsch (1964), Needham (1973), Laponce (1981), and van Leeuwen-Turnovcovä (1989a, 1989b, 1990, 1991). The asymmetry is associated in many languages with a difference in the extent to which the roots are productive in word formation, with 'right' words more prolific than 'left' words.1 This ratio of productivity applies to semantics as well, with 'right' words generally extending into more semantic domains than 'left' words. Russian serves as a prime example because it represents an extreme case of these differences. The Russian root from which the adjective pravyj 'right' is built, -prav-, is extraordinarily productive in comparison with the Russian root -lev-, which forms the adjective levy'] 'left'. The root -prav- combines with numerous affixes, especially prefixes, and covers a range of semantic fields, while -lev- is very limited in these respects. Accordingly, these roots and the words derived from them in Russian will provide the data for this case study. What is the basis for these asymmetries? To date there are few detailed semantic studies of the roots for 'left' and 'right' and their derivatives in any language. Van Leeuwen-Turnovcovä's works survey the range of IndoEuropean (IE) languages, including Slavic; encompass diachronic and synchronic data; and are insightful in their descriptive breadth as well as detail. However, they remain outside of any theoretical framework of semantic analysis. They leave unanswered questions as to the mechanisms engendering semantic change, the relative status held in any one language by the different

300

Alan Cienki

meanings of the 'left'- and 'right'- words, and the relations of the words to each other. Why do some meanings of the 'left'- and 'right'-roots arise in more different words than other meanings? The present study is informed by cognitive linguistic theory, one of the basic insights of which is that cognitive categories commonly exhibit a radial structure, with a prototype as the central case, and non-prototypical cases related to it - to a greater or lesser degree - via "family resemblances". Linguistic categories, being cognitive categories, have been found to exhibit the same structure in many, if not most, instances. This includes the semantics of lexical and grammatical items. Studies in cognitive linguistics have also revealed the important role of metaphor in synchronic as well as diachronic semantic extension, and how diachronic semantic analysis can thereby help explain contemporary polysemy patterns (as demonstrated in Sweetser 1990). It is hoped that this work on the semantics of word roots will complement previous cognitive semantic analyses which have, for the most part, focussed either on whole words or on grammatical and lexical affixes.

1.2.

Weak polarity

Research in the field of cognitive linguistics has shown that the metaphorical usage of spatial terms is normally based on the transformation of an "imageschema", some structure or pattern which organizes our experience (Johnson 1987; Lakoff 1987). However, the left-right spatial axis is very weakly polarized in terms of the human body and our daily conscious functioning: it plays a much less significant role than do the up-down and front-back axes. Of these three axes of orientation, it is the one: - in terms of which our bodies are most symmetrical The physical and functional differences between our front and back sides, and upper and lower halves of our bodies, are clear; but the physical and functional differences between our left and right sides, particularly with regard to external form, are far from clear, especially when considered non-scientifically. - most easily reversed The axis is weakly polarized in that "[t]here is no equivalent to gravity that would help us to distinguish left from right, [...]" (Laponce 1981: 7), and no parallel to our experience of usually moving forward, in the direction we are facing, versus backward. Many people, therefore, confuse left and

The left/right polarity in Russian

301

right, either occasionally or habitually, as a result of our bilateral symmetry and spatial experience along the left-right axis. - learned the latest Johnston (1988: 196) cites a number of studies based on different languages showing that children properly used the terms for On' and 'under' at age two, and for 'back' and 'front' at age four. In comparison, Corballis and Beale (1976: 142) discuss developmental studies which show that most children could reliably differentiate between left and right parts of their own bodies by age six, but that performance on other objects continued to improve even up to age ten. - most subject to referential confusion between inherent and relative directionality Inherent and relative direction coincide much more often in the up/down axis, for example. Given the standing or sitting upright positions in which social encounters canonically take place, what is up for you is up for me, and the same for down. This is not the case with left and right. This study will consider the questions: What is the basis for the strong semantic polarity of the two terms here given their relatively weak markedness in the spatial realm? Why, especially, has -prav- and its cognate roots in other languages been so productive, but not -lev- and its cognates? We will begin with an overview of the synchronic data from Russian.

2. 2.1.

Contemporary uses of the Russian roots -lev- and -pravSynchronic analysis of -prav-

In Russian, the productivity of the root -prav- in word formation illustrates the language's complex derivational morphology. There are several large nests of words (mostly nouns, verbs, and adjectives) built on -prav-, and the most common of these words in contemporary usage are listed in Appendix A.2 While the categories in Appendix A are based on this author's intuitive groupings, they echo and supplement the list of meanings given in Gribble (1981: 52):3 "right, rect-, just, correct, ortho-; direct, rule." These categories are clearly not distinct, but blend into each other to varying degrees, e.g., the categories TO CONTROL and TO RULE, GOVERN. The polysemous nature of many of the -prav- words places some of them into several categories. A listing of polysemous words from Appendix A along with

302

Alan Cienki

greatly simplified versions of their definitions as given in the Academy of Sciences contemporary dictionary is given in (1). The meanings of each word are connected by lines which indicate my interpretation of their "chaining structure" (Lakoff 1987), the connections between meanings in the semantic network that each word represents. Dotted lines are used when the Academy dictionary listed the meanings under two separate entries, e.g., pravyj^ and pravyJ2- The size of the type indicates the order in which the meanings appear in the Academy dictionary entry, larger type having primacy. Horizontal and vertical placement of meanings in a network is significant only in that the primary meaning(s) of a word can be considered to be central to each network (even when the other meanings are all given "to one side" of the primary meaning). (1)

a.

pravyj spatial

just b.

right

political right

correct

pravit' straighten

correct

govern c.

sharpen

control

popravit' straighten

improve

correct

repair d.

pravil 'nyj real

true

correct

evenly proportioned

The left/right polarity in Russian

e.

303

pravda justice

truth

really

law

f.

upravljat'

govern

• control

direct

g.

napravit' sharpen

h.

pravo

i.

ispravit'

direct

send

improve

j.

vypravit'

correct

repair

k. 1.

zapravit' straighten

make fit for use

cope with

control

season (food)

spravit'sja refer to

There is clearly a great deal of overlap between the semantic networks of these words. Given that simplified versions of the meanings of the monosemous words in the Academy dictionary are all represented in at least one of

304

Alan Cienki

these networks, a composite of all of the networks, as in (2), shows the complexity of the semantic category represented by -prav-. Where the size of the type is larger in (2) than in (1), it indicates that the given meaning, which was not primary for any polysemous word, was the simplified meaning of a monosemous word. (2)

A model of the semantic network of -pravpolitical right ι spatial right

evenly proportioned

just(ice)

a right - lafw govern cope with — carry across Two principles determine which meanings have a more central position in the model. One is the contemporary frequency of the meanings in different word forms, which can be seen in the number of words in each category in Appendix A. The other is based on Geeraerts' (1985: 139) observation about the synchronic prototype of a semantic category, that "within each cluster, central senses can be determined [...] by their logical preponderance with regard to the other senses (i.e. by the question whether those other senses can be derived from them)." A consequence of this structuring is that those meanings which are frequently part of a polysemous word network, e.g. 'correct', have a more central position in the model. The model is constructed as if the individual word networks were superimposed. Meanings that are intuitively close, such as 'correct' and 'repair', are thereby shown as more strongly connected (with thick double lines representing three connections in polysemous networks) than meanings that are not necessarily seen as closely related ('correct' and 'spatial right'). Note that 'spatial right', therefore, is nearly independent from

The left/right polarity in Russian

305

the rest of the network, connected only by the polysemy of the adjective pravyj (which, recall, the Academy dictionary divided into two separate word entries, and hence the dotted line connection). Semantic similarity also plays a role in horizontal and vertical placement of meanings in the network such that spatial proximity is intended to roughly reflect closeness of meanings. Another consequence of the model's structuring is that the meanings that are more centrally placed are the ones which are more abstract ('correct', 'truth', 'govern') or general ('straighten', 'repair') as opposed to the more peripheral ones which are domain specific ('political right', 'season (food)', 'carry across'). Not surprisingly, there are more interconnections with the broad, abstract concepts associated with -prav- than with the domain specific applications. While the semantic network for the root reflects the radial structure shown in many previous studies to be characteristic of cognitive categories (see citations in Lakoff 1987), its actual cognitive reality remains to be tested empirically.4

2.2.

Synchronic analysis of -lev-

The root -lev-, on the other hand, is the source of only a small group of words, and is most common only as an adjective, levyj (or as a related adverb). Most of the metaphorical extensions of the root appear in different applications of this adjective, as shown in Appendix B. Informal linguistic interviews conducted with native speakers of Russian in Moscow in the summer of 1993 revealed a range of colloquial and slang usage of the adjective levyj, but no such parallel usage of the adjective pravyj. Levyj in the meaning of 'illegal' and its affiliated sub-categories in Appendix B has been popular in colloquial usage since the early part of this century, as attested by my middle-aged Russian informants. While this usage was apparently too colloquial to appear in the references used to compile the Appendix, it does appear in Marder (1992: 200). The most recent semantic development of levyj appears in the category marked WRONG: usage in the sense of the two sub-categories, STRANGE and NOT AUTHENTIC, was restricted to young people up to the early 30's in age.5 The few polysemous words built on -lev- are given in network fashion in (3), following the same principles used with the polysemous -prav- words. The complex network for -lev- represented in (4) was constructed using the

306

Alan Cienki

same methods as for the semantic network of -prav-. (3)

a.

levyj used illegally

political left

spatial left

wrong

illegal

weak

strange

obtained illegally

not authentic

b.

nalevo spatial left

(4)

illegal

A model of the semantic network of -levweak

political left

./^

strange

spatial left. ^ss,.

^^ wrong illegal ^^^ obtained illegally

used illegally not authentic

2.3.

Synchronic comparison of -lev- and -prav-

The 'spatial left' meaning appears more integrally connected in the small network of -lev- than 'spatial right' does in the -prav- network. This reflects the fact that while the adverbs nalevo and sleva go beyond the spatial meaning to that of 'illegal', the morphological counterparts built on -prav- (napravo and sprava) do not extend to the semantic realm of legality. Besides the antonymy of 'spatial left' and 'spatial right', there are other broad antonymy relations to be found between words built on -lev- and -prav-, and a comparison of the models in (3) and (4) brings these out. Whereas some -lev-

The left/right polarity in Russian

307

words bear the meanings 'wrong', 'illegal', 'weak', and 'not authentic', various -prav- words have meanings relating to 'correct'/'true', 'just', 'control', and 'real'. The oppositions are therefore sometimes approximate at best. The misaligned antonymy of morphological counterparts will be discussed further in the Conclusions section. A comparison of the distilled meanings in the -lev- and -prav- networks also reveals that -lev- words, dominated by the adjective levyj, refer more to qualities, while -prav- words, many of which are verbs and nouns, refer to actions and (abstract) entities. An examination of the etymologies of these roots will help reveal the reasons behind this disparity in semantic and morphological development.

3. Historical development of the roots -prav- and -lev3.1. Development of -prav- through the llth century According to Pokorny (1959,1: 854) the root -prav- developed from I-E *prouo 'inclined forward'. Through the Common Slavic form pravu the root developed semantically into the notions (1) 'located ahead' expressed by Slavic pirv- 'first' and the prefix pra- 'pre-' or 'proto-', and (2) connection to the point in front via a straight path, expressed by Slavic prav-. This semantic origin of -prav- is the key to its productivity in word formation. Van Leeuwen-Turnovcovä (1990, 1991) discusses how the concept 'straight' functions not only as a structural element of geometry, but was an organizing principle in many domains of proto-Indo-European society, and remains so in many of the successor cultures today, including the Slavic ones. The factors she discusses, included in the historical overview of -prav- below, help explain the productivity of the root and its extension into words meaning, for example, 'control', 'correct', and 'true'.6 These factors include the following: - The use of the plow, which dates back at least to the Iron Age (Gimbutas 1971), not only increased the amount of usable farmland, but also served to distinguish the farmland from the surrounding wilderness. Land plowed in straight lines is thus marked off as territory that is controlled by humans (really 'by man', as men did most of the plowing). Cf. Common Slavic pravu. 'straight, even', and Old Russian praviti 'to make straight' (Cyganenko 1970: 364-365) and 'control, steer' (Sreznevskij 1989 [1893-1912] (henceforth SDJa) volume II, 1345).7

308

Alan Cienki

- The division of (farm)land along straight lines for fair distribution among those who worked the land (be it extended or nuclear families). Cf. Old Russianpravyj 'straight' and'correct'\pravu 'justly' (SDJall, 1348; 1352). - The straight forms used in building construction. Gimbutas (1971: 81) notes that even the earliest settlements verified as Slavic, the Zhitomir sites dating from the 6th century AD, consisted of semi-subterranean square houses. Houses in the 6th-7th century Volyntsevo site were "either built of horizontal logs, or constructed of vertical posts interspersed with wattleand-daub walls" (ibid.: 90) (emphasis added). Cf. Russian pravilo, a tool for making straight lines in construction and masonry (also, a rudder on a boat). These straight forms are repeated on a larger social scale in the change from an agrarian society to an urban one: in the organization of cities, furthering man's control of the natural world, as van LeeuwenTurnovcovä (1990, 1991) asserts. - The ordering of society, cf. Russian pravilo 'rule'. Van Leeuwen-Turnovcovä (1991: 295) discusses the logical development from 'straight' as a way of structuring order in physical space and 'straight' as a principle defining order in society. Cf. also Old Russian upraviti and Modern Russian upravljat' 'to control' and 'to rule, govern' among other meanings (SDJa II, 1345-1347); and the institution which maintains the ordered society, pravitel'stvo 'government'. We thus see an example of the tendency common in historical semantics for a word or root with a concrete meaning (here -prav- with the spatial meaning 'straight') to develop abstract metaphorical meanings (Sweetser 1990; Traugott 1982). Note, furthermore, that the two parallel forms built with the agentive suffix -lo differ only in stress: pravilo and pravilo. The former is a native Russian word, while the latter derives from Old Church Slavic (OCS). They follow the pattern in Russian whereby borrowings from OCS often have abstract designations, while cognate native Russian forms have more concrete referents. - Conforming to rules as being correct. As with English correct (< IE reg'to move in a straight line', and from which English right), there is a historical metaphor of straightness and order behind Russian provil'nyj 'correct'. This adjective was formed from the noun pravilo 'rule', and its Old Russian predecessor pravilmyi meant 'conforming to Church rules' (cf. English being in line with the rules) and therefore also 'lawful, legitimate' (SDJa II, 1344). - Many straight things, according to a folk-theoretical view of the world, are

The left/right polarity In Russian

309

also firm, solid, and consistent; similarly, things believed to be true are solidly consistent with known facts or one's perception of reality. Russian pravda 'truth' (Old Russian pravida) is derived from the addition of the no longer productive noun-forming suffix -da (Old Russian -idd) to the root prav-. Given that the suffix indicated an abstract noun with the quality determined by the meaning of the root, the Old Russian noun pravida came to have a wide range of senses (SDJa lists 22), among them: 'truth', 'justice', 'correctness', 'an oath', 'a right', 'court', and 'a set of laws'.8 The last sense appears in the name of Russia's first legal code, the Russkaja Pravda, dating from the eleventh century. Van Leeuwen-Turnovcovä (1991: 296) notes that law for the Slavs in the early Middle Ages, was common law; the law, customs, and morals were in harmony. Kaiser (1980: 4) concurs that, "[...] in its infancy Russian law came much closer to the system of consensual norms which governs those societies without a developed state structure [...]." The purpose of the justice system was the restoration of (social) order (bringing things 'into line') that had been disturbed by some action, and compensation rather than the administration of punitive measures (van Leeuwen-Turnovcovä 1990: 259). In a punitive justice system, which Kaiser characterizes as 'vertical', normally a third party, an institutionalized higher authority, metes justice. She notes that the early medieval Russian legal system, however, was 'horizontal' in structure, and observes: "Societies where this kind of justice prevails are usually homogeneous, small, and bound by kinship relations" (1990: 7). It is not surprising, then, that in a society sharing a world view sufficiently uniform such that consensus could serve as the basis for the legal system, the concepts 'correct', 'truth', and 'law' were based not only on the same root, -prav-, but were represented by the same word, pravda. That the root of the word also meant 'straight' neatly supports Johnson's (1993) observations about moral principles that are considered shared, stable, and unproblematic: they are viewed by society as absolute. It should be mentioned that by the 11th century, there already existed several prefixed verbs and nouns built on -prav-. Most of these prefixes are spatial in origin, and the words formed with them had spatial meanings and, like most of their contemporary descendants, metaphorically derived meanings. For example: - from iz- Out of - ispraviti 'to correct' [< 'to straighten out'], 'to return', and others (SDJa I, 1132-1134); - from na- 'on(to)' - napraviti 'to direct, send' (SDJa II, 311);

310

Alan Cienki

- from u- 'away' - upraviti 'to send, direct', 'to extend', 'to rule' and others (SDJa II, 1244-1245); - frompo- (simple perfective prefix) -popraviti 'to correct', 'to send off', 'to fulfill'(SDJa II, 1196).

3.2. Spatial right It becomes evident from this brief historical overview that the spatial concept of 'right' was not part of -prav-'s early semantic development. In fact there was another root in Old Russian and other Slavic languages which referred to the right side, namely desn- (· pravil 'nyj 'correct'

'really'

(limited to historical reference to Russkaja Pravda)

The left/right polarity in Russian

313

developed from a root meaning 'straight', -lev- is believed to have been developed from IE *laiuo-s 'crooked, bent' (Pokorny 19591: 652). Unlike -prav-, -lev- did not semantically replace a pre-existing spatial term, but did overlap with a parallel root meaning 'left' in Old Russian, suj-. It is worth noting that the semantic history of suj- also parallels that of -lev-: according to Pokorny (1959 I: 914), suj- comes from IE *seu- 'to bend, turn'. L'vov (1966: 269-272) supposes that both roots came into Common Slavic and both meant 'left' by the ninth century. As the Slavic languages differentiated between the tenth and the fifteenth centuries, suj- (and its variants) was no longer common to all of them, but survived only in some dialects and languages. In the other languages, including Russian, -lev- predominated to mean 'left', and suj- eventually became obsolete. A partial schematic time line of the semantic history of -lev- is represented in (6).

(6)

-levlevSa (?) left-hander'

Common Slavic lea 'left', 'bent?'

levyi —> levyj 'spatial left' levyj (early 20th C.) levyj

'political left' levy] (early 20th C.)

3 u ο

''illegal ^s. levyi (late 20th C.) levyj 'wrong' —> (narrower contexts)

'not authentic' (colloq.) levyj (late 20th C.) 'strange' (colloq.)

The etymological meanings of -lev- and -prav-, therefore, exhibit antonymy of a sort other than the usually symmetric (in bodily terms) opposition of left and right. The two concepts 'straight' and 'bent' can be considered opposites as well, along a semantic scale of forms. As part of this study, several native speakers of Russian in Moscow were asked whether they could discern any concept that held together the various contemporary senses of the adjective levyj. It is interesting to note that, unaware of the word's etymology, several cited the notion of otklonenie ('bowing/bending away') from some standard or norm.

314

Alan Cienki

3.5. On metaphors from 'bent' and 'straight' While I am not proposing that people hold the histories of words in their heads, I do believe the data for both -lev- and -prav- show how a root's earlier meaning (here 'bent' and 'straight', respectively) can be maintained and developed as associations between words in a network. The spatial concepts 'straight' and 'bent' allow for logical metaphors for understanding 'right' and 'wrong': while I suspect most people could agree on a standard or prototypical image of what physically straight looks like, I doubt they could agree in the same way on the prototypical image of 'bent'. There are many different shapes into which something can be bent, whereas 'straight' is more likely to be viewed as an absolute form. Similarly, that which is correct (say, a correct answer), from a folk-theoretical perspective, has an absolute quality, as opposed to the many ways in which something can be wrong. The metaphorical connections between the concepts 'bent' and 'wrong', and between 'straight' and 'correct',11 are not only reinforced and maintained through daily experiences, but are also reflected and reinforced through other linguistic metaphors that have made for examples of polysemy similar to those seen for -lev- and -prav-. One example is the Russian root -kriv- 'crooked, curved', for example. Its spatial meaning is apparent in words like krivoj 'crooked' and krivit' 'to twist, contort', while metaphorical extensions of this meaning appear in krivotolki 'false rumors, idle gossip' and expressions like krivit' dusoj 'be hypocritical' (literally, 'to twist the soul'). The root -rjad- Order; row', on the other hand, gives rise to porjadok Order', and also porjadocnyj 'honest, decent'. (See also van Leeuwen-Tumovcova 1991: 295, 298.) In this connection, I would suggest that, in addition to representing two spatial image-schemas, the opposition 'straight' versus 'bent' represents a pair of complementary force-dynamic image-schemas (Talmy 1988), and reflects the ideas of resistance to some force versus submission. The metaphorical reflections of this are apparent in the different meanings of Russian words from the roots -prjam- and -gib-. The root -prjam- 'straight', for example, appears not only in the adjective prjamoj 'straight', but also in uprjamyj 'stubborn'. This contrasts with -gib\- 'bend, flexible' as in gibkij 'flexible, supple', and the related root -gib*}- 'perish' as in gibnut' 'to perish'. (See also Cienki 1998a.)

The left/right polarity in Russian

3.6.

315

Handedness

Many languages and cultures reflect a correlation between straight forms, positive qualities, and strength, on the one hand, and between crooked ones, negative values, and sometimes weakness, on the other hand. This polarity is associated in most cultures with the strength/weakness polarity of most people's hands. Corballis (1983: 13) suggests that the proportion of natural left-handers in human populations is 10-12%. Of the numerous studies cited by Laponce (1981) for different countries over the past century, seldom does the percentage of left-handers exceed 10-12%. Looking at the other side of the coin, van Leeuwen-Turaovcova (1990: 145) claims that 80% prefer using the right hand with many activities, leaving a generous 20% figure for lefthanders. Whatever the exact figures, it is no surprise that right-handers make up the majority in most societies, stemming from the functional asymmetry of the human brain (Corballis and Beale 1976), and practices in many cultures which reinforce uniform dominance of the right hand. The asymmetry in strength between most people's right and left sides, therefore, provides an embodied, experiential basis for the metaphors and symbolism (Ivanov and Toporov 1965) related in many cultures with left and right. Left/right metaphors are orientational metaphors (see Lakoff and Johnson 1980, ch. 6 for others) that draw on both spatial and force-dynamic image Schemas.

3.7.

Political left and right

Laponce (1981) discusses the source of the political connotations of the terms for 'left' and 'right' in so many languages today. In late 18th century France, the revolution brought down the vertically symbolized power of the monarchy and gave rise to a more democratic, 'horizontal' political order. The new French parliament in Versailles was spatially organized such that the clergy, the aristocracy, and supporters of the king were on the right side when viewed from the front of the hall, and those against the monarchy, for democracy and for change, were on the left. The terminology of left and right as aligned with the respective ideas of change versus status quo quickly spread to the description of politics outside the parliament, and from there throughout France, Europe, and the rest of the world. Laponce (1981: 52) notes the two main channels of dissemination were the language of parliamentary democracy and the language of socialism, the latter of which was clearly the

316

A Ian Cienki

prime vehicle for introducing left/right terminology into Russian and Soviet politics. 'Left' and 'right' are, of course, terms of relative as well as inherent directionality. The same referent, in spatial or political terms, could be described as on the left or right depending on the relevant frame of reference. In political terms, however, 'left as change' and 'right as status quo' constitute stable elements in the polarity. Viewed in force-dynamic terms (Talmy 1988), the contrast between a steady-state force (at rest or in motion) versus a shifting force has interesting repercussions with the earlier meanings of the terms for 'right' and 'left' in many IE languages, namely 'straight' and 'bent' respectively. In terms of naive, "billiard-ball" physics, inertia will keep a ball rolling on a straight path if the force of another ball doesn't knock it and bend its path. Thus while the political usage of the terms originated from metonymy based on the political views of those on the two sides of the French parliament, it is reinforced (in Russian and some other languages) by the metaphors, discussed earlier, which draw on the 'straight' and 'bent' image-schemas. This particular interaction of metonymy and metaphor, here metaphor from metonymy,12 is not only a phenomenon in the history of Russian, as the following contemporary example shows. In an editorial in one Russian newspaper13, an author makes the following comment on nationalist extremists: (7)

Sleva - Svoboda s bombami, sprava on-the-left - Freedom with bombs-INSTRL, on-the-right Porjadok s tankami. order with tanks-INSTRL On the left there is Freedom with bombs, on the right there is Order with tanks.'

Thus order is associated here with the right as well as with strong control.

4. Conclusions The low significance of the left/right axis in our spatial orientation is reflected in the fact that in Russian and other Indo-European languages, the terms identifying these lateral sides were adopted from a different spatial distinction, that of straight versus bent. The historical association of the "bent" and "straight" roots, -lev- and -prav-, with the left/right spatial axis appar-

The left/right polarity in Russian

317

ently developed in part in connection with force-dynamic images relating to weakness and strength. The experiential gestalts conveyed by -lev- and prav- extended to the metaphorical development of these roots, expressing the realms of social norms of goodness, legality, and truth. In the spatial senses of -lev- and -prav-, there is a semantic and morphological parallelism in the adverbs referring to motion/direction to or from, which is typical of the system of spatial expressions in Russian, as shown in (8). (8)

Russian adverbs of lateral motion/direction and location left right

to that side nalevo napravo

from/on that side sleva sprava

However, the semantic parallelism breaks down when speaking about some kind of metaphorical motion or direction with reference to the two different sides. For example, the opposite of rabotat' nalevo 'to work illegally or unofficially' or 'to moonlight' is not *rabotat' napravo; the morphologically parallel form napravo is not used to mean 'correct'. This could reflect the fact that the metaphorical meaning related to -lev- here does not stem from the meaning 'spatial left', but rather from the notion of 'bent' discussed earlier. Spatial left and right often oppose each other in our experience in a symmetric way, as mirror opposites (think of human and animal bodily forms). 'Straight' and 'bent', however, are asymmetric antonyms, and their metaphors do not oppose each other in a parallel fashion. The root -lev- is, in fact, so limited in the variety of its word formation that there are few cases where morphological forms parallelling those of -prav- exist in opposing semantic domains (witness the many different prefixed forms of -prav- and the dearth of prefixed -lev- words). The adjective levyj is the one -lev- word that has expanded into the most semantic fields, however the contexts in which is it antonymous with pravyj are limited. The phrase levoe delo is not a true antonym ofpravoe delo 'righteous cause': The former would actually refer to a business matter On the side' that is probably illegal, rather than an 'unrighteous cause'. The antonym some of my young Russian informants produced for levye krossovki 'not quality, name-brand sneakers' was not *pravye krossovki, but rodnye krossovki, literally 'native sneakers', i.e., made where they are supposed to be made. Where morphologically parallel forms of -lev- and -prav- words are antonymous, it is because they oppose each other along a common semantic axis (cf. Lindner 1982),

318

Alan Cienki

e.g., in the spatial semantic field (levsa 'left-handed person', pravsa 'righthanded person'), and the metonymically-related political field (polevet' 'to shift to the left, become more radical', popravet' 'to shift to the right, become more conservative'). This historical overview of the semantic development of the roots -levand -prav- supports Geeraerts' (1985) conclusions on the structure of semantic change, that semantic categories structured in prototype fashion have a dynamic character that is an essential aspect of their nature. Briefly put, the prototypical structure of cognitive categories - with some concepts more central and others more peripheral - allows words, roots, and affixes to adapt and develop semantically in new directions according to changing circumstances. The meaning 'straight' appears to have been central in the radial structure of the semantic category -prav- for most speakers by the early period of Old Russian; by the Modern Russian period, however, earlier peripheral meanings (e.g., 'control', 'truth') gained currency, and formed a network of meanings that are now central to -prav- and have served as sources for the development of new meanings. One could imagine a three-dimensional synthesis of the two-dimensional models in (2) and (5), whereby turning the synchronic network in (2) on its side would allow one to see the diachronic history (5) trailing behind it, and therefore see how the central and peripheral roles of the root's meanings have changed. We have also seen how metaphors, apparently used earlier in a root's history as a source of semantic extension, can remain as active conceptual metaphors (cf. Sweetser 1990). The polysemy that developed for -prav- by the Old Russian period was based on the metaphors such as CORRECT is STRAIGHT and TRUTH is STRAIGHT. These metaphors can be viewed as derived from the folk theory of truth as absolute ("There is one and only one correct way to understand reality" [Lakoff 1987: 294]), and from our visual experience; one example is that just as a flat reflective surface (e.g., water, a mirror) reflects a more accurate image of the world as we see it than a curved or wavy surface does, so does truth correspond better to what is known or perceived to be real than falsehood. Various idioms reflect the same metaphor today. For example, the idiom krivit' dusoj 'to be hypocritical' ['to bend the soul'] is based on the logical counterpart of the metaphor that TRUTH IS STRAIGHT, namely FALSEHOOD IS BENT. Other idioms and sayings related to truth in contemporary Russian reflect the combination of spatial and force-dynamic image-schemas mentioned earlier in the notion that something straight is often solid, and firm or hard. Thus in Russian, an idiom meaning to speak what

The left/right polarity in Russian

319

you think directly and frankly is rezat' pravdu 'to cut the truth', metaphorizing truth as a solid object. A saying that reflects truth as not only hard, but also sharp, is pravda glaza kolet 'truth stabs the eyes', said when someone tries to object to critical, but correct, comments made about them (cf. English the truth hurts). Poetic metaphor can draw on the same image-schemas: (9)

U Krivdy - sto lazeechek, u Pravdy at falsehood-GEN - hundred loophole-GEN-PL, at truth-GEN ni odnoj; u Krivdy —put' izvilistyj, u Pravdy-put' not one-GEN; at falsehood-GEN - path winding, at truth - path prjamoj. straight. 'Falsehood has a hundred loopholes, Truth - not one; Falsehood's path is winding, Truth's path is straight.' (Bednyj, in M. L'vov 1984: 208)

These lines employ both an OBJECT metaphor and a SPATIAL metaphor. The first is that TRUTH IS SOLID (having no loopholes). The second involves what Johnson (1993: 36-39) calls the LOCATION version of the EVENT STRUCTURE metaphor, namely that ACTIONS ARE MOTIONS ALONG PATHS from one state/location to another. The verse adds detail to this, specifying that truth follows a straight path. If 'speaking' is the action that is presumably implied, and THE PATH OF TRUTH is STRAIGHT, then SPEAKING THE TRUTH IS MOTION ALONG A STRAIGHT PATH14 between speaker and listener. An obvious inference that derives from this, then, is that since a straight line is the most direct path between two points (the best way to get from A to B), telling the truth is the best way for two parties to communicate. As a final point, the Russian data show a derivative metaphor from ACTIONS ARE MOTIONS ALONG PATHS and its component PURPOSES ARE DESTINATIONS, namely DIRECTING AN ACTIVITY IS KEEPING IT MOVING

ALONG A PATH. This metaphorical structuring helps explain the intuitive logic behind some of the examples of polysemy seen earlier, such as the fact that bothpravif' and upravljat' mean 'to drive' as well as 'to rule, govern'. The issue as to whether this metaphor's underlying PATH schema must necessarily be straight is pursued in a separate study (Cienki 1998b).

320

Alan Cienki

Appendix A: Examples of contemporary uses of the Russian root -pravThe words in Appendices A and B are grouped by general categories, and the categories are ordered according to the quantity of words representing them. More specific subcategories are indented underneath more general categories. Illustrative compound words (with a root in addition to -prav- or -lev-, respectively) are included in some categories, indented, and followed by a more literal translation in curly brackets. The list is not exhaustive, but, I believe, representative. CORRECT

prav, pravyj - [adj.] right, correct pravil'nyj- [adj.] right, correct pravil'nost' - [noun] lightness, correctness pravka - [noun] correcting (proofreading) pravota - [noun] tightness, correctness pravil' - [verb] to correct (something written) [see also 'to control'] popravljat'/popravit'- [verb] to correct popravljat'sja/popravit'sja- [reflexive verb] to correct oneself popravimyj - [adj.] rectifiable (error) popravka - [noun] 1. correction. 2. amendment popravocnyj - [adj.] correction [attrib.] vypravljat'/vypravit' - [verb] to correct (a manuscript, proofs, etc.); to rectify, straighten out vypravljal'sja/vypravit'sja - [reflexive verb] to get straightened out, straighten oneself out ispravljat'/ispravit' - [verb] to correct ispravimyj- [adj.] repairable, rectifiable nepopravimyj - [adj.] irreparable; irretrievable ispravitel'nyj - [adj.] corrective, remedial ispravlenie- [noun] 1. repairing, correcting. 2. correction perepravljat'/perepravit' - [verb] to correct [colloq.] spravedlivyj - [adj.] correct, valid pravoslavie - [noun] the Orthodox faith {'right-praising'; caique from Greek} pravoslavnyj - [adj.] orthodox; member of the Orthodox Church pravomernyj- [adj.] legitimate {'right-measured'} pravopisanie - [noun] orthography, spelling {'correct-writing'} pravosudie - [noun] justice {'right-judgment'} PHYSICALLY SET STRAIGHT, CORRECT

popravljat'/popravit' - [verb] to straighten, adjust, set right vypravljat'/vypravit' - [verb] to straighten vpravljat'/vpravil' - [verb] 1. to set (a bone, a joint, etc.). 2. to tuck into zapravljat '/zapravit' - [verb] to tuck in, tuck under opravljat'/opravii' - [verb] 1. to straighten, adjust. 2. to mount, set in a mount oprava - [noun] mount, setting; frame

The left/right polarity in Russian

321

opravljat'sja - [reflexive verb] to straighten one's clothes opravka - [noun] a spindle used to support material being milled podpravljat'/podpravit'- [verb] to fix up, straighten raspravljat'/raspravit' - [verb] to smooth out, straighten, spread out raspravljat'sja/raspravit'sja- [reflexive verb] to get smoothed out RETURN TO ITS ORIGINAL OR PROPER STATE OR ORDER; MAKE FIT FOR SOME PURPOSE; CORRECT; WORKING popravljat'/popravit' - [verb] to repair, fix, mend popravimyj - [adj.] not beyond repair(of a situation) popravka- [noun] 1. repair. 2. adjustment ispravljat'/ispravit' - [verb] repair ispravnost' - [noun] good condition, good working order ispravnyj - [adj.] 1. in good condition, in good working order. 2. conscientious, industrious napravljatVnapravit' - [verb] to sharpen ['set straight'?] nepopravimyj - [adj.] irreparable, irretrievable nepopravimost' - [noun] irreparability HEALTH popravljat'/popravit' - [verb] to restore (one's health) popravljat 'sja/popravit 'sja - [reflexive verb] to get well popravka - [noun] recovery opravljat 'sja - [reflexive verb] to recover (from an illness) WEIGHT

popravljat'sja/popravit'sja-

[reflexive verb] to put on weight

SITUATION

spravljat'sja/spravit'sja-

[reflexive verb] to cope with, handle

FOOD

pripravljat'/pripravit' - [verb] to season, flavor priprava - [noun] seasoning, dressing, condiment zapravljat'/zapravit' - [verb] to season with zapravka - [noun] seasoning TRANSPORTATION

dozapravljat'/dozapravit'- [verb] to refuel (a plane) dozapravljat'sja/dozapravit'sja - [reflexive verb] to be refueled dozapravka - [noun] refueling zapravlat'/zapravit' - [verb] to put fuel in zapravljat'sja - [reflexive verb] to take on fuel, refuel zapravka - [noun] refueling zapravocnyj- [adj.] refueling [attributive] JUST, MORALLY CORRECT; JUSTICE pravyj - [adj.] right, just pravda - [noun] justice pravil'nyj- [adj.] proper

322

Alan Cienki po pravu - [PP]rightfully,deservedly pravednyj- [adj.] 1. pious, religious. 2. just, righteous pravednik/pravednica - [noun]righteousman/woman pravdist/ka - [noun] crusader, advocate of a cause; evangelist pravednost' - [noun] righteousness ispravljal'/ispravit' - [verb] to reform ispravljat'sja/ispravit'sja- [reflexive verb] to reform, mend one's ways opravdyvat'/opravdat' - [verb] 1. to justify (prove to be right). 2. to excuse. 3. to acquit opravdyvat'sja/opravdai'sja - [reflexive verb] 1. to justify oneself/itself. 2. to (try to) prove one's innocence. 3. to prove to be correct opravdanie- [noun] 1. justification. 2. excuse. 3. acquittal opravdatel'nyjprigovor- [NP] verdict of "not guilty" spravedlivyj - [adj.] just, fair spravedlivo - [adv.] fairly, justly spravedlivost' -justice, fairness uprava - [noun] justice [colloq.] BREAKDOWN OF JUSTICE

raspravljat'sja/raspravit'sja - [reflexive verb] to deal with (severely or cruelly) rasprava - [noun] harsh treatment, reprisals FINDING OUT WHAT IS CORRECT

spravljat'sja/spravit'sja- [reflexive verb] 1. to inquire. 2. to consult (a book, dictionary) spravocnik - [noun] reference book; directory spravocnyj - [adj.] reference, information [attrib.] spravka- [noun, usually pi.] 1. reference. 2. information. 3. certificate DIRECT, SEND; DEPART FROM A STARTING POINT

napravljat'/napravit' - [verb] 1. to direct. 2. to aim or point something ['set straight']. 3. to send napravljat'sa/napravit'sja - [reflexive verb] to head for, make for napravlenie- [noun] 1. direction. 2. trend. 3. (military) axis napravlennost' - [noun] direction, orientation napravlennyj [adj.] directional (radio) otpravljat' /olpravit' - [verb] to send, dispatch otpravljai'sja/otpravit'sja - [verb] 1. to leave, start, set out. 2. to proceed from (in one's thinking) otpravitel'/nica- [noun] sender otpravka - [noun] dispatch, shipment [colloq.] otpravlenie- [noun] 1. sending, dispatch. 2. departure. 3. item of mail otpravnoj - [adj.] 1. dispatch, shipping [attributive]. 2. initial: otpravnoj punkt - starting point perepravljat'/perepravit' - [verb] 1. to carry across, ferry across. 2. to forward perepravljat'sja/perepravil'sja - [reflexive verb] to cross pereprava - [noun] 1. crossing (of a river). 2. place to cross (a river)

The left/right polarity in Russian

323

DIRECT(WITH A PURPOSE)

napravlenie- [noun] 1. assignment. 2. order, permit napravlennyj - [adj.] purposeful otpravljat' - [verb] to discharge, perform otpravlenie - [noun] 1. exercise, performance, discharge. 2. [pi.] functions (of an organism) spravljat'/spravit' - [verb] to celebrate (an occasion) [colloq.] upravljaemyij - [adj.] guided (missile) TO CONTROL

pravit' - 1. [verb] to drive, steer (a car). 2. [verb] to sharpen, hone ['make straight, right'] upravljat'/upravit' - [verb] 1. to govern, control. 2. to operate (a machine); drive (a car). 3. to conduct (an orchestra). 4. to govern (grammatically) upravlenie - [noun] 1. driving (a car); conducting (an orchestra). 2. mechanical control upravljat'sja/upravit'sja - [reflexive verb] 1. to finish (with). 2. to cope with, deal with, handle CONTROLLED

pravil'nyj - [adj.] regular (verb, facial features); proper (fraction) pravil'nost' - [noun] regularity vypravka - [noun] bearing, carriage upravljat'sja - [reflexive verb] (of a vehicle) to ride, handle (a certain way) TO RULE, GOVERN [cf. DIRECT]

pravit' - [verb] to rule, govern pravitel' - [noun] ruler, one who rules pravilo - [noun] a rule pravitel'stvo - [noun] government pravitel'stvennyj - [adj.] governmental pravlenie - 1. [noun] government, rule. 2. [noun] management, board of directors zapravljat' - [verb] to boss, run [colloq.] zapravila - [noun] boss;ringleader,instigator [colloq.] upravljaf' - [verb] 1. to manage, administer. 2. to rule, govern upravljajuscij - [adj.] manager upravitel' - [noun] manager, steward [obscure] upravlenie- [noun] 1. management. 2. government. 3. administration, board, agency uprava - [noun] (pre-revolutionary) council, board upravlenceskij - [adj.] administrative TRUE pravda - [noun] truth pravda - [adv.] really, admittedly, to be sure; [interjection in questions] isn't that so pravo - [particle] really [colloq.] vpravdu - [adv.] really [colloq.] zapravskij - [adj.] real, true, regular [colloq.] pravdivyj - [adj.] truthful pravdivost' - [noun] truthfulness, veracity

324

Alan Cienki opravdyvat'sja/opravdat'sja - [reflexive verb] to be realized, come true pravdopodobie- [noun] plausibility {'truth-resembling,' verisimilitude} pravdopodobnyj - [adj.] plausible pravovernost' - [noun] orthodoxy {'true-faith'} pravovernyj-[adj.] 1. orthodox. 2. faithful {'true-believer'}

LAW

pravo - [noun] law (as a subject) pravovoj - [adj.] legal, relating to the law pravoved - [noun] lawyer, jurist {'law-knower'} pravovedenie - [noun] science of law, jurisprudence {'law-knowledge'} pravomocnyj- [adj.] competent {'law-able'? 'truly-able'?} pravonarusenie - [noun] offense {'law-violation'} pravonarusitei - [noun] offender; lawbreaker {'law-violater'} pravoporjadok - [noun] law and order {'law-order'} pravosposobnost' - [noun] (legal) capacity {'law-capable'} pravosposobnyj - [adj.] (legally) capable A RIGHT, SOMETHING ALLOWED

pravo - [noun] a right vprave - [adv.] having a right (to be or do something) prava - [pi. noun] (driver's) license na pravakh - [PP] as, in the capacity of bespravie - [noun] absence of rights bespravnyj- [adj.] without rights SPATIAL RIGHT

pravyj- [adj.] right pravsa - [noun]right-handedperson vpravo - [adv.] to the right napravo - [adv.] to the right; on the right sprava - [adv.] from the right; on the right pravobereznyj - [adj.] situated on the right bank 'right-bank' pravostoronnij - [adj.] right-side [attrib.] 'right-side' POLITICAL SIDE

pravyj - [adj.] right, status quo; [plural] the left pravet Vpopravet' - [verb] to shift to the right, become more conservative

Appendix B: Examples of contemporary uses of the Russian root -lev- spatial side levyj - [adj.] left levsa - [noun] left-handed person vlevo - [adv.] to the left nalevo - [adv.] to the left; on the left sleva - [adv.] from the left; on the left

The left/right polarity in Russian

325

levobereznyj - [adj.] situated on the left bank {'left-bank'} levostoronnij - [adj.] left-side [attrib.] {'left-side'} LEFT HAND (AS LESS SKILLFUL, WEAK)

(delat' cto-to) odnoj levoj - [NP] to do something easily, "with one hand tied behind one's back" {'with one left [hand]'} [colloq.] POLITICAL SIDE

levy] - [adj.] left, left-wing; [plural] the left levacki - [adj.] leftist level Vpolevet' - [verb] shift to the left, become more radical levizna - [noun] leftism levak - [noun] left-winger [pejorative] WRONG

levy] - [adj.] wrong (side of material; path) vstat' s levoj nogi - to get up "on the wrong side of the bed" {'from the left leg/foot'} (idti/guljat') nalevo - [adv.] to cheat on one's spouse {'to go/walk to the left'} [colloq.] STRANGE, NOT RIGHT [very colloquial] levyj (celovek) - [adj.] strange (person) ILLEGAL, FOR ONE'S OWN PROFIT

levyj - [adj.] illegal, bootleg [goods] (rabotat', prodat') nalevo - [adv.] (to work, sell) illegally levak - [noun] one who works illegally for his/her own profit [colloq.] USED FOR ILLEGAL PURPOSES [colloquial]

levyj (magazin, masina, ...) - [adj.] _?_ (store [a "front"], car, ...) OBTAINED BY ILLEGAL MEANS [colloquial]

levyj (stol, masina, ...)- [adj.] illegally obtained (table, car [a "hot" car], ...) NOT AUTHENTIC, BAD QUALITY [very, colloquial] levyj (televizor, Ribok) - [adj.] sham/bad imitation of (television, Reeboks)

Acknowledgments Earlier versions of this paper were presented at the Third International Cognitive Linguistics Conference (Leuven, Belgium, July 1993) and at the annual meeting of the American Association for the Advancement of Slavic Studies (Honolulu, November 1993). I am grateful to members of those audiences for their questions, and to Katja Rakhilina for comments on a draft of the paper. I also appreciate the suggestions received (especially from Nina Arutjunova and Andrej Kibrik) when I presented a talk on this material at the Russian Academy of Sciences Institute of Linguistics in Moscow (June 1994).

326

Alan Cienki

Notes 1. See Laponce's (1981: 42) survey of the length of dictionary entries for 'left' and 'right' root words and their derivatives in twelve languages. Chinese is the only exception in the list. 2. The main sources of the data in the appendices are the Soviet Academy of Sciences Slovar' russkogo jazyka [Dictionary of Russian] (1981-1984), Kuznecova and Efremova (1986), and Katzner (1984). 3. Meanings given for other roots to be discussed are also from Gribble (1981). 4. An experiment which might address this problem would involve having native Russian-speaking subjects judge whether specific words represent better or worse examples of the use of -prav-. The object would be to see if words (such as pravit'), which encompass several meanings given here as central in the model, are in fact judged to be more prototypical examples of the use of -prav- than words (such as zapravit') expressing meanings on the periphery. 5. This usage correlates with the current drastic increase in the amount of imported western goods for sale, especially in the major cities, and an increasing consumer-consciousness. In particular, young people in metropolitan areas who are being inundated with western advertising on television do not want just any new clothes or electronic conveniences, they want "the real thing", brand name products. 6. I view van Leeuwen-Turnovcovä's analysis as providing examples of daily life experiences which help reveal the motivation behind the semantic extensions which occurred. I have included the Russian examples for comparison, even though not all of them stem directly from the situations she discusses. 7. SDJa attests dates of textual evidence of a word's use with a given meaning with quotes from, and citations of, Old Russian manuscripts. 8. The connection between 'straightness' and 'resistance to force' will be discussed further below. 9. The root desn- survives today in the adjectival form with the meaning 'spatial right' in South Slavic (Bulgarian and Macedonian desen, Croatian and Serbian desni, and Slovene desni). 10. These uses are also less likely to appear in SDJa since they are more characteristic of spoken than written language. 11. English wrong (< IE *u???er- 'to turn, bend') and correct (< IE *reg- 'to move in a straight line'), of course, are examples of the same historical metaphor. 12. See Goossens (1990) on the ways in which metaphor and metonymy can interact. 13. Nezavisimaja gazeta [Independent Newspaper], 11 November 1992. 14. George Lakoff (personal communication) and his students have observed the same metaphor in English, e.g., "I can't get a straight answer from him".

The left/right polarity in Russian

327

References Arutjunova, Nina D. 1991 Istina: Fön i konnotacii [Truth: Background and connotations]. Logiceskij analiz jazyka: kul'turmye koncepty [The Logical Analysis of Language: Cultural Concepts], 21-30. Moscow: Nauka. Avanesov, Ruben I. (ed.) 1988Slovar' drevnerusskogo jazyka (XI-XIV vv.) [Old Russian Dictionary (9th14th C)]. Moscow: Russkij jazyk. Buck, Carl Darling 1949 A dictionary of selected synonyms in the principal Indo-European languages. Chicago: University of Chicago Press. Cienki, Alan 1998a Slavic roots for 'straight' and 'bent': Experimental gestalts, conceptual metaphors, and cultural models as factors in semantic change. In: Alan Timberlake (ed.), American Contributions to the Twelfth International Congress ofSlavists, 298-313. Bloomington, IN: Slavica. 1998b STRAIGHT: An image schema and its metaphorical extensions. Cognitive Linguistics 9: 107-149. Corballis, Michael C. 1983 Human Laterality. New York: Academic Press. Corballis, Michael C. and Ivan L. Beale 1976 The Psychology of Left and Right. Hillsdale, NJ: Lawrence Erlbaum Associates. Cyganenko, Galina P. 1989 Etimologiceskij slovar' russkogo jazyka [Russian Etymological Dictionary]. Kiev: Radjans'ka akola. Dal', Vladimir 1955 Tolkovyj slovar' zivogo velikorusskogo jazyka. [Defining Dictionary of the Living Russian Language] 3 vols. [1880-1882] Moscow: Gosudarstvennoe izdatel'stvo inostrannykh i nacional'nykh slovarej. Fritsch, Vilma 1964 Links und Rechts in Wissenschaft und Leben. Stuttgart: Kohlhammer. Geeraerts, Dirk 1985 Cognitive Restrictions on the Structure of Semantic Change. In Jacek Fisiak (ed.), Historical Semantics: Historical Word-Formation, 127-153. New York: Mouton. Gimbutas, Marija 1971 The Slavs. New York/Washington: Praeger. Goossens, Louis 1990 Metaphtonymy: The Interaction of Metaphor and Metonymy in Expressions for Linguistic Action. Cognitive Linguistics 1: 323-340. Gribble, Charles E. 1981 Russian Root List. Columbus, OH: Slavica.

328

Alan Cienki

Ivanov, Vjaceslav and Vladimir Toporov 1965 Slavjanskie jazykovye modelirujustie semioliceskie sistemy [Semiotic Modeling Systems of the Slavic Languages]. Moscow: Nauka. Johnson, Mark 1987 The Body in the Mind: The Bodily Basis of Meaning, Imagination, and Reason. Chicago: University of Chicago Press. 1993 Moral Imagination: Implications of Cognitive Science for Ethics. Chicago: University of Chicago Press. Johnston, Judith R. 1988 Children's Verbal Representation of Spatial Location. In Stiles-Davis et al. (eds.), Spatial Cognition: Brain Bases and Development, 195-205. Hillsdale, NJ: Lawrence Erlbaum. Kaiser, Daniel H. 1980 The Growth of the Law in Medieval Russia. Princeton: Princeton University Press. Katzner, Kenneth 1984 English-Russian, Russian-English Dictionary. New York: John Wiley & Sons. Kegler, Dietrich 1975 Untersuchungen zur Begriffsgeschichte von Istina und Pravda im Russischen. Frankfurt/M.: Peter Lang. Kuznecova, Ariadna I. and Tat'jana F. Efremova (eds.) 1986 Slovar' morfem russkogo jazyka [A Dictionary of Russian Morphemes]. Moscow: Russkij jazyk. Lakoff, George 1987 Women, Fire, and Dangerous Things: What Categories Reveal About the Mind. Chicago: University of Chicago Press. Lakoff, George and Mark Johnson 1980 Metaphors We Live By. Chicago: University of Chicago Press. Laponce, J. A. 1981 Left and Right: The Topography of Political Perceptions. Toronto: University of Toronto Press. Lindner, Sue 1982 What Goes Up Doesn't Necessarily Come Down: The Ins and Outs of Opposites. Papers from the Regional Meeting of the Chicago Linguistic Society 18: 305-323. L'vov, Andrej S. 1966 Ocerki po leksike pamjatnikov staroslavjanskoj pis'mennosti [Essays on the Lexicon of Monuments of Old Slavic Literary Texts]. Moscow: Nauka. L'vov, Mikhail R. 1984 Slovar' antonimov russkogo jazyka [A Dictionary of Russian Antonyms]. Moscow: Russkij Jazyk. Marder, Stephen 1992 A Supplementary Russian-English Dictionary. Columbus, OH: Slavica. Markey, T. L. 1982 Indo-European Etyma for 'Left, Left-Handed' and Markedness Reversal. The Mankind Quarterly 23: 183-194.

The left/right polarity in Russian

329

Matthews, William K. 1960 Russian Historical Grammar. London: University of London/ Athlone. Mondry, Henrietta and John R. Taylor 1992 On lying in Russian. Language and Communication 12: 133-143. Needham, Rodney (ed.) 1973 Right and Left: Essays in Dual Symbolic Classification. Chicago: University of Chicago Press. Pokomy, Julius 1959 Indogermanisches etymologisches Wörterbuch. 2 vols. Bern: Francke. Sanskij, Nikolaj M., Valerij V. Ivanov, and Tamara V. Sanskaja (eds.) 1971 Kratkij etimologiceskij slovar' russkogo jazyka [A Short Russian Etymological Dictionary]. Moscow: ProsveSoenie. SDJa: see Sreznevskij (1989) Slovar' russkogo jazyka [Dictionary of the Russian Language] 1981-1984 4 vols. Akademija nauk SSSR. Moscow: Russkij jazyk. Slovar' russkogo jazyka XVIII veka [Dictionary of the Russian Language of the 18th Century] 1984Akademija nauk SSSR. Leningrad: Nauka. Sreznevskij, Izmail I. 1989 Slovar' drevnerusskogo jazyka [SDJa] [Old Russian Dictionary]. 3 vols. Moscow: Kniga. Sweetser, Eve E. 1990 From Etymology to Pragmatics: Metaphorical and Cultural Aspects of Semantic Structure. Cambridge: Cambridge University Press. Talmy, Leonard 1988 Force Dynamics in Language and Cognition. Cognitive Science 12: 49-100. Traugott, Elizabeth Closs 1974 From Propositional to Textual and Expressive Meanings: Some SemanticPragmatic Aspects of Grammaticalization. In Winfred P. Lehmann and Yakov Malkiel (eds.), Perspectives on Historical Linguistics, 245-271. Amsterdam: Benjamins. van Leeuwen-Tumovcovä, Jifina 1989a Anmerkungen zur Semantik einiger 'Substandard'-Eigenschaften im tschechischen und deutschen Kulturkreis. Zeitschrift für slavische Philologie 49 (1): 132-154. 1989b Semantik und Symbolik von 'links' und 'rechts'. In Norbert Reiter (ed.), Sprechen und Hören: Akten des 23. linguistischen Kolloquiums, Berlin 1988, 573-585. Tübingen: Max Niemeyer. 1990 Rechts und Links in Europa: Ein Beitrag zur Semantik und Symbolik der Geschlechterpolarität. Berlin: Osteuropa-Institut. 1991 Warum ist das Recht gerade? Zeitschrift für slavische Philologie 51: 291 -313. Vasmer, Max 1958 Russisches etymologisches Wörterbuch. 3 vols. Heidelberg: Carl Winter.

'Unnatural barriers': Why metaphor matters (or, linguistics meets the geopolitics of law) David Delaney and Michele Emanatian

Metaphors in law are to be narrowly watched, for starting as devices to liberate thought, they end often by enslaving it. Benjamin Cardozo (quoted in Bosmajian 1992)

1.

Introduction

On March 7,1938, Carl and Carrie Johnson were ordered by the judges of the Missouri Court of Appeals to vacate the home they had purchased in Kansas City. The reason was that the Johnsons were Black and the people who had sold them the house had signed a contract promising never to sell their property to Black people. The neighbors took the Johnsons to court and won. Hence the court order to vacate. The event itself was rather unremarkable in those days, and the decision of the judges, unfortunately, rather commonplace. What interests us 60 years later is how the decision was reached and justified. What we would like to address in this paper is the role that metaphor and metaphorical reasoning played in adjudicating disputes concerning racial segregation in the U.S. in the mid-20th century. We use metaphorical analysis in order to illuminate episodes in legal and social history, in order to understand how people shape the world we live in. Most of the readers of this volume will have already rejected the traditional notion of metaphor as a 'mere' literary device, a conscious rhetorical strategy for getting a point across. Even conventional ("dead") metaphors are now recognized as potentially parts of systematic conceptual mappings from one domain of experience to another. In the last few years scholars from different disciplines seeking to probe systematic ways of reasoning have begun to explore the idea that people think, partly, through such conceptual metaphors. How people do this is the big question. How do we make inferences about the "target" concept based on our knowledge of - or beliefs about - the "source"

332

David Delaney and Michele Emanatian

concept? What role do these largely unconscious inferences play in thinking? And even for those metaphors which are consciously selected, how is it that they guide thoughts, such that they are effective rhetorically?1 The present study applies a metaphor analysis to judicial reasoning about racial segregation and urban change. Unlike many studies of metaphor, our analysis is confined to an actual corpus, texts from a particular historical period. We are particularly interested in how judges understood causation in these matters and how this understanding influenced their determination of the applicability of legal doctrine to specific states of affairs or cases. Legal reasoning, of course, is thought to be above such mutable cultural-rhetorical processes as metaphorization.2 Our story will be told in three steps. First, we will present a sketch of the social, historical and legal background. Next, we will give a preliminary account of the metaphorical mapping which is revealed through analysis of judicial opinions of the period of study. Then, we will present a more detailed analysis of how the metaphors worked, how judges thought through metaphor - or, perhaps, how the metaphors spoke through the judges.

2. Shifting geographies of race The historical-empirical focus of our study is racial residential segregation in U.S. cities in the period 1930 to 1945, and, more specifically, attempts to reinforce or modify patterns of segregation through legal action and litigation; that is, through legal argument. For most of American history Black people were overwhelmingly a rural people and, until the early decades of the 20th century they were overwhelmingly concentrated in the southern states where their ancestors had been held as slaves. The first decades of this century witnessed a rather dramatic demographic shift which was the result of what is referred to as the Great Migration. This migration had two principal components: Black people moving from the South to the North, and moving in both regions from rural to urban areas. As a result, by the 1920s and '30s most northern cities had significant numbers of newly arrived Black residents (Henri 1976; Marks 1989). For many white people the increasing presence of Blacks in the cities of the North was felt to be problematic. Consequently, the new arrivals were not permitted to live wherever they chose in the city but were compelled, by a number of means, to live in a limited number of specific and usually over-

'Unnatural barriers': Why metaphor matters

333

crowded neighborhoods. This state of affairs in turn created serious problems for Black people. The practice of confining Blacks to specific sections of cities resulted in the emergence of racial ghettos as a distinctive feature of the American landscape. (See, for example, Kusmer 1976; Spear 1967.) Legal practices, by which we mean social practices which employed legal means instrumentally to achieve desired social ends, played a prominent role in the formation and maintenance of these patterns of segregation. There were also efforts by Black people and their allies to resist the processes of segregation and ghettoization, and some of these efforts took a specifically legal form. We now look at some of the legal aspects of what we call the geopolitics of race and racism of the period.

3.

Law and the geopolitics of race

One principal legal device that white people utilized in their efforts to shape and revise patterns of segregation was the RESTRICTIVE COVENANT. By way of serviceable definition a restrictive covenant - or more precisely a racially restrictive covenant - was an agreement among neighboring property owners not to sell their property to Black people or other proscribed groups for a specified period, say 21, 50 or 99 years. A restrictive covenant is a contract, a legal document. As such it binds the signatories to its terms. Like any contract it was considered to be enforceable should one of the parties break their promise (see for instance, Vose 1959; Clark and Perlman 1948). Often enough someone would try to break the contract and attempt to sell or rent the restricted property to Black people. And often enough when this happened other parties to the agreement would sue. Here 'to sue' means to ask a judge to enforce the terms of the contract and 'to enforce' means, among other things, requiring the family who bought the property to vacate, perhaps forfeiting the property and the purchase price, perhaps having to pay damages as well. What we have, then, is a court action or a case.3 The judges had a certain amount of discretion in these cases. So, for example, if the defendant offered what the judge considered to be good reasons, then the covenant would not be enforced, the contract would be invalidated and the restrictions removed. There were at this time a number of arguments that would typically be advanced by defendants in such cases. Among these were claims that the

334

David Delaney and Michele Emanation

covenants were unconstitutional or contrary to public policy. In the period under study these arguments never worked. One that occasionally did work relied upon something called THE DOCTRINE OF CHANGED CONDITIONS. It is toward an understanding of this doctrine as a defense and justification that our metaphorical analysis is directed. The immediate point here is that the outcome of the case was determined by judicial assessment of the applicability of the doctrine. The Doctrine of Changed Conditions, like other legal doctrines, is a loose body of legal principles, rules, or rules of thumb that facilitated interpretation of contracts, and thereby, the adjudication of disputes. As a defense against the enforcement of racially restrictive covenants, asserting the applicability of the doctrine was essentially pleading exceptional circumstances. In practice it went something like this: Defendants would admit that they had, indeed, entered into an agreement with the neighbors and that under normal circumstances they would be bound by its terms. They would then assert that conditions in the neighborhood had changed to such an extent since the contract was originally signed that its purpose could no longer be achieved. To enforce a contract under these circumstances would be unfair. Plaintiffs, in turn, would argue that conditions had not changed to the extent alleged by defendants, that the purpose could still be accomplished and that the contract should, therefore, be enforced. Now, in these situations what counted as 'conditions' and what counted as 'change' were determined with reference to the terms and purposes of the contract: 'conditions' here meant racial composition of a neighborhood, or more specifically, blackness. 'Change' meant, in effect, an increase in the number of Black people living in an area. In fact, the precise boundaries of the area and the number of people that would count as constituting 'change' were at the very heart of the conflict. Nevertheless, in assessing the applicability of the Doctrine of Changed Conditions attorneys and judges had to offer some kind of evidence. That is, they had to advance interpretations of the dynamics of urban transformation. They had to give some indication of what 'conditions' had been at some time in the past, of what 'conditions' were like now, and of how they had or hadn't changed (Delaney 1993). As we will show, the participants in these legal disputes understood the processes of demographic change metaphorically. They understood these processes in terms suggestive of a then-prevalent theory of urban sociology

'Unnatural barriers': Why metaphor matters

335

called Urban Ecology which, as the name suggests, conceptualized these social processes in terms of natural or ecological processes.

4.

'Invasions' and 'barriers'

In their assessment of demographic change, and, therefore, their determination of the applicability of the Doctrine of Changed Conditions, judges principally relied upon two conceptual metaphors that were central to an Urban Ecological understanding of urban transformation. We argue, however, that these two metaphors implicate the larger system of which they are key components. The two principal metaphors are: Black property ownership is invasion

and Restrictions on black property ownership are barriers. In the cases we have examined, judges make statements such as the following: During this time the negroes continued to invade the neighborhood. (Pickel et al. v. McCawley et al., 1931: 859) A meeting was called to discuss means to protect the neighborhood against negro invasion. (Pickel et al. v. McCawley et al., 1931: 860) Unless they protected themselves against it the block would be invaded by colored residents. (Porter et al. v. Johnson et al., 1938: 534) The officers [of the property owners association] in their desperate efforts failed to obtain sufficient signatures to block the stampede of negroes coming east of Vanderventer avenue. (Pickel et al. v. McCawley et al., 1931: 860) Other like terms included infiltration, penetration, migration, encroachment and influx. Remember, though, that the actions referred to with these terms are simply those related to buying a house. In these cases, the neighborhood is conceived of as a container of sorts for its white residents. Black people are clearly seen as both alien and menacing, competing and incompatible with whites. The mere presence of Blacks constitutes a degradation of the locality. In this framework, then, how did judges regard the role played by the covenants themselves? The restrictive covenants - and of course the enforcement of covenants - were seen as barriers to the invasion. Judges wrote in their opinions, for example:

336

David Delaney and Michele Emanatian

It [the covenant] furnishes a complete barrier against the eastward movement of colored population into the restricted area - a dividing line. (Grady v. Garland, 1937: 819) Defendant's property [...] although lying across the street from the remainder of the property involved in the contract, served as a buffer for plaintiff's property. It kept Negroes just that much farther away from plaintiff's property. (Clark v. Vaughan, 1930: 787) The officers evidently did their very best but the barrier had been broken and the influx of negroes continued. (Pickel et al. v. McCawley et al., 1931: 861)

Thus, the contracts which were the objects of dispute in these cases could be seen as providing the necessary protection against the invasion. However, as we will show, they could also be seen as unnatural barriers against the natural succession of populations in a city.

5.

The ecology of race

At first glance the metaphors of invasion, infiltration, barriers and dividing lines seem to be drawn from the domain of geopolitics or military strategy. This may ultimately be accurate, but only indirectly, for the concepts appear to have reached legal discourse from natural history by way of Urban Sociology. Here we juxtapose a larger passage from one of our cases with a passage from an urban ecological text of the same period in order to highlight the shared conceptualization of the city and the processes of urban transformation. First, from the case of Hundley v. Gorewitz: [I] n the natural growth of a city, property originally constructed for residential purposes is abandoned for homes of more modern construction in more desirable locations, for a serious decline in values would follow unless the way was open either for use of the property for business purposes or for the housing needs of a lower income class. And it [the Doctrine of Changed Conditions] is also applicable where removals [that is, white out-migration] are caused by the constant penetration into white neighborhoods of colored persons. For in such cases to enforce the restriction would be to create an unnatural barrier to civic development and thereby to establish a virtually uninhabitable section of the city. (1942: 24, emphasis ours)

Here, of course, the judge is explaining causation and urban change in terms of racial antipathy: negro penetration causes white out-migration. This

'Unnatural barriers': Why metaphor matters

337

is a natural process; consequently, interfering with it, at least in this case, is unnatural. Compare this excerpt from the 1938 piece Population Succession in Chicago: 1898-1930 by urban ecologist Paul Cressey: [The] common process of succession involves a cycle of invasion, conflict, recession, and reorganization. These successive stages are interrelated, and they recur in the movement of all groups in the city... Mass invasions set in after the initial invaders have established themselves in the new area. Such movement may involve merely a gradual transition which slowly replaces the older population, or it may take place with such rapidity as to be thought of in terms of a stampede. (1938: 62)

Cressey then goes on to state that "the direction of such movement is influenced by ecological barriers" (1938: 62). Urban Ecology provided a way of understanding how cities work. During the period covered by our study this theory had achieved what an historian of the discipline has called, "near monopoly status in American Sociology" (Szacki 1979). Urban Ecology (also referred to as the Chicago School of Urban Sociology) was most closely associated with the work of Robert Park and his colleagues and students at the University of Chicago (Raushenbush 1979; Lai 1990; see Park 1916, 1925, 1936, 1939a, 1939b). As developed there Urban Ecology is a rather elaborate metaphorization of urban form and process whereby the city and urbanization are understood in terms of Plant Ecology and natural history. The city is conceptualized as a kind of "superorganism", "the habitat of civilized man". It was, we argue, a version of Urban Ecology that judges relied upon in assessing the dynamics of urban change in order to determine the applicability of the Doctrine of Changed Conditions, and therefore, to decide whether or not to enforce a restrictive covenant.4

6.

The ecology of plants

We now turn to the domain of Plant Ecology employed by Urban Sociologists in their efforts to understand the city. Plant Ecology was, and is, the scientific study of the relationship of plants to their environment and to each other. It tries to understand spatial patterns or distributions of plants within a community, and the processes and mechanisms of change over time (Weaver and Clements 1938; Daubenmire 1968).

338

David Delaney and Michele Emanation

Urban ecologists relied upon several notions borrowed from Plant Ecology: First, with respect to form and pattern, the terms COMMUNITY or HABITAT name assemblages of organisms and their functional relationships with each other and with the physical characteristics of their site; NICHES or NATURAL AREAS are more specialized environments within a larger community. The notion of NICHE concerns the functional match of species traits and site conditions. Second, with respect to process, the key term is SUCCESSION. SUCCESSION in Plant Ecology refers to the process whereby different species in sequence occupy a given site over time as the result of changed site conditions. The process of succession tends toward an equilibrium stage called CLIMAX. The climax stage as well as antecedent stages have characteristic species that are referred to as 'dominants' and 'subordinants'. It is with respect to the more precise mechanisms of change or succession that the utility of ecological metaphors for adjudicating restrictive covenant cases is most obvious. We will mention four of the most important concepts that carried over into legal reasoning: COMPETITION - in Plant Ecology this refers to competition among species for water, light, space or other favorable site conditions. MIGRATION - for plant ecologists this refers primarily to methods of seed dispersal. INVASION - this occurs when one species begins to successfully displace another from a particular site. This, of course, is the first phase of succession. The precise dynamics of invasion are strongly conditioned by the presence or absence of what plant ecologists call BARRIERS, of which there are a variety of forms. Plant Ecology is an extremely rich conceptual framework. Some notions (such as soil, plant formations, and disturbance climax) had no obvious analogue to city life and were not used by urban ecologists. However, all of the concepts highlighted above were used in fashioning an understanding of the generic city and of particular cities and events. In the world of the urban ecologists, then: THE CITY is a NATURAL COMMUNITY or HABITAT. THE NEIGHBORHOOD is a NICHE or NATURAL AREA. CATEGORIES OF PEOPLE (races, ethnic groups, classes, occupations) are NATURAL GROUPS or SPECIES.

'Unnatural barriers': Why metaphor matters

339

COMPETITION, according to Park, brought about, first, an orderly distribution of the population, and, second, a differentiation of species within a habitat. As we've seen, demographic changes within a city were accounted for by the natural processes of succession, migration, and invasion, as these are conditioned by the presence or absence of ecological barriers. The endpoint of these processes is a climax stage characterized by equilibrium, stability and order. This was the image of how cities worked that was most prominent in the period of our study and which we believe informed judicial thinking in restrictive covenant cases, whether as a matter of conscious deliberation or not. This way of understanding cities did not compel one decision over others. Rather, judges used these notions to interpret 'facts' as presented by attorneys. Covenants, and thus, enforcement, could be seen as 'unnatural' or 'natural' according to what judges viewed as the natural processes of competition and succession.

7. Entailment of urban ecology metaphors We return now to the metaphors used by judges in considering the applicability of the Doctrine of Changed Conditions to restrictive covenant cases. While there seems to be an implicit, nearly wholesale borrowing of the ecological understanding of the city from Urban Ecology, only two principal metaphors explicitly enter into the judicial record. These are, again, Black property ownership is invasion. Restrictions on black property ownership are barriers. In the Invasion metaphor, we have Blacks buying property in covenanted areas being understood in terms of the ecological concept of invasion. Black people who buy property in a 'white neighborhood' are the INVADERS, the species who are not part of the community. The white neighborhood is the TERRITORY INVADED. Purchase of property and subsequent habitation by blacks is the PROCESS OF INVASION itself. Invasion of a plant community opens up the possibility of COMPETITION between the invaders and the Original' inhabitants (if water, nutrients, light, space, etc. are insufficient for all). Blacks and whites are seen as competing for space, or rather, for decent affordable housing. Successful invasion requires the efforts of PIONEERS, those plants who migrate into a community first; if they succeed in establishing themselves, MASS INVASION will follow. The Black families who purchased coven-

340

David Delaney and Michele Emanation

anted property are the pioneers, and it is the fear of subsequent mass invasion which leads to the courtroom. Invaders MODIFY THE ENVIRONMENT in new ways. The changed conditions they introduce may have an impact on the growth or even the continued existence of the plants Originally' in the community. This, of course, is the very essence of the cases to be decided. The claim was that Blacks, by owning property in a 'white neighborhood', and through occupying that property, were altering the property values in negative ways. The second metaphor, of Barriers, provides a way to understand the utility of the covenants. Restrictions on Black property ownership, that is, the covenants themselves, are barriers to invasion. A barrier obstructs or prevents the passage of something from one area to another. If that 'thing' is dangerous or undesirable, then a barrier serves to PROTECT an area and its inhabitants from the danger or unpleasantness. In Plant Ecology, a barrier can, for example, protect a climax community from incursion by foreign species. In our case, legal contracts are seen as protecting white residents from the encroachment of Blacks into the neighborhood, and/or protecting property in the neighborhood from Black ownership, and all that that suggested. In the plant world, barriers may be more or less complete (or extensive); more or less successful (that is, barriers may be circumvented); and more or less permanent. The extent to which restrictive covenants had - or should have - these traits was, in fact, contested in the court cases. The notion of barriers presupposes a directionality of movement of the migrating species. The success of a covenant was assessed in terms of whether it stood against the DIRECTION OF ADVANCE of the migrating Negroes. In Plant Ecology, barriers can be physical (topographic, such as a mountain range or river) or biological (such as the presence of alkaline soil or grazing animals). DISTANCE itself can be a barrier to the migration of new species into a community. In the judicial decisions considered here, if the distance between the covenanted tracts and the areas of Black occupancy was large enough, then the covenant-barrier was considered to be a BUFFER zone. In plant communities, the ABSENCE OF A NECESSARY FEATURE (e.g. pollinating insects) can be an ecological barrier to invasion. This, of course, was the whole idea of restrictive covenants: that there be no property available for purchase by Blacks. The types of barriers to plant migration just mentioned are NATURAL BARRIERS. But it is also possible to have UNNATURAL BARRIERS: hu-

'Unnatural barriers': Why metaphor matters

341

mans may selectively destroy or renew parts of plant communities, for example, in agricultual practice. That is, they may interfere with natural processes of succession. For some of the judges, this was the very issue on which the decision turned: were the covenants natural or unnatural barriers to invasion? The city, as the "natural habitat of civilized man", "just grew up" - subject to natural processes of change. Was Black migration a natural process, unnaturally restricted by covenants? Or were the covenants a natural barrier to an encroaching foreign species? Notice that there are several concepts of the source domains - of Plant and Urban Ecology - which do not carry over to the restrictive covenant cases. It is most striking that the courts did not explicitly or implicitly use the concepts of SPECIES, CLIMAX, or DOMINANCE in their discussion of the issues. Presumably such notions were too vulgar, and too obviously ran counter to ideas of equality found elsewhere in American legal discourse.

8.

Conclusion: Why metaphor matters

We conclude with a few brief remarks. We would first like to draw attention to what we call "the filter effect" (cf. Black's "screen", 1981). The initial source domain of our analysis, Plant Ecology, is, as we have discussed, a rather rich and elaborate set of concepts that natural scientists use to understand relations among species and their environment. Much of the richness and subtlety of this domain is lost in the transfer to Urban Ecology. Yet, Urban Ecology is itself rather full-bodied compared to the use to which it was put by judges. What we see here is a serial impoverishment or successive winnowing of not only the number of concepts, but also the conceptual richness of the concepts that are carried over. We note again that Plant Ecology itself drew on concepts that were originally associated with social phenomena, such as military activity and economics. So in a sense what we've described is the process of metaphorization coming full circle. In the words of geographer David Harvey (as quoted in Martin 1991), we have the "implanting of social imagery on representations of nature so as to lay a firm basis for reimporting exactly that same imagery as natural explanations of social phenomena". Thus the use of plant ecological metaphors in legal reasoning induces a "naturalization effect". The metaphors highlight the conceptualization of Black people as not simply 'invaders', but as environmental conditions for whites. 'Negro penetration' is regarded as a mass process, which makes it

342

David Delaney and Michele Emanatian

harder to see specific real estate transactions as arrangements between the actual humans involved in the deal. Moreover, segregation itself is regarded as a natural process, and as the result of natural processes. In the wider social conflicts of which these cases were simply one manifestation, Black people were fighting against segregation and ghettoization. An Urban Ecological understanding of urban change foreclosed the possibility of challenging segregation per se, while occasionally allowing for judicial review of particular events. To naturalize here means to de-politicize, in the sense of rendering segregation immune to political intervention.5 Both sides in these disputes, however, participated in this naturalization process because of the limitations imposed by the Doctrine of Changed Conditions. After 1945 judges began to be persuaded by arguments other than those required by the Doctrine of Changed Conditions. Specifically, they decided that enforcement of racially restrictive covenants violated the U.S. Constitution (Vose 1959). Thus the questions of whether and how 'conditions' had or had not 'changed' became moot, and the ecological metaphors used to answer those questions became irrelevant. Finally, we wish to stress that the decisions reached in these cases were consequential. They were clearly consequential to the parties involved. Thousands of people like Carl and Carrie Johnson were put out of their homes or confined to substandard housing. More generally, the body of law developed through these cases created, we might say, a formidable barrier to the use of legal action in the fight against racial inequality and injustice. It is in this regard above all that metaphor and metaphorical reasoning matters.

Acknowledgements We would like to thank the participants in the Third ICLC at Leuven for helpful and encouraging comments. We would also like to express our appreciation to Drew Ross and Norm Rosenberg for their insightful observations.

Notes 1. Prominent attempts to deal with some of these issues include: Beneke's (1982), Lakoff's (1987) and Johnson's (1987) rhetorical analyses of a man's justifications for rape, in which the metaphor Sex Is A Physical Force, along with many other cultural assumptions, provides a rationale for an 'appropriate' response to an attractive woman; Johnson's treatment (1987) of Selye's discovery of physio-

'Unnatural barriers': Why metaphor matters

2. 3.

4.

5.

343

logical stress through a re-metaphorization of the body, replacing the prevailing Body As Machine understanding with the novel Body As Homeostatic Organism; and Emily Martin's demonstration (1987) that American medicine's treatment of the birthing process as factory production influences actual practice in the delivery room. Seminal studies of the relevance of a cognitivist understanding of metaphor to the practice of legal reasoning can be found in Winter 1989, 1992. For the benefit of those not versed in the peculiarities of the Anglo-American system of adversarial justice, we should mention that a CASE here is a formal proceeding in which one party, the plaintiff, presents arguments about why the contract should be enforced and the other party, the defendant, presents arguments about why the court should refrain from enforcing the contract. As an empirical or historical issue it is not essential to our interpretation that state or lower federal court judges had direct knowledge of current urban sociological theories - though, given Park's prominence and his efforts to popularize academic sociology, such knowledge is not out of the question. Rather, Urban Ecology itself may have been simply a rather elaborate way of articulating conventional understandings - understandings that elite members of white society, like judges, shared and gave voice to when required to justify their actions. Readers may recognize this effect in structuring contemporary understandings of, for example, poverty, patriarchy, and racism.

Cases cited Clark v. Vaughan (292 P.783, 1930) Grady v. Garland (89 F.2d 817, 1937) Hundley v. Gorewitz (132 F.2d 23, 1942) Pickel et al. v. McCawley et al. (44 S.W.2d 857, 1931) Porter et al. v. Johnson et al. (115 S.W.2d 529, 1938)

References Beneke, Timothy 1982 Men on Rape. New York: St. Martin's Press. Black, Max 1981 Metaphor. In: Mark Johnson (ed.), Philosophical Perspectives on Metaphor. Minneapolis: University of Minnesota Press. 63-82. Bosmajian, Haig 1992 Metaphor and Reason in Judicial Opinions. Carbondale & Edwardsville: Southern Illinois University Press. Clark, Tom and P. Perlman 1948 Prejudice and Property: an Historic Brief Against Racial Covenants. Washington: Public Affairs Press.

344

David Delaney and Michele Emanation

Cressey, Paul 1938

Population Succession in Chicago: 1898-1930. American Journal of Sociology 44: 59-69. Daubenmire, Rexford 1968 Plant Communities. New York: Harper and Row. Delaney, David 1993 Geographies of judgment: legal reasoning and the geopolitics of race. Annals of the Association of American Geographers 83: 48-65. Henri, Florette 1976 Black Migration: Movement North, 1900-1920. Garden City, New York: Anchor Press. Johnson, Mark 1987 The Body in the Mind. Chicago: University of Chicago Press. Kusmer, Kenneth 1976 A Ghetto Takes Shape: Black Cleveland, 1870-1930. Urbana: University of Illinois Press. Lakoff, George 1987 Women, Fire and Dangerous Things. What Categories Reveal About the Mind. Chicago: University of Chicago Press. Lakoff, George and Mark Johnson 1980 Metaphors We Live By. Chicago: University of Chicago Press. Lai, Barbara 1990 The Romance of Culture in an Urban Civilization: Robert E. Park on Race and Ethnic Relations in Cities. London: Routledge. Marks, Carole 1989 Farewell - We're Good and Gone: the Great Black Migration. Bloomington: University of Indiana Press. Martin, Emily 1987 The Woman in the Body. A Cultural Analysis of Reproduction. Boston: Beacon Press. 1991 The egg and the sperm: how science has constructed a romance based on stereotypical male-female roles. Signs 16.3: 485-501. Park, Robert E. 1916 The city: suggestions for the investigation of human behavior in the urban environment. American Journal of Sociology 20: 577-612. 1925 The urban community as a spatial pattern and moral order. Publications of the American Sociological Society 20: 1-14. 1936 Succession: an ecological concept. American Sociological Review I: 171179. 1939a (1952)The city as a natural phenomenon. In: Hughes, Everett C., et al. (eds.), Human Communities. The City and Human Ecology. Glencoe, IL: The Free Press. 118-127. 1939b Symbiosis and socialization: a frame of reference for the study of society. American Journal of Sociology 45: 1-25.

'Unnatural barriers': Why metaphor matters

345

Raushenbush, M. 1979 Robert E. Park: A Biography of a Sociologist. Durham: Duke University Press. Spear, Allan 1967 Black Chicago: The Making of a Negro Ghetto. Chicago: University of Chicago Press. Szacki, J. 1979 A History of Sociological Thought. Westport, CT: Greenwood Press. Vose, C. 1959 Caucasians Only: The Supreme Court, the NAACP and the Restrictive Covenant Cases. Berkeley: University of California Press. Weaver, John E. and Frederic E. Clements 1938 Plant Ecology. 2nd ed. New York: McGraw-Hill. Winter, Steven 1989 Transcendental nonsense, metaphoric reasoning and the cognitive stakes for law. University of Pennsylvania Law Review. 1105-1238. 1992 Death is the mother of metaphor. Harvard Law Review 105: 745-772.

Through as a means to metaphor Joseph Hilferty 1.

Introduction

Many - if not most - linguistic frameworks treat prepositions as semantically vacuous grammatical morphemes whose primary function is to signal grammatical relations. Such approaches to grammar generally view prepositional usage as inherently arbitrary. Proponents of Cognitive Linguistics, however, have called both of these assumptions into question. In fact, there is now a substantial body of research indicating that, far from being semantically empty, prepositions (and related phenomena) are actually highly polysemous.1 These studies suggest that prepositional polysemy manifests itself as radial categories, that is, networks of interrelated, yet different, senses. Very briefly, one sense, the prototype, possesses special prominence within the network, in that it is the most representative conceptual instance of the category. Other, less representative senses are related to the prototype (or to intermediate sense-nodes) by virtue of the family resemblance they bear to one another. More specifically, some senses in the network are related by virtue of literal similarity; others are associated via metonymy; and still others are connected on the basis of metaphor (for a complete account, see Lakoff 1987). In the present paper, I will concentrate on this latter type of linkage and show that through^ metaphorical uses signifying MEANS draw on its prototypical spatial sense.

2. The prototypical spatial image of through In order to lay the preliminary groundwork for the analysis, it might be useful to say a few words about the various theoretical constructs of the framework I will be adopting here. In Cognitive Grammar (Langacker 1987, 1990, 1991, 1992a), meaning is equated with conceptualization (as it is in other cognitively oriented frameworks). Specifically, it is claimed that the semantic structure of a linguistic expression - termed a predication - gains its meaningfulness through the imposition of a profile on a base. The necessary conceptual context evoked by a linguistic expression constitutes the base of a predication; the profile, on the other hand, is that portion of the base which the

348

Joseph Hilferty

linguistic expression conceptually designates. Take, for instance, the predication VERTICALITY. Such a predication takes as its base the domain of oriented space and evokes as its profile the UP-DOWN axis. This is represented in Figure 1, where the heavy line indicates the expression's profile. »p

right *^

.-» back

front

down Figure 1. Predication VERTICALITY

Before getting on to the prototypical meaning of through, it is important to stipulate some sort of base (in the sense outlined above). Both Lakoff (1987) and Johnson (1987) have talked at some length about a SOURCE-PATHGOAL schema grounded in bodily experience that underlies our conception of moving from one place to another. The logic of the schema is that all spatial trajectories consist of the following components: (A) a starting point (i.e., an origin or source); (B) an end point (i.e., a destination or goal); (C) a series of contiguous points linking the source with the goal (i.e., a path); and (D) directed movement by a mover, away from the source and towards the goal. In other words, in order to get from (A) to (B), one must traverse (C). Moreover, movement through space necessarily implies movement through time: the further one advances along (C), the more time has passed (see also DeLancey 1982). Following Johnson (1987: 28, 114), we can represent this schema as in Figure 2, where the vector connects the two endpoints, leading from the source to the goal.

source

path

goal

Figure 2. Source-Path-GOAL schema

We are now in a position to describe the prototypical sense of through. Being a path-preposition, through profiles the PATH portion of a trajectory (as

Through as a means to metaphor

349

opposed to goal-prepositions such as to, which profile the GOAL portion of a trajectory, or source-prepositions such as/rom, which profile the SOURCE portion of a trajectory).2 Though it might be stretching the term a bit, it is useful to think of path-prepositions as possessing imperfective Aktionsart, inasmuch as they do not mark a definite starting or end point (cf. Hawkins 1984, 1987). This is illustrated, for instance, by sentence (1), where it is impossible to ascertain the origin or the destination of the bus's trajectory.

(1)

The bus went through the tunnel.

In a somewhat artificial (yet helpful) analysis of this example, through's trajector3 (i.e., the bus) can be said to trace a fairly linear path that takes it into, along, and out of the landmark (i.e., the tunnel). The artificiality of this analysis becomes manifest upon assessing examples such as Your brother just walked through the door. In such a sentence, the landmark (the door, which actually stands metonymically for the doorway) is conceived as two dimensional and therefore lacks depth. Hence, the trajector (i.e., my brother) can certainly be said to cross the doorway, but in no way can it be said to go along it (Bruce Hawkins, pers. com.). To give a more accurate portrayal of the prototypical spatial through, I shall characterize its semantic image, or imageschema, as one in which the trajector simply traverses the landmark. This is depicted in Figure 3, where, with regard to the landmark, the trajector's linear path takes it sequentially from a relationship of noncoincidence, across one of coincidence, to that of noncoincidence (cf. Hawkins 1984).

Int Figure 3. Trajector traverses landmark

Space 1

fan

Space 2

Figure 4. Medium between landmark's entrance and exit

350

Joseph Hilferty

Note that a prototypical through-landmark is a bounded MEDIUM (as defined by Hawkins 1988) that separates two adjoining spaces: one which is located at the threshold of the landmark's entrance and another which is situated at the threshold of its exit (see Figure 4). Another way of contemplating this configuration is to construe through's landmark as constituting a means of getting from one location to another. Notice also that the trajector first comes from Space 1, then encounters the landmark, and finally reaches Space 2. With regard to movement along this trajectory, Space 1, the landmark itself, and Space 2 correspond to Time], Time2, and Time3, respectively. This falls out from one of the underlying assumptions of the SOURCE-PATH-GOAL schema: the further along a trajectory one is, the more time has passed. The importance of these observations will become more apparent in Section 4.4

3.

Through as a metaphorical means

It is hardly a new insight to suggest that the abstract senses of prepositions are often metaphorical (see, e.g., Hanssen 1911). Until fairly recently, however, most analyses have remained intuitive, rather than truly explicative. Brugman's (1981) and Lindner's (1981, 1982) studies on prepositions (or verb particles, as the case may be) have shown that the metaphorical senses of such grammatical morphemes rely heavily on image-schemata from the domain of space (see also Ben ware 1993). In what follows, I shall show that through's metaphorical senses denoting MEANS are motivated by its prototypical spatial image of TRAVERSAL. One of the observations made in Section 2 is that the through-landmark represents a path that connects one space with another. I contend that the examples in (2) are analogous to this situation. (2)

a. b.

John and Mary are related through a common ancestor, Anne and I are united through our love for music.

Ontologically, in (2a), John and Mary represent two separate entities that are "connected" by means of an intervening third entity (i.e., a common ancestor). Much the same can be said for (2b). It is conceivable that Anne and I would not have any relationship at all if it were not for the love for music that we have in common. That is, our love for music acts as a means to connect us. Examples such as these lead me to posit the following metaphor: CONNECTIONS BETWEEN ENTITIES ARE PATHS. Following the convention

Through as a means to metaphor

351

established in Lakoff (1987) and Lakoff and Kovecses (1987), the mapping of this conceptual metaphor may be stated as follows: Source domain Target domain: Ontological correspondences

Epistemic correspondences

PATHS CONNECTIONS BETWEEN ENTITIES Space 1 corresponds to an entity. Space 2 corresponds to another entity. The path (i.e., the landmark) corresponds to the connection. Source: Paths unite distinct spaces with one another. Target: Connections unite distinct entities with one another.

The first type of mapping, called ontological correspondences, describes relationships that exist between structures in the source and target domains. The second type of mapping relates to epistemic correspondences; these correspondences represent knowledge links that go across domains. Continuing on, sentence (3) exhibits a somewhat different variation of this image: (3)

John can get his answer to me through either Sean or Eileen.

In this case, John cannot give his answer directly to me; instead, he must use either Sean or Eileen as a go-between. Because Sean and Eileen are understood as intermediaries, they constitute the means by which John can send his answer to me. Though it would hardly be controversial to term me as the goal in this example, I suggest that John can be regarded as the source, inasmuch as he is the starting point for his answer's trajectory. Hence, it seems natural in this case for the intermediaries to be marked by a path-preposition. This metaphorical use of through, termed here INTERMEDIARIES ARE PATHS, can be given the following mapping: Source domain: Target domain: Ontological correspondences:

PATHS INTERMEDIARIES The path corresponds to the intermediary. The trajector corresponds to the transferred entity.

352

Joseph Hilferty

Epistemic correspondences:

Source: Paths link a spatial origin and destination. Target: Intermediaries link a sender and a receiver.

Variations on the above conception can be found in examples (4) below. Though quite similar, the general meaning of these examples is slightly different in each case: whereas the meaning of sentence (3) is basically one of transfer, sentence (4a) is about how a goal can be attained and (4b) expresses how an event came about. (4)

a. b.

Unfortunately, to order the dissertation Al would have to go through Professor Craddock. We met through a matrimonial agency - we met through my mother.

In (4a), Professor Craddock stands between Al and the dissertation. In this utterance, it appears that, if Al really wanted the dissertation, Professor Craddock would have to mediate for him in some way. That is, Professor Craddock constitutes the means by which Al can achieve a certain purpose (viz., placing the order). This situation is diagrammed in Figure 5. Example (4b) presents a more complex situation. Here, the speaker cheekily states that a matrimonial agency - which turns out to be his mother - mediated his meeting his spouse. In other words, the speaker's mother acts as a means to bring the two parties together (see Figure 6). These two particular uses of through, which I shall call MEDIATORS ARE PATHS, can be mapped in the following way: Source domain: Target domain: Ontological correspondences:

Epistemic correspondences:

PATHS MEDIATORS The path corresponds to the mediator. The trajector corresponds to the mediated event or action. Source: A path lies between a spatial origin and destination. Target: A mediator intervenes between two separate entities.

Through as a means to metaphor

353

Figure 5. MEANS for Al to get a dissertation

Figure 6. Mother as MEANS to bring both together

It should be noted that, although mediators tend to be human, what is mediated is some sort of action or event. By this I mean that it would be inexact to say the husband and wife of example (4b) are the mediated entities per se: what is mediated is their meeting each other. The same can be said in (4a): Al is not what is mediated, but rather the paper work needed to order the thesis. Mediators therefore act as intervening agents, as it were, that help to bring about a certain state of affairs.

4.

Causal Means and Procedural Means

Let us now turn our attention to two further metaphorical senses, both of which are closely related to each other, as well as to those discussed in the previous section. The first of these senses that I will discuss at length is the use of through in causative sentences.5 Generally, these are sentences with active-voice intransitive verbs that denote a change of state, as in (5). (5)

a. b.

The patient died through the doctor's negligence, He got proficient through plenty of practice. Event Space

State 1

S".

\

State!

Doctor's Negligence

Γ

/

s

Time,

>

Time,

Figure 7. Through in causative sentences

Tim 63

354

Joseph Hilferty

Crucial to understanding this use of through are the conceptual metaphors STATES ARE LOCATIONS (e.g., I'm in a state of ecstasy) and CHANGE OF STATE IS CHANGE OF LOCATION (e.g., Things went from bad to worse). I will not go into the details of these conceptual metaphors here, which have been discussed in Lakoff and Turner (1989); without further ado, I will use them to help specify the internal structure of examples such as those in (5). First, consider sentence (5a): DIE profiles a process whereby a patient undergoes a change of state, from the state of LIVING to the state of LIFELESSNESS. The cause of the patient's death is the doctor's negligence, which intervenes between the above-mentioned states. Figure 7 schematizes this scenario. In such cases, we conceive the entire event chronologically. First, at Timej there is the assumption that the patient was originally alive. Then, there is the Event Space at Time2, when the doctor's negligence takes place. This in turn causes the patient's death at a later Time3. This conceived sequence is also true of (5b). First, we assume that the person in question was not proficient. This state is followed by a time of plenty of practice, after which the person becomes proficient. Following Jane Espenson (pers. com.), I will call this use of through CAUSAL MEANS ARE PATHS: Source domain: Target domain: Ontological correspondences:

Epistemic correspondences:

PATHS CAUSAL MEANS Space 1 corresponds to State 1. Space 2 corresponds to State 2. The path corresponds to the causal situation. Movement from Space 1 to Space 2 corresponds to changing from State 1 to State 2. Source: The landmark is located in space between Space 1 and Space 2. Target: The causal situation is located in time between State 1 and State 2.

The next use of through to be examined is that in which the through-phrase designates the means by which an action can be realized (or a state can be perpetuated, as we shall see below). Consider the following examples: (6)

a. b.

Through sports, children learn how to handle defeat as well as victory. His uncompromising style of journalism has caused many to try and suppress him through civil and criminal proceedings.

Through as a means to metaphor c.

355

Money raised through the phone surcharge could be used to buy new equipment and pay worker salaries in the communications center.

Intuitively, these examples are very closely related to those in (5). They are slightly different, however, in that all have transitive verb(al)s or, at least, have transitive counterparts (for example, raised, in (6c), has a transitive counterpart in raise). I will call the conceptual metaphor that subsumes the above examples: PROCEDURAL MEANS ARE PATHS. In what follows, I will show that this metaphor corresponds with the metaphorical mapping means are paths to destinations mentioned in Lakoff (1990: 57; see also Turner 1991: 204). Let us start by comparing the following examples, taken from a pair of bumper stickers in California: (7)

a. b.

Rehabilitation through reincarnation. I support peace through superior firepower and total retaliation.

Despite its nominal head, (7a) does not present much of a problem, since rehabilitation is derived from a transitive verb. Furthermore, it clearly implies a change of state. On a fairly literal level, the heavy-handed implications of this "modest proposal" go something like the following: First, there are criminals who are alive and committing serious offenses; they go to jail for rehabilitation but to no avail: they relapse into a life of delinquency. They are, in other words, metaphorically located in a state of nonrehabilitation. Second, to rehabilitate these offenders once and for all, it is advocated that they should be put to death, in the hope that they can be reincarnated. Their reincarnation constitutes the Event Space. Third, after (or simultaneously with) the criminals' reincarnation, they are reborn as rehabilitated citizens who should be able to function within the norms of society. Thus, they are metaphorically located in a state of rehabilitation. Sentence (7b), on the other hand, proves to be more problematic (at least at first blush). For one thing, this sentence is syntactically ambiguous, since the prepositional phrase can be construed as being adjectival or adverbial. This ambiguity is represented in the following parses:6 (8)

a.

I support [NP peace [pp through superior firepower and total retaliation]].

356

Joseph Hilferty b.

I [VP support peace [pp through superior firepower and total retaliation]].

The main problem, however, is that neither SUPPORT, nor PEACE - depending on one's parse - predicate a change of state (after all, both are atelic notions). This prompts the following question: if, according to the present account, the //irowg/z-landmark (i.e., ihrough's object) is conceived as separating two entities, what two entities are they? To find out, we will have to scratch beneath the surface a bit. I would speculate that the throughlandmark separates an entity and its privative opposite (in the sense described in Lyons 1977). On the adjectival interpretation represented in (8a), the through-landmark stands between nonpeace (i.e., war, unrest, etc.) and peace. That is, through helps to denote a way (or means) to perpetuate the state of peace, while foreclosing the state of nonpeace. Thus, if this characterization is correct, Space 1 corresponds ontologically to the privative opposite (i.e., nonpeace), the path to the means (i.e., superior firepower, etc.), and Space 2 to the perpetuated state (i.e., peace). The adverbial interpretation can be described in a similar fashion. In this case, the landmark (i.e., superior firepower, etc.) is located between the nonsupporting of peace and the supporting of peace. Hence, the through-phrase conveys the way in (or the means by) which the bumper-sticker writer sustains the process of supporting peace, while foreclosing its privative opposite. Ontologically, then, Space 1 corresponds to the nonsupporting of peace, the path to the means (i.e., superior firepower, etc.), and Space 2 to the act of supporting peace. Source domain: Target domain: Ontological correspondences:

Epistemic correspondences:

PATHS PROCEDURAL MEANS Space 2 corresponds to an action or a state. Space 1 corresponds to the privative opposite of the above-stated action or state. The path corresponds to the procedural means. Traversing the path corresponds to realizing an action or perpetuating a state. Source: A path is a way of getting to Space 2. Target: A means is a way of procuring a state of affairs.

Through as a means to metaphor

357

The analysis presented above bears some clarification on two points. First of all, it should be noted that I am not asserting that with the PROCEDURAL MEANS metaphor the conception is always one of going temporally to a state of affairs from its privative opposite. This does seem to be the case when the prepositional subject (in the sense of Leys 1989) - be it verbal or nominal - expresses a change of state, as in (la). It certainly cannot be asserted, however, for examples such as (9a) below, where no change of state is predicated. The second point to be clarified is that, in stating that the f/twwg/i-landmark separates two entities, I am not claiming that both entities are profiled an equal footing. On the contrary, I suggest that, whereas one entity receives explicit mention, the privative opposite lies in the realm of the implicit. Consider, for example, the sentences in (9). (9)

a. b.

Peace is maintained through superior firepower and total retaliation. In traditional, clan-oriented societies, individuals were kept in line through shame.

In sentence (9a), the threat posed by superior firepower, etc. clearly mediates the maintenance of peace. Furthermore, the sentence does seem to suggest that, if it were not for superior firepower, etc., peace would not be maintained. The same can be said of (9b), where it is understood that shame was a means for keeping people in line. It naively follows that in the absence of shame, people would not be kept in line (unless, of course, it were replaced by some other means). Such a situation can be seen most clearly when the main clause denotes a goal-oriented activity, such as in (10). (10)

IMPORTANT: Only through correct use will you get 100% out of your appliance.

In (10), correct use does not separate the obtainment of 100% and the obtainment of 0% (its polar opposite). Rather, it separates the obtainment of 100% and the obtainment of anything less than 100% (which is tantamount to not obtaining 100%). These intuitions lead me to believe that through1^, landmark stands metaphorically between a profiled entity and its privative opposite, which is connoted within the expression's base.

358

Joseph Hilferty

5. Some constraints With regard to the PROCEDURAL MEANS use of through, the action or state that the prepositional phrase modifies is intuitively understood as a goal. What I am essentially suggesting is: just as paths are conceived as ways of getting to destinations, means are conceived as ways of achieving ends. Some evidence for this analogy is presented in the following attested example, where there exists a certain parallelism between route to and through. (11)

On the campuses of elite universities, students quickly learn the grammar and protocols of power - that the route to moral superiority and premier griping rights can be gained most efficiently through being a victim - which explains academia's search for what one critic calls the "unified field theory of oppression."7

The above-mentioned analogy is interesting not only for what it tells us about conceptual structure (it adds more proof for the primacy of the domain of space), but it is also interesting for the constraints it should entail. The main prediction is that, if the action or state predicated by the finite clause cannot be thought of as an end, then the resulting utterance should be unacceptable. This constraint accounts for the contrast in grammaticality judgments of the following examples. (12)

a.

The troops were able to surround the castle through superior tactics, b. *A moat surrounds the castle through superior engineering.

Sentence (12a) is perfectly acceptable because the surrounding of the castle is clearly construed as a perfective event, involving two basic configurations: one in which the troops have not surrounded the castle and another in which they have. The troops' superior tactics merely acts as a "conduit", so to speak, so that the final configurational state can be attained from the previous one. Example (12b), on the other hand, is quite different. Though the water within a moat may flow around somewhat, moats themselves are static entities, in that they remain fixed in one place. Thus, only one configuration is predicated in (12b), which is the basis for its stative quality. More important, I believe, is the fact that moats are inanimate objects. This makes it very hard to think of them as entities that seek to accomplish ends. It therefore makes sense that, just like (12b), example (13) below also turns out to be unacceptable, despite

Through as a means to metaphor

359

the fact that it is possible to interpret that two states are involved (i.e., one where the moat existed and another where it no longer does). (13)

*A moat once surrounded the castle through superior engineering.

Finally, I would like to examine what appears to be a necessary constraint for the CAUSAL MEANS ARE PATHS metaphor. It will be seen that, in and of themselves, image-schematic constraints (e.g., the Invariance Hypothesis8) may not be enough to explain the grammatical behavior of the following example: (14)

*The fish died through pollution.

The people I have consulted all judge example (14) to be an unacceptable utterance. But why should this be so? The answer, I believe, is tied to the fact that sentences of this type are most felicitous when the through-phrase suggests the occurrence of some sort of action or event. (Recall that this is stipulated in our ontological correspondences: "The path corresponds to the causal situation.") With this in mind, I would contend that sentences (15a) and (15b) are fully acceptable because both refrigeration and research convey action-like concepts. (15)

a. b.

The milk went sour through a lack of proper refrigeration. Computer technology has evolved through constant research.

The above observation appears to be borne out in full by the grammatical judgments that accompany the following examples: (16)

Laura got rich through a. ?*stamps. b. collecting stamps. c. philately. d. ?*gold. e. gold mining. f. dealing in gold.

Of the above variations, those sentences that are judged to be fully grammatical are precisely those containing through-phrases that express an action or an activity. Examples (16a) and (16d), on the other hand, are of dubious acceptability because they denote physical objects.9 This hypothesis accounts

360

Joseph Hilferty

for the lack of acceptability of sentence (14). Pollution denotes the resulting byproduct of some contaminating activity and not the activity itself. Because it is rather difficult (if not impossible) to construe pollution as representing a nominalized process, the sentence is judged to be semantically ill-formed.10

6.

Conclusion

This paper has shown that prepositional metaphor is clearly not beyond the pale of linguistic inquiry (contrary to what Vandeloise 1990 seems to intimate with regard to Lakoff's 1987 treatment of over). In fact, the present account reveals that through is remarkably coherent in its metaphorical manifestations, in that there exists a high degree of family resemblance between all the uses contemplated in this study. This lends credence to the assertion that through is the name of a network of interrelated, yet distinct, senses. The center of this network appears to be the spatial sense, discussed in Section 2, in which through expresses the traversal of a bounded MEDIUM. The metaphorical senses described in sections 3 and 4 all draw heavily on the basic facets of through^ prototypical spatial image; hence, their shared family resemblance and their overall metaphorical coherence. In the course of this study, I have paid special attention to two closely related conceptual metaphors: CAUSAL MEANS ARE PATHS and PROCEDURAL MEANS ARE PATHS. These are interesting metaphors because of the constraints that appear to be involved. On the one hand, the CAUSAL MEANS metaphor seems to obey constraints that go beyond considerations of image-schema "unification" (ä la Invariance Hypothesis). As shown in Section 5, through cannot express all causes; rather this use tends to work best when it introduces an action. It is hard to see how this observation can be formulated in strictly image-schematic terms, though it certainly can be explicitly stated in the mapping. On the other hand, the PROCEDURAL MEANS metaphor is interesting because it apparently observes constraints placed upon it by intuitions carried over from its source domain: the semantic well-formedness of utterances of this type depends principally on whether or not the element modified by the through-phrase can be conceived as an end. Enticing as it is, this working hypothesis will no doubt require further scrutiny. In closing, I would like to stress that by no means am I claiming that I have explained all of the metaphorical senses of through (to do so would certainly

Through as a means to metaphor

361

fill a book-length monograph). I believe, however, that the analyses proposed here are on the right track. It has been shown - convincingly, I hope - that the metaphorical senses of through are highly motivated by its prototypical spatial meaning. Though there may be some surprises, I am confident that (at least in its prepositional usage) most of through's other abstract senses will follow this pattern. If this prediction is correct, then the totality of through's distribution will manifest itself not as a haphazard myriad of unmotivated uses, but as a remarkably coherent natural category.

Acknowledgments Many thanks go out to Nicole Delbecque, Al Muth, and Sarah Taub for their critiques on earlier versions of this paper. I would also like to acknowledge Joe Grady and George Lakoff for suggesting examples (5b) and (14), respectively. Of course, all errors and shortcomings are strictly of my own making.

Notes 1. See, for example, Brugman 1981, Brugman and Lakoff 1988, Cuyckens 1991, Dewell 1994, Geeraerts 1992, Hawkins 1984, Lakoff 1987, Langacker 1992b, Lindner 1981, 1982, Taylor 1988, 1989, inter alia. 2. Cf. Taylor 1993. It should be noted that I have conveniently disregarded directional trajectories stressing goals (e.g., towards) and sources (e.g., away from). See Jackendoff (1983) for further discussion of these types spatial prepositions. 3. It should be noted that the term trajector has acquired an ambiguous value in the literature. Here I am definitely not employing the term in its syntagmatic sense advocated by Langacker; I am instead using it as an equivalent for what Herskovits (1986, 1988) calls the "located object" and Vandeloise (1986) the "target" in a spatial relation. 4. It should be noted that I am not claiming that through is necessarily monosemous in the domain of space. I am inclined to believe that it is polysemous, given examples such as the following: (i)

a. b. c.

Protestors marched through the streets of our nation's capital. Blood circulates through the body. There's a rusty old nail sticking through this board.

In some cases, through's landmarks are conceived as either UNBOUNDED, as in (ia), or CIRCUITIVE, as in (ib) (cf. Hawkins 1984). Other cases, such as (ic), differ further from what I suggest is the prototype, in that they lack the attribute of MOTION (cf. Langacker 1987: 239 for a discussion of a similar example).

362

Joseph Hilferty

Though certainly worth investigating, a full account of such data would lead us astray from the main points of this study. 5. The bulk of my explanation of the CAUSAL MEANS ARE PATHS conceptual metaphor is based on King's (1988) account of the German preposition durch. 6. This syntactic ambiguity can also be shown by contrasting (iia) and (iib): (ii)

a. b.

Peace through superior firepower and total retaliation is what I support. Through superior firepower and total retaliation, I support peace.

In (iia) the through-phrase clearly forms part of a subject NP and therefore gains an adjectival interpretation. On the other hand, in (iib), where the through-phrase is fronted, the PP can only modify the VP and thus allows only an adverbial interpretation. 7. This quotation is from p. 12 of Charles J. Sykes' A Nation of Victims: The Decay of the American Character (New York: St. Martin's Press, 1992). 8. Cf. Brugman 1990, Lakoff 1990, Turner 1990, 1991. 9. The only plausible way to comprehend these two examples is by the metonymy THE OBJECT FOR THE ACTIVITY. On such an interpretation, stamps would stand for collecting stamps, and gold could stand for dealing in gold. In terms of cognitive processing, however, this seems to be a costly interpretation to recover, especially if sufficient situational context is lacking, as is the case here (cf. Sperber and Wilson 1986). 10. Actually, this also explains why a sentence such as *He opened the door through a key is ill-formed. When through1?, object denotes an inanimate physical object, hearers try to understand the preposition in its TRAVERSAL sense. In this example, such an interpretation of course leads to an anomalous image, since the person opening the door is much larger than the key he is using. (The above example is due to Stef Grondelaers.)

References Benware, Wilbur A. 1993 Representing prepositions: New High German urn. Linguistics 31.1: 135-157. Brugman, Claudia M. 1981 Story of Over. M.A. thesis. University of California at Berkeley. Reproduced by Indiana University Linguistics Club, Bloomington, IN, 1983. (Published as: The Story of Over: Polysemy, Semantics, and the Structure of the Lexicon. New York: Garland Publishing, 1988). 1990 What Is the Invariance Hypothesis? Cognitive Linguistics 1.2: 257-266. Brugman, Claudia and George Lakoff 1988 Cognitive topology and lexical networks. In Steven Small, Garrison W. Cottrell and Michael K. Tanenhaus (eds.), Lexical Ambiguity Resolution: Perspec-

Through as a means to metaphor

363

tives from Psycholinguistics, Neuropsychology, and Artificial Intelligence, 477-508. San Mateo, CA: Morgan Kaufmann Publishers, Inc. Cuyckens, Hubert 1991 The Semantics of Spatial Prepositions in Dutch: A Cognitive-Linguistics Exercise. Doctoral diss. Universitaire Instelling Antwerpen. DeLancey, Scott 1982 Aspect, transitivity, and viewpoint. In Paul J. Hopper (ed.), Tense-Aspect: Between Semantics and Pragmatics, 167-183. Amsterdam: John Benjamins Publishing Company. Dewell, Robert B. 1994 Over again: Image-schema transformations in semantic analysis. Cognitive Linguistics 5.4: 351-380. Geeraerts, Dirk 1992 The semantic structure of Dutch over. Leuvense Bijdragen 81.1-3: 205-230. Hanssen, Federico 1911 Observaciones sobre la preposicion para. Bulletin Hispanique XIII: 40-^43. Hawkins, Bruce W. 1984 The Semantics of English Spatial Prepositions. Doctoral diss. University of California at San Diego. 1987 The profile imposition hypothesis. Manuscript. 1988 The natural category MEDIUM: An alternative to selection restrictions and similar constructs. In Rudzka-Ostyn (1988), 231-270. Herskovits, Annette 1986 Language and Spatial Cognition: An Interdisciplinary Study of the Prepositions in English. Cambridge: Cambridge University Press. 1988 Spatial expressions and the plasticity of meaning. In Rudzka-Ostyn (1988), 271-297. Jackendoff, Ray 1983 Semantics and Cognition. Cambridge, MA: The MIT Press. Johnson, Mark 1987 The Body in the Mind: The Bodily Basis of Meaning, Imagination, and Reason. Chicago: The University of Chicago Press. King, Robert Thomas 1988 Spatial metaphor in German causative constructions. In Rudzka-Ostyn (1988), 555-585. Lakoff, George 1987 Women, Fire, and Dangerous Things: What Categories Reveal about the Mind. Chicago: The University of Chicago Press. 1990 The Invariance Hypothesis: Is abstract reason based on image-schemas? Cognitive Linguistics 1.1: 39-74. Lakoff, George and Zoltän Kovecses 1987 The cognitive model of anger inherent in American English. In Dorothy Holland and Naomi Quinn (eds.), Cultural Models in Language and Thought, 195-221. Cambridge: Cambridge University Press.

364

Joseph Hilferty

Lakoff, George and Mark Turner 1989 More than Cool Reason: A Field Guide to Poetic Metaphor. Chicago: The University of Chicago Press. Langacker, Ronald W. 1987 Foundations of Cognitive Grammar, vol. I: Theoretical Prerequisites. Stanford, CA: Stanford University Press. 1990 Concept, Image, and Symbol: The Cognitive Basis of Grammar. Berlin: Mouton de Gruyter. 1991 Foundations of Cognitive Grammar, vol. II: Descriptive Application. Stanford, CA: Stanford University Press. 1992a Prepositions as grammatical(izing) elements. Leuvense Bijdragen 81.1-3: 287-309. 1992b The symbolic nature of Cognitive Grammar: The meaning of of and o/-periphrasis. In Martin Pütz (ed.), Thirty Years of Linguistic Evolution: Papers in Honour of Rene Dirven, 483-502. Amsterdam: John Benjamins Publishing Company. Leys, Odo 1989 Some remarks on spatial prepositional structure. In Frans J. Heyvaert and Frieda Steurs (eds.), Worlds Behind Words: Festschrift Droste, 77-84. Leuven: Leuven University Press. Lindner, Susan J. 1981 A Lexico-Semantic Analysis of English Verb-Particle Constructions with Out and Up. Doctoral diss. University of California at San Diego. Reproduced by Indiana University Linguistics Club, Bloomington, IN, 1983. 1982 What goes up doesn't necessarily come down: The ins and outs of opposites. Chicago Linguistic Society 18: 305-323. Lyons, John 1977 Semantics, vol. 1. Cambridge: Cambridge University Press. Rudzka-Ostyn, Brygida (ed.) 1988 Topics in Cognitive Linguistics. Amsterdam: John Benjamins Publishing Company. Sperber, Dan and Deirdre Wilson 1986 Relevance: Communication and Cognition. Oxford: Basil Blackwell Ltd. Also published by Harvard University Press, Cambridge, MA, 1988. Taylor, John R. 1988 Contrasting prepositional categories: English and Italian. In Rudzka-Ostyn (1988), 299-326. 1989 Linguistic Categorization: Prototypes in Linguistic Theory. Oxford: Clarendon Press. 1993 Prepositions: Patterns of polysemization and strategies of disambiguation. In Cornelia Zelinsky-Wibbelt (ed.), The Semantics of Prepositions: From Menial Processing to Natural Language Processing. Berlin: Mouton de Gruyter. Turner, Mark 1990 Aspects of the Invariance Hypothesis. Cognitive Linguistics 1.2: 247-255. 1991 Reading Minds: The Study of English in the Age of Cognitive Science. Princeton, NJ: Princeton University Press.

Through as a means to metaphor

365

Vandeloise, Claude 1986

1990

L'espace enfranqais (S4mantique des prepositions spatiales). Paris: Editions du Seuil. (Expanded Eng. version: Spatial Prepositions: A Case Study from French. Trans, by Anna R. Bosch. Chicago: The University of Chicago Press, 1991.) Representation, prototypes, and centrality. In Savas L. Tsohatzidis (ed.), Meanings and Prototypes: Studies in Linguistic Categorization, 403-437. London: Routledge.

Is metaphor really a one-way street? One of the basic tenets of the cognitive theory of metaphor put to the test Olaf Jäkel 1.

Introduction

This report introduces a survey of the understanding of metaphor carried out with native speakers of German in the autumn of 1992.1 It starts by establishing (in section 2) the basic theoretical tenet of the Cognitive Theory of Metaphor that is to be tested empirically. Then (in section 3) the make-up of the investigation is explained: In this section, the linguistic material is introduced, the test design is described and some information is given about the group of informants. In the main section (4), the most important results of the investigation are presented and discussed. The paper finishes with a summary and conclusion (in section 5), reassessing the initial hypothesis in the light of the empirical evidence.

2.

The unidirectionality hypothesis

The cognitive theory of metaphor2 as developed by George Lakoff und Mark Johnson endows metaphor with the important cognitive function of explanation and understanding. Conceptual metaphors are basic cognitive structures by means of which one subject matter is understood in terms of a completely different domain of experience. A metaphor of the type "X is Y" thus links two different conceptual domains, one of which (Y) functions as source domain, the other (X) as target domain of the metaphorical mapping. Because of this explanatory function of metaphor, I will call the two elements X and Y explanandum and explanans respectively. In a metaphor of the formal pattern "X is Y", the source domain provides Y as explanans for a partial explanation of the explanandum X. This explanation will be partial because a metaphorical explanans provides no more and no less than an imaginative model of the explanandum. Like any model taken in isolation, it has an effect of bias and simplification. It focusses on certain aspects of the explanandum while disregarding others.3

368

Olaf Jäkel

One further central claim of this theory is that, in general, abstract and complex target domains (X) are metaphorically conceptualized drawing on source domains (Y) which are concrete, simply structured and can be directly experienced through the senses: "The metaphors come out of our clearly delineated and concrete experiences and allow us to construct highly abstract and elaborate concepts, like that of an argument." (Lakoff and Johnson 1980: 105). Exceptions notwithstanding, this rule applies as a general preference or "tendency": "[...] the tendency to understand the less concrete in terms of the more concrete."4 In contrast to other - for example interactionist - theories of metaphor, it is strongly emphasized that the relation between the elements X and is irreversible; the metaphorical projection is said to have an unmistakable direction: "We would also like to explain the fact that the mappings are unidirectional: bodily experience is a source of vocabulary for our psychological states, but not the other way around."5 This unidirectionality hypothesis has been advocated right from the beginnings of the cognitive theory of metaphor in 1980, and essentially it has not changed to this day. The latest version of the hypothesis advocated by the father of the theory is as follows: "Metaphor allows us to understand a relatively abstract or inherently unstructured subject matter in terms of a more concrete, or at least a more highly structured subject matter." (Lakoff 1993: 42). The fact that the tenet of unidirectionality or directionality of metaphor is quite a bit older than the cognitive theory of metaphor is shown by a look at a theory of metaphor from the early sixties. The German linguist Harald Weinrich uses the term image fields ("Bildfelder") to express exactly what is meant by Lakoff's and Johnson's conceptual metaphors. Thus, in an argument against the Aristotelian theory of metaphor, he speaks of image fields, which, being traditional and social products, as a rule are unidirectional [einsinnig]. [ . . . ] Here, tradition has favoured one direction of metaphor, and with this directionality [Gerichtetheii], metaphor has unfolded into an image field, thus becoming not just a stylistic, but a linguistic reality.6

To conclude this introduction, there follows a final and definite formulation of the unidirectionality hypothesis: Unidirectionality hypothesis (U): As a rule, metaphor ("X is Y") links an abstract and complex target domain (X) as explanandum with a

Is metaphor really a one-way street?

369

source domain (Y) as explanans which is more concrete, simply structured, and open to experience through the senses.7 This unidirectionality hypothesis, which represents one of the fundamental pillars of Lakoff's and Johnson's theory of "experientialism"8, is put to the test in the following. In its very general form, the hypothesis permits no direct empirical testing. To do this, we have to derive empirically testable hypotheses from the unidirectionality hypothesis.9

2.1.

Empirically testable hypotheses

One derived hypothesis that implicitly underlies most investigations carried out within the framework of the cognitive theory of metaphor, though it is seldom voiced explicitly, is the following: Textlinguistic hypothesis (T): Texts/statements on abstract matters should be much richer in metaphors than those dealing with concrete matters. This textlinguistic hypothesis is a direct consequence of the unidirectionality hypothesis. A large number of lexical-semantic metaphor studies speak in favour of T. For a methodically watertight verification of this textlinguistic hypothesis, however, proper corpus-based analyses are needed, which would also permit statements about frequency. First attempts at such a systematic "onomasiological cognitive metaphorology" have been submitted10 and ought to be developed further. To permit another test of the general unidirectionality hypothesis based on independent evidence, another empirically testable hypothesis can be established on the basis of U: Hypothesis concerning the understanding of metaphor (M): Metaphors deviating from the generally preferred direction of transfer should prove harder to understand. The proposition expressed in this hypothesis might be regarded as the natural reason for the general phenomenology as described by the unidirectionality hypothesis. If competent speakers find metaphors with a certain direction of transfer easier to understand than all others, then this type of metaphor

370

Olaf Jäkel

best fulfills its communicative and explanatory function. Thus, because of its functional success, this type of metaphor will be sanctioned by the evolution of natural language. It is this empirical, and in principle falsifiable, hypothesis (M) concerning the understanding of metaphor that was to be tested in the survey presented in the following. The next section describes the make-up and procedure of this investigation.

3. 3.1.

Make-up of the investigation Linguistic material

For an investigation of the linguistic understanding of metaphor and in particular for a test of the hypothesis M, the linguistic material to be used had to meet two requirements. Firstly, in order to provide a fair starting point for comparison, generally unfamiliar metaphors were needed, which would require the same active and conscious processing. Familiarity with some metaphors would obviously have influenced the test subjects' understanding. Secondly, it had to be metaphors with diverse directions of transfer. The idea of artificially generating metaphors was dismissed in order not to expose the investigation to the criticism of unduly influencing its results from the start by suggestive constructions. The first of the two conditions already excluded everyday language metaphors from the test material. However, going back to poetic aphorisms, i.e., the realm of innovative metaphors, it was possible to meet both demands. Thus, the following ten example sentences provide the linguistic material for the investigation (an English translation is given):11 (1)

Phantasie ist etwas, was sich manche Leute gar nicht vorstellen können. 'Imagination is something some people have no idea of.'

(2)

Die Ehe ist ein Bauwerk, das jeden Tag neu errichtet werden muß. 'Marriage is a building that has to be put up anew each day.

(3)

Musik ist Liebe auf Suche nach einem Wort. 'Music is love in search of a word.'

Is metaphor really a one-way street?

371

(4)

Das Weib ist ein vernünftiges Märchen. 'Woman is a reasonable fairy tale.'

(5)

Wünsche sind das Rauschen in den Kronen der Träume. 'Desires are the rustling in the dream tops.'

(6)

Der Kompromiß ist ein guter Schirm, aber ein schlechtes Dach. 'Compromise is a good umbrella, but a bad roof.'

(7)

Persönlichkeit ist vom Geiste gebundene Seele. 'Personality is soul bound by spirit.'

(8)

Tanzen ist die Poesie des Fußes. 'Dancing is the poetry of the foot.'

(9)

Illusionen sind die Schmetterlinge des Lebensfrühlings. 'Illusions are the butterflies of life's spring.'

(10)

Eine Universität ist ein Ort, wo Kieselsteine geschliffen und Diamanten getrübt werden. 'University is a place where pebbles are polished and diamonds are dulled.'

These stimuli were in all likelihood unfamiliar to the test subjects. Therefore, to be understood all of them would require active and conscious processing. The diverse directions of transfer exhibited by these metaphors will be explained in section 4.1. Sentence (1) represents a nonmetaphorical example for a counter check.

3.2.

Design of the questionnaire

The survey was carried out using an eleven page questionnaire. Instructions were given on page 1; there followed ten standardized sheets, one for each of the example sentences. The presentation of the stimulus sentence at issue at the head of the page was followed by three blocks of questions. In the first of these blocks, a yes-or-no-decision had to be made in answer to the question "Did you understand this statement?" In the case of a positive answer, informants were asked to provide a written paraphrase of the alleged meaning. The second block inquired into the degree of difficulty of understanding as judged by the subjects. The following five-point scale was to be ticked off:

372

Olaf Jäkel

"very difficult" - "difficult" - "so-so/medium" - "not that difficult" - "quite easy". If they decided for "very difficult" or "difficult", test persons were asked to try and give some reasons for the difficulty of understanding. In the third set of questions, candidates were asked to judge the appropriateness of some given attributes to describe the stimulus sentences. For their assessment, informants were offered the standard five-point scale: "very apt" - "apt" - "don't know" - "not so apt" - "inapt". In this manner, the following 25 attributes had to be considered (English equivalents are given in brackets): geistreich ('witty') borniert/dumm ('stupid') poetisch ('poetic') witzig ('funny') überflüssig ('superfluous') dunkel ('dark/obscure') nichtssagend ( ' empty/trite') einseitig ('one-sided') verworren ('confused') schlüssig/klar ('clear') floskelhaft ('stereotyped') aphoristisch ('aphoristic') schön ('beautiful')

richtig/treffend ('right') stark wertend ('evaluating') erhellend ('elucidating') interessant ('interesting') provozierend ('provocative') abgedroschen ('hackneyed') abgehoben ( ' lofty') tiefsinnig ('profound') philosophisch ('philosophical') verzerrend ('distorted') metaphorisch ('metaphorical') blumig ('fancy/flowery').

As can be seen from this list, "metaphorical" was one of the attributes to be assessed. Apart from this, the term "metaphor" was not mentioned at all in the whole questionnaire. In the instructions informants were asked to take some time to answer the questionnaire. Performance was not controlled, i.e., participants were given the questionnaire to hand it back only when they had finished. From responses in the course of the survey it became clear that, at least in some cases, the time needed by the test subjects considerably exceeded the test designer's estimate. However, it was considered important not to place the respondents under pressure of time in that complex task. In general, it can be assumed that there were no random tick offs. More on that issue in the following section (3.3.).

Is metaphor really a one-way street?

3.3.

373

Informants (sample)

Test subjects were 39 native speakers of German, 23 of them female and 16 male, aged between 20 and 30.12 After some years of work in their jobs they were attending the Hamburg "Hansa-Kolleg" (a kind of 6th form-college of further education) to get their "Abitur" (German university entrance qualifications). At the time of the survey in autumn 1992 they were in their second year (of three) at that college. The choice of this relatively homogeneous group of informants ensured two things: On the one hand, a certain level of reflection necessary for the sensible response to sophisticated questions concerning linguistic understanding was guaranteed. On the other hand, the test subjects had no specialist knowledge of psychology, linguistics, and far less the theory of metaphor, which might have distorted the results of the investigation. Because of these points, the chosen group can be considered a reasonable sample for the intended test of the hypothesis M concerning the understanding of metaphor. The return rate of the questionnaires (handed out to 49 persons) was 80 percent. This high return rate of a quite time-consuming and labour-intensive questionnaire already shows that the test subjects were exceptionally motivated and willing, exhibiting a high degree of cooperation.13 Their positive attitude is also reflected in some of the qualitative results, which partly reveal an extreme eagerness to understand. Thus it can be assumed that the questionnaires were completed sensibly and carefully. The survey using the questionnaire described above supplied not only extensive qualitative answers to the two open questions, but also a sum total of 10,530 raw quantitative data, not all of which are used in the following presentation of results.

4.

Results

The results of our investigation are presented in two sections. Firstly, exploiting the first two blocks of questions from the questionnaire, we treat (in 4.1.) the central issue of the understanding of metaphor. Following this main section, we present (in 4.2.) some excerpts from the assessments of attributes.14

374

Olaf Jäkel

4.1.

Concerning the understanding of metaphors

Before we can introduce the results of the survey concerning the empirical test of the unidirectionality hypothesis, we have to supply the classification of the stimulus material used in terms of the diverse directions of metaphorical transfer. The following schematic presentation uses S for source domain (explanans), T for target domain (explanandum), A for abstract and C for concrete.*5 Thus, five types (I-V) of metaphor can be distinguished (English translations of the stimulus sentences are once more given): (I)

Metaphor with direction of transfer CONCRETE ->· ABSTRACT: Metaphor T:A

Die Ehe

ist

'Marriage

is

(6)

Der Kompromiß

ist

(10)

'Compromise Eine Universität

is ist

'University

is

(2)

(II)

S:C

ein Bauwerk, das jeden Tag neu errichtet werden muß. a building that has to be put up anew each day.' ein guter Schirm, aber ein schlechtes Dach. a good umbrella, but a bad roof.' ein Ort, wo Kieselsteine geschliffen und Diamanten getrübt werden. a place where pebbles are polished and diamonds are dulled.'

Metaphor with direction of transfer ABSTRACT ->· ABSTRACT: Metaphor 1

S:A

T:A I

I

(3)

Musik

ist

is

(5)

'Music Wünsche

sind

'Desires Persönlichkeit 'Personality

are ist is

(7)

Liebe auf Suche nach einem Wort. love in search of a word.' das Rauschen in den Kronen der Träume. the rustling in the dream tops.' vom Geiste gebundene Seele. soul bound by spirit.'

Is metaphor really a one-way street?

375

(III) Metaphor with direction of transfer ABSTRACT -» CONCRETE: .

Metaphor

ι

I

(4)

T:C

I

Das Weib 'Woman

S:A

I

ist is

, ,

ι

I

ein vern nftiges M rchen. a reasonable fairy tale.'

(IV) Metaphor with direction of transfer ABSTRACT/CONCRETE -> CONCRETE: Metaphor

1

ι

T:C

ι

S:A/C

ι

ι

A

(8)

Tanzen 'Dancing

ist is

C

die Poesie des Fu es. the poetry of the foot.'

(V) Metaphor with direction of transfer ABSTRACT/CONCRETE -)· ABSTRACT: Metaphor ι

T:A

ι

1 1

(9)

Illusionen 'Illusions

sind are

1

S:A/C C

die Schmetterlinge the butterflies

1 1

1

A

1

des Lebensfr hlings. of life's spring.'

The last two types are hybrid forms with source domains that combine concrete as well as abstract elements. We will look at the peculiarities of these hybrid types later. However, the types I to III are more crucial to our central issue. The metaphors under I are "well-formed" according to the unidirectionality hypothesis (U): An abstract target domain is metaphorically explained in terms of a concrete source domain. By contrast, the types II and III constitute "violations" of the direction of transfer preferred according to U, in that they make use of an abstract source domain as explanans. The fact that examples of type III are really scarce - in both poetical metaphors and everyday language - can be regarded as a first and partial corroboration of U.16 As more examples can be found of type II metaphors, we will in the following concentrate on a direct comparison between the understanding of groups I and II.

376

Olaf Jäkel

The prediction from the hypotheses U and M that is to be tested here can now be put in concrete terms relating to the types of metaphorical transfer: Metaphors of type I should be the easiest to understand, whereas the understanding of those of types II and III should prove much harder. The results of our investigation clearly show this prediction to be true. Table 1 lists the single metaphorical stimuli according to the degree of difficulty exhibited in their understanding. Thus we get a conspicuous ranking from stimulus (6), which 100 percent of the participants claimed to understand, to stimulus (4), which appeared comprehensible to not quite half (46.2%)17 of the informants. The assessment of the stimuli as "rather difficult" versus "not so difficult" correlates completely with these results, supporting the same ranking.18 The nonmetaphorical stimulus (1) is displayed at the head of the table to permit comparison. The column on the far right of the table identifies the types of direction of metaphorical transfer. Here we can clearly see a forming of groups within the ranking. Both the three metaphors (6, 2, 10) of type I and the three examples (5, 7, 3) of type II have among themselves very similar values of understanding as well as assessments of difficulty. On the understanding scale, group I comprises all three front runners, while group II is far behind, occupying the lower ranks, with stimulus (4) of type III bringing up the rear. The hybrid types (8/IV and 9/V) are in a good mid-position. We shall now discuss the quantitative results presented in table 1 in detail, drawing on some material from the qualitative parts of the survey. The metaphors of type I came up with values of understanding of just under 100%; as did only the nonmetaphorical aphorism (1) equally big. Understanding was found easy: The average of the difficulty values exactly matched the value of the nonmetaphorical stimulus (1), with an easiness ratio considerably higher. The paraphrases of the examples (6, 2, 10) formulated by the informants show a high degree of concurrence, coming close to a consensus. In contrast, the metaphors of type II were much less well understood, with values ranging between 74.4% and 56.4%. Accordingly, understanding was felt to be rather difficult. From the reasons given by the participants for this difficulty, some exemplary statements shall be quoted here, which were formulated in like manner for all the three examples (5, 7, 3) (translation by the author): "Because to me the words make no sense in this combination." "I cannot say why. I understand the words, but cannot see any connection." "To me somehow this is too abstract."

Is metaphor really a one-way street?

377

Table 1. Ranking of metaphors according to the degree of difficulty in understanding Understood

rather difficult - not so difficult in %

(1) "Phantasie ist etwas, was sich manche Leute gar nicht vorstellen können."

97.4

5.1-61.5

No metaphor

(6) "Der Kompromiß ist ein guter Schirm, aber ein schlechtes Dach." (2) "Die Ehe ist ein Bauwerk, das jeden Tag neu errichtet werden muß." (10) "Eine Universität ist ein Ort, wo Kieselsteine geschliffen und Diamanten getrübt werden."

100.0

2.6-79.5

Concrete -> Abstract

94.9

5.1-89.7

Concrete-> Abstract

94.9

7.7-74.4

Concrete ->· Abstract

(8) "Tanzen ist die Poesie des Fußes."

92.3

7.7-64.1

Abstract/concrete Concrete

(9) "Illusionen sind die Schmetterlinge des Lebensfrühlings."

84.6

23.1-48.7

Abstract/concrete Abstract

(5) "Wünsche sind das Rauschen in den Kronen der Träume." (7) "Persönlichkeit ist vom Geiste gebundene Seele." (3) "Musik ist Liebe auf Suche nach einem Wort."

74.4

33.3-38.5

Abstract -> Abstract

66.7

48.7-35.9

Abstract ->· Abstract

56.4

53.9-18.0

Abstract ->· Abstract

46.2

61.5-20.5

Abstract ->· Concrete

Stimulus

(4) "Das Weib ist ein vernünftiges Märchen."

Direction of metaphorical transfer (source domain -> target domain)

"It comes close to the bounds of imagination." "Because I could read several meanings into it." "Very vague statement, very much open to interpretation." The problems with this type of metaphor pointed out in the last two quotes can be confirmed by a look at the paraphrases formulated by those subjects who ticked off the statements (5), (7), and (3) as "understood": These para-

378

Olaf Jäkel

phrases constitute such a heterogeneous picture that the subjectively felt "understanding" ought to be questioned critically as regards its intersubjective stability. What is the communicative value of expressions that make an astonishing number of hearers understand "something", without leading to some reasonably common understanding shared by its interpreters? Metaphor (4) of type III records the highest negative values in the understanding questions. Only less than half (46.2%) of the respondents claimed to have understood the statement. With 61.5% it also has the highest difficulty scores of all stimuli. The reasons given for this were very similar to those found with type II. Again with this stimulus (4), the alleged "understanding" of those 46.2% must be questioned, as here, too, paraphrases were quite varied and inconsistent, often remaining "vehicle-immanent": Subjects expounded on the source domain of the metaphor without ever establishing a link with the explanandum of the target domain. To conclude this presentation we shall take a short look at those metaphors of types IV and V. They occupy mid-positions in the judgments of understanding, somewhere between the extreme groups treated so far. With their source domains combining concrete and abstract elements, these hybrid types permit no clear prediction from the hypotheses U and M. Seen in retrospect, it might be presumed that they are understood relatively well (92.3% and 84.6%) because of their concrete elements. For stimulus (9), which records a significantly higher degree of difficulty of 23.1 % - with reasons given closely resembling those under II and III -, the subjectively felt "understanding" would again have to be seen in perspective. As with types II and III, we find here relatively heterogeneous paraphrases, in this case even with a striking discrepancy between more positive and more negative interpretations of the metaphorical statement. Finally, stimulus (8) presumably owes its rather high understanding values at least in part to the application of a unique comprehension strategy, which is not yet accounted for in any way by the concrete/abstract classification. The - mostly concurring - paraphrases reveal that in the understanding of (8), participants reconstruct a complete analogy: Sprache : Poesie (in contrast to Prosa) = Fuß : Tanzen (i. c. t. Gehen) [language : poetry (i. c. t. prose) =foot: dancing (i. c. t. walking)]. In addition, a complex tertium comparationis is identified (quotes from the subjects' paraphrases, translated by the author): "Artistic way of expressing feelings/emotions/beauty", "aesthetic treatment", "creative", "nontrivial", "not utilitarianist".

Is metaphor really a one-way street?

379

The fact that this metaphor seems to invite such a strategy of understanding might well be regarded as a small late victory by Aristotle, who already in his writings on rhetoric classifies "metaphors built on grounds of analogy" as a special type of metaphor besides those with diverse directions of transfer.19 Even though metaphors of this type seem to be quite rare20, the qualitative results of our survey call for some theoretical caution or modesty. The phenomenon of metaphor is indeed too complex and multifarious for all its aspects to be settled completely in a statement of simple laws or regularities. Nevertheless we can sum up this section as follows: The results of our survey presented so far impressively corroborate the derived hypothesis (M) concerning the understanding of metaphor, as all investigated metaphors that did not follow the preferred direction of transfer proved more difficult to understand than those "directionally well-formed" ones of type I. And so the unidirectionality hypothesis (U) as the basis of M may also be considered empirically confirmed.

4.2.

Comparison of attributions

Having concluded the central part of the inquiry dealing with the actual empirical test of the unidirectionality hypothesis, we can in the following present some of the quantitative results of the survey concerning the attribution to the metaphorical stimuli of certain given qualities. From those 25 attributes to be assessed in the third set of questions (see section 3.2.), ten have been selected here. The assessments are compared especially for the two groups of metaphors of type I and of type II, in order to gain further insight into the characteristics of the most comprehensible versus the least comprehensible metaphors. The results are presented in a table for each single attribute and then briefly commented on. Tables 2-10 each in turn present in their left half the metaphors (2, 6, 10) of type I and in the right half those (3, 5, 7) of type II. Beside each metaphor two percentages are given, representing the "appropriate" "inappropriate" votes.21 In the assessment of the attribute "schlüssig/klar" ('clear') there are considerable differences between type I and type II.22 From two thirds to 82.1% of the participants found this attribute appropriate for type I, while it was judged inappropriate for type II by a weaker though still absolute majority (51.3-66.7%). These results obviously confirm the group-specific judgments

380

Olaf Jäkel

Table 2. Assessment of the attribute "schlüssig/klar" ('clear') "Die Ehe ..." "DerKompromiß..." "Eine Universität..."

82.1- 7.7 79.5-12.8 66.7-18.0

"Musik ..." "Wünsche..." "Persönlichkeit..."

18.0-64. l 20.5-66.7 28.2-51.3

Table 3. Assessment of the attribute "richtig/treffend" ('right')

"Die Ehe . . ." "Der Kompromiß . . ." "Eine Universität . . ."

89.7- 5.1 74.4-15.4 43.6-20.5

"Musik . . ." "Wünsche . . ." "Persönlichkeit . . ."

15.4-38.5 43.6-25.6 35.9-23.1

Table 4. Assessment of the attribute "verworren" ('confused') "Die Ehe..." "DerKompromiß..." "Eine Universität..."

2.6-89.7 5.1-84.6 7.7-82.1

"Musik..." "Wünsche..." "Persönlichkeit..."

41.0-43.6 33.3-51.3 33.3-48.7

on difficulty of understanding. Statements that are regarded as clear are understood better and with more ease. The attribute "richtig/treffend" ('right') applies to type I in the eyes of the majority (43.6-89.7%). It is less appropriate (approving votes between 15.4% and 43.6%) for type II, though here we find a varied group picture with relatively balanced votes and a striking amount of uncertainty: With shares of between 30.8% and 46.1%, the "don't know" vote is extremely high. This uncertainty effect can easily be explained, as judging a statement as "right" presupposes an understanding of that statement. Indeed, the low understanding values of type II would not suggest the prediction of unanimous disapproval of the attribute "right", but rather of high "don't know" values. First in a series of possible antonyms of "schlüssig/klar", the attribute "verworren" ('confused') indeed scores reciprocal votes, at least for type I. For this type of metaphor, the attribute is rejected almost unanimously as inappropriate (by 82.1-89.7%). By contrast, it is more appropriate for type II (at least for between a third and 41% of the participants), with disapproving and approving votes rather balanced. The attribute "abgehoben" ('lofty') records assessment values very similar to those of "verworren". The absolute majority (59.0-89.7%) of subjects finds this term inappropriate for type I. It is judged more appropriate for type II, again with relatively balanced votes.

Is metaphor really a one-way street?

381

Table 5. Assessment of the attribute "abgehoben" ('lofty') "Die Ehe..." "Der Kompromiß..." "Eine Universität..."

5.1-89.7 5.1-79.5 12.8-59.0

"Musik..." "Wünsche..." "Persönlichkeit..."

43.6-33.3 41.0-38.5 35.9-35.9

Table 6. Assessment of the attribute "tiefsinnig" ('profound') "Die Ehe..." "DerKompromiß..." "Eine Universität..."

33.3-38.5 51.3-35.9 20.5^8.7

"Musik..." "Wünsche..." "Persönlichkeit..."

46.2-30.8 56.4-20.5 59.0-20.5

Table 7. Assessment of the attribute "philosophisch" ('philosophical') "Die Ehe..." "DerKompromiß..." "Eine Universität..."

25.6-51.3 35.9^6.2 18.0-61.5

"Musik..." "Wünsche..." "Persönlichkeit..."

33.3^3.6 48.7-33.3 64.1-10.3

Type II is deemed "tiefsinnig" ('profound') (by 46.2-59.0%); type I by contrast less so, though with a heterogeneous picture (20.5-51.3%) and relatively balanced votes. Profoundness seems to be a popular quality to be attributed even without a full understanding of the statement in question. By calling it "profound", something positive can be said about some phrase that is not fully understood and which may to a degree really be incomprehensible. The majority (46.2-61.5%) of participants does not regard type I as "philosophisch" ('philosophical'). For type II the term is thought more appropriate, although votes (between a third and 64.1% approval, between 10.3% and 43.6% disapproval) are both relatively balanced and heterogeneous. Apart from these rather rational attributes, the questionnaire offered some others which asked for an aesthetic or stylistic assessment of the stimuli. The following three attributes reveal in their assessment by the subjects interesting correlations with the evaluations of understanding. The attribute "schön" ('beautiful') is appropriate for type II in the opinion of most subjects (48.7-74.4%), and rather not appropriate (only 12.8-38.5% approval) for type I. Once more, judgments on type I are relatively balanced. The contrary assessment of the two types is even more indisputable where the attribute "poetisch" ('poetic') is concerned. The absolute majority of participants thinks this term appropriate for type II (61.5-71.8%), but not for

382

Olaf Jäkel

Table 8. Assessment of the attribute "schön" ('beautiful') "Die Ehe..." "DerKompromiß..." "Eine Universität..."

38.5-41.0 33.3-41.0 12.8-66.7

"Musik..." "Wünsche..." "Persönlichkeit..."

51.3-28.2 74.4-18.0 48.7-28.2

Table 9. Assessment of the attribute "poetisch" ('poetic')

"Die Ehe . . ." "Der Kompromiß . . ." "Eine Universität . . ."

28.2-51.3 28.2-51.3 23.1-61.5

"Musik . . ." "Wünsche . . ." "Persönlichkeit . . ."

61.5-18.0 71.8-15.4 61.5-25.6

Table 10. Assessment of the attribute "blumig" ('fancy/flowery') "Die Ehe..." "DerKompromiß..." "Eine Universität..."

12.8-71.8 5.1-69.2 10.3-61.5

"Musik..." "Wünsche..." "Persönlichkeit..."

56.4-28.2 59.0-20.5 25.6-51.3

type I (51.3-61.5%). In this, two different causal relations between the assessments of understanding and of poeticity might be conceived: On the one hand, a statement can be difficult to understand because it appears particularly poetical. On the other hand, as with the assessment of a statement as "profound", the attribution of poetic qualities can compensate a lack of comprehensibility, and thus conceal or justify a lack of understanding. The empirical results admit either of the two explanations, while substantiating the obvious correlation of poetic quality and difficulty of understanding. Just like "poetical", the attribute "blumig" ('fancy/flowery') is rejected as inappropriate for type I by an absolute majority (61.5-71.8%) of subjects. By contrast, it is more appropriate for type II, though the approval of this fluctuates between 25.6% and 59.0%, thus remaining heterogeneous. To conclude this section, we shall consider the attribute "metaphorisch" ('metaphorical'). Unlike with the preceding attributes, assessments will now be presented not exclusively for the metaphors of types I and II, but for all of the ten stimuli. In Table 11, metaphors are listed according to the size of their metaphoricity values as scored in the survey. Stimulus (1), which is included for comparison purposes, is classified correctly as nonmetaphorical by 61.5% of the subjects. With the exception of example (7), all metaphors are identified as metaphorical by a majority of

Is metaphor really a one-way street?

383

Table 11. Ranking of metaphors according to the degree of metaphoricity assessed by subjects Stimulus

(1) "Phantasie ist etwas, was sich manche Leute gar nicht vorstellen können." (9) "Illusionen sind die Schmetterlinge des Lebensfrühlings." (5) "Wünsche sind das Rauschen in den Kronen der Träume." (6) "Der Kompromiß ist ein guter Schirm, aber ein schlechtes Dach." (10) "Eine Universität ist ein Ort, wo Kieselsteine geschliffen und Diamanten getrübt werden." (2) "Die Ehe ist ein Bauwerk, das jeden Tag neu errichtet werden muß." (3) "Musik ist Liebe auf Suche nach einem Wort." (4) "Das Weib ist ein vernünftiges Märchen." (8) "Tanzen ist die Poesie des Fußes." (7) "Persönlichkeit ist vom Geiste gebundene Seele."

Attribute "metaphorical": appropriate - inappropriate in % 10.3-61.5 76.9-12.8 69.2-10.3 66.7-15.4 61.5-20.5 61.5-25.6 53.9-18.0 46.2-12.8 41.0-38.5 30.8-35.9

votes; in fact, except for stimuli (4) and (8), by an absolute majority of up to 76.9% (for example 9). Stimulus (7) records a slight predominance in favour of "rather not metaphorical", together with a high amount of uncertainty: A third of the participants remained undecided, voting "don't know". Concerning the groups of metaphors I and II, no homogeneous tendency could be established in the assessment of these stimuli as "metaphorical". A consistently lower metaphoricity of type I metaphors as against higher values with type II might have seemed conceivable. The results of our survey paint a different picture: While the examples (2, 6, 10) of type I have high scores (around two thirds of votes), those (3,5,7) of type II record highly different metaphoricity values. Thus, we can establish no connection between difficulty of understanding and degree of metaphoricity. On the whole, the comparison with the correct identification of stimulus (1) as nonmetaphorical shows that the group of informants make relatively competent use of the attribute "metaphorical".

384

5.

Olaf Jäkel

Summary and conclusion

Following our detailed presentation and assessment of results, the essential outcome of this inquiry will now be given in summary. The empirical survey of the understanding of metaphor introduced in this paper establishes a number of substantial differences between those metaphors which follow the preferred direction of transfer from concrete source domain to abstract target domain, as the unidirectionality hypothesis (U) has it (type I), and those with different directions of transfer (types II/III): - Metaphors of type I are understood best, much better than those of type II/III.23 (Corroboration of hypothesis M concerning the understanding of metaphor.) - The understanding of type I metaphors is found easy, in some cases even easier than that of nonmetaphorical aphorisms; the understanding of metaphors of type II/III by contrast is regarded as rather difficult. - Metaphors of type I are to a great extent interpreted unanimously, thus enabling intersubjective understanding. Those heterogeneous attempts at interpreting metaphors of type II/III make any common and shared understanding dubious, if not impossible. - Type I metaphors are typically assessed as "clear" and also as "right"; those of type II/III by contrast are rather termed "confused", "lofty", "profound", or "philosophical". - In contrast to metaphors of type I, those of type II/III are rather assigned aesthetic or stylistic qualities: They are considered "beautiful", "poetic", and "fancy/flowery". - No correlation can be established between difficulty of understanding and degree of metaphoricity: Comprehensible statements do not necessarily have to be less metaphorical, and the reason for lack of comprehensibility is not in the fact that some statement is metaphorical, but in the kind of metaphor that is employed. With the hypothesis (M) concerning the understanding of metaphor being empirically corroborated by the results of our survey, the unidirectionality hypothesis (U) as the basis of M may also be considered empirically confirmed. As regards the interpretation of U, however, an important clue lies in the possibility of (for instance analogical) metaphors which are not at all adequately analysed by mere description of their direction of transfer. It is a

Is metaphor really a one-way street?

385

fact that such metaphors exist - at least in the innovative realms of rhetoric and poetry - and are understood as well. This should make theoreticians of metaphor bear in mind that the directionality as stated in U is nothing more and nothing less than a tendency or preference: an efficient rule with great explanatory power, though not without exceptions. The conception of metaphor as a kind of one-way street suggests a rigidity of fixed, rule-governed direction of transfer which simply does not exist in linguistic reality. Perhaps this complex reality is matched better by an alternative image: Linguistic understanding, pictured as a big, wide river with a powerful current from the concrete and better known towards the more abstract and unknown, easily carries the common rafts of everyday language along in its own natural direction. But it takes only little courage to hoist a poetic sail and go against the flow. And maybe some sailors only mean to cross the river, just for the fun of it. As long as one has some fairly seaworthy vehicle to go by, there is hope: The benevolent river of understanding will carry even weak vessels and daring constructions.

Notes 1. An extended version of this paper was prepublished by C.L.E.A.R. Cognitive Linguistics: Explorations, Applications, Research 1: 1993, University of Hamburg. I thank Stephen W. Telfer (of Fife) for checking my English; all remaining mistakes are mine. 2. The most important contributions to the cognitive theory of metaphor are Lakoff and Johnson (1980), Lakoff (1987 and 1993), Lakoff and Turner (1989), Sweetser (1990), and Johnson (1987 and 1992). 3. Lakoff and Johnson (1980: lOff.) treat this focussing effect of metaphor as Highlighting and Hiding. 4. Lakoff and Johnson (1980: 109). Cf. also Sweetser (1990: 25 and 31). 5. Sweetser (1990: 30), with original highlighting. Cf. Lakoff and Turner (1989: 132). 6. Weinrich (1963: 315), translated by the author, with original highlighting. A more comprehensive treatment of the merits of this and other predecessors of the cognitive theory of metaphor will be provided in Jäkel (1997). 7. The terms "abstract" und "concrete" are used by Lakoff and Johnson at crucial points within their theory, unfortunately without ever being explicitly defined. In a traditional, Kant-oriented, philosophical sense I understand "concrete" as "open to experience through the senses", and "abstract" as "not concrete"; cf. Hoffmeister (1955: 9f and 354), as well as Kant's (1800) Logik, §16. Incidentally, this understanding tallies with the common linguistic dichotomy "concrete

386

Olaf Jäkel

(noun)" vs. "abstract (noun)"; cf. Lewandowski (1984: 24f. and 563). That the terms "concrete" and "abstract" should rather be located on a continuous scale is discussed elsewhere (Jäkel 1997). 8. Cf. especially Johnson (1987 and 1992) as well as Lakoff (1987). 9. In this I follow Popper's (1935) standard description of a critical-rationalist approach to the empirical testing of theories. 10. E.g. Jäkel (1990, 1993a, 1995 and 1996). Concerning the methodology of an "onomasiological cognitive metaphorology" see Jäkel (1997). 11. The aphorisms are by the following authors: (1) Gabriel Laub, (2) Andre Maurois, (3) Sidney Lanier, (4) Peter Hille, (5) Martin Kessel, (6) James Rüssel Lowell, (7) Ludwig Klages, (8) John Dryden, (9) Peter Sirius, (10) Robert Green Ingersoll. 12. With the exception of one 44-year-old informant. 13.1 want to take this opportunity to thank again all the participants of the survey for their trouble and care in answering especially those qualitative questions. Although these explicitly formulated answers cannot nearly be enlarged on here, I have benefited greatly from them. Regina Jäkel was - apart from all else an invaluable help in the distribution, collecting, and primary evaluation of the questionnaires. 14. The results of a repeat test cannot be reported here due to lack of space. See Jäkel (1993b, 1997). 15. Concerning the problems with the antonymical terms "abstract" vs. "concrete" cf. footnote 7. 16. Even in sentence (4), "Das Weib" might be regarded as a more abstract general term. Then (4), too, would come under the category of II. The results of the survey do not contradict such a view. 17. All percentages given in this paper have been rounded to one decimal place. 18. In this column, the label "rather difficult" combines the percentages of the answers "very difficult" and "difficult", and the label "not so difficult" includes the percentages of "not that difficult" and "quite easy". The share of the answer "so-so/medium", which is not presented here, accounts for the balance to 100%. 19. Aristotle, Rhetoric, book III, chapter 10 (141 la); also cf. his Poetics, chapter 21 (1457b). I have explained elsewhere that Lakoff and Johnson with their sweeping criticism do not always do justice to the Aristotelian theory of metaphor (Jäkel 1997). 20. It seems very likely that this kind of analogy metaphor occurs exclusively in the innovative realm of poetic literature and rhetoric, but not at all in everyday language. 21. Here the labels integrate once more: "appropriate" includes the questionnaire votes for "apt" and "very apt", while "inappropriate" includes "not so apt" and "inapt". The shares of the neutral answer "don't know", which are not given in the presentation, account for the margin to 100%.

Is metaphor really a one-way street?

387

22. Here and in the following, I use 'type and 'type as handy abbreviations for "the metaphors (2, 6, 10) of type I" and "the metaphors (3, 5, 7) of type " respectively. 23. No doubt this difference would turn out more extreme with a less willing and cooperative group of informants. Still more pronounced differences between the metaphors of type I and II/III would have to be expected if, instead of a questionnaire survey without pressure of time, controlled reaction time experiments were done to test the understanding of the same (or some similar) stimulus material. The fact that, even without such artificially-induced pressure of time, we were able to establish characteristic differences between the understanding of the diverse types of metaphor makes our investigation all the more valid and convincing.

References Aristotle 1959 The 'Art' of Rhetoric. Transl. by John Henry Freese, London. 1969 Poetics. Transl. by John Warrington, London: Dent. 1980 Rhetorik. Transl. and ed. by Franz G. Sieveke, München: Wilhelm Fink. 1982 Poetik. Transl. and ed. by Manfred Fuhrmann, Stuttgart: Philipp Reclam. 1983 Von der Dichtkunst. Transl. and ed. by Olof Gigon, München: dtv. Hoffmeister, Johannes 1955 Wörterbuch der philosophischen Begriffe. Hamburg: Meiner. Jäkel, Olaf 1989 'Der handgreifliche Intellekt': zur Metaphorik geistiger Tätigkeiten. Grazer Linguistische Studien 32: 5-19. 1990 'Der handgreifliche Intellekt', oder: MENTAL ACTIVITY IS MANIPULATION: Kognitiv-linguistische Untersuchung einer grundlegenden konzeptuellen Metapher der englischen Alltagssprache. Thesis, published in the paper series of the Graduiertenkolleg Kognitionswissenschaft, University of Hamburg. 1993a 'Economic Growth' versus 'Pushing up the GNP': Metaphors of Quantity from the Economic Domain. L.A.U.D. 1993 (C:24), University of Duisburg. 1993b Is Metaphor Really a One-way Street? One of the Basic Tenets of the Cognitive Theory of Metaphor Put to the Test. C.L.E.A.R. Cognitive Linguistics: Explorations, Applications, Research 1: 1993, University of Hamburg. 1995 The Metaphorical Concept of Mind: MENTAL ACTIVITY IS MANIPULATION. In: John R. Taylor and Robert E. MacLaury (eds.) Language and the Cognitive Construal of the World. Berlin/New York: Mouton de Gruyter, 197-229. [Prepublished by L.A.U.D. 1993 (A:333), University of Duisburg.] 1996 Metaphorical Scenarios of 'Science'. In Martin Pütz and Reni Dirven (eds.), The Construal of Space in Language and Thought. Berlin/New York: Mouton de Gruyter, 649-678.[Prepublished by C.L.E.A.R. Cognitive Linguistics: Explorations, Applications, Research 6: 1995, University of Hamburg.]

388 1997

Olaf Jäkel Metaphern in abstrakten Diskurs-Domänen: Eine kognitiv-linguistische Untersuchung anhand der Bereiche Geistestätigkeit, Wirtschafi und Wissenschaft. Frankfurt a. M.: Peter Lang.

Johnson, Mark 1987 The Body in the Mind. The Bodily Basis of Meaning, Imagination, and Reason. Chicago/London: University of Chicago Press. 1992 Philosophical implications of cognitive semantics. Cognitive Linguistics 3.4: 345-66. Kant, Immanuel 1800 Logik, in: Werkausgabe, ed. by Wilhelm Weischedel, Vol. 6, Frankfurt/M. (1981): Suhrkamp. Lakoff, George 1987 Women, Fire, and Dangerous Things: What Categories Reveal about the Mind. Chicago/London: University of Chicago Press. 1990 The Invariance Hypothesis: is abstract reason based on image- schemas? Cognitive Linguistics 1.1: 39-74. 1993 The Contemporary Theory of Metaphor. In: Andrew Ortony (ed.) Metaphor and Thought. Cambridge: Cambridge University Press, 202-251. Lakoff, George and Mark Johnson 1980 Metaphors We Live By. Chicago/London: University of Chicago Press. 1989 More than Cool Reason: A Field Guide to Poetic Metaphor. Chicago/London: University of Chicago Press. Lewandowski, Theodor 1984 Linguistisches Wörterbuch. Heidelberg: Quelle & Meyer. Popper, Karl R. 1935 Logik der Forschung. Tübingen 1984: J.C.B. Mohr. Sweetser, Eve E. 1990 From Etymology to Pragmatics: Metaphorical and Cultural Aspects of Semantic Structure. Cambridge: Cambridge University Press. Weinrich, Harald 1963 Semantik der kühnen Metapher. In: Harald Weinrich. 1976, 295-316. 1976 Sprache in Texten. Stuttgart: Klett.

Syntax and semantics

Samoan as an active zone language1 Kenneth William Cook

1.

Introduction

Langacker (1984), working in the framework of Cognitive Grammar, observes that there is a discrepancy between the choice of subject and what he calls the "active zone" of English examples like (1): (1)

a. b.

David blinked. I'm in the phone book.

Although it is actually David's eyelids which blink, and the name and phone number of the speaker which are in the phone book, in both (la and Ib) the clause selects as subject the people involved (David and the speaker) rather than the active zones (i.e., the facets of an entity that most directly participate in a given relationship). Such discrepancies are presumably motivated in that they allow linguistic expressions to focus on entities which are cognitively salient, e.g., humans vs. nonhumans, wholes vs. parts, concrete objects vs. abstractions, etc.2 The examples in (1) are from English, which is a language which allows (or even favors) the coding of inherently salient entities (instead of active zones) in the syntactically prominent positions of subject and object. Presumably, however, active zones themselves are also cognitively salient since they are most centrally involved in events, and it is logically possible that some language would favor coding active zones rather than inherently salient entities in syntactically prominent positions. Based on my own observations and those of Duranti (1981, 1990, 1994), Duranti and Ochs (1983, 1990), Ochs (1988), Shore (1982), Mosel (1991) and Mosel and Hovdhaugen (1992), I would say that Samoan is such a language. Specifically, Samoan is an ergative language in which active zones often bear the syntactically prominent grammatical relation of absolutive. In the next section I will describe the ergative case marking system of Samoan and argue for the syntactic prominence of the absolutive. In the following two sections I will demonstrate the coding of active zones in Samoan clauses

392

Kenneth William Cook

and contrast the expression of transitive agents as possessors with their expression as ergatives. Finally, I will conclude the paper with a discussion of the tendency in Samoan to code events "from the inside out."

2.

Ergative case marking and the syntactic prominence of the absolutive

Since Samoan is an ergative language, it has in its repertoire of clauses the ergative construction exemplified in (2a). In this construction the agent-like participant is an ergative marked e? However, as will be shown below, transitive agents can also be coded as possessors. The patient-like participant in the ergative construction is an absolutive and is unmarked or (less frequently) marked Ό. The central participant of an intransitive clause like (2b) is also an (unmarked) absolutive.4 (2)

a.

b.

Na tipi e le tama le ufi. (ergative) PAST cut ERG the boy the yam The boy cut the yam.' 'Ua alu le tama 'i Samoa, (intransitive) PERFgo the boy DIR Samoa 'The boy has gone to Samoa.'

Evidence that the absolutive relation is syntactically prominent comes from observations concerning quantifier float and what in Cook (1991) I have called "modem verb agreement".5 Absolutives are the only nominals in Samoan that can participate in both of these phenomena no matter where they occur in a clause. In contrast, ergatives, which are presumably syntactically salient to some degree, must be in immediate postverbal position in order to participate in these phenomena. As illustrated in (3), an absolutive can be associated with the floating quantifier 'uma 'all' whether it occurs immediately after the verb or later in the clause. (3)

a.

b.

Na 'ave 'uma e le tama tusi.6 PAST take all ERG the boy book The boy took all the books.' Na 'ave 'uma tusi e le tama. PAST take all book ERG the boy The books were all taken by the boy.'

Samoan as an active zone language

393

An ergative, however, can be associated with the floating quantifier 'uma only if it occurs in immediate postverbal position. (4)

a.

E uli 'uma e a'uuö la'u ta'avale. IMP drive all ERG my friend my car 'My friends all drive my car.' b. *E uli 'uma la 'u ta 'avale e a'u uo. IMP drive all my car ERG my friend 'My car is driven by all my friends.'

Similarly, the verb agrees in number with the absolutive whether the latter occurs right after the verb or later in the clause. This is illustrated in (5a, b). Sentences (5c, d) show that an ergative must be in immediate postverbal position in order to trigger verb agreement. (5)

a.

b.

c.

d.

Na tutuli e le ali'i 'avefe'au. PAST send(PL) ERG the chief messenger The chief sent the messengers.' Na tutuli 'avefe'au e le ali'i. PAST send(PL) messenger ERG the chief 'The messengers were sent by the chief.' Na tutuli e ali'i le 'avefe'au. PAST send(PL) ERG chief the messenger 'The chiefs sent the messenger.' Na tuli (*tutuli) le 'avefe'au e ali'i. PAST send (send(PL)) the messenger ERG chief 'The messenger was sent by the chiefs.'

In sum, I take the fact that only absolutives participate freely in these two phenomena as evidence that absolutives are syntactically prominent in Samoan.

3.

The coding of active zones in Samoan clauses

The most basic expression involving an active zone in Samoan would be an intransitive clause like (6) which consists of only a verb and an absolutive. (6)

Tiga taliga! hurt ear '(My) ears hurt!'

394

Kenneth William Cook

This sentence is said in anger and is tantamount to a command to be quiet. Notice that the body part rather than any individual is mentioned in this expression. Sentence (6) does not overtly specify whose ears are hurting. However, as illustrated in (7a-c), body parts are often coded as absolutives while the individuals that possess those parts are expressed as (syntactically peripheral) possessives of those absolutives. I will henceforth refer to this pattern as the "possessive-absolutive construction." (7)

a.

b.

c.

'Ua ma'αϊ tele lou ulu. PERF sharp very your head 'You were very clever.' (lit. Your head is/was very sharp.) (Milner 1966: 117) 'Ua gau lona vae. PERF break his leg 'He broke his leg.' (lit. His leg has/is broken.) (Milner 1966: 78) 'Ua tete lota moa. PERF tremble my solar plexus Ί am afraid.' (lit. My solar plexus trembles.) (Milner 1966: 146)

This type of coding occurs even in imperatives. Consider, e.g., (8a-c). (8)

a.

b.

c.

Savaliou vae! walk your(PL) leg 'Get a move on!' (lit. Walk your legs.) Va'ai leleiou mata! look/see well your(PL) eye 'Look (open your eyes).' (lit. Look well your eyes.) (Milner 1966: 308) Salapu lou gutu! shut-up your mouth 'Shut up (your mouth)!'

Samoan as an active zone language

395

Sentence (8c) is interesting in that salapu, which is a Samoanization of English shut up, takes on a body part (lou gutu 'your mouth') even though the original English expression does not mention one. This detail suggests that the coding of active zones as absolutives is productive. With respect to coding humans and their body parts, Samoan contrasts in an interesting way not only with English but also with Romance languages such as Spanish. As can be seen in declarative sentences like (9a, b), English mentions the person as subject and as possessor of the body part; Spanish mentions the person only as subject, and as we have seen in (7), Samoan mentions the person only as a possessor. Thus, if the person is to be mentioned at all, and if it is to be mentioned either as a subject/absolutive and/or as a possessor, then the three languages exhaust the logical possibilities.7 (9)

a. b.

The student raised his hand, El alumno levanto la mono. the student raised the hand 'The student raised his (lit. the) hand.'

In addition to body parts, the mind and heart (as the seat of emotions as opposed to the physical organ) are also coded as active-zone absolutives: (10)

a.

b.

'Ua tumu lona mäfaufau i le fa'anoanoa. PERF full his mind with the sad 'His mind was filled with sadness.' (Milner 1966: 47) 'Ua tigä lo 'u loto 'i ana 'upu. PERF hurt my heart by his(PL) word was hurt by his remarks.' (lit. My heart hurts because of his words.) (Milner 1966: 264)

Similarly, in telling one's age, the absolutive is the word tausaga 'years' and the predicate is the number of years; the person is merely a possessor of the years (Mosel 1991: 301): (11)

E fä sefiiluo'utausaga. IMP four ten my years T m forty years old.' (lit. My years are forty.)

396

Kenneth William Cook

As shown in (12), the verbs iai 'exist' and leai 'not exist' are used to predicate possession (Mosel 1991: 302). The item possessed is in the absolutive, and the possessor is expressed as a possessive determiner of the absolutive, or it is coded as a locative prepositional phrase. (12)

a.

b.

4.

E leai sa'u penitala. IMP not-exist my pencil Ί don't have a pencil.' (lit. My pencil doesn't exist.) E iai i te Oe le kl? IMP exist LOG you the key 'Do you have the key (with you)?' (lit. Does the key exist at you?)

The coding of transitive agents as possessors instead of ergatives

The verbs in the majority of the previous Samoan examples are intransitive. However, this same possessive-absolutive construction is also possible (and very common) with transitive verbs. In sentences (13a-c), for example, the patient is coded as an absolutive, and the agent as a possessor of the patient. The fact that Samoan favors this type of coding is somewhat surprising given that one would expect agents in an ergative language to be coded as ergatives per se, which of course at times they are. The contrast between the two coding patterns of agents will be dealt with below. (13)

a.

b.

c.

Olo'ogau tolo a tamaiti. PROG chew cane of children The children are chewing sugar cane.' (lit. PRO be chewing the sugar cane of the children.)8 (Milner 1966: 78) Ό Ιο Ό su 'e s 'u g luega. PROG search my work Ί am looking for a job.' (lit. PRO be looking for my job.) (Milner 1966: 75) 5 falle fono a le nu'u. PAST do the council of the village

Samoan as an active zone language

397

'The village held a council.' (lit. PRO held the council of the village.) (Milner 1966: 69) Samoan has an alienable/inalienable contrast in possessive expressions. In (13a-c), the possession is alienable (marked a). Since the alienable category is the dominant one (the one of control), one might think that agents coded as possessors must be expressed only as alienable possessors, but as illustrated in (14a, b), agents are also expressed as inalienable possessors (marked o). As elsewhere, the choice between alienable and inalienable marking is determined by the relationship between the possessor and the possessed item. For example, the possession of one's own (false) teeth in (14a) is inalienable; thus the possessor is marked o. Likewise, possession of a boat falls into the inalienable category in (14b).9 (14)

a.

Onatu'uifo

lea

nifo o Lelei ...

b.

then put down then ABS teeth of Lelei... 'Lelei took out his false teeth (lit. Then PRO put down the teeth of Lelei...) (Milner 1966: 1) 'Ua tu'u o lätouva'ai le matäfaga. PERF leave of them boat on the shore They have left their boats on the shore.' (lit. PRO have left their boats on the shore.) (Milner 1966: 136)

The expression of transitive agents as possessors contrasts with their expression as ergatives. Compare (15a, b), in which the agent is respectively coded as a possessor and as an ergative.10 (15)

a.

Na

fufulule

ta'avalea le tama.

PAST wash the car of the boy The boy washed his car.' The boy's car was washed.' '(Previously mentioned agent) washed the boy's car.' b.

Na fufulu e le tama le ta 'avale. PAST wash ERG the boy the car The boy washed the car.'

398

Kenneth William Cook

With respect to the contrast between the possessive-absolutive and ergative constructions exemplified in (15a, b), Duranti and Ochs (1990) state the following: The use of an Absolutive NP with a genitive focuses on the Object or the result of an action and presents the Agent as not necessarily responsible for the creation or pursuit of the Object, whereas the ergative NP with a canonical [transitive] verb highlights the human participant (Agent) as a willful and responsible actor whose actions may directly affect an object.

Duranti (1990, 1994) explores the use of this and other contrasts in the political arena of the Samoan fono (i.e. the village council). He finds that possessive agents are used in situations in which "the speaker wants to mention and at the same time deemphasize someone's contribution to a given task or achievement ..." (1990: 656), while the ergative agent is used in acts of assigning responsibility, i.e. in accusations and instances of praise. He also claims that the frequency with which a speaker uses ergatives correlates positively with his political weight on the village council (1990: 661). A more powerful member is in a position to do more accusing and praising than is a less powerful member. As pointed out by Mosel (1991: 299) and Mosel and Hovdhaugen (1992: 422), if the agent is simultaneously a possessor of the patient, then the possessive-absolutive construction is most commonly used. Note also that sentence (15a) can be understood in three different ways. If there is no agent supplied in the previous context, then the sentence is understood as either The boy washed his car' or The boy's car was washed.' If an agent is introduced earlier into the discourse, such as by means of a question like (16a) or (16b), then (15a) is understood as an assertion of what that previously mentioned agent did. (16)

a.

b.

le ä le mea na faie Seve? PRED the what the thing PAST do ERG Seve 'What did Seve do?' (lit. What is the thing that Seve did?) le ä le mea a Seve na fai? PRED the what the thing of Seve PAST do 'What did Seve do?' (lit. What is Seve's thing that PRO did?)

Samoan as an active zone language

399

The questions in (16a, b) illustrate that a question like 'What did Seve do?' can be coded with either the ergative construction (16a) or the possessiveabsolutive construction (16b). Notice that (16a, b) also lend themselves to what in Relational Grammar would be called a "possessor ascension" analysis, i.e., the subject (Seve) of a relative clause such as the one in (16a) could be said to "ascend" into the main clause and become the possessor of the head noun of the relative clause, producing a sentence like (16b). While such an analysis will work for pairs like (16a, b), it will not work for sentences like (15a) since this sentence does not contain a relative clause out of which a nominal could ascend.11 Apart from the issue of possessor ascension, sentence (16b) also serves to illustrate that possession in a strict sense of owning an item is not necessary in order for the agent to be coded as a possessor. The questions in (17a, b) are similar to question (16b). However, the relative clause in (17a) is optional and there is no relative clause in (17b). Again, the head noun and its possessor do not depend on the existence of a relative clause out of which the possessor would ascend. Hence a possessor ascension analysis would not be able to account for these sentence types. (17)

a.

b.

le ä lau tala (na fai)? PRED the what your statement past make 'What did you say? (lit. What is your statement (that PRO made)?) le ä lou manatu? PRED the what your thought 'What do you think? (lit. What is your thought?)

Sentences (17a, b) also serve to illustrate the tendency in Samoan to code the product of an activity (rather than the person carrying out that activity) as an absolutive. Duranti (1981: 172-3) observes that possessive phrases can also code benefactives and recipients, or "goals" as he calls them. Duranti's examples in (18a, b) illustrate this point.12 (18)

a.

...fai se kakoupe'epe'e! (benefactive) make some our cream '... make some cream for us all!' (lit. make some of our cream.)

400

Kenneth William Cook

b.

. . . 'avaku lau fagu! (recipient) take your bottle ' . . . [I] am giving you [this] bottle.' '... [You] take the bottle for you[rself].' (lit. Take your bottle.)

Duranti and Ochs (1990) also point out that the idiom fai se to'alua 'be married' (literally 'do a spouse') does not just prefer but requires the possessive-absolutive construction. Sentence (19) illustrates this point. (19)

a.

'{Ja fai sou to'alua? PERF do your spouse 'Are you married?' (lit. Have PRO done your spouse?) b. *'Ua faie Oe se to'alua? PERF do ERG you a spouse (Are you married?)

5. The tendency in Samoan to code events 'from the inside out" The possessive-absolutive construction (with either a transitive or intransitive verb) is different from typical constructions of English (and many other languages) in which the possessor would be coded as subject (as it is in the glosses of many of these sentences). This coding discrepancy between English and Samoan has led Mosel (1991) to say that English is a "personoriented language" and Samoan is an "anti-person oriented language".13 Also with respect to Samoan coding patterns, Brad Shore (1982: 173) has commented that "Samoans commonly talk about actions and feelings as if the body were a decentralized agglomeration of discrete parts, each imbued with its own will." To these observations I would add my own proposal, which is that the tendency to code active zones as absolutives is a manifestation of a general tendency in Samoan to code events "from the inside out", i.e., from the core of the event moving out to the participants which possess the core, act upon it, or receive it as benefactives or recipients. The ergative case marking of Samoan, I would claim, is another manifestation of this tendency. Following Langacker (1991: Chapter 9), I would analyze transitive events as being conceptually layered, and suggest that at the heart of a transitive

Samoan as an active zone language

401

event, we find a minimal core, or thematic relationship, as Langacker calls it. This core consists of a participant such as the door in (20a) undergoing a process or finding itself in a static situation. Transitive event (20b) is more complex than state/process (20a) in that it involves an outer layer of an agent that brings about the inner state or process. (20)

a.

b.

'Ua matala le faitoto'a. PERF open (v.i.) the door 'The door is open/has opened.' (state/process) (Milner 1966: 137) Na tatala e le faiä Oga le faitoto 'a. PAST open (v.t.) ERG the teacher the door The teacher opened the door.' (transitive event)

As Langacker claims, the core of the event is conceptually autonomous in that it is much easier to conceive of a participant in a state or undergoing a process without making reference to an agent than it is to think of a transitive agent acting without making reference to what it is acting upon. The agent and what it does, then, are conceptually dependent on the core relationship. As a result, the conceptual autonomy of the absolutive makes the absolutive (as opposed to an ergative) central to the construal of the event.14 What an ergative language does with respect to case marking, then, is start at the core of the event and move out, marking first the absolutive and then the ergative (if the clause includes one). The absolutive, whose role is determined by the verb, is usually unmarked in ergative languages, while the ergative, which, in some sense, is an agentive add-on in the portrayal of the event, receives special marking. This general description of case marking in ergative languages fits well the case marking facts of Samoan. In this way, ergative case marking, like the coding of active zones as absolutives, is another manifestation of the general tendency in Samoan to code events "from the inside out". The observations that this paper has presented raise at least two questions: (1) What other languages or types of languages show a tendency to code active zones (rather than their possessors) in prominent syntactic positions? and (2) Is the correlation between coding active zones as absolutives and ergative case marking something peculiar to Samoan or does it generally occur in other ergative languages?15 The answers to these questions lie in the domain of future research.

402

Kenneth William Cook

Notes 1. Earlier versions of this paper were presented at the First International Conference on Oceanic Languages, which was held at the University of the South Pacific in Port Vila, Vanuatu in July of 1993, at the Third International Cognitive Linguistics Association Conference in Leuven, Belgium in July of 1993, and at the Seminar on Indigenous Languages conducted by Ricardo Maldonado at the Autonomous National University of Mexico in Mexico City in October of 1993. I thank Alessandro Duranti and three anonymous readers for written comments. The usual disclaimers apply. 2. Langacker (1984) made this observation as part of an analysis of certain grammatical phenomena (e.g., object-to-subject raising) as cases of lexical polysemy; i.e., there are variants of predicates that select different facets of a scene as the subject or object of a clause. For example, as illustrated in (i), the predicate easy has at least two variants; one selects a process as subject and the other selects the patient ofthat process. (Langacker (1993a) maintains this analysis of raising constructions.) (i)

a. b. c.

To fix Hondas is easy. Hondas are easy to fix. Hondas are easy.

This is similar to accounting for David blinked and David's eyelids blinked as manifestations of two variants of blink, one which selects the active zone (the eyelids) as subject and another which selects the possessor of that active zone (cf. (la)). 3. See Ochs (1982) for the sociolinguistic factors that determine whether or not the ergative marker is encoded. 4. Unless otherwise indicated, the orthography of the Samoan sentences is that of Milner (1966). A g represents a velar nasal, and an inverted comma indicates a glottal stop. The abbreviations used in the morpheme-by-morpheme glosses are as follows: ABS: absolutive DIR: directional ERG: ergative IMP: imperfect INF: infinitive LOG: locative

NOM: nominative PERF: perfect PL: plural PRED: predicative PROG: progressive

5. See Ochs (1988: 110-15) and Mosel and Hovdhaugen (1992: 764) for further discussion of the syntactic centrality of the absolutive in Samoan. 6. As can be anticipated, if both the ergative and the absolutive are plural and the ergative precedes the absolutive, ambiguity results:

Samoan as an active zone language (i)

403

Na 'ave 'uma e tamaiti tusl. PAST take all ERG children books 'The children all took books.' or The children took all the books.' 7. Langacker (1993b) presents a reference point analysis of the construction exemplified in (9b). 8. PRO is used here to fill the empty subject slot in the English glosses. 9. See Mosel and Hovdhaugen (1992: 282-90) for more on the alienable-inalienable distinction in Samoan. 10. Fillmore (1968: 14) observes that "a common feature of ergative systems is that the 'genitive' form is the same as the ergative ..." See Anderson (1977: 336ff) for discussion on why this should be the case. 11. Clark (1976: 116-119) discusses "possessor ascension" phenomena in a number of Polynesian languages. For these languages, Clark (1976: 119) finds a special rule of ascension to be unnecessary. 12. Duranti's examples capture the k speech style in which what is normally written t is pronounced [k], and both written g and n are pronounced as a velar nasal. In the t style, g represents a velar nasal and n an alveolar nasal. 13. At the presentation of this paper at the Third International Cognitive Linguistics Association Conference (see footnote 1), Frederick Newmeyer observed that on a continuum from person-oriented to anti-person oriented, English would not be located at the far (person-oriented) end of the continuum since there are other (European) languages that are even more person-oriented. It should also be noted that English does occasionally code an active zone as subject or object. Consider, for example, the idiom "shake a leg" (cf. (8a) above) and the first line of the Battle Hymn of the Republic: "Mine eyes have seen the coming of the glory of the Lord", in which both the subject and object are more active-zone-like than the subject and object of the less poetic "I have seen the Lord coming in all his glory". An anonymous reader has also pointed out that writers of (presumably English) fiction have long known about active zone coding. 14. To illustrate the concept of "conceptual autonomy", Langacker (1991: 246) draws analogies in phonology and morphology. Vowels are autonomous vis-ä-vis consonants in that they are the starting point for constructiong a syllable. Similarly, morphological roots form the core of more complex forms, while affixes are dependent upon those roots. 15. At the presentation of this paper at the First International Conference on Oceanic Linguistics (see footnote 1), Frantisek Lichtenberk commented that the coding of agents as possessors is common among Polynesian languages. My impression, based on limited exposure to Tahitian, Maori and Hawaiian, is that the accusative Polynesian languages, such as the ones just mentioned, do have agents coded as possessors but do not code transitive events (in context-free independent clauses) with the patient in the accusative and the agent (only) in the possessive (cf. the Samoan sentence type exemplified in (15a)).

404

Kenneth William Cook

References Anderson, Stephen R. 1977 On mechanisms by which languages become ergative. In Charles N. Li (ed.), Mechanisms of Syntactic Change. Austin: University of Texas Press, 317363. Clark, Ross 1976 Aspects of Proto-Polynesian syntax. Auckland: Linguistic Society of New Zealand. Cook, Kenneth W. 1991 The search for subject in Samoan. In Robert Blust (ed.), Currents in Pacific Linguistics: Papers in Austronesian Languages and Ethnolinguistics in Honour of George W. Grace. Canberra: The Australian National University, 77-98. Duranti, Alessandro 1981 The Samoan Fono: A Sociolinguistic Study. Pacific Linguistics B-80. Canberra: Australian National University Press. 1990 Politics and grammar: agency in Samoan political discourse. American Ethnologist 17: 646-666. 1994 From Grammar to Politics. Berkeley: University of California Press. Duranti, Alessandro and Elinor Ochs 1983 Word order in Samoan discourse: a conspiracy toward a two-consitutent pattern. Lecture presented at the Discourse Seminar at the Summer Institute of Linguistics, University of California, Los Angeles. 1990 Genitive constructions and agency in Samoan discourse. Studies in Language 14: 1-23. Fillmore, Charles J. 1968 The case for case. In Emmon Bach and Robert T. Harms (eds.), Universals in Linguistic Theory. New York: Holt, Rinehart and Winston, 1-88. Langacker, Ronald W. 1984 Active zones. Berkeley Linguistics Society 10: 172-88. (reprinted in Langacker 1991) 1991 Concept, Image, and Symbol: The Cognitive Basis of Grammar. Berlin: Mouton de Gruyter. 1993a Raising and transparency. Paper presented at the Third International Cognitive Linguistics Association Conference, Leuven, Belgium. 1993b Reference-point constructions. Cognitive Linguistics 4: 1-38. Milner, George B. 1966 Samoan Dictionary. London: Oxford University Press. Mosel, Ulrike 1991 The Samoan construction of reality. In Robert Blust (ed.), Currents in Pacific Linguistics: Papers in Austronesian Languages and Ethnolinguistics in Honour of George W. Grace. Canberra: The Australian National University, 293-303. Mosel, Ulrike and Even Hovdhaugen 1992 Samoan Reference Grammar. Oslo: Scandinavian University Press.

Samoan as an active zone language Ochs, Elinor 1982 1988 Shore, Bradd 1982

405

Ergativity and word order in Samoan child language. Language 58: 646-671. Culture and Language Development: Language Acquisition and Language Socialization in a Samoan Village. Cambridge: Cambridge University Press. Sala'ilua: A Samoan Mystery. New York: Columbia University Press.

Two transitive construction frames in Spanish: The prepositional and the non-prepositional accusative Nicole Delbecque

In this short presentation it is impossible to go into all the similarities and dissimilarities between the prepositional accusative and the other constructions which are characterized by the same preposition, viz., the directional construction and its dative counterpart. Given the space allowed, this paper only considers the conceptual distinction between the prepositional and the bare accusatives. Although it is not possible, either, to examine previous analyses in detail, it is worthwhile to recall that the norm, as put forth by the Spanish Academy (R.A.E. 1973: 345), simply stipulates that the direct object is introduced by the preposition a when it designates an individuated person; so-called "exceptions" to the rule are allowed in both directions: there are uses where the preposition unexpectedly appears and, on the other hand, there are cases in which the preposition does not appear, though it normally would be expected. In fact, in a modern-essay corpus, 86 of 335 prepositional accusatives have an inanimate referent. This high number of cases which contradict the rule casts doubt on the relevance of the traditional criterion. Instead of relating the prepositional accusative to semantic properties of the direct object, viz. [animate] and [individuated], or invoking an analogy with the nominative and with the dative, I propose that Spanish has developed two transitive construction frames, one prepositional, the other not, each of which has its own semantics. The point of departure for the present study is that the meaning of a construction does not only result from the interaction between the semantics of the verb and that of the event participants, but that it also, most crucially, hinges upon the meaning imposed by the construction frame as a holistic cognitive unit. This principle of cognitive grammar, according to which abstract meaning is integrated in syntactic constructions, leads us to believe that the presence vs. absence of the preposition a cannot be satisfactorily explained in terms of determinisms which operate at the level of the direct object complement. I suggest that these two constructions yield a

408

Nicole Delbecque

different conceptualization of the global argumentative structure. In the non-prepositional configuration, the accusative is plainly integrated in the relational predication, instantiated by the VP, whereas the prepositional configuration signals the relevance of a proper reference domain, which is not bounded by the verbal predication. To demonstrate the grammaticization of this schematic opposition, let us begin by examining constructions which preclude the insertion of the preposition before the direct object. The question arises whether the obligatory absence of a, i.e., the "low" profile, coincides with the fact that the internal scope of the predication imposes a limit to the number of interconnected entities. We, then, shall turn to the more complex conceptualization of the prepositional construction, i.e., the "high" profile, in order to verify whether it, indeed, stands in opposition to the bare accusative. The first question to raise concerns the impossibility of using the preposition a in some cases. Up to what point can we say that the exclusion of the prepositional accusative is due to the fact that the predicational relation is semantically incompatible with the "high" profile transitive frame? The answer seems to be that this is true to the extent that the direct object is not grounded outside the verbal predication, i.e., insofar as Nj (the direct object) can only be conceived of as fully and exclusively integrated within the verbal predication. This observation applies to the presentative quantifier hay 'there is/are', which is, at the same time, incompatible with the preposition a and with the clitic le. The presentative function of the predication makes it impossible to signal, at the same time, the existence of an independent reference domain for the N j entity. Take, for instance, sentence (1): (1)

Hay aqui (*a) tres hombres que quieren verte. 'Here are three men who want to see you.'

The same restriction holds for other verbs, in particular for perception verbs, when they function as presentatives, e.g. (2): (2)

Solo veo / conozco un hombre capaz de componer tal obra. Ί only see / know of one man able to compose such a work.'

The preposition cannot be used, either, when the direct object possesses prepositional content, viz., when it refers to a third order entity in the sense

Two transitive construction frames in Spanish

409

of Lyons (action, process, situation). The temporal character of such a direct object (implicit in infinitives and deverbal nouns) renders it uncategorizable; since it is impossible for it to apply to a categorizable cognitive domain, the use of the preposition is excluded. Notice that deverbal nouns allow the preposition when they are construed as non-processual; compare, for example, (3) and (4): (3)

A esta situation [...] me he enfrentado [...], atormentado por una sensation de malestar tan viva e irremediable como la que acompano en mijuventud el descubrimiento de la realidad del pah. This situation I have confronted, tormented by a sensation of uneasiness as vivid and uncurable as the one which accompanied in my youth the discovery of the reality of the country.' (J. Goytisolo)

(4)

[...] los temerosos que solo miran a la conveniencia de no concitarse la enemistadde las autoridades [...] 'the cowards who only look at the convenience of not exciting the animosity of the authorities' (R. Gallegos)

This observation can be stated in more general terms: the use of a is impossible when a continuous, i.e., a non-discrete, kind of reference is established, as in (5). Both in example (5) and in (6), the quantified nominal Nj expression represents a complement type traditionally called an "internal" or "cognate" object. It precludes the use of the preposition, no matter how definite its form is, since, conceptually, the nominal is part of the verbal predication; by this I mean the direct object is integrated in the verbal concept. (5)

Maria bebe coda dia dos litros de agua. 'Maria drinks two liters water every day.'

(6)

Padece el SIDA en su estado incipiente. 'He suffers from AIDS in its incipient form.' (El Pais)

The systematic inclusion of some nouns in the verbal domain gives rise to compound verbal predicates, e.g., tener/tomar conciencia 'to have/take conscience', tener ganas 'to feel like', tener sentido 'to make sense', convocar plaza 'to advertise a job', sacar partido/beneficio 'to take advantage', sacar conclusiones 'to draw conclusions'. The preselected noun can be accompanied by determiners and other modifiers, e.g., (7):

410

Nicole Delbecque

Domain NO

Figure I. POSSESSION: non-prepositional construction

Domain NO

Domain N2

Figure 2. TRANSFER: ditransitive non-prepositional construction

(7)

Cumple lafuncion de director general. 'He fulfils the function of general director.'

The productivity of the verb tener 'to have' is noteworthy in this respect: the possessed entity (Nj) is denied any conceptualization outside the predicational relation, since the latter asserts its inclusion in the N0, i.e. in the subject domain, as shown in Figure 1. The same applies to predicates which express a form of TRANSFER (material or communicative transfer) by means of the [N0 V N^ a N2] construction, e.g. (8): the object (Nj), transferred from N0 to N2, is conceived of as starting off in the NQ domain (see Figure 2). This conceptualization of N j does not allow the interpretation of the prepositional accusative frame. (8)

Comunico la noticia a sus colegas. 'He communicated the news to his colleagues.'

Many constructions present the split expression of a verbal concept in V and Nj, in which Nj specifies the object implied by the verbal predicate. The direct object thus acts as an internally bounded domain, though not in the sense that the N] entity would constitute the last element of the verbal

Two transitive construction frames in Spanish

411

process: in (9), for instance, it is not the case that the subject referent first reads and then goes on reading till he ends up reading the newspaper; the concept of 'newspaper' is connected with the reading process from the start on: it defines the domain of the verbal scope, by specifying a subset (in this example 'newspaper' vs 'book' or 'journal', for instance). (9)

lee el periodico 'he reads the newspaper'

(10)

escribe una carta 'he writes a letter'

Instead of being affected, as in (9), the direct object entity may be effected, as in (10). The distinction between affected and effected entities, related to the difference between imperfective and perfective processes, has to do with a subdivision of N j argument roles. It does not, however, interfere with the delimitative function of the bare accusative as such: at any rate, Nj functions as paradigmatic limit of the extension of the verbal process. The affected/effected distinction does not seem to be pertinent for the nonprepositional frame. The prepositional frame, on the other hand, only takes affected entities as direct object. The verbs which can be construed with an affected non-prepositional Nj can also be constructed prepositionally; the reverse, however, does not hold: in my data, a number of verbs constructed with prepositional accusative are never constructed otherwise, e.g., alegrar 'to gladden', aliviar 'to alleviate', asustar 'to frighten', cansar 'to tire', deslumbrar 'to dazzle', elogiar 'to praise', enfadar 'to make angry\fatigar 'to tire', felicitar 'to congratulate', halagar 'to flatter', irritar 'to irritate', molestar 'to annoy', reconfortar 'to comfort'. This fact probably explains why traditional grammar considers the prepositional accusative as a property of individuated animate referents. There undoubtedly exists a statistical tendency in this direction. In fact, this only indicates that animation and individuation combine more easily with the "high", marked transitive profile and, conversely, that it is more difficult to make them compatible with the low transitive profile. Without negating the symptomatic value of these properties, the present analysis aims to disclose the divergent underlying conceptual orientations responsible for this state of affairs. Example (11), with inanimate and generic prepositional accusatives, invites us to examine the global conceptualization of the predicational relation.

412

(11)

Nicole Delbecque

Hermogenes, al reducir el ser al aparecer, degüella a la verdad en la cuna; como contrapartida, el admitir como unica proposition posible la que formula el hombre por siy ante si, hace verdadero [...] tanto a lo que es verdad [...] como a lo que no lo es. 'Hermogenes, upon reducing being to appearance, decapitates the truth in the cradle; in contrast, admitting as the only possible proposition the one which man formulates for himself and in regard to himself, makes truthful [...] both what is true [...] and what is not.' (CJ. Cela)

Up till now we have seen that the impossibility of using the prepositional accusative with certain verbs is due to the fact that the entity designated by the direct object depends exclusively on the domain instantiated by the predicational axis, viz. subject NP - VP. Now we go on to consider the prepositional accusative. The prepositional accusative indicates that the direct object represents more than just a domain bounded by the propositional structure: it signals the relevance for the Nj entity of a domain situated at discourse level, i.e., in the mental space which includes the entities and relations supposedly shared by the speaker and the hearer (as the communicative base at the moment of speech). From this situation emerges a scene which is more complex than the one instantiated by the non-prepositional construction. My proposal partially continues the line of thought defended by Granville Hatcher (1942: 42), who speaks about the "autonomization" of the direct object with respect to the predicate. It substantially differs from this view, however, in that I advocate a holistic approach in which the grammar is conceived in terms of construction frames which yield distinct conceptual representations (cf. Goldberg 1992). If it is true that the use of the preposition is not primarily related to the properties of Nj but instead derives from the global conceptual image, then it should be possible to detect syntactic clues which corroborate this interpretation. The high frequency of the prepositional accusative in constructions with double predications deserves special attention: the expansion on the primary clausal predication which is interpreted as a second predication evokes the domain with which the affected entity gets associated. The complexity of the construction has repercussions at the level of the Nj profile. Following Langacker, in the predicational relation instantiated by the VP, Nj constitutes the landmark (LM), designating a participant which is less salient than N0 (the grammatical subject), i.e., the trajector (TR). The second predication instanti-

Two transitive construction frames in Spanish

413

Domain N 1

Domain NQ

TR

ο

1m

LM

Figure 3. Complex prepositional accusative frame ates another domain: in this domain, Nj functions as the most salient participant (the trajector in lower case: tr), with respect to another entity (or entities) which relates to it as landmark (1m, also in lower case). The secondary predicational relation can be realized by means of a subordinate relative clause or an infinitive; for instance, vi caer al avion Ί saw the plane crash'. However, the presence of a verb is not an absolute requisite for having a second predication, since the exact interconnection between Nj as a trajector (tr) and its landmark (1m) is not crucial for the domain of the primary TR-LM profile (in capitals), which is expressed by the primary construction. Needless to say, not all modifiers are susceptible to triggering a second(ary) domain. Take, for instance, the relative clause which takes the subjunctive; for example: Busco una secretaria que sepa ingles Ί am looking for a secretary who knows English'. This means that the direct object (Nj), as a reference point for the grammatical subject (N0), may receive more ample grounding from another entity which need not necessarily refer to a conceptual domain which would not be bounded by the primary predication. However, any modification which instantiates a true secondary predication, as, for instance, in (12a), implies projection on a second conceptual domain and, thus, differs fundamentally from a modification which does not. The kind of conceptual reduplication which corresponds to the prepositional accusative can be visualized by means of Figure 3, which stands in opposition to Figure 4. With certain verbal concepts, the use of the prepositional accusative can be related to the fact that they yield a secondary predication. Here, we limit

414

Nicole Delbecque

Domain NQ

V

TR

LM

Figure 4. Complex non-prepositional accusative frame

the illustration to the notions of (1) INCLUSION, (2) SUBSTITUTION, (3) SEARCH, (4) PERCEPTION, (5) NAMING, (6) CAUSATIVITY. (1) INCLUSION. The verb tener 'to have' does not only fit the above mentioned construction, it can also be used in a construction frame characterized by a second predication. The same can be said of other verbs which express a form of INCLUSION: e.g., abarcar 'to include', captar 'to capt', conservar 'to conserve', guardar 'to guard', incluir 'to include', poseer 'to possess'. Explicit inclusion of Nj in an appropriate reference domain triggers the use of the preposition. The inclusion can be spatial, as in (12a) or metaphorical as in (12b, c). A second predication in the form of a relative clause (in the indicative), or a prepositional phrase, confers an abstract character to the inclusion relationship: it signals a sort of equivalence, e.g. (12b, c). (12)

a. b. c.

tiene a su madre cerca / consigo / asu lado 'he has his mother nearby / with him / next to him' tengo alpresidente por un hombre honrado Ί take the president for an honorable man' estos estados tienen al espanol como idioma oficial 'these states have Spanish as official language'

Compare the above with (13): (13)

tiene el pelo rojo 'he has red hair' (lit.: he has the hair red)

The difference can also be appreciated by comparing (14a) and (14b):

Two transitive construction frames in Spanish

(14)

a. b.

415

C *Este mes) tiene un hermano enfermo I en el extranjero '(This month) he has a brother ill / abroad' Este mes tiene a un hermano enfermo / en el extranjero This month he has a brother ill / abroad'

The coextensiveness of Nj and N0 (due to the relational character of Nj) explains the impossibility of adjoining a temporal delimitation to the (14a) sentence; the necessity of relating such a delimitation with the predication of N} is incompatible with the non-prepositional frame; it is, however, congruent with the prepositional frame, e.g., (14b). Contrary to what is generally believed, the prepositional accusative need not have a referential value, in the sense that it can just instantiate an attribute or quality of an entity, since this attribute or quality is, by definition, grounded in a "secondary" domain. The referentiality of N j is not a property of the prepositional accusative but rather it depends on the concept predicated by the verb. When the latter entails a dynamic interpretation, it implies an agentive grammatical subject: N0 brings about a change of state in Nj, making it enter the domain of N0, e.g. (15). This implies access to a referential entity. On the other hand, the static concept of possession does not exclude nonreferentiality of the direct object: for instance, in (16), the grounding of the non-referential Nj, a un maestro 'a teacher' surfaces in the oblique en el 'in him'. (15)

La ventaja de esta conception "totalizadora" de la literatura sobre la vision sociologista de Rama esta en que aquella abarca tambien a esta ultima, aunque despojandola. The advantage of this "globalizing" conception of the literature over the sociologist vision of Rama is that the former also includes the latter, although stripping it.' (Vargas Llosa)

(16)

Υ hasta para divertirse, tuvo en el a un maestro. 'And even to amuse himself, he had in him a teacher.' (quotation from an Argentine novel)

As for negation, in the majority of the cases, it has N] in its scope and N j is not grounded in any other domain outside of the one evoked by the N0-Nj relationship, e.g. (17). However, when N j further remains unspecified, there still remains the possibility of an all-encompassing reference: in (18), the verbal predication signals, on the one hand, the non-existence of the N0-Nj

416

Nicole Delbecque

relationship and, on the other, it leaves room for multiple Nj domains outside this relation ("en el mundo" 'in the world'). (17)

El coronel no tiene quien le escriba. The colonel does not have anybody who writes to him (to write to him).' (Garcia Märquez)

(18)

No tiene a nadie (en el mundo). 'He does not have anybody (in the world).'

(2) SUBSTITUTION. The verbs reemplazar 'to replace', su(b)stituir 'to substitute', suplir 'to replace', express the substitution of one entity for another. Mathematically, the verbal predicate situates one entity [Y] in a domain so as to occupy the pace of another entity [X]. Such a configuration is expressed by means of a two- or three-argument frame, as shown in (19) and (20), respectively. The three-argument frame imposes a dynamic reading, while the other does not. In both, the object of the substitution, the substituted entity [X], appears as Nj; it is prepositional in (19), whereas in (20) it is not. Whatever the chosen domain, there always exists a precedence relationship of entity [X] over entity [Y], but it is conceptualized in a different way. In the two-argument frame, the substituting entity [Y] is profiled as the primary clausal figure (N0), and entity [X] as the secondary clausal figure (Nj). Conceptually, the substitution of the latter for the former presupposes a domain for Nj. This is precisely the image yielded by the prepositional accusative: elaboration by a second predication is possible (the italicized part of (21)). The three-argument structure presents another image: the role of entity [Y] is roughly that of means or instrument: it is relegated to an oblique position. The primary figure is now assigned to an agent (N0), so that the relationship between [X] and [Y] becomes dependent upon the action that emanates from the agent. This means that the predication subordinates the [X]-[Y] relationship, here Nj-N 2 , to the scope of N0. This conceptualization precludes the use of the prepositional accusative, as in (22). (19)

[entity Y] N0 + [SUBSTITUTION] V + a + [entity X] N! (+ second predication)

(20)

[agent] N0 + [SUBSTITUTION] V + [entity X] N! + por/con + [entity Y] N2

(21)

[...] una realidad "ersatz" que sustituyepuntualmente a la realidad de la vida que vivimos.

Two transitive construction frames in Spanish

417

'an "Ersatz" reality which punctually substitutes the reality of the life we are living' (Martinez Estrada) (22)

[...] querer sustituir "totalmente" la realidad con ficciones "totales" [...] 'we want to substitute reality "totally" with "total" fictions' (Vargas Llosa)

(3) SEARCH. The notion of searching is expressed by verbs such as buscar 'to look for', necesitar 'to need', querer 'to want', desear 'to desire', preferir 'to prefer', elegir 'to elect'. By grounding Nj in a sort of "attribute" which acts as second predication, the notion of searching can be conceived in two ways: retrospectively or prospectively. With the prospective meaning, the entity designated by N t has propositional content and, thus, cannot be prepositional. Kleiber (1981: 293), in an analysis of French, calls this a "virtual" direct object, in order to reflect the fact that it does not have truth value at the moment t of speech, although it may end up having one at a moment t+i, in a "possible world", e.g. (23). The subjunctive mood selection in the relative clause corroborates the prospective reading. The prepositional accusative, on the other hand, triggers the indicative, e.g. (24); it would indeed, be contradictory to indicate that Nj does not simply emerge from the verbal predication without at least recognizing that a relevant reference domain is operative for Nj outside the transitive construction proper. (23)

Juan {busca / necesita / quiere}{-/ unos / los} amigos (que sean) fiables 'Juan {looks for / needs / wants}{- / some / the} friends (which would be) liable'

(24)

Juan {busca /necesita /quiere} a {-/unos /los} amigos (que son) fiables 'Juan {looks for / needs / wants} ("a") {-/ some / the} friends (who are) liable'

(4) PERCEPTION. In the construction known as "accusativus cum infinitive", Nj, the direct object of the perception verb, functions as the subject of the second predication, usually an infinitive or a gerund, e.g. (25). With verbs which express mental perception, such as conocer 'to know', considerar 'to consider', estimar 'to estimate', apreciar 'to appreciate' desdenar 'to disdain', menospreciar 'to despise', the second predication takes another form;

418

Nicole Delbecque

the result is a static perspective instead of a dynamic one, e.g. (26). Again, the prepositional accusative applies also to inanimate nominals, provided the construction attributes a proper domain to the N| entity. (25)

Ese sanguineo valenciano al que vemos desafiando bravuconamente a sus contrincantes [...] 'This warm-blooded Valencian man whom we see challenging his opponents with bravoure.' (Vargas Llosa)

(26)

A los temas netamente literarios [...] nosotros los hemos considerado como elementos constituyentes de una nation [...] 'The clearly literary themes, we have considered them as elements constituting a nation.' (Martinez Estrada)

(5) NAMING. Construed prepositionally, verbs such as calificar 'to qualify', caracterizar 'to characterize', denominar 'to denominate', especificar 'to specify', explicar 'to explain', tratar 'to treat', tachar 'to qualify', express the denomination of Nj by N0, like the prototypical llamar 'to call'. The proper concept of name-giving presupposes the conceptualization of a relevant domain: Nj is more than the simple patient of the name-giving or evaluating activity. The change which affects Nj may be characterized in terms of transition from a state of implicit identity to that of explicit identification. The second predication enhances the discourse salience of Nj. The prepositional frame is used, with no matter which referent: (27) is an example with inanimate Nj, while (28) contains a indefinite Nj: (27)

[...] y a esto le Human transigir (Unamuno) 'and this they call compromise'

(28)

El caso es que mi nuevo consultante me pregunta si al llamarle a algo "secundario" [ . . . ] 'The fact is that my new consultor is asking me if by calling something "secondary" [...]' (Unamuno)

(29)

Tambien al lenguaje que aqui llamo cratiliano alude Max Scheler [...] 'Max Scheler alludes also to the language which I here call Cratilian' (Cela)

When the non-prepositional frame is used, e.g. (29), llamar 'to call' receives a presentative interpretation (cf. supra).

Two transitive construction frames in Spanish

1—

Domain

Vi

V2/t

Nj/t-i

419

Domain Nj/t+i

Figure 5. Complex DATIVE causative frame Domain

Domain

Nj/t+i

Nj/t-i

Vi

V 2 /t

Figwre 6. Complex ACCUSATIVE causative frame

(6) CAUSATIVITY. The infinitive complement (V2) of the causative verb (Vj) can be considered as a form of second predication. Depending on V j (llevar 'to bring' vs. hacer 'to make'), V2 will be prepositional or not. This difference entails conceptual differences which are beyond the scope of this paper. The causative verb marks N0 as the instigator (prototypically but not necessarily animate), responsible for the involvement of N j in the action expressed by V2. Nj, in turn, instantiates the role of patient of V j and a role which can go from agent to experiencer or patient of V2. Affected by a change of state due to Vj, N j appears to be the most salient participant in this complex event structure. When V2 is introduced by the preposition a, it is conceived of as metaphorical extension of a goal (cf. the directional frame), e.g. (30). (30)

a. b.

Su gestion ha llevado a lafamilia a la ruina 'His management has led the family to the ruin' Su gestion ha llevado a lafamilia a arruinarse 'His management has led the family to ruin itself

The function of N| seems to vacillate between the accusative and the dative, with a shade of indetermination between them. The difference is presented in Figures 5 and 6. With verbs of movement, such as llevar 'to lead', traer 'to bring', con-

420

Nicole Delbecque

ducir 'to conduct', empujar 'to push', poner 'to put', cliticization of Nj can be done both by means of le as by means of Ιο/la. This flexibility can probably be explained by the fact that these verbal concepts convey the lightest presuppositional load with respect to the degree of intervention on part of the N! referent in the double event structure, i.e., with respect to both the Vj and the V2 predication. The use of le leaves the option between both conceptualizations open: (31) can be interpreted both according to Figure 5 and to Figure 6; (32) however, imposes the reading depicted in Figure 6. (31)

[una mcion] que, no obstante, le[=la disciplina] ha llevado a repartir patentes y ejercer veto s [...] '[a function] which, nevertheless, has led it [=the discipline] to distribute licences and to exercise vetoes [...]' (Cela)

(32)

[S. Dali] empuno el fiero lean del superrealismo por las oniricas melenas, y lo trajo a rendirse sumisamente, como perrito de estrada, a los pies del ramo de perfumeria, cosmeticos [ . . . ] '[S. Dali] grasped the fierce lion of surrealism by the oniric mane, and brought it to surrender submissively, like a show dog, at the feet of the branch of perfumery, cosmetics [...]' (Salinas)

On the other hand, the verbal concepts expressed by hacer 'to make', dejar 'to let', permitir 'to permit', impedir 'to prevent', prohibir 'to prohibit', are used to direct one's attention towards a volitional referent; therefore, the clitic which corresponds to Nj is le. This tendency to instantiate Figure 5 coincides with the fact that the construction does not become agrammatical when Nj is omitted, e.g. (33). (33)

Un comentario del "Pueblo " [...] le incita a Alfonso Paso a hablar de la censura [...] Ά comment from "Pueblo" [...] incites Alfonso Paso to speak about the censorship' (Sanchez Ferlosio)

The causative predication sometimes gets instantiated with a second predication which is not verbal at all. In this type of construction, with verbs such as hacer 'to make', dejar 'to \et\poner 'to put', volver 'to turn', the use of the prepositional accusative seems to be generalized, e.g. (34). (34)

Ha vuelto loco a su entorno. 'He has driven his surroundings crazy.'

Two transitive construction frames in Spanish

421

This analysis does not only apply to complex constructions as the above. Its validity can also be demonstrated for simple constructions, i.e., those which lack a second predication. The prepositional frame signals the relevance of a reference domain proper for Nj, which is not bounded to the verbal predication. The presence of the preposition profiles the relationship between NO and N! as one of "duality", as opposed to their "unitary" character when the preposition is lacking. Among the transitive prepositional constructions which reflect this conceptualization, let us just mention the following types of relationships: (1) N0 causes a change of state in Nj; (2) N0 exerts a constructive or destructive action on Nj; (3) the N0-N] relation is hierarchical: next to the spatial or symbolic ordering, we also find inclusion and exclusion relationships; (4) the relationship expresses material or communicative transfer oriented towards a goal (Nj); (5) the goal-oriented perceptual relationship also gets instantiated by means of the prepositional frame. For the verbs which allow both transitive constructions, the choice of the prepositional frame or of the non-prepositional one obeys the same principle as the one we have just described. With some verbs, the choice yields semantic glidings (e.g., perder 'to lose' vs. perder a 'to ruin'). Generally, it can be observed that the prepositional accusative occupies a more salient thematic position in discourse. The existence of the irregular se-passive construction of the [se (leAo) V a Nj] type, as opposed to the regular [se V N0], can also be explained along the same lines, e.g. (35). These aspects of the question unfortunately exceed the scope of this paper. (35)

[...] se estudia con algo mas de interes la mistica, pero es principalmente por razones religiosas, y se la quiere explicar, a ella, cumbre de selection, como unfenomeno de caracter popular, [se (le/lo) V a Nil 'mysticism is studied with somewhat more interest, but it is principally for religious reasons, and it is being explained, mysticism, creme de la creme, as a phenomenon of a popular character.' (D. Alonso)

For now, I hope to have demonstrated that the use of the prepositional accusative cannot be satisfactorily clarified by means of the syntactico-semantic category of the direct object nor by the type of verb. In its different manifestations, i.e., in complex constructions as well as in simple ones, respectively with and without a second predication, and in the sui generis se-passive, the

422

Nicole Delbecque

prepositional accusative signals that the relationship the verbal predication establishes between N0 and Nj does not prevent the latter from being grounded in a cognitive domain which is not bound to N0 and the VP. The condition which stipulates that N j must be grounded outside the Ng-Nj relationship in order for the prepositional frame to apply, accounts, on the one hand, for the occurrences of specific animate nominals as bare accusatives, and, on the other, for occurrences of the prepositional accusative with non-referential animate nominals and with inanimate ones, indefinite as well as definite. In the field of Spanish syntax, there is a large gap between the norm and actual use. The predominance of the normative conscience over observation of the facts does not facilitate a revision of the norm. As a matter of fact, there is a form of blindness which leads native speakers, even language professionals, to come up as defenders of the norm. The attitude which, a priori, condemns all that does not conform to the "rules" prevents them from asking themselves where the so-called "deviations" they blame possibly come from. The letter of complaint, which (36) reproduces as a fragment, is written by a teacher of Spanish (as a mother tongue) and directed to the ombudsman of El Pais. It is just one example, out of many, of the lack of adequacy between the scholastic normative attitude and linguistic competence as reflected not only in journalistic texts but also in the writings of acknowledged, prestigious authors. It is about time to remedy this discrepancy. (36)

"En Malcolm X, Lee propone a la enigmätica y fortisima personalidad de este lider revolucionario como una de las claves [...]". Sobra la preposicion a, porque "la enigmätica ..." es complemento directo. El mismo error, senala Ballesta [la autora de la carta, profesora de lengua espanola], se repite mäs adelante cuando se escribe: "Considere a esa salvajada [...]" ' "In Malcolm X, Lee proposes ("a") the enigmatic and very strong personality of this revolutionary leader as one of the keys [...]". The preposition a is not needed, because "the enigmatic ..." is a direct complement. The first error, points out Ballesta [the author of the letter, a Spanish language teacher], is repeated further on when one finds written: "I considered ["a"] this savagery [...]"' (El Pais 14 de marzo de 1993)

Two transitive construction frames in Spanish

423

References Bello, Andrei 1847 Gramatica de la lengua castellana destinada al uso de los americanos. Edicion crftica de Ramon Trujillo. Tenerife. Cano Aguilar, Rafael 1981 Estructuras sintacticas transitivas en el espanol actual. Madrid: Gredos. Comrie, Bernard 1978 Aspect. Cambridge: Cambridge University Press. Cuervo, Rufino Jose 1886 Diccionario de construction y regimen de la lengua castellana. Tomo I. Institute Caro y Cuervo 1953. 1893 Diccionario de construction y regimen de la lengua castellana. Tomo II. Instituto Caro y Cuervo 1954. De Kock, Josse 1992 Entre corpus et grammaire normative: le regime direct prepositionnel en espagnol. I.T.L. Review of Applied Linguistics 95-96, 49-89. De Kock, Josse, Carmen Gomez Molina and Nicole Delbecque 1990 Gramatica espanola: Ensenanza e investigation. ILL Gramatica didactica. Salamanca: Ediciones Universidad de Salamanca. 1991 Gramatica espanola: Ensenanza e investigation. HI. Textos. Salamanca: Ediciones Universidad de Salamanca. Delbecque, Nicole 1994 The Spanish prepositional accusative. A matter of frame semantics. Departement Linguistiek K.U. Leuven. Preprint nr. 151 (108 biz.). 1999 La transitivite en espagnol: deux constructions plutöt qu'une. Verbum XXI 1: 49-65. Delbecque, Nicole and B6atrice Lamiroy 1992 The Spanish dative: a problem of delimitation. Leuvense Bijdragen 81, 1-3: 113-161. 1996 Towards a typology of the Spanish dative. In W. Van Belle and W. Van Langendonck (eds.), The Dative. Descriptive Studies. Amsterdam: J. Benjamins. 73-117. Fernandez Ramirez, Sanchez 1986 Gramatica espanola. 4. El verbo y la oration. Edicion preparada por Ignacio Bosque. Madrid: Arco Libros. Garcia, Erica 1975 The Role of the Theory in Linguistic Analysis: The Spanish Pronoun System. Amsterdam: North-Holland. Hatcher, Anne Granville 1942 The Use of "a" as a Designation of the Personal Accusative in Spanish. Modern Language Notes 57, 421-429. Kleiber, Georges 1981 Verbes virtuels et propositions relatives: specificite et non specificit6. Travaux de Linguistique et de Literature XIX. 1: 293-311.

424

Nicole Delbecque

Lyons, John 1977 Semantics. Cambridge: Cambridge University Press. Molho, Maurice 1980 Sur la grammaire de l'objet en espagnol. Travaux de Linguistique et de LitteratureXVllll: 213-225. Real Academia Espaüola. Comision de Gramätica 1973 Esbozo de una nueva Gramätica de la lengua espanola. Madrid: EspasaCalpe. Roegiest, Eugeen 1990 La tipologia sintäctica del objeto transitive en espanol. Verba 17: 239-248. Salvä, Vicente 1847 Gramätica de la Lengua Castellana. Madrid: Arco/Libros 1988.

Swedish abstract transitional phrases: An in-between phenomenon in the linguistic system Lena Ekberg

1.

Introduction

As a linguistic phenomenon, Swedish complex predicates of the type illustrated in (1) are a clear-cut support for one of the basic assumptions of Cognitive Grammar, namely, that traditional linguistic levels such as syntax, lexicon, and morphology rather than being absolutely distinct levels are nondiscrete categories with focal points and overlapping boundaries (see Langacker 1987).1 The Swedish complex predicates in question are called abstract transitional phrases (or ATPs), as they denote a transition from one abstract "space" into another, i.e. a transition into a state or an activity:2 (1)

a. b. c.

d.

fallai sömn fall into sleep3 'fall asleep' gä till an/all go to attack 'attack' komma till insikt come to knowledge 'realize' raka i panik happen-to-come into panic 'panic'

This paper is based on a larger study (Ekberg 1989), in which I examine the semantic, syntactic, and lexical status of Swedish ATPs, as well as the function of ATPs in relation to lexically equivalent simple predicates. The study is confined to phrases which denote transitions between specifically human states or activities (cf. (1)). It is furthermore restricted to phrases containing the verbs falla 'fall', gä 'go', komma 'come', raka 'happen-tocome' (which are the most frequent verbs in ATPs), in combination with the prepositions / 'into', pa On', till 'to', ur Out of. In the present paper I will

426

Lena Ekberg

concentrate on ATPs containing the preposition 'into' and till 'to', as these prepositions are by far the most frequent ones in ATPs. ATPs constitute an in-between phenomenon not merely as regards their syntactic-semantic properties but also as regards their formation. One of the aims of this paper is thus to show that syntactic and semantic regularity is not inconsistent with lexicalization. Although ATPs are syntactically regular in formation as well as semantically fully compositional, they are nevertheless to a certain degree lexicalized (i.e., they can be regarded as idioms). Further, as regards formation (and meaning), ATPs simultaneously behave like syntactically productive phrases and like derived words. Syntactically these constructions are fully productive; in practice, however, there are gaps in the formation. In the first place, these gaps seem to be due to the specific semantic content of the "derivational element" of an ATP, i.e. the combination of verb and preposition. I will however argue that also prototypicality plays a role in the formation of these syntactically productive phrases. My data show that prototypical ATPs, at least partly, "govern" the formation of ATPs. Thus, in the second place, there are gaps in the formation due to the nonexistence of prototypical ATPs. ATPs are verbal periphrases with a clear aktionsart function. They constitute a systematic way of changing the aktionsart of a nontransitional simple predicate (cf. arbeta 'work' with komma i arbete 'come into work'), as well as a way of emphasizing a specific aktionsart already expressed by a synonymous predicate (cf. somna 'fall asleep' with/a//a i sömn 'fall into sleep', where the latter expression emphasizes the momentaneousness of the event "to fall asleep"). As argued by Quesada 1994 verbal periphrases expressing aktionsart typically belong to the intermediate component in the linguistic system, which he claims to be a third focal point intermediate between grammar and lexicon (see also Mathews 1990: 54f.). The form and function of elements belonging to the intermediate component are more systematic than those of purely lexical elements, but less systematic than those of purely grammatical elements. Further, elements belonging to the intermediate component are characterized by a more abstract meaning than purely lexical elements, but have a more concrete, or specific, meaning than purely grammatical elements (cf. Bybee and Dahl 1989 and Talmy 1988 as regards lexical vs. grammatical meaning). In the subsequent sections I will elaborate the intermediate status of ATPs, arguing that ATPs are in-between syntactic constructions and lexical units (sect. 2), in-between syntactic constructions

Swedish abstract transitional phrases

427

and idioms (sect. 3), and, finally, in-between syntactic constructions and derived words (sect. 4).

2.

ATPs as in-between syntactic constructions and lexical units

Semantically, ATPs are word combinations with a unitary verbal meaning, i.e., they refer to a single event, although the semantic representation is complex. The fact that ATPs (often) can replace (or be replaced by) a simple verb (or a simple verb construction) can be taken as an indication that these verb phrases are lexical units; cf. (2). In Langacker's terms, ATPs and the corresponding simple verb constructions illustrate alternate paths of composition (cf. Langacker 1987: 476; 1991: 370). (2)

a.

b.

folia i sömn: somna fall into sleep : sleep + INGRESSIVE 'fall asleep' : 'fall asleep' gä till an/all : (börja) attackera go to attack : (begin to) attack 'attack' : 'attack'

Being lexical units, ATPs can be coordinated with simple, transitive verbs: (3)

Armen gick till an/all mot och besegrade fienden. the-army went to attack against and defeated the-enemy 'The army attacked and defeated the enemy.'

Further, like words, ATPs are anaphorical islands,4 i.e., parts of the ATP cannot be referred to by pronouns, and pronouns cannot occur inside an ATP, cf. (4). Thus lexical units, whether simple words or phrases, obviously cannot contain specific references. (4)

a. *Honföll i sömn som vor djup och behaglig. she fell into sleep which was deep and pleasant b. * Forst foil han i sömn en gang, sedan foil han i den en first fell he into sleep one time, then fell he in it one gang till time more

There are, consequently, several indications that ATPs are lexical units. However, ATPs share properties with syntactically formed verb phrases too. Not

428

Lena Ekberg

only are ATPs identical to ordinary verb phrases in their surface structure, but they are also semantically equivalent to concrete verb phrases with a locative PP, cf. (5): (5)

a. b.

gä till affaren 'go to the shop' falla i sjön 'fall into the lake'

Both types of phrases denote a transition in which the subject of the verb phrase is the Theme (the person who moves) and the noun phrase in the preposition phrase is the Goal (or Source) of the movement. ATPs are futhermore semantically decomposable just like their concrete equivalents. This is due both to the fact that the meanings of verb and prepositions are regular metaphorical extensions of the concrete use of these items,5 and to the fact that there are regular syntagmatic relations between the parts of an ATP. Hence, the verbs that are also used with a concrete sense preserve their meanings as regards volitionality, aktionsart, and orientation (see further sect. 3). Thus it seems that ATPs can be regarded simultaneously as lexical units and as semantically ordinary V+PP constructions, like the ones illustrated in (5) above.

3.

ATPs as in-between syntactic constructions and idioms

As regards the grammatical-lexical status of ATPs, there is evidence that these phrases constitute an in-between phenomenon, in more than one respect. Thus ATPs have properties in common not only with syntactic constructions but also with idioms and productive derivational patterns. Like productive syntactic constructions, ATPs are regularly formed; there are restrictions on the formation of ATPs but these are strictly semantic. In fact, these restrictions are regular and predictable, and show that the parts of an ATP potentially may enter the grammar as independent elements. The semantic conditions which must be fulfilled if an ATP is to be considered well-formed can be regarded as a kind of semantic concord between verb and nominal (cf. Daniels 1963: 23, König 1973: 109). Thus, the verb and the noun generally cannot have conflicting values for the feature ±volitional: (6)

a. *falla i strejk fall into strike

Swedish abstract transitional phrases

429

b. *gä i sömn go into sleep On the other hand, like idioms, ATPs have constraints on syntactic operations (topicalization, cleft, etc.) which indicate that ATPs are to a certain degree lexicalized. Thus, all of the following examples are either highly questionable or clearly ungrarnmatical in Swedish. (7)

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

Topicalization

HSömn foil hon i. sleep fell she into Cleft UDet var sömn som hon foil i. it was sleep that she fell into Pseudo-cleft HVad var dethonfölll? what was it she fell into Wh-question Vart foil hon? - *I sömn. where fell she? - into sleep Variation in species Fienden gick till anfalll ^anfallet the-enemy went to attack/the-attack Variation in number Fienden gick till attack! *attacker the-enemy went to attack/attacks

However, ATPs are not as syntactically frozen as idioms are: this becomes clear when we apply a number of idiom criteria to ATPs, comparing them both with syntactic constructions and with "pure" idioms. Like the equivalent concrete verb phrases, the noun in an ATP may be extended with attributes, and construed with an article, indefinite or definite (8a, b). A similar extension of the noun in a pure idiom with the same syntactic-semantic form is not possible (8c). (8)

a.

Jonas gick till ett ovanligt häftigt angrepp mot det nya Jonas went to an unusually violent attack against the new forslaget. proposal b. Lisa gick till det röda huset. Lisa went to the red house c. *Lisa gick till djup hotten med problemet. Lisa went to deep bottom with the-problem (gä till botten med 'go to the bottom of sth')

430

Lena Ekberg

The application of idiom criteria to ATPs shows that these phrases lie somewhere at the midpoint of a lexicalization scale where one pole is made up of "pure" idioms and the other of "pure" syntactic constructions. The flexibility as regards the form and content of the noun, however, makes the ATPs vary their position along the lexicalization scale. In its extended form (cf. 8a) an ATP is closer to the "pure" syntactic pole than to the "pure" idiom pole. Finally, there is at least one property which ATPs share neither with ordinary VPs nor with idioms: the proposition in an ATP is invariably followed by a bare noun. This is different from ordinary V+PP constructions, where the noun in the PP is almost always construed with an article (cf. 8b). The naked form of the noun is an indication that some kind of semantic merger has taken place. In a recent syntactic description of ATPs (Platzack 1995), it is argued that complex predicates of this type are ordinary V+PP in the overt syntax, whereas they function as atoms in the semantic component, like idioms and simple words. Arguments supporting the analysis of ATPs as ordinary V+PPs are for instance: (i) The possibility to modify an ATP with an adjective, cf. (8a). (ii) The fact that an ATP never takes an NP-complement, but only a PP-complement. This indicates that an ATP is syntactically distinct from ordinary transitive verbs. The complement of an ATP must always be introduced by a preposition, as in (9a), whereas a corresponding verb often takes an NP-complement, as in (9b): (9)

a.

b.

Han gick till onfall mot fienden. he went to attack against the-enemy 'He attacked the enemy.' Han angrep fienden. he attacked the-enemy 'He attacked the enemy.'

The fact that ATPs have meanings similar to simple verbs supports the semantic treatment of these phrases as atoms. In terms of recent principle and parameter approaches to syntax, this is described as an incorporation of the noun into the preposition. The naked noun in fact forces an incorporation analysis, given this theory, since a bare noun cannot otherwise be identified (see Baker 1988: 119). Thus far I have accounted for the similarities of ATPs with on the one hand ordinary syntactic constructions, and on the other hand pure idioms. But,

Swedish abstract transitional phrases

431

as a syntactic-semantic pattern, ATPs also show similarties with derivational patterns in word formation.

4.

ATPs as a derivational pattern

ATPs appear in morpho-syntactic groups, constituted by a common verb and preposition, cf. the schema in (10). As I will argue in this section, the verbpreposition combination semantically behaves like a derivational affix. (10)

falla i N fall into N

gaiN go into N gä till N go to N

komma i N come into N komma till N come to N

roka i N happen-to-come into N

As is clear from the schema in (10), there are gaps within the morpho-syntactic grouping. Thus/a//a 'fall' and raka 'happen-to-come' do not appear with till 'to', whereas gä 'go' and komma 'come' appear with both 'into' and till 'to'. This seems arbitrary since falla 'fall' in its concrete use is fully compatible with the preposition till 'to'. There is a further semantic grouping within these morpho-syntactic groups in that a specific combination of verb and preposition tends to be construed with a certain (semantic) type of noun;6 cf. (11)

falla i sömn I dvala I slummer I koma I trans ... fall into sleep / doze / slumber / coma / trance gä till onfall I attack I drabbning I kamp l offensiv l reträtt... go to attack / assault / clash / battle / offensive / retreat

Primarily, this semantic grouping is due to the semantic content of the specific verb-preposition combinations. Thus, just as other derivational elements, the verb-preposition combination has specific semantic properties that determine which type of noun it is compatible with (cf. Fanselow 1988). Secondly, the semantic grouping is due to the existence of prototypical patterns for the formation of new ATPs. This means that a potential, semantically well-formed, ATP which does not adhere to a prototypical ATP-pattern is less likely to be formed than an ATP which adheres to such a pattern. I will argue that it is merely at a comparatively general level that the properties of the verbpreposition combination determine which type of nouns are compatible with

432

Lena. Ekberg

it. At a more specific level, prototypical patterns of ATPs will determine from what specific semantic domains these nouns are taken. The morpho-syntactic grouping, as well as the semantic grouping, give rise to lacunae which resemble those found in word formation (see, e.g., Marchand 1964; Fanselow 1988). There are consequently three types of lacunae in the formation of ATPs: a. Those due to a nonexisting verb-preposition combination, cf. (12). In other words, certain combinations of verb and preposition do not function as a productive "derivational element" (cf. the single instance of the combination falla till 'fall to', \iz.falla tillföga 'submit', 'yield', which is totally frozen and noncompositional). (12)

a. *falia till N fall to N b. *raka till N happen-to-come to N

b. Those due to the semantic content of the specific verb-preposition combinations, cf. (13): (13)

a. *fallai konflikt fall into conflict b. *ga till gräl go to quarrel c. *gä i

onfall

go into attack d. *räka i balans happen-to-come into balance c. Those due to the nonexistence of a prototypical ATP-pattem, cf. (14) (# indicates that the expression is semantically acceptable but not attested): (14)

a. üfallai bräk/gräl I stria fall into row / quarrel / fight b. #komma i sorg come into sorrow c. #raka i förtjusning happen-to-come into delight

Swedish abstract transitional phrases

433

Of these three types of gaps, those illustrated in (14) come closest to being "arbitrary" gaps. In the following subsection, I will investigate the semantic content of the verb-preposition of ATPs, showing why the phrases in (13) are unacceptable. Section 4.2 presents some data as regards the prototypical patterns of ATP, which in turn will explain why the phrases in (14) are not likely to be formed, although they are semantically well-formed. Section 4.3, finally, summarizes the restrictions on productivity given by the principle of semantic concord, the specific content of the verb-preposition combination, and the existence (or rather nonexistence) of prototypical ATP-patterns.

4.1.

The semantic content of the "derivational element" of ATPs

The invariant verb-preposition combination in an ATP can be regarded as a derivational morpheme, as the properties of this combination are equivalent to the semantic content of a derivational morpheme in word formation (see Fanselow 1988). In the following I will consider the meaning of both the separate verb and the verb-preposition combination in ATPs. The verb gä 'go' generally denotes the volitional (momentary) initiation of an action, an activity, or (more seldom) a state when used in an ATP.7 More specifically, the combination ga in 'go into' denotes the volitional (momentary) initiation of an activity or a state - and thus is combined with nouns denoting activities or states (cf. (15)), but not with nouns denoting momentary actions, cf. *gä i onfall 'go into attack'. (15)

a. b.

gä i närkamp I debatt I strejk ... go into infighting / debate / strike gä i konflikt I konkurs I exil ... go into conflict / bankruptcy / exile

The combination gä till 'go to', on the other hand, denotes either the volitional (momentary) initiation of an action - cf. (16a) - or the initiation of an activity which is conceptualized as an action (a point), cf. (16b): (16)

a.

gä till handling / attack / offensiv

b.

go to action / attack / offensive gä till kamp /förhandling / val goto fight /negotiation /election

434

Lena Ekberg

The difference between gä i 'go into' and gä till 'go to' is obviously due to the preposition, which in the context of an ATP preserves the semantic properties from its basic spatial domain; cf. (17), where till 'to' and i 'in'8 are used in concrete contexts. (17a) shows that the use of till 'to' causes the following noun to be interpreted as zero-dimensional, whereas (17b) shows that the use of / 'in' causes the same noun to be interpreted as three-dimensional. (17)

a. b.

Hangick tilstaden Odim he walked to the-town Hangick i staden 3dim he walked in the-town (i.e., he walked around in the town)

Given that activities and states are metaphorically interpreted as three-dimensional (i.e., CONTAINERS), whereas actions are interpreted as zero-dimensional (i.e., as POINTS) - noted, e.g., by Lakoff and Johnson (1980) - it is expected that gä i 'go into' will be used to designate the beginning of an avtivity or state, whereas gä till 'go to' will designate the beginning of an action - which is exactly the case in ATPs. Cf. the implicational pattern in (18): (18)

a.

b.

De har gatt i strejk they have gone into strike They have gone on strike.' De har gatt till an/all they have gone to attack 'They have attacked.'

—> -»

They are on strike. They were on strike,

—>· -*»

They attacked. They are attacking.9

The verb komma 'come' is unspecified for volitionality, and is furthermore goal-oriented, i.e., focuses on (the achievement of) the goal of the movement. In fact, komma seems to be unspecified for volitionality because it is goal-oriented. Using komma nothing is asserted regarding the initiation of the movement. Komma neither asserts that the movement started as the result of an act of volition, nor does it exclude this construal. In this respect, komma is significantly different from/alia 'fall' and raka 'happen-to-come', which explicitly deny the involvement of a volitional initiator. Besides being goal-oriented, komma (just like English come) is deictically grounded, in

Swedish abstract transitional phrases

435

that the change of location indicated by komma is prototypically seen from the speaker's location at the goal (cf. Radden 1995 for the case of come}. In extending the use of komma to denote change of abstract state, the verb will prototypically indicate that the viewpoint from which the state is observed is at the goal. Due to this viewpoint of meaning, komma is well suited to denote achievements of intended mental states. When komma is used the achievement of the goal can be interpreted either as momentary or as durative, i.e., the part of the event which is in focus can be conceived of as either a point without extension, cf. (19), or as an extended line, cf. (20): (19)

Honkom till medvetande kl 8.20. she came to consciousness at 8.20 'She regained consciousness at 8.20.'

(20)

Hon kom langsamt till medvetande. she came slowly to consciousness 'She slowly regained consciousness.'

Both komma i and komma till may be combined with a noun denoting either an action, an activity, or a state. The combination komma i 'come into' expresses that the final part of the source-state merges into the entering of the goal-state, whereas komma till 'come to' expresses that the final part of the source-state (which lies within the orientation of komma) continuously leads up to the goal-state - but not into it. Cf. the figure below (the continuous line indicates focus): (21)

a. b.

komma i come into komma till come to

When an ATP with komma designates a momentary transition - illustrated in (22) - the final part of the source-state plus the beginning of the goal-state are, conceptually, compressed to a point. (22)

a.

Hon kom plötsligt i förlägenhet. she came suddenly in embarrassment 'She suddenly became embarressed.'

436

Lena Ekberg b.

Hon kom till medvetande kl. 8.20. she came to consciousness at 8.20. 'She regained consciousness at 8.20.'

The verb/alia 'fall' is merely combined with i 'into'. The combination folia i 'fall into' designates the nonvolitional, momentary transition into a mental/physiological process or a state: (23)

a. b.

folia fall folia fall

i begrundan l grät ... into meditation / crying i förväning l sömn I trans ... into surprise / sleep / trance

Also räka 'happen-to-come' is merely combined with / 'into'. In comparison with/ oak tree —> low branch —>· (paw —>·) tiger. (13)

And then at last, just when Tommy was sure his tiger was lost and gone forever - he happened to look out of his window at the bare oak tree. And there, hanging over a low branch, and waving one paw to say, 'Come and get me - I thought you were lost!' was the tiger. (Jackson 1960: 18)

The choice to conform to experiential iconicity may seem automatic until we actively choose to go against it. In the same way as the consequences of other text-strategic decisions, however, the choice to conform or not to experiential iconicity is also visible in the text.

536

Tuija Virtanen

TEMPORAL SUCCESSION

nar/////ins/////des/////exp/////arg

LOGICAL SUCCESSION

Figure 1. The relative positions of five types of text on a scale of 'temporal' vs. 'logical' succession in discourse (nar = narration; des = description; arg = argumentation; ins = instruction; and exp = exposition).

3. Text types The discussion of iconically motivated ordering in the text may be related to text-internally characterizable text types, such as the narrative, descriptive, instructive, expository or argumentative types of text (cf., e.g., Werlich 1976; see also Longacre 1983). Adverbial placement varies according to text type (see e.g. Virtanen 1990; 1992d). We have seen that initial position is typically reserved for adverbials of time signalling textual boundaries in narratives, and adverbials of place and other locative expressions in descriptions of space. It is possible to draw a scale ranging from texts that are typically arranged temporally to others that rather conform to a temporally iconic, logical succession in discourse, with several types in between (cf. also Virtanen 1992d: 329-331). In outlining Figure 1,1 have started out from Werlich's (1976) five types of text, regarding them as prototypical abstractions of unitype texts. Most texts are, of course, multitype, i.e. they consist of more than one type of text. Very briefly, narratives - at the one extreme of the scale - are, implicitly or explicitly, organized temporally, and deviation from temporal sequentiality demands explicit signalling in the text. Argumentation, in the sense of certain culture-specific patterns of rather openly presenting evaluative discourse, may be placed at the other end of the scale. I am not here concerned with the non-temporal organizing principles of expository and argumentative texts. However, some of the various logical connectors, which may but need not be used to structure argumentation or expository texts, indicate a temporally iconic logical succession in discourse, for instance, the enumerative conjuncts./zrsf, then, next, and finally. As pointed out above, a fundamental aspect of a successful instructive text is its strict conformity to experiential iconicity, which makes explicit markers of time, in their primary function, unnecessary. They may, however, be used to signal other aspects of text organization. Instructive texts vary from

Adverbial placement and iconicity

537

more narrative-like texts that make use of temporal signals as explicit markers of textual boundaries, to more description-like, or even exposition-like texts using enumerative conjuncts instead. Expository texts, again, range from 'expository narratives' (e.g. biographies) to blends of exposition and argumentation. In between, we may find another flexible category that may still be distinguished from the others, i.e. description. It often follows what may be called 'generic' time, and in a normal multitype text, it may therefore be relatively easily adjusted to match the time orientation of the main type of text. Descriptions, the most 'neutral' category on this scale, vary from 'suggestive', typically found in multitype narratives, to more expository ones. On a high level of abstraction, the scale thus displays a gradual change from the explicitly signalled temporal text strategy to an implicit one, as we proceed from left to right. Similarly, as we proceed from right to left, we may discern a gradual change from a large variety of potential signals of logical succession in argumentative and expository texts - signals which are seldom merely enumerative - to a smaller array of such logical markers, such as so in narrative.3 Markers of a locative text strategy appear essentially in the middle area of this scale - i.e., in descriptions and expository accounts of spatial relationships, and route instructions. But none of these three types of texts - description, instruction, and exposition - need to display a locative text strategy; they can be organized around another notion or entity central to the discourse topic (for 'discourse topic', see Brown and Yule 1983: 71ff.). Further, narratives may rely on implicitness in signalling temporal succession and manifest a locative text strategy, which moves participants from one scene to another. What we often find in such texts, however, is a blend of explicit markers of the two strategies. The scale in Figure 1 is included in the present paper because it may be considered in terms of experiential iconicity. Hence, narratives, instructions, and descriptions may be temporally and/or spatially iconic. Locatively structured instructions and place descriptions occur in many languages. And despite cultural variation, the structure of a prototypical narrative is highly iconic. Thus, at least as long as the picture of the world on which the iconicity in the text is based remains relatively constant cross-linguistically, such iconic types of text may be assumed to have a more universal character than most expository texts and in particular argumentative texts, which are rather organized according to various culture-specific patterns, and may thus be sug-

538

Tuija Virtanen

gested to manifest a lower degree of universality. As pointed out above, however, expository and argumentative texts may, but need not, manifest a series of enumerative markers of logical succession in discourse, which may be iconic of markers of the temporal succession in narrative. But this relation represents a different kind of iconicity, 'text-typical iconicity', pertaining to text types and markers of text structure characteristic of different types of text: These texts do not conform to experiential iconicity, in the sense of a match between the text and our experience of the world. Finally, TirkkonenCondit (1992) uses the term 'processual iconicity' to refer to reflections of the argumentative process developing around anticipated disagreement, in the sequential organization of written argument.

4.

Iconicity in text structure

In this section I shall view adverbials of time and place signalling a temporal or locative text strategy in terms of iconicity reflecting the hierarchy of text structure. To figure out the hierarchic organization of a number of texts, I have used established models of text structure, which are typically content-based, as well as paragraphing tests (see Virtanen 1992d). If (a) the content-based analyses of a near-prototype text according to several different models, (b) native speakers' paragraphing of the same text, and (c) my analysis of its text-strategic signals all point in the same direction at a given stage in the text, then that particular juncture in the text may be assumed to constitute a textual boundary of some kind. Further, we may consider a given point in a text an important boundary if all, or the large majority, of the subjects agree on their paragraph markings at that particular point and if it also seems to be a major boundary according to one or several different models of text structure and if, further, it manifests explicit signals of one or several cooccurring text strategies. The analysis is fully reported in Virtanen 1992d and I can here only note a tendency which Givon (1983: 18; 1984: 245; cf. also Givon 1985) pays attention to in his discussion of topic continuity: Fewer markers or markers of smaller size seem to appear at lower-level boundaries while clusters of markers or markers of larger size tend to occur at major boundaries of a text. A text often manifests a complex interplay of cooccurring or combined text strategies. We have seen above that initially placed adverbials of time

Adverbial placement and iconicity

539

and place may create cohesion and coherence and signal boundaries between different textual units. While they thus function as markers of a text strategy, they may also perform a host of other textual tasks because of their position in the text. What is of interest in this connection is how these adverbials may be used to mark the hierarchy of text structure; they may contribute to the signalling of textual boundaries as minor or major. Major boundaries may be signalled with the help of clusters of markers of one or several cooccurring text strategies, or markers of larger size than those indicating a minor boundary. Minor boundaries may then be signalled with the help of only one marker, which may also be smaller in size. Examples of size are one evening vs. then in a narrative, or in Trafalgar Square vs. here in a place description. The text fragment (14), from the middle of a children's story, illustrates the difference in the number of markers. What can be observed in this example is a combination of a temporal and participantoriented text strategy. Lexically weighty explicit markers of the two cooccurring strategies appear at the outset of a major boundary in the text (i.e. the adverbial marker one evening and the full-NP reference to one of the main participants, the woman). This boundary may be compared with the following, less important boundary in the text, where we find a temporal marker, on Tuesday, while the participant is referred to using a pronoun, which does not function as a textual marker. Later in the same text, example (15), we find a minor textual boundary, and very little linguistic material is then needed in the temporal marker. (14)

One evening the woman found that there was a bit of milk left over after supper. "I may as well give it to those skinny, scraggly, scrawny cats," she decided. She poured it into a pan and put it in the garden. That was on Monday. On Tuesday, she ordered a whole extra quart of milk from the milkman. By mistake, of course. (Rowand 1966: 154)

(15)

Then, of course, she couldn't throw it away - because she knew how cats feel about FISH. (Rowand 1966: 155)

Another way of indicating the hierarchy of text structure is the use of different kinds of markers at major and minor boundaries. A discussion of such instances, as well as text strategies realized through non-adverbial signals, must, however, remain outside the scope of the present paper and the reader

540

Tuija Virtanen

is again referred to Virtanen (1992d) for an extensive discussion of the interplay of text strategies in this and other texts. In addition to the size of the marker, or the number of markers, at the outset of major and minor textual units, the hierarchy of text structure may also show up in the information status of strategy-marking adverbials. I have shown elsewhere that clause-initial adverbials may have a variable range of information status on a scale reaching from given to new information. Characteristically, however, such adverbials contain information which may be situated in the middle of such a scale, rather than the extreme ends of textual^ given information or new information proper. This obviously reflects their multifunctional role in the text: They may act as signals of a text strategy, mark textual boundaries, and constitute a starting point for a new textual unit (see Virtanen 1992a; 1992d). The information status of the strategy-marking adverbials seems to contribute to the signalling of textual boundaries as minor or major in the sense that informationally newer material may be used to indicate the outset of a major textual unit. In contrast, minor boundaries may then be marked with elements that are more given. Compare, in this light, the information status of the initial adverbials one evening and then in (14) and (15), p. 539 above. In (16), below, the second cluster of locative adverbials conveys information that is more given than the information contained in the first cluster, which indicates a major boundary in the text and is therefore more highly specified lexically. (16)

In Redhall Bank Road, off the south side of Lanark Road, is an eccentric block of quarriers' cottages by Sir James Gowans, c. 1850, built, unusually for this area, of a type of Kentish Rag stone. Nearby in the same road is Millbank, an 18th-century house with pediment, roundel and stone stair. (Hamilton 1978: 160)

We may of course assume that more material at the outset of a major textual unit may also easily lead to newer elements being included in the marker or markers. Further, adverbials conveying information of similar status may have a different source of givenness or familiarity when they signal a major or a minor boundary in the text. In other words, the source of givenness of the strategy marker may be less far away and thus assumed to be more readily accessible if the boundary is a minor one. Markers of major boundaries, again, often

Adverbial placement and iconicity

541

refer to a 'topic entity' (Brown and Yule 1983: 137) or to an activated narrative 'schema' (de Beaugrande and Dressier 1981: 90-91, 184) in the types of texts for which I have given examples. Hence, in Text (2), p. 528f. above, we find, for instance, the adverbial cluster upstairs, in the Tool Room introducing a new major textual unit and the prepositional phrase in one corner signalling the next minor boundary within that major unit of text. All three adverbials can be analysed similarly, for example, as conveying 'inferrable' information (Prince 1981). The mention of the Tool Room at the outset of the preceding sentence helps the text-receiver make the inference at the following textual boundary (cf. 'rooms have corners'). In contrast, the first cluster of adverbials must be inferred from the topic entity, i.e. the museum (cf. also the discussion in Virtanen 1992a; 1992d). Cognitively, it may be hypothesized that the appearance of lexically weighty signals and clusters of markers at major boundaries in the text allows the text receiver to (temporarily) drop from her/his working memory many of the aspects that have been kept activated over minor boundaries. The text producer is presumed to help the text receiver by signalling the relevant text-strategic aspects of the new major textual unit - which need to be reactivated or created at that point in the text (cf. e.g. the discussion of the short-term, working memory, and the 'episodic text memory' in van Dijk and Kintsch 1983: 246ff.; see also e.g. Givon 1990: 940-941; Stark 1988: 301). It seems that contrary to the content, signals of text structure need not enter the longer-term memory systems (cf. e.g. Chafe 1987). The signalling of major and minor boundaries with different means on the textual surface is a way of conveying the hierarchy of text structure through the perforce linear expression of language: Through the use of more and/or informationally newer linguistic material - adverbial or other - at the outset of major textual units, as compared to the beginning of minor units, the text producer can make use of iconicity. In other words, the number/size and information status of these markers then forms an iconic reflection of the size of the boundary.

5.

Summary: On the notion of 'iconicity'

I have discussed clause-initial adverbials of time and place from the perspective of two kinds of iconicity: First, the obvious - and perhaps a basic kind of - iconicity in which the ordering of clause elements or parts of text conforms

542

Tuija Virtanen

to the order of what they depict, as experienced by the text producer. In addition to such experiential iconicity, I have also discussed iconicity manifested in the signalling of text structure. Other phenomena, too, are iconic, and some of them may perhaps be derived from the simple, concrete and obvious principles of experiential iconicity. In Text (2), the strategy of signalling the outset of a new unit in the text, i.e. a new move that the tourist needs to make to visit the museum in the intended order, through an initial locative such as on the I. side in (17), below, is obviously not followed when no further move is intended. Hence, nearby in (17) does not appear initially in the sentence as it is possible to contemplate all the models from the very spot on which the tourist is assumed to be standing already. At the same time, however, this arrangement allows the text producer to indicate that the models form a single sight, only necessitating one stop: All the models are grouped together in the text by being named one after the other (in an order which would invite yet another discussion). (17)

On the I. side is a VICTORIAN BEDROOM with models of a farmer's wife in a half tester bed with her new baby. Her small son and the midwife stand nearby. (McGregor Eadie 1981: 69)

If, indeed, iconicity, as it seems, is inherent in language, and the possibility of going against iconicity of one kind may be explained with the help of other iconicity principles - such as the grouping of elements close to each other on the textual surface as in (17), to indicate that they form a unit - then what we need to do is to figure out the different kinds of iconic reflections we typically find in text and discourse. Otherwise the notion of 'iconicity' may prove too broad, encompassing such a wide range of phenomena that it risks losing its explanatory value in the analysis of authentic texts.

Notes 1. It is useful to make a terminological distinction between typographical units such as paragraphs and textual units of various kinds as the two do not necessarily coincide. 2. I shall not discuss the difficult question which of the two notions, spatiality or temporality, in fact implies the other. Let me, however, note in passing that the temporal succession of events in story-time and text-time may be viewed as a spatial metaphor. But it is equally possible to regard the flow of time as a primary notion in narrative: We live in time, our speech emerges incrementally

Adverbial placement and iconicity

543

in time, and we make sense of the world around us by structuring it in terms of events that succeed each other in time. For discussion of narrative as a pattern for organizing experience or imagination, see, e.g., Björklund 1993; Fleischman 1990: 94ff.; Ong 1982; White 1980. Consider also the discussion in Enkvist 1981; 1987; and in Fleischman 1991. For discussion of 'localism', see, e.g., Lyons 1977: 669, 718ff., and Levinson 1983: 84-85; cf. also Levinson 1992. 3. The persuasive goal of argumentative texts, in fact, often demands a high degree of implicitness. Östman (1987: 104) points out that the unmarked situation of persuasion "would be an avoidance of markers - especially explicit markers that might indicate that you are in the process of persuading somebody." Cf. also the discussion in Virtanen 1992c.

References Texts Boase, W. (reteller) 1983 Three Bears. London: Walker Books. Hamilton, A. 1978 Essential Edinburgh. London: Andree Deutsch. Jackson, K. 1960 The Bedtime Book of 365 Stories. Feltham: Hamlyn. McGregor Eadie, P. 1981 Blue Guide: The Channel Islands. London & Tonbridge: Ernest Benn. Scott, P. 1983 Travel Diaries of a Naturalist I. London: Collins. Piper, M.R. 1970 Sunset Oriental Cook Book. Menlo Park, CA: Lane Books. Robertson, I. 1981 Blue Guide: Cyprus. London & Tonbridge: Ernest Benn. Rowand, P. 1966 'The cats who stayed for dinner.' B. Ireson. (ed.) The Faber Book of Nursery Stories. London: Faber & Faber. 152-156. Wilson, T. 1972 Great Rice Dishes of the World. Manchester: World Distributors.

Studies Björklund, Martina 1993 Narrative Strategies in Cechov's 'The Steppe': Cohesion, Grounding and Point of View. Abo: Abo Akademi University Press. Brown, Gillian and George Yule 1983 Discourse Analysis. Cambridge: Cambridge University Press.

544

Tuija Virtanen

Chafe, Wallace L. 1987 Repeated verbalizations as evidence for the organization of knowledge. Preprints of the Plenary Session Papers: XlVlh International Congress of Linguists, Berlin/GDR, August 10th-l5th, 1987. Berlin: Akademie der Wissenschaften der DDR. 88-110. Chatman, Seymour 1983 Story and Discourse. Narrative Structure in Fiction and Film. Ithaca & London: Cornell University Press. Cornell Paperbacks. (Istpubl. 1978 by Cornell University Press). Clark, Herbert H. and Eve V. Clark 1977 Psychology and Language: An Introduction to Psycholinguistics. New York etc.: Harcourt Brace Jovanovich. De Beaugrande, Robert and Wolfgang Dressier 1981 Introduction to Text Linguistics. London & New York: Longman. Enkvist, Niels Erik 1975 Tekstilingvistiikanperuskäsitteitä. Jyväskylä: Gaudeamus. 1976

Notes on valency, semantic scope, and thematic perspective as parameters of adverbial placement in English. In N.E. Enkvist and V. Kohonen (eds.), Reports on Text Linguistics: Approaches to Word Order, 51-74. Publications of the Research Institute of the Abo Akademi Foundation 8. Abo.

1981 1987

Experiential iconicism in text strategy. Text 1.1: 97-111. A note towards the definition of text strategy. Zeitschrift für Phonetik, Sprachwissenschaft und Kommunikationsforschung 40. l: 19-27. Connexity, interpretability, universes of discourse, and text worlds. In S. (ed.), Possible Worlds in Humanities, Arts and Sciences: Proceedings of Nobel Symposium 65, 162-186. Research in Text Theory, 14. Berlin, New York: Walter de Gruyter.

1989

Erlich, V. 1969 Firbas, Jan 1979 1986

Russian Formalism: History - Doctrine. Slavistic Printings and Reprintings 4. The Hague, Paris: Mouton. 2nd, revised ed. first printed 1965; Isted. 1955. A functional view of Ordo Naturalis'. Brno Studies in English 13: 29-60. On the dynamics of written communication in the light of the theory of functional sentence perspective. In C.R. Cooper and S. Greenbaum (eds.), Studying Writing: Linguistic Approaches, 40-71. Written Communication Annual 1. Beverly Hills etc.: Sage Publications.

Fleischman, Suzanne 1990 Tense and Narrativity: From Medieval Performance to Modern Fiction. London: Routledge. 1991

Discourse as space/Discourse as time: Reflections on the metalanguage of spoken and written discourse. Journal of Pragmatics 16.4: 291-306.

Genette, Gerard 1972 Figures III. Paris: Seuil.

Adverbial placement and iconicity

545

Givon, Talmy 1983 Topic continuity in discourse: An introduction. In T. Givon (ed.) Topic Continuity in Discourse: A Quantitative Cross-Language Study, 1-42. Typological Studies in Language 3. Amsterdam, Philadelphia: John Benjamins. 1984 Syntax: A Functional-Typological Introduction. Vol. I. Amsterdam, Philadelphia: John Benjamins. 1985 Iconicity, isomorphism and non-arbitrary coding in syntax. In J. Haiman (ed.), Iconicity in Syntax, 187-219. Typological Studies in Language 6. Amsterdam, Philadelphia: John Benjamins. 1990 Syntax: A Functional-Typological Introduction. Volume II. Amsterdam, Philadelphia: John Benjamins. Grimes, Joseph E. 1975 The Thread of Discourse. Janua Linguarum, Series Minor 207. The Hague, Paris: Mouton. Labov, William 1977 Language in the Inner City: Studies in the Black English Vernacular. Oxford: Basil Blackwell. (Copyright 1972 by The University of Pennsylvania Press.) Levelt, W.J.M. 1981 The speaker's linearization problem. The Psychological Mechanisms of Language. London: The Royal Society and The British Academy. 91-101. (1st publ. in Philosophical Transactions of the Royal Society of London B: 295. 305-315.) Levinson, Stephen C. 1983 Pragmatics. Cambridge: Cambridge University Press. 1992 Primer for the field investigation of spatial description and conception. Pragmatics 2.1: 5-47. Linde, Charlotte and William Labov 1975 Spatial networks as a site for the study of language and thought. Language 51.4: 924-939. Longacre, Robert E. 1983 The Grammar of Discourse. New York, London: Plenum Press. Lyons, John 1977 Semantics. Vol. II. Cambridge etc.: Cambridge University Press. Ong, Walter J. 1982 Oral remembering and narrative structures. In D. Tannen (ed.), Analyzing Discourse: Text and Talk, 12-24. Georgetown University Round Table on Languages and Linguistics 1981. Washington, D.C.: Georgetown University Press. Osgood, Charles E. 1980 Lectures on Language Performance. Springer Series in Language and Communication 7. New York etc.: Springer Verlag.

Östman, Jan-Ola 1987

Pragmatic markers of persuasion. In J. Hawthorn (ed.), Propaganda, Persuasion and Polemic, 91-105. London: Edward Arnold.

546

Tuija Virtanen

Prince, Ellen F. 1981 Toward a taxonomy of Given-New Information. In P. Cole (ed.), Radical Pragmatics, 223-255. New York: Academic Press. Rimmon-Kenan, Shlomith 1983 Narrative Fiction: Contemporary Poetics. London & New York: Methuen. Schiffrin, Deborah 1990 Between text and context: Deixis, anaphora, and the meaning of then. Text 10.3: 245-270. Stark, Helen A. 1988 What do paragraph markings do? Discourse Processes 11.3: 275-303. Tirkkonen-Condit, Sonja 1992 Articulation of relational propositions: A tool for identifying an aspect of text comprehension. In A.-C. Lindeberg, N.E. Enkvist and K. Wikberg (eds.), Nordic Research on Text and Discourse: NORDTEXT Symposium 1990, 173184. Abo: Abo Academy Press. Van Dijk, Teun A. and Walter Kintsch 1983 Strategies of Discourse Comprehension. New York etc.: Academic Press. Virtanen, Tuija 1990 Adverbial placement as a criterion in text typology. In W. Banner, J. Schildt and D. Viehweger (eds.), Proceedings of the XlVth International Congress of Linguists (Berlin/GDR, August 10th-15th, 1987). Berlin: Akademie-Verlag. Vol. Ill: 2247-2250. 1992a Given and new information in adverbials: Clause-initial adverbials of time and place. Journal of Pragmatics 17.2: 99-115. 1992b Temporal adverbials in text structuring: On temporal text strategy. In A-C. Lindeberg, N.E. Enkvist and K. Wikberg (eds.), Nordic Research on Text and Discourse: NORDTEXT Symposium 1990, 185-197. Abo: Abo Academy Press. 1992c Issues of text typology: Narrative - a 'basic' type of text? Text 12.2: 293-310. 1992d Discourse Functions of Adverbial Placement in English: Clause-Initial Adverbials of Time and Place in Narratives and Procedural Place Descriptions. Abo: Abo Akademi University Press. Werlich, Egon 1976 A Text Grammar of English. Heidelberg: Quelle & Meyer. White, Hayden 1980 The value of narrativity in the representation of reality. Critical Inquiry 1.1: 5-27.

Pragmatics

Nominal vs. temporal interpretation in discourse Alfons Maes and Leonoor Oversteegen 1. Introduction The central question in this paper is to what extent nominal and temporal interpretative phenomena in discourse are analogous. Firstly, three apparent similarities between nominal and temporal reference will be put forward. Subsequently, ontological differences between temporal and nominal referents and their linguistic realizations in discourse will be discussed. These differences are used as a kind of search light in discovering, describing and explaining three discourse phenomena, in which nominals and temporals behave differently. Treating nominal and temporal interpretative phenomena in discourse in the same way is inspired by apparent similarities as illustrated in (1): (1)

He went home.

In interpreting this sentence, the addressee will have to attach the nominal expression he to some contextually or situationally available discourse referent. Similarly, the past tense has to be attached to some contextually or situationally available point of reference. Nominal and temporal anaphoric expressions are alike, in that both presuppose a referential framework within which interpretation takes place. This framework accounts for two interpretational steps: the addressee must establish an initial reference point, and (s)he must maintain and change referents and reference points as the discourse proceeds.

2. Three apparent similarities When one surveys recent theories on nominal and temporal phenomena in discourse, it turns out that temporal phenomena are interpreted or re-interpreted in terms of characteristics which were traditionally developed in the field of nominal phenomena, and not vice versa. We will discuss three commonalities in the treatment of nominal and temporal constructions:

550

Alfons Maes and Leonoor Oversteegen

-both can be described in terms of phoric characteristics -both can be described in terms of focus and centering theory -both can be described in terms of their dependence on real world knowledge.

2.1.

Phoric characteristics

A straightforward way in which nominal and temporal interpretation can be said to be analogous is given in the tense as an anaphor tradition. This tradition more or less restricts temporal interpretation to a semantic process (cf. Hinrichs 1981, 1986; Partee 1984; see Oversteegen 1992: 48^9), an approach that can be compared to the strict view on nominal anaphor resolution, in which only the antecedent and the anaphor are involved. Within this tradition, several parallels between nominals and temporals can be described. Partee (1984) lists the following cases of parallel behaviour between nominals and temporals. Nominals as well as temporals can be used without linguistic antecedent, i.e., their interpretation can be dependent on situational interpretation: (2)

a. b.

She left me. / didn 't turn off the stove.

NOM TEMP

Nominals as well as temporals can have definite antecedents, as in (3a) and (3b): (3)

a. b.

5am is married. He has three children. Sheila had a party last Friday and Sam got drunk. When John saw Mary, she crossed the street. At 3 pm June 21st, 1960, Mary had a brilliant idea.

NOM TEMP

Nominals as well as temporals can have indefinite antecedents, as in (4a) and (4b): (4)

a. b.

Pedro owns a donkey. He beats it. NOM Mary woke up some time during the night. She turned on the light. TEMP

Nominals as well as temporals can have bound variables as antecedents: (5)

a. b.

Every woman believes that she is happy. Whenever Mary telephoned, Sam was asleep.

NOM TEMP

Nominal vs. temporal interpretation in discourse

551

Nominals as well as temporals can be used to construct so-called donkeysentences: (6)

a. b.

Every farmer who owns a donkey beats it. NOM Whenever Mary telephoned on a Friday, Sam was asleep.TEMP

These examples can all be interpreted as instantiations of anaphoricity. However, there is an important difference between the nominal and the temporal cases: whereas in the case of nominals the antecedent and anaphor are realised by the noun phrases themselves, and the triggers of anaphoricity are properties of NPs such as (in)definiteness, gender etc., in the temporal examples the anaphoric relata are more complex. The examples show that at least temporal adverbials and temporal subclauses are involved in the anaphoric relation. This observation conveys a first glimpse of the central difference between nominals and temporals: in general, the referents of temporals are ontologically more complex than nominals, and consequently more linguistic apparatus is required to bring about temporal meaning in discourse.

2.2.

Focus and centering

A second similarity between nominal and temporal interpretation in discourse can be found in the idea that both nominal and temporal expressions are interpreted with respect to an attentional component. Take the following discourse: (7)

a. b. c. d.

John went over to Paul's house. On the way he stopped by the grocery shop for a bottle of whisky. He bought a Johny Walker black label because he knew that his friend was depressed. But unfortunately, the bottle failed to cheer him up.

Constructing a coherent representation of this text requires a description of the transitions between and the accumulation of the subsequent segments in terms of continuation and shift of nominal and temporal entities. At any particular moment in discourse a particular nominal and a particular temporal entity functions as the focus of attention (see for example Grosz and Sidner

552

Alfons Maes and Leonoor Oversteegen

1986, Webber 1988, Brennan et al. 1987, Poesio and Kameyama 1992, Grosz, Joshi and Weinstein 1995). As far as the interpretation of nominal referents is concerned, the addressee should be able to identify the evoked referents and attach the incoming information or events to these referents. This cognitive activity has been proved to be a major task in processing discourse (e.g., Garrod and Sanford 1977: 88, Cloitre and Bever 1988: 295 and Anderson and Hastie 1974: 512) and it is reflected in discourse models and representations (e.g., Seuren 1985, Kamp and Reyle 1993). With regard to temporal interpretation, addressees should be able to keep track of the temporal indices of the successive events. This is reflected within theoretical frameworks such as discourse representation theory DRT (Hinrichs 1981 and 1986, Kamp and Reyle 1993) or Augmented Transition Network structures (Polanyi and Scha 1984). For both nominal and temporal entities, a focus component has been postulated, containing the entities, properties and relations that are most salient at a certain point. For the example mentioned, the nominal focus component should not simply include nominal entities like the referents of John and Paul, but also a focal hierarchy between the entities, thus representing a prediction as to what entity will be the most likely focus to be continued in the following utterance. Likewise, the temporal focus should not only contain the points of reference mentioned in the discourse, but also propose a focal hierarchy representing the most probable time at which to interpret the following event.

2.3.

Nominal/temporal interpretation and real world knowledge

A third obvious similarity between nominal and temporal interpretation in discourse concerns the integration of incoming information with existing knowledge in the process of interpretation. Often linguistic information is not sufficient for the interpretation of the nominal or temporal structure. In those cases, real world knowledge is employed to come to an interpretation. Consider the following examples: (8)

John fell. Paul pushed him.

(9)

a. b.

John called Paul, because he wanted to know something, John called Paul, because it was his birthday.

Nominal vs. temporal interpretation in discourse

553

For both temporal interpretation in (8) and nominal interpretation in (9), reasoning based on world knowledge is decisive. According to our real world knowledge, the relationship between falling and pushing in (8) causes us to adopt a temporal ordering of events. Similarly, the interpretation of he in (9a) as referring to John and his in (9b) as referring to Paul, depends on world knowledge of the social conventions in using telephones.

3. 3.1.

The ontological difference between nominal and temporal entities Introduction

Whereas the previous section focused on similarities between nominal and temporal expressions, this section will stress their differences. It will be argued here that nominal and temporal expressions in discourse differ with respect to referential autonomy or, conversely, with respect to context dependency. This claim can be substantiated by comparing the minimal devices for nominal and temporal reference, pronouns and tense respectively. (10)

Hij kwam de bar binnen. He entered the pub.

The minimal triggers for nominal and referential interpretation in (10) bring about referential (who?) as well as temporal (when?) indeterminacy. But the minimal nominal device, i.e. the unstressed pronoun, carries itself categorical information about the referent, namely its number and its gender. The minimal temporal device, i.e. tense, does not contain any intrinsic information on the time interval referred to. It only offers deictic information: it conveys the relationship between the point in time at which (10) is uttered (the point of speech) and the interval referred to. Consequently these two devices carry fundamentally different information about their referents. This observation might induce doubt about the adequacy of the concept of tense as an anaphor. This initial comparison is illustrative for the difference between nominal and temporal interpretation in discourse: nominal interpretation comes about more autonomously than temporal interpretation. The linguistic realization of temporals is more complex given the complex temporal ontology: for the interpretation of tensed sentences, we need at least two types of information

554

Alfons Maes and Leonoor Oversteegen

sources. The first decides on the temporal structure of events and states (see section 3.2 below) and the second bears upon the relationship between the events and states and a reference point, for example the point of speech (see section 3.3 below).

3.2. 3.2.1.

Temporal referents versus nominal referents Introduction

Nominal referents can be conceived of as discrete entities mainly triggered by NPs. Nominal interpretation often is envisaged as an activation process of entities within semantic networks. These are structured by ontological relations, based on prototypes and stereotypes. The network relations are instrumental in interpreting anaphoric relations and consequently in explaining these interpretations. An entity, for example, which has been introduced as a sparrow can be referred to later as the bird because both mental objects are connected in the semantic network by a relationship of hyponymy. Temporal referents are not classified that easily. As we said, there are two essentially different types of temporal referents, which we will discuss subsequently. 3.2.2.

Events and states as referents

The most straightforward referent of a temporal construction like a tensed verb phrase is some kind of event. In an event some kind of movement or change may be expressed, change in time, as Russell (1903) called it. This change in time is due to the flow of time. So, events have an internal temporal structure, which is brought out in language; this enables us to refer to a certain part of the event (e.g., by means of an inchoative) or to indicate that the endpoint of the event is not necessarily reached (by means of a progressive). Of course, nominals may express events as well. But even if they do, these expressions do not incorporate the flow of time in the same way as temporals do. Compare the following sentences: (11)

(12)

Zij vergaderden en vergaderden en vergaderden. They conferred and conferred and conferred. ?Zy hielden een vergadering en een vergadering en een vergadering. They had a conference and a conference and a conference.

Nominal vs. temporal interpretation in discourse

555

Whereas (11) indicates that the meeting extends over a long period, (12) is simply odd. The meetings do not add. The difference between (11) and (12) can also be put in terms of event type. The event typology that started a tradition was the verb classification of Vendler (1957). Since, it has been argued convincingly (e.g., in Verkuyl 1972) that event type is not a verbal property. Examples of the four Vendlerean classes are given below. STATES (express no change): to be alive, to be in the garden ACTIVITIES (express unlimited change): to walk in the garden, to dance (for hours) ACCOMPLISHMENTS (express limited change): to eat an apple, to walk a block (*for hours) ACHIEVEMENTS (express punctual change): to explode, to reach the top In recent literature, usually a tripartition is made consisting of states, processes and events. In this paper, we solely distinguish between states and events instead of these four classes. Roughly, our states correspond to the two Vendlerean classes states and activities while our events consist of the classes of accomplishments and achievements taken together. In this terminology, (11) is a state, and (12) an event. The distinction made here between events and states may be reminiscent of a similar distinction in the nominal realm. Two major lexical classes of nouns have been acknowledged traditionally: count nouns and mass nouns. Count nouns, which can be quantified over, resemble events and mass nouns, which cannot be quantified over, resemble states (cf. Krifka 1987). There are two important differences between nominals and temporals, however. First, the syntactic unit that leads to the determination whether a nominal or a temporal is mass or count (i.e. state or event) is different. In case of a nominal this is largely lexically determined. In case of a temporal the event type cannot be established before all kinds of components of the entire sentence have been taken into account. Compare the following sentences. (13)

a. b.

Roderick drank a cup of coffee (*for hours), Roderick drank coffee (for hours).

(14)

a.

Steven stuurde een klacht naar ledere commissie die de stad rijk was. Steven sent in a complaint to every committee in town, Steven stuurde een klacht naar iedere commissie die in het leven werd geroepen.

b.

556

Alfons Maes and Leonoor Oversteegen

Steven sent in a complaint to every committee that was called into being. In (13a) the direct object is a count noun, this leads to the sentence expressing an event. In (13b) the direct object is a mass noun. As a result the entire sentence expresses a state. The fact that the indirect object in (14a) denotes a closed set and in (14b) an open set is crucial to the determination of event type: sentence (14a) is considered to express an event and (14b) is considered to express a state. Event type, as it seems, is determined compositionally. Whereas the first difference between nominal and temporal reference types was intra-sentential, the second difference is inter-sentential. The event type property of a sentence often defeasibly determines its temporal relation to a neighbouring sentence, as is demonstrated in (15). (15)

a. b.

Allard came in. The director was sitting behind his desk, (overlap} Allard came in. The director got up. (sequence)

The state-character of the second sentence in (15a) determines the overlap relationship between the two sentences, whereas the event character of the second sentence in (15b) is responsible for a continuation relationship. This kind of inter-sentential influence does not occur in the interpretation process for nominals: two nominals may be anaphoric (partly or entirely identical reference-wise) or non-anaphoric; the mass-count distinction for nominals can not trigger other types of relation. 3.2.3.

Reference points as referents

A second kind of referents of temporal expressions are points of reference. They were originally introduced by Reichenbach (1947) in order to be able to distinguish between several linguistic expressions that denote the same relation between event and speech time, like a simple past, a present perfect and a pluperfect. The reference point is introduced by tense. Tense establishes a reference point with respect to the point of speech. But tense in itself is not specific about the position of the reference point; it may be further specified (or identified) by temporal adverbs. We already mentioned the parallel between nominal and temporal anaphoric approaches. In all approaches to tense as an

Nominal vs. temporal interpretation in discourse

557

anaphor, it is this reference point that is considered to be the anaphor in an anaphoric relation. Simple past tense introduces a reference point preceding the point of speech. Therefore, past tense sequences can be interpreted anaphorically. Perfect, not being a tense but an aspect, merely introduces an event and not a reference point. Aspect determines the relation between events and existing reference points. Therefore, a perfect cannot be used anaphorically. It is difficult to consider reference points as having a substance which is comparable to nominal referents, i.e., discrete real world objects - whether concrete or abstract - with a conventionalised meaning. Rather, they should be considered as technical devices that are part of the representation structure. What exactly would a reference point be? It is no more than a position in time, getting its "meaning" from its place in the ordering and its relations to any or all of the following: (i) other reference points, (ii) the point of speech and (iii) the calendar. If a reference point maintains a relationship with the latter, it will inherit some kind of semantic content.

3.3.

Conclusion

Nominal and temporal expressions have different referential properties, notwithstanding the apparent similarities. Whereas unambiguous nominal expressions have only one type of referent, i.e. (sets of) entities - be it mass or count - temporal expressions have two of them, i.e. (sets of) events/states and reference points. Events/states are comparable with entities with respect to their properties and semantic content. Events/states can be attributed to reference points, as truth conditional semantics has it, and these reference points are incomparable to any semantic dimension of nominals. Theoretically, it would have been conceivable that we would have to indicate the spatial or temporal dimension of an entity as expressed by a nominal, but this does not happen to be the case in our target languages. Consequently, we call temporal expressions "schizophrenic" with regard to reference.

4.

Three linguistic consequences

We have claimed that the ontological differences described in section 3 provide us with explanations for several linguistic phenomena. We will substantiate this claim by giving three examples.

558

Alfons Maes and Leonoor Oversteegen

4.1. The conception of focus and centering 4.1.1. Nominal vs temporal focus Focus plays an important role in many theories on reference. Focus is conceived of as the abstraction of the participants' focus of attention. In the conception of Grosz and Sidner (1986) it is meant to entail those objects, properties and relations that are most salient at a certain point. The focus component itself is parasitic upon the intentional structure of discourse, in that pushes and pops of salient elements in discourse are largely determined by the intention underlying the discourse segments. In order to be able to constrain the reader's search for possible referents of definite NPs and pronouns, the focus stack should be the place where possible referents are listed and hierarchised on the basis of their accessibility for future reference. For hierarchising the elements in the focus stack, readers rely on contextual triggers, such as syntactic position, recency of mention, persistence of referents and the presence of competing candidates. The way in which these triggers guide the internal organisation of the focus component in discourse can be systematised and modeled by rules and constraints. The centering approach is one way of modeling this process. This approach is presented by, e.g., by Grosz, Joshi and Weinstein (1995) and Brennan, Friedman and Pollard (1987). The core of the centering approach is a set of rules and constraints regulating the relationship between what the discourse is about and the linguistic choices made in terms of grammatical function, syntactic structure and type of referring expression. A similar focus component can be proposed for temporal interpretation, as is done in Webber (1988). She proposes a dynamically changing temporal focus to capture a similar intuition that at any point in the discourse, there is one temporal entity that is most likely to stand in an anaphoric relation with the reference time of the next clause. 4.1.2. Differences between nominal and temporal focus Contextual clues in determining the order of foci Nominal and temporal foci are fundamentally different with respect to the way in which the order of foci can be determined or calculated. The ordering of nominal referents can be seen as a process guided by fixed criteria, such as syntactic position, thematic position, types of referring expression used

Nominal vs. temporal interpretation in discourse

559

etc. So, determining John as the referent of he in (16b) and as the focus of attention is primarily suggested by the syntactic and thematic position of John in (16a). (16)

a. b. c. d.

John went over to Paul's house. On the way he stopped by the grocery shop for a bottle of whisky. He bought a Johny Walker black label because he knew that his friend was depressed. But unfortunately, the bottle failed to cheer him up.

For temporal interpretation, it is much harder to imagine what characteristics of the utterance can function as the organising principles of reference points in a temporal focus stack. To start with, it is hard to imagine what properties of reference points would have a say in the preference ordering of temporal foci. Being semantically empty, reference points don't have thematic status. As for syntactic clues, apart from recency in the linear order of utterances, the only relevant syntactic information deciding on the position of a reference point with respect to the stack would be the answer to the question whether the reference point is evoked by a main clause or by a subclause. As may be obvious from the following example, however, for the complex sentence (17) both the continuation of the main clause (a) and the continuation of the subclause (b) are acceptable. (17)

Jane bleefthuis omdat Tarzan de auto total loss had gereden. Jane stayed at home because Tarzan had wracked the car. a. Zij wilde trouwens graag Baywatch zien. She enjoyed being able to watch Baywatch. b. Hij reed hem tegen een boom. He drove it into a tree.

Even relative clauses may be continued, as (18b) and (19b) demonstrate: (18)

De agent die hetproces verbaal had opgemaakt, was ziek. The policeman who had made the report was ill. a. Hij lag met griep in bed. He was down with the flu. b. Hij had late dienst gehad die nacht. He had been on late duty that night.

560

(19)

Alfons Maes and Leonoor Oversteegen

De agent, die 's ochtends niet ontbeten had, accepteerde een broodje. The cop, who had not had breakfast in the morning, accepted a sandwich. a. b.

Het smaakte als poetskatoen. It had a taste of cottonwool. Hij was te laat opgestaan. He had woken up too late.

The above data show that the most plausible contextual clue in determining temporal focus, syntactic position, does not yield any predictive value. The cognitive value of the focus of attention Both nominal and temporal focus components should guide local coherence in that they direct the transitions between meaningfull segments in discourse and signal discourse continuations and shifts. From a cognitive point of view, the relational network provided by the successive nominal focus stacks in discourse links up to global coherence in that it enables readers to deduce the global discourse topic. The reason for that is that the succession of local foci can be related in a meaningful way to the global topic of discourse: a local focus is the topic of discourse itself or it is related to the topic of discourse in a way which is suggested by the organisation of local foci in successive focus stacks. It is not clear how local temporal foci contribute to the global temporal focus. As with nominal focus, temporal foci contribute to the local coherence of discourse, as is convincingly argued for in Webber 1988. Unlike nominal focus, however, the successive temporal foci do not link up with the global temporal focus. It is not clear how a global temporal focus is related to, or brought about by the dynamically changing local temporal foci consisting of types of events, speech points and reference points.

4.2.

Overspecification of referents

The ontological differences we discerned can explain a difference in the use of so-called overspecified expressions referring to nominal and temporal referents. Look at the following examples: (20)

Clinton^ has announced to cut the military budget. It is the first time this president,· has dared to do so.

Nominal vs. temporal interpretation in discourse

(21)

561

When the washing machine^ first came on the market, this fantastic and truly revolutionary aid for the housewife^ was available only to very few. (Bosch 1983: 150)

The use of the anaphoric NPs in these examples can be considered to be overspecified: as the referent behind the coreferential NPs clearly is the focus of attention, the use of a strong anaphoric device (i.e. the nominal NPs) is not necessary in order to simply identify the intended referent. That's why they can be called overspecified. In other work (see Maes 1996, Maes and Noordman 1995), this overspecification is claimed to have a non-identificational function: in (20) for example it is used to modify the interpretation of the referent: the use of this president enriches the semantic representation of the referent by activating the class-interpretation of the referent. It is striking that this type of overspecification is natural for NP-expressions referring to nominal entities, whereas it is unusual for NP-expressions referring to temporal entities. This can be validated by two observations, (i) Many NP-expressions, pointing back to an explicitly introduced temporal referent, suggest non-identity. When we look at the following examples, the identity between the coreferential nominal NPs in (22) is taken for granted, whereas this is not the case with the coreferential temporal NPs in (23): (22) (23)

Yesterday the train to Vlissingen^ was derailed. The intercity^ landed in a ditch. Yesterday, the train to Vlissingen was derailed. It landed in a ditch on the 30th November?i.

(ii) NP-expressions, which clearly overspecify or modify temporal referents, as opposed to simply identifying them, are often awkward, as can be seen in (24): (24)

On November 22 1963^ John Kennedy got killed in Texas. On this bright afternoon-^, he was shot in his car.

These observations fit in with the fact that in a large corpus of written discourse (see Maes 1996), none of the demonstrative nominal anaphors is an overspecified temporal expression. The only temporal NPs are simple identifiers, such as at that moment, in that year, etc. There are two ontological differences between nominal and temporal referents which can explain these differences in the occurrence of an overspecificational effect. First of all, nominal entities - unlike temporal ones - appeal

562

Alfons Maes and Leonoor Oversteegen

to real world objects with Stereotypie features and characteristics attached to them. They occupy a relatively fixed place in the semantic network within the mental lexicon. Consequently, the ontology of nominal referents is suitable for being semantically modified by these attributes, whereas the ontological emptiness of temporal reference points makes such modification unnatural. A second characteristic of referents which is indispensable for the overspecificational effect is that the identification of the referents can be taken for granted. If this is not the case, the anaphoric expression is simply taken to be necessary in order to identify the referent. So the effect of overspecification lives on the fact that readers expect the following sentences in the examples mentioned to be about the referent which is the focus of attention. This expectation is exactly what all theories of nominal focus have incorporated: continuation of the same focus is considered to be the default situation. This default focus is mostly expressed by a pronoun, which acts as a simple placeholder for the entity in focus. For temporal focus, such an identity default does not exist. The default temporal focus is ordered through the precedence relationship with the current focus; this suggests continuation in time: time appears to move on as the discourse proceeds. This default underlies Dowty's discourse principle of temporal interpretation TDIP (Dowty 1986). Thus, contrary to nominal focus, this continuation default by no means expresses identity of focus. Consequently, a temporal focus can hardly be said to be established, in the sense in which a nominal focus can be said to be established.

4.3.

Perfect and indefmiteness

The previous topic concerned established referents, i.e. referents already introduced in the discourse. Nominals and temporals were shown to differ because of the ontological difference of their referents. The ontology gives no reason to expect a discrepancy between nominals and temporals with respect to new referents. Both nominals and temporals have linguistic forms indicating that a referent is newly introduced. Indefinite NPs and Perfects have exactly this in common: they are both considered to introduce a new discourse referent. The former an entity, the latter an event.1 Note (i) that we can refer anaphorically to an event in Perfect, and (ii) we can expand an event in Perfect. As we have seen, and as has been assumed in much of the literature since Reichenbach, a Perfect does not introduce a

Nominal vs. temporal interpretation in discourse

563

reference point. For this reason indefinites and Perfect usually both introduce their referents in the discourse. And for the same reason, usually they don't designate an already mentioned referent. Sometimes they do however. Since the function of an indefinite NP is establishing a discourse referent, it might be expected that indefinite NPs referring to an already mentioned entity, again, have to serve another purpose. And they seem to do so, according to the relevant literature (Hinds 1977; Ushi 1986; Sanders 1990). Ushi (1986) argues that an indefinite expression can very well be used coreferentially with a preceding expression and that it then serves to present the already mentioned referent in a new light and from a different perspective. Take the following example: (25)

Then she (i.e. Mrs. Miller) met Miriam^. It was snowing that night. [...] A long line stretched in front of the box office; she took her place at the end. The line seemed to be taking its own time and, looking around for some distraction, she suddenly became aware of a little girl standing under the edge of the marquee ^

The referent Miriam is re-identified from the speakers point of view as a little girl standing under the edge of the marquee: the indefinite NP represents the perspective of the protagonist. A similar shift of perspective can be accomplished with the help of an introductory temporal: (26)

De Telegraaf heeft het nieuws geweigerd. De redactieleden meenden dat de bron onbetrouwbaar was en ze besloten unaniem tot nietplaatsing. 'De Telegraaf has refused the news. The members of the editorial staff thought the source was unreliable and they decided unanimously not to put it in the paper. a. b.

Dat speet ze later. Later, they were sorry Dat heeft ze later gespeten. Later, they have been sorry.

Whereas (26a) represents the story in a continuous way, (26b) involves a temporal rupture: the Perfect cannot adhere to a reference point present in the context since those are reference points in the past. Moreover, the content of (26b) is interpreted rather than presented. It is as if the speaker is reflecting

564

Alfons Maes and Leonoor Oversteegen

on the paper's reaction, while the rest of the story was not written from a transparent perspective.

5.

Conclusion

The rather ambitious goal of this paper was to discourage a strong parallelism view on nominal and temporal expressions. We hope to have shown that the observations made in section 2 concerning similarities between the behaviour of nominal and temporal expressions may find a counterbalance in the phenomena presented in section 4, which stress the differences in the behaviour of nominal and temporal expressions. These differences were explained in terms of ontology: the structure of the domain of reference of nominal vs. temporal expressions. These ontological differences trigger many predictions - more of them than the few that have been presented in this paper.

Notes 1. We are aware of the fact that recent studies on temporal discourse semantics often have Perfect as a stative operator, that is: a device to turn any event into a state. We will not comment on the correctness of this regularity for English but for Dutch it certainly does not hold.

References Anderson, John and Reid Hastie 1974 Individuation and reference in memory: proper names and definite descriptions. Cognitive Psychology, 6: 495-514. Bosch, Peter 1983 Agreement and Anaphora. A Study of the Role of Pronouns in Syntax and Discourse. London: Academic Press. Brennan, Susan E., N.W. Friedman and Carl J. Pollard 1987 A centering approach to pronouns. In Proceedings of the 25th Meeting of the Association for Computational Linguistics, 155-162. Stanford, California. Cloitre, M. and Thomas G. Bever 1988 Linguistic anaphors, levels of representation and discourse. Language and Cognitive Processes, 3, 293-322. Dowty, David 1986 The effects of aspectual class on the temporal structure of discourse: semantics or pragmatics? Linguistics and Philosophy, 9, 37-61.

Nominal vs. temporal interpretation in discourse

565

Garrod, Simon and Anthony J. Sanford 1977 Interpreting anaphoric relations: The integration of semantic information while reading. Journal of Verbal Learning and Verbal Behavior, 16, 77-90. Grosz, Barbara J., Aravind K. Joshi and Scott Weinstein 1995 Centering: a framework for modelling the local coherence of discourse. Computational Linguistics 21: 203-225. Grosz, Barbara J. and Candace L. Sidner 1986 Attention, intentions and the structure of discourse. Computational Linguistics, 12, 175-204. Hinds, John 1977 Paragraph structure and pronominalization. Papers in Linguistics, 10, 77-99. Hinrichs, Erhard 1986 Temporal anaphora in discourses of English. Linguistics and Philosophy, 9, 63-82 Kamp, Hans and Uwe Reyle 1993 From Discourse to Logic. Dordrecht: Kluwer. Krifka, M. 1987 Nominal reference and temporal constitution: towards a semantics of quantity, in J. Groenendijk, M. Stokhof and F. Veltman (eds.) Proceedings of the Sixth Amsterdam Colloquium. 153-173. Lascarides, Alex 1990 Knowledge, causality and temporal representation. Linguistics, 30, 941-973. Maes, Alfons 1996 Nominal Anaphors, Markedness and the Coherence of Discourse. Peeters: Leuven. Maes, Alfons and Noordman, Leo 1995 Demonstrative nominal anaphors: a case of non-identificational markedness. Linguistics. 33(2). 255-282. Oversteegen, Leonoor 1992 Tense and temporal structure. Think, 1, 46-59. Partee, Barbara 1984 Nominal and temporal anaphora. Linguistics and Philosophy, 7, 243-286. Poesio, Massimo and Migumi Kameyama 1992 Temporal centering, (manuscript). Reichenbach, Hans 1947 Elements of Symbolic Logic. New York: The Free Press. Russell, Bertrand 1903 The Principles of Mathematics. Cambridge: Cambridge University Press. Sanders, Jos6 1990 Expliciet of niet? Referentie-bepalende factoren bij personen in nieuwsberichten [Explicit or not? Factors determining reference to persons in news reports]. 77T, 9,159-180. Scha, Remco and Livia Polanyi 1988 An Augmented context free Grammar for Discourse. Proceedings of the 12th International Conference on Computational Linguistics. Budapest. 573-577.

566

Alfons Maes and Leonoor Oversteegen

Seuren, Pieter A.M. 1985 Discourse Semantics. New York: Basil Blackwell. Ushie, Y. 1986 'Corepresentation' - A textual function of the indefinite expression. Text. 6. 427^46. Vendler, Zeno 1957 Verbs and times. The Philosophical Review. 66. 143-160. Verkuyl, Henk J. 1972 On the Compositional Nature of the Aspects. Dordrecht: Reidel. Webber, Bonnie L. 1988 Tense as a Discourse Anaphor. Computational Linguistics, 14, 61-73.

Markedness and prototypical speaker attributes Willy Van Langendonck

0. In this paper, I would like to deal with the notion of markedness in an experiential framework. Following Mayerthaler (1980; 1988), I will derive unmarked categories from prototypical speaker attributes. Adapting and extending this concept, I will apply it to phonology and to the nominal area of personal pronouns and proper names. Attention will also be paid to the much discussed phenomenon of markedness assimilation. As Henning Andersen (1989: 11) puts it: "Markedness has been hailed by some as one of the most significant conceptual advances of twentieth-century linguistics, and it has been denounced by others as an entirely vacuous concept". This may partly be due to differences in linguistic background. But even if one's attitude is positive, there are differences of opinion or terminology. To enhance confusion, affiliated notions have popped up in recent times. The notion 'unmarked' appears to have something in common with, e.g., prototypes, stereotypes, default, basic level, ground (as opposed to figure). We will limit ourselves to some considerations about unmarkedness and prototypicality. 1. As for markedness itself, we can set out from the common view that the marked - unmarked relation is an asymmetric one that applies to certain oppositions. This asymmetry is especially embodied in the fact that the unmarked member can stand for the supercategory. For instance, the unmarked adjective long can be used in the wider sense of the supercategory length (as in how long ...; a meter long), comprising the specific senses long 'above the norm' and short 'under the norm'. On what kind of oppositions exist, the two pioneers of markedness theory did not agree. For Jakobson every opposition was a binary one or reducible to one, while Trubetzkoy developed three types of (phonemic) oppositions: privative (e.g. +/— voice), gradual (referring to different degrees of some gradient property, e.g., i-e - ce) and equipollent (each member has a mark that the others lack, e.g. p -1 - k) (Andrews 1990: 14). Obviously, Trubetzkoy's trichotomy comes closer to the basic tenets of Cognitive Linguistics because of the recognition of scalarity.

568

Willy Van Langendonck

It is true that the notion of unmarked member resembles that of prototypical member. However, prototypes are the best examples of some category or lexical item, whereby the number of members or variants is usually big and even undetermined (cf. Geeraerts 1989). So we do not have to do with a contrast set as in the case of markedness. For instance, we can posit that the unmarked grammatical relation is the subject, and on the other hand, that the prototypical subject is an agent-topic combination. More specifically, markedness concerns two or more different words, classes, features, phonemes or other expressions, while prototypicality pertains to the different applications of the same class or item. Hence we can speak of the prototypical bird, but not of the unmarked bird. By contrast, we should speak about the (un)marked member dog in such a pair as dog - bitch. On the other hand, the unmarked member could be seen as a kind of metonymic prototype. According to Lakoff (1987, 288-289), a prototype effect can arise in a metonymic way: "Given category B, where A is either a member or subcategory of B, suppose that A metonymically 'stands for' B. That is, it is either a social stereotype, or a typical case, or an ideal, or a submodel, etc. Then, A will be a best example of B". Thus, long (A, specific sense) is the best example of length (B), and can therefore stand metonymically (pars pro toto) for the wider sense. Other connections between 'unmarked' and 'prototypical' can be observed. Sometimes, there is a special relationship between the unmarked category or element of a pair or set and the prototypical member of a class. For example, take grammatical relations again. It is obviously not a coincidence that the prototypical subject is the intersection of agent and topic and that, at the same time, the agent is the unmarked semantic role and the topic the unmarked member in the pragmatic pair topic - comment. Another example is the indirect object construction. We could argue that the pattern exemplified by / gave you a book (without a preposition) is the prototypical indirect object pattern, and then also that the prepositionless indirect object is unmarked with regard to a PP counterpart as in / gave a book to you. So unmarkedness and prototypicality seem to go together in a so-called idealized cognitive model such as the definite human agent-topic constellation (see Lakoff 1982; 1987: ch. 4). It follows that the difference between unmarkedness and prototypicality is not so essential but rather complementary. Indeed, it appears that both unmarked and prototypical items usually display a greater lexical or text fequency. This similarity between the two concepts is exploited by Janda (1996). She tries to explain markedness in terms of prototypicality: "Markedness correlates with distance from the prototype of a category, with

Markedness and prototypical speaker attributes

569

the least marked elements closest to the center, and the most marked elements in the most peripheral positions" (p. 209). Further, "The position of the prototype in the radial category explains the priority of the unmarked element" (p. 216). This explanation still leaves us with the question as to why such phenomena as center and periphery should exist at all as conceptual tools in language. 2. Maybe this question can be brought closer to an answer if we look for markedness (or center - periphery) phenomena outside language. Battistella (1990) devotes his last chapter to extralinguistic markedness phenomena in various cultures, without, however, intending to explain linguistic markedness phenomena on this basis. If we want to get to an explanation, a further and most important step forward seems to me the grounding of linguistic markedness phenomena in an experiential reality. It can still be observed that a lot of recent publications adhere to some objectivist view of markedness (Andrews 1990; Kiefer 1989), denying any isomorphism between linguistic and extralinguistic categories, and distinguishing between linguistic and extralinguistic meaning. This is especially obvious in the realm of semantic markedness, which in the objectivist approach is seen as a kind of structural complexity, whereby unmarked = jc, and marked = + n. In my view, the experiential approach of Mayerthaler (1980; 1988) does more justice to the linguistic facts. To determine what is semantically marked, Mayerthaler resorts to the biological, perceptual and cultural attributes of the PROTOTYPICAL SPEAKER (PS), also including factors of the particular speaker situation. His definition is as follows: "A semantically less marked category (sem(Ki))' — def category Ki that reflects prototypical speaker attributes" (Mayerthaler 1988: 9). For instance, since the PS is animate and is at the summit of the hierarchy of the animate via propagation of his own kind, it is true that a semantically more animate subject is unmarked vis-ä-vis a less animate one (ibid.). Another striking example is the markedness relation among spatial dimensions. In an objectivist view, we would expect one-dimensionality to be least marked and three-dimensionality to be most marked. Yet, the reverse appears to hold. The reason is simply that we ourselves consist of three dimensions and live in a three-dimensional world (Mayerthaler 1980: 23). This squares with the order of acquisition of the three basic local prepositions: in > on > at. However, at least one qualification to this interesting approach is in order. It seems to me that the experiential basis is not biology, let alone scientific

570

Willy Van Langendonck

biology, but rather the way the PS prototypically perceives himself, other people and his natural and cultural environment. Finally, we might try to extend this experiential approach to phonology. In my view, it is plausible to consider vowels unmarked with respect to consonants. This can be traced back to the fact that the vocalic feature-values [-1- voice, + sonorant, + continuant, 4- syllabic] are better perceptible and pronounceable than their negative counterparts. Together, they constitute an ideal cluster, a kind of idealized cognitive model (cf. Van Langendonck 1986). 3. Since markedness has to do with a contrast set, markedness relations are often reflected in specific formal characteristics where the above asymmetry is embodied. The marked counterpart of an opposition often displays more formal bulk (the iconicity principle, whereby 'zero form' plays an important part). Syncretization and defectivation have to do with gaps in the marked value. In the case of extension, the form of the unmarked value is used for the marked counterpart. The unmarked form also shows up when an opposition is neutralized. More or less the same battery of formal characteristics seems to apply to prototype effects (Battistella 1990: 43). Often it is claimed that the formal tests we mentioned are independent of the establishment of markedness relations or that they are merely symptoms. It is true that they do not always point in the same direction. Furthermore, their application to phonology and to the level of the sign may differ. The argument of formal bulk cannot be applied to phonology for obvious reasons. On the other hand, Andrews (1990: 143) claims that neutralization applies only to phonology but syncretization only to morphology. However, I do not wish to endorse her "narrowly defined markedness theory" (p. 147-157) and would like to attach some more weight to the above formal diagnostics. Though interesting for both markedness and prototypicality, the use of independent (but still linguistic) evidence such as facts of linguistic change, language acquisition, aphasia and linguistic typology is much more delicate and intricate. For one thing, "... when systems change, existent markedness relationships are disturbed. Very often these disturbances lead to markedness reversal" (Tomic 1989: 3). Another problem is constituted by formal bulk, especially by zero sign. Can we speak of zero sign when a category is not expressed in a given stage of language acquisition? For it might be that the category is still entirely absent at that stage. For example, what does it mean that in the acquisition of Dutch, the indefinite article surfaces one month ear-

Markedness and prototypical speaker attributes

571

Her than the definite one? Therefore, a better criterion is to find out which value of some category is first acquired in all its forms and functions. 4. Finally, we have to mention a remarkable and somewhat embarrassing phenomenon in the realm of markedness. Typical of markedness relations is their contextually conditioned reversibility and assimilation, also found outside language (Battistella 1990: 5, 7; but see Garcia, this volume). As Janda (1996: 218) puts it: "The cooccurrence of unmarked elements is frequently referred to as 'markedness assimilation', and the cooccurrence of marked elements has been labeled both 'markedness assimilation' and 'markedness reversal', the reason being that in a marked context, the appearance of more marked elements is unmarked, so the value appears to be reversed". She then calls both phenomena 'markedness alignment' and reduces it to the metaphorical mapping of Cognitive Linguistics: "Similar mapping operates between categories, producing the alignment of central members with central members and of peripheral members with their marked counterparts" (p. 218). This time, the author looks for an extralinguistic explanation: the postulation of this type of mapping would be "consistent with what we know about mapping functions between neural nets in the brain" (p. 218). Nonetheless, the existence of such principles as markedness assimilation and reversal raises the problem of circularity. As Battistella (1990: 114) observes, in this way it becomes difficult to know "which principle applies when". Let me adduce an example from phonology. Although a moment ago, I claimed the plus-value of the feature [+/- voice] to be unmarked, Trubetzkoy and the Prague School contended the reverse, exemplifying their thesis by the obstruent pairs b-p, d-t, g-k etc. We would now say that with obstruents, a markedness reversal takes place, since in this case, all four features [+/— voice, +/— sonorant, +/— continuant, +/— syllabic] display the negative values. At the same time, we can observe here a markedness assimilation or alignment of plus-values and minus-values, respectively. This appears to constitute a case of iconicity in that similar values assimilate to each other and form a cluster or 'gestalt'. However, in order to know which values are primary and which are reversed, the independent criterion of PS attributes has to be resorted to. We can observe then that, independently of context, the PS will perceive [+ voice] better than [— voice], so we can hold that in this case, the plus-values are primary. Summarizing my approach to assessing markedness relationships, I would like to set out from the PS and to see whether the result of this analysis

572

Willy Van Langendonck

squares with the battery of linguistic markedness criteria. This battery can be considered a prototype in itself, i.e. they need not be deemed necessary and sufficient conditions. It appears that especially the occurrence of iconic congruence relating to formal bulk is not obligatory. This optionality also goes for markedness alignment, where iconicity is involved as well. The alignment may or may not occur. 5. Let us illustrate this approach by an investigation of nominal categories (pronouns and proper names) and the corresponding features. First, it will be argued that the unmarkedness of such features as referentiality, specificness, definiteness, concreteness, humanness and semantic singularity can be grounded outside language, i.e., in PS attributes. From this, the unmarkedness of personal pronouns and proper names with regard to common nouns will be derived. Least marked seem to be singular personal pronouns. 5.1. The PS and other people and things he observes are in the first place referents, not senses. Hence, reference is unmarked, sense is marked. Indeed, sense can be seen as an abstraction from reference. Looking for nominal categories that are only referential and lack a sense, we encounter pronouns (at least personal pronouns) and proper names. It does not make sense to ask: 'What do you understand by he or sheT or 'What is the (synchronic) meaning of John or California? 5.2. On the other hand, personal pronouns and proper names are mostly called inherently definite-specific. This is confirmed by at least one test for definiteness. In a number of languages, only definite-specific NPs can appear as an 'afterthought' (Van Langendonck 1979; see also Löbner 1985: 299300), compare (1)

a. He did that, he / Peter/ the prisoner! b. *He did that, a prisoner, some prisoner!

The primary unmarked value of the opposition [+/— definite] in NPs appears to be [+ definite] (Van Langendonck 1979; Mayerthaler 1980: 33; 1988: 8). The following criteria can be adduced for this claim. (a) PS attributes: definiteness is associated with the obvious property of the PS that he is presupposed to exist and to be unique and acquainted with the hearer, at least in the speech situation.

Markedness and prototypical speaker attributes

573

(b) Formal bulk: many (pro)nominal categories have zero sign (an indication of unmarkedness), e.g., personal proper names (John, Mary), settlement and country names (London, Britain), month names (June, July), etc., and also all kinds of pronouns, except of course the indefinite ones. However, the latter appear to consist of more than one morpheme: some-body, any-one, no-body, some-thing, any-thing, etc. Furthermore, in many languages, indefinite but not definite pronominalization needs two morphemes, e.g., in French en un, as in il en a donne un, un livre 'he has given one, a book', in Dutch er een as in hij heefi er een gegeven, in Goethe's German deren eins as in er hat deren eins gegeben (as opposed to English he has given one). This again marks the complexity of 'indefinite'. (c) Syncretization (typical of a marked status): whereas in the Dutch singular there is but one form for the indefinite article (een), there exists a neuter (het) and a non-neuter form (de) of the definite article. Another syncretism is to be found in such Dutch patterns as ik heb (de) kaas gegeten Ί have eaten (the) cheese', where the [+/— holistic] distinction is syncretized with indefinite nonspecific, not with definite NPs: a. definite [holistic]: ik heb de kaas (op)gegeten Ί have eaten (up) the cheese' b. definite [partial]: Ik heb van de kaas (* op)gegeten Ί have eaten part of the cheese' c. indefinite nonspecific: ik heb (* van) kaas gegeten Ί have eaten (part of) cheese * ik heb kaas opgegeten Ί have eaten up cheese' (but indefinite specific: ik heb van een bepaalde kaas gegeten Ί have eaten part of a specific cheese) (d) Defectivation: the marked value may lack a subclass. As has been mentioned, proper names are inherently definite. There is no indefinite subclass. Likewise, there is only one indefinite class of pronouns (somebody, something, etc.) whereas all others are definite (personal, demonstrative pronouns, etc.). (e) Extension: in indefinite NPs with so-called absolute superlatives, the definite article may show up: these people suffered the greatest hardship possible. (f) Neutralization: in case the opposition [+/— definite] is neutralized, definite forms like it and there surface, e.g. it is raining, there exist people who ... Furthermore, the term 'definiteness' itself is often used for the whole feature [+/- definite]. These two kinds of neutralization again point to the

574

Willy Van Langendonck

unmarkedness of [+ definite]. In French, the definite article le/la can be used in such expressions as c'est leprintemps 'it is spring'. (g) In language acquisition, it has been shown that the complete mastering of the indefinite article lasts longer than that of the definite counterpart (cf. Karmiloff-Smith 1979, for French). Finally, even in recent logical analyses, it is contended that definite NPs, unlike indefinite NPs, are not quantifiers (Löbner 1985; Krifka 1992). Translated into (a more objectivist) markedness theory, this means indefiniteness is marked. When the two unmarked categories, referentiality and definiteness, go together, they form a natural combination and constitute a case of unmarkedness assimilation. According to the same alignment principle, when there is only a sense and no referent, a definite NP is marked, viz. the predicate nominal in such he-sentences as Mohammed was the greatest. Unmarked again is, e.g., Mary is a woman, where [— referential] and [— definite] go together. Another reversal occurs in the direct object. Since definiteness naturally goes together with subjecthood, it does not with the direct object, so in this position, definite is the marked value. In a number of languages, e.g., Turkish, a definite direct object does receive a special mark (Comrie 1989: 132). 5.3. Other unmarked features are humanness and concreteness. It is to be expected that since the PS can perceive concrete but not abstract entities, the category 'concrete' will be unmarked. The subclass 'human' of concrete things is obviously least marked. Again we see that proper names and personal pronouns fit this feature. First and second person pronouns inherently refer to humans; the third person pronouns he and she too, at least when they are stressed. Except for temporal names like June or 7993, all proper names refer to concrete things. The least marked proper names are undoubtedly personal names. They normally have zero expression for all the features in question. 5.4. It is also often claimed that proper names inherently display singular number. Therefore, if we take proper names and singular personal pronouns together, we can posit that they constitute the least marked nominal class, since they share such PS properties as referentiality, definite specificness, concreteness (and mostly humanness) and semantic singularity (but see Battistella 1990: 84, 112 and my reply to this in Van Langendonck 1994a). This is further corroborated by the criterion of the extension of the unmarked.

Markedness and prototypical speaker attributes

575

The above unmarked class allows for a systematic synchronic extension to its marked counterpart, i.e. common nouns: we can say, e.g., the I of the narrator, a new I, a new me, a new you, a he, a she; a John, a Mary, another London, etc. 6. Conclusion. I have tried to make sense of an experiential approach to markedness, adducing the PS to decide on primary unmarkedness, while allowing for reversals and assimilations. Adhering to the cognitive thesis of meaning-form correspondence, I have taken the usual formal markedness criteria seriously and found out that they fit an experiential approach much better than an objectivist one.

Acknowledgments I thank the audience of the Cognitive Linguistics Congress for their interesting remarks on the version read to them, and also Caroline Greenman for her native speaker advice. For a more elaborate version concentrated on English, see Van Langendonck 1994b.

References Andersen, Henning 1989 Markedness theory - the first 150 years. In: Tomic, O. (ed.): 11-47. Andrews, Edna 1990 Markedness Theory. Durham: Duke University Press. Battistella, Edwin L. 1990 Markedness. The Evaluative Superstructure of Language. (SUNY Series in Linguistics). Albany: State University of New York Press. Comrie, Bernard 1989 Language Universals and Linguistic Typology. Syntax and Morphology. 2nd ed. Oxford: Blackwell. Garcia, Erica C. this volume A cognitive reinterpretation of markedness reversal. Paper read at the Third International Cognitive Linguistics Conference (Leuven 1993). Geeraerts, Dirk 1989 Prospects and problems of prototype theory. Linguistics 27: 587-612. Janda, Laura 1996

Unpacking markedness. In E. Casad (ed.). Cognitive Linguistics in the Redwoods. Berlin/New York: Mouton de Gruyter: 207-233

576

Willy Van Langendonck

Karmiloff-Smith, Annette 1979 A Functional Approach to Child Language. A Study of Determiners and Reference. Cambridge University Press. Kiefer, Ferenc 1989 Towards a theory of semantic markedness. In: Tomio, O. (ed.): 121-138. Krifka, Manfred 1992 Definite NPs aren't quantifiers. Linguistic Inquiry 23/1: 156-163. Lakoff, George 1982 Categories and cognitive models. Paper no. 96, Ser. A. LAUT (Linguistic Agency University of Trier). 1987 Women, Fire, and Dangerous Things. What Categories Reveal about the Mind. University of Chicago Press. Löbner, Sebastian 1985 Definites. Journal of Semantics 4: 279-326. Mayerthaler, Willi 1980 Ikonismus in der Morphologie. Zeitschrift für Semiolik 2: 19-37. 1988 Morphological naturalness. Ann Arbor: Karoma. Tomic, Olga 1989 Introduction. In: Tomio, O. (ed.): 1-10. Tomic, Olga (ed.) 1989 Markedness in Synchrony and Diachrony. Berlin/New York: Mouton de Gruyter. Van Langendonck, Willy 1979 Denniteness as an unmarked category. Linguistische Berichte 63: 33-55. 1986 Markedness, prototypes and language acquisition. In: L. Beheydt (ed.). Langage enfantin = Cahiers de l'Institut de linguistique de Louvain 13: 41-78. 1994a Review-article of E.L. Battistella, Markedness: The evaluative superstructure of language (Albany: SUNY, 1990). Lingua 94: 189-195. 1994b Markedness, experientialism and nominal categories. In: K. Carlon, K. Davidse & B. Rudzka-Ostyn (eds.). Perspectives on English. Studies in honour of Professor Emma Vorlat. Leuven: Peelers, 331-344.

Computational Linguistics

Implementing cognitive semantics - overview of the semantic composition processes and insights into the grammatical composition processes Kenneth Holmqvist 1.

Introduction

Although Langacker's 1987 Cognitive Grammar is a very powerful full-scale model of language, there have been few attempts to implement it on computers. We think that the reason for this is twofold: First, cognitive grammar is presented as a theory of the language as a static entity. It does not focus on the linguistic and semantic processing that an individual has to perform when understanding and producing speech.1 Therefore, anyone who wants to implement language understanding based on cognitive grammar must first devise a processual counterpart to it. The second reason that implementations of cognitive grammar are and will be rare is the general complexity and open-endedness of the theory. It is simply impossible to make any larger computer implementation of cognitive grammar without first interpreting virtually all elements of the theory into some firmer form. Domains, constructional schemata, schema types, scanning modes, predications, valence composition etc. must therefore be given computational counterparts. This is not only difficult, but it also involves the danger of completely rebuilding the theory so that it is adapted more to the needs of the computer than to the linguistic reality behind cognitive grammar. In our project, we base our implementation work on the computational model of cognitive grammar developed in Holmqvist 1993a. This computational model consists of three main components: representations of schemata, semantic composition processes, and mechanisms for valence suggestion and updating of the schema population.

2. 2.1.

Part I: Semantic representations and composition processes Semantic representations

First, there is a computer model of so-called 'image schemata'. Different authors in the cognitive linguistics tradition attach different meanings to the

580

Kenneth Holmqvist

term 'image schema' (in particular, Langacker 1987 contrasts against Lakoff 1987, 1989). In our model of Langacker's 1987 schemata (compare figure 1 below, showing the schema [KNIFE]), a schema consists of: 1. a matrix of domains, ordered by centrality values 2. a list of parts ordered by their saliences 3. a list of wholes ordered by their saliences [KNIFE] has [CUT] and [SILVERWARE] as wholes, because [KNIFE] functions as a part in them. In the same way, [BLADE] and [HANDLE] are parts of [KNIFE]. Parts and wholes are in themselves schemata, and they can therefore be described with the same three elements. Of course, just saying that the [BLADE] is a part of [KNIFE] is not sufficient. We must also characterize this part-whole relation closer. For instance, the relative sizes of the blade and the knife must not deviate outside of certain limits. The relative spatial position of the blade on the knife must also be correct, i.e. the blade must be correctly attached. Holmqvist 1993a discusses these and other questions about the part-whole relations in schemata and suggests implementable solutions. Most of the content substance of a schema resides in the domains, which collectively form the matrix. Typical domains are: colour, spatial form, material, age, temperature, profession, emotion, etc. Domains are made up by dimensions. The colour domain has three dimensions (hue, saturation and brightness), while temperature has one dimension. Spatial form has a varying number of dimensions, because of the varying number of forms. Profession and emotion domains probably have rather abstract dimensions. The schema predicates in each domain in its matrix. In the spatial form domain, [KNIFE] predicates a ID-directed form. In the colour domain, [KNIFE] may predicate a brown colour. A predication such as a ID-directed form or a colour picks a subset of the total number of dimensions in the domain and attributes values to the dimensions in that subset. For example, a ID-directed form attributes a value to the spatial extension dimension.2 Dimensions and their values need not refer to measure units in the external world. Their main function is to serve in a mathematical region activation model underlying the representation. In all domains, a predication is equated with a region. The simplest case is that of regions in colour space. By using a mathematical model where regions are characterized by prototypical points, the model allows for variability. There is thus a range of brown colours just like a range of ID-directed forms. The variability is constrained by there

Implementing cognitive semantics

581

Wholes (abstract domains) Are also schemata

Predication GoofigunliDmu/locatjpnal

Dimensions with values

Arc auo schemata

Domains Set of dimensions with • metric and « region mccnjlliuin

Domains ordered by cbcir Generality

Figure 1. Core of the computational representation of [KNIFE].

being other regions than the predicated one in the domain. If the colour dimension values (i.e., the colour point) of the knife would move outside of the brown region, another region would be more saliently activated and there would be a conflict between that region and the brown predication in the schema. All semantic constraints regarding predications are intended to be withheld by this region activation model. The example here has concerned the noun (more precisely: thing) [KNIFE], but the same representation form also applies to other schema type. A verb (i.e: process) like [CUT] also has its parts, wholes and domains. The domains of [CUT] are different from that of [KNIFE] in several ways, the most important of which being that [CUT] has temporal dimensions in its predications. These temporal dimensions are what makes [CUT] processual.

582

Kenneth Holmqvist

Figure 2. Superimposition for the construction and evaluation of composites. [ROPE], [STRETCHED], [ACROSS] and [ROAD] superimpose during the evaluation of two valence relations.

2.2.

The semantic composition processes

The second part of the computational model consists of a process for the construction and evaluation of composite schemata (in Langacker's 1987 terminology this is called 'accommodation'). The guiding principle behind this process is that of image superimposition: The composite schema is formed from the individual schemata by superimposing them. Figure 2 shows the superimposition of [ROPE], [STRETCHED], [ACROSS] and [ROAD] to form [(A) ROPE STRETCHED ACROSS (THE) ROAD]. We are working with image superimposition, because schemata having predications in different domains can successfully be viewed as images. Of course, in the computer the superimposition process operates on schema structures of the sort just presented.3 The computational superimposition process subdivides into a number of smaller operations on schemata. The first of these operations is called domain identification. It takes the matrices of the input schemata and decides which domains are central enough in all input schemata to be kept in the composite schema. For instance, if we superimpose [SALLY] onto [UNDER]4, the matrices of the two schemata and the composite would be: [SALLY]: Emotional, Spatial, Age, Colour, Profession,... [UNDER]: Spatial

->· [SALLY UNDER]: Spatial

Implementing cognitive semantics

583

That is, in the composite [SALLY UNDER] only the spatial domain remains. [SALLY]: Emotional, Spatial, Age, Colour, Profession,... [FURIOUS]: Emotional, Colour -> [SALLY FURIOUS]: Emotional, Colour If [SALLY] is instead superimposed onto [FURIOUS], the composite probably contains the emotional and colour domains. [STONE]: Spatial, Material, Age, Colour, . . . [FURIOUS]: Emotional, Colour -> [STONE FURIOUS]: Colour In the two previous cases, the centralities of the domains in the composite matrix are central. However, if a [STONE] is superimposed onto [FURIOUS], only the non-central colour domain remains. In Holmqvist 1993a, a mathematical function for the calculation of the composite matrix is discussed, but here it is sufficient to point out that the domain centralities play the main role. Domain identification has the property, of course, to disambiguate the schemata, as the examples show. But domain identification also serves for anomaly detection. If as in the third example only a non-central domain remains in the composite, then the two input schemata are too different to be superimposed, and an anomaly will then be signalled. Once the domains of the composite have been sorted out, the actual superimposition can start. In a common domain, each of the input schemata have a predication. These predications must now be joined into one, as in figure 2. A number of operations jointly perform this welding task. Sometimes it is only a matter of seeing that two predications are both ID-extended, as in the temporal domain examples of Figure 3. However, often one of the two predications must be possible to turn properly before they can be successfully superimposed, as in Figure 4. Also, the superimposition must map the dimensions in one predication onto the proper dimensions of the other predication. When there is only one dimension of the same type there is no problem. See Figure 5. But when there are several dimensions of that type in the other predication, as in Figure 6, another mapping principle must be adopted.

584

Kenneth Holmqvist

tvelwe ο'dock

day

a.. .ago

TIME

a...ago

_v

7

ISUPERIMPOSmONl

I NOT SljPEmfPOSABLEl

Figure 3. Non-matching and matching predications.

along

HOB

HOW?

\

JSUPERIMPOSITIONl

HOWS

HORll

(NOT SUPERIMPQSABLE

HOWS

Figure 4. Turning and tilting as a result of superimposition.

Ion«

rope

Extension

Extension Diameter Posrtion(s) Direction

Schema

Dimensions

u

the flagpole

along

the pole

Figure 5. Dimensional mapping according to type.

Implementing cognitive semantics long

585

wafl

Schema

Dimensions

Extension A: 2:nd largest value • Extension B: latßect value Extension C: least value Direction A: vertical

Extension

Figure 6. Dimensional mapping according to relative size within type.

John

walk/TR snake

walk.TR

input leva

fart levtt

Figure 7. Superimposition at input and part levels.

The solution suggested in Holmqvist 1993 a, 1993b bases on the thorough investigating of dimensional mapping principles made by Lang et al. 1991. For Figure 6 it means that the extension dimension in [LONG] should map onto the extension dimension in [WALL] having the largest value. Even if we are successful in mapping dimensions, turning and adjusting the predications at the input (matrix) level, there are also predications to be superimposed at the part level. Figure 7 shows how an unproblematic superimposition of ID-directed spatial predications at the input level is followed by different superimpositions of parts. Of these two cases, only the left one is successful: Every part of [WALK].TR is assigned a part in [JOHN]. There is no mismatch like there is between [SNAKE] and [WALKj.TR.

586

Kenneth Holmqvist above

input leva

the shoals

Ο Ο the shoals

parts atthe whole level

Figure 8. Superimposition at the whole level.

If one looks at a larger number of cases where parts are superimposed (cf. Holmqvist 1993a), it seems as if superimposition of parts consists of two operations: First, each part in one schema must be mapped onto a part in the other schema. Second, each such pair of parts must be superimposable. Thus, the legs and feet of [JOHN] must be attached at the same place, have the same directionality and extension as the legs and feet of the generic walker [WALK].TR, as well as having basically the same domains in their matrices. The wholes are in a way the opposites of the parts. Superimposing wholes in a similar way as we just saw parts being superimposed is a mechanism for dealing with important but slightly subtle semantic effects. Take as an example the difference between "John used to row above the shoals" and "John used to row above the hills". Figure 8 depicts the situation. [ROW] predicates a processual motion path, placed spatially above the [SHOAL] or [HILL] predications by [ABOVE]. At the input level, this works fine in both cases, because there [HILL] and [shoal] have the same predications. However, when we enter the whole level of each, they differ. A whole could be said to be the surrounding of each of its parts. The whole of [SHOAL] thus also has [(SEA-)XVATER] as a part (i.e., [SHOAL] and [(SEA-)WATER] are siblings), while [HILL] has [air] as a sibling. [ROW] has a number of sibling schemata, among them [(SEA-)WATER]. When [ROW], [ABOVE] and [SHOAL] superimpose, the [(SEA-)WATER] schema of [ROW] will coincide with the [(SEA-)WATER] schema of [SHOAL]. This is of course fine, since [(SEA-)WATER] easily superimposes onto itself. However, when [ROW], [ABOVE] and [HILL] superimpose, [(SEA-)WATER] will coincide with [AIR], and these two schemata do not superimpose easily.

Implementing cognitive semantics

587

whole level

input level Figure 9. Ascent.

The basic principle here is the same as when parts were superimposed (Figure 7): When schemata are mapped onto each other, it is done because they coincide in the composite schema. When two schemata coincide, it must be possible to superimpose them into being one schema. Before we go on, let us summarize the basic operations during schema composition: First the matrix of the composite is calculated from the matrices of the parts. Then in each domain the input level predications are superimposed, which concerns their extension, direction, dimensions, dimensional values etc. Finally parts are mapped onto parts, sibling schemata onto sibling schemata, and the resulting pairs are superimposed. Let us also point out that in this process there is no classical type-checking of the kind that the direct object to a verb must be a noun. An error like "Sam hit goes" is detected during domain identification because [GOES] has domains very dissimilar from the domains of [HiTj.LM. The two schemata do not superimpose easily. Along with the superimposition operations we have seen so far, there are also a number of operations that alter the processing in various ways. One such operation is ascent, illustrated by the question what "it" in (1) refers to. (1)

I saw headlights coming straight at me, but I was able to get out of its way.

One answer would claim that it cannot be the headlights, because they superimpose badly onto [COME].TR. However, the whole [CAR] of [HEADLIGHTS] has the domains that allow it to superimpose onto [COMEj.TR. The ascent operation substitutes a schema with its whole.

588

Kenneth Holmqvist

Figure 10. A type 2 split of [GUN].

Another altering operation is the split. The splitting operation appears in different situations, where there is need to split or divide a schema. The hedge [FAKE] performs a split of type 2 by cutting off the processual wholes of schemata. The remaining meaning of a [GUN] to which [FAKE] has been applied is a gun of the same spatial form, material etc. and with the same parts (trigger, barrel etc.) as any gun. However, the fake gun will have no ability to shoot, because it lacks the processual wholes. There is no room here for further exemplifications of the semantic operations for altering schemata, but Holmqvist 1993a discusses several more: Descent, the reversed ascent, forced installation of structure, predication scaling, correspondence reassignment and lexical reevocation (which occur in reinterpretations such as Garden-Path sentences) and predication mapping metaphors. One of the major questions in our project is to investigate the order of invocation between these semantic operations during schema composition in a natural language parser.

3.

Part II: Grammatical valence suggestion and updating of the schema population

In order to put the semantic operations into a parser, we need mechanisms for incoming morphemes, valence suggestion and management of the composite schemata. Here we shall have a closer look at the mechanisms for "grammar without any semantics" devised in Holmqvist 1993a that we are now in the process of implementing. Using a mathematization of Behaghel's principle, it can perform sentence segmentation without referring to any semantic distinctions such as V, N, Adj etc. Additionally, this grammatical model is very easy to add to the accommodation process just described.

Implementing cognitive semantics

589

3.1. Introduction The input to the parser is a stream of morphemes. The morphemes may originate either from text or from speech. At this stage we do not plan to include any phonetic information in the stream other than the time passing between incoming morphemes. It is important that the stream should be viewed as such. It may not be the case that the parser decides the rate of incoming morphemes by what speed it is able to consume them at. The parser must consume morphemes at the external rate.5 Placing this simple demand on the time complexity of the parser has as a consequence that most present parser algorithms are excluded. One natural algorithm which satisfies the complexity demand is incremental parsing combined with interrupts caused by incoming morphemes. The overall algorithm can be described as simply as this: - Suggest valence relations (i.e. composites) and evaluate them. - When a new morpheme arrives, interrupt and update. All lexically evoked schemata and all composites are kept in a schema population (SP). For each incoming morpheme, there is one generation of the schema population, i.e., we change SP generation at every incoming morpheme. In the time between incoming morphemes, i.e., when valence relations are being suggested and evaluated, two SP generations are kept available: The current SP generation belonging to the last incoming morpheme, and the previous SP generation belonging to the second to last incoming morpheme. It is from the schemata in these two SP generations that new composites should be suggested.

3.2.

Grammatical expectations and behaghelian distance

Let us now for a while turn over to the grammatical expectations. Together a schema and a grammatical expectation are intended to simulate what Langacker 1991 calls a constructional schema. The role of a grammatical expectation is to suggest valence relations between its own schema and other schemata in the previous or the current schema populations. In order to do this, grammatical expectations make use of order information in the morpheme stream. Figure 11 shows a number of grammatical expectations situated in a morpheme stream.

590

Kenneth Holmqvist Morpheme stream order



77. Various grammatical expectations in a morpheme stream.

The hexagons mark the evocation position of the expectations. The rectangles mark positions (site areas) where it is likely that schemata can be found that suit the parts of the evocating schema. For example, [MARY] seems proper to superimpose onto the LMl ('recipient') part of [GIVE]. It should be obvious that grammatical expectations are devised after so-called conceptual dependency. A part (like 'recipient') in a schema for which another schema is to be found is called a site, hence the name site area for the positions where they are expected to be found. Nothing says that [MARY] should be found two morphemes away from [GIVE] or that [SOME MONEY] should be found nine morphemes away. There is no fixed distance from the evocation position at which a schema for a given site should be found. Instead, grammatical expectations are supposed to pick the closest schema (Holmqvist 1993a). Closeness between schema in the morpheme stream is however defined to follow Behaghel's principle of closeness between morphemes and semantic closeness in schemata. Defining such a behaghelian closeness yields a behaghelian distance. In parsing the morpheme stream of Figure 1 1, the behaghelian distance from the evocating morpheme evolves as in Figure 12. We can there see that at the shortest distances, we find [GIVES] (distance -1), (0), (0), [MARY WHO] (0) and [MARY WHO SAW A HOUSE] (0). These will therefore be the schemata best suited to be superimposed onto the LMl part of [GIVE], from the viewpoint of grammatical expectations. In the parser, schemata are updated incrementally as we change SP generations. Because of this it is necessary to calculate the behaghelian distance incrementally. The incremental calculation of the behaghelian distance is done in the following way: 1. Initially, i.e., at morpheme arrival, the distance of the expectation evoked at the new morpheme to the current morpheme is 0. 2. When a new morpheme arrives, the distance of all expectations of previously arrived morphemes (those in SP t _j; cf. below) to that current morpheme is incremented by 1.

Implementing cognitive semantics

591

Distance distance from [AHOUSE]

33

\

2 --

1 -·

SAW]

O

0 -{GIVES] WHO]

<

/[MARY WHO, SAW AHOUSE] -I 1

ISOMEMONEYI \·

Figure 12. The evolving behaghelian catching distance from the expectations evoked by the morpheme to later morphemes and composites. It is assumed that catches to form [GIVES], that catches to form [MARY WHO], that catches to form [SAW], that [SAW] catches [MARY WHO] and [A HOUSE] to form [MARY WHO SAW A HOUSE], which is then caught by [GIVES] etc.

3. When a schema Sj (such as [GIVE]) is being caught by a schema S2 (such as [+S]) to the right of Sj, all expectation distances in S t to the current morpheme are decremented by the number of morphemes included in 82 (here: 1). 3.1. If the schema S2 would have remaining expectations (which is not the case with [+S]), the distance of all expectations in Sj must also be incremented by the number of site areas in the expectations of S2. 4. When an expectation of schema Sj (such as [GIVE]) catches schema S2 (such as [MARY WHO SAW A HOUSE]) to its right, the catching distance of the expectation in SI is reduced by the number of morphemes in S2 but one. In our example, the distance from [GIVES] to is 6, so the catching distance to [MARY WHO SAW A HOUSE] is 5-6+1=0. 4.1. If the schema S2 being caught by Sj has remaining expectations (the case with [MARY WHO SAW]), the distance of expectations in S} to schemata following S2 will have to be incremented by the number of remaining site areas in the expectations of S2. In our example, [MARY WHO SAW] expects schemata for one site area to the right, so the catching distance from [GIVES] to schemata following [MARY WHO SAW] is 1, not 0.

592

Kenneth Holmqvist

caching distance = 1 -> binding o»sy-l md=2/3 catching die»« = -!-» binding energy* 1-1 = 0

Figure 13. Ideal catching distances vs. binding energy.

The incremental distance calculation in 1. through 4.1. concerns rightward distances only. When distances for leftward expectations are calculated, it is not the expectations but the schemata in the SP that hold the distance value. They are, however, updated in a similar way, not described here but in Holmqvist 1993a. 3.3.

Ideal catching distance vs. binding energy vs. mel

Ideally, a rightward expectation should catch schemata at the catching distance 0. In Figure 12, the expectations evoked by could catch , , [MARY WHO], [MARY WHO SAW A HOUSE], and [SOME MONEY].

The question then arises: Should expectations only be allowed to catch schemata at the ideal catching distance 0? My own answer is no, and it is not only because of the abundance of "grammatical illegalities" in understandable morpheme streams originating from spoken language. The major reason is that allowing non-ideal catching distances has clear advantages in the parser we are developing. In our parser, the latest morphemes may always try to catch schemata to its left before schemata to its left may have their try on the schemata evoked by new morpheme (cf. Holmqvist 1993 a: 261 and below). In Figure 13, the current morpheme first attempts to catch schema [B] and [A]. Supposing that [C] is a verb, [B] a verb prefix and [A] a noun for the agent, should clearly be allowed to catch [A] at distance 1. Instead of not allowing expectations to catch schemata at distance 1, we sum up all deviances from the ideal catching distance in the binding energy value. After has caught schemata to its left, the verb prefix morpheme can try to catch schemata to its right. It will then find schema [AC], and according to rule 4 the catching distance will be —1. In the composite [ABC], this negative catching distance is added to the former binding energy 1 to give the new binding energy 0.

Implementing cognitive semantics

593

distance =« 0 - 2 + 1 = -1 [(ACXBD)] catching binding energy = 1 + 1 -1= 1

[CD]

bindBg energy -0 caching distance«! - 2 + 1*0 binding energy = 0 + 0+ 0=0 md = l

Figure 14. Binding energy versus mel. Binding energy is calculated as the sum of all catching distances involved in building a composite. Mel is defined as the quotient between the number of morphemes in a composite and the total number of morphemes from first to last morpheme in the composite.

Binding energy is the sum of all catching distances of the suggested valence relations that keep the composite together. If there are "holes" in the composite, like in [AC], the binding energy will be greater than 0. But if morphemes inside the hole catches the non-compact composite, the hole can be "filled" by that morpheme. The binding energy is reduced due to the negative catching distance from such morphemes. If we only allowed expectations to catch at the ideal distance 0, this would not be possible. High binding energy will not necessarily indicate "holes" in the composite. It also indicates that a composite may have non-behaghelian composition structure. For instance, the composite [AC] in Figure 13 has binding energy 1 and a hole. We could also mark that it has a hole by dividing the number of morphemes (2) include in [AC] by the number of morphemes over which [AC] stretches (3). We then get the mel (morpheme explanation, local) quotient %. A mel value lower than 1 indicates that there are holes in the composite. If the mel value is higher than 1, the same morpheme occurs more than once in the composite. Although the composites [(AC)(BD)] and [(AB)(CD)] in Figure 14 both have mel = 1, they have different binding energy. In forming [(AC)(BD)], schemata have twice been caught at distances 1. The binding energy is decremented by the hole being filled, but it still remains 1. In [(AB)(CD)], however, all schemata have been caught at distance 0, hence the binding energy is also 0.

594

Kenneth Holmqvist

3.4.

Duplicate filter

When grammatical expectations are allowed to catch all schemata at all distances, the first problem is that a composite may include the (same occurrence of) the same morpheme more than once. In Figure 12, what would stop from catching [GIVES] to form [GIVE(GIVES)]? The first answer is that such composites have mel > 1. So we disallow composites with mel > 1 from being formed. The second answer is that many composites with duplicate morphemes end up with a negative catching distance. Such is the case with [(AB)(ACD)], which has a binding energy of — 1. These composites are also not allowed to be formed. Third, some composites, like [(AC)(CD)] have mel = 1 and a binding energy of 0, and still they are duplicates. Therefore the parser always looks for endpoint identities: The first and last position of one schema are checked against the first and last position of the other schema. If there is a match, the composite is not allowed to be formed. These three criteria sort away most composites with duplicates but not every single one. One that slips through is [((AB)E)(BD)], with mel = 1 and binding energy = 1. Still, we think that our filter is sufficiently reliable. In order for two schemata to form a composite that slips through the filter, at least one of the schemata must have a high binding energy, which (as we will soon see) means that it is unlikely to survive. And on top of that, the accommodation process that evaluates the suggested valence relations is highly likely to dispense of all composites with duplicates that our filter does not remove, since these duplicates are virtually always semantic nonsense.

3.5.

The schema population updating

It is now time to look at the overall administration of the grammatical half of the composition process. Figure 15 depicts the previous and current morpheme streams at the time when a new morpheme has just arrived. Observe that the current SP only includes the schema of the last morpheme. The previous SP contains all composite and simple schemata from earlier parsing with the exception of [JOHN +ED].

The schemata are ordered according to their values on important parameters. A schema with a higher binding energy is considered less valuable than

Implementing cognitive semantics SP«,

SP.

IJOHNSAW]

« O l l

PAW]

0 0 1 1 / 3

POBNUB)

1 0 1 ΙΟ

o-«= 0-=«

595

Huron

ο

0 0 1 1 / 4

ο ο ι ίο

It ED] «=O

ISBB1

I

HOHN]

1 0 1 10

Ο

0 I 10

Figure 15. Previous and current SP directly after the new morpheme arrived. Previous morphemes were: . Each composite is one entry (row). The left column shows the current grammatical expectation and the morphemes included in the composite. The middle column shows the catching distance, the binding energy, the mel and meg values in this order. The right column is for the image superimposition (here empty).

SP generation shift. Interrupt on / morpheme / arrival

v,,

Leftward: Expectations of SP, schemata catch SPt-i schemata.

V Rightwant Expectations of SP(_i schemata catch SPt schemata.

Accommodation of suggested valence relations.

Figure 16. Overall processual structure of SP updating.

a schema with lower binding energy. Also, the more morphemes included in a composite, the better. The number of included morphemes relative to all morphemes so far appearing in the morpheme stream is measured by meg (morpheme explanation, global). It is also better to have a lower distance than a higher. Finally, of course, the better the accommodation process succeeds in superimposing the schemata in the composite, the better. Together, these values decide the survivability value of schemata in the SP. Those with a low survivability are likely to be pruned off, because the SP has limited size. Only schemata that have a high survivability value are likely to be part of future larger composites, and hence only these schemata are kept. Holmqvist 1993a discusses in more detail how these values measure the survivability of schemata.

596

Kenneth Holmqvist

The overall processual structure of SP updating is shown in Figure 16. The example in Figure 16 depicts the situation directly after an SP generation shift caused by the arrival of the morpheme . After the completion of a generation shift, the parser forks into two parallel processes. One handles the accommodation of the valence relations being suggested by expectations. Here we shall only focus on the other process, which administrates the grammatical expectations. First, schemata in SPt (in Figure 15: SP) with leftward expectations suggest valence relations to schemata in SP^j. The schema [ANNIE] has no leftward expectations. If it had, it would suggest valence relations in the order of distance to schemata in SP t _j: first to [JOHN SAW], then to [SAW], [+ED], [JOHN SEE], [SEE] and finally to [JOHN]. The composites then formed would immediately be inserted into SP, and its valence relations accommodated by the parallel process. When all leftward expectation in SPt have had their try on schemata in SP t _j, it is time for the schemata in SP t _ t (i.e., SP) with rightward expectations to suggest valence relations to schemata in SPt. This is again done in order of short distance: First [JOHN SAW] catches [ANNIE] to form [JOHN SAW ANNIE], then [SAW] catches [ANNIE] to form [SAW ANNIE]. When all SP schemata at distance 0 have suggested valence relations, the composites are inserted in SP. After the two composites have been inserted, the rightward-expecting SP schemata at distance 1 ([JOHN SEE] and [SEE]) can suggest valence relations not only to [ANNIE] but also to [JOHN SAW ANNIE] and [SAW ANNIE]. In this case several of the resulting composites will be filtered off because of having duplicate instances of the same morpheme . However, in other cases, such as when we have a row of rightward expectations ("A big blue balloon"), it is necessary that early schemata () can catch composites built later on ([BIG BLUE BALLOON]). Since there are no schemata with rightward expectations in SP at distances higher than 1, no more grammatical valence suggestions will be made. The parser can spend all its time on accommodation, while awaiting the arrival of the next morpheme. It should be noted that all distance and other values are updated in the composites formed by the suggestion mechanisms. Also, remember that the SP has a finite size. When it gets full, the schemata with least survivability are pruned off. In our current implementation we prune off four schemata every time there are at least twelve of them.

Implementing cognitive semantics

597

When the next morpheme finally arrives, both the valence suggestion and the accommodation processes are interrupted. Their work is quickly but neatly put away, so that it can be continued in the new SP. Now we first move all SP^ schemata over to SPt. There will thus not only be a [JOHN SAW ANNIE] in SP, but also a [JOHN SAW] and a [SAW]. Of course, these schemata will have their values properly updated and hence be sorted by their proper survivability values. Some of them will be pruned off in the new competition in SP although they could live in SP. After the SP^ schemata have been copied over to SPt, we make a new SPt, and the old SPt gets to be the new SP^. In the new SP, we put the schema and expectations evoked by the new morpheme, along with their proper initial values. We then have a situation like that in Figure 15, only one morpheme further ahead.

3.6.

Interrupt time and morpheme arrival speed

It is now interesting to notice what happens if morphemes arrive at different speeds. If morphemes keep coming at a slow pace, all grammatical valence suggestions are made, even by the last rightward expectations in the ordering. But if morphemes stream in very quickly, the suggestion mechanism may not have time to perform all possible valence suggestions. The morpheme arrival speed thereby makes a difference in what composites are created. We saw in the previous example, based on Figure 15, that both [JOHN SAW ANNIE] and [JOHN SAW] ended up in SP . Now, assuming that the morpheme stream continues with "and so he walked over to her". Then the composite [JOHN SAW ANNIE] is the proper one. If instead the morpheme stream continues with "and Susie", then [JOHN SAW] is the schema for the closest future, properly suited to catch the coming composite [ANNIE AND SUSIE]. While the once so promising [JOHN SAW ANNIE] is realised to have been just a temporary misunderstanding. This is not only interesting because of SP size economy. It must also be a goal that the schema in the SP with the currently highest survivability should correspond to what a human language understander has in his or her linguistic short term memory at the time of that morpheme. It does not seem correct that in receiving the morpheme stream "John saw Annie and Susie", he or she would at some time be thinking that John saw only Annie.6 The way we are building our parser is such that it can be so calibrated that fast morpheme streams interrupt the valence suggestion already at an

598

Kenneth Holmqvist

early stage. It can even be the case that [JOHN SAW ANNIE] is never created when , and arrive quickly. [JOHN SAW] will then remain the most salient schema even in SP. Although we do not yet have access to linguistic investigations showing this, it seems likely that the time between morphemes and is considerably shorter in "John saw Annie and Susie" than in "John saw Annie, and so he walked over to her".

3.7.

Project goals and current project work

The project 'Conceptual Engineering' has as its goal an experimental implementation of the model described above. In particular, we are interested in investigating the processual order between the superimposition mechanisms in the implementation, such as domain identification, predication mapping, ascent etc. It is also highly relevant that the valence suggesting mechanisms and the schema population have the proper processuality. During these first three months of our project, we have been implementing the morpheme stream and schema population mechanisms. They will later serve as the frame under which implementations of the superimposition mechanisms can be investigated. We work in a Smalltalk environment, in order to achieve ease and speed in our experimental implementation. In the test runs we have been doing so far, it is clear that the parser has a good capability to perform segmentation of the morpheme stream without any reference to semantics. Of course, when behaghelian distance is the only segmentation principle, the parser suggests some composites that are semantic nonsense. When the accommodation mechanism has been implemented, it is easy to add the valence relation evaluation to the survivability value of the composite. The nonsensical composites will then be pruned off in competition with semantically better composites. In parallel with our implementation work, we also try to refine the model. At this stage we are looking at experiments with discourse markers which help an addressee to recognise structural units in discourse and identify the relations between those units (cf. composition, evaluation, relevance and phatic markers in Holsänovä 1993). We hope that this work will allow us to specify an updating mechanism for the schema population that is also to some extent aware of the discourse structure.

Implementing cognitive semantics

599

Notes 1. As opposed to, for instance, generative grammar, in which the language description, the language reception model and the language production models are all one and the same. 2. Holmqvist 1993a gives a survey of different predications and their dimensionality. For spatial form predications, Lang et al. 1991 gives a similar and very systematic formalization, which can be translated into the more general framework of predications and region activation (see Holmqvist 1993b). 3. The superimposition process for accommodating valence relations can be compared with Centner's et al. 1989, in press 'structural mapping'. 4. More precisely, onto [UNDER].TR, the trajector of [UNDER]. The same remark applies also to the following examples. 5. This means that the parser should have a roughly linear time complexity in terms of the number of incoming morphemes. 6. With the exception of language learners at an early stage receiving unnaturally slow morpheme streams from reading or listening.

References Gentner, Dedre 1983 Structure-Mapping: A Theoretical Framework for Analogy. Cognitive Science 7,155-170. Gentner, Dedre, Falkenheimer, Brian and Forbus, Ken 1989 The Structure-Mapping Engine: Algorithm and Examples. Artificial Intelligence 41, 1-63. Gentner, Dedre, Markham, Arthur B. (1995) Similarity is Like Analogy. In: C. Cacciari (ed.), Similarity. Brussels: Brepols. Holmqvist, Kenneth 1993a Implementing Cognitive Semantics, Lund: Department of Cognitive Science. 1993b Two Dimensional Representations Compared. Lund: Department of Cognitive Science. HolSanovä, Jana 1993 Discourse operators in language understanding. Lund: Department of Cognitive Science. Lakoff, George 1987 Women, Fire, and Dangerous Things. Chicago: University of Chicago Press. 1989 Some Empirical Results about the Nature of Concepts. Mind & Language 4, 103-129. Lang, Ewald, Carstensen, Kai-Uwe and Simmons, Geoffrey 1991 Modelling Spatial Knowledge on a Linguistic Basis. Heidelberg: SpringerVerlag.

600

Kenneth Holmqvist

Langacker, Ronald 1987 Foundations of Cognitive Grammar, vol. I. Stanford: Stanford University Press. 1991 Foundations of Cognitive Grammar, vol. II. Stanford: Stanford University Press.

Subject index abductive change, 76 absolutive, 392 abstract transitional phrase, 425-442 accusative, 407 actance variation, 449 action chain, 457,459 active zones, 391 activeness, 463 adjectives, 99 adverbial cluster, 534 of place, 527-542 of time, 527-542 placement, 527-542 affected, 411 affectedness, 457,466 aktionsart, 245, 426 alienable/inalienable, 397 ambiguity pragmatic, 150, 153 analogical extension, 75 anaphoricity, 551 animacy, 494-498 antonymy, 97 aphasic patients, 5 archilexeme, 63-64, 68 argument transfer, 127 arguments core ~, 492 aspect, 228 associative names, 39 attached contact, 178 attachment, 178 autonomous syntax, 5 autonomy, 7 autonomy hypothesis, 6 autonomy thesis, 3,6, 9, 17 barriers, 335

base, 347 basic act, 114 Basic English, 115 basic level, 38,44 basic vocabulary, 114-115 body part names, 177 brand names, 37, 54 Bulgarian, 93 categorical perception, 6 categories radial ~, 347 categorization, 76,94 causativity, 419 clauses biactant ~, 493,498 ditransitive ~, 447 climax, 338-340 Coeur d'Alene, 165 coextensiveness, 415 cognitive commitment, 4, 6 colour basic colour terms, 67 competition, 338, 339 complementation English ~, 505 complex category, 170 computer implementation, 579 conceptual autonomy, 401 conceptualization, 63, 66 conduit metaphor, 120 connectionist models, 16 construction grammatical ~, 8 syntactic ~, 450 construction frames, 407 constructional meaning, 138-139 container schema, 101, 107 containment, 29

602

Subject index

context, 31 continuous, 409 contrast semantic ~, 97 contrastivity, 97 conventionality, 149,150, 156 conventionalization and predictability, 156 correspondences epistemic ~, 351 ontological ~, 351 count, 180 covering surfaces, 174 culture and language, 35 Czech, 85 declaration, 512 deduction natural ~, 23, 30 definiteness, 572, 574 deontic, 506 deontic modality, 509 direct mode, 471,479-483 direct speech, 486 discursive prominence, 455 dissociations, 5 domain, 137 control, 116 force-dynamics ~, 116 human interest ~, 117 spatio-temporal ~, 116 double predications, 412 Dutch, 50 elaboration, 172 empathy, 464 encyclopaedic knowledge, 98 energy flow, 129 entity nominal and temporal ~, 552 entrenchment, 44, 149 epistemic, 506, 508

modals, 476, 486 qualifiers, 475 ergative, 392 event-related brain potentials, 5 events, 554 exaptation, 75-76 experiential, 567,569,570, 575 experiential iconicity, 531 expression overspecified ~, 560 extension, 172 metaphorical ~, 229 figure/ground, 459 filter effect, 341 fixation principle, 24, 31 focus nominal ~, 558 of attention, 551, 560 temporal ~, 558 force-dynamic, 314-317 formal system, 4 formation pattern, 440 frame, 146,148 verbal ~, 149,154 frame semantics, 154 free indirect mode, 479—483 French, 50 functions central ~, 491,493 generative grammar, 4 generative linguistics, 3 genitive, 398 geometry relational ~, 25 gestalt, 61, 68 give causative, 126 double object, 130 figurative extensions, 118 impersonal reflexive, 138 impersonal subject, 123, 138

Subject index literal, 114 passive, 123 reflexive, 118,122,138 gradable complementaries, 107 grammatical expectations, 589 grammaticalization, 225 grounded, 415,422

603

markedness assimilation, 571 meaning lexical and grammatical ~, 230 meaning of the ditransitive construction, 452 mental space, 471,475,479-483,486 mereology, 26 metaphor, 7, 10, 11,40,54,314-319 heterosemy, 227 conceptual ~, 367 human body, 174 conduit ~, 120 understanding, 367 iconicity, 527-542 metaphor in law, 331-342 experiental ~, 531 metaphor in reasoning, 331-342 idioms, 429 metaphorical projection, 66 image schema, 60-62, 64, 68, 99, 100, metonymic mapping, 147,154,158 103,300,314,315,349,580 metonymy, 141,147,148,153,155,156, image superimposition, 582 158 inclusion, 414 minimalist program, 15 indefiniteness, 562 modality, 471 indirect objects, non-actantial, 454 deontic ~, 471-473 Indo-European, 76 epistemic ~, 471-479 information status, 540 model-theoretic approaches, 17 intentional and unintentional events, 499 modularity, 13, 15 intentional causation, 508 interpretation naming, 418 nominal ~, 553 natural order, 532 temporal ~, 553 naturalization effect, 341 invariance hypothesis, 359, 360 networks of Schemas, 170 invasion, 335, 338, 339 nominalization, 141,148,151,156,158 non-discrete, 409 land, 175 object, 253,256, 391 landmark, 170 direct ~, 448, 491, 493, 496, 497, language acquisition, 114 501 language and culture, 54 indirect ~, 448,491,493,496,497, lexicalization, 63,430 499, 501 lexicography, 143 nouns, 253-256 lexicon, 15,17 syntactic coding, 493 Macedonian, 90-93 ontology, 30 markedness, 466, 500-501, 567-570, nominal ~, 553 574 temporal ~, 553 syntactic ~, 491,530,531 opposition, 62,63 textual ~, 531 ordo naturalis, 532

604

Subject index

overspecification of referents, 560 part-whole relation, 29 participants central ~, 449,458,462 particles, 225 perception, 417 perceptual deictic, 10 perfect, 562 perfectivity, 133-134 personal pronouns, 572,574 perspective, 471,479, 483,486, 563 phonology, 570,571 place, 253 place nouns, 253-256 plant ecology, 337-341 Polish, 85 polysemy, 59, 150, 153, 155, 158, 227, 402 competing motivations, 136-138 prepositional ~, 347 possession, 410 possessive-absolutive construction, 394 possessor ascension, 399 pragmatics, 15, 17,24, 31 predictability, 149,152,154,155 prefixes, 165 prepositions, 347 English ~, 225 presentative, 408,418 profile, 347 proper names, 572-574 propositional content, 408 prototype, 300, 304, 318,491,494 prototypical speaker, 567, 569 quantifier float, 392 racial segregation, 331-342 reconstruction semantic ~, 57-59 reference domain, 408,421 reference point, 455,480, 556

referent discourse ~, 549 nominal ~, 552, 554 temporal ~, 554 referential, 415,422 relational grammar, 399 relational nouns nominality, 249-272 relational predication, 408 repetitive or scattered contact, 184 salience, 37,43,47, 53, 391,458 saliency, 464 Salish, 165 satellites, 226 scale schema, 101,104 schema, 170 source-path-goal, 348, 350 search, 417 second predication, 412,414,417,418 secondary predication, 413 semantic fields, 438—441 semantics conceptual ~, 4 formal ~, 23 model-theoretic ~, 4 semantics of space, 23 sentential complements, 505 Serbo-Croatian, 86 Slovak, 81-86 Slovene, 86-90 source domain, 367 space, 23 Spanish, 395 clauses, 492 triactant clauses, 447 spatial prefixes, 170 spatio-relational, 167 states, 554 subjacency, 14 subject, 391,448,491 subject function, 492

Subject index subjectification, 471,473,474,479,483, 484,486 subjectivity, 471 substitution, 416 succession, 338,339 support, 29 supporting surfaces, 175 target domain, 367 tense

as anaphor, 550,553 text strategy locative ~, 527 temporal ~, 527 text structure hierarchy of, 538 text types, 536 textual fit, 531 that clauses, 505 thematic relationship, 401 f/zere-constructions, 10-13 thought, 485,486 to infinitives, 505

topicality, 455,463,464 topology, 26 trajector, 170 trajector-landmark, 187 trajectory, 181 transfer, 410 transitive prototype, 493, 499 transitivity, 129-130,494, 495, 498 unidirectionality hypothesis, 368 337, 339,341

urban ecoiogy) 335)

vagueness, 150,151 valency, 5 30 verb agreement, 392 verb-particle constructions, 237 verbs phrasal ~, 237 volition, 498-500 water, 175 word order, 392 world knowledge, 31

605

Cognitive Linguistics Research Edited by Rene Dirven, Ronald W. Langacker and John R. Taylor Mouton de Gruyter · Berlin · New York This series offers a forum for the presentation of research within the perspective of "cognitive linguistics". This rubric subsumes a variety of concerns and broadly compatible theoretical approaches that have a common basic outlook: that language is an integral facet of cognition which reflects the interaction of social, cultural, psychological, communicative and functional considerations, and which can only be understood in the context of a realistic view of acquisition, cognitive development and mental processing. Cognitive linguistics thus eschews the imposition of artificial boundaries, both internal and external. Internally, it seeks a unified account of language structure that avoids such problematic dichotomies as lexicon vs. grammar, morphology vs. syntax, semantics vs. pragmatics, and synchrony vs. diachrony. Externally, it seeks insofar as possible to explicate language structure in terms of the other facets of cognition on which it draws, as well as the communicative function it serves. Linguistic analysis can therefore profit from the insights of neighboring and overlapping disciplines such as sociology, cultural anthropology, neuroscience, philosophy, psychology, and cognitive science. 1 Ronald W. Langacker, Concept, Image, and Symbol. The Cognitive Basis of Grammar. 1990. 2 Paul D. Deane, Grammar in Mind and Brain. Explorations in Cognitive Syntax. 1992. 3 Conceptualizations and Mental Processing in Language. Edited by Richard A. Geiger and Brygida Rudzka-Ostyn. 1993. 4 Laura A. Janda, A Geography of Case Semantics. The Czech Dative and the Russian Instrumental. 1993. 5 Dirk Geeraerts, Stefan Grondelaers and Peter Bakema, The Structure of Lexical Variation. Meaning, Naming, and Context. 1994. 6 Cognitive Linguistics in the Redwoods. The Expansion of a New Paradigm in Linguistics. Edited by Eugene H. Casad. 1996. 7 John Newman, Give. A Cognitive Linguistic Study. 1996.

8 The Construal of Space in Language and Thought. Edited by Martin Pütz and Rene Dirven. 1996. 9 Ewa Dabrowska, Cognitive Semantics and the Polish Dative. 1997. 10 Speaking of Emotions: Conceptualisation and Expression. Edited by Angeliki Athanasiadou and Elzbieta Tabakowska. 1998. 11 Michel Achard, Representation of Cognitive Structures. 1998. 12 Issues in Cognitive Linguistics. 1993 Proceedings of the International Cognitive Linguistics Conference. Edited by Leon de Stadier and Christoph Eyrich. 1999.