New Vistas in Grammar: Invariance and Variation, Proceedings of the Second International Roman Jakobson Conference, New York University, Nov. 5-8, 1985 (Current Issues in Linguistic Theory)

NEW VISTAS IN GRAMMAR

AMSTERDAM STUDIES IN THE THEORY AND HISTORY OF LINGUISTIC SCIENCE General Editor E.F. KONRAD KOERNER (University of Ottawa)

Series IV - CURRENT ISSUES IN LINGUISTIC THEORY

Advisory Editorial Board Henning Andersen (Los Angeles); Raimo Anttila (Los Angeles) Thomas V. Gamkrelidze (Tbilisi); John E. Joseph (College Park, Md.) Hans-Heinrich Lieb (Berlin); Ernst Pulgram (Ann Arbor, Mich.) E. Wyn Roberts (Vancouver, B.C.); Danny Steinberg (Tokyo)

Volume 49

Linda R. Waugh and Stephen Rudy (eds) New Vistas in Grammar: Invariance and Variation

NEW VISTAS IN GRAMMAR: INVARIANCE AND VARIATION

Edited by

LINDA R. WAUGH Cornell University

and STEPHEN RUDY New York University

JOHN BENJAMINS PUBLISHING COMPANY AMSTERDAM/PHILADELPHIA 1991

Library of Congress ataloging-i η-Publication Data International Roman Jakobson Conference (2nd : 1985 : New York University) New vistas in grammar : invariance and variation : proceedings of the Second Interna­ tional Roman Jakobson Conference, New York University, Nov. 5-8, 1985 / edited by Linda R. Waugh and Stephen Rudy. p. cm. ~ (Amsterdam studies in the theory and history of linguistic science. Series IV, Current issues in linguistic theory, ISSN 0304-0763; v. 49) Includes bibliographical references. 1. Language and languages ~ Variation ~ Congresses. 2. Grammar, Comparative and general - Congresses. 3. Discourse analysis — Congresses. 4. Pragmatics ~ Congresses. 5. Typology (Linguistics) -- Congresses. I. Waugh, Linda R. II. Rudy, Stephen. III. Title. IV. Series. P120.V37I56 1985 415--dc20 91-27655 ISBN 90 272 3585 6 (pb.; alk. paper) CIP © Copyright 1991 - John Benjamins B.V. No part of this book may be reproduced in any form, by print, photoprint, microfilm, or any other means, without written permission from the publisher.

Contents

Preface Stephen Rudy Introduction Linda R. Waugh

ix 1

I. T H E Q U E S T I O N O F I N V A R I A N C E The Concept of Structure in Contemporary Linguistics Edward Stankiewicz Grasping the Nettle: Variation as Proof of Invariance Erica Garcia Invariant Meaning: Alternative Variations on an Invariant Theme Yishai Tobin

11 33 61

II. I N V A R I A N C E AND G R A M M A T I C A L C A T E G O R I E S Toward a Universal Calculus of Inflectional Categories: On Roman Jakobson's Trail Igor č Towards a Typology of Verbal Categories Howard I. Aronson Two Types of Markedness and their Implications for the Concep­ tualization of Grammatical Invariance Rodney B. Sangster The Role of Animacy in Language Change: From Dative to Genitive in Middle Indo-Aryan Monique Monville-Burston Invariance and Mutation in Acatec Mayan John S. Robertson

85 111

133

153 167

Vi

CONTENTS

Variation, Invariance, Hierarchy, and Integration as Grammatical Parameters Pierre Swiggers Invariance and Markedness in Grammatical Categories Madeleine Newfield and Linda R. Waugh

185 221

III. G R A M M A R AND DISCOURSE Tense-Aspect and Hierarchy of Meanings: Pragmatic, Textual, Modal, Discourse, Expressive, Referential Linda R. Waugh On the Concept of Time: Prolegomena to a Theory of Aspect and Tense in Narrative Discourse Nils B. Thelin On the Projection of Equivalence Relations into Syntagms Henning Andersen Invariance in Grammar, Variation in Discourse: Discussion Flora Klein-Andreu

241

261 287 313

IV. G R A M M A R A N D P R A G M A T I C S Deixis and Shifters after Jakobson Herman Parret Praguean Structure and Autopoiesis: Deixis as Individuation Cornells H. van Schooneveld Shifters and Non-Verbal Categories of Russian Olga T. Yokoyama English Speech Act Verbs: A Historical Perspective Elizabeth Closs Traugott Grammar and Pragmatics: The Two Axes of Language and Deixis Edna Andrews

321 341 363 387 407

V. T Y P O L O G Y A N D U N I V E R S A L S Two Approaches to Language Universais Joseph Greenberg Invariance and Variation: The Dimensional Model of Language Universais Hansjakob Seiler Classical and Modern Universais Research: Their Philosophical Background Elmar Holenstein

417

437

451

CONTENTS

Language Typology and Diachronic Linguistics Thomas V. Gamkrelidze Language Universals in Relation to Acquisition and Change: A Tribute to Roman Jakobson John A. Hawkins Paralinguistic Universais and Preconceptual Thinking in Language Ivan Fónagy

Vii

465

473 495

Preface After Roman Jakobson's death on July 18, 1982, a group of his colleagues decided that a fitting tribute to his memory would be an international conference devoted to a major topic of his scholarly work, namely the role of invariance and variation in grammar. Rather than being a specific record of his work in the field, it would reflect the state of contemporary scholarship on the subject while acknowledging the legacy of Jakobson as a major figure in modern linguistics. Linda Waugh's "Introduction" to the present volume details the theoretical issues that eventually became the focus of the con­ ference. An organizing committee was formed for the conference, consisting of Joseph H. Greenberg (Stanford), Stephen Rudy (NYU), C. H. van Schoonveld (Indiana), Edward Stankiewicz (Yale), Linda R. Waugh (Cornell), and Dean S. Worth (UCLA). It was decided to hold the conference at New York University, and Stephen Rudy was appointed the secretary of the committee. He and Linda Waugh served as the conference's coordinators. The conference, which was entitled "New Vistas in Grammar: Invariance and Variation", took place on October 10-13, 1985. After welcoming remarks by the Dean of the Facuilty of Arts and Sciences at NYU,  Duncan Rice, and opening remarks by the conference coordinators, three public lectures were given by Edward Stankiewicz, Hansjakob Seiler, and Igor Mel'čuk. In the present volume those lectures are incorporated by topic into the panel sessions which made up the bulk of the conference and which form the parts of the present volume: L The Question of Invariance; II. Invariance and Grammatical Categories; III. Grammar and Discourse; IV. Grammar and Pragmatics; and V. Typology and Universais. In two instances papers delivered at the conference were not included in the volume (Henrik Birnbaum's "Gesamtbedeutung — A Reality of Lan­ guage or a Linguistic Construct?"; Maurice Gross' "Projecting a Lexicon Grammar on Texts"). Several scholars who had been invited to attend were unable to do so, but their papers have been included here (Ivan Fónagy, Thomas V. Gamkrelidze, Joseph H. Greenberg, and Pierre Swiggers). For each of the conference's five panels, official discussants were invited to give their reactions to the papers, which had been distributed in advance. In the present volume it was possible to include only selected comments by the

χ

PREFACE

discussants, but we would like to thank all of them for their active partici­ pation: Edna Andrews (Duke), Paul Hopper (Carnegie Mellon), Flora KleinAndreu (SUNY Stony Brook), Henry Kučera (Brown), Madeleine Newfield (Cornell), Hansjakob Seiler (Köln). The coordinators of the conference would like to thank those individuals and institutions that made both the conference and the publication of its proceedings possible. A. Richard Turner, former Dean of the Faculty of Arts and Sciences at NYU, enthusiastically supported the idea of the confer­ ence, and his successor, C. Duncan Rice, provided generous financial sup­ port. The conference was also funded in part by private sources and by a grant from the National Endowment for the Humanities, a United States federal agency which supports the study of such fields as history, philosophy, literature and languages (NEH Grant Award RD-20623-85). The publication of the present volume was made possible by a subsidy from the Roman Jakobson and Krystyna Pomorska Jakobson Foundation, Inc., of New York. The proceedings of this conference are related broadly to two other international conferences held in tribute to Roman Jakobson. They are: Language, Poetry and Poetics. The Generation of the 1890's: Jakobson, Trubetzkoy, Majakovskij (Proceedings of the First Roman Jakobson Collo­ quium at the Massachusetts Institute of Technology, October 5-6, 1984), edited by Krystyna Pomorska et al. (Berlin: Mouton de Gruyter, 1987); and Roman Jakobson (Proceedings of the Conference on Roman Jakobson: Linguistics and Poetics held at the Università di Roma "La Sapienza" in November 1986), edited by Pietro Montani and Massimo Prampolini (Rome: Editori Riuniti, 1990). This volume is dedicated to the memory of Krystyna Pomorska Jakobson, without whose support the conference could not have happened and whose devotion to Roman Jakobson's scholarship was the chief animator of all of these proceedings. Stephen Rudy New York University

Introduction

Linda R. Waugh Cornell University

1. Within the last few years, there has been renewed interest in the semantics of grammatical categories, in particular, in morphological and syntactic categories such as tense, aspect, mood, voice, word order, and the like.1 This work has inevitably questioned the unity of the categories examined, for grammatical categories occur in such a wide variety of contexts, with such evident differences in interpretation both within a language and across languages, that it remains problematic whether they may all be viewed as manifestations of the same phenomenon. So, the fundamental question of sameness within difference (equivalence in difference) or invariance in the midst of variation, and conversely, variation correlated with invariance, is thus a central one in grammatical semantics. It seems to be accepted by the majority of linguists that grammatical categories such as person, tense, number, and case recur within a given language and are shared by a number of languages, and that even certain specific categories such as third person, past tense, plural number, and accusative case can be defined in such a way as to be applicable to a variety of uses in diverse languages and to different contexts in one language, even when they have non-uniform stylistic, pragmatic, and sociolinguistic values. Thus, the notion of passive voice, whether within one language or across languages, assumes, if only implicitly, a correlation with some common denominator, despite distinct forms, uses, and interpretations across con­ texts. This viewpoint is based on the assumption, whether directly stated or not, that there is some communality, or invariance, which lies behind the variation. In this most general formulation, the assumption of invariance has played an important role in the analysis of grammatical categories. Thus, for example, debates about whether English has one present perfect or two or

2

LINDA R. WAUGH

three (cf. McCoard 1978) all assume that invariance is to be posited with respect to variation, whether it is one invariant or two or three invariants. To be sure, there is disagreement as to whether the present perfect is to be treated as a single form associated with a single meaning (and of course the exact nature of that single meaning is also open to debate) or as a set of two or more homonymous forms each associated with its own specific meaning, but no one has assumed that every occurrence of the form called present perfect is different in kind from every other. In other words, the assumption of homonymy, though it enlarges the number of forms, nonethe­ less presupposes the concept of invariance. But invariance is a fundamental principle of linguistic research not just on the level of grammatical semantics. Indeed, it arises also in the study of phonology, syntax, and lexical semantics. Thus, the feature [ + nasal] is assumed to correspond to some invariant physical property of nasality, whatever the differences in frequency, intensity, duration, etc., that occur within the contexts in which [+ nasal] is present. In like fashion, any discussion of whether "bachelor" is one word or two or three accepts in principle the notion of invariance. And studies of basic word order involving subject, verb and object assume the equivalence of all those linguistic elements that can function as subject, verb or object, respectively. 2. While invariance in this most general formulation is implicitly or explicitly at the basis of most work in grammatical semantics, there is a much stronger, and thus more interesting, hypothesis about invariance with respect to grammar, first advanced by Roman Jakobson in his studies of Russian grammatical categories in the 1930s and elaborated by him in later studies.2 Invariance is the leitmotif which, as Jakobson himself said (1971b, 1981), unifies his work in such diverse domains as phonology, grammatical analysis, poetics, language acquisition, aphasia, and semiotics. He repeatedly pointed out that "invariance in the midst of variation" is one of the cardinal problems in linguistics, first addressed in the 1890s, and destined to be one of the major themes at issue (implicitly and explicitly) in modern thinking; and in fact, his name is the one probably most linked with this theme in twentieth-century linguistics. For Jakobson, the questions of invariance and variation are central to the issues of the structure of language and of its usage, since communication and intersubjectivity in general require an agreement to endow linguistic forms with perceptual and conceptual constants. Moreover, linguistic creativity — in particular, the use and understanding of a given form in a new context — requires that some inherent value be associated with that form. If this were not the case, neither speaker nor hearer would be able to produce or interpret a new usage. This is particularly true of grammatical categories.

INTRODUCTION

3

Grammatical categories, whether morphological or syntactic, are those which are obligatory for the construction of well-formed (grammatical) utterances in a given language; grammatical meanings are those meanings which are necessarily conveyed by a user of the language (1959b). Thus, in the Russian finite verb, the grammatical category of tense is necessarily conveyed, providing the user with a choice between the past and the nonpast tenses. Lexical categories are optional in the sense that the user may decide whether a given lexical category is to be conveyed or not. The differences between languages reside, then, not in what can be conveyed, but in what must be conveyed —whether a given concept is grammaticalized or not. Grammatical categories are not only those which are obligatorily and thus inevitably expressed, they are also those which evidence the most tightly constrained conceptual structure. According to Jakobson, for each gram­ matical category there is a single, general invariant meaning, a Gesamtbedeutung, a "common denominator of meaning". Any particular, specific meaning or interpretation of that category — its basic and marginal mean­ ings, its literal and figurative meanings, its frequent and infrequent meanings — are contextually determined variants, correlated with, but not the same as, the general invariant. Thus, according to Jakobson, the perfective in Russian has as its conceptual invariant a concern with the absolute com­ pletion of the narrated event (1932, 1957: 137); specific instantiations — such as inchoative event, quick events occurring in a sequence, shortlasting event — are its contextual variants. A grammatical invariant, then, is a general to be defined in relation to a hierarchized series of contextual specifics. Secondly, Jakobson claimed that invariant meanings, rather than being associated with grammatical categories as isolated and absolute entities, are based on the interrelated concepts of opposition and markedness — e.g., the perfective aspect in Russian possesses its invariant meaning insofar as it is opposed to the imperfective. And in normal declarative sentences, the perfective is the marked term in this relationship. Grammatical paradigms are underlain by sets of recurrent binary oppositions where one pole of the opposition is marked (more focused on a narrowly specified and delimited conceptual item) and the other unmarked. 3 A grammatical invariant, just as a phonological invariant, is relational and oppositional in nature. Thirdly, Jakobson advanced the theoretical and thus methodological claim that, in general, singleness of form is associated with singleness of meaning, except in certain, well-defined cases of homonymy. This principle has come to be known as the one form-one meaning hypothesis in relation to Jakobson's work, but it has also been variously labeled in general linguistic research: for example, "one-to-one symbolization" (Chafe 1970), "isomor-

4

LINDA R. WAUGH

phism" (Haiman 1980), "formal determinism" (Sangster 1984), and so forth. 4 A grammatical invariant, then, is closely and in many cases iconically related to the form(s) which carry it. Whether one agrees or disagrees with Jakobson on these questions of grammatical semantics, it is fair to say that his work has been seminal and is to be reckoned with by anyone seriously concerned with this area. Indeed, his profound influence will be evident in many of the papers included here, although the specifics of his hypotheses are still the subject of much controversy. 3. Ever since the 1930s the debate over the semantic analysis of grammatical categories has repeatedly returned to questions concerning invariance and variation. One common focus of discussion has been the nature of the invariants proposed. Many scholars have worried that the invariants are so general, vague, or abstract, that they may cover not only all the possible uses of a given category, but also predict impossible ones. This becomes particularly apparent when comparing what seems to be the same grammati­ cal category, but which nevertheless evidences systematic differences in closely related languages. Moreover, when there are competing analyses, what evidence is to be used to verify a particular one or to decide between them? What constitute confirming or disconfirming data with respect to a particular invariant posited? And where do such methodological issues as the question of economy, the role of formalization, the status of quantitative data, and so forth fit in? A second area of discussion has been the question of how much and what kind of variation exist in grammatical systems. In other words, how far must contextual meanings diverge before they may be viewed as different invariants associated with homonymous forms? What in fact is the crucial evidence for homonymy? Polysemy has been advanced by some as proof of the presence of wide-ranging but still related variation, while for others it is proof of homonymy. A further focus is the question of how to formulate an invariant which holds not just within one grammatical category but across categories and how to understand the differences it manifests. Simi­ larly, questions arise about the claim that the same invariants show up in quite different grammatical systems or even in both grammatical and lexical systems. A third issue has to do with the relation between invariance and variation: is either one of these to be seen as more central than the other? Furthermore, there is the question of the pathway between invariance and variation, and of the prediction from the interaction of the invariant and its context to the particular variants which exist — and vice versa. Is the basic meaning to be used as that variant from which the others can be derived in some way?

INTRODUCTION

5

Do, in fact, the variants derive one from another, or do they derive from the invariant — and in any case in what way would such derivation operate? Furthermore, many studies assume a hierarchy within the variants, such as basic (nuclear or core) vs. marginal, literal vs. figurative, (proto)typical vs. non-(proto)typical; others either treat all contextual variants on an equal footing, or conversely ignore all but the basic, literal, (proto)typical variants. This leads to the question of what kinds of interaction with the linguistic context exist and the problem of competing analyses based on whether, for example, the sentence or the discourse is taken as the contextual factor to be considered. What indeed are the contexts which are important for gram­ matical analysis? These issues are implicated further in the question of how invariance and variation are to be related to other major dichotomies of language — in particular system and usage (code and message, langue and parole, competence and performance), grammar and lexicon, morphology and syntax. Do the variants belong to usage only or are they to be seen as integral to the system; is the lexicon treatable in terms of invariance/variation in the same way that grammar is; is morphology essentially different from syntax in the ways in which it is structured with respect to invariance? A further set of issues revolves around the question of opposition and markedness. What is the role of binary oppositions and markedness in grammatical systems? What about ternary or η-ary oppositions or continua? How is markedness to be defined? Are all grammatical systems based on opposition and markedness between grammatical pairs, as Jakobson claimed? And can grammatical semantics be done in terms of semantic features at all? There are also a number of questions concerning the importance of the formal correlates of grammatical categories as determinants in grammatical semantics. Is the difference between morphological and syntactic categories a crucial one? What about periphrasis and grammatical words (particles, clitics, auxiliaries) as members of paradigmatic grammatical classes? And what about the ways in which allomorphy may be correlated with semantic relations between grammatical categories? In addition to these language-internal questions, there are the crosslinguistic issues of invariance. Various theories of universais assume crosslinguistic invariance and address the range of cross-linguistic variation and its limits. Typological studies, on the other hand, attempt to categorize languages according to this variation and to make universal statements about it. This brings up the question of what the relevant parameters are for typologies with regard to grammar, what the limits on variation are, what possible types of changes there are, and so forth. There also emerges the question of variation with respect to the usage of language, that is, the pragmatic consideration, and in particular the relation

6

LINDA R. WAUGH

of language to the speech situation in which it occurs. This is particularly important with regard to those categories which most explicitly bring in the speech situation, such as shifters (deictic categories) in the verb, noun, pronoun. And finally, there is the "human factor": what is the relation between invariance and the mind, between invariance and our perceptual system? What role does invariance, and perforce grammatical invariance as one of its central components, play in human thought and human life as a whole? 4. The present volume is centered around these issues of invariance and variation in grammar. It concerns in particular the present understanding of invariance and variation as general theoretical questions of grammar, as reflected in the synchronic and diachronic analyses of specific languages, as evidenced in questions pertaining to the relation between grammar and discourse and between grammar and pragmatics, and as realized in work on typology and universais.

Notes 1. As an example of this interest we could point to the textbooks on tense, aspect, and mood and modality published in the "Cambridge Textbooks in Linguistics" series (Comrie 1976, 1985, Palmer, 1986). 2. See Jakobson 1932, 1936, 1953, 1957, 1958, 1959a, 1959b, 1960, 1962, 1965, 1976, 1977, Jakobson & Waugh, 1979. See also Waugh 1984. 3. See Newfield & Waugh (this volume) for further discussion concerning the nature of markedness in Jakobson. 4. See Anttila 1972, Matejka 1975, Parret 1980, Haiman 1980, Waugh & Newfield ms, for further discussion.

References Anttila, Raimo. 1972. Introduction to Historical and Comparative Linguistics. New York: Macmillan. Chafe, Wallace. 1970. Meaning and the Structure of Language. Chicago: University Press. Comrie, Bernard. 1976. Aspect. Cambridge: University Press. . 1985. Tense. Cambridge: University Press. Haiman, John. 1980. "The Iconicity of Grammar: Isomorphism and Motivation." Language 56: 515-40. Hopper, Paul & Linda R. Waugh. 1987. "Invariance in Linguistic Theory." Semiotics 1982, ed. by J. Deely & J. Evans, 81-90. Lanham: University Press of America. Jakobson, Roman. 1932. "Structure of the Russian Verb." In 1984, 1-14. . 1936. "Contributions to the General Theory of Case: General Meanings of the Russian Cases." In 1984, 57-103.

INTRODUCTION

7

. 1953. Results of a Joint Conference of Anthropologists and Linguists. In 1971a, 554-567. . 1957. "Shifters, Verbal Categories, and the Russian Verb." In 1971a, 130-147 and 1984, 41-58. . 1958. "Morphological Observations on Slavic Declension (The Structure of Russian Case Forms)." In 1984, 105-133. . 1959a. "Linguistic Glosses on Goldstein's 'Wortbegriff'." In 1971a, 267-271. . 1959b. '"Boas' View of Grammatical Meaning." In 1971a, 267-271. . 1960. "The Gender Pattern of Russian." In 1971a, 184-186 and in 1984, 141-143. . 1962. "On the Rumanian Neuter." In 1971a, 187-189. . 1965. "Structure of the Russian and Ukrainian Imperative." In 1984, 33-40. . 1971a. Selected Writings II: Word and Language. The Hague: Mouton. . 1971b. "Retrospect." In 1971a, 711-724. . 1976. "Spatial Relationships in Slavic Adjectives." In 1985, 68-72. . 1977. "The Grammatical Buildup of Children's Language." In 1985, 141-147. . 1981. "My Favorite Topics." In 1985, 371-376. (Originally presented at the Accademia Nazionale dei Lincei in Rome in 1980, on the occasion of receiving the Antonio Feltrinelli Prize for Linguistics and Philology.) . 1984. Russian and Slavic Grammar: Studies, 1931-1981. Edited by L. R. Waugh and M. Halle, Intro. by L. R. Waugh. Berlin-New York: Mouton. . 1985. Selected Writings VII: Contributions to Comparative Mythology. Studies in Linguis­ tics and Philology, 1972-1982. Edited by S. Rudy, with a Preface by L. R. Waugh. Berlin: Mouton. & Linda R. Waugh. 1979. The Sound Shape of Language. Bloomington: Indiana University Press. Second edition, 1987, Berlin: Mouton de Gruyter. Matejka, Ladislav. 1975. "Crossroads of Form and Meaning." IJSLP 20: 93-120. McCoard, Robert. 1978. The English Perfect: Tense Choice and Pragmatic Inferences. Amster­ dam: North Holland. Newfield, Madeleine & Linda R. Waugh. this volume. "Invariance and Markedness in Gram­ matical Categories." Palmer, F. R. 1986. Mood and Modality. Cambridge: University Press. Parret, Hermann. 1980. "Postscript to the One Form-One Meaning Controversy" (debate with M. Bowerman, R. Kirsner, J. Whyte, W. Zwanenburg and M. Alinei). Quaderni di Semantica 1: 339—345. Sangster, Rodney . 1984. Roman Jakobson and Beyond. Berlin-New York: Mouton. Waugh, Linda R. 1984. "Introduction." In Jakobson 1984, ix-xvi. & Madeleine Newfield. ms. "Iconicity and the Morpheme: Toward a Model of the Lexicon." To appear in Lingua.

PART ONE THE QUESTION OF INVARIANCE

The Concept of Structure in Contemporary Linguistics Edward Stankiewicz Yale University 1. The study of structure, i.e., of the interrelations of the various elements of language, has been one of the central concerns of modern linguistics and has largely defined the direction of its development. It has given rise to a number of structuralist schools and has spilled over into several related disciplines, such as anthropology, semiotics, and poetics, for which it has become a method of inquiry and a badge of identity. The term "structuralist" itself was apparently first adopted by the Prague Circle, which sought to describe language as a network of relations and treat it in conjunction with its multiple functions. A "functional" and "structural" approach to language consequently became the hallmark of the Prague Circle, affecting the overall course of 20th-century linguistics. The rise of structuralism as an intellectual trend and method of research was not limited to the science of language. It emerged around the same time in a number of disciplines in reaction to the blunt empiricism of the late 19th century, which treated even logic as an inductive science and which combined a strong belief in observable facts with an equally strong distrust of abstraction and inference. However, a new current of thought was by then in the making and, according to the testimony of Charles Sanders Peirce, it made itself felt first in the abstract sciences, such as mathematics, logic, physics, astronomy and music. "Their ideal," wrote Peirce, "was the universal and abstract.. .in contradiction to the ideas formed upon outside experience" (Peirce 1966: 263). The new ideal soon gained sway also in linguistics, which had been dominated for some time by the Neogrammarians, for whom historical change and the accumulation of "real objects and facts" {einzelne Faktoren) were the only true objects of linguistic research. The forerunners of contemporary structuralism (such as Michel Bréal, Ferdinand de Saussure, and Jan Baudouin de Courtenay), who were them­ selves outstanding comparativists, rejected the piecemeal methods of the Neogrammarians by shifting attention to the synchronic state of languages

12

EDWARD STANKIEWICZ

that would provide an insight into language as a whole, rather than to those of their elements that throw light on their origin or historical development. Within this new synchronic approach the question of the relation of the parts to the whole and of the interdependence of the elements of a system could not but impose itself with particular force. The view of language as a system and the questions it entailed received the sharpest formulation in the Cours de linguistique générale of Saussure, and it is this book (published by Saussure's students in 1916) that became the vade mecum of most structural linguists. The problems raised in the Cours had already been anticipated in Saussure's youthful study on the Indo-European vowels {Mémoire sur le système primitif des voyelles dans les langues indoeuropéennes), which changed our whole view of Proto-IndoEuropean while it advanced some basic proposals on the interdependence of its elements. The words used by Antoine Meillet with regard to this book have since become one of the most quoted passages in contemporary linguistics and a kind of expression of structuralist faith. "Le Mémoire," wrote Meillet, "apportait par innovation, un système cohérent qui embras­ sait tous les faits. Dès lors il n'était plus permis d'ignorer, à propos d'aucune question, que chaque langue forme un système où tout se tient." (1927: 475). The claim that language is a tightly organized system in which "tout se tient," and which Saussure subsequently compared (in the Cours) to a game of chess, gave rise, nevertheless, to some serious reservations on the part of his followers. It was in effect Saussure's own disciple, Charles Bally, who pointed out that language is not as closed as claimed by his master since it combined different styles, older and innovating forms, and a variety of dialectal and foreign words (1932: 9ff). Edward Sapir, you may recall, put it more simply when he declared that "all grammar leaks." The problems of structure and of the types of relations that hold between the elements of a language were then put on the agenda of various schools of linguistics which gave them quite different solutions, particularly since the Cours had defined them in highly suggestive but largely inadequate terms. While most structuralists would readily accept the basic, yet age-old proposals of the Cours, namely that language is a system of signs consisting of two facets, a signons and signatum, that it is a tool of social interaction endowed with a conventional force, and that there is a difference between langue and parole, they differed from Saussure on a number of issues, including the treatment of language as a structured whole. The import of the various structuralist schools that sprang up in Europe and the United States between the two World Wars and after may in fact be judged by the content they imputed to the concept of structure and by the type of relations they attributed to language.

STRUCTURE IN CONTEMPORARY LINGUISTICS

13

Although the structural approach has in recent times been challenged by other approaches, there is little doubt that the concepts and methods devel­ oped by the former have deeply modified the science of language. From the perspective of time, furthermore, it would appear that among the various structuralist schools it was the school of Prague and its direct and indirect followers who have made the most lasting contributions to our understand­ ing of language. The significance and continuing vitality of that school is largely due to the work of Roman Jakobson, who transplanted the structur­ alist ideas of Prague to these shores and who gave them the most encompassing and compelling interpretation. Foremost among them is a cluster of concepts which grew out specifically from concerns with structure, and which include such notions as invariance and variation, opposition and hierarchy, markedness and neutralization. It is not by chance that these notions have become the catchwords and the stock in trade of most practic­ ing linguists and have found their way into a number of disciplines. These concepts and the problems they helped to illuminate have opened new perspectives into the science of language and will no doubt continue to guide its research. One such perspective was recently pointed out by Jerzy Kuryłowicz. He wrote Il est clair que le structuralisme dit classique, la doctrine la plus robuste qui ait succédé à l'école néogrammairienne, ne saurait être borné à la description et l'histoire de langues particulières. Il est en même temps évident que justement dans le domaine du comparativisme il est loin d'avoir épuisé toutes les possibilités. (1977: 5). The possibilities which this "robust" and "classical" structuralism has opened up obviously do transcend the field of comparativism, though at this point one can hardly envisage their direction or limits. But since the problems of structure pertain to the fundamental relations of language, one can hardly assume that their investigation will cease, no matter under what name such studies will ultimately evolve. In the following remarks I shall give a short survey of the history of ideas that have gravitated in one or another form toward the problems of struc­ ture, and then offer some proposals of how the overall problems of structure might be tackled and further pursued. 2. The questions of structure, like most European grammatical ideas, were first advanced by the Greeks, who formulated the concept of phonetic invariants (stoikheia), of grammatical categories, and of parts of speech. Together with the description of paradigms and cases, they posited the question of oppositions and the difference between marked and unmarked cases (ptōsis). They were also the first to recognize the two-fold nature of the linguistic sign and of the diverse functions of language. The concept of

14

EDWARD STANKIEWICZ

language as a structured whole arose most clearly during the Renaissance when the new national languages came to vie with each other and with the classical languages. It was at that time that the phonetic and grammatical properties of a language became the subject of extensive research, and that each language began to be seen as an ordered, autonomous, and systematic whole (una struttura diversa or adunanza regolata [Stankiewicz 1982: 177ff]). The true forerunner of modern linguistics was, no doubt, Wilhelm von Humboldt. It is curious that Noam Chomsky claimed to find an affiliation between himself and von Humboldt, for the ideas of the latter were in every respect different from the logistic and universalist approaches of the eight­ eenth century grammarians in whom Chomsky justly saw his true prede­ cessors. Imbued with the esthetic ideas of the Romantics (who saw an inseparable bond between the content and form of a work of art), Humboldt rejected the Rationalist belief in concepts that existed prior to and apart from their material expression and in expression which did not serve some functional end. The unity of meaning and form was for him an unshakeable axiom which determined the so-called "genius" of a language and which accounted for the diversity of the languages of the world. Consequently he also rejected the eighteenth century belief in a universal language which was propounded by the "philosophical" or "rational" grammars. "Il serait entiè­ rement chimérique," he wrote, "de vouloir former de toutes des différentes qualités une même langue universelle, qui deviendrait vide, si elle faisait abstraction des caractères distinctifs, et contradictoire, si elle admettait tous à la fois." (1906a: 308). In recognizing the basic diversity of linguistic systems he did not, however, deny that they shared a number of common, universal traits which he tried to capture by studying languages of the most diverse linguistic types. As a consequence he replaced the concept of an a priori universal grammar with that of empirically attested linguistic universais. The study of individual languages also led him to emphasize, in the same way as Saussure, the interdependence of the various elements of a language and the sense of cohesion they bring to the whole: "Es gibt nichts Einzelnes in der Sprache," he wrote, "jedes ihrer Elemente kündigt sich nur als Teil eines Ganzen an" (1969: 2). Although he shared with his era some typically idealistic ideas (such as the belief in the preformation of languages, or in the striving of so-called "inner form" for phonetic realization), his concept of structure remained throughout flexible and open-ended: "Die gramma­ tischen Elemente," he asserted, "bilden dagegen mit dem Character des Ganzen nicht gleich sichtbar verbundene Teile" (1906b: 372). The above-mentioned ideas of Humboldt hardly found an echo in the contemporary German milieu, where a one-sided historicism and the dis­ covery of Sanskrit came to replace Humboldt's typological and structureoriented approach. The harbinger of the new "archeological" and historical

STRUCTURE IN CONTEMPORARY LINGUISTICS

15

approach was the well-known booklet of Friedrich Schlegel {Über die Weis­ heit und Sprache der Indier, 1808) which found in the Ablaut of Sanskrit and other Indo-European languages the expression of a primordial "organic" {synthetic) unity which it deemed to be intrinsically superior to the "mechanical" nature of the modern analytic languages (especially of French). Influenced by Georges Cuvier's "principe de corrélation des for­ mes" and by the latter's belief that the fragment of any extinct biological organism may provide an insight into its original complete structure {"déchiffrer et restaurer [le tout]...par un seul fragment d'os" [Lepschy 1962: 179]), Schlegel became the initiator of that comparative method for which the study of the modern languages became merely a means to an end, namely to the reconstruction of the lost protolanguage. "The decisive point which will illumine all," Schlegel wrote, "is the inner structure or comparativegrammar which will throw light on the genealogy of languages in the same way in which comparative anatomy has illuminated the history of higher nature" (1808: 15). The subsequent fate and achievements of comparative grammar are, I believe, well known. The gradual perfection of the comparative method gave us an ever more reliable and precise instrument for the reconstruction and genetic comparison of languages, but diverted our sight from the actual, synchronic nature of languages, as well as from the problems of language as a whole. By the time the Neogrammarians came around, linguistics had become a science of "blind" phonetic laws, of limitless and continuous change, of star-studded pages of reconstructed forms (in the words of A. H. Sayce 1890: xvi, "an empty clatter of roots and suffixes"), and of catalogues of haphazard grammatical forms. The study of abstract relations, wrote Hermann Paul, could not yield any fruitful results; the task of the linguist is "die Wirksamkeit der Faktoren isoliert zu behandeln." (1960: 24). By identifying the study of sounds with that of physiology and the problems of analogy with that of psychology, the Neogrammarians had in fact lost the sense of their own discipline. It is not surprising that in calling for a renewal of linguistics, Saussure insisted above all that it must become a science "en elle-même et pour elle-même". This science was, according to the Cours, bound to recognize two principal truths: first, that language is a system of signs (and therefore also a part of the more general science of sémiologie), and second, that it is a system of oppositions which define the very character of the linguistic elements. The importance of going beyond the mere observation of facts was similarly emphasized by Saussure's con­ temporary, Baudouin de Courtenay. "The linguist," Baudouin wrote, "must not be content with the registration of haphazard facts but must attempt to discover the genuinely distinctive properties of language...the laws hidden in the depth, in the intricate combination of various elements" (1972: 276).

16

EDWARD STANKIEWICZ

Contemporary linguists have repéatedly paid tribute to the pioneering work of Saussure, while Baudouin's work was hardly known in the West and received short shrift even from such linguists as Nikolai Trubetzkoy. Yet it is apparent that the basic insights into the structure of language came from Baudouin, whose approach was, to begin with, more flexible and profound than that of Saussure. The differences between the Swiss and the Polish linguist stemmed partly from the fact that Saussure's theory had a sharper polemical edge and was largely divorced from his practice (he wrote little and only on historical problems), whereas Baudouin combined descrip­ tive and historical work and achieved a more significant synthesis by deepen­ ing linguistic analysis. In reacting to the historical methods of the Neogrammarians, Saussure attributed a systematic character only to synchrony, declaring diachrony to be a matter of contingency and chance; at the same time, he set up a gulf between langue and parole, ascribing the latter to the idiosyncrasies of the speaker. He hardly touched upon problems of phonology and relegated syntax to the realm of parole. Though Saussure emphasized more strongly than Baudouin the importance of phonological and grammatical oppo­ sitions, he interpreted them in a highly formal and schematic way. The phonemes of a language, he claimed, "ont une valeur purement oppositive, relative et négative" (1949: 164), while in language as a whole, "il n'y a que des différences sans termes positifs" (1949: 166). The disregard of the positive, phonetic and semantic properties of language made him treat all the elements of language as equally abstract algebraic symbols, and prevented him from seeing their diverse roles in the hierarchy of a system. His separation of abstract "valeurs" from their underlying substantive properties affected profoundly the field of phonology, which was seen to consist, on the one hand, of purely relative and psychological images acoustiques, and, on the other hand, of an extralinguistic physical substratum. Such a dualistic conception of language continued to dominate at least two schools of structuralism which came in a direct or circuitous way under the influence of the Cours: Louis Hjelmslev's glossematics and American descriptivism. The first took the more idealist road by recognizing as relevant only the study of relations devoid of semantic content and phonetic expression; whereas the second, with its more pragmatist and behaviorist traditions, saw as its foremost goal the identification of linguistic units, reducing the question of their relations to the lowest denominator, to that of their linear distribution. As a result, the first lost sight of the specificity of language, treating it on a par with any system of signs, while the second never went beyond the study of "items and arrangements", which it treated in a similar, homogeneous way on all levels of language. Baudouin de Courtenay was in every respect more flexible than Saussure.

STRUCTURE IN CONTEMPORARY LINGUISTICS

17

In the first place, he saw language itself as a variable, dynamic and open system both in the makeup of its parts and in its social and historical layers. Unlike Saussure, who identified synchrony with statics and diachrony with dynamics, Baudouin insisted on the presence of dynamics in synchrony, and on the significance of invariance in diachrony. Historical change was not for him a matter of parole and accidental drifts, but was, on the contrary, closely related to the stratified organization of the linguistic code. "There is a limit," he wrote, "to the changeability of sounds," for certain features and oppositions persist and are immune to change throughout the history of a language. The relation of the stable and variable elements of a system was discussed by him in two masterful works {History of Latin Phonetics and Survey of the History of the Polish Language (1893, 1922), which became the foundation of a new structural approach to linguistic diachrony. Another insight which grew out of his work on phonetic alternations (and which he elaborated together with his student Mikołai Kruszewski) was the recog­ nition of the interplay between the phonological and morphological elements of a language. By adopting a consistent functional approach he was able to show that the variants of sounds which are in one stage determined by their phonetic environment acquire new morphological functions when the orig­ inal phonetic conditions are changed. He thus went beyond the phonetic biases of the Neogrammarians, as well as beyond contemporary "systematic phonology", which attempts to define all variants of morphemes in terms of phonetic rules. The interplay of phonetic and morphological change was discussed by Baudouin in a number of other pioneering studies. Finally, Baudouin was, the first to pin down the nature of phonological relations, by showing that they do not involve global units, but units that are defined by their common and distinctive traits. "Requirements of scientific analysis," he wrote, "do not allow us to stop with the phonemes. The phonemes consist of ultimate psychological (articulatory and acoustic) elements," which are "interconnected with each other" and which "form a system of oppositions" (1972: 267). The psychological bias of his time did not, however, allow him to explore the consequences of his insight, and he too failed in the end to see the inseparable link between the invariants and variants of a phonemic system. Baudouin's work had a decisive influence on the Prague Circle, which pursued his ideas to their logical conclusion and taught us to see language as a complex of levels and as a network of formal and functional relations. That Baudouin was the primary influence on Prague is apparent from the very areas of its research, which included phonology and morphophonemics, the relation of synchrony and diachrony, linguistic diffusion and the broadest approach to the questions of structure. But it is, of course, thanks to their own work and especially that of Jakobson, that the linguists of Prague

18

EDWARD STANKIEWICZ

imposed on contemporary linguistics a new stamp by converting its early emphasis on description and synchrony into a quest for predictive and universal laws. The absorbing concern with structural relations which define the very elements and levels of a language was tellingly described by Jakobson in one of his "Retrospects": "Perhaps the strongest impulse toward a shift in the approach to language and linguistics was for me, at least, the turbulent artistic movement of the early century.... Those of us who were concerned with language learned to apply the principle of relativity in linguistic operations [and learned] from cubism that everything is based on relationship with the interaction of parts and wholes" (1962: 631-632). The structured patterning of phonological systems was formulated by Jakobson as early as in 1929 (in the Remarques sur l'évolution phonologique du russe comparée à celle des autres langues slaves) when he pointed out the implicational relations of phonemic oppositions according to which the presence of a category A (such as pitch) presupposes the existence of a category  (such as length), or excludes the presence of another category (such as the presence of stress excludes the occurrence of length). Although some of the established correlations have proven to be too extreme, the problem of interdependence of elements was given a precise and compelling form, which was subject to empirical verification on the languages of the world. The formulation of implicational laws required thus at the same time an advance in typology, where the hypothetical and empirical aspects would abet and complement each other. The above-stated implicational laws acquired even greater explanatory force with the introduction of the concept of markedness. According to this concept the marked members of an opposition are endowed with features that are absent in their unmarked counterparts and are less likely to combine with other features or categories; they are also more susceptible to historical change. It is easy to see that underlying the concept of asymmetry is the notion of symmetry since there is, apparently, a balance in the amount of information linguistic forms may convey. Jakobson's consistent application of distinctive features revealed, finally, not only the relational nature of phonemic oppositions, and the interdependence of variants and invariants, but also their hierarchical organization, which defines the limits of phono­ logical systems. As a result the phonological systems of natural languages appeared themeselves to be variant realizations of universally available phonemic possibilities. Two more studies by Jakobson have, in my opinion, a special bearing on the problems of structure, as they specify the relations between various linguistic components and levels. One of them ("Prinzipien der historischen

STRUCTURE IN CONTEMPORARY LINGUISTICS

19

Phonologie" 1931) raises the question of the interdependence between phonemic variants and invariants which are liable to switch their functions in the historical development of a language, whereas the other study ("The phonemic and grammatical aspects of language in their interrelations", 1949) makes it clear that the phonemic system of a language cannot be fully comprehended apart from its role in morphology. Thus certain pho­ nemes or features carry a heavy morphological load (e.g., the role of dental consonants in the English inflectional endings), whereas the others are used for purely lexical ends. These works of Jakobson, as well as those of a number of linguists whom I have here ignored for lack of space, have significantly advanced our understanding of language by pointing up the intimate unity of elements and relations both in their synchronic and diachronic stages. 3. After this survey of the concept of structure as it evolved in the history of our science, I shall now advance some ideas of how we may broach the problem of structure in language as a whole, i.e., with regard to the relations between the various levels and the relations between the elements of each particular level. My discussion will of necessity be schematic because of the limitations of time and because some of the problems require further and far more detailed analysis. Students of so-called "rich systems" — of which language is one — have defined them as consisting of a number of parts that interact with each other in complex ways. The interrelations between the parts (the so-called "intercomponential relations") are in such systems looser than the relations that hold between the elements belonging to the same part (the "intra-componential relations") (Simon 1962: 468ff). This definition applies eminently to language if we substitute the notion of "level" for that of "part". The primary problem we face is the determination of the kinds and number of levels, since there is no general agreement on the matter. A lack of functional criteria would make it almost insurmountable. A case in point is the treatment of morphophonemics, which some linguists have defined as a level that is no longer phonology and not yet morphology. Some approaches to transformational grammar have done away with morphology, and have lumped together phonology and morphophonemics; word forma­ tion and lexicology are treated by some scholars as different levels. Since function is in my opinion the only solid criterion for the discrimi­ nation of levels, I shall distinguish only three traditional levels: a level of phonology, a level of morphology (in the broad sense of the word), and a level of syntax. Phonology deals with the sound system of a language (i.e. with those elements which pertain to its "second articulation" and which are known in Russian linguistics as niže znaka). It is the level that includes

20

EDWARD STANKIEWICZ

both the phonemes and their variants, or in other terms, the distinctive, as well as the redundant, configurational, and expressive features. Morphology is the study of those elements which carry a meaning, or in Peirce's terms, those which have an "interpretant". This level encompasses the grammatical and lexical elements of a language, or those which hinge on the word as the carrier of grammatical and/or lexical meanings. The word itself is, however, a heterogeneous and ill-defined unit and requires for its interpretation refer­ ence to the underlying grammatical structure of a language. The level of syntax, finally, deals with the various types of sentences and their internal relations. The three linguistic levels are hierarchical and transitive: they form socalled "structures emboîtées" in which the lower levels are in the service of the higher levels without relinquishing their own autonomy. Consequently, each level can be, and has historically been, described both in terms of its internal structure (its intra-level relations) and in terms of its relations to the neighboring level(s), and to language as a whole. The hierarchical status of the three levels finds further reflection in their relative complexity and in the freedoms and constraints each of them pre­ sents. The phonological level is maximally coded and allows minimal inter­ ference on the part of the speaker. The morphological level is far more complex, involving both meanings and forms, and allowing various degrees of freedom and choice. It is minimally coded in the field of word formation and maximally constrained in its grammatical categories. But even in this area it allows the substitution of marked by unmarked terms, and a choice between synonymous forms (e.g. such as the use of modals for commands). The level of syntax allows a whole gamut of freedoms because the limits of a sentence are not strictly defined. Furthermore, the form of a sentence may vary according to verbal or situational context (as in the use of ellipsis), according to the relation of the speakers, and according to the prevailing written or spoken norms. It is not by chance that syntax has often been invoked to illustrate phenomena of linguistic creativity. Despite their functional and formal differences, the three levels of language exhibit a pervasive isomorphism. Thus each level is heterogeneous, compris­ ing different components and functions; each contains primary and obliga­ tory elements, and elements which admit a greater or lesser degree of variation; each comprises monofunctional units, as well as units which carry more than one function; and each consists of a number of subsystems which vary in their hierarchy and mutual relations. It is the hierarchy, parallelisms and interdependence of the levels that make language into a cohesive, as well as a highly flexible system. I shall now develop some of these points beginning with the level of phonology.

STRUCTURE IN CONTEMPORARY LINGUISTICS

21

4. A century of inquiry into phonological systems has given the field of phonology an almost finished form and has made it, as Jakobson put it, into "a methodological model for all other areas of linguistic analysis" (1960: 428). Thus there is no need for me to repeat what I have in part stated above and what are its indisputable methodological achievements. What I would like to emphasize, however, is the coexistence of complemen­ tary and asymmetrical relations and the articulation of sound systems into partial patterns. The coexistence of different and partial patterns is to some extent due to the fact that in addition to the marked and unmarked members of opppositions, phonological systems include equipollent and neutral terms. To the former belong phonemes which are equally marked or unmarked (e.g., the labial vs. the velar consonants, or the stressed vs. the long vowels), whereas the latter include phonemes which remain outside a given phonemic opposition or which combine the features of the opposite terms (e.g., syllabic sonants, which combine the features of vowels and consonants). Another source of internal asymmetry is the lack of correspondence between markedness and neutralization. While such a correspondence often does prevail, it is by no means a general law. Thus, for example, we find more prosodically marked (stressed or long) vowels than unmarked (unstressed or short) vowels, which are easily prone to reduction and loss. Further asymmetries arise in the paradigmatic and syntagmatic behavior of phonemes. Thus vowels and sonorants are generally neutral with respect to voicing, whereas in some languages they behave syntagmatically like voiced consonants (e.g., the voicing of final voiceless conosnants before η in Greek and Japanese or before initial vowels in southern Polish dialects). There are, furthermore, reversals in the markings of some phonemes. Thus, nasal consonants are optimally diffuse (m, n), whereas nasal vowels are optimally compact (œ ą); palatal consonants are optimally spirants (š, ž) , whereas nonpalatal consonants are optimally stops. Another source of linguistic asymmetry lies in the relation between the distinctive and redundant features, for the wider the scope of phonemic distinctions (within a given set or subsystem of a language), the narrower the range of redundant variation. Otherwise, the redundant (concomitant) features regularly support the distinctive features in sharpening the various phonemic oppositions (e.g., long vowels are regu­ larly tense, while short vowels are lax; stops are mellow, while spirants are strident; back vowels are rounded, while front vowels are unrounded). Some distinctive features or phonemes are, on the other hand, inherently multi­ functional in that they simultaneously perform distinctive and configurational functions (e.g., the prosodic features of most languages mark the unity or boundaries of words; the English phonemes h and ŋ signal respect­ ively the beginning and end of a word). Thus, the phonemic system of a language resembles, in effect, a typological field with a number of coexisting

22

EDWARD STANKIEWICZ

elements and functions exhibiting parallel or complementary relations. The coexistence of heterogenous elements and relations imparts to the system a sense of order and tension, or, in other words, a state of unstable equilibrium which opens up possibilities for their historical transformation. The multifunctionality of phonological elements bring us in turn to the question of the relation of phonology to morphology, or more specifically, to the role of phonemic elements in the grammatical system of a language. The use of such elements is highly selective and is known to vary from language to language, as well as within each individual language. Thus the Semitic languages make the widest grammatical use of their vowels, while English employs in its desinences only the dental consonants -d/-t, -n, -z and the velar ŋ. Similarly, Russian makes foremost use of its vowels in the inflection of nouns, but gives precedence to consonants in the endings of the verb. The linguists of Prague, who took up these problems in the footsteps of Baudouin, treated them primarily from the viewpoint of pho­ nology (they even coined the term "morphological phonology"). However, Baudouin was closer to the truth when he spoke of the "morphologization" of phonetic distinctions. The difference in viewpoint is by no means trivial for it has a bearing on the relation between levels, which we shall also encounter when we move from morphology to syntax. The point is that phonology provides but the means for grammatical distinctions, but it is morphology that restructures them for its own ends by assigning to them specific grammatical functions and by placing them in relation with the other morphological devices of a given language (such as the various affixes, morphophonemic alternations and the zero, which is pertinent only to morphology). The difference between the morphological and phonological functions of sounds is not, it should be added, absolute, for there is no rigid borderline between the automatic and non-automatic alternations of sounds. 5. I shall pass now to the level of morphology, which is far more complex than phonology because it includes such diverse components as form and meaning, grammatical categories, lexicon and word derivation. My purpose will be, as above, to indicate the interdependence of the "parts". Traditional grammars have oscillated between an emphasis on one or another of the morphological components: some have concentrated on the grammatical meanings (in particular on the parts of speech), whereas some modern scholars (beginning with Schleicher) have attempted to reduce mor­ phology to the study of forms and to slip in surreptitiously reference to meaning. The view that morphology is the study of the "word" has a far stronger claim since the word is essentially a nexus of lexical and grammatical meanings, and it is precisely these two types of meanings which make up the semantic system of a language. However, since the words themselves

STRUCTURE IN CONTEMPORARY LINGUISTICS

23

fall into different classes we must turn to the latter, i.e., to the parts of speech, for an analysis of the morphological structure of a language. Grammatical practice (and the very names of the classes) has frequently led to the belief that the parts of speech are categories of syntax. This view is misleading since different parts of speech may perform the same syntactic function (e.g., the noun and the infinitive may function as subjects), while the same part of speech may perform different syntactic functions (e.g., the participle functions as an attribute, and the finite verb as a predicate). The traditional division of words into parts of speech retains its validity, provided we define them in consistent grammatical terms and state their mutual relations and hierarchy. In the first place we should exclude from the level of morphology those words and forms that have a strictly syntactic function. These include the so-called syncategorematic words (conjunctions, prepositions, and modal qualifiers) and forms that function as equivalents of sentences (i.e., interjections, phatic formulas, the vocative and imperative). Morphology will thus include those classes of words which we define, by analogy with phonology, as bundles of grammatical categories. The extreme poles of these classes are those which are maximally marked. They include, on the one hand, pronouns, which are totally grammatical and lack lexical meaning, and, on the other hand, adverbs, which are almost totally lexical and lack grammatical categories (except for the comparative). Pronouns are not only totally grammatical, but a class of classes, for they comprise nominal, adjectival, adverbial and verbal forms (e.g., English I, you; mine, his; here, whence; to be, to do). The fact that pronouns duplicate, as if in miniature, the principal parts of speech — minus their lexical meanings — could make it appear that they are redundant and exist only to tax the memory of a speaker. In truth, they are the most universal categories of language, which establish a direct link between the speaker and the context and can thus dispense with any lexical information. A sentence such as I am here now can, no doubt, be rendered in any language of the world. The universality of pronouns may serve to remind us that language is inseparably rooted in the speech act, and that egocentricity is one of language's foremost characteristics. It is not my purpose to go through the various parts of speech, but rather to highlight those features which tie them together as opposite as well as complementary sets. A comparison of the Russian noun with the nominal pronouns may illustrate my point. While the nominal system expresses the grammatical categories of gender, number and case, the nominal pronoun involves, in addition, such concepts as the distance of the referent from the speaker (in such forms as on "he", tot "that", ètot "this"), the concepts of question and quantification (in such forms as kto "who", čto "what"; nékto "someone", któ-to "somebody", ničtó "nothing") and the reflexive (sebé,

24

EDWARD STANKIEWICZ

sebjä "oneself"). On the other hand, it neutralizes some of the distinctions of gender, number and case (cf. the overlap of the genitive-accusative and dative-locative singular in the Russian personal pronou ja "I"), and reverses the markings of the singular and plural (the plural of the noun is marked, while the plural of my "we", vy "you" is unmarked with respect to ja "I", ty "you"). A similar inverse relation marks the expression of number in nouns and in the grammatical quantifiers (among which I include the numerals, which traditional grammars treat as a separate part of speech). In some Slavic languages the plurals of nouns distinguish collective/noncollective and counted/noncounted forms, but these distinctions are the exception rather than the rule. The quantifiers express these oppositions quite regularly in addition to a number of other quantitative notions (such as the inclusive both vs. noninclusive two, the distributive each vs. the nondistributive some). On the other hand, it is well known that they neutralize most grammatical oppositions carried by the noun (number, gender, and case) or may be converted, like the adverbs, into purely lexical terms. The principle of neutralization and complementarity also defines the relations between grammatical categories belonging to one and the same class. Thus the Slavic noun expresses in the singular the concepts of ani­ mation, sexual gender and case, whereas the plural tends to neutralize some (or all) genders and/or cases, but introduces, by compensation, new gram­ matical or lexical distinctions: e.g., in Polish, the virile gender and expression of pejoration, and in Russian, distinctions of a lexical kind (e.g., mexi "bellows''/mexà "pelts", tórmozy " obstacles" /tormozá "brakes", soboli "sables" (animals)lsoboljá "sables" (pelts), zúby "teeth"/zúb j a "gears", listy "sheets"/list'ja "leaves", kámni "(precious) stones"/kamén'ja "stones", koleni "knees'/koléna "generations; figures").1 The interplay of grammatical and lexical meanings in the plural takes us to the broader problem of the relation between the two basic components of morphology, grammar and lexicon. The interdependence of these compo­ nents can be assumed in advance, even though it has, so far, received all too little attention. Suffice it to note that two of the major grammatical categories of language — gender and voice — are inherent in the very lexical meanings of the nominal and verbal stems (in the latter only indirectly through the transitive or intransitive meanings of the verb). This interdepen­ dence comes conspicuously to the fore when we turn to that aspect of the lexicon which is more or less systematic, namely to word formation. This interdependence can be defined by two mutually related rules: (1) the more grammatical a part of speech, the less room it allows for word derivation, and (2) the more marked a given grammatical category, the stronger its tendency towards lexicalization. The first rule is best illustrated

STRUCTURE IN CONTEMPORARY LINGUISTICS

25

by the above-mentioned difference between the adverbs and pronouns: the former are minimally grammatical, but consist almost totally of derived words, whereas the latter are maximally grammaticalized, but exclude lexical derivation almost completely. A similar correlation holds to a lesser extent between the noun and the verb: the former exhibits a relatively small number of grammatical categories, but a rich roster of derivational types and forms, whereas the verb has the most complex system of grammatical categories (mood, tense, person, aspect, gender, number), but hardly admits derivation of verbs from verbs, or expresses derivation only in conjunction with gram­ matical distinctions (as in the prefixation of verbs, which in the Slavic languages is concomitant with the expression of perfectivity). The second rule applies to the marked grammatical categories of all parts of speech and yields forms of a hybrid, inflectional-derivational character. Thus, most participles can function syntactically as adjectives; infinitives can function as nouns; gerunds and the marked (predicative) forms of adjectives may function as adverbs (cf. the Russian forms sidja "sitting", gljädjä "in view of", ne smotrjä "regardless"; slábo "weekly", sil'no "strongly", trudno "difficult"; (comp.) slabée, sil'née, trudnée). The marked (oblique) cases of nouns likewise lean towards derivation (e.g., Russ. vécerom "in the evening", noc'ju "at night", iz domu "from the house", ótrodu "from birth", s ispúgu "from fear'; v soldàty "into the army", v studénty "become a student'). The conversion of marked grammatical forms into lexical derivatives is confirmed by a number of historical facts; e.g., the adverbs and adverbial predicates of such languages as French and Lithuanian were originally neuter adjectival forms (e.g., French mieux "better", moins "less", pis "worse" < Latin melius, minus, peius; Lith. geĕa "well", šlãpia "wet", saldù "sweetly", gražu "beautifully"), whereas the original Latin plurals of neuter nouns have yielded in French and Italian distinctions of a lexical or augmentative kind (e.g., Fr. brasse "fathom"/bras "arm", cervelle "brains" (as food)/cerveau "brain", feuille "leaf "/feuil (techn, "lining"); Ital. la frutta "fruits" /il frutto "fruit", la fossa "cavern, grave" /il fosso "ditch", la buca "hole, depression"/ il buco "(button)hole" (Rohlfs 1949: 85). The compensatory rules of inflec­ tion and derivation find further expression in the fact that grammatical distinctions are optimally expressed in simple stems, while they are subject to various types of neutralization in derived forms (e.g., the neutralization of aspect in Russian denominal verbs or of the singular/plural opposition in derived abstract nouns). Let us now turn to the question of the relation between grammatical meaning and form. The traditional (and Saussurean) belief in the arbitra­ riness of the linguistic sign was, as you may recall, persuasively challenged by Jakobson, according to whom a principle of iconicity pervades the system of grammar inasmuch as semantically marked categories are conveyed by

26

EDWARD STANKIEWICZ

formally marked elements, such as the addition of phonemic features, gemi­ nation, reduplication or the accumulation of morphemes (1966). The prin­ ciple of iconicity no doubt explains a number of grammatical phenomena (especially in the agglutinative languages), but is far less fruitful for the synthetic languages, in which single morphemes contain bundles of diverse grammatical meanings. It is, furthermore, contradicted by a number of forms. Thus, for example, the German and Russian marked predicative adjectives are shorter than the corresponding unmarked attributive forms; the Slavic imperatives and the aorist exhibit truncated, shorter forms than those of the present tense. A correlation between grammatical meaning and form does nevertheless exist, and it is perhaps more significant though less obvious than is assumed in Jakobson's principle of iconicity. This correlation finds outward expression in the different makeup of various grammatical categories and parts of speech, as well as in their morphophonemic alter­ nations. Thus the pronouns express their distinctions by means of suppletion; the nouns of most European languages express their grammatical meanings by means of "synthetic" suffixal forms; the verbal system, which is grammati­ cally the most complex, employs a whole plethora of morphological devices which it assigns to various grammatical categories; e.g., the Slavic verb uses ambifixation and infixation for the formation of aspects; single suffixes for the expression of the infinitive, gerund and the participle; agglutination for the expression of the past tense; and analytic constructions for the expression of the passive and mood. Equally diversified is the use of morphophonemic alternations, which participate with the endings in the expression of gram­ matical meanings. Thus the English noun admits almost exclusively alter­ nations of consonants (in such forms as wife/wives, house/houses, wreath/ wreaths), whereas the verb makes primary use of alternations of vowels (e.g., sing/sang/sung; shine/shone, bring/brought); the Russian declension employs alternations of stress, whereas the conjugation favors alternations of conson­ ants. Serbo-Croatian inflection makes vast use of alternations of stress, which in its declension moves from the stem to the endings, and in its conjugation, primarily from the endings to the stem. A similar correlation between grammatical meaning and morphophonemic alternations has been observed in a large number of languages (Kubrjakova and Pankrac 1983: 103ff). 5. Let us now pass to the level of syntax. In recent years, and mostly thanks to transformational grammar, this level has been given precedence over all other levels of language and has served as a springboard for far-reaching conclusions about the general nature of language. There is some reason to doubt, however, whether this approach will survive the proverbial test of time, as it has over the last two decades or so undergone a series of drastic revisions, relinquishing in the process some of its basic original claims.

STRUCTURE IN CONTEMPORARY LINGUISTICS

27

Among the major shortcomings of transformational grammar, I consider the insistence on a formal syntax without functions, the separation of form and meaning, the lack of a morphological level, and the methodological split of competence and performance. But it is precisely on the level of syntax that linguistic performance looms large since the production of a sentence is inevitably a goal-directed, intentional act which is rooted in speech and in situational contexts. Thus, even the declarative sentence, which is primarily set on the cognitive function, i.e., on the so-called "states of affairs", can dispense neither with the speech act as its basic coordinate, nor with an actual or imaginary addressee. The linguistic counterpart of the declarative sentence is, therefore, the interrogative, which constitutes with the former a dialectical unity. The essence of speech is, in other words, the dialogue, which involves an exchange and switching of roles between addressor and addressee. It is only in poetry, which creates its own context and reference, that the situational context can be suppressed, for even schizophrenics who have lost the capacity for speech may continue to write verse, as we know from the example of Hölderlin. The social character of speech becomes even more apparent when we turn from the declarative sentence to those types of sentences that are totally oriented toward the speech act, and have as their sole purpose the interaction of the participants. I have in mind the types of utterance that serve appella­ tive, phatic and emotive functions (e.g., the expressions come here!; how do you do?; oh, what a beautiful morning! or Russ. ax ty! "oh my!" dušen'ka "my dear"; idi-ka "go"; bóže moj "oh my God!"). What is interesting about these functions is that they constantly overlap (as in the use of interjections and diminutives in the vocative) and color even the most matter-of-fact declarative sentences. The socializing types of utterances we may consider as marked since they are inseparable from the concrete speech act, whereas the declarative sentence transcends the hi et nunc of the speech act in order to tell us something about the world. The declarative sentence is thus the bridge between language and reality (in the broadest sense of the word) since its role is to convert meaning into reference by pointing to actual or possible events. This conversion can be achieved only by means of modality or through the use of indexical symbols, both of which involve the position of the speaker. "Semantic reference" takes place on Peirce's view when, and only when, there is a conjunction of the "Secondness" of an indexical sign with the movement of linguistic signs, or "Thirdness," thereby bringing the latter to a close in a way which links it into the former" (Dewey 1946: 91). The pivot of the declarative sentence is, then, the predicate, which involves the morphological categories of mood, person and tense.

28

EDWARD STANKIEWICZ

I shall ignore henceforth the highly diversified aspects of syntax and concentrate on the simple sentence as a carrier of those categories which, together with the category of voice, make up its nuclear and indispensable elements. These elements employ in most languages the fundamental mor­ phological categories of the verb, but it is only in the sentence that the latter acquire concreteness and referential value since they appear here in specific grammatical functions (such as active or passive; present or past) and enter into a relation with the other constituents of the sentence. This is also the reason why grammatical tradition has found it expedient to resort on the level of syntax to the categories of predicate, subject or attribute in place of the morphological concepts of verb, noun or adjective. The primary and universal category of predication is mood, which estab­ lishes the truth function of the sentence or its mere possibility. However, this typological invariant appears to us only in the guise of its variants, since the languages of the world implement it in a number of different linguistic forms. Thus, for example, English distinguishes affirmative and assertorial statements, whereas other languages project "possibility" in far more intricate and idiosyncratic ways (e.g. Turkish, which distinguishes presumptive, hortative, desiderative and contrafactual moods). Linguistic "truth" is rarely a simple statement of fact, for it is colored by the attitudes and hesitations of the speakers through the intrusion of sentential qualifiers (such as indeed, really, of course, surely, apparently) or tag questions {isn't if, n'est-ce pas', neužéli). The predicates of most languages carry, in addition, the category of tense, which indicates the position of the speaker with respect to the narrated event. This near-typological invariant presents, in turn, a number of more specific variants, especially in the expression of the past, which is universally more complex than that of the future. No less fundamental is the category of person which designates the participant of the narrated event which may coincide with the participants of the speech act or which designates a third person, i.e., one which is outside the speech event. As a rule, the person of the verb is in concord with the subject of the sentence, but the third person may also mark the absence of a subject (as in Latin pluit "it rains", Italian piove "it rains", English there is, Russian nétu déneg "there is no money"). It is thus a neutral category which may, but need not, signal the presence of a subject. Since the expression of person cannot be omitted, languages that fail to render it consistently in the verb depend in impersonal constructions on the use of an unspecified subject, i.e., the third person demonstrative pronoun (e.g., Engl, it rains, Germ. es gibt "there is", French il y a "there is"). The expression of person brings us, in turn, to the category of voice with which it is inseparably linked. If the former three categories establish the

STRUCTURE IN CONTEMPORARY LINGUISTICS

29

relation of the speech act to the narrated event, and ultimately the factual import of the sentence, voice defines the nuclear structure of the sentence by specifying the relation of the narrated event to its participants (or in logical terms, the relation of the predicate to the arguments). The fact that the declarative sentence does not require a subject has led some scholars to the belief that the predicate alone makes up the basic component of the sentence. This view cannot be maintained, for impersonal sentences constitute special, semantically marked cases: they indicate specifically the irrelevance of a subject (such as the absence or presence of a certain quantity; e.g., Russian nétu "there ain't", pribylo vody "there was an increase of water"; Spanish no hay "there isn't"; Polish niema "there is not", było ich trzech "there were three of them") or phenomena beyond human control (e.g., Russ. nel'zjä "it is not allowed", móžno "it is possible", nádo "it is necessary", mne strax "I am afraid", mne xólodno "I am cold"; Italian bisogna "it is necessary'; fa freddo "it is cold"). The pivotal importance of voice for the nuclear structure of the sentence was forcefully stated by Sapir: "Speech consists of a series of propositions. There must be something to talk about and something must be said about the subject of discourse once it is selected.... The most common subject of discourse is a noun, and the thing predicated of a subject is generally an activity ... in other words, the verb.... No language wholly fails to distinguish noun and verb.... It is different with other parts of speech. Not one of them is imperatively required for the life of language" (1921: 126). The central constituents of a sentence are then the subject and the predicate in which the former is normally the topic of discourse and the latter its comment. We must distinguish, in turn, two types of predicates: (1) static predicates, which designate a state and involve a subject and predicate, and (2) dynamic predicates, which designate events that involve a subject and one or more objects (i.e., two or three place predicates). The former is rendered by means of an auxiliary plus a nominal form (Germ, mein Bruder is gut "my brother is good'; mein Bruder ist ein Soldat "my brother is a soldier'; Russ. moj brat xoròs "my brother is good'; mojá doč" učitel'nica "my daughter is a teacher'), or through an intransitive verb (the boy stands or sleeps), whereas the latter utilize a transitive verb (Lat. doceo pueros grammaticam "I teach the children grammar'; Engl. I teach him Latin; Russ. ja dal emú knígu "I gave him a book"). It is the latter construction which exhibits the sentence in its fullest form since it makes room for the principal satellites of the predicate: the subject and object, or the agent and patient(s) of the narrated event. The difference between the two types of sentence is basic but not absolute. Thus the transitive verbs (such as the English hang, burn, break, run, walk) acquire the meaning of state when the object is

30

EDWARD STANKIEWICZ

ignored, whereas the auxiliaries turn, become, grow, wax are far more dynamic than the auxiliary is. Similarly, Russian intransitive verbs become dynamic when they express a goal, i.e., when they shift from the imperfective to the perfective (cf. Russ. xodil po úlice "he walked down the street" vs. peresël úlicu "he crossed the street"). Grammatical voice, which conveys the relation of patient and agent, is a linguistic invariant which acquires speci­ ficity only through its variants. Thus most languages express the agent/ patient relation by marking the patient either through the use of the passive voice or through a special form of the object, whereas some European languages (Slavic, Yiddish, Italian) introduce, in addition, a third, neutral category of voice, the so-called reflexive, which leaves the agent/patient relation unmarked or expresses it only through the verbal context (e.g., Russ. sobáka kusáetsja "the dog has the habit of biting" vs. Sónja mòetsja "Sonya washes herself"; posúda mòetsja Sònej "the dishes are washed by Sonya"; Ital. il giornale si vende "the newspaper is being sold" vs. la ragazza si vende "the girl sells herself"). Ergative languages, on the other hand, mark the agent itself for the expression of the patient (through an ergative or other oblique case) or leave the expression of the patient unmarked (through the so-called absolute case). The predicate and its two major satellites, the subject and object, constitute the core, or "grammatical" nucleus of the sentence; but its other constituents are by no means as purely "semantic", as claimed by Kuryłowicz (1960), for they too are subject to more or less strict rules for the expression of their particular syntactic functions. But since these functions are not as tightly structured as those of the nuclear components, they depend on the support of such syntactic exponents as prepositions, conjunctions, word order and intonation. It is the interplay of these exponents with each other and with the morphological resources of a language which secures the cohesion of a sentence and the communicative efficacy of the message. The interplay of these elements comes clearly to the fore in the use of word order, which may vary its rules depending on the presence of other syntactic exponents. Although all languages have canonical word orders, which define the positions of the major syntactic constituents, these orders can be reversed for the sake of stylistic or configurational functions once the primary relations are unambiguously expressed by means of grammatical cases and/ or prepositions. As a final resort there is always the context of the speech act which, to quote again Sapir, is "that background of mutual understand­ ing which is essential to the complete intellegibility of all speech" (1921: 97). 6. The foregoing remarks suggest the following conclusions. No analysis of language can afford to ignore its semiotic and functional aspects. It is function that determines the articulation of language into levels and the

STRUCTURE IN CONTEMPORARY LINGUISTICS

31

mutual relations of their components. The components of each level interact because they support or complement each other in their functions. The various levels form, in turn, a hierarchical, pyramidal structure in which the higher levels make use of the lower ones, which they partially transform and enrich with their own components and functions. A well-known example is the Latin sound i which is a phoneme on the phonological level, a morpheme on the morphological level, and an exclamatory sentence on the syntactic level. But although each lower level is subordinate to a higher level, it never loses its autonomy and distinctive structure. Thus it is not merely a means towards an end, but also an end in itself. The reciprocal relation of means and ends is apparently a feature of all complex systems, as was pointedly brought out by Immanuel Kant in his discussion of biological structures: "An organic product of nature is one in which each goal is at the same time also a means," and in which "each part exists for the sake of the other parts and all of them for the sake of each" (19001978, vol. 5: 376; vol. 3: 15). For students of poetry this formulation should have an immediately recognizable ring, though it is useful to keep in mind that the relations that characterize one type of system are not necessarily homologous with those that obtain in other systems. The same caveat applies when we study the various languages of the world. Although all of them have basically the same design, the variant realizations of this design are by no means less significant than their invariants. In this tension between typological variation and invariance Emile Benveniste saw a fundamental linguistic "truth," and I shall round up my paper by quoting him in full: Il faudra se pénétrer de cette vérité que la réflexion sur le langage n'est fructueuse que si elle parte d'abord sur les langues réelles. L'étude de ces organismes empiriques, historiques, que sont les langues demeure le seul accès possible à la compréhension des mécanismes généraux et du fonctionnement du langage. (1966: Avant-propos).

Notes 1. The translations of Russian, Polish and Lithuanian examples here and on the following pages are approximate, because we are dealing with subtle nuances.

References Bally, Charles. 1932. Linguistique générale et linguistique française. Paris. Baudouin de Courtenay, Jan. 1893. Iz lekcij po latinskoj fonetike. Voronež. . 1922. Zarys historii języka polskiego. Warsaw.

32

EDWARD STANKIEWICZ

. 1972. "Phonetic Laws". A Baudouin de Courtenay Anthology, ed. by E. Stankiewicz. Bloomington, Ind: Indiana University Press. Benveniste, Emile. 1966. Problèmes de linguistique générale. Paris. Dewey, John. 1946. "Peirce's Theory of Linguistic Signs, Thought and Meaning". The Journal of Philosophy 43. Godei, Robert. 1957. Les sources manuscrites du Cours de linguistique générale de F. de Saussure. Paris. Humboldt, Wilhelm von. 1906a. "Essai sur les langues du nouveau Continent." Gesammelte Schriften, vol. 3. Berlin. . 1906b. "Grundzüge des allgemeinen Sprachtypus". Gesammelte Schriften, vol. 7. Berlin. . 1969. "Uber das vergleichende Sprachstudium in Beziehung auf die verschiedenen Epochen der Sprachentwicklung". Werke, vol. 3: Schriften zur Sprachphilosophie, ed. by Α. Flitner and K. Giel. Darmstadt. Jakobson, Roman. 1929. Remarques sur l'évolution phonologique du russe comparée à celle des autres langues slaves. Reprinted in 1962/1971, 7-116. . 1931. "Prinzipien der historischen Phonologie". Travaux du cercle linguistique de Prague 4, 247-67. In French version in 1962/1971, 202-220. In English version in 1978, Readings in Historical Phonology, ed. by P. Baldi and R. Werth, 103-120. University Park: Penn. State Press. . 1949. "The Phonemic and Grammatical Aspects of Language in their Interrelations". Reprinted in 1971, 103-114. . 1962. "Retrospect". In 1962/1971, 629-658. . 1962/1971. Selected Writings I: Phonological studies. The Hague/The Hague-Paris: Mouton. [First edition/Second, augmented edition]. . 1966. "Quest for the Essence of Language". Reprinted in 1971, 345-359. . 1971. Selected Writings II: Word and language. The Hague-Paris: Mouton. Kant, Immanuel. 1900-78. Gesammelte Schriften, vol. 3, vol. 5. Berlin. Kubrjakova, E. S. and Pankrac, Ju. G. 1983. Morfonologija v opisanii jazykov. Moscow. Kuryłowicz, Jerzy. 1977. Problèmes de linguistique indo-européenne. Wroclaw-Warsaw-Cracow. Lepschy, Giulio . 1962. "Osservazioni sul termine struttura" (A proposito di Sens et usage du terme structure dans les sciences humaines et sociales, édité par R. Bastide). Atti della Scuola Normale Superiore di Pisa 31, 3/4. Meillet, Antoine. 1937. Introduction à l'étude comparative des langues indoeuropéennes. Paris. Paul, Hermann. 1960. Prinzipien der Sprachgeschichte. Tübingen. Peirce, Charles Sanders. 1966. Selected Writings, ed. by Ph. P. Wiener. New York. Rohlfs, A. 1949. Historische Grammatik der italienischen Sprache, Vol. 2. Berne:Francke. Sapir, Edward. 1921. Language. New York. Saussure, Ferdinand de. 1949. Cours de linguistique générale. Paris. Schlegel, Friedrich. 1808. Über die Sprache und Weisheit der Indier. Heidelberg. Simon, H. P. 1962. "The Architecture of Complexity". Proceedings of the American Philosophi­ cal Society 106:6. Stankiewicz, Edward. 1982. "The 'Genius' of Language in Sixteenth Century Linguistics". Logos Semantikos, 177-189. Berlin-New York-Madrid. Trnka, Bohumil. 1966. "On the Linguistic Sign and the Multilevel Organization of Language". Travaux linguistiques de Prague 1. University of Alabama Press.

Grasping the Nettle: Variation as Proof of Invariance Erica C. Garcia Leiden University But I tell you, my lord fool: out of this nettle, danger, pluck we this flower, safety. Hotspur (Henry IV, i, II, iii)

1. Introduction: The paradox* The problem posed by the search for invariance is frequently seen as a forced choice between the two equally uncomfortable horns of an unpleasant dilemma, namely the sacrifice of either descriptive elegance or psychological truth, cf.: Both in the psychology of semantics and in the psychology of phonology/ mor­ phology we have the following choice: on the one hand, non-redundant, economic descriptions which eliminate "surface entities" or allo-forms by reducing them, via derivation rules, to "basic" entities; on the other hand, descriptions which merely state the relations between partly similar "surface entities" and qualify as redundant or non-eliminative by virtue of not trying to reduce the recurrent similarities to a common, "basic" source. To all appearances, it is the latter alternative which represents the way that the (untrained) human mind operates. This does not mean, however, that the former alternative is without scientific value (Itkonen 1983: 282). A more ambitious and interesting position would be one where the analyst has his cake and eats it too, by postulating an invariance that not merely underlies but in fact motivates variation. 1 That this goal is worth striving for is suggested by the fact that the very notion of "variation" presupposes, after all, the identification within diversity of a constant sameness: "vari­ ation" would otherwise reduce to mere difference, and pass unnoticed. The relation between invariance and variation is, therefore, an issue that lies at the very heart of linguistic theory, since linguistic analysis should yield a finite, ideally non-arbitrary description of whatever it is that accounts for the infinitely variable, creatively productive, free whimsical ad hoc use of language. Conversely, open-ended communication among human beings

34

ERICA GARCIA

presupposes the infinite, and especially the variable, context-sensitive exploi­ tation of finite resources. The fundamental reason, then, for assuming that any linguistic unit must make a constant and invariant contribution to communication are (cognitive) considerations of economy: the principle of invariance can be viewed as a particular instantiation of that distinctness (Saussure 1959: 120; Haiman 1983: 814) on which all of language depends. Only second in importance is the methodological motivation: the principle of invariance provides an analytic control for the description of variation, necessarily open-ended. It is invariance that preserves the analyst from lapsing into ad hoc and arbitrary enumeration of particular facts; a point made by Jakobson, with unsurpassable clarity, already in 1936 (Jakobson 1966: 51-53).

2. Critique of "invariance" as analytical principle The principle of invariance is nonetheless far from universally accepted: insightful analyses continue to be proposed where no Gesamtbedeutung is postulated for the forms isolated (cf. Smith 1981, Wierzbicka 1980), though the practice can certainly not be blamed on ignorance or unawareness of Jakobson's work. The failure of the "invariance" approach to find support in general analytical practice thus brings to mind Chesterton's quip to the effect that Christianity has not failed: it just has never been tried yet. But one should then inquire what the reason may be for so many scholars failing to try it. In some cases an explanation can perhaps be found in many linguists' unwillingness to give up certain preconceptions as to the nature of language, which determine their view of linguistic analysis. Take, for instance, some reactions to Bolinger's Meaning and Form (1977). Both Geis (1979: 685) and Newmeyer (1983: 113, 119), express sincere puzzlement that a scholar of Bolinger's sensitivity and intellectual caliber should care so little about contributing to the "formal theory of grammar". The idea has apparently never occurred to them that there may be a causal connection between these two facets of Bolinger's personality. But even among those who eschew the modular tenets of generative grammar there are some scholars who seriously question the general validity of the invariance principle. In these cases the difficulty seems to reside, fundamentally, in the conviction — or fear — that invariance is bought at the expense of variation; more particularly, that by postulating invariants one perforce does violence, or less than full justice, to the complexity of the data. After all, there is no direct evidence for the postulated invariant, whereas the varied interpretations of a form appear to be fairly accessible. Listing the latter would at least keep one from postulat-

VARIATION AS PROOF OF INVARIANCE

35

ing ad hoc, uncontrollable entities. It is no accident, then, that we find many scholars demanding of an invariant meaning not only that it explain the various exploitations of the form — i.e., that it account for all its uses — but also that it predict the non-occurrences as well — i.e. that it account for its uses only. Wierzbicka (1980: 157-8) quotes with approval son's critique of Jakobson and remarks herself: Formulas like " + peripheral, -affected" are not self-explanatory (indeed they are certainly less so than traditional labels like "manner" or "instrument"), and one could stretch them in many different ways to make them fit the facts. Even if we limited ourselves to blind imitation of Jakobson's way of using and interpreting his formulas, we would still often go wrong because these formulas have very limited predictive power, and insofar as they do have predictive power, the predic­ tions they generate are often incorrect. ...A person who does not know Russian cannot learn to use the Russian cases on the basis of Jakobson's formulas. (Wierz­ bicka 1980: xv). She then compares the divergent use of Russian and Polish "instrumental" and remarks: This shows that Jakobson's unitary formula is too vague, too general to be empiri­ cally adequate. If it fits all the uses of the Russian instrumental, it also fits many sentences in which the instrumental cannot be used. Insofar, therefore, as it is verifiable, it does not pass the test. (Wierzbicka 1980: xvi). The obvious answer is, of course, that even if the Polish and the Russian "instrumental" should happen to have the same meaning, the languages as wholes are different, so that, fitting into different systems, these allegedly identical Instrumentals must perforce come to be used differently. But such an answer may still not satisfy the critics: Wierzbicka at least appears to expect from linguistic analysis an explicit account of usage that i. can be directly converted into a teaching method (1980: xvi); ii. establishes a biunique link between meaning and message so that no message conveyed by form X could also (approximately) be conveyed by form Y, perfect distinctness being thereby guaranteed (for an earlier version of this objection, cf. Bazeli 1966: 332); iii. allows all non-uses of a form to follow from its meaning. Such an account would no doubt be fully testable, and falsifiable by means of counterexamples. But in order to achieve this degree of explicitness and predictive power the analysis would necessarily have to be formulated in terms of finite, well-defined configurations. And it is (or should be) clear where that road leads: up a tree. As Itkonen (1983: 314) soberly puts it: "We are confronted...with the general fact that the degree of formalization is inversely proportional to the degree of descriptive adequacy." But, even if the analyst could somehow succeed in accounting for all and

36

ERICA GARCIA

only the uses of the form, another objection awaits the postulated invariant: no longer that it is inadequate and/or ad hoc, but that it is circular. We thus find Birnbaum asking: In other words, is there not an element of circular reasoning inherent in an alleged overall semantic category set up as a generalized abstraction of a variety of actually ascertainable functional meanings.. .which, once it is claimed to have been estab­ lished, then in turn is being invoked as the basis from which the contextual semantic variants are said to be derived? (Birnbaum 1978: 29) Birnbaum confines himself to posing the question — but that such a question could still be asked in 1978 is proof that the relation between variant and invariant remains far from clear.

3. Validating invariants And for excellent reason: the fundamental analogy underlying "invariant"type analysis, i.e. phoneme is to allophone as meaning is to contextual realization (Jakobson 1964: 365) has unfortunately not been accompanied in structural linguistics by a discussion of how the obviously different substances involved in the second vs. the first articulation (Martinet 1965) affect the realization process. To name one small yet critical difference: homonymous morphemes are conceivable: not so homonymous phonemes. Any claim of invariance — and consequently any assignment of variant contextual realizations to a particular invariant meaning — demands, thus, some principled basis for deciding that homonymy is (or not) involved. Indeed, all the truth in the initial section of the "Beitrag", plus all the elegance of the postulated system, plus even the possible correctness of the analysis itself, do not make up for the one flaw in that great paper: the total absence of a principled discussion of how variants do result from the interaction of context and invariant meaning (Miller 1985: 149).2 Jakobson's followers appear to have sensed this shortcoming and have attempted to make up for it. Among the best work in this tradition we may unquestionably count Waugh (1976): however, when we look at her actual discussion of variants (1976: 97, 99, 101; cf. also Sangster 1982: 70 ff) what we find, every time, is a strong claim that such and such is a contextual variant of the particular invariant meaning under discussion — but no description of how and why it is that. The risk of circularity looms large, indeed, if invariants are justified by their ability to account for variants whose characterization follows from those very invariants (Birnbaum 1984: 414). A safeguard from arbitrariness is obviously needed, and the first and basic

VARIATION AS PROOF OF INVARIANCE

37

one — given the distinctive nature of linguistic units — lies, as Jakobson never tired of pointing out, in the oppositional nature of linguistic value.3 This is why, perhaps, a control on the necessarily abstract invariants the analyst must postulate, has been sought in the larger system itself (Jakobson 1966: 62-63 ff; Wierzbicka 1980: xiv). Consider, for instance, Sangster's remark: Of course, the invariant conceptual features uncovered will have to be confirmed by a demonstration of their recurrence in other grammatical or lexical categories than the one in which they were originally determined. If this subsequent step can be successfully performed, then we have extremely strong evidence for the reality of the semantic invariants proposed for a given language (Sangster 1982: 78). However, this progression to "intercategorial invariance" depends, itself (Waugh 1976: 83), on the correctness of certain basic assumptions concerning the nature of language, such as: finite number of distinctive features, binarism of oppositions, and overall systematicity. And for these there is no independent evidence. So we run around in ever larger circles. More is needed, then, if our analyses are not to be reconcilable with any kind of data whatsoever. Some means must be found of making our analysis vulnerable to disconfirming data — in short, of deriving testable predictions from the postulated invariant. 4 And this must be done in such a way that a deterministic algorithm is not involved, since the link between invariant and variation should not be psychologically implausible. In particular, account must be taken of inference, i.e. human beings' ability to leap to specific conclusions (different ones in different contexts). In short, the con­ nection established between invariant and variant must reflect the variability which context necessarily introduces into the inferential process (Itkonen 1983: 193). Now it is precisely because of the failure of certain predictions that doubt has again been cast: not, mind you, on a specific analysis (as would have been reasonable, since what fails must be what is tested, i.e. the analysis) but on the principle of invariance itself (which, after all, makes possible the predictions that constitute the test). Timberlake challenges the invariance of the Russian Perfective and Imperfective on the basis of questionable predic­ tions, which rest on the (unmotivated) assumption that an invariant meaning should be usable (Timberlake 1982: 327) "to predict the syntagmatic effect of aspect on other categories". But though one can — and probably should — disagree with much of what he says, it is to Timberlake's great credit that he has demanded of an invariant analysis that it have teeth that bite into the data in intersubjectively controllable fashion. Alas, "invariant" analyses have tossed so far between the Scylla of circularity and the Charybdis of subjective arbitrariness. In

38

ERICA GARCIA

these dire straits a heroic decision is in order: we might as well grasp the nettle, and seek firm ground in the shifting sands of use.

4. Skewed use as a control We may and must assume that what links invariant meaning to variable exploitation is the communicative appropriateness of the former to an intended message. That is, using a particular meaning allows the speaker to hope that the hearer will be able creatively to infer — from that meaningin-context — the particular message the speaker intends. In short, the meaning used must be congruent with the message intended. It will further be granted that no message — intended or inferred — can be reduced to a determinate function of the several meanings which contrib­ ute to its communication. From this it follows that no part of an utterance — specifically, no single meaning — can stand in a one-to-one relation to any part of the message. Conversely, all meanings involved will contribute to (or at least not detract from) the global message conveyed. In short, utterances will owe their communicative success to their coherence, i.e. to the cooperativeness of the various meanings. The notion that syntagmatic distribution can be explained by paradigmatic value is, of course, not a new one: cf. Jakobson's discussion of so-called "government" of cases by prep­ ositions (Sangster 1982: 76 ff). This insight, however, has consequences for the quantitative aspect of syntagmatic distribution as well. Note that a message will be the more easily inferred the more all clues point in the same direction (cf. Ziff 1984: 33, 34). It follows, then, that most of the time the speaker, if he is coherent, will produce more than one clue to his intention, and that congruent clues will tend to co-occur, since communicative success will select for cohering/ coherent co-occurrences. Finally, if different linguistic units — i.e. different meanings — lend themselves to different communicative uses, they will necessarily show a preference for different contexts. However, in view of the indirectness of the contribution of meaning to message, there is no reason to expect that this preference will manifest itself as a matter of absolute yes/no, i.e. as comp­ lementary distribution. On the contrary, the very openness of the communi­ cative use of language guarantees the dysfunctionality of such hard and fast rules. But what is to be expected — and, as will be shown below, can be observed — is quantitative skewing of usage: a cline of preference for one form (alternative) vs. the other, across a range of contexts. In short, what goes under the name of "syntactic variation" in sociolinguistics — i.e. the relative frequency of different forms in different environments and contexts

VARIATION AS PROOF OF INVARIANCE

39

— is also something to be accounted for by our analysis. We therefore fully agree with Naro when he says: From the point of view of linguistic theory, it would be much more interesting if patterned behavior in linguistic variation could be shown to be a result of linguistic competence, rather than just a part of it, in the same way as the regular behavior of a coin upon flipping is a result of its physical structure (Naro 1980: 165) The kind of explanation that sociolinguistic analysis has failed to provide for variation in usage (Naro 1980, García 1985b) can be achieved precisely by taking seriously the implications of an invariant analysis. If the relation between "invariant" and "variation" — both qualitative and quantitative — is, fundamentally, one of communicative appropriate­ ness, we have, in what people are observed to do most often, a clue to the semantic substance of the invariant itself.5 That is, quantitative data can tell us something about the qualitative side of our analysis, and in this fashion provide a control on the postulated meanings.

5. A test case a. The analysis At this point a picture is worth a thousand words. We shall present a specific analysis of specific forms, and make specific predictions, on the basis of the postulated (invariant) meanings, as to quantitative skewings in the use of forms. Our discussion will involve Spanish forms occurring in prepositional phrases, and often characterized as "third-person pronouns". The following decontextualized examples present the basic referential facts in pre-analytical fashion: (1) Compró un libro para sí "He bought a book for himself" (2) Compró un libro para él "He bought a book for himself/him" (3) Le habló de sí "He i spoke to him j about himselfi/j" (4) Le habló de él "Hej spoke to hinij about himselfi/j\himk" (5) De casualidad, Juan halló un libro valioso de por sí "By chance, John found a book valuable in itself (6) De casualidad, Juan halló un libro valioso para él "By chance, John found a book useful to him"

40

ERICA GARCIA

At issue in all these cases is the identity of the pronoun's referent, or rather, which of the various (third) persons present in the context is referred to with one or the other form. It is true that this kind of problem does not resemble the familiar "invariant/variant" situation, in the way in which the reading of si is "contextually dependent." But it is certainly the case that: i. invariant meanings can be postulated for si and él; ii. si and él are variably frequent in reference to different "antecedents"; iii. the grouping of the prepositional phrase — a contextual factor — strongly influences what pronoun is used to refer to a given entity. Finally, in referentially equivalent situations, a qualitative difference is observed depending on what form is used to refer to the "same referent" — just as the so-called "predicative function" in Russian takes on a different flavor depending on whether an Instrumental or a Nominative is used. Now, our analysis aims at accounting for these varied facts in an invariant way, i.e. on the basis of single specific meanings for si and él. The analysis follows: SÍ = THIRD PERSON él = DEICTIC + GENDER + NUMBER6

Note that no opposition is assumed to hold between si and él (as it is by Alarcos Llorach 1970: 152; Academia 1974: 203, 208; Fernández 1951: 221; Schmidely 1983: 155). The forms differ, however, in that él, unlike si, is marked for gender and number, while si, unlike él, parallels the first and second person singular in a number of morphological oppositions. What the formal paradigm suggests, then, is that si is as personal as are the first and second person mi and ti, despite the necessarily negative character of a third-person pronoun (Benveniste 1966: 256).

b. The application Invariant meanings do not specify their application in actual communi­ cation, any more than a hammer's structure specifies the driving in of nails, or an oar specifies rowing. The connection between tool and specific (com­ municative) task is in all cases made by the creative imagination of the tool user. Just as a hearer must work out a specific interpretation for a form in context, so also must the speaker work out the applicability of the meanings available in view of the specific message he wishes to convey (the "pro­ duction" context). Both speaker and hearer, in short, must "interpret" the

VARIATION AS PROOF OF INVARIANCE

41

meaning, i.e. work out the particular, local, appropriateness of the (invariant) tool provided in the language. Now, if a given communicative problem keeps recurring, and is solved in the same way by different speakers, a strategy may develop, i.e. a standard, generalized, conventional (but not arbitrary!) way of exploiting a meaning. It is undoubtedly the case that what are called "grammatical" meanings are characterized precisely by highly regular strategies of use (Garcia 1975: 376). This should not lead us, however, to ignore the invariant meanings that motivate the strategies. No matter how regular and firmly entrenched in usage these latter may be, they are, after all, creative solutions to ad hoc problems, i.e. specific exploitations of a tool for a specific communicative goal. The communicative goal or problem may, of course, vary in generality: it may be posed by the culture (cf. García & Otheguy 1983) or it may be given by human nature itself (cf. the relevant discussion of the tertium comparationis in Seiler, this volume). In the specific case under discussion, the forms si and él, because of their meanings, turn out to be useful when it comes to making the hearer think of a referent. That si should be suited to this task follows from the meaning of person: just as mi and tí single out the first and second person (singular) respectively for reference as non-centrally involved in an event7, si ( = THIRD PERSON), by failing to single out the first and second persons, invites the inference that somebody/something/anything else is involved. That él, ella, etc. should be suited to the task of "making the hearer think of a referent" is a natural consequence of the coupling of Gender and Number information (i.e. attri­ butes of a referent) with the steering of the hearer's attention to an identifi­ able referent, which is what Deixis implies. That both tools are suited to the same general task does not imply, however, that they will perform equally well for all referents, under all circumstances. After all, a hammer and a screwdriver are both appropriate when it comes to driving thin, elongated, pointed metal objects into wood: but the difference in their shapes makes us use the one for nails, the other for screws. How is the labor divided between si and éP. The meaning of él, ella, etc. i.e. DEIXIS + Gender + Number specification, results in these forms being usable only when the referent is of the specified gender and number. But this means, in turn, that the gender and number of the referent must be known and/or relevant. "Generic" reference to an unspecified referent will not be a task for which él is particularly appropriate. And indeed, si "alter­ nates" (i.e. competes communicatively) not only with él, ella, etc., but also with uno ( una) "one" and with Usted "You (formal address, grammatically third person)". 8 Now, suppose that gender and number of the referent are known, and

42

ERICA GARCIA

that a bona fide9 third person is involved: there may still be local, contextual conditions making either si or él appropriate. Si's (person) meaning is limited to a failure to specify the first and second persons: hereby si becomes totally dependent on the context for the identifi­ cation of its referent, if any. Now the only thing left in the hic-et-nunc of the actual act of communication, once speaker and addressee are ruled out, is the act of communication, i.e. the discourse itself. And it is here that si finds its referent. This also gives us the difference between si and él: as a deictic, él is positively defined, and points out its referent, via gender and number information, within a set relevant to and given by the speech situation, él therefore presupposes the possibility and the need to differentiate among referents. Among the personal pronouns, si is negatively defined, and points out by default, presupposing the immediate givenness of its referent within the speech situation itself.10 Given the different value of si and él it is hardly surprising that in decontextualized utterances (examples (1) through (6) above), si must find its referent within the boundaries of the clause within which it appears, while él can transcend those boundaries. In actual discourse contexts, however, si can refer beyond the sentence. Elsewhere (Garcia 1983: 192-202) we have discussed some instances of that use. But the difference between the two forms can best be appreciated (and does the most damage to the traditional "sentence gram­ mar" analysis in terms of reflexivity), when we examine si and él as alternants for si's majority use, i.e. in reference to the subject. Consider the following pair: Mallea, Resentimiento 113 Pero nada veía el hombre del cetro. Más que su delirio. Se le oía de cerca, por más que se dictara, según su ley, para sí solo, el susurro "...." "But the man with the sceptre saw nothing. Besides his delusion. One could hear him from close by although, according to his law, he dictated his murmur for himself [si] alone: '....'" Cortázar, Vuelta II: 174 Esta ùltima (foto) el cronopio se la guarda para él y está contentísimo con esa foto. "This last photograph the cronopio keeps for himself [él] and he is as pleased as Punch with it." The context makes clear that the man referred to via si in the passage from Mallea is i. speaking to himself; ii. alone; iii. self-contained: even if others were present, he would ignore them. Only one third person human is referred to in the context and, what is more,

VARIATION AS PROOF OF INVARIANCE

43

that person is the only one relevant to the event of "dictating". Hence the use of si, which at once depends on the immediate givenness of the referent, and stresses its uniqueness. Not so in the example from Cortázar. Though only one third person is physically present (the cronopio who has taken five passport photos of himself in a booth, and is keeping a sixth for himself), the larger context — to which the taking of the passport photographs is relevant — involves more than just the cronopio. He is in competition with, on a par with, and thus forms a set with, the consulate for which he is filling out a travel application, and which will get the remaining five pictures, but for the sixth one, the one he keeps for himself, he sticks his tongue out. The contrast between cronopio and consulate is underlined, thus, by the difference between the photographs allotted to each party. The cronopio, in short, is perceived in contrast to other relevant persons, and this makes it appropriate to point him out by means of the deictic él. Consider also the following pair: Sabato, Abaddón 510 Se miraron entre sí en muda consulta y Nicolás,fijandocon severidad sus ojos sobre él, (...) exigió que nombrara la característica más típica de don Sierra. "They looked at each other in mute consultation and Nicholas,fixinghim severe­ ly, ... demanded that he should name don Sierra's most characteristic trait." Cortázar, Fuegos 61 Reunidos en el comedor, miraban al lugar vacío de tía Clelia, se miraban entre ellos, vacilando. "Together in the dining room, they looked at Aunt Clelia's empty place, they looked at each other, hesitating." These examples differ i. as to the number of things looked at; ii. in the tense of the verb. In the first example there is only one entity looked at: the "they" that exchanged the look of intelligence, while in the second example "they" first look at Aunt Clelia's empty place, then, at a loss, at each other. The net result is that "they" are the only ones that come into consideration with respect to looking in the first case, while in the second they compete with Aunt Clelia's place. This difference is underscored by the tense of the verb: the preterite {miraron) in the first example confirms that there was just one mutual look of telepathic understanding among the brothers, as a result of which Nicholas posed his test question, while the imperfect (miraban) of the second example suggests several (different) interchanges of glances, which is highly plausible given the situation: Aunt Clelia's brother, and her nieces and nephew are trying to find a way of concealing the fact of her death

44

ERICA GARCIA

from "Mamá", Aunt Clelia's invalid sister. Among whom the glances were interchanged is thus self-evident in the first example: all those consulting one another, namely, all the brothers, who in fact think "as one man". But in the second example several different combinations are possible and likely: hence si resp. él are appropriate to, and at the same time help to suggest, the two different situations. In short: the natural division of labor following from si's meaning, vis à vis el's, is that si will be used for a unique(ly) context-given referent; él for one that needs/deserves to be singled out in contrast to others. The strategies just described and exemplified are not, however, an alternative to an invariant-meaning analysis, for the simple reason that they do not — and cannot — exhaust the meaning of the forms, which remains available for novel applications (cf. Hofstadter 1980: 669 for insightful discussion of the cognitive value of invariants). 11

6. Quantitative testing In the preceding section we have argued that i. sí and él have different invariant meanings; ii. these meanings will most naturally lend themselves to different com­ municative ends, with the consequence that the forms will be used under different contextual circumstances. We have also presented qualitative data — i.e. minimal pairs — in support of our contention. In appealing to this type of explication, however, we run the risk of justifying ad hoc every instance of use of a form via arbitrary appeal to uncontrolled traits of the context. This danger can be avoided by having recourse to experimental data, i.e. confronting native speakers with con­ trolled pairs of situations differing in traits that match the postulated differ­ ence between si and él. We now present native subjects' response data for a number of contextual factors; we explain in each case how these should (or not) favour the use of si, given i. the postulated meanings of si and él, and the strategic deployment that follows therefrom; ii. the "forced choice" to which the experimental subjects were reduced: they had to use a third-person pronominal form. 12 In order to dispel effectively any lingering suspicion that si might, after all, merely be a 'reflexive' pronoun we begin by showing that both si and él can refer to a non-subject — in particular, to an Accusative or Dative object of the verb, referred to by means of the clitic pronouns lo or le.

45

VARIATION AS PROOF OF INVARIANCE

Consider the frames under (7): (7a)

Alabándole sus poemas, lo entusiasmé y lo arranqué de "By praising his poems I filled him with enthusiasm and carried him [Acc. lo] beyond (himself)" (7b) Perico es muy vanidoso, y lo ùnico que me tolera es que le hable de "Pete is very vain, and only allows me to talk to him [Dat. le] about (himself)" 13

The two frames differ as to the case of the "antecedent" pronoun: Accusative vs. Dative, plus, of course, the lexicon appropriate to that case: arrancar lo "tear him out of something, carry him beyond", vs. hablar le "talk to him of".14 The results appear in Table 1. We observe that the informants have produced a much higher percentage of si when referring to an Accusative than to a Dative Object.15 If our hypothesis as to si vs. él is correct, and the choice of pronoun is indeed due to the difference between Dative and Accus­ ative, it should follow that the Accusative is more context-given than a Dative.16 And indeed, Jakobson's characterization of Accusative and Dative as Central vs. Peripheral immediately comes to mind: the Accusative is more centrally involved in the event to which the prepositional phrase is relevant, and hence scores a higher percentage of si. Table 1.

Use ofsi vs. él depending on case (oblique) of antecedent.

Frame

Condition

sí

7a 7b

arrancar Accusative de hablar Dative de

41 5 15 32 X2--= 30.356

él

Total

% sí

46 47 p 2pl/sg > 3sg > 3pl This formula gives the hierarchical relations holding, in Yaka, between associated persons which are alike in subjecthood, and giving way to (1) pluralization (taking the grammatical person of the element left from the > sign), and (2) relegation of the hierarchically inferior member (at the right of the > sign). 4.3.2 For integration as manifested in structures, I will take the example of relative clauses, with special reference to French. Relative clauses — what­ ever their formal status 49 — should preferably be defined as propositions integrated within a sentence.50 This integration is evident at various levels: 1. that of sentence prosodics; 2. that of word order (e.g. SVO — S ( +relative clause) VO; S ( +relative clause) V 0 ( + relative clause, etc.); 3. that of syntactic-semantic features (agreement with the features of the antecedent, e.g. la maison où il se trouvait LOC

LOC

"the house where he was" l'employé qui nous a conseillé de déposer une plainte ANIMATE

ANIMATE

"the employee who advised us to file a complaint"; 4. that of word forms: je lui ai demandé cela "I asked him that" *je lui ai demandé cela qui se passait là-bas *"I asked him that which was happening over there" je lui ai demandé ce qui se passait là-bas "I asked him what was happening over there" The integration of the relative clause within the main clause calls for two

GRAMMATICAL PARAMETERS

207

types of analysis, the first of which concerns the nature of the integration, whereas the second concerns the formal marking of the integration. As to the nature (or "depth") of the integration, a basic distinction should be made between uniplanar extraction as in: je vous apporterai le livre dont je vous parlais "I'll bring for you the book I was talking about" and biplanar extraction, as in l'homme que je dis qui est venu "the man who I said who came" [je dis que] [l'homme est venu] [je dis que] l'homme [qui est venu] l'homme [que je dis] [qui est venu] As to the formal marking of the integration of French relative clauses, a major distinction must be made between: 1. relative clauses introduced by a variable relator 2. relative clauses introduced by an invariable relator (the so-called relatives phrasoïdes): — —

il était une fois un homme que rien ne lui plaisait "there once was a man that he liked nothing" des feux qu'il faut appeler les pompiers pour les éteindre "fires that it's necessary to call the fire station to put them out"

The variable relator varies according to 1. the opposition animate/inanimate (contrast à qui — à quoi) 2. its function in the subordinate clause: — — — — — —

subject: qui predicate: que direct object: que indirect object: à qui/à quoi locative/time-indication: où prepositional object: de qui/de quoi — dont other prepositions (à, pour, par...) + qui/quoi

3. the opposition individualized/non individualized: contrast lequel/laquelle ~ qui In modem spoken French, the invariable relator que (the "relative particle") has more and more replaced the variable relators (or relative "pronouns"). As major determining factors in this process, I would see

208

PIERRE SWIGGERS

1. the fact that the invariable relator is psycholinguistically less complex; 2. the fact that the "relative phrasoïde", which allows the speaker to separate the two basic functions of the relative pronoun, viz. linkage and syntactic marking, provides for greater precision of content within the subordinate clause. Compare (A)

La Russie dont le bolchevisme entrave la civilisation "Russia whose bolshevism impedes civilization" ( = "relatives phrasoïdes") I. La Russie que son bolchevisme entrave la civilisation IL La Russie que le bolchevisme entrave sa civilisation III. La Russie que son bolchevisme entrave sa civilisation

3. the parallelism with other subordinate clauses (causal, final, ... clauses), where the syntactic integration is also marked by an invariable element. 5. Before offering some concluding remarks, it may be worthwhile to point out that the parameters of hierarchy and integration should be viewed as open, dynamic parameters. A long-term view of the history of languages shows how hierarchies can be reversed (see e.g. the evolution of the phonemic inventories of the various Indo-European languages, or the emergence of prepositions governing specific cases) or leveled out (neutralization and specialization of variants). The same is true of integration, and as an example I would cite the case of Rhaeto-Romance stuvair/stuv er/stuer/stueir.51 This verb testifies to the possibility of reversal of level-integration: as a matter of fact, the Latin syntagm est opus has been absorbed 52 within morphology, and has been "morphologized" here as a verb. The result of this process is different within the three areas of Central Romance. Old French has retained only impersonal forms (estuet, estovoit, estut, estovra, estovroit, estuisse, esteust; reconstructed infinitive *estooir < *estopere). In Old Italian, we also find exclusively impersonal verb forms: estuf (Old Piedmontese), stor (Old Genovese), stovesse (Old Pavanese), astove (Old Milanese) and stoverá/ stourá (Old Venetian). The case is different in Rhaeto-Romance, where we have a full conjugation of the verb stuvair "must, to be obliged". In Romansh, 53 we find the following conjugation in the present tense: Sursilvan: stoi/stos/sto/stuein/stueis/ston "Oberengadinisch": stogi/stoust/stu/stu(v)ains/stu(v)ais/staun "Unterengadinisch": stogi/stoust/sto/stuvain/stuvais/ston. Examples: tü stoust gnir "you must come" ti stos gie buca far quei "you don't need to do that" el sto star a casa "he must stay home". Whatever may have been the cultural factors determining this process, 54 it is interesting to note that

GRAMMATICAL PARAMETERS

209

1. the process involves an important category-shift (possibly through the link established with a second person verbal form: est opes > estopes). Such a category shift was favored by (1) the vague semantic load of opus, and (2) a number of phonetic processes affecting final vowels in Gallo-Romance, Rhaeto-Romance and North-Italian dialects. 2. the "incorporating" shift gives way to a paradigm: estopeo, estopes, estopei, etc. 3. the process reflects an analogy with other changes: as an example one can cite the case of the impersonal verb convenit which in Dolomitic and Friulian has become a fully regular verb (cógno, cognes, cogna, cognòn, cogna', škoñ, skoñi, škoñ, skuñúη, skuñedi, škon).55 4. the process involves analogically built formations (e.g. ston: based on pon < potent; staun; based on staun < stant; through the entire paradigm there are traces of analogy with the verb dovair "to have to"). 56 At the end of this process, a syntactic idiom has become a variable morphological unit, and a former lexical unit {opus) has become part of a new verb stem. Syntax has been turned on its head, and personal deixis has become inflectional (stögi/ stoust...) instead of semi-clitic and exocentric {mihi/tibi...est opus). We have here a spectacular case of the reversal of levelintegration, and as I have pointed out, such a reversal brings about a productive pattern.

6. In this paper I have tried to explore a pervasive topic in Roman Jakobson's linguistic thinking: the dialectics of invariance and variation. As will be evident from the preceding discussion, the dialectics is a fundamental one for our grasping of language and for the construction of a comprehensive theory of grammar. The interplay between variation and invariance has an incidence on all levels of language and language processes, and there is no single fact of language that can be properly understood without reference to variational and invariant patterns. I have tried to illustrate this by analyzing two fundamental categories of language, viz. hierarchy and inte­ gration. I have focused on the linguistic embedding of these principles, viewed here as parameters of linguistic description. It goes without saying that much more could (and should) have been taken into account: I will limit myself here to indicating the philosophical (duality; forms of "empirical rationalism"; process thinking), cultural (patterning of values; language as a cultural good), and cognitive-psychological (lateralization of speech; binarism in reasoning) perspectives.57 I would like to conclude by pointing out that variation and invariance are the specific implementation of an over­ riding property of language as a semiotic system: its relative efficiency. As noted by Linda Waugh,

210

PIERRE SWIGGERS

The efficiency of language is the efficiency of the overall system, thus implying the relative efficiency of the component parts in terms of each other. However, there is an interrelation between the parts of the system and the lack of any absolute balance between different functions, as in any social system. Efficiency need not imply a razor-edged simplicity. In fact, efficiency in one area (i.e., a small number of features and their combinations into phonemes and sequences) may lead to non-efficiency in another area (i.e., a large number of homonyms or long words). Efficiency in morphology or in the number of words may lead to longer or more intricate syntactic constructions. There is always a trade-off. It is, in essence, a question of hierarchy. The hierarchy of the units from the largest to the smallest is diverse. Furthermore, the economy of one level is accompanied by lesser economy of other levels. (Waugh 1976: 19) This relative efficiency springs from the nature of language itself: as a medium (in the etymological sense) of form and meaning, of referentiality (including se1f-referentiality) and expressivity, of directness and indirectness, of tradition and innovation (recreativity rather t h a n creativity), of culture and personality.

Notes 1. This is the notion proposed in Swiggers (1981a, b). For a discussion of the methodological implications, see Swiggers (1983). 2. See especially the framework adopted by Lepschy (1970, 1986). On the social aspects of "scientific communities" in twentieth-century linguistics, see Murray (1983). 3. On this surface diversity, see Jakobson's remarks (1973:12): "At first glance, linguistic theory of our time seems to offer a stunning variety and disparity of clashing doctrines. Like any age of innovative experimentation, the present stage of reflections on language has been marked by intensive contentions and tumultuous controversies. Yet a careful, unprejudiced examination of all these sectarian creeds and vehement polemics reveals an essentially monolithic whole behind the striking divergences in terms, slogans, and techni­ cal contrivances. To use the distinction between deep and superficial structures that is current today in linguistic phraseology, one may state that most of these allegedly irreconcilable contradictions appear to be confined to the surface of our science, whereas in its deep foundations the linguistics of the last decades exhibits an amazing uniformity. This communality of basic tendencies is particularly impressive in comparison with the substantially heterogeneous tenets that characterized some earlier epochs of this discipline, in particular, the nineteenth and the early years of the twentieth century. Factually, most of the recent discord is based partly on dissimilarities in terminology and style of presen­ tation and partly upon a different distribution of linguistic problems chosen and pointed out by single scholars or teams of inquirers as the most urgent and important". 4. See especially Bloomfield (1926, 1927, 1933, 1939, 1945, 1958, 1962), Harris (1951, 1962, 1981a, b — in the latter collection, above all the articles "Distributional Structure", "Morpheme Alternants in Linguistic Analysis", "Discontinuous Morphemes", "From Morpheme to Utterance", and "Structural Restatements"), and Chomsky (1957, 1975, 1977, 1981, 1982).

GRAMMATICAL PARAMETERS 5. 6.

7.

8. 9. 10.

11.

12.

211

For very accessible selections, see Whitney (1971), Saussure (1968-1974), Sapir (1921, 1949), Jespersen (1922, 1933a, 1933b), Jakobson (19712, 1972a, 1981, 1984). As is well known, Jakobson's work has been instrumental in showing the fallacious identification of synchrony versus diachrony with statics versus dynamics, and in viewing changes as intrinsic components of a dynamic synchrony. See the succinct statement in Jakobson (1972:23): "The diachronic linguistics of today examines the succession of dynamic synchronies, confronts them, and in this way, delineates the evolution of a language in a wider historical perspective, with due attention not only to the mutability of the linguistic system but also to its immutable, static elements. The concentration upon the system and the application to diachrony of the same analytic principles as those employed in synchrony has enabled the diachronic research of our time to achieve impressive results in the field of internal reconstruction; and, on the other hand, when focusing upon the historical stratification of linguistic systems, explorers observe new, significant affinities between this stratification and the synchronic patterning of languages. Present-day linguistics could hardly adhere to the reminder which was quite opportune half a century ago, when it was necessary to emphasize and to set the tasks of descriptive linguistics: 'The opposition of diachronic and synchronic shows up in every point'." The opposition between two types of linguistic reflection, which is sketched here, should be interpreted in a dynamic way, in order to provide room for "switching interests": a typical case in point is the work of Charles F. Hockett, partly technique-oriented (see Hockett 1942, 1947, 1948, 1950, 1952), and partly devoted to the characterization of language(s). For the latter orientation, see Hockett (1959, 1961, 1966, 1977). See especially the appraisals of Jakobson's work by Holenstein (1975), Waugh (1976), and more succinctly by Mounin (1972: 137-153) and Swiggers (1979). See also the essays in Armstrong & van Schooneveld (eds. 1977) and A Tribute to Roman Jakobson 18961982 (1983). See Jakobson (1966), and also (1941, 1960, 1963, 1969, 1970). On purposive behavior as manifested in language structure, see Silverstein (1976, 1985a, b). On functionalism, see Silverstein (1980), who recognizes three theoretical uses of the term "function" in recent linguistic research: "The first may be characterized as abstract, 'sentence'-internal distribution (what I will call referential function2), which has been the most influential view in American linguistic and psychological circles. The second may be characterized as use of signal forms for purposive, intentional social effect (what I will call pragmatic function1), which is motivated by most of the work on the anthropology and sociology of language, particularly in the ethnography of speaking approaches, and by much of the newer philosophy of speech acts. The third may be characterized as the distribution of forms as indexicals, both mutually in defining discourse cohesion, and with respect to the nonlinguistic context of language use (what I will call pragmatic function2), a view now once more beginning to come into its own, after a lapse of interest". In Swiggers (1986a) I have proposed a further reduction to two fundamental contents: (1) function as correspondence or as principle of constitution/compositionality (= functionc), and (2) function as (way of) use, or as effective procedure ( = functionc). On iconicity, see Haiman (1980), Mayerthaler (1980), Pesot (1980) and Ross (1980). Iconicity, defined as a strict correlation between referential categories and linguistic categories, is most apparent in sex-gender correspondences, and time-tense relationships, though non-marking or deviant (contradictory) marking are frequently attested. For the use of iconicity in universal grammar, see Swiggers (1989b). On the arbitrariness of the linguistic sign, see Jakobson (1959, 1963, 1966, 1972a: 711-722), Jakobson & Waugh (1979), Gamkrelidze (1974), Swiggers (1984a). On Saussure's defi­ nition of "l'arbitraire du signe", and on the history of the topic, see Engler (1962, 1964), Koerner (1972), Coseriu (1967), and Stéfanini (1975).

212 13. 14. 15. 16. 17. 18. 19. 20.

21.

22.

23.

24. 25.

26.

27.

PIERRE SWIGGERS See the goals defined in Greenberg, Ferguson & Moravcsik (eds. 1978), Comrie (1981), and Shopen (ed. 1985). See Seiler (1973, 1982, 1983), Seiler & Lehmann (eds. 1982) and Seiler & Stachowiak (eds. 1982). On Item-and-Arrangement and Item-and-Process techniques, see Hockett (1954), Mat­ thews (1974: 226-7, 235), and Swiggers & Van den Eynde (1988a). I have cited the text as it is printed in Benveniste (1966:19). On the part-whole relationship, see Waugh (1976: 56-63). For "normalization" as inherent in language evolution, see Coseriu (1952, 1958). For a similar idea in Antoine Meillet's work, see Swiggers (1986b, c). For a fuller description of French verb morphology, see Swiggers & Van den Eynde (1984, 1988b). The forms cited here are orthographic forms. Granted even that several of them are divergent notations of a single phonological sequence, the high amount of variation is still impressive. I am grateful to Gilles Roques (C.N.R.S., Nancy) for the fruitful exchange of ideas on this problem of historical morphology. Bréal (19084: 11): "Une tendance de l'esprit qui s'explique par le besoin de clarté, c'est de substituer des exposants invariables, indépendants, aux exposants variables, assujettis" (see e.g. the case of analytical comparatives and superlatives). Bréal (19084: 26): "Nous appelons "répartition" l'ordre intentionnel par suite duquel des mots qui devraient être synonymes, et qui l'étaient en effet, ont pris cependant des sens différents et ne peuvent plus s'employer l'un pour l'autre" (see the example on p. 28: "Le Savoyard emploie les noms de père et mère pour ses parents, au lieu qu'il garde pour le bétail les anciens mots de pâré et de mâré"). Bréal (19084: 43-44): "L'allemand moderne a une espèce de verbes qu'on peut appeler "dépréciatifs", car ils expriment l'action en y joignant une idée de mésestime et d'ironie. Ils sont terminés en -eln. Ainsi de klug, "intelligent", on forme klügeln, "faire l'entendu, subtiliser"; de Witz, "esprit", on forme witzeln, "faire le bel esprit, dire des balivernes"; de fromm, "religieux", on forme frömmeln, "faire le cagot". Quelquefois le verbe en -eln est tiré directement d'un autre verbe: deuten, "interpréter"; deuteln, "subtiliser sur un texte". L'idée dépréciative est entrée après coup dans cette désinence, qui n'avait à l'origine aucune signification fâcheuse. La formation en -eln vient d'anciens substantifs en -el ... comme parmi ces substantifs il y en avait quelques-uns à sens diminutif, tels que Würfel, "dé"; Schnitzel, "copeau, rognure"; 'Àugel, "ocellus", cette circonstance a suffi pour colorer la désinence verbale d'une nuance particulière. Dire que ce sont des produits de l'analogie est une explication insuffisante: l'esprit populaire a multiplié ces verbes parce que l'irradiation y avait fait entrer une signification spéciale". See Bréal (19084: 91-97). "Comme les pierres d'un édifice qui, pour avoir été jointes longtemps et exactement, finissent par ne plus composer qu'une seule masse, certains mots que le sens rapproche s'adossent et s'appliquent l'un à l'autre. Nous nous habituons à les voir ainsi accolés, et en vertu d'une illusion dont l'étude du langage offre d'autres exemples, nous supposons quelque force cachée qui les maintient ensemble et les subordonne. Ainsi s'établit dans les esprits l'idee d'une "force transitive" résidant en certaines espèces de mots" (Bréal 19084: 194). For a diachronic application see Swiggers (1985). Bréal (19084: 205): "J'ai autrefois proposé d'appeler du nom de contagion un phénomène qui se présente assez souvent, et qui a pour effet de communiquer à un mot le sens de son entourage. Il est bien clair que cette contagion n'est pas autre chose qu'une forme particulière de l'association des idées". On the distinction between allomorphs and morpheme alternants, see Swiggers & Van den Eynde (1988c).

GRAMMATICAL PARAMETERS 28.

29.

30.

31.

32. 33. 34. 35. 36.

37.

213

This typology owes much to a number of methodological treatments of morphological problems, such as Hockett (1947), Voegelin (1947), Nida (1948, 19492), and Matthews (1974). See the important observation by Silverstein (1976: 27): "Every linguistic sign is an icon of the linguistic sign type, and in this sense every linguistic sign trivially incorporates an iconic mode". Seiler (1983) subsumes various types of relationships under the heading "Possession": social relationships, implements of material culture, part-whole relationships, agent-action and object-action relations, and spatial orientation. While this inclusion can be questioned on a priori grounds, Seller's data-based approach makes it clear that there is a set of techniques (which stand in various types of distribution to each other) for the expression of this complex domain of relationships. Moreover, the use of the techniques is subject to two interacting and opposed forces: one which sets off the EGO from "its possessions", and another which integrates "items" within the personal sphere. This dynamic model — which has interesting links with Thorn's "catastrophe theory", and with Piaget's developmental psychology {accommodation vs. assimilation) — can thus throw interesting light on the diachrony of languages (e.g. development of a possessive verb; development of specific classes of possessive pronouns; the evolution of the link between possession and location; the primitive semantics of verbs such as German besitzen, French posséder, etc.). This term is a direct translation of the French term dispositifs used in Blanche-Benveniste (et al. 1984). I am grateful to Karel Van den Eynde (University of Leuven) for several discussions on problems of descriptive method. For a methodological reflection on the descriptive procedures outlined here, see Swiggers & Van den Eynde (1988a). Note that no exclusive use is made here of semantic features. A description of this type would complement the extant dictionaries of verbal construc­ tions (see e.g. Gross 1975; Busse & Dubost 1977). On hierarchy and integration in Jakobson's science of language, see Waugh (1976: 13-17, 21-22, 51-61, 85-89, 96-101). See Waugh (1976: 48-9, 98). For the metaphor, see Meigret (1550: 19): "Le langage, l'oraison, le parler, ou propos est un bâtiment de vocables, ou paroles ordonnées de sorte qu'elles rendent un sens conve­ nable et parfait". "One of Jakobson's characterizations of the universal principles of language is done in terms of a necessary, uninvertable hierarchy, which as a rule (in most languages) is composed of the following parts and wholes: DISTINCTIVE FEATURES, PHONEME, SYLLABLE, MORPHEME, WORD, PHRASE, CLAUSE, SENTENCE, UTTERANCE, DISCOURSE are a l l f o u n d i n a l l

38. 39.

40.

41.

42.

languages" (Waugh 1976: 60). And going back to some of Jakobson's (1958) observations. See Gamkrelidze (1976, 1979, 1981), Gamkrelidze & Ivanov (1973, 1984); similar con­ clusions were arrived at, independently, by Hopper (1973, 1977a, b, 1982). See also Haudricourt (1975). See also the essays in Shevoroshkin & Markey (eds. 1986), and Vennemann (ed. 1989). For an appraisal, see Mayrhofer (1983: 146), Rosén (1984: 77-78), and Szemerényi (1985: 11-12); for a general assessment, see Haider (1985). See Gamkrelidze (1979: 286-287); for the typological argumentation, see Gamkrelidze (1981), with reference to Greenberg (1970) and Melikišvili (1976). On the status of/b/ in Indo-European, see Mayrhofer (1983: 146), Szemerényi (1985: 11-12) and Swiggers (1989a: n. 64). Let me briefly note that my counterproposal (see Swiggers 1989a, in which the following

214

43.

44. 45. 46. 47.

48.

49. 50.

51. 52.

53.

54.

PIERRE SWIGGERS Proto-Indo-European consonant system is posited: I. b, d, g: - + Lenis, + Voice; II. bh, dh, gh: + Fortition, + Voice; III. p, t, k: + Fortis, -Voice) also appeals to hierarchy, and provides room for integration, viz. with respect to data deriving from language contacts (see Swiggers 1989a: epilogue, and the reference to Watkins 1955). See Jakobson (1972b: 31): "To me the principle of the two different levels: morpheme and word is a universal, although there are languages, like Chinese, with monosyllabic words. These, however, have elaborated the concept of the difference between empty words and full words which corresponds to the difference between grammatical mor­ phemes and lexical morphemes". On the word as a unit of language, and its relation with the pars orationis, see the contributions in Swiggers & Van Hoecke (eds. 1986). See Swiggers (1986d). As a consequence, the French form au (/o/) should be regarded, in spite of the parallelism with à la, à un(e)..., as a single word. This type of structure can occur on both the level of the verbal predicate and that of pronominal forms (cf. Schwartz 1988). The verb form thwééndá is in the selective mood, and there is a tonal bridge with the following yááku. The form yilumbu is in the determined case. For a description of case and mood in Yaka, see Van den Eynde (1968: 13-26), and Van den Eynde & Kutumisa Kyota (1984). In the following I will not deal with tonal problems raised by the syntactic constructions under discussion. For the examples to follow, I will restrict myself to the "concomitance-interpretation", taking the plural verb forms as referring to a combination of two singular persons (thus thwéénda'. "I with you"; the possible translation "we with you" will not be considered here). For a more elaborate analysis, see Swiggers & Kutumisa Kyota (1988). For a succinct typological statement, see Swiggers (1984b: 71-76); see also the impressive synthesis by Lehmann (1984). Compare Damourette and Pichon's (1911-1940: vol. IV, 174 & 1296) use of the term "subordonnées intégrames adjectiveuses ptérosynaptiques" as a designation of relative clauses. See the comprehensive study by Jud (1946-47). According to Jud (1946-47: 36, 47), the process goes back to Late Latin: "[*ESTOPERE] muss wohl wegen seiner territorialen Verbreitung über Graubünden, dessen Verbalflexion einen so typisch archaischen Charakter trägt und die meisten frühmittelalterlichen Neuerungen im Konjugationsbereich des gallofranzösischen und galloitalienischen Raumes nicht übernommen hat, ferner über Nordfrankreich-Limousin, das mit dem oberit.-bündnerromanischen Raum seit dem 7. Jh. keinen geographischen Kontakt mehr besitzt, ins Spätlatein hinaufreichen"; " 1 . Die stammbetonten bündnerromanischen For­ men von "stuvair" beruhen auf lat.  von *ESTÕPET; 2. Sie sind durchkonjugiert und stehen so in scharfem Gegensatz zu den oberitalienischen-nordfr. unpersönlichen Formen (afr. estuet); 3. Die Eigenwilligkeit des ganzen bündnerromanischen Formensystems von "stuvair", wie ich sie hier dargelegt habe, bestätigt mit aller Deutlichkeit, dass die Schaffung des Verbums [*stopere], wie übrigens A. Tobler selbst leicht angedeutet hatte, ins Spätlatein hinaufreichen muss". For Old French estuet, see Tobler & Lommatzsch (1954: col. 1429-1431). Note also the conjugation of the verb in Dolomitic dialects (Bregaglia, Poschiavo): stç (1-3 person sg.), um stç,  stué, šton. See Jaberg & Jud (1928-1940, maps 351 and 1391 "bisogna"); Gartner (1883: 163; 1910: 263). Jud (1946-47) points to the role of Christian moral theory: "Man würde sich nicht wundern, wenn das unpersönliche opus est, oportet, debet, usw. in der kirchlichen Vulgär­ sprache zunächst eine beträchtliche Verbreiterung seines Gebrauchs erfahren hätte. Ander-

GRAMMATICAL PARAMETERS

215

seits dürfen wir nicht vergessen, dass die christliche Lehre mit Nachdruck die Auffassung vertritt, jeder Einzelne bestimme im diesseitigen Leben durch sein sittliches Verhalten selber sein Schicksal nach dem Tode. Das persönliche "ich muss das tun" wäre dann das Korrelat der inneren freudigen Zustimmung des einzelnen Gläubigen zum Sittengesetz seiner Kirche, gilt doch die Nichtbeobachtung grundlegender moralischer Prinzipien beim Christen nicht nur als ein Vergehen gegen den Staat, sondern noch mehr als ein Gott zugefügtes Unrecht, eine Sünde, für deren Auslöschung die Kirche dem Fehlbaren ein streng geregeltes Verfahren der poenitentia auferlegt" (55). See also von Wartburg (1955: 381-382, s.v. "opus"). 55. For Friulian, see the paradigms given by Marchetti (19854: 296-297): ' scuen, tu scuegnis, al scuen, 'o scugnìn, a' scuégnin (also: 'o scugni, tu scùgnis, al scugne, 'o scugnìn, 'o scugnìis, a' scùgnin); cf. Pirona (1983: 993). 56. See Jud (1946-7: 44-45): "Es scheint mir näherliegend, dass das Präsens von 'stovair' sich an dasjenige seines gleichbedeutenden 'dovair' angeglichen hat: dem stouva 1 ent­ spricht im Schlussvokal daia 1, stouvas 2 : daias 2, stouva 3 : daia 3, daiva 3; stouven 6: daien 6, dain 6. Diese zweisilbigen Formen daia — und damit auch stouva — sind genau gleich zu beurteilen wie die von E. Löfstedt aus dem Spätlatein verzeichneten debeat, oporteat in der Bedeutung von debet, oportet (Syntactica, p. 130, 491). Mit Recht bemerkt Löfstedt: 'Wenn so viele Vorschriften, Satzungen, Verordnungen, kurzum so viele Aus­ drücke für das, was geschehen soll, regelmässig im Konjunktiv erscheinen, so ist es kaum verwunderlich, dass unter Umständen auch das Verbum des Sollens in den Modus des Sollens tritt'." 57. For the cognitive-psychological perspectives, see now the thought-provoking study of Ivanov (1978).

References Armstrong, D. &  H. Van Schooneveld. eds. 1977. Roman Jakobson. Echoes of his Scholarship. Lisse: Peter de Ridder. Benveniste, E. 1963. Coup d'oeil sur le développement de la linguistique. Paris: Académie des Inscriptions et Belles-Lettres. —. 1966. Problèmes de linguistique générale I. Paris: Gallimard. Blanche-Benveniste, C , J. Deulofeu, J. Stéfanini, & . Van den Eynde. 1984. Pronom et syntaxe. Paris: SELAF. Bloomfield, L. 1926. "A Set of Postulates for the Science of Language". Language 2.153-164. (Repr. in IJAL 15. 195-202, 1949; Joos ed. 1957. 19-25; Hockett ed. 1970. 128-138). —. 1927. "On some rules of Panini". Journal of the American Oriental Society 47.61-70. (Repr. Hockett ed. 1970. 157-165). —. 1933. Language. New York: Holt. —. 1939. "Menomini Morphophonemics". Travaux du Cercle linguistique de Prague 8.105-115 (excerpts in Hockett ed. 1970. 351-362). —. 1945. "On Describing Inflection". Festschrift für M. Blakemore Evans (Monatshefte für deutschen Unterricht 37.4-5) 8-13. —. 1958. Eastern Ojibwa: Grammatical Sketch, Texts and Word List. Ann Arbor: University of Michigan Press. —. 1962. The Menomini Language. New Haven: Yale University Press. Bréal, M. 19084. Essai de sémantique (Science des significations). Paris: Hachette (18971). Busse, W. & J. P. Dubost. 1977. Französisches Verblexikon. Stuttgart: Klett-Cotta.

216

PIERRE SWIGGERS

Chomsky, N. 1957. Syntactic Structures. The Hague: Mouton. —. 1975. The Logical Structure of Linguistic Theory. Chicago: University of Chicago Press. —. 1977. Essays on Form and Interpretation. Amsterdam: North Holland Publishing Company. —. 1981. Lectures on Government and Binding. Dordrecht: Foris. —. 1982. Some Concepts and Consequences of the Theory of Government and Binding. Cambridge, Mass.: M.I.T. Press. Comrie, B. 1981. Language Universais and Linguistic Typology. Syntax and Morphology. Oxford: B. Blackwell. Coseriu, E. 1952. Sistema, norma y habla. Montevideo: Universidad de la República (German translation "System, Norm und Rede", in E. Coseriu, Sprachtheorie und allgemeine Sprach­ wissenschaft, 11-101. München: W. Fink). —. 1958. Sincronía, diacronía e historia. El problema del cambio lingüístico. Montevideo: Universidad de la República. —. 1967. "L'arbitraire du signe. Zur Spätgeschichte eines aristotelischen Begriffes". Archiv für das Studium der neueren Sprachen 204.81-112. Damourette, J. & E. Pichon. 1911-1940. Des mots à la pensée. Essai de grammaire de la langue française. Paris: d'Artrey. Engler, Rudolf. 1962. "Théorie et critique d'un principe saussurien. L'arbitraire du signe". Cahiers Ferdinand de Saussure 19, 5-66. —. 1964. "Compléments à l'arbitraire". Cahiers Ferdinand de Saussure 21. 25-32. Gamkrelidze, T. 1974. "The Problem of'L'arbitraire du signe'". Language 50, 102-110. —. 1976. "Linguistic typology and Indo-European reconstruction". Linguistic studies offered to Joseph Greenberg on the occasion of his sixtieth birthday, vol. II, ed. by A. Juilland, 399-406. Saratoga: Anma Libri. —. 1979. "Hierarchical Relationships of Dominance as Phonological Universais and their Implications for Indo-European Reconstruction". In Studies in Diachronic, Synchronic, and Typological Linguistics. Festschrift for Oswald Szemerényi on the occasion of his 65th birthday, ed. by B. Brogyanyi, 283-290. Amsterdam: J. Benjamins. —. 1981. "Language Typology and Language Universais and their Implications for the Reconstruction of the Indo-European Stop System". Bono homini donum. Essays in Historical Linguistics, in Memory of J. Alexander Kerns, ed. by Y. Arbeitman & A. Bomhard, 571-609. Amsterdam: J. Benjamins. — & V. Ivanov. 1973. "Sprachtypologie und die Rekonstruktion der gemeinindogerma­ nischen Verschlüsse". Phonetica 27. 150-156. —. 1984. Indoevropejskij jazyk i indoevropejcy. Tbilisi: Izd. Tbilisskogo Univ. (2 volumes). Gartner, Th. 1883. Raetoromanische Grammatik. Heilbronn: Henninger. —. 1910. Handbuch der rätoromanischen Sprache und Literatur. Halle: Niemeyer. Greenberg, J. 1970. "Some Generalizations Concerning Glottalic Consonants, Especially Implosives". International Journal of American Linguistics 36.123-145. —, Ch. Ferguson, & E. Moravcsik. eds. 1978. Universals of Human Language. Stanford: Stanford University Press (4 volumes). Gross, M. 1975. Méthodes en syntaxe. Régime des constructions complétives. Paris: Hermann. Haider, H. 1985. "The Fallacy of Typology. Remarks on the PIE Stop-system". Lingua 65.1-27. Haiman, J. 1980. "The Iconicity of Grammar". Language 56.515-541. Harris, Z. S. 1951. Methods in Structural Linguistics. Chicago: University of Chicago Press. —. 1962. String Analysis of Sentence Structure. The Hague: Mouton. —. 1981a. Grammar of English on Mathematical Principles. New York: Wiley & Sons. —. 1981b. Papers on Syntax. Ed. by Henry Hiż. Dordrecht: Reidel.

GRAMMATICAL PARAMETERS

217

Haudricourt, Α. 1975. "Les mutations consonantiques (occlusives) en indo-européen". Mé­ langes linguistiques offerts à Emile Benveniste, 267-272. Louvain: Peeters. Hockett, Ch. F. 1942. "A System of Descriptive Phonology". Language 18.3-21 (Repr. Joos ed. 1957. 97-108). —. 1947. "Problems of Morphemic Analysis". Language 23. 321-343 (Repr. Joos ed. 1957. 229-242). —. 1948. "Potawatomi". International Journal of American Linguistics 14.1-10, 63-73, 139-149, 213-225. —. 1950. "Peiping Morphophonemics". Language 26. 63-85 (Repr. Joos ed. 1957. 315-348). —. 1952. "A Formal Statement of Morphemic Analysis". Studies in Linguistics 10, 27-39. —. 1954. "Two Models of Grammatical Description". Word 10, 210-231 (Repr. Joos ed. 1957. 386-399). —. 1959. "Animal 'Languages' and Human Language". Human Biology 31, 32-39. —. 1961. "Linguistic Elements and their Relations". Language 37, 29-53. —. 1966. "What Algonquian is Really Like". International Journal of American Linguistics 32. 59-73. —. ed. 1970. A Leonard Bloomfield Anthology. Bloomington: Indiana University Press. —. 1977. The View from Language. Selected Essays 1948-1974. Athens: University of Georgia Press. Holenstein, E. 1975. Roman Jakobsons phänomenologischer Strukturalismus. Frankfurt am Main: Suhrkamp. Hopper, P. 1973. "Glottalized and murmured occlusives in IE". Glossa 7.141-166. —. 1977a. "The Typology of the PIE Segmental Inventory". Journal of Indo-European Studies 5.41-53. —. 1977b. "Indo-European Consonantism and the New Look". Orbis 26, 57-72. —. 1982. "Areal Typology and the Early IE Consonant System". The Indo-Europeans in the Fourth and Third Millennia, ed. by E. C. Polomé, 121-139. Ann Arbor: Karoma Publishers. Ivanov, V. 1978. Čet i nečet. Asimmetrija mosga i snakovych system. Moskva: Sovjetskoe radio (German translation by W. Petri: Gerade und ungerade. Die Asymmetrie des Gehirns und der Zeichensysteme. Stuttgart: Hirzel, 1983). Jaberg, K. & J. Jud. 1928-1940. Sprach- und Sachatlas Italiens und der Südschweiz. (8 volumes) Zofingen: Ringier. Jakobson, R. 1941. Kindersprache, Aphasie und allgemeine Lautgesetze. Uppsala. (Reprint from Språkvetenskapliga Sällskapets i Uppsala Förhandlingar 1940-42). (Repr. in Jakobson 19712: 328-401; English translation: Child Language, Aphasia and Phonological Universais. The Hague: Mouton, 1968.) —. 1958. "Typological Studies and their Contribution to Historical-comparative Linguis­ tics". Proceedings of the Eighth International Congress of Linguists, Oslo 1957, 17- 25 (Repr. in Jakobson 19712: 523-532). —. 1959. "Zeichen und System der Sprache". (Repr. in Jakobson 1972a: 272-279). —. 1960. "Linguistics and Poetics". Style in language, ed. by T. Sebeok, 350-377. Cambridge, Mass.: M.I.T. Press. 1963. "Implications of language universais for linguistics". Universals of Language, ed. —. by J. Greenberg, 208-219. Cambridge, Mass.: M.I.T. Press (Repr. in Jakobson 1972a: 580-592). —. 1966. "Quest for the Essence of Language". Diogenes 51, 21-37 (Repr. in Jakobson 1972a: 345-359; French translation "A la recherche de l'essence du langage", Diogène 51.22-38, 1965). —. 1969. "Linguistics in its Relation to Other Sciences". Actes du Xe Congrès international des linguistes, Bucarest, 28 août - 2 septembre 1967, vol. I, 75-111. Bucarest: Ed. de l'Académie (Repr. in Jakobson 1972a: 655-696).

218

PIERRE SWIGGERS

—. 1970. "Language in Relation to Other Communication Systems". Linguaggi nella società e nella tecnica, 3-16. Milano (Repr. in Jakobson 1972a: 697-708). —. 19712. Selected Writings I: Phonological studies. The Hague-Paris: Mouton (19621). —. 1972a. Selected Writings II: Word and language. The Hague-Paris: Mouton. —. 1972b. Louvain Lectures. In Aspects of the theories of Roman Jakobson, ed. by M. Van Ballaer, Memoir, Katholieke Universiteit Leuven, 1973. —. 1973. Main Trends in the Science of Language. London: George Allen & Unwin. —. 1975. Coup d'oeil sur le développement de la sémiotique. Bloomington, In: Research Center for Language and Semiotic Studies. —. 1981. Selected Writings VII: Contributions to Comparative Mythology. Recent Studies in Linguistics and Philology. Retrospections. Bibliography. Berlin-New York-Amsterdam: Mouton. —& L. Waugh. 1979. The Sound Shape of Language. Brighton: Harvester Press. Jespersen, . 1922. Language, its Nature, Development and Origin. London: George Allen & Unwin. —. 1933a. The System of Grammar. London-Copenhagen: Allen & Unwin—Levin & Munksgaard. ——. 1933b. Linguistica. Selected Papers in English, French and German. Copenhagen: Levin & Munksgaard. Joos, M. ed. 1957. Readings in Linguistics. Washington: American Council of Learned Societies. Jud, J. 1946-47. "Altfrz. estuet. bündnerrom. stuver, stuvair" Vox Romanica 9, 29-56. Koerner, E.F.K. 1972. Contribution au débat post-saussurien sur le signe linguistique. Introduc­ tion générale et bibliographie raisonnée. The Hague: Mouton. Lehmann, Ch. 1984. Der Relativsatz. Typologie seiner Strukturen, Theorie seiner Funktionen, Kompendium seiner Grammatik. Tübingen: Narr. Lepschy, G. 1970. A Survey of Structural Linguistics. London: Faber & Faber (It. original: La linguistica strutturale. Torino: Einaudi, 1966). —. 1986. "European Linguistics in the Twentieth Century". Studies in the History of Western Linguistics in Honour of R.H. Robins, ed. by Th. Bynon & F. R. Palmer, 189-201. Cambridge: University Press. Marchetti, G. 19854. Lineamenti di grammatica friulana. Udine: Società filologica friulana. (19521). Matthews, P. H. 1974. Morphology. An Introduction to the Theory of Word Structure. Cam­ bridge: University Press. Mayerthaler, W. 1980. "Ikonismus in der Morphologie". Zeitschrift für Semiotik 2, 19-37. Mayrhofer, M. 1983. "Sanskrit und die Sprachen Alteuropas. Zwei Jahrhunderte des Widerspiels von Entdeckungen und Irrtümern". Nachrichten der Akademie der Wissenschaft in Göttingen, Phil.-Hist. Klasse, Jg. 83.5, 121-154. Meigret, L. 1550. Le tretté de la gramère françoèze. (Reprint Genève: Slatkine, 1972). Melikišvili, I. 1976. Relationship of Markedness in Phonology (In Georgian). Tbilisi: Mecenereba. Mounin, G. 1972. La linguistique du XXe siècle. Paris: Presses universitaires de France. Murray, S. 1983. Group Formation in Social Science. Edmonton: Linguistic Research Inc. Nida, E. Α. 1948. "A System for the Identification of Morphemes". Language 24, 414-441. —. 19492. Morphology. The Descriptive Analysis of Words. Ann Arbor: University of Michi­ gan Press (19461). Pesot, J. 1980. "Ikonismus in der Phonologie". Zeitschrift für Semiotik 2, 7-18. Pirona, G.A. 1983. Il nuovo Pirona. Vocabolario friulano. Udine: Società filologica friulana (anastatical reprint of 1935 edition). Roques, G. 1985. "La conjugaison du verbe vouloir en ancien français". Actes du IVe colloque international sur le moyen français, ed. by A. Dees, 227-265. Amsterdam: Rodopi.

GRAMMATICAL PARAMETERS

219

Rosén, H. 1984. Review of Y. Arbeitman & A. Bomhard eds. 1981, Bono homini donum. Essays in Historical Linguistics, in Memory of J. Alexander Kerns. Bulletin de la Société de Linguis­ tique de Paris 79.2, 76-91. Ross, J. 1980. "Ikonismus in der Phraseologie". Zeitschrift für Semiotik 2. 39-56. Sapir, E. 1921. Language. An Introduction to the Study of Speech. New York: Harcourt-Brace. —. 1949. Selected Writings of Edward Sapir in Language, Culture, and Personality. Ed. by D. G. Mandelbaum. Berkeley: University of California Press. Saussure, Ferdinand de. 1968-1974. Cours de linguistique générale. (Critical edition by Rudolf Engler). Wiesbaden: Harrassowitz (2 volumes). Schwartz, L. 1988. "Asymmetric Feature Distribution in Pronominal 'Coordinations'". Agree­ ment in Natural Language. Approaches, Theories, Descriptions, ed. M. Barlow & Ch. A. Ferguson, 237-249. Stanford: Center for the Study of Language and Information. Seiler, H. 1972. "Universals of Language". Leuvense Bijdragen 61. 371-393 (Repr. in Seiler 1977. 207-229). —. 1973. "Zum Problem der sprachlichen Possessivität". Folia Linguistica 6. 231-250 (Repr. in Seiler 1977. 187-206). —. 1977. Sprache und Sprachen. Gesammelte Aufsätze. München: W. Fink. —. 1982. "Inherent vs. Established Relation, Proximity vs. Obviation, and Two Types of Cahuilla Kinship Expressions". International Journal of American Linguistics 48, 185-196. —. 1983. Possession as an Operational Dimension of Language. Tübingen: Narr. —& Ch. Lehmann. eds. 1982. Apprehension. Das sprachliche Erfassen von Gegenständen. Teil I: Bereich und Ordnung der Phänomene. Tübingen: Narr. —& F. J. Stachowiak, eds. 1982. Apprehension. Das sprachliche Erfassen von Gegenständen. Teil II: Die Techniken und ihr Zusammenhang in Einzelsprachen. Tübingen: Narr. Shevoroshkin, V. & T.L. Markey. eds. 1986. Typology, Relationship and Time. A Collection of Papers on Language Change and Relationship by Soviet Linguists. Ann Arbor: Karoma. Shopen, T. ed. 1985. Language Typology and Syntactic Description. I: Clause Structure. II: Complex Constructions. III. Grammatical Categories and the Lexicon. Cambridge: Cambridge University Press (3 volumes). Silverstein, M. 1976. "Shifters, Linguistic Categories, and Cultural Description". Meaning in Anthropology, ed. by K. Basso & H. Selby, 11-55. Albuquerque: University of New Mexico Press. —. 1980. "The Three Faces of 'Function': Preliminaries to a Psychology of Language". Proceedings of a Working Conference on the Social Foundations of Language and Thought, ed. by M. Hickmann, 1-34. Chicago: Center for Psychosocial studies. —. 1985a. "The Functional Stratification of Language and the Problem of Ontogenesis". Culture, Communication, and Cognition: Vygotskian Perspectives, ed. by J. V. Wertsch, 205-235. Cambridge: University Press. —. 1985b. "Language and the Culture of Gender: At the Intersection of Structure, Usage, and Ideology". Signs in Society: Psychological and Sociocultural Studies in Semiotic Mediation, ed. by E. Mertz & R. Parmentier, New York: Academic Press. Stéfanini, J. 1975. "Tradition grammaticale et arbitraire du signe". Cahiers de linguistique, d'orientalisme et de slavistique 5/6, 373-386. Swiggers, P. 1979. "Roman Jakobson". Romaneske 4.1, 10-20. —. 1981a. "The History-writing of Linguistics: A methodological note". General Linguistics 21. 11-16. —. 1981b. "Comment écrire l'histoire de la linguistique?". Lingua 55, 63-74. —. 1983. "La méthodologie de l'historiographie de la linguistique". Folia Linguistica His­ torica 4, 55-79. —. 1984a. "L'arbitraire du signe linguistique". Neuphilologische Mitteilungen 85, 401-404.

220

PIERRE SWIGGERS

—. 1984b. "Typological and Universal Linguistics". Lingua 64. 63-93. —. 1985. "Transitivité sémantique dans l'adverbe. Un exemple arménien". Neuphilologische Mitteilungen 86, 316-318. 1986a. "La linguistique fonctionnelle du Cercle de Prague". Philologica Pragensia 22, —. 76-82. —. 1986b. "La linguistique historico-comparative d'Antoine Meillet: Théorie et méthode". Cahiers Ferdinand de Saussure 39, 181-195. —. 1986c. "La linguistique historique devant la variation: le cas de Meillet". Recherches sur le français parlé 7, 61-74. 1986d "Le mot: unité d'intégration". Etudes de linguistique générale et de linguistique —. latine offertes en hommage à Guy Serbai, 57-66. Paris: SIG. 1989a. "Towards a Characterization of the Proto-Indo-European Sound System". The —. New Sound of Indo-European, ed. by Th. Vennemann, 177-208. Berlin-New York: Mouton de Gruyter. —. 1989b. "La Grammaire de Port-Royal et le parallélisme logico-grammatical". Orbis 33, 29-56. —& . Van den Eynde. 1984. "French Verb Morphology. Some Problems of Morphophonology". Paper presented at the Fifth international phonology meeting (Eisenstadt, 1984). — & —. 1988a. "Item-and-Arrangement vs. Item-and-Process Models: The problem of abstraction and abstractedness in linguistic Description" (manuscript). — & —. 1988b. "La morphologie du verbe français". ITL 77-78, 151-251. — & —. 1988c. "Allomorphes et morphèmes alternants". Bulletin de la Société de Linguis­ tique de Paris 73, 319-322. —— & W. Van Hoecke. eds. 1986. Mot et parties du discours) Word and Word Classes / Wort und Wortarten. Louvain: Leuven University Press-Peeters. — & Kutumisa Kyota. 1988. "Concomitance-mediated Plurals and Grammatical Person Hierarchy in Yaka" Folia Linguistica 22, 387-395. Szemerényi, . 1985. "Recent Developments in Indo-European Linguistics". Transactions of the Philological Society 1985, 1-71. Tobler, A. & E. Lommatzsch. 1954. Altfranzösisches Wörterbuch. 3. Band. Wiesbaden: Steiner. Van den Eynde, K. 1968. Eléments de grammaire yaka. Phonologie et morphologie flexionnelle. Kinshasa: Publications de L'Université Lovanium. —& Kutumisa Kyota. 1984. "Procédés tonals de construction syntaxique en yaka (Zaïre)". Discussion Papers. Fifth International Phonology Meeting, June 25-28, 1984. ed. by W. Dressler, . Pfeiffer, & J. Rennison, Wiener Linguistische Gazette, Beiheft 3, 68-72. Vennemann, Th. ed. 1989. The New Sound of Indo-European. Berlin-New York: Mouton de Gruyter. Voegelin, Ch. F. 1947. "A Problem in Morpheme Alternants and their Distribution". Language 23. 245-254. von Wartburg, W. 1955. Französisches Etymologisches Wörterbuch. 7. Band. Basel: Zbinden. Watkins,  1955. "The Phonemics of Gaulish. The Dialect of Narbonensis". Language 31. 9-19. Waugh, L. 1976. Roman Jakobson's Science of Language. Lisse: Peter de Ridder Press. Whitney, W.D. 1971. Whitney on Language. Selected Writings of William Dwight Whitney. Ed. by Michael Silverstein. Introductory essay by Roman Jakobson. Cambridge, Mass.: M.I.T. Press.

Invariance and Markedness in Grammatical Categories Madeleine Newfield and Linda R. Waugh Cornell University Roman Jakobson's 1957 study, "Shifters, Verbal Categories and the Russian Verb", has been extremely fruitful, to judge from the papers gathered in this section (and in section four of this volume), for each paper in one way or another echoes some facet of that work. In line with the general theme of invariance amidst variation, the present contributions are concerned with the nature of the semantic invariants that underlie the grammatical structures of the world's languages: the limits on the universality of grammatical categories, the possible types of formal reflections of these universal semantic categories, the evidence that change brings to bear on these universal ques­ tions, the relation between markedness and invariance, and the correlation between invariance/variation and code/message (langue/parole). The follow­ ing represents some of our reactions to the ways these issues are addressed here. 1. Our first concern is with the search for invariance in the sense of universal, cross-linguistic typological generalizations (see the contributions by Mel'čuk and Aronson). Since Jakobson's schema (1957) provides a feature-like analy­ sis of the semantic qualities of (verbal) grammatical categories, an analogy with distinctive feature analysis in phonology might offer some insights (cf. the discussion by Aronson). Jakobson posits factors such as the narrated event (En), the speech event (Es), and the participants in either of these events (Pn or Ps) as necessary for the analysis of all verbal grammatical categories. We would thus expect these basic semantic qualities, like phono­ logical distinctive features, to have cross-linguistic relevance. However, the grammatical categories themselves correspond to phonemes, and hence would not be expected to be universally applicable in the same way. Just as the phoneme /a/ in two languages may have different distinctive feature make-ups in accordance with the phonological system of each language, so the category mood in various languages may not always correspond to the same basic semantic qualities because of the differences in its usage and

222

MADELEINE NEWFIELD AND LINDA WAUGH

because of its interaction with the other grammatical categories in the system of each language. Other difficulties with the universal characterization of grammatical categ­ ories arise because of the lack of consistency and rigorous definition in traditional category labels. A possible bias may be introduced by relying on the category names and classifications presented in source grammars of specific languages. We may ask to what extent the terminological samenesses and differences in the literature represent real notional or semantic same­ nesses and differences. It is not clear, for example, that what has been called mood in Russian, French, and Hopi necessarily refers to categories with the same general meaning, i.e., that require the same features in their definitions. In the literature, mood covers a considerable range — epistemic and deontic meanings, conditional and futurity senses, evidential and quotative to vol­ itional and imperative uses. While arguments can be made for the likeness of all these, 1 there are also some clear semantic differences between them, which may be understood in terms of whether the speech event and its participants or another narrated event are inherently refered to. 2 Note that these various traditional senses of mood are not all characterized alike in Jakobson's original schema, where mood, evidential, and status are given separate definitions, and note furthermore that they are construed differently in each of the two proposed revisions here. Indeed, there are substantial differences in the definitions of a number of grammatical categories in the conceptions of the verbal "calculus" by Jakobson, Mel'čuk, and Aronson. Although Jakobson devised this schema as a category system accessible to all researchers, the divergent conclusions reached by Mel'čuk and Aronson particularly with respect to the relational categories (e.g., PnEn) show that it may have very different applications. This is disturbing, if we want it to have universal validity. The difficulties with labeling and definition are augmented by the fact that even within the system of a given language, what are labeled as different categories may not be clearly distinguishable from one another. For example, after carefully considering the distinction between voice, mode, and aspect in the semantics of Jakarta Malay verbal prefixes, Wallace concluded that the category labels do not capture the semantics of the prefixes (1979, see also 1982 and Waugh 1975). Aronson shows in numerous examples that there are similarities and links between the various categories (and defi­ nitions) he proposes. These may be indicative not only of the relatedness between the categories in terms of their defining qualities, but also of coalescences and overlappings. Certainly such blurrings of categorial distinc­ tions are not surprising when we take into account a more dynamic perspec­ tive. There are well known instances where, for example, aspect develops into tense, and we would expect to find some confusion if we were to

INVARIANCE AND MARKEDNESS

223

examine a language at an intermediate stage in this development. Moreover, in his contribution to this volume, Robertson uses Mayan examples to illustrate the diachronic stability of the semantics of -le if a wide-ranging enough definition of its general meaning is chosen. However, his example might alternatively demonstrate a change in the status of voice, since over time the once passive -le loses its reference to the participants of the narrated event, and informs only about the narrated event itself. This categorial difference does not seem to be crucial for Robertson, but would be for Mel'čuk and Aronson (and Jakobson). Summing up this general point: while we do not intend to dispute the important insights gained in seeking universal definitions of grammatical categories, it would be wise to avoid interpreting the latter as absolutes. This is especially true when we use these labels both for particular categories in a given language at a specific stage in its development, and as metalinguis­ tic universal cover terms. 2. Thus far, we have been considering only the semantic side of grammatical categories, but there are further issues to consider if we take into account their formal exponents. It is notable that while both Mel'čuk and Aronson pay relatively little attention to questions of form, the final section of Jakobson's original study (1957: 143-146) made an insightful correlation of the semantics of the Russian verbal categories with the manner of their formal expression, whether by periphrasis, particles, morphology (including stem alternation), or a zero sign (see also Jakobson 1939). In general, the kinds of grammatical meaning under discussion may be represented by a variety of formal means, and not always by a formal manifestation of the word that the meaning applies to. The gender or number of a noun, for example, may be formally revealed only by a marker affixed to an adjective or verb which agrees with that noun. Furthermore, some categories are manifested only in certain kinds of sentences or usages, as, for example, noun gender in English, which is present only when the word is replaced by a pronoun. In a paper which deserves more attention, Whorf contrasts such "covert" categories to "overt" categories where the marker is generally present (1945: 88-90) and emphasizes that grammatical information may be realized not only as morphology, but also through word order and other types of syntactic patterning, lexical selection, or even the avoidance of certain co-occurrences. The importance of these non-morphological types of grammatical infor­ mation is shown both synchronically and diachronically by MonvilleBurston's paper in this section; she demonstrates convincingly the role of animacy in the change from dative to genitive in middle Indo-Aryan. Mel'­ čuk, however, insists on linking grammatical categories to morphological

224

MADELEINE NEWFIELD AND LINDA WAUGH

expression (the presence of a formal marker on an inflected word). It is not clear on what basis he eschews covert and/or non-morphological expression. This is akin to Sangster's claim (this volume) that by definition the morpho­ logical tenses of English constitute the central system and the periphrastic ones a more complex overlay. While we agree that form is an important determinant of the categories to be dealt with in any semantic analysis and while it is clear that iconicity is an important property of linguistic systems (see Waugh & Newfield ms.) — more important than many linguists admit — it is still a question as to whether iconicity exists in any given case. It is clear from work on the French tense system, for example, that the morpho­ logical simple past may at one time have been more central than the periphrastic compound past, but that in modern French the situation is reversed (see Waugh 1987). A related issue is that of how to treat inherent (fixed) grammatical categories, such as noun classes (e.g., gender) or verbal classes (e.g., transitiv­ ity), in addition to the familiar part of speech categories (lexical classes). Both of these types were called selective by Whorf (1938, 1945, 1956: 125-133):3 this denotes "a grammatical class with membership fixed, and limited as compared with some larger class....Certain semantic and gram­ matical properties are assured in the word by selecting it from a certain class of fixed membership not coterminous with the whole vocabulary" (1945: 93). These are to be differentiated from modulus categories, "which modify...a word already allocated to a delimited class" (1938: 113) and are "generally applicable and removable at will" (1945: 95), examples being case, tense, aspect, voice, etc. Mel'čuk takes exception to the idea that the inherent categories are grammatical categories at all. He claims that they are rather "syntactic properties" defining "lexemic classes", since, according to him, they have "nothing to do with meanings", being characteristics which are "neither expressible nor changeable". And yet, the inherent categories, like his inflectional categories, represent obligatorily specified qualities, and obli­ gatoriness has been seen to be the primary defining characteristic of gram­ matical categories, by Mel'čuk and others (Jakobson 1959, Matthews 1974, Bybee 1985). The argument that they are not expressible rests on Mel'čuk's apparent requirement that a category be overt (morphological) in the same word where its meaning lies; but, as mentioned above, this is overly strict. Nor do the fixed values of inherent categories for a given word necessarily make them mere meaningless syntactic markers, as Mel'čuk implies. Noun gender, for example, potentially provides much information. In many cases, it corresponds to the experiential classification of animacy or sex. And in English, at least, gender is also, as Whorf suggested, a covert linguistic category, not wholly predictable from the referential world, as

INVARIANCE AND MARKEDNESS

225

shown by the use of on the one hand "she" for boats, cars, countries, and cities and on the other hand "it" for babies and small animals. Even such seemingly referentially based distinctions as gender ones are given a linguis­ tic stamp. 4 While the inherent (selective) categories are semantically relevant, they are also different from the non-inherent (modulus) categories. The former, because of their invariant association with lexical items, are by nature tied to the linguistic system (the code), and don't appear to be determined by either the speech event or the narrated event; that is, their intrinsic semantic nature does not depend on aspects of their usage in message contexts. This accounts for their fixed quality. In contrast, what makes for the variation in application of the modulus (inflectional) categories is the very fact that they depend on context for their semantic fulfillment. They are intrinsically linked at least to the content of the message (the narrated event), if not also to the circumstances of the message (the speech event), as when they are shifters proper. Consequently, what the modulus categories seem to share, in contrast to the inherent ones, is an indexicality in their functioning — the necessity of pointing either to elements of the speech event or to other elements in the narrated context (see Anttila 1975). We can exemplify this difference if we compare (e.g., in Russian) gender in nouns, where it is an inherent and fixed category, with gender in verbs, pronouns, and adjectives, where it is a modulus category. In the latter, the specification of gender is obligatory but variable, depending on the indexical relations entailed by grammatical agreement. The modulus type of gender thus involves, in the verb for example, a reference to the participants of the narrated event (Pn), as in Jakobson's original proposal (1957), whereas the selective variety of gender, in the noun, includes a reference to the code. This then accounts for Aronson's alternative formulation, which suggests that an additional basic quality is needed, what he calls "characterizing with reference to the lexicon".5 This phrasing is somewhat confusing, however, since it obscures the fact that the inherent category, being obligatory, involves aspects of grammatical meaning, not just lexical meaning. 3. There are further issues concerned with the difference between lexical and grammatical meaning. Mel'čuk seems at first glance to imply that the lexicalgrammatical difference is based only on considerations of form. Similarly, he ties the definition of derivational meaning to the irregularity of its formal expression as compared with inflection.6 This, along with his admission that these distinctions are often blurred, needs elaboration; we still lack precise details about the exact nature and qualities of combination of these kinds of meaning. Mel'čuk himself emphasizes the gradual nature of the concept of inflectional category (in connection with the gradual nature of its regu-

226

MADELEINE NEWFIELD AND LINDA WAUGH

larity). In our opinion, this gradualness needs to be extended beyond the area of inflection; we envision the whole lexical-derivational-inflectional question not as strict differences but as a continuum (see Sapir 1921: ch.5, also Bybee 1985). Certainly, clustering of characteristics typical for one or another type of element or meaning are evidenced along the continuum: the propensity for a11-or-none analyses of this question in the literature shows that the linguistic data are not evenly distributed along it. But the continuum also manifests transitional types; and change involving movement along the continuum sometimes creates hard-to-classify intermediate stages. 4. One of the concepts in Jakobson 1957 which had been promulgated a quarter of a century earlier by him and Trubetzkoy 7 and which has had much success in the more than quarter of a century since,8 is that of markedness. Indeed, one of the assumptions underlying the contributions of Sangster, Robertson, and Monville-Burston in this volume is that mark­ edness is the main means by which the grammatical meanings in oppositions such as perfective-imperfective, animate-inanimate, singular-plural are defined. While we are in agreement about the primary importance of mark­ edness as a structuring principle, it should be pointed out that there has been considerable confusion concerning its definition (see Jakobson 1974 and Jakobson & Waugh 1979a for a similar point); moreover, in some cases (e.g., in the paper by Sangster) the notions of mark, invariant and feature are not well enough differentiated from each other. Below, we will attempt to clarify these concepts.9 In the linguistic literature the term feature has taken on two correlated, but different, senses. On the one hand, it has been used to refer to a given opposition as a whole, be it phonological (e.g., the voiced-voiceless feature) or grammatical (e.g., the animate-inanimate feature) — or even lexical, in those cases where a lexical relation is defined in oppositional terms, e.g., optimal extension vs. reduced extension in the spatial terms "high" — "low", "deep" — "shallow", "wide" — "narrow", "far" — "near" (see Jakobson 1976). In such cases, the opposition is necessarily binary in nature. But, on the other hand, and somewhat confusingly (as Fischer-Jørgensen has complained — see 1975: 146), one of the two poles of a given opposition has also been called a feature. Thus, "the feature of voicing" may denote the voiced-voiceless opposition as well as the voiced pole alone (and likewise for "the animacy feature"). That the term feature has been used in these two senses is due to the fact that each of the poles of an opposition necessarily implies both the other pole and the opposition as a whole. As long as this mutual implication is recognized, the term feature can stand for both the opposition and each of its binary choices without confusion (see Jakobson 1974, Jakobson & Waugh 1979a: 27).

INVARIANCE AND MARKEDNESS

227

Nevertheless, there has sometimes been a tendency, when speaking of one pole of an opposition, to treat that pole as if it existed in isolation, as if it were autonomous rather than relational. Thus "the animacy feature" (or "the extension feature") is treated as if it had to do with animacy alone (or extension alone) rather than with the animate-inanimate (or extensionnonextension) opposition, of which animacy (or extension) is but one of two choices. Given such usage, and the attendant confusion, our suggestion is that the term feature be reserved for one of the two poles of an opposition, and the term opposition be used for the opposition as a whole: thus, the two features, voiced and voiceless, make up the voiced-voiceless opposition. This is the usage we will follow here. In like fashion, the term mark (and correlated terms markedness, marked, unmarked) has also been used somewhat confusingly. In the usage we prefer, mark designates a restrictive concept: in any given opposition, there is a privative relation between a marked pole and an unmarked one; i.e., one of the two features (as just defined) is the marked one. For example, either voicing or voicelessness, either animacy or inanimacy is the marked feature; in obstruents, the voiced feature is the marked one, while the voicelessness feature is the unmarked one. Which is marked depends on which is the more constrained (focalized) of the two. "The constraining, focusing charac­ ter of the marked term of any grammatical opposition is directed toward a more narrowly specified and delimited conceptual item. In the dyads of distinctive features the marked term is opposed to the unmarked one by its closer concentration on a certain, either positive or negative perceptual sound property polar to that of the unmarked term, and is accordingly characterized by a restriction of occurrence to specific sequential or concur­ rent contexts" (Jakobson & Waugh 1979a: 94; see also Waugh 1982). In other words, the unmarked term is less limited in its contextual variation: it occurs in a larger number of different contexts, and exhibits a larger number of different variants; moreover, it tends to be more frequent and to be learned earlier in child acquisition than its opposite (see Greenberg 1966 for a discussion of these and other factors). 10 To sum up, then, a given opposition is made up of two features, either of which may be the marked one (may be the mark); and markedness correlates with the fact that one of the two terms is more narrowly specified and delimited, which in turn is related to contextual variation. Thus, the animacy opposition comprises two features — animate and inanimate — and either the animate feature or the inanimate is the marked one, as evidenced by contextual variation. Contextual variation, and its correlated opposite invariance, are very gen­ eral properties of language, since they pertain to all levels and all units, not only phonological, morphological and syntactic, but also both oppositional

228

MADELEINE NEWFIELD AND LINDA WAUGH

and non-oppositional (non-binary) lexical categories. Since our concern is with oppositional categories, we will leave aside the much more difficult question of invariance in non-oppositional parts of the lexicon. For any given opposition, what is invariant is the definition of the opposition as a whole, and thus the two features which make up the opposition. For example, the opposition voiced-voiceless is invariantly, and relationally, characterized as more vs. less (including no) voicing. This means that the voicing feature invariantly designates more voicing than the voicelessness feature; voiced-voiceless is not presence vs. absence of voicing only, since in some contexts there may be some voicing with /p/; but there is always more voicing with /b/. In a given context, this may mean no voicing for the voicelessness feature versus some, perceptually discernable amount, for the voicing feature; or, it may mean some residual amount of voicing for the voicelessness feature (as when unvoiced obstruents occur between vowels) versus a comparatively stronger, discernable amount for the voicing feature. (Some of these differences are captured through the concept of voice onset time; see also Jakobson & Waugh 1979: 138-142.) Thus, the invariance of the opposition as a whole, as well as of each feature of the opposition, is defined with respect to the relationship between the two choices. The association of a given phonological entity (e.g., phoneme) with one distinctive feature or the other is also invariant. For example, in the oppo­ sition between /p/ and /b/ in Russian, /p/ is invariantly voiceless and /b/ voiced: that is, in any context where the two can both occur, /b/ is invariantly more voiced than /p/. Of course, in contexts where the voiced-voiceless opposition is not possible — e.g., in word-final position — this characteriz­ ation is not viable, since in this case neither a voiceless /p/ nor a voiced /b/ occurs, but rather a neutralized or "incomplete" phoneme (see Trubetzkoy 1939, Jakobson & Waugh 1979) which is externally similar to voiceless /'p/, due to markedness. The marked or unmarked status of a given feature is a function of the context in which it is found and thus may vary according to the context. Whether, for example, it is voicing or voicelessness which is marked is dependent on the context: in obstruents voicing is marked but in resonants voicelessness is marked. Thus:

Obstruents Resonants

Voiced marked unmarked

vs. vs. vs.

Voiceless unmarked marked 11

(e.g., /b/ vs. /p/) (e.g., /n/ vs. /N/)

This correlates with the fact that on the one hand voiced nasals, liquids and vowels and on the other hand voiceless stops and fricatives exhibit the wider contextual variation discussed above than their marked counterparts.

INVARIANCE AND MARKEDNESS

229

Now, in grammatical oppositions, the same situation obtains. 12 Each grammatical element of an opposition is endowed with a conceptual invariant (general meaning, Gesamtbedeutung) which is defined in terms of the conceptual choice between the two features (e.g., one vs. more-than-one for singular-plural). In addition, there is a privative opposition between the presence of a mark and its absence. The unmarked pole evidences wider contextual variation than the marked one, but this has more striking conse­ quences than in phonology. Since the marked pole signals χ (the mark), the unmarked pole signals non-x. Now, non-x is interpreted in two major ways: as the non-signalization of x and as the signalization of non-x. 13 Thus, as Holenstein (1976: 131) has said, any opposition is composed of two major sub-oppositions, "between a positive and a negative term and between an indefinite and a definite one":

1. 2.

unmarked negative (signalization of non-x indefinite (non-signalization of χ

vs. vs. vs. vs. vs.

marked positive signalization of x) definite signalization of x)

The negative (minus) interpretation of the unmarked pole is often charac­ terized as its basic (core, nuclear) meaning {Grundbedeutung). It constitutes the meeting ground of the paradigmatic and syntagmatic axes: it is that contextual variant which is both the result of its opposition to the marked pole (since the basic meaning is that one where the mark is explicitly denied) and the least contextually bound (since it can occur in a greater variety of contexts and needs only minimal context). This relation of negative vs. positive is one of mutual exclusivity. As to the indefinite meaning, it includes the most generic meaning, in which the category as a whole is designated without specific subdivision into negative and positive; the relation of indefinite vs. definite is that of set-subset (ground-figure). All of these contextual possibilities may be exemplified by the present tense in many languages (see Comrie 1985, Waugh 1975, 1982, MonvilleBurston & Waugh 1990). Its basic meaning is reference to present time (here's my brother); its generic meaning can be evidenced by omnitemporal {man is a wolf for man) and atemporal {two and two are four) uses. 14 With this as a point of departure, we may now turn to the phenomenon of markedness reversals, the existence of which is disputed by Sangster. In his discussion of gender in Russian, Jakobson (1960: 185) defines a mark­ edness shift for the neuter: "The neuter, which is a specified, marked category in the case-forms, proves to be the least specified — the unmarked gender — among the caseless forms." The neuter is marked in the case-forms

230

MADELEINE NEWFIELD AND LINDA WAUGH

(nouns, long form of adjectives) because it signals a lack of sex reference: neuter nouns are used for inanimates or for generic designations of animate beings. The unmarked gender in case-forms is the masculine since it may refer to either sexual or asexual referents. In caseless forms (short form of adjectives, preterit form of verbs) on the other hand, the neuter is the unmarked pole and thus evidences more contextual variation: it may relate the adjective or noun either to neuter subjects or to a lack of a substantival headword (impersonal sentences). In these contexts, the masculine is marked since it necessarily signals that there is a subject of masculine gender (see also Jakobson 1957). Thus:

Case-forms Caseless forms

Neuter marked unmarked

vs. vs. vs.

Masc unmarked marked 15

The specifics of the contextual variation are dependent on the type of word: they relate to types of referents in nouns and attributive (long-form) adjec­ tives, and to types of subjects in predicate forms (short-form adjectives and preterit of the verb). "Such shifts in the distribution of marked and unmarked categories in the caseless forms, as compared to the case-forms, are quite natural" (Jakobson 1960: 186). To take the example of number, also discussed by Sangster: nouns, pronouns, adjectives and verbs in many languages often evidence the oppo­ sition between singular and plural. As a first approximation, we can say that singular-plural is invariantly associated with the opposition between one (entity or perception) and more-than-one. Now, in some contexts, morethan-one is the marked term and one the unmarked term; in other contexts, it is the reverse. Simplistically speaking, the unmarked term is the one which is used for the more generic meaning: it stands for the category as a whole, or occurs when number is unknown or not at issue. It is, presumably, this line of reasoning about markedness reversals which is the basis for Stankie­ wicz' claim (this volume) that singular is marked in first and second person while plural is marked in third person, 16 thus:

first, second person third person

one (Sg) marked unmarked

vs. vs. vs.

more-than-one (Plur) unmarked marked

We can now return to our first approximation of the meaning for singularplural and note that it was not quite accurate: when singular is unmarked, one is its basic meaning; when it is marked, one is its general meaning. Likewise, when plural is unmarked, more-than-one is its basic meaning;

INVARIANCE AND MARKEDNESS

231

when it is marked, more-than-one is its general meaning. The difference lies in the presence or absence of the indefinite meanings (exemplified below with the generic variant), thus:

1. marked

2. unmarked

Singular one

vs. vs.

unmarked

basic = one generic = category as a whole

vs.

marked

Plural basic = more-than-one generic = category as a whole more-than-one

That is, when markedness is used as the structuring principle of the relation between invariance and variation, then all of the semantic possibilities are elegantly subsumed under one general formulation: the four seemingly dis­ tinct properties mentioned by Sangster (unmarked singular, marked singular, unmarked plural, marked plural) are defined through the shift in the mark­ edness relation pertaining to one opposition. Given this, the terminology of "reversals" is perhaps unfortunate since it seems to imply, wrongly, that one of the poles is originally the unmarked one and that something happens to it, somewhere, to cause it to become marked in a derivative way. That is not the implication here: we do not assume that something starts out unmarked and becomes marked; a given contextualization is either marked or unmarked. However, just as with contextual meanings, there may be a hierarchical relationship between the contextual markings: some may be more basic, others more marginal. In the light of this, we propose that we abandon the term markedness reversal and that we use instead Jakobson's more neutral term: "shift". We will adopt this usage in our discussion below. Markedness shifts are found within a given language at one time, over time (language change), 17 and across languages. Their existence has been noted by a variety of researchers.18 Sangster's discussion of markedness shifts is clouded by various prob­ lems. 19 Firstly, just as in early work on phonology, 20 he assumes that the unmarked pole of a feature is always that pole which is associated with the absence of the given quality, e.g., that [- voice] inevitably means that voicelessness is unmarked. That is, "-" is used to stand both for the absence of a given property and for the unmarked status of that same property, thus leading to an unfortunate conflation of the two different conceptual domains. 21 But as we have shown, in vowels [- voice] is the marked feature, thus necessitating a differentiation of the two concepts, mark and feature (see Jakobson 1974). Secondly, in the discussion of grammatical and lexical

232

MADELEINE NEWFIELD AND LINDA WAUGH

oppositions, the expression "marked or unmarked for" a given feature is used. But, this is inaccurate, /b/ and /m/ in Russian are not marked for voicing; rather they are both associated with the voicing feature, which is the marked pole of the voiced-voiceless opposition in obstruents (thus in /b/, voicing is a mark) and the unmarked pole in resonants (thus in /m/ voicing is not a mark). Likewise, a plural noun is not marked for reference to more-than-one; rather it is associated with the more-than-one pole of the opposition and this may be the marked or the unmarked pole depending on the concurrent and sequential context. Thirdly, Sangster assumes that if for a given element a given feature χ is the unmarked one, then χ is irrelevant to that item. But, again, as we can see with our example above, this is inaccurate. Voicelessness in /p/ is unmarked, but it is hardly irrelevant, since it differentiates /p/ from /b/. Likewise, in Russian, the possibility of reference to sex in masculine case forms is the unmarked pole of the opposition, but in certain contexts this reference is quite relevant, as witnessed by the use of masculine gender for male beings (Jakobson 1960). Lastly, association with the unmarked pole of a given semantic feature must be differentiated from not participating in a particular opposition at all. To take another example, in pronominal systems there are various oppositions which pronouns of one type participate in and others don't. Thus, to use Garcia's data (this volume), the singular el and the plural ellos are associated respectively with the unmarked and marked poles of the singular-plural opposition. But this opposition is irrelevant for si; si is neutral with respect to singular-plural. Thus:

singular-plural

el singular unmarked

ellos plural marked

si irrelevant/neutral

Two different claims are at issue here: on the one hand, the association of el with the unmarked pole of the singular-plural opposition; on the other hand, the irrelevance/neutrality of this same opposition for si. To claim that el and si are both "unmarked for" the singular-plural opposition is to do away with this important difference. The collapsing of unmarkedness with irrelevance/neutrality has, unfortu­ nately, been furthered by Jakobson's own work on the Russian case system (1936, 1958) in which he gave names not to the opposition as a whole, but to the marked pole only. This has led to the unfortunate tendency to talk solely in terms of marks rather than using the concepts of binary opposition, invariance, and mark, for grammatical and lexical meaning. Thus, semantic oppositions are sometimes defined as unary, not in polar terms: rather than

INVARIANCE AND MARKEDNESS

233

one vs. more-than-one and animate-inanimate, and so on, terms such as restrictedness and extension are used without any discussion of what their opposites are. The possibility of markedness shifts is thereby excluded, by definition.22 All of this is based on the conflation of feature with mark and mark with invariant. What is called a mark in Sangster's paper is what Jakobson originally meant by invariant (Gesamtbedeutung): it is the conceptual prop­ erty which is invariantly associated with a given element. But then, invariance and markedness, rather than being seen as constituting "two integral proper­ ties of linguistic structure" which "force us to use strictly relational, topologi­ cal definitions" (Jakobson 1972: 76), have been collapsed into one. And what is not defined, paradoxically, in Sangster's discussion, is the semantic content of the unmarked member of a grammatical opposition, since for him to be unmarked is to have no defining characteristic. There is no definition of nonrestrictedness or nonextension; we know what the perfective (dimensionality) or plural (plurality) is, but not what imperfective or singular is. Nor do we know how to differentiate imperfective from singular. So, rather than making two independent and correlated claims — for example, (1) that the perfective is invariantly associated with the dimensionality pole and the imperfective with the nondimensionality pole of the dimensionality vs. nondimensionality opposition, and (2) that the dimensionality pole is the marked one (in such and such a context) — the claims are collapsed into one. Invariance is defined not as association with one pole or other of an opposition, but rather with only the marked feature. And thus the allegation is made — as in Sangster's paper — that to discuss markedness reversals (shifts) is to deny invariance. This is a misconception: to discuss markedness shifts is to deny that a given pole of an opposition is invariantly the marked one. Invariance is maintained for the association of a given grammatical category with one pole or the other of an opposition. We suggest, then, that just as a phonological system is built out of a series of distinctive oppositions such as voiced-voiceless, tense-lax, etc., so a grammatical system should be conceptualized as being based on a set of semantic oppositions, such as animate-inanimate, dimensionality-nondimensionality, one vs. more-than-one. And while both systems associate such oppositions with the relation marked-unmarked, either pole may be the marked one and the markedness relation may shift depending upon the context. 5. These questions about the status of the contextualizations of grammatical categories lead us to our fifth topic, namely, the relationship between invari­ ance-variation and code-message (langue-parole, competence-perform­ ance). While, as Sangster suggests, the issue of the line of demarcation

234

MADELEINE NEWFIELD AND LINDA WAUGH

between code and message is an important one for theoretical linguistics, it seems to us that the two are seen in his discussion as too separate. In fact, there has been an unfortunate tendency in much linguistic work in the 20th century to define linguistics in various partial ways, for example, as a linguistics of code {langue, competence) only, while questions pertaining to message (parole, performance) have been deemed to be outside of the province of linguistics proper. But just as language is not to be equated with the code only, since both code and message are interdependent manifes­ tations of language, so a truly general linguistics should encompass both facets (witness the recent interest in sociolinguistics, discourse analysis, conversation analysis, and pragmatics). In like fashion, general meanings and contextual variants are two interrelated properties of signs. Both are necessary for a sign to work for communication and both are linguistic phenomena. The general meaning is not only part of the code, but is actualized in particular messages (and thus is neither merely potential nor abstract). In like fashion, there are many facets of contextual meanings which belong to the code as well as to the message; for example, as shown above, in oppositional relations, the particular contextual meanings which a category can evidence as well as the hierarchization of these meanings is determined by the marked-unmarked relationship. Moreover, the context which is at issue is perforce a linguistic one. The contextuaUzation of language is an extremely important issue which cannot be swept aside by claiming that contextual variants are only a facet of messages, are nonlinguistic, or are purely pragmatic. This does away with the properly linguis­ tic aspects of contextual meanings and curtails linguistic analysis. Thus, a too radical division is made between so-called inherent properties of linguistic elements and their external contextualizations. And yet, inherent properties often refer to those very contextualizations — e.g., in deictic categories (Jakobson 1957, Mel'čuk and Aronson, this volume), in highly context-sensitive elements (Waugh 1979b), and in the semantics of syntax (Waugh 1976c, 1977). Moreover, the relations between basic and general, basic and secondary (marginal), literal and figurative meanings, between types of meanings (see Waugh, this volume) are defined by criteria intrinsic to the given category. Furthermore, they are part of the code and at the same time they manifest themselves through contexts. It is part of the grammar of English that the past tense may have as two common variants reference to past time and hypotheticality. And it is also codified that the former is the basic variant while the latter is a more contextually dependent variant. An analysis of the past tense in only its past time uses would be a partial one, just as an analysis which assumes that all the contextual appli­ cations of the past tense are equal would be incorrect. We agree with Sangster that all contextual variants must be used and none should be left

INVARIANCE AND MARKEDNESS

235

aside no matter how strange or deviant. But we insist also that contextual variants themselves are hierarchized and a delimitation of this hierarchy should be made since it is linguistically relevant. We plea, then, for a recognition that the semantic structure of a language encompasses both invariance-variation and marked-unmarked in their mut­ ual dependence and hierarchization. Notes 1. 2. 3.

4.

5.

6.

7. 8. 9.

10. 11. 12.

13. 14.

See Chung and Timberlake (1985; 241-250), Lyons (1978); Bybee (1985), and Palmer (1986). These subtypes of mood may be differentiated, as Chung and Timberlake (1985) suggest, in terms of actual/non-actual events, and the inclusion of "source" and/or "target". Lyongs (1968: 274), for example, calls the parts of speech "primary grammatical categor­ ies" and all others such as gender, case, person and tense, "secondary grammatical categories". There may be some variation depending on the kind of realization of the category. For example, in German, Mädchen "girl" (neuter) gets the expected neuter agreement on an adjacent attributive adjective, but is typically referred to now by a feminine pronoun; likewise, in French, professeur, even when referring to a woman, takes a masculine attributive adjective, but a predicate adjective or a pronominal substitute may be feminine. What seems to be at issue here is the greater distance of the predicate or the pronoun from the noun; discourse factors, such as a proper name in the larger discourse context, may also play a role. A similar rationale may explain the difference between Jakobson's and Aronson's concep­ tions of voice. Voice is not equivalent to transitivity, but rather corresponds to it as a modulus variety in contrast to the apparently inherent nature of transitivity. He also quite correctly characterizes derivation in terms of its non-obligatory nature. Later, however, he counters the initial impression by emphasizing that "the distinction between inflectional meanings and derivational meanings is essentially dependent on their semantic nature". Jakobson (1932, 1936, 1939), Trubetzkoy (1939, 1975), Viel (1984). See for example Belletti et al. (1981), Eckman et al. (1986). And since the concept of markedness which Sangster addresses is within the Jakobsonian theoretical perspective, because of lack of space we will not address the interesting issue of how to correlate it with other theoretical perspectives. See also Jakobson (1932, 1936, 1957), Waugh (1979a, 1982.). See Jakobson (1960), Andersen (1968, 1972), Shapiro (1974), Jakobson & Waugh (1979a, b), Waugh (1979a). For a discussion of grammatical markedness, see Jakobson (1932, 1936, 1957, 1959, 1974), Greenberg (1966), Trubetzkoy (1975), Comrie (1976), Holenstein (1976: 121-37), Waugh (1976b: 89-100), Jakobson & Waugh (1979a), Jakobson & Pomorska (1980), Waugh (1982). Jakobson (1932, 1936, 1939, 1957), Waugh (1982); cf. Andrews (1985), who confuses the absence of a marking with its irrelevance (see discussion below). For other examples see Waugh (1975, 1982), Comrie (1976), Monville-Burston & Waugh (1991). Due to limitations of space, we will not be discussing the "hypostatic" use of the unmarked term, e.g., the historical present.

236

MADELEINE NEWFIELD AND LINDA WAUGH

15. We leave the feminine aside here; for the most part, it is marked vis-à-vis the masculine. 16. We could link this with a change in the English pronominal system today: he being used more and more for purely masculine sex reference and they taking over the function of generalized usage. This is perhaps an example of a shift in the markedness of singularplural in the third person. 17. As to the claim that markedness shifts can only happen in language change but not in language synchrony (Andrews 1985:178), let us remember that change originates in synchronic variation (dynamic synchrony). 18. For other examples of such shifts in semantics, see the 1930 letter by Jakobson to Trubetzkoy (in Trubetzkoy 1975), quoted in Jakobson & Waugh (1979a) and Waugh (1982), as well as Haiman (1975, 1985a: 148ff, 1985b), Shapiro (1976, 1983), Silverstein (1976), Dik (1978: 164 and section 5.7), Givón (1979: 108), Waugh (1982), Bailey (1982: 44, in press), Witkowski & Brown (1983), Andersen (this volume). 19. Although it should be pointed out that he is not alone and that this conflation also resides, for example, in earlier work by the second author (Waugh 1975, 1976a, c, 1979b) as well as in the seminal studies by van Schooneveld (see 1978). 20. For example, Jakobson, Fant, & Halle (1952), Jakobson & Halle (1956), Jakobson (1959); but cf. Jakobson (1960), Jakobson & Waugh (1979) where this assumption is no longer made. 21. In early phonological work this led to hesitations as to the naming of some of the oppositions since the marked term was given first: abrupt-continuant or continuantabrupt? 22. See Sangster, this volume, as well as earlier work by Andrews (1984a, b, 1985).

References Andersen, Henning. 1968. IE *s after i, u, r,  in Baltic and Slavic. Acta Linguistica Hafniensia 11, 171-190. —. 1972. "Diphthongization". Language 48. 11-50. Andrews, Edna. 1984a. A Theoretical Foundation for Markedness: Asymmetry in language from a mathematical perspective. Indiana University Doctoral Dissertation. —. 1984b. "Markedness Theory: an Explication of its Theoretical Basis and Applicability in Semantic Analysis". Memorial Volume in Honor of J. Daniel Armstrong, ed. by Charles Gribble et al. Columbus, OH: Slavica. —. 1985. "Markedness Reversals in Linguistic Sign Systems". In Memory of Roman Jakobson: Papers from the 1984 MALC, ed. by G. Youmans & D. Lance. Columbia, MO: Linguistics Area Program, 169-180, Anttila, Raimo. 1975. The Indexical Element in Morphology. Innsbruck: Institut für Sprachwis­ senschaft der Universität Innsbruck. Aronson, Howard. this volume. "Towards a Typology of Verbal Categories". Bailey,  J. N. 1982. On the Yin and Yang Nature of Language. Ann Arbor, MI: Karoma. —. in press. "Markedness-reversal and the Pragmatic Principle of 'Reading between the Lines in the Presence of Marked Usage'". Tübingen Working Papers in Linguistics, no. 12, 1-71. Belletti, Andriane et al. 1981. Theory of Markedness in Generative Grammar. Pisa. Bybee, Joan. 1985. Morphology. Amsterdam: John Benjamins. Chung, Sandra & Alan Timberlake. 1985. "Tense, Aspect and Mood". Language Typology and Syntactic Description, Vol. 3: Grammatical categories and the lexicon, ed. by T. Shopen, 202-58. Cambridge: University Press.

INVARIANCE AND MARKEDNESS

237

Comrie, Bernard. 1976. Aspect. Cambridge: University Press. —. 1985. Tense. Cambridge: University Press. Dik, Simon. 1978. Functional Grammar. Amsterdam: North-Holland. Eckman, Fred, Edith Moravcsik, & Jessica Wirth. 1986. Markedness. New York: Plenum. García, Erica, this volume. "Grasping the Nettle: Variation as Proof of Invariance". Givón, Talmy. 1979. On Understanding Grammar. New York: Academic Press. Greenberg, Joseph. 1966. Language Universais. The Hague: Mouton. Haiman, John. 1975. "Neutralization and Markedness Assimilation: Future and Subjunctive in Hua". Oceanic Linguistics 14. 119-127. —. 1985a. Natural Syntax: Iconicity and Erosion. Cambridge: University Press. —. 1985b. Introduction. Iconicity in Syntax, ed. by J. Haiman, 1-7. Amsterdam: John Benjamins. Holenstein, Elmar. 1976. Roman Jakobson's Approach to Language. Bloomington: Indiana University Press. Jakobson, Roman. 1932. "Structure of the Russian Verb". In 1984, 1-14. —. 1936. "Contribution to the General Theory of Case: General Meanings of the Russian Cases". In 1984, 57-103. —. 1939. "The Zero Sign". In 1984, 151-160. —. 1957. "Shifters, Verbal Categories, and the Russian Verb". In 1971, 130-147 and 1984, 41-58. —. 1959. "Boas' View of Grammatical Meaning". In 1971, 489-96. —. 1960. "The Gender Pattern of Russian". In 1971, 184-186 and 1984, 141-143. —. 1971. Selected Writings, Vol. II: Word and language. The Hague: Mouton. —. 1972. "Verbal Communication". In 1971, 81-92. —. 1974. "Mark and Feature". In 1985, 122-124. —. 1985. Selected Writings, Vol. VII: Contributions to comparative mythology. Studies in linguistics and philology, 1972-1982. Berlin: Mouton. —. 1987. Selected Writings, Vol. VIII: Major works 1972-1982. Berlin: Mouton. —, Gunnar Fant, & Morris Halle 1952. Preliminaries to Speech Analysis. Revised edition, 1963, Cambridge: MIT Press. Also in Jakobson 1987. — & Morris Halle. 1956. Fundamentals of Language. Second, revised edition, 1971. The Hague: Mouton. — & Krystyna Pomorska. 1980. Dialogues. English Edition, 1983: Cambridge, MA: MIT Press. — & Linda R. Waugh. 1979a. The Sound Shape of Language. Bloomington: Indiana Univer­ sity Press. Second, augmented edition, 1987, Berlin: Mouton de Gruyter. Also in Jakobson 1987. — & —. 1979b. "An Instance of Interconnection between the Distinctive Features". In Jakobson 1985, 59-61. Lyons, John. 1968. Introduction to Theoretical Linguistics. Cambridge: University Press. Matthews, Peter. 1974. Morphology. Cambridge: University Press. Mel'čuk, Igor, this volume. "Toward a Universal Calculus of Inflectional Categories: on Roman Jakobson's Trail". Monville-Burston, Monique, this volume. "The Role of Animacy in Language Change: from Dative to Genitive in Middle Indo-Aryan". & Linda R. Waugh. 1991. Multivalency: The French Historical Present in Journalistic Discourse. Discourse-Pragmatics and the Verb: Evidence from Romance, ed. by S. Fleischman & L. Waugh. London: Routledge. Palmer, F. R. 1986. Mood and Modality. Cambridge: University Press. Robertson, John, this volume. "Invariance in Change: a Case Study from Acatec Maya".

238

MADELEINE NEWFIELD AND LINDA WAUGH

Sangster, Rodney. 1982. Roman Jakobson and Beyond: Language as a system of signs. Berlin: Mouton. . this volume. "Two Types of Markedness and their Implications for the Conceptualization of Grammatical Invariance". Sapir, Edward. 1921. Language. New York: Harcourt, Brace & World. Shapiro, Michael. 1974. "Markedness and Distinctive Feature Hierarchies". Proceedings of the 11th International Congress of Linguists, Vol. 2, ed. by L. Heilman, 775-81. Bologna: il Mulino. . 1976. Asymmetry: an inquiry into the linguistic structures of poetry. Amsterdam: NorthHolland. . 1983. The Sense of Grammar: Language as semeiotic. Bloomington: Indiana University Press. Silverstein, Michael. 1976. "Hierarchy of Features and Ergativity". Grammatical Categories in Australian Languages, ed. by R. Dixon, 112-171. Canberra: Australian Institute of Aborigi­ nal Studies. Stankiewicz, Edward, this volume. "The Concept of Structure in Contemporary Linguistics". Trubetzkoy, Nikolai. 1939. Grundzüge der Phonologic Travaux du cercle linguistique de Prague 7. . 1975. N. S. Trubetzkoy's Letters and Notes. Ed. by R. Jakobson. The Hague: Mouton. Van Schooneveld, Cornells H. 1978. Semantic Transmutations. Bloomington, IN: Physsardt Publications. Viel, Michel. 1984. La notion de "marque" chez Trubetzkoy et Jakobson: Un Episode de l'histoire de la pensée structurale. Lille: Université Lille III. Wallace, Stephen. 1979. "Voice, Mode, or Aspect? The Semantics of Verbal Prefixes in Jakarta Malay". Contributions to Grammatical Studies: Semantics and Syntax, ed. by L. Waugh & F. van Coetsem, 151-178. Leiden: Brill. . 1982. "Figure and Ground: The interrelationships of linguistic categories". Tense-Aspect: between Semantics and Pragmatics, ed. by P. Hopper, 201-223. Waugh, Linda R. 1975. "A Semantic Analysis of the French Tense System". Orbis 24. 436-485. . 1976a. "Lexical Meaning: the prepositions en and dans in French". Lingua 39. 69-118. . 1976b. Roman Jakobsons Science of Language. Lisse: Peter de Ridder Press. . 1976c. "The Semantics and Paradigmatics of Word Order". Language 52.1, 82-107. . 1977. A Semantic Analysis of Word Order. Leiden: Brill. . 1979a. "Markedness and Phonological Systems". LACUS V. Columbia, S.C.: Hornbeam Press. . 1979b. "The Context-Sensitive Meaning of the French Subjunctive". In Contributions to Grammatical Studies: Semantics and Syntax, ed. by L. Waugh & F. van Coetsem, 179-228. Leiden: Brill. . 1982. "Marked and Unmarked: A Choice between Unequals in Semiotic Structure". Semiotica 38. 299-318. . 1987. "Marking Time with the Passé Composé: Toward a Theory of the Perfect". Linguisticae Investigationes XI:1, 1-47. . this volume. "Tense-Aspect and Hierarchy of Meanings: Pragmatic, textual, modal, discourse, expressive, referential". & Madeleine Newfield. ms. "Iconicity and the Morpheme: Toward a model of the lexicon". To appear in Lingua. Whorf, Benjamin L. 1938. "Some Verbal Categories of Hopi". In Whorf 1956, 112-124. . 1945. "Grammatical Categories". In Whorf 1956, 87-101. . 1956. Language, Thought and Reality. Edited by John Carroll. Cambridge, MA: MIT Press. Witkowski, Stanley & Cecil Brown. 1983. "Marking Reversals and Cultural Importance". Language 59. 569-82.

PART THREE GRAMMAR AND DISCOURSE

Tense-Aspect and Hierarchy of Meanings: Pragmatic, Textual, Modal, Discourse, Expressive, Referential1 Linda R. Waugh Cornell University 1. Introduction Many analyses of tense-aspect have focused almost exclusively on its referen­ tial meanings, in particular its capacity to denote deictic or non-deictic tem­ poral qualities of the verbal event.2 Recent work — especially that which centers on discourse analysis — has argued that such an approach is inad­ equate by showing that tense-aspect has discourse functions. However, dis­ course analysts have tended to treat referential and discourse meanings as competing hypotheses. They reject referential meanings in favor of discourse functions (see Weinrich 1973, Reid 1977, 1979). Or they claim that the former are epiphenomenal and can be derived from the latter; Hopper (1979, 1982), for example, maintains that the meaning of complete action associated with perfective verbs derives from their function for foregrounding in narrative discourse. Or they argue that such seemingly contradictory functions are proof that no general, i.e., invariant, meaning can be found which covers all of the uses (e.g., Hopper 1982, Traugott 1982, 1986a, this volume). It is the aim of the present paper to show that the French simple past (SP)3 — which has been one focus of this debate — evidences a wide variety of contextual meanings which can be categorized in the following manner: pragmatic relating to the communication situation in which the SP is found; textual relating to general and generic properties of the text as a whole in which the SP is used; modal relating to the speaker's assessment of the reality of the SP event; discourse relating to specific properties of the discourse in which the SP occurs; expressive relating to the attitude toward, opinion of, reaction to the SP event on the part of the writer; referential relating to denotational properties of the text, including various temporal factors.4 None of these should be disregarded nor treated as derivative.

242

LINDA R. WAUGH

It will be further argued that, far from being discrete and unrelated, these meanings (functions) are interconnected in various ways. On the one hand, they are the specific interpretations which the general meaning of the SP attains in given contexts. On the other hand, they are ordered in a hierarchy. The latter is founded on the structuring principles which were given their first modern formulation by Roman Jakobson (1932, 1936, 1957): invariant vs. contextual meanings and basic vs. marginal meanings. However, there is one major change: whereas Jakobson assumed that invariant meanings are always general and never specific, this is not the case here: certain of the specific meanings of the SP are themselves invariant, while others are quite variable. The present study is based on the data and analyses presented in MonvilleBurston and Waugh (1985) and Waugh and Monville-Burston (1986). Since these papers are widely available, their results will be summarized and their data merely exemplified in order to show the basis for the further conclusions to be discussed here. While previous analyses of the SP have, with few exceptions, concentrated on its literary/historical usage in narratives, the corpus upon which the Waugh/Monville-Burston studies are based came from newspapers and newsmagazines. Such journalistic uses evidence a large variety of genres (narrative and non-narrative) associated with a number of different topics and are, as against belletristic writing, accessible to and read by the general public. They represent some of the most creative and pro­ ductive examples of the SP and show that it is far from being the circum­ scribed category described in the traditional and discourse studies listed above. It was demonstrated in the Waugh/Monville-Burston articles that the SP has a general meaning which is made up of three semantic marks: 5 1. deictic reference to past time (vs. no such reference to past time in Pr); 2. dimensionalization, e.g., delimitation of a figure with clear-cut contours or dimensions (vs. no such dimensionalization, e.g., the ground upon which a figure might rest - this would be one use of Imp); 3. detachment, i.e., separation or dissociation within some universe (vs. no such detachment, e.g., close attachment to the current moment in CP — see Waugh 1987). The latter two of these can be likened to the notion of figure in perception: what is important about a figure as differentiated from a ground is the detachment (separation) within a universe of a dimensionalized entity (delimited unit). Thus, the SP invariantly signals that the verbal event is past and that, in addition, there is a unit which is detached from its context in some way and which at the same time possesses precise dimensions which define it.

TENSE-ASPECT AND HIERARCHY OF MEANINGS

243

2. General Remarks The three features associated with the SP are subdivided into two groups. On the one hand, there is the denotation of past time. In the basic uses in historical and journalistic writing, the past is referentially defined as deictically past with respect to the moment of writing; however, in literary contexts this can be superseded to mean past with respect to a (fictional) moment defined internally to the text itself. This facet of the SP's meaning — which is quite unremarkable and recurs in the other tense-aspects of French as well as of other languages — will not be studied further here, but will return later in the concluding remarks. What will be our focus of attention is the other subgrouping, namely the two correlated features of dimensionalization and detachment. Together they define a hierarchy of interpretations. At one extreme are the pragmatic, textual, and modal functions. All of this information is invariably present, no matter what the context. Moreover, it creates, as it were, an overarching context for the other interpretations of the SP, none of which are invariant. Of these, the discourse meanings (in conjunction, at times, with expressive uses) are particularly important since they occur in a great number of contexts and with a great diversity. Moreover, they often supersede the referential information, which occupies the other extreme of the continuum and which is thus highly variable and optional. It is this continuum which will be the focus in this section.

2.1 Pragmatic, Textual, Modal Meanings The pragmatic, textual, and modal meanings — all of which are invariant — are correlated in some way with the communication situation. Here, we find that detachment is the predominant feature. Thus, for example, the SP is necessarily associated with the written language, not with speaking. More­ over, it invariably signals that there is a separation between the writer and the reader, for it is used only in those types of writing (novels, stories, historical works, tales, legends, poems, and newspaper and magazine art­ icles) which are addressed to whom it may concern, to the public at large, and thus to any and all readers. They speak at a distance. These types of writing are monologues, requiring no oral or written response from the reader, and in which the reader has little active role. 6 In fact, in many cases, a reader may be able to decode the SP but may not be able to produce one (especially in its correct morphology): the SP is, for the majority of the French public, a passive (and late) acquisition. Thus, rather than the arche­ typal speech situation, with the SP there is a writer (rather than a speaker),

244

LINDA R. WAUGH

unknown and passive readers (rather than a known and potentially active hearer), and a unidirectional communication (rather than the possibility of exchange of roles). And, there is often no common spatio-temporal frame: the time and place of writing and the time and place of reading are usually distinct from one another. There is, obligatorily, a detachment, a distancing, a lack of immediacy, within the communication situation when the SP is used. Moreover, the text which results is detached from the writer in the sense that it is a public artifact: meant for the public, it becomes part of the public domain. It thus typically belongs to those genres meant for public consumption rather than those destined for more private reading (diaries, letters, notes, intimate correspondence — see Benveniste 1971, SimoninGrumbach 1975, 1977). In connection with this pragmatic information, the SP also has a modal function: the events it depicts are presented as detached from the author, as independent of the writer's and the reader's subjectivity (see Benveniste 1971, Weinrich 1973). In newspapers, for example, the assumption is that the journalist is merely a relator or narrator of events which occurred elsewhere and independently; thus, their recounting is part of the detached world of reportorial discourse. The journalist, so to speak, vouches for the truth of the events, since they have been verified by him/herself or someone else. Reporting presents itself as objective.7 The distancing of the text from the writer has other, textual repercussions: the text itself is presented as an integral, dimensionalized whole which has an internal consistency and cohesion of its own. It is of known length, with a beginning, a middle, and an end. It is a clearly delimited unit, to be read and understood as such. Moreover, it is a carefully prepared text, planned in advance, belonging to the formal register of language, in which the more colloquial lexical, grammatical, and syntactic constructions are avoided, and such elements as hesitations, repetitions, reformulations, and corrections do not exist. There is a necessary and inherent association between the use of the SP and special care taken in the formulation of a text which has been carefully elaborated and reflected upon. Sometimes, this elaboration and reflection may go so far as to become a studied elegance, an affectation, a preciousness in the style. Thus, the SP always belongs to a special and particular type of text, stylistically separated from other texts. Moreover, the text is easily characterizable as to genre: the SP is always associated in some way with narration. Thus, it is used when narrative plays either the dominant or a very important role (e.g., novels, stories, legends, tales, historical works) or, at least, is one part of the overall thrust (e.g., newspaper and magazine articles). 8 In the former case, the SP tends to occur in groups, to recur throughout a text, to be contrasted with the Imp only, and to be associated, at least in its basic uses, with the narrative

TENSE-ASPECT AND HIERARCHY OF MEANINGS

245

line — in the sense of a recounting of separate and sequenced events in the order they occurred (see Labov and Waletzky 1967, Reinhart 1985). This, in fact, is a general characterization of most literary-historical and some journalistic usage and is the basis for the claims in discourse studies that the SP represents a figure ( = dimensionalization plus detachment). 9 But, in newspapers, while a given article may contain truly narrative sections, the SP is not necessarily, or only, used for the narrative line. Other tense-aspects (e.g., the CP or the HPr) may constitute the narrative, while the SP occurs elsewhere. And, when so used, it often is in isolation: a given SP may be the only one of an article or may be widely separated from — and thus not directly related to — any other SP that may be used. In other cases, there may be a group of (two or more) SP's which occur in succession. They form either a narrative or some sort of parallelism which binds them together versus the surrounding text. This formal dispo­ sition of the SP in an article — and its semantic counterpart (to be discussed below) — is itself relatable to the dimensionalization and detach­ ment characteristic of it.

2.2 Discourse and Expressive Meanings The specific instantiations of detachment and dimensionalization just dis­ cussed are always accompanied by others, the most important of which are correlated with text-internal, discourse factors. They are in conformity with the fact that the text in which the SP is used is an integral whole, possessing a general architecture and an internal structure and coherence. A general characterization of its discourse functions is the following: in cases where only the SP and the Imp are used for the narrative and where the SP is repeated throughout the text, one quite frequent result is narrative fore­ grounding vs. backgrounding. 10 This is shown by example #1: (1)

La Broche de Sarah [entire article]11 Sarah Bernhardt collectionnait les admirateurs. L'un d'entre eux, particuli­ èrement généreux, COMMANDA chez le joaillier Boucheron une broche noeud, en acier repercé garni de brillants. L'actrice la PERDIT et s'en DESOLA tant que l'amoureux (l'on ne peut toujours pas révéler son nom, il a exigé le secret à tout jamais), FIT refaire une broche jumelle. Entretemps, l'habilleuse de Sarah Bernhardt retrouvait le bijou dans sa loge. Frédéric Boucheron, fondateur en 1858 de la bijouterie qui porte toujours son nom, DECIDA de conserver la seconde broche pour sa collection personnelle. (L'Express, 2/13-19/84).

[Sarah Bernhardt collected admirers. One of them, who was particularly gen­ erous, ORDERED at the Boucheron jewelers a brooch in the form of a knot, of

246

LINDA R. WAUGH

pierced steel garnished with brilliants. The actress LOST it and WAS SO UPSET by that that her lover (his name still can't be revealed, he insisted on its remaining a secret forever), HAD a duplicate brooch made. Meanwhile, Sarah Bernahrdt's wardrobe woman found the piece in her dressing room. Frederic Boucheron, the founder in 1858 of the jewelry store which still carries his name, DECIDED to keep the second brooch for his private collection.] However, when the SP occurs in isolation, or when there are other past tense-aspects (especially when three or four of them are found in a narra­ tive), or when the context is not a narrative one, the SP manifests discourse uses which are not examples of foregrounding, but are correlated with the more general notions of dimensionalization and detachment. These dis­ course meanings are related either in global terms to the overall configur­ ation of the text — in which case detachment plays the greater role — or in more local terms to the linkage of a given clause or sentence with neighboring ones. In the latter case, dimensionalization plays the greater role. While both meanings are present in all contexts, their hierarchization is evident. 2.2.1 Global Discourse Meanings With regard to the more global factors, the SP helps to bring out the configuration (or construction) of the text by marking either its ultimate limits (beginning or end), or its internal subdiv­ isions, or in more pointed manner, particular steps in its logical construction. Thus, it marks the transition from one part of the text to another. In this demarcative 12 function, the SP very often occurs in isolation and in many cases is the only one used in an article; it is always in contrast with some other basic past tense-aspect (the Imp plus the HPr and/or the CP) in the context. Furthermore, this function is usually shared with and reinforced by other elements which also signify a change (or even a rupture) of some kind in the text: formal markers (e.g., titles, subtitles, indentation for new paragraphs, signature of the journalist), punctuation marks (e.g., commas, colons), and particular lexical items (e.g., coordinating conjunctions express­ ing logical relations). It is these other elements which signal the particular interpretation to be given to the SP, for it seems in this use to be bleached of any specific semantic value. As such, it relates to the more formal, constructional characteristics of the text. And in some cases, its use is almost redundant, for the contextual supporting elements are often enough to convey the intended meaning. What saves the SP from being simply a redundant signal in these cases is the fact that the contextual supports are sometimes absent; in addition, dimensionalization often contributes a sense of highlighting (about which, more below). Thus, an isolated SP may either begin or end an article or a section of an article. In addition, it may sometimes call attention to the ideas expressed

TENSE-ASPECT AND HIERARCHY OF MEANINGS

247

at these points and help the reader follow the general development of the text. Thus, a beginning SP might serve as an introduction which gives a global characterization of the theme or it may simply set the tone. In many cases, however, it often seems as if the writer wants to provoke a sense of surprise (cf. Imbs 1960: 83) and plunge the reader into the text without any orientation. This is shown by example #2a: (2a)

L'événement: La cassette "guerre des étoiles" [first sentences of article only] Ce FUT un événement cinématographique, ce sera un événement vidéo. La Guerre des Etoiles arrive sur cassettes ... (La République du Centre, 4/14/84) [The event: The "star wars" cassette It WAS a cinematographic event, it will be a video event. Star Wars is coming on cassette tapes ...]

In their turn, SPs of conclusion may not only mark the end of a development, but also show its particular importance as: the lesson to be learned from what was said before, the moral of the article, an affirmation in the form of a resume, the revelation of important information, and so forth. Such a conclusion is often a way of condensing the action down to a uniform entity, thereby placing it at a certain distance, as shown by example #2b: (2b)

Jean-Marie le Pen, Dieu et l'école [last sentence of the article only] ... Ainsi PARLA, pendant près de deux heures, saint Jean-Marie Bouche d'Or chantre des "Français d'abord", qui appelle au "réveil" de ses concitoyens "contre les forces de mort et d'obscurantisme" parce qu'il souhaite que sa France demeure un "pays missionné". (A. R., Le Monde, 4/27/84) [Jean-Marie le Pen, God and school... Thus SPAKE, for almost two hours, Saint Jean-Marie Golden Mouth, cantor of the "French first", who calls on his fellow citizens to be "awakened against the forces of death and obscurantism" because he wishes his France to stay a "country with a mission".]

Not only the formal subdivisions but also the logical construction of the argumentation of a text may be given by the SP. Thus, one isolated SP (or at most two SPs), often accompanied by coordinating conjunctions and expressions pertaining to logical relations, may mark a stage in the logical development of the text. The switch from some other tense-aspect to the SP may be used either to define the essential logical divisions of the text as a whole, or, at the level of the sentence, to separate two verbs in the same or contiguous sentences from one another and show that they form a coherent logical progression. The particular logical connections are varied; the ones which occur the most could be gathered under the general rubric of causeeffect or point of departure-result, as in example #3:

248

LINDA R. WAUGH

(3)

Le tournoi de Bruxelles La nouvelle arme de McEnroe ... Au début du second set, McEnroe a estimé que le spectacle offert au public était de piètre qualité, et il APOSTROPHA vivement Ostoga... (A. G., Le Monde, 3/13/84). [The Brussels tournament McEnroe's new weapon ... At the beginning of the second set, McEnroe was of the opinion that the spectacle offered to the public was of poor quality, and so he UPBRAIDED Ostoga...]

Also present in the corpus are uses where the SP marks a particular moment when the reasoning is advanced by a new element in the argumentation or a change of perspective, shows the opposition between two points or two facts in an argumentation, or indicates the logical transition from general statement to an explanation for, proof of, or particular example of that general statement. 13

2.2.2 Local discourse meanings In addition to its use for marking the more global formal and logical articulations of a text, the SP is also correlated with more local properties. In such cases, it is dimensionalization which comes to the fore and which evidences a great variety of contextualization: a continuum ranging from greater or lesser foregrounding (highlighting) to greater or lesser backgrounding (downplaying) including a middle or neutral position. 14 What binds all of these together is their precision: their focus on some precisely defined element. If we start with foregrounding, the first use to be noted is that alluded to above: namely, foregrounding of the plot-line or the major, sequenced events in narratives, in contrast with the backgrounded Imp's (see example #1 above). However, the use of the SP for foregrounding is not confined to narrative contexts nor to use in groups and is not necessarily allied to the notions of narrative plot line, major events, sequencing of events, and so forth. Moreover, foregrounding may be of different sorts and of different degrees. Thus, the SP may be used to put strong emphasis on some special point (e.g., the major point of an article), but it also may put a more moderate stress on interesting facts to which the journalist wants to draw the attention of the reader, or it may focus temporarily on details judged to be worthy of interest for the moment, and so forth. Examples of the stronger focus can be found in those openings, closings, consequences, new elements, explanations, exemplifications discussed above which are dramatic in nature (examples #2a and 2b above). Lesser emphasis is evidenced by those uses of the SP which focus on facts which are only of local and

TENSE-ASPECT AND HIERARCHY OF MEANINGS

249

temporary interest in the development of an article or only moderately important in relation to the major theme, as in example #4: (4)

"Je suis opposé à toute intervention étrangère en Amérique centrale" nous déclare le nouveau président, M. Lusinchi

....Le pétrole a sans doute été un atout majeur pour le Vénézuéla. Mais en même temps, ce FUT une "malédiction", selon l'expression de Juan Pablo Perez Alfonso, l'un des pères de ΓΟΡΕΡ (Jean-Claude Buhrer, Le Monde, 3/2/84) ["I am against any foreign intervention in Central America" the new president, M. Lusinchi, tells us ....Oil has without a doubt been a major trump card for Venezuela. But at the same time, it HAS BEEN a "curse", according to Juan Pablo Perez Alfonso, one of the fathers of OPEC....] In all such examples, there is typically a switch from some other past tense (CP or HPr) to an isolated SP (or vice versa), reinforced by syntactic and lexical expressions showing mise-en-relief. In some cases, emphasis and focus on facts judged to be important either in and of themselves or at a particular moment in the text, are equivalent to simply furnishing a detail. Thus, on the continuum of dimensionalization, we are far from pure foregrounding and closer to notions like precision and specificity. It has often been remarked that the verbal process given by the SP has a beginning and/or an ending which is defined in the context and that the SP is, thus, often accompanied by indications of the date, the duration, the location of the action. There is no uncertainty about its spatiotemporal context nor about who performed it. However, more generally, the SP is used there in the discourse where the journalist draws the attention of the reader to a specific point. Thus the switch from, for example, a CP to a SP is accompanied by a narrowing of the focus onto one very particular aspect of the situation described. The impression created is comparable to the "zooming-in" of filmmakers. One specific example of this technique could be supplied from the logical uses mentioned above: introducing the illustration of an argument, bringing proof for an assertion, furnishing an example. All of these are a means of demonstrating with particular and specific facts the truth of affirmations which are general, vague, or abstract. There are other uses where the zooming-in simply marks the transition from something larger to something smaller. But there is a further variant of this: the passage from the exposition of a theme (given by the CP or the HPr) to its development in all its details and specifications (given by the SP), as shown by example #5: (5)

Un mort et trois blessés dans une collision [entire article] Lundi matin, vers 4 h 45, une voiture conduite par M.Jean-Marie

250

LINDA R. WAUGH

Chagourin, 23 ans, soudeur, domicilié au lieu-dit "Les Rethures", à Vimory, près de Montargis où il se rendait, a brusquement quitté la route à hauteur de "La Gacherie" à Gien pour aller percuter un arbre. Le vehic­ ule FUTRENVOYEau milieu de la chaussée et une automobile qui le suivait VINT s'écraser sur lui. Elle était conduite par M. François Maugard, 26 ans, élève inspecteur, habitant La Peyrousse-Fossat (Haute Garonne) qui avait à son bord Mlle M arie-Louise de Prado, 28 ans, élève inspecteur, 2, rue des Lilas, à Toulouse et Didier Galigne, 28 ans, élève inspecteur, 49, rue des Cigognes, à Ramonville-Saint Ange (Haute Garonne). Au cours de la collision, M. Chagourin a été tué et les trois autres blessés. [One dead and three hurt in a collision Monday morning, about 4:45 a.m., a car driven by Mr. Jean-Marie Chagourin, 23 years old, a welder, resident at the place called "Les Rethures", at Vimory, near Montargis where he was going, suddenly left the road opposite "La Gacherie" at Gien and crashed into a tree. The vehicle WAS THROWN BACK into the middle of the roadway and a car which was following it CAME and smashed against it. The car was driven by Mr. François Maugard, 26 years old, student inspector, living at La Peyrousse-Fossat (Haute Garonne) who hadwith him Miss MarieLouise de Prado, 28 years old, student inspector, 2 rue des Lilas, Toulouse and Didier Galigne, 28 years old, student inspector, 49 rue des Cigognes, Ramonville-Saint Ange (Haute Garonne). In the collision, Mr. Chagourin was killed and the three others hurt.] There is a striking difference between this usage (the CP to open and sometimes to close a text, the SP for internal development and details) and the demarcative use noted above (the SP for opening or closing the text, other past tense-aspects for the main body of the article). The introductions in the SP are relatively short, often laconic, well constructed, written in an attention-catching style. Often, there seems to be a wish to dramatize the situation about which they give little or no real information. They create a kind of tension which forces the reader to progress further in the article in order to understand better. On the other hand, the introductions in the CP are stylistically more neutral: they don't contain the emphatic or marked constructions often found in their SP counterparts and often give a first, rather sketchy description of what the situation was (to be filled in after with details provided in the SP). As to the conclusions, those in the SP tend to gather the text on itself; they close it but at the same time oblige the reader to understand the ending in terms of what went before and to consider the article as a whole. Those with other past tense-aspects are, on the other hand, more prospective in nature: they open the text toward the future, connect it with its present consequences, consider the extensions of the action. They treat the article less as a firmly established whole.

TENSE-ASPECT AND HIERARCHY OF MEANINGS

251

Because of its ability to be easily used in utterances meant to express precision and specificity, the SP is comparable to figures which are sharply and clearly defined, capable of being perceived in a single glance. This means that they may be of limited size.15 Precision, thus, may be coupled with smallness; and just as the notion of minuteness suggests both scrupulous accuracy and negligeable importance and just as that which is small can be unnoticed, so the SP has, in certain contexts, uses where precision and specificity are allied with lesser significance and secondariness. and even inconsequentiaîity. Thus, it often contributes to a sense of backgrounding or downplaying. This backgrounding, just as for foregrounding, is mani­ fested in a variety of ways, e.g., in relative clauses and in brief16 parenthetical remarks of various sorts, as shown by example #6: (6)

Acculés au désespoir, deux commerçants décident de se donner la mort ...Les le Linge avaient bien prémédité leur disparition qu'ils avaient réglée dans les moindres détails. Ils avaient addressé des lettres à ceux qu'ils allaient laisser. A leur voisin des pompes funèbres aussi, qui TROUVA hier leur courrier dans sa boîte postale de Sully, Ils le chargaient d'organiser leurs funérailles....(La République du Centre, 3/2/84) [Driven against the wall by despair, two shopkeepers decide io take their own lives ...The Le Linges had carefully premeditated their disappearance, which they had planned in the smallest details. They had sent letters to those they were going to leave. And even to their neighbor who works for a funeral parlor and who FOUND yesterday their letter in his mailbox in Sully. They instructed him to organize their funeral...]

In addition, it can be reinforced by lexical items showing that a secondary aspect of the theme is meant and by the type of punctuation (dashes, commas, parentheses) that accompanies secondary material and iconically recalls the dimensionalization of the SP by surrounding the clause in which it is used. While the cline from foregrounding to precision to backgrounding is represented in the corpus, there are many more examples of foregrounding and precision than of backgrounding, and in general it is foregrounding of various degrees that is best represented in journalistic discourse. To sum up, on the one hand the SP can aid in the linear development of the text: it can be used to delimit various types of sections. In this demarcative function, related to global discourse factors of wider scope, it contrib­ utes first and foremost to the more formal structure of the text. It is here that detachment seems no dominate and thus, there is less specific semantic content. On the other hand, the SP can contribute to the dense structure of a text by establishing in it levels of different depth: foreground of narration vs. background of description and commentary; level of the general vs. level

252

LINDA R. WAUGH

of the particular; surface of the text vs. projected elements (in high or low relief) which stand out; basic configuration of the text in the recesses of which one finds subordinated elements; dominant line of development vs. secondary or even subsidiary shunted elements. In these more local discourse uses of narrower scope, it is dimensionalization which often seems to domi­ nate, thus lending more specific semantic content. 2.2.3 Expressive meaning In addition to this discourse information, the SP also has an expressive function. It was said earlier that it gives modal information, since it presents the verbal event as independent of the writer's subjectivity and thus independently verifiable. While this is true, the journal­ ist's own reaction to or opinion of the events described may also be evident. It is the journalist, after all, who decides what to foreground and what not to in an article, which lexical items and syntactic means to use, how to characterize the participants and the events, and so forth. Thus, the reporter's opinion may be made known, directly, and indirectly, and the SP may be associated with expressive information which may go so far as to convey commentary about the events themselves. Thus, there is a complex interplay here: the invariant, modal information associated with objective detachment may be accompanied by an overlay of reaction and opinion (as in example #2b above). 17

2.3 Referential Meanings Lastly, there are referential meanings: these sometimes may occur alone (without discourse or expressive information), but at other times they may accompany that information. And they may also be overridden: that is, discourse or expressive information may cancel one or more of these referen­ tial properties. The SP may be used to denote an event which is seen as an integrated whole, closed on itself, indivisible, impenetrable, an event which is complete (its aspectual meaning — see Sten 1952, Imbs 1960, Martin 1971, Comrie 1976). Furthermore, as said above, it is often precisely defined as to place, time, participants: and the temporal, spatial, and pronominal terms with which it occurs in these uses are those which are not deictically defined with respect to time, place, and participants of the speech event. Rather, reference tends to be defined in extralinguistic, absolute terms (e.g., au printemps 1983 (26 mars)) or intratextually/anaphorically (e.g., JeanMarie...qui...il ...sa France, from example #2b). Referentially, the text is detached as much as possible from the communication situation. In addition, the SP may give temporal remoteness - chronological detachment from the communication situation or, perhaps as a metaphorical extension, psycho-

TENSE-ASPECT AND HIERARCHY OF MEANINGS

253

logical distance. In such cases, the CP (in conformity with the present tense of its auxiliary) may give events which are attached to the present moment or are chronologically closer to it or of present relevance. Furthermore, the SP is often used to relate the life of someone who is dead (as in a necrology), whereas the CP may be used for someone who is still alive. Lastly, the SP is also, typically, not found in quoted speech; the CP is its counterpart in such cases (for further discussion of the difference between the SP and the CP, see Waugh 1987, 1991.). In discourse usage, each of these referential meanings can be canceled: the event may be incomplete or completeness may not be at issue (example #4); the time, place, participants of the event may be deictically defined (example #6); the event may not be remote either temporally or psychologi­ cally (examples #4 and 6); the person referred to may not be dead (example #2b); and the SP can be used in quoted speech even when the original contained no SP. 2.4 Further Remarks If we focus further on the pragmatic, modal, and textual information associ­ ated with the SP, we find that they are related to its paradigmatic differen­ tiation from the other tense-aspects of French. No other one (except perhaps the compound counterpart of the SP, the anterior past) invariably conveys information of this sort. Conversely, if we focus on the more highly variable discourse, expressive, and referential interpretations of the SP, what is striking is their high syntagmatic contextualization. On the one hand, SP is eminently contrastive; 18 it contrasts first and foremost with the other past tense-aspects (including other SPs) used in the sentence (or clause) which precede or follow it and through this contrast establishes differences within the text itself. These differences are highly discourse related: it is necessary to know what tense-aspect was used before or will be used after and where in the text one is in order to determine the basis on which the contrast is being made. Such contrasts are obligatory: the SP used without another tense-aspect in the vicinity is difficult (or impossible) to interpret. 19 On the other hand, the SP is often reinforced by syntactic, formal, and lexical facets of its context. 20 As said above, these uses of the SP are close to redundancy. What information it provides is also often given in the context. 3. By Way of Conclusion: Diachronic Remarks The synchronic hierarchy of information defined here — first, pragmatic, textual and modal; second, discourse and expressive; third, referential —

254

LINDA R. WAUGH

also has its diachronic analog. The denotation of past time in addition to the referential and discourse interpretations of dimensionalization (in par­ ticular, those values generally associated with perfectivity and narration — complete, foregrounded events — again example #1), are already present in the earliest texts we have in the Old French period (11th to 13th centuries). It is also these uses which are, in the modern language, still consciously relevant: asked to give a typical example of the SP, most native readers proffer a narrative one (cf. the focus in discourse analyses on just these uses). Since the time of Old French, the SP has acquired the new feature of detachment, which, in correlation with dimensionalization, has led to new functions, in particular those which rely on the interplay of the two features: i.e., certain referential uses (e.g., remoteness), new discourse meanings (e.g., the demarcative ones) as well as the pragmatic, textual, and modal functions. These have been developing since the time when, in Middle French, the CP, which before had been a present perfect (e.g., j'ai mangé, je n'ai plus faim 'T have eaten, I'm no longer hungry"), became in addition a preterit and could be used to refer directly to past time (e.g., j'ai mangé une pomme à deux heures 'T ate an apple at two o'clock"). It thus began to encroach upon the territory of the SP and a competition was established between the two (see Waugh 1987). The result of this competition has been, as Kuryłowicz (1965) predicted, that the newer form, the CP, has expanded its terrain (in spoken French in particular), taking on the more general meaning (present perfect and past perfective).21 Correspondingly, the older form, the SP, has become more specialized (marked — see Waugh 1982) and has gradually taken on many of the characteristics which are typical of a grammatical category which may be disappearing: drop in frequency, restriction to specific kinds of texts, association with certain verbs of high frequency (such as être "be"), use more and more in isolation, further and further contrastive and redundant contextualization. But, in terms of contextual variants, it has shown a great versatility by developing a high semantic density. Moreover, this develop­ ment has resulted, as has been found in the evolution of other semantic categories (see Traugott 1982: 256, 1986a, this volume), in the elaboration of further meanings based on factors having to do with the communication situation, rather than on more objective factors. Thus, further discourse meanings — an extension, perhaps, of its primordial function of fore­ grounding — have been created, in addition to pragmatic, textual, and modal functions. And even in the referential domain, the meanings are more subjectively construed: witness the notion of remoteness, especially in its psychological interpretation. 22 We have here an example of the diachronic extension of contextual variants into predictable domains on the basis of general meaning.

TENSE-ASPECT AND HIERARCHY OF MEANINGS

255

What is striking, moreover, is that it is dimensionalization and detachment which have shown this evolution, not denotation of past time, the textual variant of which belongs to the oldest written Old French documents and is subordinated to its referential use. This is due to the fact that the deno­ tation of time, especially past and present, is the dominant feature of this part of the French verbal system: all other facets are subordinated to it. Only the future tense-aspect predictably evidences serious competition between temporality and modality (see Fleischman 1982a, b, 1983). In the SP, reference to the past remains as a base line on which these other, competing functions are an overlay. Thus, the history of the SP is one in which three interrelated factors are relevant: little change in one feature (denotation of past time), the addition of a new feature (detachment) which combines with a preexisting one (dimensionalization), and the expansion of these latter two into new domains roughly characterizable as being non-referential and associated more with the communication situation in some way. The change in the SP has thus occurred jointly both to its invariant general meaning and to its invariant and variant specific meanings. It is thus further testimony to the interplay of invariance and variation which is characteristic of all grammatical categories. Notes 1. I would like to thank Roger Wales and the Department of Psychology at the University of Melbourne, as well as Bliss Carnochan and Morton Sosna and the Humanities Center at Stanford University for the hospitality they afforded me while I was writing this paper. 2. This is as true of traditional analyses as it is of studies based on propositional or Reichenbachian semantics of some sort. See e.g., Sten 1952, Imbs 1960, Klum 1961, Reichenbach 1966, Schogt 1968, Martin 1971, Comrie 1976, 1985 for a general discussion, and Guenther, Hoepelman and Rohrer 1978, Vet 1980, Kamp and Rohrer 1983, Dahl 1984, De Both-Diez 1985, Vet and Molendijk 1986 for focus on the French past tenseaspects. 3. The following abbreviations will be used in this paper: SP for simple past, CP for compound past, Imp for imperfect, Pr for present, HPr for historical present. 4. Given the diversity of ways in which the terms pragmatic, textual, modal, discourse, expressive, and referential are used in the grammatical literature, they should be inter­ preted for the present paper only in terms of the definitions given here. 5. For a definition of mark, see the contribution of Newfield and Waugh in this volume. 6. In those very special cases where the SP is spoken — for example, in oratory style (e.g., in certain types of political speeches like the address of the President to the nation, in funeral orations, in discourses at solemn occasions and the like) and in newscasts on the radio and TV — there is a written or a memorized text which is read aloud with declamatory intonation, presented as a monologue to an audience of unknown hearers obligatorily distant from the speaker and without any possibility of interrupting or answering.

256 7.

8. 9.

10.

11.

12.

13.

14. 15.

16. 17.

18. 19.

LINDA R. WAUGH In the fictional use of the SP, that it is fictional discourse is signaled by other textual means, not by the SP alone. The fictional world is, however, set up as detached from the writer (and the narrator) and the events are presented as true of this world. That there may be one or more narrators separate from the writer in fiction is another manifestation of the detachment of the SP. In this latter case, narration may be synecdochically or metonymically related to the general theme of the article. Much more work needs to be done on fictional narrative, especially on more complex literary works than heretofore, before we can fully characterize this use of the SP (see in this light Saunders 1969, Waugh 1987). That is, the SP gives the plot-line, the major or central sequential events of the narrative, and is used with punctual, dynamic, or kinetic verbs, while the Imp gives supportive or subsidiary background information such as description, characterization, commentary, or unsequenced events, and is often associated with stative or durative verbs (see Weinrich 1973, Reid 1977, 1979, Hopper 1979, 1982a, and Wallace 1982). In the following examples (taken from Waugh and Monville-Burston 1986), all finite verbs are in italics in the French original and in their English translations (provided by the present author and Monique Monville-Burston). Any SP, and its equivalent in English, are in capitals. These uses of the SP are reminiscent of demarcative/delimitative signals — also called, in the phonology of certain languages, border or boundary signals — which serve to mark the limits of various units (morphemes, words, phrases, clauses, sentences, para­ graphs, etc.- see Jakobson and Waugh 1979). Such uses may be related in general to the narrative SP. There is, in all these cases, a change of some sort: in narrative, the rupture is contextually evidenced as separate, temporally sequenced events, whereas here it is rather (perhaps metaphorically related) the logical separation of one discrete step in the argumentation from another. The SP of conclusion and that of consequence are also closely related to each other as contextual variants — what differs is the type of context in which they occur. Many discourse analysts assume that there are different levels of grounding (from fore­ grounding to backgrounding): e.g., Grimes 1975, Jones and Jones 1979, Fleischman 1985. Perceptual, especially Gestalt, psychologists remark that the optimal figures are (among other things) closed, structured, small. See Zusne 1970: 113-124, Rock 1975: 253-263, Wallace 1982: 214, and Reinhart 1985. This propensity of the SP to give small figures may be associated iconically with its form, which is small and compact in comparison with the Imp and especially the CP: fit vs. faisait and a fait, fut vs. était and a été, parla vs. parlait and a parlé,and so forth. Their brevity may also be iconic: that which is minor in importance is also restricted in length. Weinrich (1973), thus, is wrong in insisting that the writer's attitude is always one of detachment: first, the SP gives information about how the events are presented, not about how the writer feels; and second, the expressive information conveyed may well indicate the writer's opinion/reaction. This contrast itself may be a result of the detachment ( = separation from the context) of a unit which retains its own, particular integrity ( = dimensionalization). Thus, for example, when it occurs in the vicinity of Imp's for description, it takes on a punctual interpretation; used after a CP for setting the scene, it serves for the development; when in a narrative given in the HPr or the CP, it may open the text, close it, give some interesting detail, or make a parenthetical remark; where a different tense-aspect is used to give the cause of an event, the transition to the SP gives the consequence; and so forth.

TENSE-ASPECT AND HIERARCHY OF MEANINGS

257

20.

Marked syntactic constructions (inversions, dislocations, parallelisms, verb-subject word order, and so forth), conjunctions expressing a logical relation, emphatic vocabulary, the material disposition of the text (typography, punctuation, blank spaces, differentiation between titles, subtitles, text, and author's name) — all of these help to underline the contrastive value of the SP and to show its relation to the construction of the text. 21. Furthermore, the Imp as well has been enlarging its domain, most notably since the 19th century with the addition of narrative uses which had originally been the province of the SP and some of which, interestingly, have to do with opening and closing texts {imparfait d'ouverture, de clôture). 22. The attempts by grammarians in the 16th century to establish a 24-hour rule for the remotness of the SP can thus be seen as doomed since this more objective factor was necessarily to lose out to more subjective ones, as well as to non-referential ones (cf. Dahl 1984).

References Benveniste, Emile. 1971. "Correlations of Tense in the French Verb". Problems of General Linguistics, 153-61. Coral Gables: University of Miami Press. Comrie, Bernard. 1976. Aspect. Cambridge: Cambridge University Press. . 1985. Tense. Cambridge: Cambridge University Press. Dahl, Osten. 1984. "Temporal Distance: Remoteness Distinctions in Tense-Aspect Systems". Explanations f or Language Universals, ed. by . Butterworth et al, 105-122. Berlin:Mouton. De Both-Diez, Α.-Μ. 1985. "L'aspect et ses implications dans le fonctionnement de l'imparfait, du passé simple, et du passé composé au niveau textuel". Langue française 67. 5-22. Fleischman, Suzanne. 1982a. The Future in Thought and Language. Cambridge: Cambridge University Press. . 1982b. "The Past and the Future: Are they Coming or GoingV BLS 8. 332-34. . 1983. "From Pragmatics to Grammar: Diachronic reflections on complex pasts and futures in Romance". Lingua 60. 183-214. . 1985. "Discourse Functions of Tense-Aspect Oppositions in Narrative: Toward a Theory of Grounding". Linguistics 23. 851-82. Grimes, Joseph. 1975. The Thread of Discourse. The Hague: Mouton. Guenther, F., J. Hoepelman, and  Rohrer. 1978. "A Note on the Passé Simple". Papers on Tense, Aspect, and Verb Classification, ed. by C. Rohrer, 11-36. Tubingen: Gunter Narr Verlag. Hopper, Paul. 1979. "Aspect and Foregrounding". Discourse and Syntax: Syntax and semantics 12, ed. by T. Givón, 213-41. New York: Academic Press. . 1982. "Aspect between Discourse and Grammar: An introductory essay for the volume". Tense Aspect: Between semantics and pragmatics, ed. by P. Hopper, 3-19. Amsterdam: John Benjamins. Imbs, Paul. 1960. L'emploi des temps verbaux en français moderne. Paris: Klincksieck. Jakobson, Roman. 1932. "Structure of the Russian Verb". Russian and Slavic Grammar: Studies 1931-1981, ed. by L. Waugh and M. Halle, 1-14. Berlin: Mouton. . 1936. "Contribution to the General Theory of Case: General meanings of the Russian cases". Russian and Slavic Grammar: Studies 1931-1981, ed. by L. Waugh and M. Halle, 59-104. Berlin: Mouton. . 1957. "Shifters, Verbal Categories, and the Russian Verb". Selected Writings II, 130-147. The Hague: Mouton. Also in Russian and Slavic Grammar: Studies 1931-1981, ed. by L. Waugh and M. Halle, 41-58.

258

LINDA R. WAUGH

and Linda R. Waugh. 1979. The Sound Shape of Language. Bloomington, Ind: Indiana University Press. Second ed., 1987, Berlin: W. de Gruyter. Jones, Larry B. and Linda K. Jones. 1979. "Multiple Levels of Information in Discourse". Discourse Studies in Meso-American Languages, vol. 1 (Discussion), ed. by Linda K. Jones, 3-28. Dallas, Texas: Summer Institute of Linguistics. Kamp, Hans and Christian Rohrer. 1983. "Tense in Texts". In R. Bäurle, . Schwartze, and A. von Stechow (eds.), Meaning, Use and Interpretation of Language, 250-69. Berlin: Walter de Gruyter. Klum, Arne. 1961. Verbe et adverbe. Stockholm: Almqvist and Wiksell. Kuryłowicz, Jerzy. 1966. "La nature des procès dits 'analogiques'". Readings in Linguistics II, ed. by Eric Hamp et al, 158-174. Chicago: University of Chicago Press. Labov, William & J. Waletzky. 1967. "Narrative Analysis: Oral versions of personal experience". Essays on the Verbal and Visual Arts, ed. by J. Helm, 12-44. Seattle: University of Washington Press. Martin, Robert. 1971. Temps et aspect: essai sur l'emploi des temps narratifs en moyen français. Paris: Klincksieck. Monville-Burston, Monique et Linda R. Waugh. 1985. "Le passé simple dans le discours journalistique". Lingua 67.121-70. Reichenbach, Hans. 1966. Elements of Symbolic Logic. NY: Free Press. Reid, Wallis. 1977. "The Quantitative Validation of a Grammatical Hypothesis: The passé simple and the imparfait". Columbia University Working Papers in Linguistics 4.59-77. (also in The North Eastern Linguistic Society 7. 315-33). . 1979. The Human Factor in Grammatical Analysis: Passé simple and imperfect. Ph. D. Dissertation. Columbia University. Reinhart, Tanya. 1985. "Principles of Gestalt Perception in the Temporal Organization of Narrative Texts". Linguistics 22:6. Rock, Irvin. 1975. An Introduction to Perception. New York: Macmillan. Saunders, H. 1969. "The Evolution of French Narrative Tenses". Forum for Modern Language Studies 5. 141-161. Schogt, Henry. 1968. Le Système verbal du français contemporain. The Hague: Mouton. Simonin-Grumbach, Jenny. 1975. "Pour une typologie des discours". Langue, discours, société, ed. by Julia Kristeva et al, 85-121. Paris: Editions du Seuil. . 1977. "Linguistique textuelle et l'étude des textes littéraires: à propos de Le Temps de H. Weinrich". Pratiques 13. 77-90. Sten, Holgar. 1952. Les temps du verbe fini (indicatif) en français moderne. Kobenhavn: Munksgaard. Traugott, Elizabeth. 1982. "From Propositional to Textual and Expressive Meanings: Some semantic-pragmatic aspects of grammaticalization". Perspectives on Historical Linguistics, ed. by Winfred Lehman and Yakov Malkiel, 245-272. Amsterdam: John Benjamins. . 1986a. "From Polysemy to Internal Semantic Reconstruction". BLS 12. . this volume. "Speech Act Verbs: A Historical Perspective". Vet, Co. 1980. Temps, aspect, et adverbes de temps en français contemporain. Genève: Librairie Droz. and Arie Molendijk. 1986. "The Discourse Functions of the Past Tenses of French". Temporal Structures in Sentence and Discourse ed. by V. Lo Cascio and C. Vet, 133-160. Dordrecht: Foris. Wallace, Stephen. 1982. "Figure and Ground: The interrelationships of linguistic categories". Tense Aspect: Between semantics and pragmatics, ed. by P. Hopper, 201-223. Amsterdam: John Benjamins. Waugh, Linda R. 1982. "Marked and Unmarked — A Choice between Unequals in Semiotic Structure". Semiotica 38:3/4. 299-318.

TENSE-ASPECT AND HIERARCHY OF MEANINGS

259

. 1987. "Marking Time with the Passé Composé: Toward a theory of the perfect". Linguisticae Investigationes XI:1. 1-48. . 1991. "Discourse Functions of Tense-Aspect in French: Dynamic Synchrony". In Verbal Aspect in Discourse, ed. by N. Thelin, 159-187. Amsterdam: John Benjamins. and Monique Monville-Burston. 1986. "Aspect and Discourse Function: the French simple past in newspaper usage". Language 62. 846-878. Weinrich, Harald. 1973. Le temps. Paris: Editions du Seuil. (1964. Tempus. Stuttgart: Verlag M. Kohlhammer). Zusne, Leonard. 1970. Visual Perception of Form. New York: Academic Press.

On the Concept of Time: Prolegomena to a Theory of Aspect and Tense in Narrative Discourse1 Nils B. Thelin University of Oldenburg To Frithiof Rundgren, outstanding Semitist and linguist It is the researcher's moments of grace when scattered parts of the complex nat­ ure observed fall beautifully into place in a simple and clarifying pattern of entirety. Fate's true reward, though, for his patient labor is to control the mechanisms for recreating this pattern, when scattered again, and adjusting it to those new, larger patterns which will continue to surround it time after time. (NBT)

1. Time, eternity, and the distinction defìnite:indefìnite The theoretical issue of temporal structure in a broader sense has a long tradition in the study of language and literature. In his Poetics (ch. 20, 1457a) Aristotle ascribed to the verb a function of "temporal determination" which is missing in the noun (by indicating present, future or past time). The German term Zeitwort, denoting the grammatical category of verbs, reflects, of course, the same old insight. Treatments of the concept of time in ancient philosophy likewise convey observations of interest for understanding the meaning of time as an essential poetic-linguistic problem. By excluding past and future from the meaning of eternity (though not yet speaking of "eternal present"; cf. Ricoeur 1983: 46; see below), Plato (Timaeus: 37c), for example, established in a sense a first approximation of the distinction so crucial to modern aspectology, viz. the one between time-relatedness and non-time-relatedness of reported events (cf. Thelin 1978: 30-31, 65-66; 1984b, 1985: 173fT.). The subsequent theological discussion of eternity and time as two different

262

NILS . THELIN

forms of existence, i.e. the life of God, the Creator, and of man, the creature, respectively, develops further the Platonic distinction by paying more atten­ tion to the psychic activities of man involved in perception and conception of time. Consequently, according to Gregory of Nyssa (quoted here in the interpretation suggested by Callahan 1960: 61), ... the nature that is self-sufficient, eternal, and all-embracing is not in place or in time, but is before and above these. This nature is not measured by ages, nor does it run along with time, but it isfirmlyfixedin itself and is not divided into past and future. For past and future are affections proper to created beings, and refer to the memory and anticipation of the life that is divided by time. Thus Callahan's merit is to have pointed out that the position taken by Gregory of Nyssa indeed mediates between the Neoplatonic concept of time as a divine principle of soul (Plotinus) and the psychological understanding developed by St. Augustine in his well-known discussion of time as an important complement to the traditional metaphysical view (see below). The conceptual-linguistic relevance of the religious distinction between time and non-time as an expression of the two kinds of being represented by man and God, respectively, is evidenced e.g. by the Sanscrit perfect in the hymns of the Vedas. As shown by Renou (1925), the Vedic perfect (though situating events in the past) was reserved for the Gods, for principal parts of the narration to mark their general and solemn character. The same author points, correspondingly, to the special treatment given the desinences of the perfect in Panini, relating them rather to nominal suffixes. It is to Kristeva's (1969: 273ff.) credit that she has shown the universal significance of such historical facts by bringing them into logical connection with the non-time-related use of the imparfait in French and thus giving support to the idea that the organization of temporal perspective in narrative discourse is based on an aspectual distinction, namely the one between timerelatedness and non-time-relatedness (cf. Thelin 1984a: 226ff., 1984b: 264ff., 1985: 180). There exist, as we shall see, in fact, different degrees of abstraction subsumed under the general concept of non-time-relatedness, but Kristeva is perfectly right in stating that her "hors-temps" or "fautemps", marked by the imparfait, ...n'est pas un temps narratif. Il ne raconte aucune histoire ni se réfère à aucun fait représentable, même si "le récit" semble mimer des faits... Si tout temps est un temps de récit, le fautemps n'est pas un temps: il abandonne le récit et ses modalités au présent, à l'aoriste, au futur... Sans préciser le temps et le lieu d'action, cet imparfait du "je" hors-jeu, hors-temps, fautemps, le désigne comme non localisable et non temporel.

ON THE CONCEPT OF TIME

263

Against this background it becomes evident that linguistic theories of aspect and tense have paid relatively little attention to the development of the concept of time itself, proceeding usually from the tacit assumption that its basic meaning is more or less self-evident. The true temporal essence of aspect, for example, has thus until today been obscured as a rule by the strong influence of superficial morphological oppositions and by its con­ fusion with verb semantics (e.g. in the spirit of the Aristotle-Ryle-KennyVendlerian tradition, accepted in most Anglo-Saxon aspectology; cf. Thelin 1985: 173) or related semantico-syntactic conditions (as in the case of the socalled teorija predel'nosti "terminativity theory", persisting in most Soviet aspectology; cf. Thelin 1984c: 227). In reality, explorations of the historical origin of the aspect category (cf. Thelin 1978: 15ff.) indicate rather that the distinction implicit in the possibil­ ity of relating events concretely to the time axis, or, conversely, abstracting or even completely removing them from it (in the former case tense distinc­ tions are still possible, in the latter — usually not; cf. Thelin 1984b: 227) is of a fundamental conceptual-semantic nature and can be traced back to the very process involved in the rise of aspect, viz. changing (under the influence of the growing tense distinctions; cf. Paul 1970: 274ff.) an original more general definite/indefinite character of the verb (corresponding to the one with nouns; cf. Thelin 1978: 15fT., 19) to a specific definiteness/indefiniteness related to time (see below). The above understanding is in accord with Rundgren's (1959: 89, 91, 307) idea of "pure" vs. "localized" or "temporalized" aspect in his analysis of the conditions in Semitic, with Koschmieder's (1929, 1934: 13ff., 228ff.) definition of aspect as an expression of "Verzeitlichung der Verbaldetermination" and Menges' (1968: 128) similar understanding of the aspectual oppositions in Ural-Altaic. Furthermore, conditions of definiteness/indefiniteness with objects (cf. Kramsky 1972) have recently been shown by Chvany (1983: 75ff.) to be generalizable and applicable also to processes and situations. She demonstrates thus the relevance of their specification as to time and space (in terms of "Saliency/Non-Saliency") for their relation to such basic discourse oppositions as backgroundforeground and theme:rheme. The ancient insight of man's fundamental ability to concretize or abstract "reality" in relation to time has not, however, with rare exceptions, been reflected in a more systematic way in modern theories of aspect and tense. Our suggestion (supported thus by certain evidence from Indo-European as well as non-Indo-European languages) to consider the opposition ±time(rel­ atedness) as the basic conceptual-semantic opposition in the universal system of temporal distinctions, conditioned, actually, the first attempt at a consist­ ent hierarchical-componential analysis of aspect capable of bridging this very gap in the tradition of temporal studies (cf. Thelin 1978, 1985: 173ff.).

264

NILS . THELIN

There certainly have been, in the past, various non-systematic observations of the non-time-related use of Slavic imperfective aspect forms with the socalled "generalized-factual" meaning as distinguished from the invariant "concrete-factual" meaning of perfective aspect f orms (cf. Rassudova 1968). Bondarko's (1967, 1971:10ff., 31ff.) feature " ± localization in time" and Maslov's (1973: 82) explicitly universal understanding of the distinction discussed were important, but mostly overlooked, steps on the rough road towards a proper understanding of the essence of aspect semantics. However, Bondarko's distinction occupied a rather subordinate status in his system of oppositions and could not be brought into a hierarchical-logical relation­ ship either with a likely diachronic process of aspect development (like the one indicated above) or with the system-inherent distinction ± totality, which specifies further, in fact, exactly the given type of time-relatedness in cases of the feature +time (i.e. as total vs. partial reference to events associated with the time axis; see below). Maslov, on the other hand, was unaware of the true temporal quality involved in the distinction "generalizedfactual": "concrete-factual". The same condition holds, indeed, for both Forsyth (1970: 5-6) and Comrie (1976: 113). Pettersson's (1972) suggestion to interpret Bondarko's opposition as equivalent to his own opposition " ± Activity" was due to his confusing basic aspect semantics with verb semantics and had the inevitable consequence that his opposition " ± T i m e " became inconsistent not only with the facts it was supposed to explain (see the criticism in Thelin 1978: 57ff.), but also with the basic insights into the mechanisms of man's perceiving "reality" in relation to time gained within the empirical framework sketched above and further developed below. Before we are in a position to formulate adequate linguistic or poeticlinguistic models accounting for the functions of aspect and tense in narrative discourse, we believe it is necessary to explore first more incisively the concept of time proper and its phenomenological status. This will also be the main concern of this paper.

2. In the footsteps of St. Augustine Quid est enim tempus? St. Augustine's {Confessions XI, 14, 17) classical question was accompanied by a first comment as humble as it was frank: "If nobody asks me I know; if someone puts the question and I have to explain, I do not know any more". Judging by most modern theories concerning functions of time in grammar, we linguists apparently can still profit not only from the open-minded attitude and clear path of reasoning demonstrated by St. Augustine (and referred to below in its essential parts on the basis of the elegant exposition offered by Ricoeur 1983: 19ff.).

ON THE CONCEPT OF TIME

265

2.1 The physical and psychical ontology of time Opposing the famous skeptical argument that "time does not exist" (viz. because "future is not yet, past is no more, and present does not remain"), St. Augustine questions how something can be measured that does not exist: "Yet, Lord, we perceive the intervals of time; we compare them with each other, and we call some longer, some shorter. We measure further how much one time is shorter than another" (16, 21). But by what means {quo pacto; 15, 18) does one actually measure time? Searching for an answer to this crucial question, St. Augustine works his way through a long chain of observations and statements. To begin with, he observes that it is only about past and future that we can say that they are long or short. Thus he concludes that present has no extension. He then goes on to modify the absolute understanding of time demonstrated by the skeptical view above, considering time rather as a quality assigned to events reported as existing "future things" and "past things". "If future things and past things exist", he continues his discussion, "I want to know where they exist" (18, 23). Proceeding from the idea that past and future represent the activities of narration and prevision, respectively, St. Augustine now takes a decisive step towards the psychological dimension of time by correlating these activities with the functions of memory and expectation, respectively. By assigning to both of them existence in mind, he is now in a position to corroborate his preliminary assumption that "past things" and "future things" exist, and to situate them in the mind in the form of images. Consequently, on this psychological level he not only solves — in his way — the ontological dispute about time, but also removes, at least at first glance, the previously stated opposition (extension:non-extension) between past and future, on the one hand, and present, on the other: "... my childhood, which is no more, is in past time which is no more; but its image...I retain in present time because it is still in memory" (18, 23). Formulating thus his "three-time-theory" (including "present of the past", "present of the present", "present of the future'), St. Augustine, in fact, gives expression to the idea of the hierarchically superior status of the psychical activities underlying ("present" in) all enunciations (irrespective of their reference to past, actual or future conditions), i.e. the permanent psychological present, or, in other words, the moment of speech as the obligatory primary point of reference ( = point of departure) for expressing all kinds of temporal order (cf. Thelin 1978: 37ff., 41ff.). His "(present of the) present" is, consequently, understood as vision or attention, i.e. the psychological presence of actual perception as distinguished from memory and expectation in cases of "(present of the) past" and "(present of the) future", respectively.

266

NILS . THELIN

2.2 Time as measure: mental extension, motion and duration The careful reader cannot have overlooked the fact that St. Augustine — in spite of the development of his argumentation towards a psychological dimension of the concept of time — at this point is still far from answering the introductory question, by what means time can be measured. Nor have we any explanation for the circumstance that "present" according to the first observation has no extension, whereas his superior "psychological present" does appear to have it. Turning directly to the question of measurement, St. Augustine (21-31) states that time is measured "as it passes". But what passes is "present", he continues, and "present" has no extension. Now, since time can be measured, it must still have extension, but extension of what? The solution to the contra­ diction must be then that extension of time is an extension of mind (distentio animi). This idea of St. Augustine's that time and its extension is a function of man's perceptual-conceptual analysis of "reality" — however useful its underlying general assumption may be — appears, so far, to proceed from the same absolute and unrelated use of the concept of time as demonstrated above in the skeptical argument for the non-existence of time. One is entitled thus to ask: is it, after all, time that is measured in the first place, or is it not rather something else that is measured with the aid of time? Could not exten­ sion be understood as a feature inherent in this "something else" perceived and measured, or, at least, as correlated to its specific nature? The concept of time as measure had, as a matter of fact, already been related by Aristotle in his Physics (11.219b2, 12.221al) to motion, without being identified with it. The concept of extension had an interesting prehis­ tory, too. According to the Stoic view, represented by Chrysippus (cf. Callahan, op. cit.: 63), time was understood more precisely as "an extension (diastēma) that accompanies the motion of the universe". Gregory of Nyssa (ibidem: 60) combined a similar understanding with the concept of measure by referring to time "as an extension accompanying or running along with the universe of becoming and as being, in addition, the measure of all motion in the universe". It is an irony of fate that St. Augustine arrived at his psychological reductionist solution of distentio animi immediately after an incisive dis­ cussion of time and motion which he abandoned just a few steps from the discovery of the essence of their interrelationship, prematurely confessing his failure (cf. Ricoeur, op. cit.: 32ff.). In fact, he had argued most convinc­ ingly for an understanding of time as related not only to the movements of celestial bodies (because he did not preclude that their velocity could be variable), but to movements of any bodies whatever. Liberating thus the concept of time from its compulsory association with regularly repeated

ON THE CONCEPT OF TIME

267

movements, he was, potentially, in a position to give it the generalized status of a flexible pyschological instrument for comparing all kinds of motion, namely in regard to their duration: 'The motion of a body is one thing, what helps us measure its duration is another; who would now not under­ stand which of them deserves to be called time?" Still St. Augustine failed to recognize what this "what" is. He had previously asked the question "whether it is the motion (motus) itself that is the day, or the duration (mora) within which it is accomplished, or both" (23, 30). Later he had even stated that in order to distinguish the moment where the interval of time begins and where it ends, it is necessary to "mark" (notare) the place from which a moving body departs and where it arrives, since otherwise we cannot say "in how long a time, from one point to another, the movement of the body or its parts has been accomplished" (24, 31).

2.3 The remaining discord between St. Augustine's "mental extension" and cognitive-pragmatic (aspectual) conditions of time Ricoeur (op. cit.: 33, 40) has correctly pointed to the difficulty arising from St. Augustine's basically fruitful spatial view, on the one hand, and his situating time as extension exclusively in the mind (see above), on the other. This difficulty, as well as the general vacillation between his psychological postulate and his observations about "reality" (as the one mentioned at the end of the preceding section), causing thus his irresolution, goes back to his defective initial argumentation: "if time can be measured, it must exist; now, time can be measured only in past and future, not in present; past and future thus have extension, present not; if we hold that present is time, too, it must have extension; this extension then exists in the mind for present as well as past and future". Consequently, the asymmetry between extension and measurement on one level is not really solved but rather replaced by symmetry on another level, i.e. it cannot be accounted for by any consequent correlation of psychological time (as being generally extensible and measur­ able) and "real" time (as being only in part extensible and measurable). Ignoring thus the inability of his abstract psychological model to explain systematically the (linguistic) use of time, more specifically the reason why "real" present indeed cannot be used for extension or measurement (at least in cases of time-relatedness), St. Augustine, in fact, retires before an aspectual obstacle, namely the distinction ± totality and the natural, pragmatically and perceptually based, incompatibility of the conceptual-semantic feature, in reality conditioning measurement, viz. -{-totality (i.e. total reference to or survey of an action reported as related to time; see below), with the present, i.e. with a moment of speech coinciding with the action reported.

268

NILS B. THELIN

In the latter case only — totality (i.e. partial reference to an action reported as being related to time) is possible, i.e. the action can be referred to only in respect to that specific part or phase of it that coincides with the moment of speech. For understandable reasons an action reported as going on simultaneously with the moment of speech is not surveyable in its totality and cannot be "extended" or measured either. This is why the total aspectual meaning of Slavic perfective verb forms in the present is incompatible with actual present meaning (and rather transposed to denote future meaning) or why simple forms in English cannot appear in the present in their time-related total aspectual use, but only in their various non-time-related uses: "eternal", "gnomic", "generic", "habit­ ual" etc. The actual present meaning can thus as a rule be expressed only by imperfective and "continuous" forms, respectively, conveying exactly the aspectual meaning of partial reference. The "reportive" and "performative" use of English simple forms in the present tense is due to a specific mechanism of temporal transposition, representing actual events as total (indeed, an "aoristic" use) and having them thus immediately vanish into the past as a prerequisite for their sequentialization (cf. Thelin 1982). Simple forms of English verbs, denoting genuine states, can express time-relatedness also in the present, but not totality, since states are not analyzable in terms of parts (phases) or totality (see section 2.5).

2.4 Aspectual delimitation and time as relative measure Thus unconscious of crucial interrelations between perceptual-conceptual and pragmatic constituents of the semantics of time, St. Augustine must face certain contradictions not explained by his restricted model (furnished in the first place to prove the existence of time, viz. as mental extension), but potentially handled within his general psychological understanding and correct observations of temporal functions. That is why he fails to see that, generalizing the concept of time from being exclusively a measure of duration based on regular motion to being a measure for all kinds of motion, he introduces instead of the concept of time as absolute measure the more abstract concept of time as relative measure. Whilst not penetrating the latter's mode of functioning, St. Augustine remains captivated by the cosmological idea of some necessary standard measure, but since he would not recognize absolute constancy with physical movements, his generalized con­ cept of time brings him into a seeming dilemma. This difficulty is connected with his inability to realize that the procedure correctly observed by him, viz. that time is distinguished by "marking" the initial and final point of a movement (the duration of which is measured; see above), implies an oper-

ON THE CONCEPT OF TIME

269

ation that is of a more fundamental, namely an aspectually delimiting character (expressed by the feature + totality referred to above) and can take place without a subsequent comparison between different delimited movements (actions) in regard to their duration. It is important to state that aspectual delimitation is a relational mechanism conditioning the perception and conception of temporal order between move­ ments (actions) as well as their preceding/resulting states (see section 4.4), and as such offers a natural point of departure for a comparison also with respect to their relative temporal extension. If regularly repeated physical movements did play a decisive role for the development of the concept of time (and on this point there seems to be general agreement), then it cannot have been in the sense of an absolute standard (as it would have been in St. Augustine's mind) because the development and refinements of instruments for measuring such constancy has continued to the present day, but precisely in the sense of relative extensions in time. This fundamental relational mechanism is evi­ denced, as a matter of fact, by the scale of linguistic expressions for time units according to the corresponding relative range of temporal extensions (day, month, year). Implicitly St. Augustine gives support to this relative under­ standing of time by assuming, hypothetically, changes in the velocity of movements serving as temporal standards (23, 30). He even maintains that motion may stop, time may not, and asks whether we do not, in fact, measure stops as well as movements (24, 31; see section 2.5).

2.5 Time as duration of motion between states, and the aspectual distinction ± totality Unable to incorporate his keen observations about "reality" within his reductionist model of distentio animi, St. Augustine never arrives at an integrated, consistent theory of time. With critical distance and accumulated knowledge it is our privilege to approach the problem of time as treated by him and to explore where he went wrong. All the relevant factors involved in the concept of time seem to be there (motion, extension, points, delimi­ tation, velocity, constancy/variability, duration, measure, comparison), but their mutual hierarchical dependency remains unclear in the shadow of St. Augustine's thesis of three times and its idea of mental extension as an isolated analytic instrument. Therefore they cannot, so far, constitute any systematic device for deriving (explaining) the meaning of time. Quid est enim tempus? Now, by using and organizing the factors already present in St. Augustine's discussion we could, indeed, formulate at least a tentative definition of time that can serve as a feasible point of departure for a subsequent more comprehensive analysis: time is a measure of the

270

NILS . THELIN

variable extension, i.e. duration, o f a movement between two points according to its velocity. Related to extensions in a world in motion, points should be understood now as states (old and new states, i.e. states preceding movements and states resulting from them, respectively). The concept of state is further related to perceptual as well as to pragmatic conditions. The concept of pragmatic state, e.g. the intended result achieved by a movement (action), should, we suggest, be distinguished from the concept of perceptual state. A pragmatic state is always also a perceptual state but not the other way around. The fact that we can thus judge the velocity of a movement also between two (pragmatic) states is conditioned by the perceptual strategy of segmentizing motion into inter­ mediate (perceptual) states or phases (cf. Bergson 1911: 18, 32, 171-172, 288, 317-318, 323; Johansson 1950, 1964; Miller & Johnson-Laird 1976: 83ff; Newtson et al. 1978; Thelin 1984a: 231, 1985: 163ff.). This important con­ dition has, as indicated above (see section 2.3, 4.4, 4.7-9, 5), direct bearing upon the aspectual opposition ± totality and gives empirical support to our equipollent interpretation of the feature - totality as underlying the meaning of "process", viz. as "partiality" or, in other words, reference to one part or phase of the movement (action) reported as related to time (cf. Thelin 1978: 31ff., 110ff., 1980: 432ff.). We have pointed to the interesting fact that the use of the partitive case with objects in Finnish may convey exactly the same pro­ cessual meaning of the verbal action as expressed by English "continuous" forms or Slavic imperfective forms in their time-related use. True (pragmatic) states do not, however, possess any internal structure that can be segmentized or partialized. Hence the simple explanation why true stative verbs in English do not as a rule take "continuous" forms, viz. because the latter express "process" exactly by means of partial reference (—totality). States can be measured, however (St. Augustine's observation with respect to "stops" is correct), but only in terms of surrounding movements, i.e. as the extension between the final point of the preceding movement and the initial point of the subsequent one. In a corresponding way states (lacking extension and phases of their own which would allow for their direct association with time) can be related indirectly to time when reported as being defined by sur­ rounding time-related movements (events) or as potentially changeable by such (future) movements. The missing partiality with states is thus a conse­ quence rather of their inability to be totalized by their own.2 3. The perceptual-conceptual hierarchy of matter, space, and time: localist prerequisites for a theory of aspect and tense Proceeding from the view that linguistic theory somehow must account for the fact that the semantic complex of time and its various linguistic shapes

ON THE CONCEPT OF TIME

271

and poetic uses cannot be understood in a deeper integrative way without our paying attention to how time is perceived and conceived by man in a context of action (interaction) and linguistic reproduction of action, we would like to suggest now a rather concrete localist approach to this concept and the preconditions for its development. The idea of deriving temporal mean­ ings from spatial ones is not new, of course, and appears to be justified by, among other things, the fundamental perspectivai function they have in common, and the higher degree of abstraction represented by the former. Further, it is well-known that temporal meanings in the form of aspectual features in many languages are (or have been) conveyed by local (prep­ ositional or adverbial) expressions (cf. Middle English "he waes on huntunge", German "er war am Jagen", "dabei sein etwas zu tun", French "être en train de faire", English "to be in process, motion, at work", etc.). Tense meanings are often expressed by verbs of locomotion, and so on. The general direction of evolution from the concrete concepts of space and motion to the more abstract concept of time could thus at least be symbolized by, for example the Latin development *movimentum {moveo "move") > momentum ("movement changing state" > "moment"). In its basic features the meaning of time as the general concept underlying the semantics of the two distinct temporal categories — aspect and tense — and their development could, according to our understanding, be formu­ lated thus in the form of its hierarchical relations to matter and space: The concepts of matter and space exist in a relation of mutual dependency. The concept of space presupposes the existence of matter and beings perceiving this matter. Equally, the concept of matter presupposes the idea of extension in space. Perceived space is confined in its extension by perceived matter. The extension of matter itself or the space between various instances of matter ("things") is conceived of as distances between its/their ultimate or closest boundaries (points), respectively. Being the fundamental perceptual device for relative ("subjective") spatial perspective, the concept of distance has been generalized and submitted to conventional, so-called absolute ("objective") measurement in terms of length. Matter has two forms of existence in space: rest and motion. At rest the distances between things in space are constant, in motion — changing. Motion causes new distances (new configurations of things) and — after its returning to rest — new states. Motion implies thus change of states. Time is the concept of distance between old and new states, i.e. the duration o f a movement until it has changed an old state into a new one. The duration of a movement is in inverse proportion to its velocity if related to a given distance: the faster the movement between two states, the shorter its duration and the conceived distance between the two states. Time is, in other words, an abstraction and further development of the concept of extension from bearing upon constant distances in a world at rest, i.e. as a means of measuring them

272

NILS . THELIN

in terms of relations within world states (the spatial dimension: length), to bearing upon variable distances in a world in motion depending on the velocity of this motion, i.e. as a means of measuring them in terms of relations between world states (the temporal dimension: duration). Being the fundamental percep­ tual device for relative ("subjective") temporal perspective, the concept of dis­ tance, i.e. duration, has been generalized and submitted to conventional socalled absolute ("objective") measurement in terms of chronological time. The cognitive-pragmatic understanding of the basic semantics of time evidenced above is supported, as a matter of fact, by developments in modern physics. The revolutionary implications of Einstein's theory of relativity and, in particular, quantum mechanics (cf. Davies 1980) no longer allow for a concept of absolute time, but tend rather to stress the decisive role of the observer's (experimenter's) subjectivity, assigning thus not only to perceptual psychology but also, in fact, to linguistic semantics a special relevance in the further study of time as a general phenomenological issue, i.e. as a concept of relative measure characterizing to a certain extent all descriptions of "reality", be it the "objective" world of science or the "subjective" world of everyday life and poetics. 4. The development of temporal distinctions How is the above basic understanding of time as a relative measure of duration to be brought into a logical and genetic relationship with the linguistic categories of aspect and tense? 4.1 Primitive features of time: concreteness, cyclicity, delimitation To begin with, it appears reasonable to assume that the concept of time originally developed related to observations and reports of concrete actions and events taking place in more or less definite locations with definite agents and definite goals or experiencers, as well. Duration was the fundamental dimension of time conditioned primarily by the 24 hour cycle of day and night. Indications of longer durations (in terms of days and nights, for example, how long time the hunters had been away from home) must have been prior to the localization of smaller durations or "points" along the line of extension of, e.g., one day. This follows from the elementary condition that partial indications in regard to the cycle could be made only if the latter existed already as a concept of confined (surveyable) entirety. The further partialization or specification of temporal cycles must, naturally, have been related to an increasing complexity in perceptual-conceptual and linguistic structure.

ON THE CONCEPT OF TIME

273

It would seem appropriate to posit then that temporal references originated in indications of durations, i.e. in measurements of actions and events (and, indirectly, states) reported as taking place (holding) in concrete, definite situations, expressing what we have referred to above (in regard to a later stage of development; see below) as the aspectual feature "time-relatedness" {+time). Since measurement of time presupposes the delimitation of the duration ascribed to actions and events reported, we can further assume that primitive temporal indications were characterized by the prototype of what we have referred to above as the aspectual feature "totality" (4- totality), as well.

4.2 Pre-aspectual perspectivai differentiation in terms of wholes and parts, and the conceptual distinction of past and present time Under the circumstances indicated above, original temporal references (in this fundamental pre-aspectual sense) applied to definite actions and events only in past time, since totality is incompatible with present (i.e. actual) time (cf. also Morin 1942: 172), and the concept of future time has, apparently, to be considered as a later development (transposing the present meaning into the future by means of modality; cf. Thelin 1978: 25ff.). Now, there is strong evidence, for example, from Indo-European languages (Ibid. 15ff.) that the conceptual and linguistic distinction past:present developed later than the aspectual distinctions. In Greek the remaining basic compatibility of the aorist with all times (except for the indicative that cannot express present meaning; see below) and the function of the imperfect as a secondary variation of the present stem expressing the same "processual" (-totality) aspect meaning in the past are such pieces of evidence. It seems justified thus to posit that the aspectual meanings, underlying e.g. the Greek aorist and present stem, respectively, had existed prior to tense meanings. Under these conditions one could think of aspect in its very origin as a feature not yet associated specifically with the meaning of motion (action) and time but representing rather a more general distinction holding for man's perspective on reality, viz. in terms of rather definite patterns of wholes and parts (in concord with Gestalt psychology; cf. Thelin 1985: 163). This perspec­ tivai analysis would then have applied to actions and objects, as well (cf. the above-mentioned use of the partitive case in Finnish objects as an expression of the partial or "processual" aspect meaning). The total view of actions and events as the primary, "pure" aspectual meaning underlying primitive measurements of duration (and reflected later by e.g. the Greek aorist or Slavic perfective forms) referred then to these actions and events (at least if definite) as past (only later as future) not due

274

NILS . THELIN

to any "tense" distinction (because it did not yet exist as such; see below), but as a consequence of this total pre-aspectual view itself. The compatibility of the Greek aorist not only with past (and future) meaning but also with present meaning just seems to be a contradiction, since this latter compati­ bility is valid only for the superior (modal) structures of the semanticosyntactically complex aorist forms represented by the subjunctive, the optative, the imperative, and the infinitive, whereas the subordinate (embed­ ded) actions characterized by the total aspectual (truly aoristic) meaning could not be referred to as actually taking place in the present, i.e. simul­ taneously with the moment of speech. This important fact is evidenced by the non-complex indicative, which could refer only to past (and future) actions, but also by the participle which denoted normally preceding actions in respect to the main action.

4.3 The pre-aspectual concept of totality: from temporal quantification to temporal qualification Proceeding from the above understanding that indications of durations in terms of natural cycles developed from entire cycles to smaller, delimited parts of cycles, we can imagine now how actions viewed in their totality could be localized along the temporal line and defined successively by such small parts of this line (or regular events firmly associated with it) that the superior meaning of totality as delimited extension, i.e. duration in time, was replaced first by a more general total perspective on actions and then through the successive process of temporalization (see Sec. 1, 4.5) by the derived meaning of total association with time, applying also to non-durational actions and changing thus the essential function of this feature from having been one of temporal quantification to being one of temporal qualification. Not the extension of an action along the time axis but its (perspectivai) relation to it thus became the essential conceptual feature of aspectual totality as it developed, prompted by the contrast exercized by the further conceptual possibility of qualifying actions also as being in process by assigning to them a partial relation to the time axis, i.e. by associating them with it only in one of their phases or parts (see below). The meaning of totality has often been referred to in aspectology as "punctuality". The latter has even been regarded as the aspectual essence of the Slavic perfective forms (cf. Dahl 1974). This understanding is not only incompatible with the hypothetical development sketched above, it is also incorrect. Exactly like Slavic perfective forms (cf. Thelin 1978: 33ff.), the Greek aorist could express totality irrespective of whether this totality was defined by its references as punctual or durational (delimitedly exten-

ON THE CONCEPT OF TIME

275

sional): ebasileuse could mean "he became a king", but ebasileuse pentëkonta etē meant "he ruled for 50 years" (cf. the corresponding observations about Modern Greek and French in Seiler 1969: 124). 4.4 The pre-aspectual concept of partiality: The essence of "processuality" Just as total indications with regard to the natural temporal cycles must have preceded partial ones, we may assume that total references to actions and events preceded the partial references underlying the "processual" meaning (see above). This seems reasonable from the above-mentioned condition that the concept of parts is conditioned by the concept of wholes, and the corresponding dependency on a total perspective exhibited by the partial perspective as prototype for the aspectual feature — totality. Actions can thus be described in their totality if referred to a past (or future) point or segment of the time axis. Such "points" of reference can be defined by other "total" actions, simultaneous with or surrounding the given action, and by the preceding or simultaneous states or processes as well. In other words, "total" actions can be defined by processes, i.e. partial references, but must not be. "Partial" actions, however, must be defined by "total" actions or other points of reference allowing for a similar "confined glimpse" (Thelin 1978: 31) into their internal structure as the essence of "processual­ ity" (expressed by the Greek present stem, Slavic imperfective forms in their time-related use, and "continuous" forms in English). Traditional expressions like "progressivity" and, especially, "durativity" (since duration, in fact, implies delimitation, i.e. totality) obscure this important condition. The circumstances referred to here are reflected by the roles fulfilled by "total" and "partial" actions or events in narrative discourse: "total" actions, constituting the plot, could theoretically dispense with "partial" actions, constituting the actual background (cf. Thelin 1984b: 226, 1985: 181, 184, 185ff.), although this is practically never the case. Narration is normally not conceivable, however, without totality, delimiting actions and events from each other and making possible in this way their succession and the progression of the story (see further section 4.10.3 for a discussion of the superior distinction between these time-related features, i.e. + totality and — totality, implicit directly or indirectly in foreground/narratio, and the feature of non-time-relatedness implicit in proper background/descriptio). 4.5 The distribution of pre-aspectual perspective, the rise of the past:present distinction, and the temporalization as genesis of aspect proper The only perspective applicable to actions in present time, i.e. referred to as simultaneous with the moment of speech, was — as indicated above —

276

NILS . THELIN

the partial one. Since the total perspective was restricted first to actions in past time, we may assume that this perspectivai distribution was the factor which also conceptually conditioned the rise of the temporal distinction past:present (and later the development of the linguistic tense category). Having initially released this process, the pre-aspect meanings — as a general (spatial) perspectivai distinction in terms of wholes and parts — were then themselves influenced by the growing concept of temporality and successively associated more firmly with the concept conditioning time, i.e. motion and its manifestations as actions and events. This was, in our understanding, how the temporalization of the original "pure" aspect or, rather, pre-aspect meanings arose and led, as a first consequence, to the proper aspect distinction ± totality, specifying actions and events with respect to the character of their relation to time (see above).

4.6 From pre-aspectual definiteness:indefiniteness to the superior aspectual distinction time-relatedness:non-time-relatedness ( ± time) Even before the temporalization, the general perspectivai distinction, which applied to actions as well as objects by referring to them as definite (concrete) patterns of wholes and parts, had probably undergone a process of generalization, abstracting from reality as instanced by directly perceiv­ able actions and objects and allowing for references to them as indefinite patterns in terms of experientially derived general conditions and classes. With regard to actions and events this feature of indefiniteness had pre­ sumedly developed as a consequence of a growing conceptual distinction between their uniqueness and non-uniqueness, respectively. Thus the tem­ poralization probably implied that the distinction definiteness: indefi­ niteness linked uniqueness of an action to the idea of a concrete or individual situation defined by specific references, associating the action with a definite segment of the time axis (i.e. the conceptual-semantic feature of time-relatedness: + time), and linked non-uniqueness of an action to the idea of general validity holding by force of e.g. its repeatability, habituality or potentiality, and defined by lack of specific references, dissociating it from the time-axis (i.e. the conceptual-semantic feature of non-time-relatedness: - time). Completing thus the system of basic aspectual distinctions, the opposition ± time was to occupy its superior hierarchical position, being specified further — in cases of + time — by the subordinate distinction ± totality in regard to the given type of timerelatedness. Thus, in distinction to our hypothesis in Thelin (1978: 18ff.), the above analysis gives conceptual-genetic priority to the more concrete opposition

ON THE CONCEPT OF TIME

277

± totality as compared with the resulting system-inherent superiority of the opposition ± time.

4.7 The rise of tense distinctions Let us return now to the rise of tense distinctions, conditioned first by the pre-aspectual distinctions and then themselves contributing in turn to their temporalization. In accord with the idea that a conceptual and linguistic distinction appears either as a consequence of further specification ("totality generates partiality") or, conversely, as a consequence of generalization ("definiteness generates indefiniteness") of original features, we may assume that the pre-tense distinction past:present developed as a further specifica­ tion, starting from the concept of past time as marked by the primary preaspect feature of totality. This is to be understood in such a way that the pre-aspectually conditioned appearance of the temporal concept of "past" was a prerequisite of contrast for present (i.e. moment of speech) to become also the temporal concept of "present" ("past generates present"). The temporal concept of present as defined in contrast with past was now associated with the only pre-aspectual feature applying to present, viz. partiality, as opposed to the feature of totality associated with the past.

4.8 The hypothetical development of temporal distinctions: summary of its conceptual-genetic stages (conceived as a process of increasing complexity with considerable over-lap) I. The concept of time as instrument for measurement of duration. 1. Conceptual features a. Basic features: definiteness, the general perspectivai distinction in terms of wholes (pre-aspectual totality) and parts (pre-aspectual partiality), motion. b. Motion-oriented features: cycles, extension, delimitation. 2. Developments a. The pre-aspectual totality perspective, applied to motion, gives rise to the past time perspective as the conceptually marked feature of temporal order. b. The pre-aspectual totality perspective as the conceptual instrument for delimiting extension in terms of cycles or parts of cycles, i.e. for temporal quantification, becomes successively the instrument for a more general total perspective on motion.

278

NILS . THELIN

II. The concept of time as instrument for pre-aspectual perspectivai differen­ tiation. 1. Conceptual features (applied to motion): definiteness, the pre-aspectual totality perspective, the pre-tense past time perspective. 2. Developments a. In accord with the general perspectivai distinction in terms of wholes and parts, motion-related pre-aspectual totality is followed by its equi­ pollent opposite partiality as an expression of the growing concept of a "processual" perspective. b. The definite perspectivai differentiation of motion (in terms of preaspectual totality:partiality) is balanced by the further possibility of indefinite references to motion.  The pre-aspectual concepts of totality and partiality, correlated to past and present motion, respectively, give rise to the conceptual pre­ tense distinction past:present. III. The concept of time as complex instrument for two interrelated but distinct modes of perspectivai differentiation: temporal quality (or "con­ tour"; cf. Hockett 1958: 237), i.e. aspect, and temporal order, i.e. tense. 1. Conceptual features (distinctions): definiteness:indefiniteness, preaspectual totality:partiality, pre-tense past:present. 2. Developments a. Correlated with the pre-tense perspectivai distinction past:present, the pre-aspectual distinction totality:partiality is associated more firmly with time as a medium for its perspectivai differentiation of motion, i.e. actions and events, not regarding their order along the time axis (as in the former case), but regarding their relation to the time axis itself, viz. their temporal quality (as opposed to primitive temporal quantity). b. The distinction definiteness:indefiniteness, accordingly, as a conse­ quence of the temporalization, now applies to the relation of actions and events to the time axis, viz. by superimposing on the distinction, defining the specific character of this relation (in terms of totality:partiality), the more general perspectivai distinction whether actions and events are defined by such specific relations to time, or whether they are not (i.e. time-relatedness vs. non-time-relatedness).  Along with the development of aspect proper, the pre-tense concept of temporal order in terms of past:present becomes successively the conceptual distinction underlying tense proper. Tense distinctions are to occupy then in the hierarchical over-all system of time semantics a position subordinate to aspect distinctions since they are dependent on

ON THE CONCEPT OF TIME

279

the latter: higher degrees of non-time-relatedness thus preclude tense differentiation completely, and time-relatedness, specified as totality, pre­ dicts the inappropriateness of (actual) present meaning.

4.9 The Indo-European trichotomy aorist-present-imperfect The state of affairs resulting from the temporalization, i.e. the two aspect meanings totality and partiality correlated to the two tense meanings past and present, respectively, characterizes in Indo-European a stage of develop­ ment represented basically by the aorist-present dichotomy. The genesis of the imperfect was a natural step to assure primarily the expression of partiality ("processuality", i.e. —totality) in past time also. Its past time "partner" — the aorist — as a consequence partly of its new role as opposite to the restrictedly past-time-related imperfect, henceforth was to lose succesively its basically tense-independent status (holding for Vedic Sanscrit, for Greek only in a restricted sense) and become an aspect-tense expression usually associated exclusively with past time (as in Slavic).

4.10 The universal aspect distinction ± time Our above-mentioned hypothesis of the distinction definiteness'.indefiniteness and its reinterpretation in respect of actions and events as a consequence of the temporalization, viz. as time-relatedness (+time) and non-timerelatedness ( — time), respectively (cf. also Thelin 1978: 19ff., 30ff.), should now be integrated further into the conceptual-genetic hypothesis of time semantics sketched so far. Without the latter distinction the universal phenomenon of aspect and its origin would remain a mystery. 4.10.1 The mechanism of temporal abstraction. It is interesting to note that Plotinus (Ennead III.7.1.20), referring to Plato's understanding of time as "an image of eternity", points to the possibility of describing the relation between them also by proceeding the other way, i.e. from time to eternity. Regarding eternity as the "archetype", Plotinus thus gives implicit support to an analysis that treats non-time-relatedness ( — time) as an abstraction of primary time-relatedness ( + time). This condition is reflected by the circum­ stance that the most abstract meanings of non-time-relatedness ( — time) refer rather to "all times" (as in cases of e.g. "eternal truths", gnomes, etc.), whereas less abstract meanings (as the "generic", habitual, and "simple denotative" meanings) may be differentiated in regard to tense. In the former case we have spoken of a complete removal from the time axis, in the latter

280

NILS . THELIN

case of different degrees of abstraction from it (see section 1; cf. Thelin 1978: 30-31, 65-66, 1984a: 227). 4.10.2 Excursus on St. Augustine, Ricœur, and the hierarchization of time. It is certainly no coincidence that Ricœur (op.cit.: 41ff.) in his insightful analysis of St. Augustine's treatment of eternity and its implications for the concept of time, independently comes to a similar conclusion, ascribing thus to his meditation on eternity a function of contrast ("idée-limite qui contraint à penser à la fois le temps et l'autre du temps"; compare our opposition ±time), intensifying "l'expérience même de la distentio au plan existentiel" (our feature + time) and causing it at the same time to approach eternity and "donc à se hiérarchiser intérieurement, à Vencontre de la fascination par la représentation d'un temps rectilinéaire". Ricœur, certainly, comes to his conclusion from his analysis of the specific conditions holding for the dinstinction between St. Augustine's concepts of the eternal Word of God (Verbum) and the temporal voice of man (vox), and assigning to the Word as instruction an internalized function, mediating between them; but his further generalization of these conditions is in har­ mony with our above-mentioned universal understanding of the distinction ± time and its constitutive role in narrative discourse: "Pérégrination et nar­ ration sont fondées dans une approximation de l'éternité par le temps, laquelle, loin d'abolir la différence, ne cesse de la creuser" (p. 52). The intensification of temporal experience achieved by the contrast of eternity and time (representing the features — time and + time respectively), is understood by Ricœur as leading to an internal hierarchization of time (i.e. our feature + time, and not - time), since he sees it in opposition to the specific concept of eternity. This is just a seeming difference between our analyses which disappears, naturally, if we (in view of a more general linguistic-poetic application) prefer to see eternity just as one of the possible meanings representing non-time-relatedness (-time), if also the most abstract one. Under these circumstances we can thus establish a pretty good equivalence between our "different degrees of abstraction from the time axis" (Thelin 1978: 65) and Ricœur's (p. 51) "hiérarchie de niveaux de tem­ poralisation, selon que cette expérience s'éloigne ou se rapproche de son pôle d'éternité". 4.10.3 ±time, strategies of narration, and the enriched temporality. It is remarkable, but not really surprising, given this background, how Ricœur's analysis, proceeding from St. Augustine's meditations on time and eternity, and our conceptual-semantic analysis of aspect, have arrived then at such a close understanding of time and the character of its constitutive role in narrative discourse. It is thus symptomatic that both analyses lead to a

ON THE CONCEPT OF TIME

281

similar generalization, connecting the temporal (basic aspectual) distinctions discovered with corresponding different strategies of narration. This circum­ stance is evidenced thus by our perspectivai interpretation of the innovative element in Čechov's prose, viz. in terms of the relations of interplay holding between foreground and background structures of narratio and descriptio, respectively (cf. Thelin 1984a: 230, 1984b: 271ff.). Proceeding from Pomorska's (1976) observation that this innovation is based on "abandoning the representation of "events", that is, the usage of perceivable, discrete units, and replacing them by "non-events" or units unmarked in the percep­ tion of the reader", we thus suggested for these two modes of narration the application of our conceptual-semantic aspect features time-relatedness and non-time-relatedness, respectively. Proposing for Čechov's type of prose the term "narrative inversion", we characterized the effect of replacing narratio (+time) with descriptio (—time) as an increasing tension (or oscillation) between them, assigning to conventional descriptive (background) structures a "dynamic-pregnant perspective". The above analysis is thus paralleled by Ricaur's generalizing conclusion: "S'il est vrai que la pente majeure de la théorie moderne du récit... est de "déchronologiser" le récit, la lutte contre la représentation linéaire du temps n'a pas nécessairement pour seule issue de "logiciser" le récit, mais bien d'en approfondir la temporalité" (our emphasis, NBT).

5. Modem aspectology and Bergson's philosophy of time To draw a parallel between the above distinction between — time and + time and Bergson's (1889; cf. also 1922: 54 ff.) concepts of "temps réel" (or "durée-qualité"), on the one hand, and "temps...materialise, devenue quantité par un developpement dans l'espace", on the other, would not seem fully appropriate to us. Bergson's temporal concept of "durée-qualité" cannot thus be considered as equivalent to the aspectual concept of backgrounding non-time-relatedness (cf. Jensen 1990: 401ff). In Bergson this concept rather implies the basic notion of permanent and indivisible continuity (cf. also Bergson 1911: 32), and as such is always constitutive of reality (motion) and also mind (cf. Morin 1942: 170). In distinction to our abstract conceptualsemantic feature of non-time-relatedness (genetically derived from timerelatedness; see above), Bergson's "durée-qualité" expresses a rather concrete hypothesis of the ontological reality underlying mental processes and their external object of perception (compare: "temps réel"), namely as permanent change. Therefore, while lacking completely the abstract concept of aspectual non-time relatedness, Bergson's time theory proceeds instead — in the shape of"durée-qualité" — from what can be called rather an intuitive

282

NILS . THELIN

prototype of the concrete concept of aspectual time-relatedness (+time) and derives from it the corresponding intellectual interpretation of the mobile and indivisible reality by means of localistic (stationary) perceptual stra­ tegies, reflected by our aspect features of totalization (+ totality) and partialization ( - totality). Understood in this way, Bergson's philosophy of time thus gives important support to our above-mentioned Gestalt-psychological analysis of the essential perceptual mechanism underlying the equipollent aspectual opposition ± totality, i.e. as an expression of total (delimiting) references in cases of state-changing events and partial (phasal) references in cases of on-going processes, respectively; cf. Bergson (1911): The mechanistic instinct of the mind is stronger than reason, stronger than immediate experience. The metaphysician that we each carry unconsciously within us... has its fixed requirements, its ready-made explanations, its irreducible propositions: all unite in denying concrete duration. Change must be reducible to an arrangement or rearrangement of parts...(p. 18) The real whole might well be, we conceive, an indivisible continuity. The systems we cut out within it would, properly speaking, not then be parts at all; they would be partial views of the whole...(p.32) ... of becoming we perceive only states, of duration only instants...(p. 288) ...The primal function of perception is precisely to grasp a series of elementary changes under the form of a quality or of a simple state, by work of condensation (p. 317). It should be evident that Bergson's references to the "whole" and to "parts" pertain here to the whole of continuity and to segmentation (of the whole) in a more general sense, respectively, covering in the latter case both totalizing (delimiting) and partializing (phasalizing) references, ie. + totality and - totality, respectively. The interesting thing is that Bergson, by speaking of "partial views" (rather than "parts'), not only gives expression to his deep understanding of the mechanisms of perception, but also supports the case for drawing on these mechanisms for a linguistic theory of verbal aspect. Understood thus as a matter of system-inherent hierarchization, Bergson's distinction between intuitive time and intellectual (materialized, spatialized) time is, as we have seen, in accord with the aspect model suggested above, and can be regarded as a system based in fact on time-relatedness (i.e. + time) and its further localistic specification (i.e. in terms of the opposition ± totality). However, if understood also as a hypothesis of the conceptualgenetic development, Bergson's distinction would conflict in this respect with the traditional assumption shared by our analysis, namely that temporal meanings have developed from spatial ones.

ON THE CONCEPT OF TIME

283

Notes 1. The main concern of the present study (explicitly stated at the end of Section 1) is the concept of time proper and its phenomenological status. The original, unabridged version of my paper, including examples from Greek, Bulgarian and Russian, has appeared in Verbal Aspect in Discourse (1990 : 91-129). I wish to express my gratitude to Flora KleinAndreu, Paul J. Hopper and John S. Robertson for their constructive discussion. I am grateful to Frank Beardow for his friendly help in improving the English of my manuscript. 2. Flora Klein-Andreu's basically favorable interpretation of my proposal in regard to the non-occurence of English "stative verbs" in the progressive form apparently requires some clarification. Her comments might perhaps give the impression that in my view states are generally non-time-related and "therefore, the distinction + / — totalization is irrelevant". The decisive condition, however, is rather that states lack the internal phasal structure characteristic of motion which allows for partial ( — totality) or total (+ totality) timerelated (+ time) aspectual references. From this primary condition one has to derive also the secondary condition that states cannot be related to time, i.e. by their own, but certainly can be so by means of surrounding state-changing events delimiting the extension of (i.e. totalizing) them, or, by means of simultaneous events or points of time confining their validity to specific segments of time and ascribing to them in this fashion a feature of (potential) changeability and, consequently, a broader contextual ("inferential") function of phase, applicable thus to aspectual partialization ( — totality) exactly as in cases of motion. If we overlook here such obvious cases where "stative verbs" in the progressive form in fact convey an activity meaning (cf. "You are being silly" = "You are behaving in a silly way"), the above interpretation would explain those seemingly problematic cases where no activity is involved but the progressive form is still possible. According to my understanding, these cases generally reveal thus a processual feature of temporariness or dynamicity superimposed on the basic stative meaning. The feature of temporary validity or intensity can be illustrated by examples like "You are looking fine tonight" and "I'm really hoping she comes", respectively. The examples suggested by Flora Klein-Andreu ("I'm understanding/liking it much better now") convey precisely the dynamicity of increas­ ing state in the "inceptive" sense correctly observed by her. The meaning of "limit" indicated here cannot, however, be associated with aspectual totality since there is no total survey implicated in these cases by surrounding events. The "limit" felt here is rather a reflection of temporary validity achieved by the aspectual interpretation of states as integrated phases of broader processes, i.e., contrarily, by (contextual) partialization. These conditions, in turn, show that the concept of "stativity" — as far as aspect semantics is concerned — is opposed in fact not by "activity" (pertaining rather to verb semantics) but by the more general concept of "motion", amenable directly to the total or partial ("processual") perspective represented by the aspectual distinction ± totality. Fulfilling in this way Flora Klein-Andreu's request for a substantiation of the analysis of "state" and " — totality" (i.e. motion seen as "process" — as opposed to " + totality", i.e. motion seen as "event"), I believe I can conclude by making the following suggestion: The two concep­ tual-semantic features "state" and "motion" constitute an equipollent opposition ( ± state) precluding necessarily their paradigmatic co-occurrence; their syntagmatic co-occurrence is possible, as we have seen, only in cases of states subordinate to (embedded into) superior meanings of time-related motion ("processes" or "events") — and this is the only concrete form of their "coherence" I can conceive of so far. Given the adequate hierarchical integration of the opposition ± state and the basic aspectual oppositions ±time and ± totality into the over-all system of predicate-semantic distinctions (for such an attempt on the basis of discourse functions, see Thelin 1990: 27ff.), the theoretical framework

284

NILS . THELIN

sketched above seems to offer thus a plausible solution also to the notorious linguistic problem of "progressive states".

References Bergson, H. 1889. Essai sur les données immédiates de la conscience. Paris. . 1911. Creative Evolution. New York. . 1922. Durée et simultanéité. Paris. Bondarko, A.V. 1967. Russkij glagol. Leningrad. . 1971. Vidy i vremja russkogo glagola. Moskva. Callahan, J.F. 1960. "Gregory of Nyssa and the Psychological View of Time". Proceedings of the XIIth International Congress of Philosophy, Venice, Sept. 12-18, 1958. Vol. 11, 59-66. Florence. Chvany, C.V. 1983. "On 'Definiteness' in Bulgarian, English and Russian". American Contri­ butions to the Ninth International Congress of Slavists, Kiev, September 1983. Vol. I (Linguis­ tics), ed. by M. S. Flier, 71-92. Columbus. Comrie, B. 1976. Aspect. Cambridge. Dahl, Ö. 1974. "Some Suggestions for a Logic of Aspects". Göteborg Contributions to the VII. International Congress of Slavists in Warsaw, Aug. 21-27, 1973 {Slavica Gothoburgensia 6), 21-35. Kungälv. Davies, P. 1980. Other Worlds — Space, Superspace and the Quantum Universe. London. Forsyth, J. 1970. A Grammar of Aspect. Usage and Meaning in the Russian Verb. Cambridge. Hockett, C. F. 1958. A Course in Modern Linguistics. Chicago. Jensen, P. 1990. "Narrative Description or Descriptive Narration: Problems of Aspectuality in Čechov". Verbal Aspect in Discourse, ed. by N.B. Thelin, 383-409. Amsterdam: Benjamins. Johansson, G. 1950. Configurations in Event Perception. Uppsala. . 1964. "Perception of Motion and Changing Form". Scandinavian Journal of Psychology 5. 181-208. Koschmieder, E. 1929. Zeitbezug und Sprache. Ein Beitrag zur Aspekt- und Tempusfrage. Leipzig-Berlin. . 1934. Nauka o aspektach czasownika polskiego w zarysie. Wilno. Kramsky, J. 1972. The Article and the Concept of Definiteness in Language. The Hague-Paris. Kristeva, J. 1969. Sēmeiōtikē: Recherches pour une sémanalyse. Paris. Maslov, Ju. S. 1973. "Universal'nye semantičeskie komponenty v soderžanii grammatičeskoj kategorii soveršennogo/nesoveršennogo vida". Sovetskoe slavjanove denie 4. 73-83. Menges, K.H. 1968. The Turkic Languages and Peoples. Wiesbaden. Miller, G. & Ph. Johnson-Laird. 1976. Language and Perception. Cambridge. Morin, H. 1942. Wikner och tidsproblemet. Stockholm. Newtson, D. et al. 1978. "Effects of Availability of Feature Changes on Behaviour Segmen­ tation". Journal of Experimental Social Psychology 14. 379-88. Paul, H. 1970. Prinzipien der Sprachgeschichte. 8th ed. Tübingen. Pettersson, Th. 1972. On Russian Predicates — A Theory of Case and Aspect {Slavica Gothoburg­ ensia 5). Göteborg. Pomorska, . 1976. "On the Structure of Modern Prose. Čechov and Solženicyn". Journal for Descriptive Poetics and Theory of Literature 1, 459-65. Rassudova, O. P. 1968. Upotreblenie vidov v russkom jazyke. Moskva. Renou, L. 1925. La valeur du parfait dans les hymnes védiques. Paris. Ricœur, P. 1983. Temps et récit. Tome 1. Paris.

ON THE CONCEPT OF TIME

285

Rundgren, F. 1959. Intensiv und Aspektkorrelation {Uppsala Universitets Årsskrift 1959:5). Uppsala-Wiesbaden. Seiler, H. 1969. "Zur Problematik des Verbalaspekts". Cahiers Ferdinand de Saussure 26. 119-35. Thelin, N.B. 1978. Towards a Theory of Aspect, Tense and Actionality in Slavic (Acta Universitatis Upsaliensis. Studia Slavica Upsaliensia 18). Uppsala. . 1980. "Aspekt und Aktionalität im Russischen". Die Welt der Slaven 25:2. 428-40. . 1982. "Universaini funkcii na temporalnost i aspektualnost i bälgarskite t.n. 'defektivni' glagoli ot tipa napisvam". Die slawischen Sprachen 1. 108-118. . 1984a. "Coherence, Perspective and Aspectual Specification in Slavonic Narrative Dis­ course". Aspect Bound: A voyage into the realm of Germanic, Slavonic and Finno-Ugrian aspectology, ed. by  de Groot & H. Tommola, 225-38. Dordrecht. . 1984b. "Komposition, Perspektive und Verbalaspekt in Puškins Prosa: Entwurf einer poetisch-linguistichen Methode". Signs of Friendship — To Honour A.G.F van Holk, ed. by J.J. Baak, 257-94. Amsterdam. . 1984c. "K tipologii glagol'noj prefiksach i  semantiki v slavjanskich jazykax". Revue des études slaves LVI:2. 225-38. . 1985. "Kognitiv-pragmatische Korrelate semantischer Strukturen in einem Zirkelmodell für prozessuele Sprachanalyse: Der Aspektspezifizierung im Slawischen zugrunde liegende Hierarchien und Komponenten". Zeitschrift für Slawistik 30:2. 153-99. . (ed.) 1990. Verbal Aspect in Discourse. Amsterdam: J. Benjamins.

On the Projection of Equivalence Relations into Syntagms Henning Andersen University of California Among the numerous stimulating insights into the nature of language which we owe to the genius of Roman Jakobson is the principle which will be the topic of this paper. Jakobson first formulated it in 1960, but he returned to it in several later works (e.g., Jakobson & Pomorska 1983: 131), confident that he had here captured a fundamental principle not just of verbal art but of all uses of language in which there is a set or focus on the aesthetic side of speech: (1)

...we must recall the two basic modes of arrangement used in verbal behavior, selection and combination.... The selection is produced on the basis of equivalence, similarity and dissimilarity, synonymity and antonymity, while the combination, the build-up of the sequence, is based on contiguity. The poetic function projects the principle of equivalence from the axis of selection onto the axis of combination. Equivalence is promoted to the constitutive device of the sequence. (Jakobson 1960: 358; his emphasis)

His own work with the analysis of poetry in more than a dozen poetic traditions corroborated his conviction that the Projection Principle, as I will call it, entails a precise, objective and fruitful technique of analysis. Many others have been similarly convinced and have followed Jakobson's lead in poetic analysis, as detailed, for instance, by Pagnini who recognizes the Projection Principle as "a basic law governing poetry" (1979: 97f.) and as acknowledged also by, for instance, Fox (1977) and Todorov (1977). Even some scholars who emphasize that there is much more to poeticity concede the basic validity of the Projection Principle; see, for instance, Koch (1982). In view of these impressive practical results and the substantial consensus among literary scholars who have applied the techniques of analysis the principle entails, it might seem superfluous to subject it to a critical review, and certainly precarious to question its adequacy. The proof of the pudding is in the eating, as the old saying goes, and if the eating is good, why question the recipe? Some may feel, however, as I do, that the Projection

288

HENNING ANDERSEN

Principle attracts attention by its very successes and deserves a closer scrutiny than it has received so far. In fact one does not have to look at it very closely to realize that it raises interesting questions from two points of view, call them theoretical and applied. The first of these has to do with the relation between the Projection Principle and other parts of, or aspects of, Jakobson's theory of language — or perhaps one should say theory of semiotics. I think most scholars familiar with Jakobson's work will agree that there is much greater coherence among different parts of his theory of language than he himself had time, or cared, to make explicit. Evidently, when one part of this theory has proven as successful through its application in linguistic (poetic) analysis as has the Projection Principle, it is not an idle question how this principle fits in with other parts of the theory, or what bearing its apparent fruitfulness might have on the theory as a whole. I will consider several aspects of Jakobson's theory of language below, and try to suggest how the Projection Principle might be seen to cohere with them. The second point of view from which the Projection Principle raises questions is what I called the applied one above. Jakobson tested the adequacy of the Projection Principle numerous times from one of two possible points of view, in the form of the question: does this principle apply to poetic texts of all kinds? But there is a complementary question which should not be neglected, and which I will consider in this paper: does the Projection Principle apply only to poetic language? The importance of the answer is obvious. If the answer is Yes, the Projection Principle can continue to be considered an essential, if not necessarily the only, determinant of poetic language. If the answer is No, the principle must be accorded a different status depending on the kinds of texts in which it is found to be reflected. In this paper I will present several kinds of evidence that suggest that this latter possibility must be faced, and that the Projection Principle in fact has a considerably wider, and very different, application than has previously been recognized. Before we go any further, however, we must make sure that we have an explicit understanding of the Projection Principle and of its extension as Jakobson saw it.

The Projection Principle Jakobson's formulation of the principle, which was cited above in (1), is tolerably clear, although in some respects it is not very precise. It is obviously

EQUIVALENCE RELATIONS AND SYNTAGMS

289

not "the principle" that is projected, but relations of equivalence. Nor is it "the poetic function" that projects these relations of equivalence, but the author or sender of the aesthetically formed message. But such cavils aside, the principle and Jakobson's claim for it are clear enough: in verbal behavior the operations of selection and combination may be executed in such a way that terms that are equivalent in the paradigmatic system of the code (grammar) enter contiguity relations in the syntagmatic chain of the message (text). This is the Projection Principle. Jakobson's thesis is that a predomi­ nance of precisely such selections and combinations will produce messages (texts) with such aesthetic qualities that they may serve a poetic function. I will refer to this as the Projection Thesis. Jakobson's discussions and illustrations make it clear that the aesthetic qualities of messages thus produced may be unpremeditated — as in the case of the unreflected preference for one wording over another ("that horrible Harry" just sounds more apt than, say, "that terrible Harry") — or they may be the result of deliberate design — as in advertizing copy or poetry. The equivalences range, in Jakobson's illustrations, from single distinctive features in patterns of alliteration or assonance, across a rich variety of phonological, morphological and lexical recurrences, to more or less complex syntactic configurations in parallel lines of poetry. And they include, accord­ ing to his programmatic statements on prose, semantic units of different extent influencing the composition of the plot, the characteristics of subjects and objects of the action, and the sequence of themes in narrative (Jakobson & Pomorska 1983: 107). Even this quite cursory enumeration of ways in which the Projection Principle may be manifested will suggest to the reflective reader that literary texts may contain more equivalences in relations of contiguity than those which are accounted for by the Projection Principle. In the first place one can see in these examples a clear difference between what might be called code defined and message defined (work-specific) recurrences. Alliteration and rime, or morphological and lexical parallelism perhaps belong to the former category. They arise through the selection of code-given equivalents which are then disposed in relations of contiguity in the poetically formed message. But when a complex semantico-syntactic construction is repeated (possibly with poignant deviations) in a later parallel line, or such a constellation of semantic units as the characterization of a person in a prose narrative is answered by a similar (or contrasting) charac­ terization of another person elsewhere in the narrative, we cannot claim to be dealing with the simple arrangement of selected equivalents in relations of contiguity. Rather, to account for such types of iteration one must recognize higher levels of textual structure whose constituent elements — far from being selected ready-made from the paradigms of some pre-

290

HENNING ANDERSEN

existing code — are complex configurations — themselves products of selection and combination — created for the nonce as correspondents in equivalence relations that are specific to the work in question. These work-specific equivalences, which Jakobson rightly recognized as an essential source of the aesthetic value of, for instance, narrative, fall outside the purview of the Projection Principle as it stands. Jakobson appar­ ently alluded to them in his discussion of another theory of his, the theory of introversive semiosis. We will return to this at a later point when we consider the relation between this theory and the Projection Thesis. Another limitation of the Projection Principle leaps to the eye in Jakobson's reference to semiotic units that influence the composition of the plot or the characterization of the persons of a narrative. It is obvious that in this formulation we are being taken across the border between properly speaking linguistic semantic units and such elements of content which — although they are ultimately expressed through the medium of language — are not by any stretch of the imagination linguistic units. There can be no question about the importance of the iteration of such semiotic units for the structure of works of literature — their contribution to the aesthetic value of literary texts is self-evident. And they were in fact recognized by literary scholars long before Jakobson made the programmatic statement cited above; see, for instance, Brown (1950), Schmid (1977, esp. 88ff.). But their existence is clearly not explained by a principle which is intended to account for the poetic function of linguistic elements. Perhaps it is Jakobson's use of the word code — instead of, say, grammar — that led to this discrepancy between the linguistic Projection Principle and these extra-linguistic examples of it. Be that as it may, we can note that the Projection Principle as it was formulated in 1960 does not capture the full breadth of Jakobson's conception. As I hope to demonstrate below, the Projection Principle is not a narrowly linguistic principle, but a general semiotic one. It is, to borrow a phrase, an accident of Jakobson's biography that it was formulated first with reference to poetic language. It needs to be recast in more general terms, but before we do this let us first determine its place relative to two central aspects of Jakobson's theory of language, his code vs. message distinction and his theory of linguistic iconicity.

Grammar, Speech, and Text Perhaps the most generally interesting thing about the Projection Principle is the fact that it explicates properties of messages by referring to the processes or operations by which messages are produced in "verbal

EQUIVALENCE RELATIONS AND SYNTAGMS

291

behavior" (cf. (1)). The principle, in point of fact, gives full recognition to all three modes of being of language: grammar (language as technique, Aristotle's dunamis), speech (language as activity, enérgeia), and text (lan­ guage as product, érgon); cf. Coseriu (1974:37). This is a notable departure from the century-long tradition of conflating two of these, of seeing only the antinomy of jazyk and reč (Baudouin de Courtenay), Sprache and Sprechen (Gabelentz), langue and parole (Saussure) or that of language and text (Hjelmslev), which Jakobson himself apparently preferred when he spoke of code and message ([1960] 1971: 41, 1971: 718ff.). Jakobson's focus on the operations by which the poetic text is created has been noted by Todorov (1977: 476) as characteristic of his theory of poetics practically since his earliest writings. It may be that this focus stands out with especial salience to the theor­ etician of literature for whom the literary text is naturally primary. But this focus on speech activity is typical also of much of Jakobson's work in general linguistics, as witnessed by his writings on aphasia and child lan­ guage. And the distinction between speech activity (the here-and-now of production, or performance) and text (the finished product, in literature, the aesthetic object, an enduring thing capable of being re-produced and experienced anew in repeated performances) is clear in Jakobson's practice, though rarely thematicized in his work (cf. 1960: 366, 371; 1973: 486). Considering that he devoted much energy to showing up the inadequacies of the Saussurean framework — particularly the famous antinomies — one could wish that he had recognized how deeply his own defacto differentiation of grammar, speech activity and text is at variance with the simple dichotomy langue/parole or, for that matter, his own code/message distinction, and how much it is superior to it. As it happens, it is perhaps only in his discussions of the Projection Principle that this logically irreducible trichotomy is given explicit expression.

Iconicity There is another thing to note about the Projection Principle. This has to do with Jakobson's understanding of poetic discourse as a complex of operations in the here-and-now which mediates between the paradigmatic system of the code and the syntagmatic chains of messages. Although he mentioned the Projection Principle repeatedly in his writings, nowhere did Jakobson dwell on the nature of this mediation. But thanks to his choice of the verb project (cf. (1)), the principle in fact defines a process of diagrammatization, in which relations of similarity or equivalence (in a grammar) become represented by relations of contiguity (in texts).

292

HENNING ANDERSEN

In his seminal paper on iconicity in language (1965), Jakobson gives examples of various dimensions of diagrammatization, though without explicitly distinguishing them. Some of his examples illustrate diagrams in grammar, as in (2), (2)

The signans of the plural tends to echo the meaning of a numeral increment by an increased length of the form. Cf. the finite verbal forms of the singular and the corresponding plural forms with longer endings: 1. je finis — nousfinissons,2. tu finis — vousfinissez,3. il finit — ils finissent', or in Polish: 1. znam (I know) — znamy, 2. znasz — znacie, 3. zna — znają. (p. 352)

where relations among the shapes of a set of desinences represent similar relations among their respective features of content. Others are textual diagrams, as when, say, the order of elements in a syntagm represents a sequence of events (e.g., Veni, vidi, vici) or the relative rank of participants (e.g., "the President and the Secretary of State attended the meeting"; p. 350), in the corresponding world of reference. But in this work the emphasis is explicitly on the systems of diagrammatization which are "patent and compulsory" in language, and not on the presumably facultative ones of poetic language. Poetic language is mentioned, and an example of figurat­ ive sound texture is offered, but this is not a central point in the article. And it is presented without any explicit statement that Jakobson recognized in this French example (3) (3)

Rumeur de la rue Réaumur ... roues ... murailles, trépidation d'immeubles .... (from Jules Romains, Les amours enfantines).

(where "a cunning intertexture" of r's and ra's reinforces a representation of an aggregate of wheels and walls and urban trepidation; p. 356f.) a diagrammatic representation of phonological equivalences, reflecting the workings of the Projection Principle. But this in fact is the semiotic essence of the Projection Thesis: it endeavors to explicate the aesthetic value of (poetic) texts as arising from the presence in such texts of diagrams of a specific kind. These can be briefly characterized in relation to the other kinds of textual iconicity Jakobson discussed — but did not attempt to delimit — as follows. Mimetic iconicity embraces those elements of a text which as images, diagrams or metaphors (I use these terms in the senses defined by Peirce, cf. 2.277) contribute to the text's representation of the world of reference it depicts. In his 1965 paper Jakobson exemplified (apart from grammatical diagrams) mainly mimetic textual diagrams. But in many other works he discussed the mimetic role of imaginai elements (sound symbolism) and of metaphoric elements (tropes in general).

EQUIVALENCE RELATIONS AND SYNTAGMS

293

Automorphic iconicity may serve as a term for the text internal equivalence relations — these too may be images, diagrams or metaphors — which were mentioned above. While mimetic textual icons represent parts of the text's matters of reference, automorphic icons have no reference outside the text; they reciprocally anticipate and recall one another. We will return to them in sec. 2.1-2. Projective iconicity, finally, is the diagrammatic representation of equival­ ence relations in the grammar by contiguity relations in a text. This is the type of textual iconicity which Jakobson saw as characteristic of the "set or focus on the message" in poetic language, and which he identified as the locus of a text's aesthetic value. As I will show in the following (sec. 1), such diagrams — which relate a text to the grammar with which it has been produced — are found also outside poetic language. Furthermore, such diagrams apparently cannot be considered vehicles of aesthetic content in and of themselves. I will argue (in sec. 2) that Jakobson in formulating the Projection Thesis failed to distinguish clearly between automorphism and projective iconicity. The term projective iconicity, which I have proposed here is insufficiently descriptive inasmuch as all diagrams can be viewed as projections. But it is motivated by the allusion it contains to Jakobson's formulation of the Projection Principle, and it is sufficiently neutral to serve as a generic term, being applicable with equal ease to strictly linguistic phenomena as well as to non-linguistic phenomena, both those which find expression through language and such as do not. In speaking about non-linguistic projection diagrams, however, the terms code and message, or grammar and text, are often awkward. Here I will use the general terms introduced by Hjemslev (1961) and speak of any (sub)system of categories as a paradigmatic and of any string of category terms as a syntagmatic. The term syntagmatic can, faute de mieux, serve both for activities (whether verbal or other) and for their results (e.g. a text).

1. The generality of the Projection Principle In the following pages, accordingly, we will consider a variety of illustrations of how equivalence relations in a paradigmatic are projected as contiguity relations in a syntagmatic so that the latter relations form diagrams of the former. The first few examples will be taken from the area of symbolic classifi­ cation. Systems of symbolic values can be manifested in a number of ways. Lévi-Strauss' analyses of myths can be mentioned as one source of excellent documentation one could draw on in this connection (see, for instance, the

294

HENNING ANDERSEN

examples in 1978). But since the question of the possible aesthetic value of projection diagrams is at issue, I deliberately avoid examples which could be construed as verbal art. Another source of illustration would be forms of dress, where one could draw on the excellent account of "la pensée bourgeoise" presented by Sahlins (1978). Here I present some examples that show how symbolic values are manifested in ritual. Examples like the ones I offer in the following can be found in abundance in the relevant anthropological literature (see, for instance, Needham 1973, 1979 with numerous further references), though — it should perhaps be noted — often not with the clear-cut differentiation of paradigmatics and syntagmatics which is emphasized here for the sake of the point to be made.

1.1 Ritual In a number of writings Fox (1971, 1973, 1974, 1975) has described major parts of the paradigmatic of symbolic values of the people of Roti, an island southwest of Timor in southern Indonesia. In (4) I list parts of this system, the binary oppositions that are most prominently relevant to the funeral rites of the Rotinese (cf. Fox 1973), which — with the omission of many essentials and elements of interpretation which would take up too much space — can be summarized as follows. (4)

Rotinese paradigmatic of symbolic values absence of ceremony outside forecourt woman west feet, tail north-south north left inauspicious

(1) (3) (5) (7) (9) (11) (13) (15) (17) (19)

:: :: :: :: :: : : : : :

ceremony (2) inside (4) house proper (6) man (8) east (10) head (12) east-west (14) south (16) right (18) auspicious (20)

The normal course of events in a funeral is as follows: The coffin is brought to the house of the deceased amid great uproar (2) and is put down outside the house, parallel to it at its west end. The coffin is then raised, carried under the roof (4), through the forecourt and up the ladder into the house (6), which stands on piles. The corpse is laid out in the men's (8) (the eastern (10)) half of the house, its head to the east (12-10) and feet to the west (14). The corpse is then placed in the coffin with the same orientation, and the mourners are admitted to the house (2). Subsequently the coffin is

EQUIVALENCE RELATIONS AND SYNTAGMS

295

brought down into the forecourt, where it may be rested on the east side (10), still with the same orientation (12-10). The deceased is then carried out of the house, feet first (11), and in this way the body is conducted in a noisy stampede (2) to the grave. At the side of the grave, which has been dug running east to west (14), the coffin is turned so that the corpse is headed westward (12-9), and in this position the coffin is lowered into the grave for its journey to the land of the dead in the west. A different rite is accorded those who have died a bad death — by drowning, falling from a tree, being stabbed, gored or bitten by man, waterbuffalo or crocodile, or in childbirth. The deceased who has died such an inauspicious death is not brought into the house proper, but either placed outside (3) or in the forecourt (5), but in this case on the west side (9). A woman who has died in childbirth remains in the women's half of the house (7), but the body is laid out north to south (13). No mourners are admitted to the house (1). Subsequently the coffin with her corpse is carried down from the house and rested on the west side of the house (9). The coffin is then carried out and to the grave, head first (12). The graves of the "bad dead" are dug running north to south (13), and the coffin is lowered into the grave headed north (15). It will be immediately apparent from the two resumes that each of these alternative funeral rites is composed largely of equivalent terms, as indicated by the even numbers (2-20) in the normal rite and the odd numbers (1-19) in the abridged, special rite applied to the "bad dead". The few deviations which stand out are well motivated. While we usually enter this world head first, the dead — in Roti as elsewhere — normally begin their journey out of this world feet first (11). It is as an inversion of this normal order of things that the "bad dead" are carried to the grave head first (12) in Roti. On the other hand the equivalence head (12) = east (10) which holds for the living (in Rotinese they are denoted by one and the same word) cannot hold for the buried who must travel west (9). By contrast the orientation of the "bad dead" conforms to their unhappy demise, 'north' (15) and 'left' (17) being denoted by one and the same word, and the left being considered inauspicious (19). To explicate these well-ordered syntagmatics in relation to the system of values which they manifest one could hardly imagine a better notion than the Projection Principle. The two alternative rites derive their ordered character and stand in such clear opposition to each other precisely because each is composed of homogeneous constituents, the normal one of unmarked terms, the special one of marked terms. Before we consider the question of the aesthetic value of these syntagmat­ ics, let us look at just one more example. It is similar to the preceding one in that it requires reference to the lay-out of a building. It differs from it

296

HENNING ANDERSEN

primarily by being composed of conjoined rather than disjunctive com­ plementary syntagmatics. In this example, I describe the church going ritual as it was practiced in Denmark from the introduction of Christianity till long after the Refor­ mation. It manifests parts of a paradigmatic of symbolic values which I list in (5). I choose for this example the church at Tapdrup in Jutland, typical in most respects of the several hundred romanesque churches in the country (cf. Trap 1962). The church is oriented west-east, the tower at the west end, the apse, in which the altar is located, at the east end. The southern (2) wall is made of sandstone ashlars (4); set in it are the south entrance at the west end — a sizable (10), arched portal flanked by decorated columns and topped by a tympanum (6) — and three fairly large (10) arched windows. (5)

Medieval paradigmatic of symbolic values north (1) : south (2) field stones (3) : ashlars (4) plain (5) : ornate (6) nature (7) : culture (8) small (9) : large (10) woman (11) man (12) left (13) : right (14) circuitous (15) : direct (16) shade (17) : sun (18) cool (19) : warm (20) moist (21) : dry (22)

The north (1) wall is built of rough field stones (3). Set in it are the north entrance, a plain (5) arched doorway (now walled up) smaller (9) than the south portal, and three windows likewise smaller (9) than their counterparts on the south wall. Formerly, when the congregation went to worship, the men (12) would enter through the south door (2) and take their places in the nave standing on the right hand (14) side of the aisle, whereas the women (11) entered by the north door (1) and stood on the left (13) side of the aisle. For the churchgoer who approached the church on a fair day a number of simple sensory experiences would be added to these complementary syntagms. The women, at least, who took the circuitous path (15) around the west end of the church would experience the palpable differences in light (17), temperature (19), and humidity (21) between the north and south side (as I can testify). These two entirely homogeneous syntagms require little comment from the point of view of interpretation. One may note that the difference in the size of doors and windows between the two sides correlates with the usual

EQUIVALENCE RELATIONS AND SYNTAGMS

297

difference in stature of women and men, and that the ornamentation of the men's entrance includes phallic elements. The different choice of building materials can be interpreted as an additional manifestation of the opposition plain (5) vs. ornate (6) in the entrances. But perhaps it is better seen as a manifestation of the opposition nature (7) vs. culture (8), which is commonly correlated with that of woman (11) vs. man (12). Both in the use of field stones (3) for the north wall and in the plainness (5) of the north entrance one can see a suppression of elements that would seem naturally called for in the grandest structure in the village, the most concentrated expression of its wealth and culture. These are marked terms like the other terms that define the women's path to worship. Like the previous example this seems to be a lucid and obviously valid example of the Projection Principle. Symbolic terms which are equivalent in the paradigmatic by virtue of their markedness value are selected and combined into two contrasting syntagmatics, each of which by its homo­ geneity can be recognized as a diagram of the respective equivalence relations. If one asks about the aesthetic value of such homogeneous syntagms, the search for an answer, I think, must start with the fact that their function is not aesthetic. It appears that in general the categories of a value system allow two maximally ordered, and from this point of view, optimal, concat­ enations — one composed entirely of marked terms, the other composed entirely of unmarked terms of the same categories. In ritual, it seems, these two possibilities are exploited functionally as a way of giving expression to culturally important categories — in Roti, the distinction between inauspicious and normal death, which is central to the Rotinese system of beliefs about the after life; or, in medieval Denmark, the socially central distinction between woman and man. It is not the pre­ ordained, regulated character of the syntagms as such that counts, but the equivalence of the contiguous terms with the category terms that govern their selection. They are thus genuinely informative of the way in which these respective categories are integrated in the paradigmatic which the rituals manifest. Being prescribed by normative social rules, such rituals are ideally totally predictable. Of course allowance may be made at certain (stipulated) points in such a syntagmatic for aesthetic elements to be introduced or superim­ posed. But the aesthetic value of the prescribed chains of homogeneous symbolic values can safely be pegged at zero. Their total predictability alone suggests this. The fact that they are stipulated by norms of proper conduct — prescriptions for doing the right thing or for doing things right — means that the extent to which a given ritual syntagmatic conforms to the norms is strictly speaking a question of ethics, not of aesthetics.

298

HENNING ANDERSEN

1.2 Grammar The second set of examples I would like to mention illustrates the working of the Projection Principle in language, but at levels of structure where the possibility of aesthetic values would seem to be entirely excluded. Several linguists have observed that in phonetic change, usually, allophones that can be considered marked in some respect arise in specific environments which either by their phonetic characteristics alone or by their specificity in some other regard can be considered marked. Mares' statement that in phonetic alterations the change is usually from unmarked to marked (1952) can be read as an understanding of this generalization. Schachter 1969 gives an explicit formulation of it. In Andersen (1968) I proposed the term markedness assimilation for this phenomenon, illustrat­ ing it with a handful of examples in addition to the problematic change of IE * s to * ś after i, u, r,  in Baltic and Slavic which was the focus of attention. Here are some more or less obvious examples; vowels may become nasalized before nasal consonants (thus in dialects of English); flatted vowels may become acute when contiguous to sharped consonants (thus in Russian); liquids may become voiceless after tense obstruents (thus in Danish); dental stops may become strident before diffuse vowels (thus in Japanese); acute consonants may become compact before the glide j (thus in Serbocroatian dialects), (p. 175). A later study offered a more explicit account, pointing out that the marked contiguous environments may be either sequential or simultaneous and may be defined with reference either to segmental or suprasegmental phonetic features or to more abstract properties such as position within the syllable or word, and recognizing the phenomenon of markedness reversal (Andersen 1972, more recently discussed as "local markedness" by Tiersma 1982). There, too, I drew a parallel between these apparently universal phonological phenomena and similar phenomena in the distribution of morphological categories and — beyond language — in the concatenation of symbolic values, for instance, in ritual (with references to Fox 1973, Needham 1967 and 1971, Rigby 1968, and Turner 1969). In Andersen (1980) I offered a large number of examples illustrating markedness assimilation with morphophonemic alternations. It can now be recognized that markedness assimilation is a manifestation of the Projection Principle. In whatever way we choose to write the syn­ chronic rules of a grammar, it is clear that in essence an allophonic variation or a morphophonemic alternation is stipulated as a correlation between two paradigms, a paradigm of the variants involved and a paradigm of their respective privileges of occurrence. Both paradigms are asymmetrical. A

EQUIVALENCE RELATIONS AND SYNTAGMS

299

paradigm of, say, allomorphs comprises one or more derived allomorphs and one basic allomorph. The corresponding paradigm of their privileges of occurrence comprises one or more narrowly specified environments and an elsewhere environment. In speech the operations of selection and combi­ nation are executed so that derived allomorphs are assigned to narrowly specified environments and the basic allomorph to the elsewhere environ­ ment(s). In other words, the terms of the two paradigms are disposed in relations of contiguity in such a way that these diagram relations of equival­ ence in the grammar. To conserve space I will not elaborate this point by rehearsing individual examples. I think, too, that I can limit myself to mentioning, as I did in (1972), that a parallel account can be given of diverse phenomena of gram­ matical government in which specified conditions require the selection of marked categories while elsewhere conditions call for the use of an unmarked category. Compare also Haiman's discussion of mood in Hua (1975) and of split ergativity and other phenomena (in 1980: 528ff). The parallelism between these linguistic examples and the rituals described in sec. 1.1 should be clear. Admittedly there are differences between them. In the performance of a ritual, for instance, the participants seem to select and combine the equivalent symbolic terms deliberately. Often they not only have the requisite technical knowledge, but can also give the rules for it and can sometimes explicate these rules by reference to an explicit body of beliefs (cf. Fox 1973:361, Rigby 1973:271). By contrast, the technical knowledge behind the linguistic examples is normally entirely implicit, and the selection of, say, the proper allomorph for a given environment usually completely automatized. But this difference is not an absolute one. We know that whenever allomorphs or allophones are invested with symbolic value (as when they are social or stylistic variables), their selection is — if not a matter of conscious deliberation — at least an intentional choice which speakers may be able to rationalize post hoc. What the linguistic examples have in common, and what they have in common with the rituals, is that they illustrate activities that are strictly governed by normative rules. This fact suggests that the proper purview of the Projection Principle which they all reflect might be normative rule governed semiotic behavior in general. With this hypothesis in mind, let us consider a final example.

1.3 Pragmatics For my last example I choose the observation, made by a number of scholars, for instance Hopper & Thompson (1980, 1982), that universally a

300

HENNING ANDERSEN

number of semantic features are employed in verbal communication in such a way that they collectively serve to characterize material presented in discourse as foregrounded or backgrounded. Hopper & Thompson present a partial paradigmatic of such features which contribute to what they call the transitivity — higher or lower — of individual sentences (presented slightly modified and without their generic category names in (6)). They note that the way in which these conceptual features are grammaticized differs from language to language, but they argue that languages universally possess morphosyntactic devices which express the degree of transitivity of clauses. (6)

Paradigmatic of conceptual categories relating to transitivity 1 participant non-action atelic non-punctual non-volitional negative irrealis Agent low in potency Object non-affected Object non-individuated

low (1) (3) (5) (7) (9) (11) (13) (15) (17) (19)

high 2 or more (2) action (4) telic (6) punctual (8) volitional (10) affirmative (12) realis (14) Agent high in potency (16) Object totally affected (18) Object highly individuated (20)

It stands to reason that such features of meaning would be selected for expression and combined in speech largely according to the communicative intentions of speakers. But the authors show, on the basis of an examination of a large sample of typologically diverse languages, that whenever there is a constraint in a language on the combination of these features in the morphosyntax or semantics of a clause in the form of an obligatory pairing of two transitivity features, "the paired features are always on the same side of the high-low transitivity scale" (1980:254). Thus, for example, in dozens of languages perfectivity (6) correlates with definiteness of the direct object (20), but no languages have been found in which perfectivity (6) correlates with indefmiteness (19) or in which imperfectivity (5) correlates with defi­ niteness (20) (p. 279). To the extent that such examples of covariation are stipulated by rules of grammar, they can evidently be compared to the examples in sec. 1.2 and taken as illustrations of how the Projection Principle applies to grammatical categories. But the wider findings concerning transitivity clearly transcend the normative rules of grammar. Despite the degree of freedom with which terms of these categories can in principle be combined, it is found that in discourse there is an overwhelming statistical predominance of high-tran-

EQUIVALENCE RELATIONS AND SYNTAGMS

301

sitivity elements in foregrounded material and of low-transitivity elements in backgrounded material (p. 288ff.). Hopper & Thompson explain this finding by pointing to the essentially pragmatic nature of the grounding distinction and recognizing "that grounding itself reflects a deeper set of principles — relating to decisions speakers make, on the basis of their assessment of their hearers' situation, about how to present what they have to say" (p. 295). Wallace (1982) has argued that the findings of Hopper & Thompson and several other scholars who have made comparable observations must be interpreted in the light of the figure-ground distinction of Gestalt psychology — though he notes, with due emphasis, that linguistic structure cannot be reduced to this contrast. It seems that this psychological perspec­ tive is entirely appropriate, and that it can now be applied to much greater advantage than fifty years ago when it was first appealed to as an expla­ nation of markedness relations in phonology (cf. Trubetzkoy 1936). But the findings of Hopper & Thompson do not concern perception and the ordering of experience, but the complementary matter of the ordering or ranking of referential material for encoding in speech. And it is from this point of view that their results are relevant in a discussion of the Projection Principle. What their results suggest is that in the casting of conceptual represen­ tations — prior to the action of any linguistic formation rules — humans, regardless of what language they speak, select and combine conceptual categories — admittedly with a certain freedom and in accordance with their communicative intentions (this is indicated by the merely statistical nature of the observed covariation), but by and large in an orderly fashion — so that the resulting linguistic representation by its clustered distribution of the terms of the corresponding language particular semantic and morphosyntactic categories diagrams the distinction between backgrounded and foregrounded material in the speaker's conceptual representation. The question to ask now is whether this apparent manifestation of the Projection Principle is compatible with the hypothesis stated at the end of sec. 1.2: is the universal differentiation of backgrounded and foregrounded information in discourse governed by normative rules? I mentioned above that to some extent the covariation of the terms of transitivity categories may be stipulated by grammar rules, that is, by language particular norms. But the universal projection of the grounding distinction as contrasting clusterings of transitivity equivalents is apparently not (or not entirely) a matter of culture specific norms. One can view it as rule governed, but one must then acknowledge that it follows rules that are not language particular but, as far as we can judge, part and parcel of a more general communicative competence.

302

HENNING ANDERSEN

1.4. Discussion In this chapter I have presented evidence that suggests that Jakobson's Projection Principle, far from being a specifically linguistic principle, can be found manifested in a variety of types of semiotic behavior. It is reflected in non-linguistic activities such as the rituals examined in sec. 1.1. It accounts for grammatical phenomena such as allophonic variation, morphophonemic alternation, and certain kinds of government (1.2), as well as the cases of obligatory co-occurrence of transitivity equivalents noted by Hopper & Thompson (1.3). And it is the device that ensures that the different grounding of material in conceptual representations is reflected in speech in an orderly fashion (1.3). In the different types of semiotic behavior we have looked at, the Projec­ tion Principle can be seen to serve one and the same function. In ritual, chains of equivalent symbolic values achieve a practical end (e.g. a funeral) and at the same time serve to make explicit how some major value category is integrated with other categories of the symbolic system. In phonetics, chains of equivalent phonetic values promote ease of articu­ lation and/or perception and besides carry information about the ranking of phonetic features in the phonological system (cf. Andersen 1979). Morphophonemic alternants index the domains in which they apply and concurrently serve to highlight the diverse phonological or morphological categories by which they are conditioned, that is, with which they are integrated in the grammatical system. The grammatically obligatory co-variation of transitivity categories subserves coherent communication and at the same time carries information about the language-particular system of grammatical categories. The clustering of transitivity equivalents in discourse, finally, has the essential mimetic function of diagramming the distinction between backgrounded and foregrounded parts of the content of a text and achieves this end by reflecting a set of conceptual categories that can be presumed to be shared by all speakers. In each case, then, the Projection Principle produces syntagmatic represen­ tations of relations in the paradigmatic of categories to which it applies. In the first four instances it thereby serves to reaffirm the constitution of some subparadigmatic of a cultural system. Only in the last example can this constitution be taken for granted. The presumably universal set of concep­ tual categories relating to transitivity serves to ground — no pun intended — the communicatively essential distinction between backgrounded and foregrounded material. The last-mentioned example can be used to shed light on the Projection Principle in two ways.

EQUIVALENCE RELATIONS AND SYNTAGMS

303

First, the statistical nature of the results presented by Hopper & Thomp­ son shows that the Projection Principle — which would ideally produce purely homogeneous syntagmatics — interacts with other factors, presum­ ably first and foremost the specific matters to be communicated. In practice it is no different with ritual or certain kinds of grammar rules. In ritual, for example, practical considerations may force the participants to alter or omit certain normative elements (cf. Rigby 1973: 271, n. 27). In verbal communi­ cation, normative rules of pronunciation, of morphophonemics, or of gram­ matical government may be left unapplied, for instance, to safeguard the correct identification of an individual lexeme. What this shows is that the Projection Principle applies as a sort of "elsewhere" or "otherwise" principle. It describes an ideal mapping from the paradigmatic axis to the syntagmatic axis which, when it is not overridden by the demands of practical life, produces syntagmatics wholly composed of equivalents and therefore totally uninformative of anything but the paradigmatics whose relations they diagram. This automatic character of the Projection Principle perhaps explains such innovations in normative behavior — I would call them deductive; cf. Andersen (1973) — as the elaboration of ritual by the accretion of additional symbolic elements to ritual syntagmatics, or the accumulation of phonetic detail rules in phonology. These are phenomena that clearly reflect a universal tendency to expand syntagmatics by additional equiv­ alences, a tendency for which a suitably general, semiotic explanation has been wanting, Secondly, the example from pragmatics shows once again that the Projec­ tion Principle has nothing to do with aesthetics. It is manifested in a variety of forms of activity whose ordered character it explains. Its highly diverse manifestations, and the fact that it regulates the selection and combination of such universal conceptual categories as those involved in grounding, show clearly enough that it is a principle of vast generality and make it very likely that it is inborn. No doubt it should be of interest to psychologists. As mentioned, Wallace (1982) has convincingly shown that the identical relations in different concep­ tual categories can be understood from the vantage point of Gestalt psy­ chology. But this is only part of the picture. The equivalences of terms belonging to different categories is a separate phenomenon, which must be accounted for by a theory of association. The Projection Principle as such, however, which concerns the mapping of mental structures of the most various kinds into ordered behavior must belong to the province of the psychology of activity. It is of relevance to the theory of cognition as well, as we will see at the very end of this paper.

304

HENNING ANDERSEN

2. Jakobson's theories of aesthetics The exposition in the preceding pages is obviously relevant to some of the questions that were raised in the beginning of this chapter. We have seen a fair amount of evidence that demonstrates the importance of the Projection Principle. But the principle does not apply only in poetic language. It is found to apply to levels of grammatical structure that are not specific to any one of the different functions of language. And it applies to different levels of conceptual structure that lie beyond the pale of gram­ mar. There is good reason to suppose that the Projection Principle is an extremely general ordering principle that regulates all kinds of semiotic behavior. As for Jakobson's Projection Thesis — the claim that it is the Projection Principle that is the source of the aesthetic value of poetically formed messages — the validity of this thesis has been put in doubt by the repeated observation that in the other areas where the Projection Principle is at work, it does not confer any aesthetic value on the resulting syntagmatics and by the consideration that in principle it cannot be a source of aesthetic value inasmuch as it merely regulates action. These findings may seem disturbing when one considers that the Projection Principle has served as part of the explicit theoretical foundation for methods of poetic analysis which have proven highly successful. It certainly seems as if something is amiss. To return to the image used above, we have, on one hand, a proven pudding and, on the other, a recipe with directions for an entirely different dish. In the following pages I will show that there is in fact a discrepancy here between theory and practice, as these findings seem to indicate. To pinpoint this discrepancy, however, I will examine another part of Jakobson's theory of semiotics to see what light it might shed on his Projection Thesis.

2.1 The theory of introversive semiosis We are in the fortunate position of having from Jakobson's hand, besides the Projection Thesis another, more general, theory of aesthetic semiotics. It is on the surface very different from the Projection Thesis. But when one looks at it at close range, the two theories turn out to be related in interesting ways. I am referring to Jakobson's theory of "introversive semiosis" (1968: 700, 704), sketched with the aim of explicating the aesthetic values primarily in non-verbal arts — music, glossolalic poetry, nonfigurative painting and sculpture. The reader who is familiar with this theory will recall that it arose out of

EQUIVALENCE RELATIONS AND SYNTAGMS

305

an attempt to analyse Peirce's concepts of icon, index and symbol into bundles of binary features. This effort, which I will not try to praise, produced — besides a misconstrual of Peirce's trichotomy — a fourth putative sign type, assumedly not envisaged by Peirce, viz signs based on "imputed similarity" (or "similitude imposée" as it was later called in French; 1979: 16). These signs Jakobson found in the elements of equivalence or contrast which anticipate or recall one another in the temporal sequences of, say, music, or which cross-reference one another in non-figurative paint­ ing, and which by their interconnections are bearers of aesthetic content in non-verbal works of art (and "to some extent also in poetry and in the bulk of representational visual arts"; thus 1971: 715). This is a very interesting theory, which invites comment from a number of points of view, only some of which will be mentioned here. First it must be said that viewed in Peircean terms there is, of course, no new, special, hitherto unnoticed type of sign involved here, no "imputed" rather than "factual similarity" (or "similitude imposée" rather than "effec­ tive"). By their actual, recognized similarity, the constituents of the iterative structures in question are icons of one another. By their contiguity, they are "really connected". This fact, together with their similarity, makes them fit to function as indexes of one another. Viewed in terms of their sign-object relations — which is the parameter relevant to the icon-index-symbol tri­ chotomy — these signs are of a common, garden variety type: they are iconic indexes. It is by the nature of their objects that they are sui generis. While the mimetic icons of a work depict, reflect or suggest elements in the matters the work represents, and projective diagrams represent relations in the code the work instantiates, these iconic indexes have their objects inside the very work in which they occur: they represent, and are in turn represented by, one another. Each set of such reciprocal iconic indexes forms an automorphic structure — to borrow a term from the theory of symmetry. It is pretty clear — as Jakobson recognized — that such structures play a cardinal role in all forms of art. For one thing, their presence or prevalence in a work of art serves to constitute it as relatively closed and internally cohesive — the reciprocity of the iconic indexes involved lend the work a character of "introversion" or, more precisely, automorphism. For another, and more importantly, each set of recurrent, equivalent elements can be experienced as a single unitary whole. This is possible because the members of the set are really connected, and because their similarity makes them appear to be of a piece. When they are so grasped, the parts of any automorphic structure together constitute a single automorphic sign (strictly a sort of diagram), which the receiver can assign an aesthetic sign-object (in Peirce's sense) and, in so doing, experience with a feeling of satisfaction, pleasure, or delight (cf. Schmid 1977: 11Of.).

306

HENNING ANDERSEN

2.2 Two theories or one? Even though Jakobson, practically as an afterthought, conceded the import­ ance of such automorphic signs also for verbal art (cf. 1971: 715), his two theories of aesthetics are almost complementary. One was conceived to account for verbal art, the other primarily for non-verbal arts. Furthermore, the two theories were conceived from different points of view. The Projection Thesis traces properties of a poetic text back to the operations by which they were created. The theory of introversive semiosis, by contrast, takes the properties of non-verbal works of art as given and tries to explicate their interpretation from the receiver's point of view. In other words, one theory focuses on encoding, the other on decoding. This difference in point of view between the two theories perhaps explains why they were not synthesized in Jakobson's work, for it does obscure their essential unity. Their complementarity, however, becomes evident the moment one looks at what they leave unsaid. Note that the focus on the receptive process in regard to works of music and non-figurative visual art leaves unconsidered the paradigmatics behind such works. This is a fateful lacuna in a theory that operates with workinternal equivalences, for it makes it present the recognition of all such equivalences as a purely subjective matter — the automorphic structures appear to be imputed (imposées) by the interpreter. Yet there can be no doubt that in these non-verbal arts, no less than in verbal art, the production of any work involves operations of selection and combination applied to repertories of semiotic means. The very recognition of equivalence relations in the automorphic structures of any work of art implies reference to paradigmatics or sets of conventions that define these equivalences. Such paradigmatics may be species general (as putatively in the case of glossolalia); they may be medium-specific (e.g., limited to pictorial arts); they may be culture-specific (e.g., characteristic of Western music), period-specific (e.g., Romantic music), or idiosyncratic (e.g., characteristic of Brahms); or they may be ad hoc, work-specific (cf. Schmid 1977: 31). Such different layers of conventions may be weighted differently in different artistic media, periods, genres, or works. But in any individual work one can discern conventions (with work-specific conventions as the limiting case) which define what in that particular work counts as equivalent. In brief, from the point of view of production, an explication of auto­ morphic signs in non-verbal art requires reference to operations of selection and combination and to paradigmatics in a way quite analogous to the situation in, for instance, poetic discourse. This being so, it is obviously of central importance, in a theory of art, as much as in the analysis and evaluation of any individual work of art, to determine to what extent such

EQUIVALENCE RELATIONS AND SYNTAGMS

307

work-internal equivalences should be viewed against the background of order relations generated by the Projection Principle. In other words, an attempt must be made to separate out the equivalences that instantiate universal, medium-specific, period-specific and other pre-existing paradig­ matics on one hand from the work-specific paradigmatics on the other. The former constitute, so to say, the aesthetically inert carrier wave against which alone the ad hoc paradigmatics created for the individual work of art can be apperceived and appreciated. In the other theory — the Projection Thesis — the opposite focus on the production of poetic discourse apparently left unasked the question of how equivalences in relations of contiguity might be interpreted. Actually the reference in this theory to the (linguistic) code effectively limits its explana­ tory power — if it is taken au pied de la lettre — to the identification of code-given (linguistic) equivalences and the recognition of such equivalences as diagrams of relations in the (linguistic) code. When viewed from the point of view of the receptive process, however, all the identifiable equivalences in verbal art seem to fall into two categories. There are those that are mere reflections of equivalence relations in some part of the complex of paradigmatics that are instantiated in verbal art. These are projective diagrams. They are as aesthetically inert as those we examined in sec. 1. And then there are those which function primarily as automorphic diagrams, that is, are conducive to being assigned aesthetic interpretants. Of course these equivalence relations cannot be identified without reference to various linguistic and extra-linguistic paradigmatics. They reflect the paradigmatics they instantiate and can be so interpreted. But they are effective because they can be subsumed into automorphic signs that have aesthetic value. It is this second category of equivalence relations that conforms to Jakobson's well-known definition of the poetic function of language as implying a set or focus on the message. For the automorphic signs require precisely such a focus — of the reader, who must correlate elements in the sequence to establish the relevant automorphic signs and infer their aesthetic value; and of the author, who must manipulate a large number of paradigmatics, in accordance with his mimetic, expressive, and conative intentions, in such a way as to create those text-internal signs which give the work its status of aesthetic object. The projective diagrams, by contrast, merely reflect relations that inhere in the respective paradigmatics. As a consequence they yield information about these relations — as I suggested in sec. 1.4 — and can be exploited by anyone who adopts a set on the code — the individual who is in the process of being socialized, the linguist or anthropologist, or the literary scholar who is interested in uncovering the grammatical, artistic, or more

308

HENNING ANDERSEN

general cultural conventions the work instantiates. But for members of the community in which the work is presented, the projective diagrams merely serve to reaffirm the received paradigmatics of values. From the preceding I think it emerges that Jakobson's two theories of aesthetics complement each other and are indeed reconcilable. But at the same time it appears that his Projection Thesis was formulated in too general terms: it aimed to explicate the aesthetically effective recurrences in (literary) works by relating them to the systems of conventions on which they depend. But by its overly general wording it takes in as well the aesthetically insig­ nificant projective diagrams. Here, then, is the reason for the apparent discrepancy between theory and practice. The thrust of Jakobson's Projection Thesis was to emphasize the dependence of a great many types of automorphic signs in verbal art on the mainly linguistic paradigmatics that are their basis. His own work in poetic analysis demonstrated the fruitfulness of this approach to poetic language, which complements other approaches to poetry, and which — quite apart from its historical relationship with a particular aesthetic, that of the Russian avant-garde — has proven applicable to the most various poetic and cultural traditions. The wide acclaim this theory has received is consonant with its aims. But the fact that a great variety of devices in verbal art cannot be defined without reference to the grammatical paradigmatics they instantiate does not mean that they have been projected from this paradigmatic in any but a very loose sense of the word project.

3. Conclusion In this paper I have examined one of Jakobson's most widely cited contri­ butions to the theory of poetics and have identified some shortcomings of his formulation of the Projection Principle. The Principle does not cover the full breadth of Jakobson's conception, for it was formulated as an account of the poetic function of language, whereas he himself found it reflected in verbal art also on levels of structure that are beyond the pale of grammar. A more serious criticism is the fact that the Projection Principle does not in any way account for certain kinds of work-specific equivalences, the elements of automorphic iconicity. These appear to fall under a different theory of Jakobson's, the theory of introversive semiosis. But this theory does not, in the form Jakobson gave it, say anything about the dependence of automorphic structures in a work of art (verbal or otherwise) on a repertory of artistic or other means of expression which would define the

EQUIVALENCE RELATIONS AND SYNTAGMS

309

equivalences in question. It is only when the two theories are recognized as complementary and are integrated, and when, in addition, the importance of work-specific conventions is duly acknowledged, that it becomes possible to mend this shortcoming of Jakobson's theories. The most serious difficulty with Jakobson's formulation of the Projection Principle arises from his use of the word project. This verb does not even begin to do justice to the creative process of poetic discourse, but rather begs the question of how an artist in fact contrives to perform the operations of selection and combination in such a way as to create an aesthetic object. What is worse, it fails to distinguish two crucially different kinds of equival­ ence relations in any poetic text, those that are generated by the Projection Principle, and which have no particular aesthetic value, and those which result from the poetic process properly speaking. This weakness, which seems most unfortunate, however, has turned out to be very fortunate in two respects. In the first place, without the strikingly simple and general term project, the Projection Principle would hardly have received the attention it has enjoyed among literary scholars. The very simplicity of the notion of projec­ tion has helped promote the aim of focusing attention on the code depen­ dence of work-internal equivalence relations and has served to put off for the time being debates on the particulars of this dependence, debates which would perhaps be premature. From a pragmatic point of view, I think one must concede that the use of the word project in the statement of the Projection Principle has been fortunate indeed. But it has been just as fortunate from a second point of view. Without the bold generalization implicit in the notion of projection, Jakobson's formulation would not have suggested the true, remarkably broad extension of the principle, which was sampled in sec. 1 above. As I have shown, his formulation adumbrates a principle of ordered action which is relevant to the most diverse aspects of semiotic behavior. It is, in fact, more than that. Since man is the measure of all things and naturally models his understand­ ing of the world on his insight into human affairs, the Projection Principle describes what for most people is the order of the world. It is in fact part and parcel of the traditional wisdom of the folk and codified in such proverbs as "Qui se ressemble, s'assemble", "Gleich und gleich gesellen sich gern", "Birds of a feather flock together", and so forth. This is not to say that the Projection Principle merely restates what the folk has known for millennia. For it is one thing to know that things have avoir-du-poids and quite another to formulate a law of gravity. What Jakob­ son's felicitous formula has turned out to capture is a fundamental semiotic principle which explains orderly rule governed activity and therefore under­ lies our preconception of the natural order of things.

310

HENNING ANDERSEN

To grasp the full significance of the Projection Principle, however, we should take it one step further and recognize that like any other deductive principle, the Projection Principle has consequences for abduction. It entails, in point of fact, a law of cognition that allows humans to hypothesize relations of equivalence on the basis of observed relations of contiguity. This converse of the Projection Principle must be the very foundation of our ability to infer their manifestations in human behavior, the multifarious paradigmatics of values of which human cultures are composed from and to integrate them into coherent patterns. Viewed from this complementary point of view, the Projection Principle opens up a perspective of such vast scope — and so far from Jakobson's original conception — that it is best left for exploration at another time.

References Andersen, Henning. 1968. "IE *s after i, u, r,  in Baltic and Slavic". Acta Linguistica Hafniensia 11. 171-190. . 1972. "Diphthongization". Language 48. 11-50. . 1980. "Morphological Change: Towards a typology". Historical Morphology, ed. by J. Fisiak, 1-50. The Hague: Mouton. Armstrong, Daniel & C.H. van Schooneveld. eds. 1977. Roman Jakobson. Echoes of his Scholarship. Lisse: Peter de Ridder Press. Bauman, R. & J. Sherzer eds. 1974. Explorations in the Ethnography of Speaking. Cambridge: Cambridge University Press. Brown, E.K. 1950. Rhythm in the Novel. Toronto: University of Toronto Press. Coseriu, Eugenio. 1974. Synchronie, Diachronie und Geschichte. Das Problem des Sprachwandels. München: Wilhelm Fink. Fox, James. 1971. "Semantic Parallelism in Rotinese Ritual Language". Bijdragen tot de Taal-, Land- en Volkenkunde 127. 215-255. . 1973. "On Bad Death and the Left Hand". In Needham, ed., 1973. 342-368. . 1974. "Our Ancestors Spoke in Pairs: Rotinese views of language, dialect and code". In Bauman & Sherzer, eds., 1974. 65-85. . 1975. "On Binary Categories and Primary Symbols". In Willis, ed., 1975. 99-132. . 1977. "Roman Jakobson and the Comparative Study of Parallelism". In Armstrong & van Schooneveld, eds., 1977. 59-90. Haiman, John. 1975. "Neutralization and Markedness Assimilation: Future and subjunctive in Hua". Oceanic Linguistics 14. 119-127. . 1980. "The Iconicity of Grammar: Isomorphism and motivation". Language 56. 115-540. Hopper, Paul J., ed. 1982. Tense-Aspect: Between semantics and pragmatics. Containing the Contributions to a Symposium on Tense and Aspect, UCLA, May 1979. Amsterdam/Philadel­ phia: Benjamins. Hopper, Paul J. & Sandra A. Thompson. 1980. "Transitivity in Grammar and Discourse". Language 56. 251-299. Jakobson, Roman. 1960. "Linguistics and Poetics". Style in Language, ed. by Thomas A. Sebeok, 350-377. Cambridge, Mass.: MIT Press. . 1965. "Quest for the Essence of Language". In Jakobson 1971. 345-360.

EQUIVALENCE RELATIONS AND SYNTAGMS

311

. 1971. "Retrospect". In Jakobson 1971. 711-724. . 1971. Selected Writings, vol. II. The Hague: Mouton. Jakobson, Roman & Krystyna Pomorska. 1983. Dialogues, tr. by Christian Hubert. Cambridge. Mass.: MIT Press. Koch, Walter A. 1982. "Poetizität: Das Triviale des Triadischen". Poetica 14. 250-298. Leach, Edmund R., ed. 1968. Dialectic in Practical Religion. Cambridge: Cambridge University Press. Lévi-Strauss, Claude. 1976. Structural Anthropology, vols. I—II. tr. by Monique Layton. N.Y.: Basic Books. Mareš, František Vaclav. (1952) 1969. Diachronische Phonologie des Ur- und Frühslavischen ( = Slavistische Beitráge 40). München: Otto Sagner. Needham, Rodney. 1967. "Right and Left in Nyoro Symbolic Classification". Africa 37. 425-452. Reprinted in R. Needham, ed., 1973. · . 1973. Right and Left. Essays on Dual Symbolic Classification. Chicago/London: University of Chicago Press. . 1979. Symbolic Classification. Santa Monica, Ca: Goodyear Publishing Co. Pagnini, Marcello. 1979. "The Science of Literature and Literary Criticism: A survey of international contributions in the field of poetry". A Semiotic Landscape, ed. by Seymour Chatman et al., 96-102 ( = Approaches to Semiotics). The Hague: Mouton. Peirce, Charles Sanders. 1931-1956. Collected Papers of Charles Sanders Peirce, vols. I-VIII, ed. by Charles Harts-horne & Paul Weiss. Cambridge. Mass.: Harvard University Press. Rigby, Peter. 1968. "Some Gogo Rituals of Purification: An essay on social & ritual categories". In E. Leach, ed., 1968. 153-178. Sahlins, M.D. 1978. Cultural and Practical Reason. Chicago: University of Chicago Press. Schachter, Paul. 1969. "Natural Assimilation Rules in Akan". UAL 35. 342-355. Schmid, Wolf. 1977. Der ästhetische Inhalt. Zur semantischen Funktion poetischer Verfahren (= Utrecht Slavic Studies in Literary Theory, 1). Lisse: Peter de Ridder Press. Tiersma, Peter Meijes. 1982. "Local and General Markedness". Language 58. 832-849. Todorov, Tzvetan. 1977. "Poétique générale". In D. Armstrong and C. H. van Schooneveld, eds., 1977. 473-484. Trap, J.P. 1962. Danmark, bd. 17: Viborg Amt, 5. udg. Copenhagen: Gad. Trubetzkoy, N.S. 1936. "Essai d'une théorie des oppositions phonologiques". Journal de psychologie 33. 5-18. Turner, V.W. 1969. The Ritual Process: Structure and anti-structure. Chicago: Aldine. Wallace, Stephen. 1982. "Figure and Ground: The interrelationships of linguistic categories". In P. Hopper, ed., 1982. 201-226. Willis, R., ed. 1975. Interpretation of Symbolism. London: Malaby Press.

Invariance in Grammar, Variation in Discourse Discussion Flora Klein-Andreu State University of New York at Stony Brook Discussion of the paper by Linda Waugh I found this paper very interesting and was especially grateful for the abundant data, taken from actual language use, which the author offers in support of her arguments. One thing that emerges very clearly from this presentation is the impossi­ bility of adequately validating individual grammatical meanings in iso­ lation — that is, independent of an equally thorough analysis of the other meanings with which they most closely compete. In this instance this would presumably involve (at the very least) the various forms that might be used in French to refer to events in the past. This is not intended as a criticism; we are dealing with work in progress, being continued precisely in the direction suggested. Rather, it is intended to point out what I consider to be one of the most important conclusions this research suggests at this particular stage — precisely because it is so thorough and well-documented — and one which also seems very much in keeping with the thinking of Roman Jakobson (as interpreted, for example, by Waugh; see e.g. 1977: 61). The need for analysis of the "system" as a whole is perhaps most obvious in appealing to such concepts as "markedness". But actually it seems to affect all aspects of the analysis — among others the use of actual examples to substantiate the semantic analyis of particular forms. This is because any analysis based on the assumption that a single meaning can lead to different interpretations, in different contexts, cannot appeal only to the obvious "appropriateness" of the meaning for the particular interpretation intended, since meaning and interpretation will necessarily be more or less different (the actual interpretation being considerably richer and more precise than the meaning). It therefore seems reasonable to expect that appropriateness ultimately must be a relative matter: that is, that the meaning in question is primarily more appropriate for conveying the desired

314

FLORA KLEIN-ANDREU

interpretation than the other available options would be. (See e.g. García, this volume, also Klein-Andreu 1983). It seems to me that the appropriate­ ness of the meaning posited here for the simple past, for many of the examples given, should emerge even more clearly when we can compare it to the meanings of the other options available for referring to events in the past. I note too the author's interesting observation that different analyses have been proposed for essentially the same forms — even though they are likewise based on the premise of semantic invariance (e.g. Reid 1979). The suggestion that the reason for this difference lies in what each analyst considers to be "important" must certainly be correct. At this point I believe an explicit comparison of the different analyses (along the lines suggested by Tobin, this volume) would be very useful in illustrating the theoretical underpinnings of each approach and their analytic conse­ quences. In this connection, too, one looks forward to the author's analysis of the entire "system" (or systems) involved here. At first blush, it would seem that her analysis of the French simple past as conveying "detach­ ment" would have the advantage, over others, of accounting quite straight­ forwardly for the fact that this form currently is found almost exclusively in relatively "formal" written language, and in other situations (such as prepared speeches) which share with it the actual detachment of the pro­ ducer from the receiver. Yet this too must ultimately depend on the analysis of the system as a whole, since the substantive content of "detachment" presumably must be constrained by that of the other meanings in the system. The presentation also suggests some questions of detail which perhaps will be answered by analysis of the system as a whole. For example, are "detachment" and "dimensionality" not just opposite sides of one and the same meaning (or semantic feature)? On the face of it this seems possible, and it would also seem to be consonant with the author's claim that different contexts tend to "bring out" one or the other. In any event, I believe it would be useful if the mechanism of this interaction of meaning(s) and context were specified somewhat more exactly. (On this question, see Kirsner 1983 and 1989.) Another question that interests me is the meaning of other compound forms such as the pluperfect and the passé antérieur (see e.g. example 11). Are the meanings of these forms simply composites of (i.e. predictable from) the meanings of their component parts? And if so, what consequences does this have for their synchronic analysis? Finally, the correlation mentioned between textual demarcations and simple past certainly would support the analysis proposed. It would be interesting, therefore, to have some figures showing the extent to which the correlation actually holds.

INVARIANCE IN GRAMMAR, VARIATION IN DISCOURSE

315

Discussion of the paper by Nils Thelin The logic of this argumentation seems very convincing. Precisely because it is, one would welcome the opportunity to confront it with some more data than are offered in this paper — perhaps along the lines of Thelin 1984. For example, is there any other evidence that the developments proposed for Indo-European were as suggested here? One reason why more actual language data would be welcome is that the proposed cognitive sequence is apparently regarded as universal. It would be interesting, then, to see why languages having apparently similar aspectual distinctions do not necessarily use them in the same way. An example of this would be the distinction between simple and progressive present in Spanish as compared to English (e.g. as summarized in Klein 1980) — or, for that matter, in older stages of English as compared to the English of today. Here too, it may be that we ultimately have to consider the relation­ ship between the aspectual distinctions and the other distinctions with which they interact, in the particular language in question. I heartily agree with Thelin's contention that in order to understand (grammatical) aspect one must distinguish it from so-called aktionsart, which, if I understand the term correctly, refers to aspectual characteristics directly conveyed by the lexicon (in this case, the meanings of particular verbs). More generally, it seems clear to me that failure to distinguish the semantic contribution of grammar from that of lexicon has been the major impediment to principled study of grammatical meaning (as I attempt to demonstrate in detail in Klein-Andreu 1983). At the same time, it must be realized that the interaction of lexicon and grammar can be quite a challenge to analytic rigor (and especially to empirical testing). A case in point is that of so-called "stative verbs" in English, which, as Thelin points out, do not normally occur in the progressive form. He proposes that this is precisely because these verbs refer, lexically, to a "state of affairs", which as such is conceived of as non-time-related, and to which, therefore, the distinction + /— totalization is irrelevant (see also Thelin 1984: 231). Intuitively, I feel this explanation is correct. This situation would be reminiscent of that of terms like fish, deer, partridge, caribou, etc., which normally do not occur in plural form, presumably because the referents are not usually thought of as individuated ( + / —) in the first place. Yet, if it is the case that lexical and grammatical meanings that are in principle highly coherent with one another (such as "state" and " — totality") do not normally co-occur, then it becomes difficult to substantiate one's analysis of such meanings. Perhaps the concept of "mark" would be helpful here, but I do not yet understand it clearly enough to offer specific suggestions. A final comment on the matter of the use of "stative verbs" in English:

316

FLORA KLEIN-ANDREU

I believe it would be useful (and could strengthen the author's arguments) to mention the fact that such verbs sometimes do occur in the progressive form (e.g. I'm understanding/liking it much better now). And it should be noted that, when they do, a limit is suggested (in fact they are usually understood in an "inceptive" sense); thus, the question of "totality" is brought into play. Such inferential "effects" as these (see also Klein-Andreu 1983: 173) lend support to Kirsner's contention that meaning-based analyses ultimately will have to specify the inferential mechanisms that lead to particular interpretations of relatively abstract grammatical meanings, in particular contexts, and in fact will have to do so in such a way that the validity of the mechanisms proposed can actually be established empirically (Kirsner 1983 and 1989).

Discussion of paper by Henning Andersen Of all the papers in this section, this one appeals most explicitly to Roman Jakobson's insights and their possible applications to various fields: eth­ nology as well as literature and linguistics. While this broad scope of interests makes it well suited to a conference devoted to honoring Jakobson, it also makes me badly suited for commenting on it. Perhaps for this reason, I have not found the author's proposals in this case as enlightening as I have in other instances. In brief, I do not see that the idea of "paradigmatic equivalence" that emerges from this presentation is particularly useful for the study of lan­ guage. For example, the author states his belief that it will account for "diverse phenomena of grammatical government in which specified con­ ditions require [my emphasis] the selection of marked categories while else­ where conditions call for the use of an unmarked category". In my opinion, the idea that the occurrence of certain forms is automatically "required" (thus, "governed") by the presence of certain others is the single worst obstacle to analysis of grammatical meaning (and thus ultimately to that of lexical meaning too). For it allows us to view the "governed" items as effectively meaningless. And these are more likely to be the grammatical elements, since their meaning is usually considerably less obvious than that of the associated lexical elements. But this necessarily affects the analysis of the lexicon too, the most egregious result being the proliferation of hom­ onyms, and even the outright ignoring of cases of contrast. (For detailed examples of this see Klein-Andreu 1983, especially p. 176.) Some of the author's statements imply that he recognizes that "paradigmatically parallel" elements are in fact independently meaningful, and that it is the coherence of their meanings that is responsible for their habitual

INVARIANCE IN GRAMMAR, VARIATION IN DISCOURSE

317

syntagmatic combination. Thus, referring to the various features Hopper and Thompson associate with + / — transitivity, he says "It stands to reason that such features of meaning would be selected for expression and combined in speech according to the communicative needs of speakers" — presumably, then, not automatically. But if this is so, then of what use is the concept of "paradigmatic parallelism" here? Are we not just dealing, once again, with "usual correlations", a characteristic of data which requires explanation, but is hardly explanatory in itself? Though I know very little about ethnology (and less about the particular cultures referred to here), the cultural co-occurrences mentioned also seem to be independent results of a common cause or at least of relatable causes (somewhat like the annual coincidence in rise in consumption of ice-cream and death by drowning). Thus, for example, if one were to assume (for the sake of the argument) that in the early Danish society described it was the men who held greatest power, it would be natural that they should take for themselves the side of the church that was most comfortable (the sunny side, which would also tend to be less humid), and that they should arrange matters so as to get there as comfortably as possible (thus, directly). If (for reasons that seem to be shared by many cultures) it was felt to be desirable for men and women to be separated in church, that would naturally leave the women with the less comfortable side, which would also have to be reached via a more circuitous route. It is also not unusual for people in relatively powerful positions to wish to show themselves socially as such — perhaps to perpetuate their power thereby. If we can make the further assumption that church-going in early Denmark was imbued with social significance (as this presentation supposes too), then it would also be natural for the men to build for themselves a more ornate entrance, of more "noble" materials. Less obvious is the reason why the men placed themselves at the right side of the aisle — thus placing themselves on the left hand of God, if the congregation is viewed as facing God (via the altar). A likely reason for this is that the altar must be on the eastern side (again, not peculiar to this culture), so that anyone wishing to remain on the warmer side of it would necessarily be to the right of the aisle. Another possible reason might be identification with the right hand of the celebrant — assuming that the priest faced the altar, having his back to the congregation. It seems to me, then, that these cultural co-occurrences might more fruitfully be viewed as independent consequences of other cultural character­ istics (in this case, that in the culture in question church-going was a social act, and that men enjoyed greater power than women), interacting with certain likely universais: that "good" things are associated with the sun, with the right hand, and with warmth. (For an analysis of linguistic meaning somewhat along these lines, see Wierzbicka 1985).

318

FLORA KLEIN-ANDREU

In general, I do not see that the concept of markedness contributes anything to an understanding of these phenomena. If anything, in the cultural examples it would seem that it is the women's lot that is "unmarked" — since in effect it seems to amount to what is "left over" from the motivated choices of the men. Finally, I do not see that the example of phonological variation is in any way analogous to the others, since where articulations are concerned none of the variants is meaningful, of itself.

References Klein, Flora. 1980. "A Quantitative Study of Syntactic and Pragmatic Indications of Change in the Spanish of Bilinguals in the U.S." Locating Language in Time and Space, ed. by William Labov, 69-82. New York: Academic Press. Klein-Andreu, Flora. 1983. "Grammar in Style: Spanish Adjective Placement." Discourse Perspectives on Syntax, ed. by Flora Klein-Andreu, 143-179. New York: Academic Press. Kirsner, Robert S. 1983. "On the Use of Quantitative Discourse Data to Determine Inferential Mechanisms in Grammar." Discourse Perspectives on Syntax, ed. by Flora Klein-Andreu, 237-257. New York: Academic Press. . 1989. "Does Sign-oriented Linguistics Have a Future?" In From Sign to Text, ed by Y. Tobin. Amsterdam: J. Benjamins, 161-178. Reid, Wallis H. 1979. "The Human Factor in Grammatical Analysis: The Passé Simple and the Imparfait". Ph.D. Dissertation, Columbia University. Schank, Roger  and Peter Childers. 1984. The Cognitive Computer. Reading, Massachusetts. Thelin, Nils 1984. "Coherence, Perspective and Aspectual Specification in Slavonic Narrative Discourse". Aspect Bound, ed. by  de Groot and M. Tommola. Dordrecht. Waugh, Linda R. 1977. A Semantic Analysis of Word Order. Leiden. E. J. Brill. Wierzbicka, Anna. 1985. Lexicography and Conceptual Analysis. Ann Arbor, Michigan.

PART FOUR GRAMMAR AND PRAGMATICS

Deixis and Shifters after Jakobson Herman Parret Belgian National Science Foundation 1. Subjectivity in language Structuralist dichotomies Structural linguistics states evidently that language is a system the structure of which has to be analyzed. Each structure, the units of which constitute each other's condition, is distinguished from other structures by the speci­ ficity of the relationships between these units. This structural principle, presented and developed by Jakobson, Karcevski and Trubetzkoy (First International Congress of Linguists, The Hague) from the late twenties on, combines with the thesis of the autonomy of langue, form, and synch­ rony with regard to parole, substance and diachrony. The combination of the two theses is problematic, especially because of the fact that the langueparole dichotomy (as well as the form-substance, and the synchronydiachrony dichotomies) has a double function in Saussure. The first func­ tion is methodological, corresponding to the classical distinction between the "object" constructed by the linguist (as a scientist), and the given data of which the object provides an explanation. However, Saussure uses the same dichotomy in order to introduce another type of distinction, a "material" distinction internal to the data: the domain of langue then is constituted by some relations observable by introspection, whereas the domain of parole is made up by another type of observable facts, namely the historical events that are the various actually realized communicative acts (Saussure uses here the comparison of the score and its execution by the musician). It should be evident that, according to the first, methodolog­ ical, interpretation of the dichotomy, only langue is systematic and struc­ tured, whereas the "material" interpretation of the dichotomy opens the possibility of the application of the structural principle to the domains of both langue and parole. It is interesting to notice that the two interpret­ ations of Saussure's dichotomizations gave rise to two types of structur-

322

HERMAN PARRET

alism in linguistics. Glossematics, with its stress on formalization and on the epistemological criteria constraining theory building, looks at linguistic structure as a scientific (re)construction, whereas "functionalism" (Jakobson's for instance) discovers structurality and systematicity within the material and observable data of the actual linguistic behavior. However, "post-structuralism", mainly in pragmatically oriented linguis­ tic disciplines, denounces the opposition of the methodological and "materiar' interpretations of the dichotomies: a central claim of pragmatics has been that langue as a theoretical object should contain a reference to parole as an empirical domain. This leads to slogans such as: parole should be reintroduced within langue, and enunciation within utterance, or, linguis­ tics has as its object the "marks" of the act of enunciation within the utterance. These formulas are paradoxical as long as the act of enunciation is considered to be a unique event, implying a particular locutor in a particular situation whereas the utterance would be invariant through the infinity of the acts of enunciation. However, a distinction should be made between the specific conditions, always original, of the enunciation produced hic et nunc, and the general fact of the enunciation, identical through the diversity of the actually realized acts. There is a general scheme of enuncia­ tion, which can be described by specifying the roles of possible locutors and interlocutors within linguistic action sequences. Moreover, the enuncia­ tion is sui-referential:that sequences are produced and understood, is mani­ fested by these sequences, that they are time-, space- and actor-bound is expressed by complex grammatical subsystems, that discourse has a subjective origin is shown by its observable effects in communication and inter­ action.

Referentiation, modality and deixis The terms "subjectivity" and "subjectivism" have a bad reputation in linguistics and in the social sciences. Nevertheless, Benveniste (1966) intro­ duced "subjectivity" within the heart of his general conception of language, and many other linguists, escaping mainstream positivism and objectivism, did the same (for a fascinating proposal in the recent "cognitive grammar framework", see Langacker 1985). However, "subjectivity" is not a very operational concept in empirical disciplines. It should rather be associated with the domains of modality, deixis and refer entiation. Reference is com­ monly looked at as a context-dependent aspect of sentence meaning: logical and linguistic semantics claim that the referential range of referring expres­ sions being fixed by their meaning in the language, their actual reference depends upon a variety of contextual factors. The context then is made

DEIXIS AND SHIFTERS AFTER JAKOBSON

323

up by ontological items (objects, states of affairs, events). Names and definite descriptions are considered to be prototypical referring expressions. But the use of, for instance, definite descriptions as referring expressions cannot be satisfactorily accounted for solely within the framework of truthconditional semantics and abstracting from the subjectivity of the utter­ ance. When a speaker employs a definite description, he indicates by means of it that he is performing the act of reference, and he assures (implicitly) the addressee that the expression will contain all the information that is required to identify the referent. The subjective origin of referentiation is even more evident in the case of referential opacity and of the de re/de dicto ambiguities. That the context of the utterance is not solely external or ontological is true in the case of referring expressions, and, a fortiori, of deictic expressions. Their meaning is even more dependent on the subjective origin of discourse. The term "deixis" originates in the notion of gestural reference, i.e. the identification of the referent by means of some bodily gesture on the part of the locutor. I will contest, further on, the relevance of a theory of deixis where the whole deictic field is associated with ostension. The deictic context is, on the contrary, centered upon the locutor's here and now — I will call this speaker-based system of deixis the egocentric one. The deictic context is not external (ontological) but subjective. From the egocentric organization of deixis, one can draw the distinction between pure and impure deixis. I and you, this and that, here and there are purely deictic (they refer to the locutor and the addressee without conveying any additional information about them). He, she and it are impure deictics, encoding distinctions based upon properties of the referent which have nothing to do with the spatio-temporal roles of the locutor or the addressee (gender and its classification, for instance). The important point is that the "purity" of deixis is measured by its proximity to the egocentric here and now, and this is why deixis comes so close to subjective modality. This proximity or remoteness is emotional, and the ego can be distancing himself attitudinally from whatever he is referring to. It is indeed a defensible view that the egocentricity of the deictic context is of its very nature subjective. Modality is the third domain that should be associated with subjectivity (although there are cases where modality penetrates the domain of deixis, as I already mentioned). The only kind of modality recognized in traditional logical semantics is that which has to do with necessity and possibility insofar as they relate to truth values. But, in the case of natural languages, the modality is more likely to be either epistemic or deontic, and utterances which are construed epistemically or deontically are subjective rather than objective. Even assertions imply epistemic commitment, and relatively few statements are modally neutral.

324

HERMAN PARRET

Productivity and creativity Saussurean structuralism — and its dichotomies — can be transcended by a theory of language (with its appropriate heuristics with regard to the empirical data) taking into account the subjectivity of utterance. If one wants to call the concept of meaning resulting from this point of view "pragmatic", then pragmatics should be reconstructing the enunciation or the conditions of production and of understanding of language sequences. Reference, deixis and modality are three interpenetrating domains that should be investigated by such a pragmatics. Subjectivity here should not be taken in any psychological or "subjectivistic" sense: the ego is not to be understood as being logically distinguishable from the attitudes expressed by the modalities, from the role it plays when performing the act of reference, nor even from its spatio-temporal location. It should be added that the ego is not a reasoning competence operating upon prop­ ositions and producing judgements, it is emotionally and passionately attitudinal. The inadequacy of structuralistic theories of language and meaning derives precisely from their being unable to reconstruct adequately subjectivity in language. It cannot be said that subjectivity, and its associated empirical domains, is an explicit priority of Roman Jakobson's linguistic investigations. How­ ever, his "phenomenological structuralism" (Holenstein 1974) transcends in many ways the framework of Saussure's axioms and their formalistic interpretation in glossematics. Jakobson partially accepts Saussure's thesis of "linguistic arbitrariness", but the autonomy of language — mainly a methodological principle — is relative and related to human psychology, culture and society. Moreover, language is not a formal entity in se, but it serves interpersonal communication. It is important then that the communi­ cative function is not a local aspect of language but a property informing and permeating the entire structure. That language is teleological (the finality of language being functional) is not a purely static principle. It should be combined with another central aspect, namely its productivity and its creativity (Waugh 1976), intensely demonstrated by the poetic (metaphoric and metonymie) use of language. However, where Jakobson's empirical investi­ gations come closest to the paradigm I present (subjectivity through enunci­ ation), is in his typology of the functions of language and his (mainly morphological) study of shifters. In comparison with Bühler (1934), Jakobson's presentation of the set of six relevant functions of language is certainly more complex, and the empirical findings more important. But it seems to me that an important feature of Bühler's conception in the Sprachtheorie is lost, namely the intrinsic link between linguistic functionality and deixis (egocentrically

DEIXIS AND SHIFTERS AFTER JAKOBSON

325

organized: I/now/here is the 'Origo des Zeigfeldes"). The importance of the speech event (enunciation) is recognized by Jakobson but its relation to the various functions of language (except for the emotive and conative functions where the relation to I and You is evident) is not made explicit. Deixis and enunciation are important facets of linguistic structure in the Jakobsonian framework, but they are still characterized as local phenom­ ena with specific empirical effects, such as shifters. It remains true that Jakobson does not go so far as to claim that "deixis may well be the most important integrating factor of the semantic system,... that meaning is essentially and inherently built on the notion of deixis". It is well known where Jakobson situates deixis: "it is built on the dialectic tension between code and message, on the anticipation of the message within the code and on the antinomy of the narrated event and the speech event" (Waugh 1976: 24). The interpretation and understanding of the meaning of deictic categ­ ories depends upon the event of enunciation, and these categories (whether grammatical or lexical) can only be differentiated by relating them to the enunciation of the discourse fragment. Enunciation enters into the meaning of shifters, or, to use Peirce's term, indexical symbols. However, two aspects differentiate Jakobson's and Peirce's approaches to these cardinal linguistic and semiotic categories. First, Jakobson characterizes the internal relationship on which shifters depend as the relation between code and message (replacing the Saussurean dichotomy of langue/parole, and taken from communication and information theory), whereas Peirce's notion depends upon a typology of signs where the internal relationship is one between the "material quality" signans, and the "immediate interpretant", signatum. This leads to a second distinction between both definitions. The content, in the Jakobsonian scheme, is a non-analyzed, substantial and positive datum, whereas in Peirce it is an item which is continuously submitted to interpretation. It is clear that this interpretative feature brings us closer to the enunciative paradigm because interpretation seems neces­ sary to the production as well as the understanding of deictic categories and deictically marked linguistic sequences. In this paper I intend to introduce some propositions which clarify the enunciative paradigm in its confrontation with deixis and shifters. How is deixis organized and how can shifters be defined once the subjective origin of discourse is accepted as a fundamental principle with far-reaching heuris­ tic and methodological consequences? Roman Jakobson's fruitful insights on these essentially pragmatic phenomena are the incentive starting point of my presentation. In a first point, I will sketch the current discussion on deixis in the philosophy of language and suggest in which way it is relevant in empirical linguistics (II); then I will discuss some recent developments in narratology and semiotics on "shifting" (III).

326

HERMAN PARRET

2. Deixis Demonstratives Pronouns and demonstratives are an intriguing and challenging topic for philosophers and linguists (this section is heavily based on Parret 1980). Anyone who pretends to have a theory of meaning and empirical evidence to support it must be able to reconstruct the meaning of demonstratives. The semantics of proper names is a well developed (I would even say an over-developed) area within the philosophy of language. The semantics of demonstratives, however, is still controversial and in many respects non­ existent. My claim is that an overall theory of meaning is relevant only if it can treat deixis and demonstratives in a valid and sensitive way. This was precisely Frege's intuition when he reached the conclusion that his otherwise truly aesthetic theory of meaning (and, in particular, its sense-reference distinction) had to be amended radically because of the semantic peculiarities of deictic categories in language. I will defend in the theoretical battlefield, a theory of deixis in which the whole domain of deictic categories is concen­ trated around I. I call this type of theory the egocentric theory of deixis, as opposed to the ostensive theory of deixis. Let me first clarify the terminology. First, the notion of demonstrative or deictic category is traditionally used for the category of words the meaning of which has, as a prerequisite, an associated demonstration (including, in most cases, the accompanying pointing). The purest examples are this and that (demonstrative pronouns, although the adverb there is very close to this/that). Here, on the other hand, is already a problematic case: do I have to associate a demonstration (by pointing) to mean something with here, or do I simply have to be somewhere as an I to mean something with here? Moreover, he, too, can be used as a demonstrative in the ostensive sense (He is a murderer). Remember, however, that this/that are not always used as demonstratives: this/that can also be anaphorical, referring to a syntagm in an earlier sequence of the discourse; and anaphora can never be demon­ stratives in the ostensive sense. In order to have distinctive notions, I will consider deictic categories to be included as a subgroup of demonstratives. I, you, now (and here, I think) are the purest examples. The referent of these expressions is dependent on the context of use, with the meaning of the expression providing a function or a means that determines the referent in terms of certain aspects of that context. The battle, certainly, will be waged on how this definition of I, you, now and here is to be made precise. There are derivations of these categories (i.e. me, my, your, at this moment, today, etc.) as well as extensions (i.e. we, yesterday, tomorrow, etc.). All these deictic entities acquire their meaning because of their intrinsic relationship to the

DEIXIS AND SHIFTERS AFTER JAKOBSON

327

context of utterance. Many terms have been used for this class of words: "token-reflexives" (Reichenbach), "egocentric particulars" (Russell), and "indexicals". I will use the following terminology: "demonstrative" as a generic term, "pure indexicals" and "pure demonstrative" as two sub-classes one of which necessitates an associated ostension or demonstration while the other one does not (see Collinson 1937, the oldest systematic account of the alternatives, in contemporary linguistic literature). demonstrative pure indexical (paradigm: I)

pure demonstrative (paradigm: this/that)

DEIXIS

OSTENSION

The distinction between "pure indexical" and "pure demonstrative" is clearcut in this abstraction, but I will argue that this distinction is not an opposition but rather an axis with a continuum of intermediate positions. The starting point remains that some demonstratives require an associated demonstration in order to determine their referents; these demonstratives are the pure demonstratives, and, in this case, the demonstrative expression refers to that which the demonstration demonstrates: that is, a linguistic rule assumes that such a demonstration accompanies the use of the pure demonstrative. On the other hand, no associated demonstration is required for pure indexicals; and any demonstration supplied in the case of pure indexicals is either irrelevant or used for stylistic and rhetorical emphasis. The speaker refers to himself when he uses I, and no amount of pointing can reverse this reference. At this stage I could conclude that the referent of a pure demonstrative depends on the associated demonstration, whereas the referent of a pure indexical depends on the context of use. But still no substantial definition of the meaning of demonstratives has been offered. Without going into empirical linguistic matters, I will state my position by reduction and in four stages, concentrating on the meaning of I, which will be seen as paradigmatic for the whole domain of demonstratives. The four theses I defend are the following: 1. I is 2. ƒ is 3. / i s 4. ƒ is

a referring expression, but not a proper name; a demonstrative, but not a pure demonstrative; a propositional function, but not a mode of identification; a designator, but not a rigid designator.

328

HERMAN PARRET

I is a referring expression It is commonly agreed that reference by language fragments is accomplished by at least the following three grammatical procedures: (1) by proprial reference (that is the reference by means of proper names as Deng-Xiaoping and Peking); (2) by descriptive reference: this is the case of definite and indefinite descriptions where lexemes are modified by the definite or the indefinite article (one should say that in these cases, the first function of the descriptive expression is always predicative; and the derived function referential, as in such definite descriptions as the Vice-Premier of China); and (3) by pronominal reference (by means of pronouns and pronominal adverbs). Linguists point out that the further combination of these three procedures is always possible; expressions such as Chaplin/with the black hat, Amsterdam/of the punk generation, any human/here, and our lovely Margareta contain combinations of proprial, pronominal and descriptive elements. But this combination, even when it creates semantic redundancy, does not eliminate the specificity of the procedures themselves (see Leys 1979). Precisely from the simple viewpoint of the economy of linguistic means, the existence of the three procedures enables us to construct an architec­ turally pleasing theory. 1. Proper names refer to individuals in their unicity, but they do so without determination by the context: the act of uttering a proper name and the context of an utterance (speaker, time, space) does not influence the semantic content of a proper name. Surely, one can discuss — and this is the burning issue in the controversy about proper names — whether there is an aspect of their semantic content which is an intralinguistic means of referring (Frege: yes; Kripke: no); but it can be accepted that the semantic content is itself context-free. The Kripkean thesis that proper names rigidly refer to their causal history (and in fact to the baptizing of the individual by its proper name) is an argument exactly against the context-boundedness of the meaning of proper names. 2. Demonstratives refer to individuals in their unicity, but they do so by determination from the context. There is no causal history of demonstratives — rather, there is a referential use of demonstratives, and this use is meaningful in context. 3. (In)definite descriptions function in many ways: they can function either as predicative expressions (non-referentially) or as referential expressions; and in the case of the latter, they can achieve reference either as a proper name or as a demonstrative. Such versatility makes the status of (in)definite descriptions truly complex. The theory of definite descriptions will be a derived theory. It does not follow from the fact that definite descriptions

DEIXIS AND SHIFTERS AFTER JAKOBSON

329

can function as proper names and as demonstratives that proper names and demonstratives are adequately characterized when they are said to be definite descriptions — we all know about Russell's difficulties when defining proper names as definite descriptions. I claim that I, indeed, is a kind of referring expression (see Holenstein 1985: 60-67, for a thorough discussion of this thesis); but it is misconstrued if regarded as a proper name. Linguists often notice that I has the syntax of a proper name, and that it is a substitution for personal proper names in subject position (provided appropriate adjustments are made to verb inflection). Nevertheless, this cannot logically mean the reduction of the meaning of demonstratives to the meaning of proper names. Anscombe, in her pioneering article, "The First Person", correctly opposes such a reduction (Anscombe 1975), and she argues that if ƒ is a proper name, its reference cannot be slight but must be associated with a criterion emphasiz­ ing the re-identification of the individual for whom it stands; and of course the application of such a criterion seems not to be an essential part of the meaning intended by the use of I. The philosophical point I would like to make is that the meaning of a demonstrative cannot be identified with the meaning of a proper name, regardless of whether you have a Russellian conception of proper names (the meaning of the proper name is the meaning as a (set of) definite descriptions(s)), or a Fregean one (the meaning of a proper name is the mode of identification of the individual), or a Kripkean one (the meaning of a proper name is its referent, namely the causal history of the named referent). In fact, all paradigmatic theories of meaning have taken the use of a proper name to be the prototypical use of language entities; this generalization goes from Russell and Frege to Kripke, Donnellan and Putnam. The dream of a language with nothing but proper names is an old dream with strong metaphysical roots — namely, the idea of a language comprised of the terms the meaning of which is exhausted by a one-to-one correspondence to particulars in reality. Russell — as in fact may semanticists and logicians — stresses the representational function of language (Russell 1905): that is, the fact that language represents objects, events and states of affairs, while referents are the conditions of (truth) value of linguistic sequences. Once demonstratives in language are taken seriously, a highly specific alternative picture of language functioning is possible. Deictic and ostensive categories are never pure representations of the world — they permit interventions in the established structure of reality, they are praxes. Language acquires argumentative, transformational and interactional force, and as such acquires practical (and not only truth) value, once the deictic and ostensive workings in language use are realized. Reality and the intersub­ jective environment are both changed by these workings. When I state that

330

HERMAN PARRET

I is a referring expression but not a proper name, I mean precisely that I does not refer representationally as proper names do. The context-sensitivity of demonstratives is not purely passive: to say I, now and here has a practical value since such fragments are constitutive with respect to the interactional and intersubjective context. To consider language capable of performing this transformational role is to re-evaluate demonstration vis-à-vis represen­ tation, or demonstratives vis-à-vis proper names. I is a demonstrative The second claim I would like to discuss is the following: / i s a demonstrative but not a pure demonstrative. This is not only a thesis about I, but also a way of hierarchizing the domain of demonstratives. Indeed, the unification of the overall domain can be realized in two different ways: there is the "ostensive" qualification of demonstratives (and Russell gave the most radical version of it), and there is the "egocentric" qualification of demon­ stratives in which the paradigm of demonstratives is I (of which Frege is a sophisticated spokesman). All demonstratives (even this/that) can be explained by their relation with I. Russell easily eliminated the problem of the heterogeneity of demonstra­ tive expressions in language (Russell 1905). According to him, all deictic and ostensive categories can be reduced to the base category this/that. The meaning of this/that, then, is neither (a) the meaning of a (grammatical) proper name (look at Russell's odd argument: proper names are always applicable to various objects, and this/that is only applicable to one object in one specific spatio-temporal situation); nor (b) a definite description such as, for instance, the object which is now in the scope of my attention (individu­ ation of this/that would still require deictic and ostensive categories in the definite description, such as now and my attention); nor (c) a general concept such as, for example, that which all objects called successively "this/that" have in common (because a general concept does not carry the meaning of the spatio-temporal uniqueness of this/that). There are three aspects of Russell's solution to the problem of the unifi­ cation of demonstratives: 1. The reduction of all so-called egocentric particles is made in one direction. The meaning of here is the place of this; of now, the time of this, and of I, the biography to which this belongs. I am, according to Russell, means this is; I am warm means this is heat or heat is here. 2. This is x, indeed, is equivalent to property x is here (is present). Thus, the circumstances of the use of deictic and ostensive categories are the direct stimuli for the linguistic act of deixis and ostension. Russell clearly identifies physical (and physico-psychological) presence and linguistic here.

DEIXIS AND SHIFTERS AFTER JAKOBSON

331

3. As a matter of fact, the meaning of this/that is reducible to the meaning of the names of what, in reality, is present for the speaker. This goes back to Russell's ontology, logical atomism, and the dream of the logically perfect language. Egocentric particles are logical proper names of expressions in one-to-one correspondance with logical atoms. The Fregean doctrine is more sophisticated (Frege 1918), and the unifi­ cation of the overall domain of demonstratives goes in the other and the right direction. To understand the motives of Frege's theory of demonstra­ tives, one could wisely review Peirce's classification of signs according to the three well known classes of symbols, indices and icons (see Burks 1949). To recall the definitions: (1) symbols are linked by the interpreter to the object with the aid of a conventional rule of great generality: for instance, red is a symbol in the sentence The table is red; (2) indices are linked to the object by the interpreter who postulates an existential relation by, for example, a gesture (pointing), as in This is a table; (3) and icons, of less importance here, are signs in which the object is exemplified, as in the diagram of a machine (see Jakobson 1966). Demonstratives, in this classifi­ cation, are indexical symbols; language entities with on the one hand, a fixed conventional or symbolic meaning, and on the other hand, a changing indexical meaning corresponding to the particular circumstances of the utterance. The difficulty Peirce thematizes with respect to demonstratives was, indeed, the fact that demonstratives always have the same symbolic meaning but with varying indexical meaning. Frege's semantics, with its sense-reference distinction, is an adequate candidate for a coherent definition of the status of indexical symbols, to use the Peircean terminology. A Fregean will argue that to understand the meaning of demonstratives, three types of knowledge — two necessary, one optional for the domain of pure indexicals — are to be presupposed in the speaking subject who is using and understanding demonstrative language: (1) knowledge of the spatio-temporal location of the utterance of the lan­ guage sequence: to know the meaning of now, I have to know when the language sequence has been uttered; to know the meaning of here I have to know where the language sequence has been uttered; (2) knowledge of the rules of linguistic use (i.e. knowledge of the symbolic meaning of demonstra­ tives): one has to know that I has the symbolic meaning of the person uttering the language sequence; and that this in this city, the symbolic meaning of the city in which the utterance of the language sequence is realized. This second type of knowledge has to be considered the essential and necessary means for the realization of the first type of knowledge — that is, the purpose of demonstrative language use; (3) the first and second types of knowledge will suffice for the production and understanding of deictic meaning; the third type of knowledge is optional for the understanding of

332

HERMAN PARRET

pure indexicals but necessary for the understanding of pure demonstratives. Indeed, to understand ostensive meaning one must have knowledge of the direction of application, either by explicit ostension or by supplementary description. This, therefore, becomes a pure demonstrative in the expression this tree (in contradistinction to this city, where this seems to be the extension of an indexical). Frege's solution when qualifying pure demonstratives, is to claim that no spatio-temporal location is possible without knowledge of the direction of application. There is a fundamental difficulty in how to consider the first type of knowledge required to understand I. Should there be something more to the meaning of I than merely the person who is here and now, or its spatiotemporal qualification? If I is an uncommunicable self-presentation — as it is suggested by Frege in Der Gedanke (Frege 1918) — then Frege's theory of language could become paradoxical. His theory of meaning is directed explicitly against the idea of private meanings. Within the framework of his realistic ontology uncommunicable meanings should be inadmissible. Mean­ ings are public because of the fact that the objectivity of the world is public. If deictic entities, such as I, have private meanings, there would be islands of subjectivistic darkness in the language: some primitive aspects of myself (and possibly of my "place in the world') would be presentable (knowable) only to myself. This could be a serious crack in Frege's theory of meaning, but the fact is that it is not. The question of what the speaker knows about himself when he says I is of no importance at all. The problem of the private part of the meaning of I quickly disappears once one admits that the meaning of I is not an epistemic meaning, not a belief, and not a thought. I argue that no special form of knowledge/belief about an object is required or presupposed in order that a person may entertain a singular proposition involving I. From this position it follows that the ignorance of the reference does not defeat the referential character of I, and, hence, of demonstratives. This is aimed at refuting so-called "Acquaintance Theories of Reference" according to which the speaker's knowledge of the referent rather than the form of reference, determines whether an utterance expresses a singular proposition containing ƒ. To grasp that ƒ is a formal referent rather than a thought or a set of belief(s), one could claim in a more technical terminology that the speaker of the language should understand / a s a propositional function and not as a mode of identification. I is a propositional function This is precisely the point where a philosopher becomes anti-Fregean. The way for / t o be a referring expression is not by being a mode of identification.

DEIXIS AND SHIFTERS AFTER JAKOBSON

333

If this is true for ƒ, it will also be true for the whole scale of demonstratives (from the pure indexicals all across the axis to the pure demonstratives). Kaplan (no date) quotes such sentences as the following examples to illus­ trate the difficulty of the Fregean viewpoint: I do not exist. I wish I were not speaking now. It will soon be the case that all that is now beautiful will be faded. I was insulted yesterday. It is possible that in Belgium, in two years only those who are actually here now, will be happy. The thesis which holds ƒ to be a propositional function (without being a mode of identification) does not mean that I and the other demonstratives have no fixed semantic rule of use which determines their referent in each context of use. There is a semantic rule which determines their referent in each context of use, but the rule does not generate the meaning of the indexical from the circumstances of evaluation (the utterance with its intrinsic meaning), but rather from the actual context of use. The rule just provides an object, a state of affairs, or an event, so the distinction between context of use and circumstances of evaluation is highly adequate. Circumstances of evaluation are the "circumstances" surrounding what is said on a given occasion of use. This is why the propositional value of utterance (and I as a propositional function) has to be distinguished from the sentential value of an utterance; that is, a propositional function has a "context of use" whereas a discourse fragment (words and syntagms, for instance) has "cir­ cumstances of evaluation". The appropriate semantic rule for I (with its extensions) could be the following: in each possible context of use, /refers to the agent of the context. This designation rule cannot be used to assign a relevant object to each circumstance of evaluation. Take, for example, I do not exist: under what circumstances would what I said be true? It would be true in circumstances in which I did not exist. Clearly, I does not refer to circumstances of evaluation, but to the agent of the context of use and the circumstances of evaluation do not involve contexts of use and agents who do not exist. Similarly, in the example I wish I were not speaking now, the circumstances desired do not involved contexts of use and agents who are not speaking. The actual context of use serves to determine the relevant individual me; and we then set up various circumstances of evaluation of that individual. Consider what is said in the sub-sentence. All that is now beautiful will be faded. I wish to evaluate the content of this sequence at some future time; however, what is the relevant time associated with now? It is the time of the context of use, to. In the sentence It is possible that in Belgium, in two years

334

HERMAN PARRET

only those who are actually here now will be happy, here and now are the place and the time of the context of use (poto), and not the set of circum­ stances of evaluation as they are determined by the sentential operators (of spatio-temporal location). Yesterday in I was insulted yesterday, is yesterday for to, because it would be today if the time were the time of the circumstance of evaluation. This leads to the formulation of the third claim: the meaning of indexicals, and especially of/, is determined with respect to the context of use. Irrelevant or inapplicable to it is the determination of the referent with respect to the circumstances of evaluation. It follows from the claim that I is in each of its utterances referential; I refers to the person who utters it, but I may have a different content in each of its utterances according to sentential modes of identification. Heuristically (according to the viewpoint of someone who "discovers" or understands / a s a meaningful expression), the order will be reversed: one understands first the relation of the indexical to the sentential circumstances of evaluation, and then its relation to the actual context of use. David Kaplan, who worked out the distinction between context of use versus circumstances of evaluation, presents the following terminology (Kaplan, no date): SENTENCE PROPOSITION

The Content of ƒ is a sentential function from possible circumstances to intentions; the Character of I is a propositional function from possible contexts to contents. Demonstratives, and in the first place I, have their Character as their meaning. As a conclusion, it should be stated that ƒ is a propositional functional function not by being a sentential mode of identifi­ cation but by being a designator of the context of use. But I is not a rigid designation rule, and demonstratives are not rigid designators.

I is a designator As I noticed before, the heuristic order in which to understand demonstrative sequences is the following: first one understands the relation of the indexical to its sentential circumstances of evaluation, and then its relation to the context of use. To reevaluate heuristics is a side-effect of a more general

DEIXIS AND SHIFTERS AFTER JAKOBSON

335

option: to build up a theory of meaning as a theory understanding. A pragmatic "communicative competence" is a competence of understanding or "discovering" the significance of discursive sequences. Thus there is an essential asymmetry between production and understanding in pragmatics. However, understanding is not a mental state or a specific experience — it is rather an extrinsic ability or an operation-in-the-world. Understanding is a practical, or interactional operation; to understand the I-sayer is to delegate to the speaker the First Person Authority. Davidson wrote recently on the special authority with which speakers attribute mental states and events to themselves (Davidson 1984). He finds an explanation for this authority in the nature of the interpretative act, for the argument is intended to show that interpretation depends on the interpreter's delegating this authority to the speaker. Interpretation and First Person Authority are dialectically intertwined. On the one hand, an explanation for this authority can be found in the nature of the interpretative act, and one can argue that interpretation depends on the interpreter's delegating the authority to the speaker. On the other hand, there is no authority which is not intersubjectively valid, thus delegation of authority is a conditio sine qua non of the subjectivity of the I-sayer. Given the nature of interpretation and understanding, First Person Authority follows. Given the special authority of the I-sayer which is not earned nor won but delegated, the act of interpretation or of understanding constitutes speaker's subjectivity. The First Person Auth­ ority, delegated by the community of interpreters to the I-sayer, consists of the fact that the I-sayer is assumed to have the right of the designation of the objects, events and states of affairs in the world, and especially the subject, event and value he means by saying I. This is why / i s a designator, not a rigid designator. I does not have to know more about himself than the role he plays in interaction and intersubjectivity; saying I (you, we, here, now), the I-sayer refers to his role in the interplay, and he intends to designate himself in this role, and he is assumed to intend so by the community. Therefore, ƒ is a designation rule or a designator, but not a rigid designator: I does not have a transparent referential function — on the contrary, I is an opaque condition on force, and an opaque principle of rational and cooperative language functioning. To conclude, I mention that the workings of the I on discourse as a whole go further than the grammatical category of demonstratives. The whole structure of language is organized around the I-sayer and his referring to the role delegated by the community. This is no doubt the essence of the pragmatic view on language, and it would be of great interest to see how this "egocentrically" oriented position on demonstratives, on demonstrative language use, on the I-sayer, would behave under testing conditions in empirical linguistics.

336

HERMAN PARRET

3. Shifters A pragmatic theory of language considers all language sequences to be modified by the I-sayer. The I-sayer holds the First Person Authority con­ sisting of shifting in and shifting out of the discourse (see Parret 1983, as well). It is my purpose to relate the analysis of the First Person Authority to a modified version of Jakobson's theory of language shifters.

Shifting in Even when the subject becomes the "object" of a self-conscious attention on the part of language turned back upon itself, it is still the case that subjectivity "shows" itself without "saying" it by means of grammaticalsentential information. To say versus to show are classical contrastive categ­ ories, exploited by Wittgenstein, Austin, Benveniste (1966), Bühler (1934; see Zeichen versus Anzeichen). When discourse is modified by the I-sayer, it is in the first place by deictic and modal evidence. Deictic modification, for instance, takes place when the I is "shifting in" in discourse (not only as a grammatical-distributional rule, but as a designation function, and even as a condition for interaction and intersubjectivity, and an "axiological" prin­ ciple of cooperation, coordination, veracity, authenticity, and so on). It is evident that the way I introduce the term "shifting in" is far from orthodox with regard to Jakobson's use of the category of shifters (Jakobson 1957). There are mainly three differences. In my view, shifters as stabilized as morpholexical categories are not what is relevant, but shifting as a discursivetextual procedure or operation; moreover, shifting does not concern only the presence of the I as grammatical-distributional rule: the workings of the I on these deeper levels of the proposition, the action unit and the actantial structure have a broader scope than just the surface grammatical categories and distributions; and finally, shifting in is in perfect balance with shifting out: there are procedures effectuated by the I of being present and being absent. This third property is almost universally forgotten in current linguis­ tic theorizing. Contemporary narratology and semiotics analyzes the dialectics of shifting in and shifting out, of "engagement" {embrayage) and "disengagement" {débrayage) within texts and other semiotic objects (of the cultural and even "natural" worlds). Engagement designates the effect of a return to the enunciation. Every engagement presupposes a disengagement operation which "logically" precedes it. "When, for example, the American president utters: 'America is a beautiful country', he operates an utterative disengage­ ment which installs in the discourse a distinct subject, distant with regard

DEIXIS AND SHIFTERS AFTER JAKOBSON

337

to the domain of enunciation. On the other hand, if the same person said: T h e American president thinks that...', it is still an utterative disengage­ ment, but one which is supplemented by a set of procedures which we call engagement and which, although they remain implicit, aim at producing, among other things, an effect of identification between the subject of the utterance and the subject of enunciation" (Greimas and Courtés 1979: 150). Engagement can be divided into actorial, temporal and spatial engagement, often united and put into place in syncretism. We can interpret engagement as the negation of the not-I, carried out by the subject of the enunciation, and aiming at an (impossible) return to the source of the enunciation. One of the main results of the semiotic analysis of narrative texts is that there are all kinds of strategies creating all the while the enunciative illusion. Engagement, indeed, is both a goal of enunciation, and it is a sort of failure, an impossibility of reaching that goal. The two "references" with the aid of which a way out of the closed world of language is sought, are a means of pinning this universe onto a totally distinct exteriority — reference to the subject (to the domain of the enunciation) and reference to the object (to the world which surrounds human beings as referent). These references succceed only, in the final analysis, in producing illusions: the referential illusion and the enunciative illusion. One other dialectic movement should be noticed: contrary to what happens at the moment of disengagement, the effect of which is to referentialize the domain where its operation begins, engagement produces a dereferentialization of the discourse that it concerns: here again, can the intensity of deixis and of subjective modality be identified — indeed, the intensity of both manifestations of the subjectivity (deixis and modality) removes the world as the objective referent out of discourse.

Shifting out It can be noticed in discourse analysis that some peculiar types of discourse are characterized by the fact that the I-sayer has a special authority to withdraw (to become absent) — this is the case, for example, with scientific and didactic discourse. The procedures of shifting out are generally less observable and truly complex and subtle: the objective way of speaking in scientific language use is, in fact, a way of hiding the originating subjectivity with its general and also specific purposes and motives. But indirect com­ munication as well, and all kinds of "deviant" communication (lying, and manipulation in particular), are cases where the speaking subject is "shifting out" of his discourse. There are techniques for simulating the absence of performativity, expressivity and subjective investment in language frag­ ments, and these are the phenomena which are of the greatest difficulty to

338

HERMAN PARRET

be described and systematically explained. The problem consists in the fact that shifting out is an evident attack on communication and solidarity. The communicative norm is that the speaker expresses his intention to be com­ municative. Shifting out can mark all three aspects of deixis: personal or actantial deixis, temporal deixis and spatial deixis: I shifts out as a person by effectuating procedures where a kind of neutral He is realized (truth, for example, is an instance of universality, and it is intrinsic to truth that it is not subject bound). Temporal shifting out realizes a kind of u-topic time where temporal indexicahty is put into brackets and idealized. Topological neutralization is equally possible when the speaking subject spatially located withdraws or manipulates the interpreter with regard to possible detection of spatial coordinates (on these points, see Parret 1985). To use semiotic terminology, we can define "shifting out" or disengage­ ment as the operation by which the enunciation disjoins and projects forth from itself, at the moment of the utterance, certain terms bound to its base structure. The utterance thus appears as a split which creates, on the one hand, the subject, the time, and the place of the enunciation and, on the other, the actantial, spatial and temporal representation of the utterance. Actantial disengagement then will consist in disjoining a not-I from the subject of the enunciation and projecting it into the utterance, temporal disengagement in postulating a not-now distinct from the time of the enunci­ ation, spatial disengagement in opposing a not-here to the place of the enunciation. It is necessary to insist on the fact that the subject of the enunciation is always implicit and presupposed, that this subject is never manifested inside the utterance. Beginning with the subject of the enunci­ ation, implicit but productive of the utterance, either the actor of the enunciation or the actor of the utterance can be projected by their installation in the discourse. In the first case, an enunciative disengagement is carried out, in the second, an utterative disengagement (as in narrations which have commonplace subjects, in so-called "objective" discourse). It will be noted that each disengagement produces a referentialization effect: a "disengaged" discourse gives the impression that the narrative constitutes the "real situation" of the discourse: it referentializes the narra­ tive. Parallel to actantial disengagement, temporal disengagement can be understood as a procedure of projecting the term not-now at the moment of the utterance, out of the domain of the enunciation. This projection permits the construction of an objective time beginning from the position that can be called the then time. By considering the then time as a zero time and by applying (starting from that zero point) the categories of concomitance/nonconcomitance (anteriority/posteriority), it is possible to construct a model of the temporality of the utterance which permits the identification of the various types of narrative discourses. By the inverse procedure, the disen-

DEIXIS AND SHIFTERS AFTER JAKOBSON

339

gaged operations can thereafter be engaged so as to produce the illusion of their identification with the domain of the enunciation (this is temporal engagement). Spatial disengagement, again, is a procedure having as its effect the expulsion out from the domain of the enunciation of the term nothere founding the "objective" space of the utterance (the elsewhere space). It is apparent that the projection of the term here, simulating the location of the enunciation, is equally possible. Secondary referential systems for the actorialization, the temporalization and the spatialization of narrative dis­ courses are frequent. In the case of spatialization, for instance, the domain of the enunciation can be installed in the utterance in the form of a simu­ lacrum, and the space of here can be inscribed in the discourse as the reported space of the enunciation. Semiotics and narratology have as an urgent task to deduce all the typological possibilities of disengagements and engagements, and their superpositions, and to apply this scheme to as many specific texts and discourses as possible. Jakobson (1957) has been at the origin of the interest in deixis in contem­ porary linguistics. The category of "shifters" proved to be highly heuristic with regard to the complex analysis of subjectivity in language. It will be evident that Jakobson's phenomenological attitude made him sensitive to the subjective origin of productivity and creativity of natural language use. Of course, he never developed his findings in a philosophical direction: he remains a genial linguist with the right intuitions and accepts that his theorizing is constrained by empirical evidence (especially morphological data). But there is no clash with results in the philosophy of language on deixis and subjectivity. Moreover, it would be interesting to study the compatibility of Jakobson's views on shifters with the pioneering work of Benveniste (for instance, 1966) and Guillaume (1929). Benveniste's theory of pronouns and demonstratives is well-known and it has been of seminal importance to pragmatics (especially that type of pragmatics called the "linguistics of enunciation'). However, Guillaume's doctrine, being more speculative and even hermetic, incorporates explicitly most of the philosophi­ cal positions I have been defending: the specificity of demonstration with regard to representation, the egocentric organization of the whole set of demonstratives, the idea of I as a designation rule dependent on the context of use, and the claim of the First Person Authority with its opaque functions and powers. There are various developments of the enunciative paradigm as it has been proposed by Guillaume: Desclés (1976) and Fraser and Joly (1979) are bright examples of analysts of deixis along the lines inspired by Guillaume. Careful investigation would show that Jakobson's view, although more limited and local, in its dependence on the one hand the phenom­ enological tradition and on the other Bühler's proposals, is in full harmony with the insights of Benveniste's and Guillaume's writings on subjectivity in

340

HERMAN PARRET

language, amd more specifically on deixis. Indeed Jakobson, Benveniste and Guillaume — all three of them marginal to orthodox structuralism — can be considered to be the godfathers of the pragmatically oriented linguistics of the future. References Anscombe, G.E.M. 1975. "The First Person". Mind and Language, ed. by G. Guttenplan. Oxford: Clarendon Press, 45-66. Benveniste, E. 1966. Problèmes de linguistique générale. Paris: Gallimard. Bühler, . 1934. Sprachtheorie. Die Darstellungsfunction der Sprache. Jena. Burks, A.W. 1949. "Icon, Index and Symbol", Philosophy and Phenomenological Research 9. 685ffs. Collinson, W.E. 1937. Indication. A Study of Demonstratives, Articles and Other "Indicators". Language Monographs. New York. Davidson, D. 1984. "The First Person Authority". Dialectica 38. 101-111. Desclés, J.P. 1976. "Représentation formelle de quelques déictiques français". Ms. Université de Paris VII. Fraser, Th. and A. Joly. 1979. "Le système de la deixis. Esquisse d'une théorie d'expression en anglais". Modèles linguistiques, 1-2. 97-151. Greimas, A.J. and J. Courtés. 1979. Sémiotique. Dictionnaire raisonnée de la théorie du langage. Paris: Hachette. Guillaume, G. 1929. Temps et verbe. Théorie des aspects, des modes et des temps. Paris: Champion. Holenstein, E. 1974. Jakobson ou le structuralisme phénoménologique. Paris: Seghers. . 1985. Menschliches Selbstverständnis. Frankfurt/Suhrkamp. Jakobson, R. 1957. "Shifters, Verbal Categories, and the Russian Verb". Selected Writings, vol. 2, 130-147. The Hague: Mouton, 1971. . 1966. "A la recherche de l'essence du langage". Diogène: Les problèmes du langage 51. 22-38. Kaplan, D. No date. "The Logic of Demonstratives". Langacker, R.W. 1985. "Observations and Speculations on Subjectivity". Iconicity in Syntax, ed. by J. Haiman, 109-150. Amsterdam/Philadelphia, J. Benjamins. Leys, O. 1979. "Was ist eine Eigenname? Ein Pragmatisch Orientierter Standpunkt". Leuvense Bijdragen 68. 61-86. Parret, H. 1980. "Demonstratives and the I-Sayer". The Semantics of Determiners, ed. by J. van der Auwera, 96-111. London/Baltimore: Croom Helm. . 1983. "Shifting In and Shifting Out: Remarks on Deictics", Proceedings of the XIIIth International Congress of Linguists (Tokyo, 1982), 1068-1071. . 1985. "Time, Space and Actors: The pragmatics of development". Developmenal Mechan­ isms of Language, ed. by C. J. Bailey and R. Harris. Oxford: Pergamon Press, 131-148. Russell, B. 1905. "On Denoting". Mind 14. 479-493. Waugh, L.R. 1976. Roman Jakobson s Science of Language. Lisse: The P. de Ridder Press.

Praguean Structure and Autopoiesis: Deixis as Individuation* C.H. van Schooneveld Indiana University 1. Introduction It is self-evident that language structure, to the Pragueans, consists primarily of relations between linguistic units along the axis of — to use Jakobson's term — replacement. It is paradigmatic. This is, for instance, strikingly obvious from Trubetzkoy's pioneering analysis of vowel and consonant systems,1 Jakobson's well-known statement in Ottuv slovník naucny2 as early as 1932, and the approaches which the latter takes to phonology in the Preliminaries, the Fundamentals and the Sound Shape.3 While Jakobson's earlier formulations regarding meaning in morphology make partial use of contiguity relations, the semantic structures he arrives at by his analyses are paradigmatic. In his revisitation of the case system in the "Nabljudenija" of 1958 and especially of verbal grammatical structure in the Shifters of the preceding year, the distinction between syntagmatic and paradigmatic relations come even more sharply to the fore.4 It is important to emphasize this since in several other trends of contemporary linguistics the notion of linguistic structure is based on the syntagmatic axis or is derived from the equalization of the paradigmatic and syntagmatic axes. This is the case in various distributionalist currents, which essentially assume that the paradig­ matic differences between linguistic units consist of the differences in their (syntagmatic) distributions. 5 As far as I can see, this view is actually a transposition into synchrony of the diachronic concept of sound law, since one can see sound law as a formulation of a diachronic shift in distribution of one or more phonemes; note that sound laws are usually formulated in terms of environments, that is, in syntagmatic terms. 6 2. Invariants In the course of the years following the appearance of the Prague School, the notion of invariant moves more and more into the foreground. While

342

CORNELIS H. VAN SCHOONEVELD

in Trubetzkoy's conceptualization a preponderant role is played by the notion of opposition, that is, of the relation(s) between the units of language, in later years the problematics are centered around what the building mater­ ials of these oppositions are. Empirically this question reduces largely to binary oppositions, one member of which (the marked one) carries an additional signalizing feature (the marking) for which the other member (the unmarked one) is unspecified. The only holdout is the ternary equipol­ lent phonological opposition ranging between compactness and diffuseness. As far as semantic oppositions in grammatical, word formative and lexical morphology are concerned, I have in my own work never encountered anything but privative binary oppositions. Linguistic structure seems to be preponderantly asymmetrical and reduceable to a limited number of recurrent invariant features, that is, linguistic entities which are components of the larger composites they are instrumental in distinguishing. Bühler's characterization Zeichen an Zeichen "signs within signs" has, in fact, an application far beyond the status of phonemes as components of morphemes. Phonemes can in turn be broken down into smaller recurrent components (Trubetzkoy's and Jakobson's Merkmale, distinctive features), and the meanings conveyed by morphemes appear to be composites of recurrent conceptual (to use Jakobson's recent term) features. I shall come back to the nature of these conceptual features later on. Paradigmatic structure is asymmetrical in that in builds up from unmarkedness to marked units of ever growing complexity, that is, with ever growing cumulations of markings. Thus, if the markings are a, b and  and the categorial value of the unit is represented by  we obtain what is shown in Table 1. To represent structure graphically, Jakobson used in "Nabljudenija" a cubic diagram, which I adapt 7 in that, in my representation (Figure 1), the first marking (a) is assigned the upper plane, the second (b), the right hand plane and the third (c), the rear plane.

Table 1.

u u u u u u u u

+0

(unmarked unit) +a +a

+b +b

+a +a

+b +b

+c +c +c +c

343

DEIXIS AS INDIVIDUATION

nominative: accusative: dative: genitive in -u: genitive in -a: locative in -ú: locative in -e:

[unmarked] (θ) + extension (exť) + extension (exť) + restrictedness (restr') + objectiveness + objectiveness + restrictedness (restr') + objectiveness + extension (exť) + restrictedness (restr') + objectiveness

+ extension (exť)

(obj') (obj') (obj') (obj')

Figure 1.

3. Are the invariants ordered? The question arises as to the order in which the features should be assigned to the cube(s). While the composites listed in the table just given constitute an ordering depending on how many features they contain, the features themselves are not ordered in this table except alphabetically, but that is an extrinsic ordering based on the alphabetic ordering of the letter symbols and not an ordering of the features by the intrinsic material of which they consist. In fact, the question is: recurrent invariant components create a structure of their composites, but do they themselves constitute a structure? This will be the case if they in turn have (even smaller) components in common; if they themselves are constituted of recurrent components. As far as morphology is concerned, this question can be answered in the affirmative. At least in Russian, and evidently also in several other languages, it seems that we have to do with six cardinal invariant components which have two characteristics. One is that they are ordered in the sense that they are arranged in an implication relationship. Each succeeding feature, besides adding its own information, implies the information conveyed by the preced­ ing feature. Thus, each succeeding feature contains the information of the preceding feature and the information of each preceding feature is contained in the next feature. As far as the objects to which the features refer are concerned, the situation tends to be the reverse. The objects referred to by each succeeding feature are specified by more properties than the object indicated by the preceding feature and constitute therefore subsets of the set of objects designated by the preceding features.

344

CORNELIS H. VAN SCHOONEVELD

It is, secondly, of the utmost importance to note that the information carried by each feature is conceived in terms of the act of perceiving the object it refers to. In fact, the previous sentence contains a contradiction if one thinks of "objects" as "objects in external reality", because it follows from that sentence that the features can indicate only perception acts. It is only indirectly, because one perception technique is better adapted to identify one type of (external) object, and another perception procedure better serves to identify another sort of (external) object, that the conceptual features are able to distinguish between and refer to — that is, to identify — objects in external reality. Thus, the only contact that language has with external reality is the fact that there is an identification act.

4. The recodification of the identification act as the basis for the ordering (hierarchy) of the semantic invariants (features): autopoiesis As I stated, each succeeding feature is more specific than each preceding feature since each succeeding feature incorporates the information given in the preceding feature. This operates apparently in the following manner. As soon as a feature is actually applied in a given identification procedure, additional information is gathered during this identification procedure. If this additional information elaborates on the identification procedure, it is codified in turn. This (re)codification creates the next feature. Thus, the inclusion hierarchy of conceptual features builds itself by codifying the application of the preceding feature. One might say that each succeeding feature equals the preceding feature plus its codified application. Since this succeeding feature again can be applied in turn, thus leading, via (re)codifi­ cation of the application, to again a new feature succeeding the preceding succeeding feature in turn, one can say that the feature hierarchy creates itself, is autopoietic, by an intermittence between the application of the code and the code itself. An identification act leads to the codification of a new feature, a new identification act applying this feature leads to the codification of a new feature, this new feature is applied in turn, and so on. According to the recent biological theory of autopoiesis proposed by the two Chilean physiologists, H.R. Maturana and F.J. Varela, the perceptual mechanism of the central nervous system is built up in the same manner. The central nervous system monitors its own perception acts and incorporates in a preconceived manner the results of these observations into the perceptual mechanism, thus creating more sophisticated perception techniques. Auto­ poiesis is based on the fact that the system can observe and register its own operations. Meaning is evidently conceived in terms of the various perception techniques leading to an identification act. In other words, it seems that

DEIXIS AS INDIVIDUATION

345

meanings, being essentially instructions to perceive, are conceived precisely in terms of the working of the mechanism of the nervous system that actually does the perceiving and identifying. In language the hierarchy of six features apparently exhausts itself with the objectiveness feature; to judge by the statements of Maturana and Varela, in the physiology of the central nervous system this autopoiesis can go on endlessly.8 5. Deixis I have now arrived at a related subject, which is deixis. It can be subdivided into two points. The first is that the Russian linguistic code distinguishes between two types of identifiers. One type of identifier is general; the identifier can be anybody, whether he/she is at the same time the encoder or decoder of a speech transmission or not. The second type of identifier constitutes a subset of the first type. It is restricted to the identifiers who are at the same time the encoders (speakers) or decoders (receivers) of a transmitted utterance. The six features are differentiated into two varieties depending on whether we have to do with an identifier who perceives in general or with the "trans­ missional" identifier, so that, at least in Russian, there exist two varieties of six features, a "perceptional" type and a "transmissional" type. In this connec­ tion, the notions of "perceptional" deixis and of "transmissional" deixis are relevant. I shall come back to these notions shortly. The second point concerns directly the range of objects to be identified. Besides indicating the identifiers, semantic structure indicates the range of identifieds. In fact the ever progressing restrictions of the ranges of identifieds are the main theme that will occupy the rest of this paper. The crucial question is whether the linguistic sign specifically indicates that the identifi­ cation of a referent has been made before encoding or whether and to what extent such an identification is not specified. This indication is done by specifying ranges of scanning within which the identified may be found. There is a progression of four degrees of specification of the ranges of identifieds from perceptional deixis to transmissional deixis to singulative perceptional deixis to singulative transmissional deixis. The first stratum of deixis makes an appearance within the hierarchy of the six conceptual features, in the distinctness feature. I shall therefore begin the discussion with a brief presentation of the six conceptual features. 6. The six conceptual (semantic) features a. Plurality. The plurality feature sets up a multiplicity of perceptions. Whether these multiple perceptions apply to more than one object or just

346

CORNELIS H. VAN SCHOONEVELD

one is a secondary matter. Plurality creates an intuitive set; the concept of singleton set is obviously a refinement invented by mathematicians. The plurality feature simply says that a given perception act has to be repeated. Whether the result is a reference to a plurality of segments in external reality or a multiple of perception operations on the same segment is not specified. Any plurality will do. b. Dimensionality. The dimensionality feature assumes a set as background, but considers only a section of this set relevant as its perceptional objective. Thus the dimensionality feature instructs the sensory apparatus to scan for (an) element(s) that is (are) set off from its background altogether, or (an) element(s) that has (have) outlines. There is a limited number of relevant elements viewed with reference to a universal set of similar elements. The common semantic denominator of all these examples is the selection of one area, coupled by one or more elements, within a larger area; in other words, the creation of a subset. The selection of a subset from a given background is guided by contextual criteria; the subset itself, just as in mathematics, can have any properties that are capable of setting off its elements from their peers. If one considers the extensional subset of mathematics as randomly selected and the intensional subset as selected according to a priori given criteria, then one can say that the dimensionality subset is neither or both extensional and intensional at the same time. The selection is according to criteria chosen ad hoc; it will involve a set which stands out conspicuously from its background. Dimensionality involves a selection on a more funda­ mentally intuitive level than is presented by the two kinds of sets that are known in mathematics. It is evident that the concept of dimensionality implies the concept of plurality; a subset is created within a (containing) set. This is to say that the two features, plurality and dimensionality, are ordered: plurality precedes dimensionality. With (c) distinctness, the language introduces for the first time a specific notification that one particular referent has been designated. The designation of one particular referent derives from the codification of the contextual effects of the application of the dimensionality feature,  Distinctness. Dimensionality identifies any subset within a containing set. There are no inflexible criteria: any set that stands out within a larger set will do. There is no generally identifiable immutable criterion which disting­ uishes the subset from its background. Dimensionality is a quantitative, not a qualitative operation; the quality which makes the subset stand out is ad hoc. The application of dimensionality results in one (sub)set being identified ad hoc. In distinctness, the ad-hocness of the choice of the set selected is codified. Ad-hocness itself becomes the new invariant. The identity of the identification act itself becomes relevant. Having to do with a given identifi-

DEIXIS AS INDIVIDUATION

347

cation act with the exclusion of all others is at the basis of the instructions for identification conveyed by the distinctness feature. The distinctness feature announces that one specific referent (the question whether the refer­ ent consists of a multiple set or a singleton set is irrelevant [unmarked]) has been identified before encoding; this identification belongs to the code; it is codal. The identification act has become codal. Distinctness has deixis of the identification (perception act). It has perceptional deixis. The distinctness feature introduces deixis into the hierarchy of conceptual features. Within this hierarchy, it is the first feature to carry perceptional deixis. d. Extension. The simplest way to distinguish one perception procedure from another is by repetition. This is why plurality is the first conceptual distinc­ tive feature; it represents the simplest way of distinguishing one type of perception technique from another. In extension we find the same thing, except that extension is derived from distinctness and gives the instruction to identify more than once a referent that has already been identified before the encoding. In other words, extension continues and amplifies the deixis initiated by distinctness. It instructs to reidentify the same referent. Extension introduces the notion of time into the semantic system. e. Restrictedness. Restrictedness elaborates on extension. Extension de facto divides the narrated situation up into an initial period and a (non-finite) second period. Restrictedness states that only the second period is relevant with the exclusion of the first period. The new subset consists of the nonfinite second period. It is evident that the concept of space is based on the concept of time, since in order to establish space one needs at least two points. Two or more points imply in turn counting, and counting implies a succession in time. Thus restrictedness develops the notion of space out of the notion of time, which is created by extension. It also should be noted that restrictedness brings up the quest for the identity of the space in which the referent is located, or rather the two spaces which are relevant for the location of the referent, three times. First in the initial period of the narrated situation the identity of the first space is asserted. Second, the identity of the first space in the ensuing non-finite period of the narrated situation, with emphasis on the continuity of the identity of this first space throughout the two periods, is established. Third, the separate identity of the second period is established and the relevance of the first period and consequently of the first space is denied. Thus restrictedness incorporates in its concep­ tualization not only distinctness but also the ensuing extension feature. f. Objectiveness. In objectiveness, both the initial and non-finite second space are considered to be of equal relevance; the referent is potentially maximally distant from either. Spatially continuous relations tend to change into a relation of mere synchronization. In objectiveness, the feature hierarchy seems to reach its limit. The referent has become maximally independent of

348

CORNELIS H. VAN SCHOONEVELD

the narrated situation, the latter all the same remaining the orientation point for the identifiability of the referent. The hierarchy of the six semantic features is represented in diagram form in Figure 2.

7. Four strata of deixis The lexicon is, in Russian, the semantic structure concerned with the nar­ rated situation. Whether we have to do with a verb, a substantive, an adjective or an adverb, they all evoke a scene in the narrated situation. Components of such a scene (referents) may be identified by any observer Plurality

Dimensionality

Distinctness

Extension

Restrictedness

Objectiveness

ref :: signatum :: means : corresponds to the horizontal lines indicate the narrated situation the broken lines in Plurality and Dimensionality indicate a non-finite set the broken lines in Restrictedness and Objectiveness indicate cancellation Figure 2.

DEIXIS AS INDIVIDUATION

349

whomsoever (whether encoder [speaker] or decoder [receiver] of the speech transmission [utterance] or not) or by a more restricted type of observer, that is to say an identifier who is at the same time encoder (speaker) or decoder (receiver) of the transmitted utterance. In the first type, we have from distinctness on to do with perceptional deixis, whereas in the second type we have to do with transmissional deixis. The latter type of deixis is the type that is called "deixis" in traditional grammar. In the first type, one can talk about an initiator of the identification procedure (the first identifier) and respondents (subsequent identifiers who are witnesses of the first identification); in the latter, the speaker and receiver are among the initiator and respondents, respectively. It may be useful to note at this point that I shall be distinguishing respondent (subsequent identifier) from receiver (addressee or decoder of an utterance [parole]). There is a second stratum of perceptional deixis, which is characteristic of several categories, among which pronouns present the most conspicuous type. In the singulatively perceptional deictic type, all identifications, the initiator's as well as the recipient's, have to be performed at the same moment. In the demonstrative pronouns the procedure is obvious. To be sure, in pronouns in general the moment of identification does not have to be the moment of encodement; in a sentence like he who kills his neighbor should be punished the referent of the pronoun he is identifiable beyond the moment of encodement. Another way of formulating singulativeness is to say that while (non-singulative) perceptional and (non-singulative) transmissional deixis generalize individual perception acts through recodification, the ulti­ mate recodification represented by singulativeness generalizes the non-generalizability, that is, the absolute individuality of the perception act. 9 This is why pronouns have no generic meaning but are purely deictic in the sense of pointing (cf. Gr. deiknumi). By singulative transmissional deixis a form refers to the fact of its own pronunciation as a unique event. It is character­ istic in relatively antique (inflectional) Indo-European languages of the categorial meaning of word formation (an operation on a preceding mor­ pheme) and of the categorial meaning of grammatical morphemes, which relate the referents of the morphemes which precede in the same word to the speech situation.

8. Are transmissionality and singulativeness ordered? Introduction Thus, four types of deixis multiply the six semantic features by four. In a relatively antique type of Indo-European language like Russian, the semantic features are implemented to constitute the signifieds of morphemes forming

350

CORNELIS H. VAN SCHOONEVELD

a word as indicated in Table 2. I am listing the optimal number of features: to be sure, only a few of them occur in a word at the same time. Lex indicates "lexical meaning", wf "word formation", and p.o.s. "part of speech" ("word category") (Table 2). We must now ask ourselves two questions. Question One: do the four types of deixis constitute a structure? If so, we must conclude following the reasoning illustrated in Figure 1 and Table 1, that they are ordered. The answer is obviously affirmative. There are two varieties of deixis marked by singulativeness as opposed to two unmarked varieties. Also, transmissional deixis carries more information than percep­ tional deixis and is therefore the latter's containing set; perceptional deixis constitutes the subset. Conversely, perceptional deixis tends to have a larger range of referents. We can represent the four types of deixis as in Figure 3. Question Two: we have now reduced the structure consisting of four varieties of deixis to two markings, transmissionality and singulativeness. Are these markings ordered in turn? The answer to this problem should begin with one preliminary question. We are now trying to classify and order two invariants which create four types of deixis. Within this structure of four deictic types, perceptional deixis may be called unmarked since it is neither transmissional nor singulative. Is it, however, the case that all conceptual features are marked in turn for deixis or are there conceptual features that are unmarked for deixis alto­ gether? The last alternative is the correct one. A type which is unmarked for deixis also exists. It is represented by the first two features, plurality (pluf) and dimensionality (dim') within the feature group which after dim' becomes perceptionally deictic with distinctness (dist')· In other words, the hierarchy to which the perceptionally deictic features belong is only two thirds percep­ tionally deictic: dist', ext', restr' and obj' are perceptionally deictic, but plur' and dim' are not. The difference between non-deixis and the two degrees of deixis consists in the information released by the encoder to the addressee (decoder) regard­ ing the ranges represented by perception situations which the receiver must scan in order to identify the referent which the speaker has in mind. In the case of unmarkedness for deixis (non-deixis) there is absolute indifference in the sign as to the objects to which the plurality and dimensionality features are actually applied as well as to the observers which witness the plurality and dimensionality features operate. Thus, in the case of the perceptional plurality feature, there is no deixis. The sign implies that any plurality will do, whether a plurality of perceptions on a single object (as is the case of Russian polite Vy "you" cf. French vous, German Sie) or a plurality of perceptions, each one on a similar object (as

plur'" dim'" dist'" ext'" restr'" obj'"

plur'" dim'" dist'" ext'" restr'" obj'"

—

plur'" dim'" dist'" ext'" restr'" obj'"

plur" dim" dist" ext" restr" obj"

plur' dim' dist' ext' restr' obj'

tat

From dist' on, single primes are perceptionally deictic. Double primes: transmissional deixis. Triple primes: singulative perceptional deixis. Quadruple primes: singulative transmissional deixis.

plur" dim" dist" ext" restr" obj"

plur" dim" dist" ext" restr" obj"

ili

plur' dim' dist' ext' restr' obj'

—

plur' dim' dist' ext' restr' obj'

hab

—

plur'" dim'" dist'" ext'" restr'" obj'"

plur'" dim'" dist'" ext'" restr'" obj'"

plur'" dim'" dist'" ext'" restr'" obj'"

plur" dim" dist" ext" restr" obj"

plur' dim' dist' ext' restr' obj'

[p.o.s.] [gender]

restr"" (agreement)

plur'" dim'" dist'" ext'" restr'" obj'"

plur" dim" dist" ext" restr" obj"

plur' dim' dist' ext' restr' obj'

[pers. pron.]

obj"" (agreement)

*The features for which the word is actually marked are italicized.

plur" dim" dist" ext" restr" obj"

plur' dim' dist' ext' restr' obj'

plur' dim' dist' ext' restr' obj' plur" dim" dist" ext" restr" obj"

[case] [number]

em

Table 2. The morphemic composition of the Latin accusative singular habilitatem "aptitude" * 0"" plur"" dim"" dist"" ext"" (lex) (wf) (..s.) (grammatical) (agreement)

DEIXIS AS INDIVIDUATION

351

352

CORNELIS . VAN SCHOONEVELD transmissional deixis

perceptional deixis

singulative transmissional deixis

singulative perceptional deixis

Figure 3.

in the case of books). Indications as to which alternative the encoder has in mind are not encoded and must be deduced by the decoder from the context. The same applies to perceptional dimensionality, which instructs the decoder to perform a subtraction operation by one or more from any plurality. Perceptional plurality and perceptional dimensionality are therefore purely mathematical operations in the sense that they apply to any objects that satisfy certain numerical conditions. With the perceptional distinctness feature, perceptional deixis makes an appearance. The code starts informing the decoder that he must limit his scanning for the referent to a restricted range which somebody, the initiator (not necessarily the encoder) has selected. The selection of the referent by the encoder cannot merely be read off from the other contextual cues. This results in arbitrariness being introduced as a contextual cue, which means in turn that instead of all contextual cues directly usable for deducing the identity of the referent, we must now look for an indirect cue. In the perceptionally deictic features which succeed perceptionally deictic distinctness in the hierarchy, that is, extension, restrictedness and objective֊ ness, the fact that a referent has been selected before direct contextual cues came into play remains semantically relevant. Once the initiator has ident­ ified a referent, this referent remains identified. Extension identifies the same referent at least twice (and may therefore be considered a perceptionally deictic plurality-like feature), and restrictedness and objectiveness specifically identify the same referent at least three and four times, respectively. As far as transmissional deixis is concerned, the range of scanning for the referent by the decoder is even more restricted by the encoder than in perceptional deixis. The decoder is now instructed to identify the referent within a range which is restricted to the encoder as a witness to the facts of the narrated situation. The encoder is necessarily one of the initiators or respondents. The decoder can now start guessing as to the identity of the

DEIXIS AS INDIVIDUATION

353

referent with more cues at his disposal. The range of his uncertainty is more restricted than in the case of perceptional deixis: the identity of the referent is restricted to those elements of the narrated situation of whose identity the encoder himself has personal knowledge. Finally, both singulative perceptional and singulative transmissional deixis have this in common that the identification of the referent must be done by the decoder (I shall come back to this point shortly) as well as by the witness to the narrated situation (initiator and respondent[s]) for singulative percep­ tional deixis, and by the witness to the speech situation (encoder and decoder, i.e. speaker and respondent[s]) for singulative transmissional deixis, respectively, at least at the moment at which the narrated situation and the speech situation, respectively, are taking place. This is not the case with non-singulative perceptional deixis and non-singulative transmissional deixis. In non-singulative perceptional deixis it is enough that the decoder knows that at some time a specific referent was identified in a specific narrated situation, and in non-singulative transmissional deixis it suffices that the decoder knows that at some time in a specific narrated situation a specific referent was identified with the knowledge of the encoder (speaker). To be sure, it suffices that the encoder knows the identity of the referent; he does not have to be present at the first identification of the referent in the narrated situation. Thus Russian idti "to go", "to arrive" (marked by ext" [determinacy]) does not imply that the speaker was present at the moment the agent arrived at the final destination, but merely that the identity of this final destination is known to him. Since singulative perceptional deixis indicates that the identification of the referent must be performed by the decoder as well as by the witness to the narrated situation at the moment that this narrated situation takes place, singulative perceptional deixis leads, among other things, to pronominality, and to anaphora. Demonstrative pronouns present a particularly clear example. Personal pronouns seem to be marked additionally by (non-singul­ ative) transmissional deixis. Singulative perceptional deixis also occurs in the numerals, in gender, in proper names, in the prepositions and conjunc­ tions, and in verbal moods. Singulative transmissional deixis, which refers in the first place to the fact of pronunciation, indicates the relation of a morpheme to other morphemes in the same word or the same sentence. It distinguishes, for instance, lexical, word formative and grammatical mor­ phemes and signalizes agreement. Any intended transmission of information, whether linguistic or not, consists of more than one perception act. The transmitter must be transmit­ ting something to a receiver. The transmitter must have perceived what he is transmitting. The receiver sets a range which he scans in order to zero in on (perceive) the transmitted. Thus the transmitter can be seen as the first

354

CORNELIS H. VAN SCHOONEVELD

perceiver, or initiator of the consecution of perception events, and the first perception by the transmitter as the initiation event in that it initiates a series of perception events.

9. Are perceptionality, transmissionality and singulativeness autopoietically ordered? Perceptionality and transmissionality It is evident that in the series non-deixis, perceptional deixis, transmissional deixis and singulativeness we have to do with a progressive contraction of scanning ranges which the encoder presents to the decoder. In perceptional deixis we start envisaging a specific (narrated) perception (identification, initiation) event, while in transmissional deixis this identification (initiation, narrated) event must entail identification by the (language-given) encoder at some time or another. In singulativeness one might say that the narrated event and the identification by the initiator and respondent must be observed simultaneously, since identification by the respondent must take place at the same time as the identification of the narrated event by anyone, the initiator included. At this point we can ask ourselves the question whether the transition from perceptional deixis to transmissionality takes place, as is the case within the hierarchy of conceptual features, via the recodification of the application of perceptional deixis. Is the transition created through autopoiesis? The transition from perceptional deixis to transmissional deixis is actually a transition from a given perception situation in general to a situation in which information is transmitted. The fact that this transmission is perfor­ med through the intermediary of language seems to me of secondary signifi­ cance. What is important is that perceptional deixis codifies an identification event that takes place in the narrated situation, that is, in any situation that can be observed. What transmissional deixis codifies is the perception of the speech event by the decoder. It is, therefore, the decoder who identifies the encoder. If the encoder identifies himself he acts as decoder. Hence, we must consider in Saussure's parole the moment of decoding (reception of the utterance by its receiver) more important than the encoding. But for what purpose does the encoder encode and the decoder decode? The encoder sends perceptional cues so that the decoder may identify the referent of the message, that is, an element in the narrated situation. Thus, what trans­ missional deixis actually does, is introduce a second identifier (the respon­ dent) of the referent; and the latter is, of course, a component of the narrated situation. Attempting a somewhat different formulation, we can say that transmissional deixis introduces the separation of two perception events.

355

DEIXIS AS INDIVIDUATION

The first perception event is the first identification event, which takes place in the narrated situation and information about which the encoder transmits to the decoder. If codified, it creates perceptional deixis. The second percep­ tion event is the identification of the narrated event by the decoder. If codified, it creates transmissional deixis. If we disregard the specifically linguistic nature of the transmission we can limit ourselves to the statement that we have to do with more than one identification of the narrated event. Distinguishing the second identification (by the respondent) from the first identification (by the initiator) implies in a linguistic transmission situation precisely distinguishing between decoder (receiver) and encoder (sender). This implies in turn, in a linguistic (speech) situation, the decodement of the sign, that is, decoding the signifier and recognizing the signified(s) it conveys. The decipherment of the message is a prerequisite for the second identification (by the decoder) of the referent. Of course, these two identification events exist in principle in any appli­ cation of the linguistic code in a communicative event, but in transmissional deixis they do not only take place in the communication event (parole), but they are incorporated into the sign (code) even before it is transmitted. Transmissional deixis, then, creates a duality (or if there is more than one decoder, a plurality) of identifications. The oppositions between the four strata of deixis are summarized in the matrix given in Table 3.

10. Transmissionality and singulativeness What does singulativeness do? It synchronizes these identifications into one single pulse in time. If such identification is done in the narrated situation, then we have to do with singulative perceptional deixis. Singulative percep­ tional deixis occurs in pronouns, proper nouns, gender, numerals and moods. Later on I shall briefly discuss, as an especially easily understood example, pronominality. It is useful to note a few points. Firstly, the idea of a single perception pulse in time means that this possibly multiple but simultaneous Table 3. deixis

initiator ≠ respondent initiator respondent

percep.

transm.

singula tive percep.

0 0

+ 0

+

0

transm.

+ +

356

CORNELIS H. VAN SCHOONEVELD

pulse is separate from any other pulse; it is separate from its peers. In other words, it seems to be characterized by a variety of dimensionality. Secondly, what is typical of dimensionality is that it indicates a subset within a containing set. But dimensionality does not indicate the size of the subset. It does not even indicate whether the subset is a singleton set or not, just as the singular (number) is unmarked for this information. In other words, as far as this information is concerned, dimensionality reverts to a preplurality situation. Plurality is merely the codal prerequisite for the autopoiesis of a subset. Whether we have to do in dimensionality with a singleton set or a proper subset has to be deduced from the context. Thus, the context informs us in the Russian phrase čaj v stakane "the tea in the glass", that the referent of the prepositional object of v stakan(e) "glass" is singular; it represents a singleton set. In the phrase v tjažëlye dni katastrofy "during the painful days of the catastrophy" the plural dni "days" tells us that the prepositional object of v "in" is a multiple. Similarly, a demonstrative pronoun like this may be connected with one single perception, that of the speaker acting as initiator. The addressee acting as recipient may not be able to identify the referent. What the demonstrative pronoun, that is, singulative perceptional deixis, tells the addressee to do, though, is to identiify at the same moment the speaker identifies. Thus the plurality of percep­ tions in singulativeness, just as in dimensionality, is optional, whereas in transmissionality, it is obligatory. Thirdly, this leads us to the relation between pronominality and singulative perceptional deixis in general. On the one hand I discuss pronominality because it presents particularly clear examples of singulative perceptional deixis. On the other, I leave personal pronouns out of the discussion since they are characterized by a combination of (non-singulative) transmissional deixis and singulative perceptional deixis. The point I want to make is that pronominality does not orient itself directly upon the speech situation. Demonstrative pronouns like this and that are usually described by linguists as indicating closeness or remoteness from the speaker. This is only indirectly correct. What pronominality does is invite all potential observers of the referent to a reunion devoted to one spell of identification of the referent for all of them at the same time. Singulative perceptionality glorifies the uniqueness of the identifier(s). When singulative perceptionality is applied in an actual parole, it creates a separate plane supporting perception relations which is visible, among others, from the plane of the speech situation. Thus the speaker and addressee can project themselves upon the singulative perception plane as the once-occurring observers of the singulative (once-occurring) perception (narrated) situation. However, these once-occurring observers may have a more general nature than speaker and addressee. Singulative perceptional reference may take on a nuance of hypotheticalness. Thus, for instance, in Russian tot čelovek

DEIXIS AS INDIVIDUATION

357

kotoryj ub'ët svoego soseda budet nakazan "that (the) man who kills his neighbor shall be punished" we have to do with a unique relationship between the subject of a relative and the subject of a main clause identifiable at any moment in time simultaneously by any group of observers. For nonsingulative perceptional deixis the simultaneousness of the identifications is not required. The projection onto the singulative perception plane from the parole plane is very much comparable to the omniscient author projecting himself (and his readers) onto the plane of the narration contained in a novel. The relation between the plane of the speech situation and the plane of singulative perception can be represented as in Figure 4. To be sure, this diagram is merely intended to illustrate the role of speaker and addressee in singulative perceptional deixis. But the speaker and addressee are just contextual elements created by a linguistic communication situation {parole). The singulative perception situation should be drawn as in Figure 5.

Figure 4.

Figure 5.

358

CORNELIS H. VAN SCHOONEVELD

What happens in singulative transmissionality? The option of a single perception is removed. We have to do with a synchronization of a multiple (at least two, like in any plurality) of perceptions. In a — not necessarily linguistic — communication situation this means, as we saw under trans­ missionality, one original identifier and one or more identifiers who follow the lead of the first identifier; initiator and respondents, one might say. In a linguistic transmission situation, initiator and respondent(s) correspond to speaker and receiver(s). When transmissionality is cumulated with singulativeness, initiator (encoder, speaker) and respondents] (decoder[s], receiver[s]) must all identify in the same time pulse. This means that they all must hear the transmission of the utterance, that is, the parole while it is being performed. The set of identifiers is restricted to those who hear the utterance while it is being pronounced. The attention of the identifiers is directed in the first place upon the signifier. Singulative perceptional deixis causes the sign to signify its own once-occurring pronunciation. Thus, singulative transmissional plurality means that a morpheme is a component of a multiple of at least two "morphemic" pronunciations (i.e. of mor­ pheme[s]): it signifies word formation. Singulative transmissional distinctness asks the question what the speaker intends (means) by pronouncing this form. It tells the addressee that the speaker invites the addressee to deduce by circumstantial evidence what significance the referent has for the speaker and addressee at the moment of speaking. It signifies grammatical meaning. Thus, the fact of a sign's being pronounced is a sign in its own right and gives rise to a new range of signifieds. The signata carrying singulative perceptional deixis (notation by quadruple primes) all deal with the signify­ ing power of an individual pronunciation of a sign. They all give information about the signifying power of the given pronunciation. To give one striking example: extension means that a referent as it is seen at the moment that information about it is relevant has minimally changed since it was affected by the influences of the original narrated situation. In other words, after the original situation to which the sign makes a reference in retrospect has taken place, the referent remains maximally the same. What will ext"" mean? The referent is the signifying power of the given pronunciation of the given sign (morpheme). The moment about which the information is given is the moment at which the signifying power of the given pronunciation takes its effect; that is, the moment of pronunciation of the given morpheme. The original narrated situation is a situation to which the morpheme refers back, in casu, an earlier actualization of the same signifying power. It is not hard to see that ext"" means agreement. The agreement ending -am of novam in the Latin phrase mensam novam "the new table (fem.acc.sg.)" or the agree­ ment ending -uju of novuju in Russian novuju knigu "the new book (fem.­ acc.sg.)" states about the signifying power of the pronunciation of the

DEIXIS AS INDIVIDUATION

359

agreeing and not of a direct grammatical morpheme that it has remained maximally the same since an earlier instantiation of the signifying power of a morphemic pronunciation, to wit, the pronunciation of the direct gram­ matical morpheme {-am in Latin mensom and -u in Russian knigu). By means of restr"" and obj"", agreement seems even to be able to distinguish between various morphological categories, such as gender, lexicon and grammar, to which the agreeing morpheme may refer back. 10 (see Table 2). In order that singulativeness may synchronize indentifications into one single pulse in time, there must be awareness on the part of the initiator of the identification and of the respondent(s) [subsequent identifier(s)] of their own roles in the identification mechanism. The participant in the identifi­ cation situation must be able to observe himself as an identifier. The result produced by this self-observation is that the participant in the identification situation singles out through synchronization a set of identification acts from identification acts that might have succeeded each other in time, as is the case in non-singulative perceptional and transmissional deixis. To bring the preceding argumentation to a conclusion: one is tempted, after viewing transmissionality as a variety of plurality, to consider singul­ ativeness as a subset of that plurality, that is, as a variety of dimensionality.

11. Conclusion: The hierarchy of deixis I submit, then, that transmissionality is a type of deixis marked by plurality and singulativeness a type of deixis marked by dimensionality. Perceptionality represents the unmarked variety of deixis. However that may be, it seems plausible that the markedness hierarchy of deixis progresses from unmarkedness for deixis to perceptional (unmarked) deixis to transmissionality to singulativeness. In this respect, it is interesting to note the markedness relations in the lexicon. In the adjectival and substantival lexica, nonsingulative perceptionality is succeeded by non-singulative transmissionality; they both give regular lexical entries. 11 I indicate these two types by single and double primes: plur' etc. and plur" etc. (see Table 2). Lexical singulative perceptional deixis (plur"' etc.) creates pronominality, whereas singulative transmissional deixis (Θ"", plur"" etc.) creates signata of an entirely different nature and leads to the concatenation: lexical morpheme — word formative morpheme — part of speech morpheme — grammatical morpheme — agree­ ment morphemes. It is evident, however, that already pronominality (plur"' etc.) has a different character than the the non-singulative types. Pronomi­ nality carries no generic meanings. Thus, the progression of four types of deixis goes from perceptional deixis to transmissional deixis to singulative perceptional deixis and finally to singulative transmissional deixis (e.g. dist',

360

CORNELIS H. VAN SCHOONEVELD

dist", dist"', dist""). The two types of perceptional deixis (e.g. dist' and dist"') are interrupted by the intrusion of transmissional deixis (e.g. dist") and the two types of transmissional deixis (e.g. dist" and dist"") are interrupted by the intrusion of singulative perceptional deixis (e.g. dist'"). The two types of non-singulative deixis (e.g. dist' and dist") come first, and then come the two types of singulative deixis (e.g. dist"' and dist""). What happens is that singulativeness doubles the opposition between perceptionality and transmissionality (see Figure 3, p. 352) just like the opposition: unmarked for dimensionality vs. marked for dimensionality in na (unmarked: θ dim') vs. ν ( + dim') is doubled by the addition of distinctness in the Russian prep­ ositions za (θ dim', + dist') and pod ( + dim', + dist'), and extension doubles the above four prepositions as the unmarked members of the opposition: unmarked for extension vs. marked for extension in po (Θ dim', θ dist' + ext'),  ( + dim', θ dist'+ ext'),  (θ dim', + dist', +ext'), and s ( + dim', + dist', + ext'); see Figure 1 on p. 343. Thus the semantic material itself also as far as cumulations of features are concerned confirms the conclusion that in the hierarchy of deixis, transmissionality comes before singulativeness. Notes *

1.

2. 3.

4.

In other work of mine written after this article, the term dimensionality has been replaced by demarcatedness, distinctness by preidentity, extension by verification and restrictedness by cancellation. Instead of perceptional deixis, I use now the term identificational deixis. N. [S.] Trubetzkoy, "Zur allgemeinen Theorie der phonologischen Systeme", Travaux du cercle linguistique de Prague I, Prague 1929, pp. 39-67; idem, "Die phonologischen Sys­ teme", TCLP 4, Prague 1931, pp. 96-116; idem, Grundzüge der Phonologie, TCLP 7, Prague 1939. English version in R. Jakobson, Selected Writings I 2 , The Hague 1971, pp. 231-233. R. Jakobson, C.G.M. Fant, M. Halle, Preliminaries to Speech Analysis, Cambridge, Mass. [1951]; R. Jakobson and M. Halle, Fundamentals of Language (janua Linguarum, 1) The Hague, 1956; R. Jakobson and L.R. Waugh, The Sound Shape of Language, Bloomington, Indiana, 1971. R.. Jakobson, "Beitrag zur allgemeinen Kasuslehre", Travaux du cercle linguistique de Prague 6, Prague 1936, pp. 240-288; English translation: "Contribution to the General Theory of Case: General Meanings of the Russian Cases", Russian and Slavic Grammar, L.R. Waugh and M. Halle, eds. (Janua Linguarum series maior 106) Berlin 1984, pp. 53-103; idem, "Morfologičeskie nabljudenija nad slavjanskim skloneniem", American Contributions to the Fourth International Congress of Slavicists, Moscow, September 1958 (Slavistic Printings and Reprintings, 21), The Hague 1958, pp. 127-156; English trans­ lation: "Morphological Observations on Slavic Declension (the Structure of Russian Case Forms)", Russian and Slavic Grammar, 1984, pp. 105-133; idem, "Zur Struktur des russischen Verbums", Charisteria ... Mathesio ... oblata, Prague 1932, pp. 74-84; English translation: "Structure of the Russian Verb", Russian and Slavic Grammar, pp. 41-58; idem, Shifters, Verbal Categories and the Russian Verb, Department of Slavic Languages

DEIXIS AS INDIVIDUATION

361

and Literatures, Harvard University, Cambridge, Mass. 1957, and Russian and Slavic Grammar, pp. 41-58. 5. E.g. Z.S. Harris, Methods in Structural Linguistics, Chicago, Illinois [1947] (e.g. pp. 5-7 ["relevant ... elements must be set up on a distributional basis."]). Taken as a point of departure by N.A. Chomsky in The Logical Structure of Linguistic Theory, New York [1975], p. 42. Note that Harris creates a vicious circle since he goes on to say: "Since this assumes that the other elements B, C, etc. are recognized at the time when the definition of A is being determined, this operation can be carried out only if it is carried out for all elements simultaneously. The elements are thus determined relatively to each other and on the basis of the distributional relations between them." Harris assumes beforehand that there is a difference between ,  and  in order to operate with his distributional basis, but then goes on to say that their differences are in their differential distributions. 6. C.H. van Schooneveld, "Neogrammarian Sound Law and Syntagmatic Structure", Papers on Linguistics and Child Language, Ruth ... Weir Memorial volume, V. Honsa and M.J. Hardman-de-Bautista, eds. (Janua Linguarum, series maior 65) The Hague [1978], pp. 245-248. 7. C.H. van Schooneveld, Semantic Transmutations, Prolegomena to a Calculus of Meaning, I: The Cardinal Semantic Structure of Prepositions, Cases and Paratactic Conjunctions ... in Russian [Bloomington, Indiana] 1978, p. 13. 8. H.R. Maturana and F.J. Varela, Autopoiesis and Cognition (Boston Studies in the Philosophy of Science, 42) Dordrecht, 1980, pp.xvii, 23, 29, 39, 41, 43-44, 46, 51, 127, 129 and 133. 9. C.H. van Schooneveld, "Programmatic Sketch of a Theory of Lexical Meaning", Quaderni di Semantica, Vol. IV, No. 2, December 1983, Bologna, pp. 162 and 168-169. 10. C.H. van Schooneveld, "Agreement in Russian", in Language and Literary Theory: In Honor of Ladislav Matejka, Ann Arbor 1984, pp. 189-214. 11. C.H. van Schooneveld, Quaderni di Semantica IV 2 , 1983, pp. 164-167.

Bibliography For further substantiation of the statements made in the present paper I refer to: Sangster, R.B., Roman Jakobson and Beyond: The quest for the ultimate invariants in language. Berlin: Mouton, 1982. van Schooneveld, C.H., Semantic Transmutations: prolegomena to a calculus of meaning, Vol. I. Bloomington, Indiana, 1978. . "The Morphemic Structure of the Slavic Word and Greenberg's Twenty-eighth Univer­ sal". Slavic Word, ed. by D.S. Worth, 443-448. The Hague, 1973. . "By Way of Introduction: Roman Jakobson's tenets and their potential". Roman Jakobson, Echoes of his Scholarship, ed. D. Armstrong and C.H. van Schooneveld, 1-11. Lisse: Peter de Ridder Press, 1977. . "The Place of Gender in the Semantic Structure of the Russian Language". Scandoslovica 23 (1977), 129-138. . "Contribution à l'étude comparative des systèmes des cas, des prépositions et des catégories grammaticales du verbe en russe moderne". Studia Slavica Hierosolymitana 2 (1978), 41-50. . "A Semantic Approach to Word Formation in Contemporary Standard Russian". American Contributions to the Eighth International Congress of S'lavists, 579-615 Columbus, Ohio: Slavica, 1978. . "A Semantic Proteus: the transitivity feature in Russian", Studia Linguistica in Honorem Valdimiri L Georgiev 377-385. Sofia, 1980.

362

CORNELIS H. VAN SCHOONEVELD

. "The Extension Feature in Russian". Slavic Linguistics and Poetics: Studies for Edward Stankiewicz on his 60th Birthday 17 November 1980, International Journal of Slavic Linguistics and Poetics, Volume XXV/XXVI. 1982, pp. 445-457. . "Contribution to the Systematic Comparison of Morphological and Lexical Semantic Structures in the Slavic Languages". American Contributions to the Ninth International Congress of Slavists, Kiev, September 1983. Volume I, 321-347. Columbus, Ohio: Slavica. . "Programmatic Sketch of a Theory of Lexical Meaning". Quaderni di Semantica 4:1 (1983), 158-170. . "Comments to other Contributions of the Roundtable". Quaderni di Semantica 4:2 (1983), 117-124. . "The Place of the Ergative Within the Category of Case". Signs of Friendship: To Honour A,G,F, van Holk, Slavist, Linguist, Semiotician, 225-256. Amsterdam, 1984. . "Agreement in Russian", Language and Literary Theory: In Honor of Ladislav Matejka, 189-214. Ann Arbor, 1984. . "Is the Vocative a Case?" Pragmatics and Linguistics: Festschrift for Jacob L. Mey 179-186. Odense, 1986. . "Jakobson's Case System and Syntax". Case in Slavic, 373-385. Columbus, Ohio: Slavica, 1986. . "Ancient Greek and Modern Russian Prepositions: A speculative comparison". Inter­ national Journal of Slavic Linguistics and poetics 31/32, 1985 (Henrik Birnbaum Festschrift), 495-514. . "Linguistic Structure and Autopoiesis". Language, Poetry and Poetics ... Proceedings of the First Roman Jakobson Colloquium, at the Massachusetts Institute of Technology October 5-6 1984, ed. by K. Pomorska et al., 123-142. Berlin: Mouton 1987. . "Paradigmatic Structure and Syntactic Relations". In The Prague School and Its Legacy, ed. by Y. Tobin, 109-122. Amsterdam; John Benjamins, 1988. . "Syntagmatic Relations and Paradigms: Tenses and moods in ancient Greek verbal structure. A semantic analysis of the ancient Greek verb system". In From Sign to Text, ed. by Y. Tobin, 99-122. Amsterdam: John Benjamins, 1989. Waugh, L.R., Roman Jakobson's Science of Language. Lisse: Peter de Ridder Press, 1976.

Shifters and Non-Verbal Categories of Russian* Olga T. Yokoyama Harvard University Roman Jakobson's penetrating analysis of shifters and verbal categories in the Russian verb (1957) has significance that reaches far beyond the immedi­ ate object of his study. The full range of linguistic phenomena that could benefit from the category-distinctions Jakobson makes still remains to be defined. Nevertheless, it is already evident that the shifters "speech event" (Es) and "participant of the speech event" (Ps) can be shown to be crucial factors in the characterization of nearly all aspects of language as a means of communication, and that they affect all of the components of its structure: lexicon, phonology, morphology, syntax, and discourse grammar. In this paper I would like to demonstrate the role of the category P s in some of these areas. I will suggest that for an accurate account of the data I present, the category P s needs to distinguish between two participants, the speaker (Ps1) and the hearer (P s 2 ). I will also propose an explicit model for the interaction of the four sets of knowledge (henceforth, "knowledge sets") involved in a discourse situation (i.e. the knowledge set of the speaker, the knowledge set of the hearer, and the knowledge sets of the current concern of the speaker and the hearer respectively), suggesting that the interaction of the participants of a Es can be best defined in terms of the intersections of a Venn diagram. 1. Shifters and referential expressions Consider first the involvement of shifters in the process of choosing a particular noun or pronoun to designate a particular referent. 1.1 Personal names Personal names represent a classic example of "code referring to code" (C/C). In most cultures, a person has several names. Russian is particularly

364

OLGA T. YOKOYAMA

rich in the possibilities it offers for naming one and the same person: the last name, first name, and patronymic can be used either alone or in combinations; social titles (e.g. Mr., Dr., Professor, Comrade) may be used (primarily with the last name); and first names run a gamut of diminutive and pejorative formations that exploit the language's rich suffixal mor­ phology. In a given Es, however, the choice of a particular name is far from being random. An obvious factor is the relationship of Ps1 and P n : if the speaker is peer or superior, and if s/he feels close to the P n , perhaps with affectionate overtones, s/he might choose the diminutive Tanjuša; whereas if s/he is younger, not terribly formal but still respectful, s/he would be expected to use the first name and patronymic, Tat' jana Mixajlovna (Pn and P s 2 , in this example, may of course be identical). But the relationship between the speaker and the person referred to is not the only thing that determines the choice of the name. A more subtle (but equally important) factor is the relationship of Ps1 and P s 2 (provided P s 2 ≠ Ρ n ). Even if the speaker is on Tanjuša terms with the woman referred to by this name, the speaker may be on distant terms with his/her interlocutor; in this case the speaker is very unlikely to refer to Tanjuša. The speaker will avoid exposing the fact that s/he is on Tanjuša terms with the referent, and will refer to her by a less affectionate form of the first name (e.g. Tanja or Tatjana), or by the first name with the patronymic {Tatjana Mixajlovna), or even "Comrade Petrova", depending on what s/he thinks the social setting dictates. 1 For an American analogue, consider the suburban housewife who refers to her husband as "Mr. Jones", when speaking to the plumber.

12 Endearing and pejorative nouns Endearing and pejorative nouns, prolifically generated by Russian wordformation processes, are characterized in the same way as personal names. On the one hand they reflect the speaker's attitude to P n (Ja nedavno videla takuju xorošen'kuju košečku "I saw such a cute kitty cat the other day"), while on the other they depend on the speaker/hearer relationship: a kindly or patronizing attitude causes adults to use many diminutives when talking to small children, a pattern which extends to other spheres as well (e.g. the speech of doctors to their patients); and in a seemingly strange reversal, a subordinate relationship of the speaker to the hearer may command the use of diminutives, especially with reference to items within the sphere of interest of a powerful or superior interlocutor, with an effect of ingratiating oneself to him (e.g. the use of Ručku požalujte "Hand (dim.), please", when a serf requests his lord's hand to kiss). The subordinate status of women, as well as their willingness to project an emotionally charged self-image, is respon-

SHIFTERS AND NON-VERBAL CATEGORIES OF RUSSIAN

365

sible for the fact that diminutives are more frequent in women's speech than in men's. To the same category belong "good" and "bad" nouns, like golubka "darling" or merzavec "bastard". But "good" nouns differ from "bad" ones in that they occur almost exclusively with reference to the addressee, whereas "bad" nouns have no such restriction. The relationship between the speaker and the hearer, when P s 2 ≠ Ρ n , is again relevant for this group of nouns: Ps1 would not refer to someone as merzavec "bastard" unless his relationship to P s 2 permitted such freedom of expression (a condition common to all "expletives"). 13 Personal pronouns Personal pronouns also offer certain choices that depend on the relationship between P s and P n , namely the choice of so-called polite forms versus informal forms. Perhaps the best known case of this sort is Japanese, which has up to a dozen commonly used first person pronoun forms (not counting literary, archaic, or dialectal ones): the choice between them depends on the social relationship between the speaker and the hearer, i.e. their age, sex, and difference in status; similar variants exist in Japanese for the other persons as well (see Harada 1976 for details). Familiar examples from European languages are mostly limited to variants in second person pro­ nouns. Thus a personal pronoun is not only a shifter by definition (as a code referring to the message, C/M), but a shifter in another sense as well, since it serves to reflect the social relationship between the speaker and the referent of the pronoun. 2. Shifters and reflexivization There is an obvious sense in which the phenomena just discussed, involving the role of P s and P n in the choice of referentially synonymous nouns and pronouns, involve "soft" (some would say "trivial") grammar, as opposed to "hard" grammar. But the role of P s in certain syntactic processes involving pronominalization leads us into grammar of the "hardest" sort. I will now demonstrate the relationship of shifters to Reflexivization in three different types of structures. 2.1 Reflexivization and pronouns in simple sentences It is often said that third person possessive pronouns in Russian should be reflexivized when they are coreferent with a third person subject. In the

366

OLGA T. YOKOYAMA

other two persons, Reflexivization is said to be "optional". Recommended examples are seen in (1): (1)

Ja posylaju moju/svoju stat'ju v izdatel'stvo. I-nom. send my self's article-c. to publisher "I'm sending my article to the publisher." On posylaet svoju stat'ju v izdetel'stvo. he-nom. sends self's article-c. to publisher "He is sending his article to the publisher".

It has been shown, however, that this is nothing but a crude prescriptive recommendation that bears little relation to the facts of Russian (see Yoko­ yama 1975, 1980, Yokoyama and Klenin 1976). It is true that pronouns are reflexivized more often than not when the subject is in the third person; but this is not always the case, nor is it true that the choice in the other two persons is truly "optional". Consider the following examples: (2)

a. Ja uvidel v zerkale *moe/svoe krasnoe lico. I-nom. saw in mirror my/self's red face-acc. "I saw my red face in the mirror." b. Ja ???moi/svoi objazatel'stva vypolnjaju. I-nom. my/self's obligations-c. fulfill "I fulfill my obligations."  Ja ??moe/svoe delo znaju. I-nom. my/self's job-c. know "I know my job." d. Ja ljublju moju/??svoju mat'. I-nom. love my/self's mother-acc. "I love my mother." e. Ja ljublju moju/???svoju mamu. I-nom. love my/self's mom-acc. "I love my mom." f. Ja ljublju moju/*svoju dočen'ku/dočurku/dočuročku. I-nom. love my/self's sweet-little-daughter-acc. "I love my sweet little girl (daughter)." g. Ja ljublju moju/*svoju Veru/Veročku/Verušu/Veririku. I-nom. love my/self's Vera/Vera (diminutives) "I love my dear little Vera."

The paradigm in (2) shows that the choice between the personal possessive

SHIFTERS AND NON-VERBAL CATEGORIES OF RUSSIAN

367

pronoun moj and the reflexive possessive pronoun svoj, with first person subject, depends on the relationship of the speaker to the hearer as appropri­ ate in a particular setting of the Es. Statements oriented towards the hearer (and particularly outsiders), like those in (2b) and (2c), preclude the occur­ rence of the personal moj, as do statements in which the speaker views himself from the outside, as in (2a). When the speaker is, on the other hand, speaking personally and is oriented towards his inner, intimate self, as in statements about loving one's family members (especially with affectionate diminutives indicating emotional attachment and an openness about express­ ing it, as in (2e) and (2f)), Reflexivization is counterindicated. Peškovskij, who was the first to point out that Reflexivization in Russian simplex sentences is not random, discussed the following striking example from Puškin's Boris Godunov (Peškovskij 1914:162): (3) Ja otravil sestru svoju, caricu, monaxinju smirennuju ... I-nom. poisoned sister-acc. self's czarina nun humble "I poisoned my sister the czarina, a humble nun..." The choice of the reflexive pronoun, says Peškovskij, indicates a distance between the speaker, the Czar Boris, as he himself perceives himself, and the Czar as he thinks the people perceive him. Boris is not really saying that he killed his sister, but only that "they" accuse him of that. Thus the speaker " I " (Ps1) is not identical to the subject of the sentence " I " (Pn), even though they refer to the same person. It is the doubling of these two selfs that characterizes the subject of sentence (3), thereby causing the Reflexivization. The same doubling occurs in (2a)-(2c), producing the same effect.2 Consider now cases with second person subject: (4)

a. Voz'mi svoj/?tvoj stakan. take self's/your glass "Take your glass." b. Voz'mite vaš/?svoj stakan. take your/self's glass "Take your glass."

These directives are addressed to a single addressee, and as directives they are addressee-oriented (cf. Jakobson 1960). The only difference between (4a) and (4b) is in the relationship between the speaker and the addressee: in (4a), the speaker uses the familiar second person singular form of address, while in (4b), the address is the formal second person plural. This difference goes along with the difference in personal versus reflexive pronouns. When the speaker identifies with the addressee, Reflexivization applies; when the speaker maintains distance, on the other hand, a personal possessive pro-

368

OLGA T. YOKOYAMA

noun is used. The distinction is subtle, and the judgment is complicated by the fact that the choice between the formal and familiar form does not necessarily reflect the real interpersonal relationship between the interlocu­ tors: the two may be very close yet they may retain the polite plural form formed years ago, and vice versa. The Reflexivization of P n , therefore, is determined by the speaker Ps1's attitude towards P n ( = P s 2 ), modified by the relationship between Ps1 and P s 2 . When the subject is in the third person, normativists are partially correct, as mentioned above, in that the possessive pronoun is much more often reflexive than personal. But personal pronouns do occur with third person subject, and grammar must explain what determines their use, as well as the relative frequency of reflexive over personal possessive pronouns. Consider the following examples: (5) a. Ja pristaval  nemu raz-drugoj prošlogo goda, I-nom. stuck to him once-twice last year kogda možno bylo s nim razgovarivat', ...so vsemi when possible was with him talk with all ètimi voprosami i zametil, čto on, nesmotrja na these questions and noticed that he-nom., despite vsju svoju svetskost' i dvadcatiletnee rasstojanie, all self's worldliness and 20-year distance kak-to črezvyčajno krivilsja. somehow especially twitched "I bothered him a couple of times last year, when it was possible to talk with him, ... , with all these questions, and I noticed that, in spite of all his social tact and a lapse of twenty years, he was somehow annoyed a great deal." b. ... mnogoe iz častnyx obstojatel'stv žizni Versilova ot a lot-nom. from private circumstances of-life of-V. menja uskol'znulo, do togo on hyl vsegda so mnoju from me slipped to that he-nom. was always with me gord, vysokomeren, zamknut i nebrežen, nesmotrja ... proud pompous closed and careless despite na poražajuščee  by smirenie ego peredo mnoju. striking as if meekness his before me "... a lot of the particular circumstances of Versilov's life eluded me, for he had always been so proud, haughty, closed and careless with me, despite his striking, well, meekness towards me." (Dostoevskij) (5a) and (5b) show how the role of P s determines the occurrence or non-

SHIFTERS AND NON-VERBAL CATEGORIES OF RUSSIAN

369

occurrence of reflexive possessives. Both sentences are uttered by a twenty year old youth about his natural father, with whom the youth has a very complex relationship. In (5a), the youth states that some of his father's reactions were inappropriate despite his father's good breeding and the fact that the subject matter that caused these reactions was twenty years old. (5b) describes the father's attitude towards the youth, as the youth perceives it: he finds his father's haughty, impersonal treatment of him contradictory to the father's otherwise meek attitude towards him, which, the youth says, is so striking to him (note also the youth's hesitation in finding the appropri­ ate word for "meekness", indicated by  by "as if"). In (5b), in other words, the speaker is preoccupied with his own introspective perception of his father, which renders it an extremely personal utterance, whereas in (5a) he is trying to understand his father's point of view. It is precisely this difference that is responsible for the Reflexivization in (5a) and the lack of it in (5b). Contexts like (5b), in which the speaker is more preoccupied with him-/ herself than with his/her sentential subject, can occur only when the speaker is prominent, and involved in a highly personal narration, i.e. generally in first person narratives, or in interpersonal communication. In many cases, however, as well as with the omniscient narrator who does not actually appear in the text as "I", there is no place for such a display of "egocentrism". Normally, the speaker effaces himself when speaking of third persons, assuming the P n 's point of view, and especially the subject's. This explains why reflexive pronouns occur with third person subjects much more often than personal pronouns, and why they are recommended by prescriptive grammars. 3

2.2 Reflexivization in embedded clauses (see Klenin 1974, Yokoyama 1980) Verbs like prikazat' "order" or pozvolit' "allow" take infinitival comp­ lements. Assuming that those infinitive clauses originate from embedded sentences whose subjects are deleted under identity with a coreferent noun in the higher sentence, their structure can be approximated as in (6): (6) Brat pozvolil Nikite i s[Nikit i -ubra- Nikiti - vešč-]s→ brother-nom. allowed N.-dat. [N. take N.'s things] Brat pozvolil Nikite i θi ubrat' svoii vešči. "My/His brother allowed Nikita to remove his things." Since Russian Reflexivization generally operates within the boundaries of one sentence (i.e. it is "clause-bounded") the occurrence of the reflexive possessive svoi "self's (acc)" is expected. As Peškovskij (1914: 162) was the

370

OLGA T. YOKOYAMA

first to mention, however, the main clause subject can reflexivize the possess­ ive phrase in the embedded clause, resulting in an ambiguous reference. Svoi in (6), then, can refer either to the "brother" or to "Nikita". Note that replacing the reflexive pronoun with a personal possessive does not resolve the ambiguity, since ego "his" would in this case be just as ambiguous. While the ambiguity of the pronouns in these Russian constructions has often been noted, it is less well known that although the reference of svoi in (6) is ambiguous, the dominant reading is that in which svoi is coreferent with the subject "brother". In (7), on the contrary, the dominant reading is that in which svoj is coreferent with the embedded (and deleted) object, i.e. Nikita:4 (7) Brat skazal Nikite ubrat' svoj stakan. brother-nom. told N.-dat. remove self's glass-acc. "My/His brother told N. to remove his glass." Both this tendency for different dominant interpretations as well as the occurrence/non-occurrence of Reflexivization in this type of embedding can be adequately explained only with reference to shifters and duplex structures. It can be shown that the difference between the main readings of (6) and (7) depends on the fact that their verbs belong to two different classes: verbs that primarily denote verbal activity (e.g. skazat' "tell", poprosit' "ask", prikazat' "order" etc.) behave as in (7), while verbs that do not imply verbal activity (e.g. predložit' "offer", prinudit' "force" or mešat' "hinder") behave as in (6). Thus, if we consider that the group denoting verbal activity represents a message referring to a message (M/M, according to Jakobson 1957), the number of perspectives the sentence can have doubles, since the embedded message must be uttered from the perspective of the person who utters it, while the main clause message is uttered from the perspective of the speaker of the whole sentence. In other words, the embedded clause of (7) can be treated in the deep structure like reported speech: Uberi svoj/tvoj stakan "Remove your glass". 5 Now within this embedded sentence, the choice between the reflexive or personal pronoun can then be determined along the lines discussed in 2.1. Given the relationship between the "brother" and Nikita, implied by the referential expressions, the likely choice is svoj. Reflexivization is thus likely to have taken place already in the embedded clause, and hence the chances of the reflexive referring to the embedded subject are great. Only in the less likely case of the brother having said Uberi tvoj stakan "Remove your glass" would the possessive pronoun refer­ ring to the reporter appear as ego. Significant in this respect is the fact that when a personal possessive is justified in the embedded sentence (as in the case of one gentleman addressing another one), vašu in the infinitive clause (e.g. Nikolaj Sergeevič imeet čest' prosit' vas ubrat' vašu koljasku s dor ogi

SHIFTERS AND NON-VERBAL CATEGORIES OF RUSSIAN

371

"N. S. has the honor of asking you to remove your carriage from the road") is clearly preferable to svoju, which is then interpreted as coreferent with N. S. On the other hand, if the brother has asked Nikita to get the brother's own glass, the embedded clause must be Uberi moj stakan "Remove my glass", which on the surface becomes ego, referring to the brother. This is the primary reading of ego. Only when the speaker has good reasons to adopt the brother's point of view so strongly as to override the M/M feature of the sentence can the possessive pronoun be reflexivized by the main subject. In this way, an appeal to the categories M/M and P n /P s explains both the pronominal references and the pattern of variation in reference judgments. Example (6), on the other hand, does not involve M/M structures: hence the relationship between P n s and P n does not affect the choice. As with third person narrative in general, the speaker in this case effaces himself and adopts the point of view of his subject. Thus in (6), the chances are much greater that the speaker takes the "brother's" standpoint and that the reflexive pronoun is coreferent with brat, while the personal possessive is coreferent with Nikita. But this is not all. The following examples indicate that the reference to Ps in the characterizations given for Reflexivization in embedded structures of this sort must be even more specific: (8)

Ja ne pozvolju tebe trogat' I-nom. not allow you-dat. touch "I won't let you touch my kids!"

moix detej! my children

(9)

Pozvol'te mne zametit', čto ja nikogda ne allow me-dat. to-note that I-nom. never not pozvoljal postoronnim vmešivat'sja v svoi dela. allowed outsiders-dat. interfere in self's affairs "Permit me to mention that I never allowed outsiders to interfere with my affairs."

The lack of Reflexivization in (8) contrasts with its presence in (9). The difference between the two sentences lies in the relationship between the speaker and the hearer: in (8), despite the combative tone of the sentence, the speaker is open with the addressee, while the statement in (9) is pompous and detached. The occurrence of the reflexive pronoun in (9) — which does not have a verb that denotes a speech act — is therefore perfectly consistent with the cases of Reflexivization in simple sentences: the reflexive depends not only on Ps1's position vis-à-vis P n , but also on the relationship between Ps1 and P s 2 .

372

OLGA T. YOKOYAMA

2.3 Reflexivization in reduced relative clauses Another case in which a reflexive pronoun is controlled by an antecedent in the higher clause involves reduced relative clauses (see Klenin 1974, Yokoyama 1979). The structures in question are as follows: (10)

Ona xotela emu skazat', cto  nee syn i she-nom. wanted him tell that at her son-nom. and pust' iščet sebe druguju, no ne mogia osilit' let seek self-dat. another-acc. but not could manage ètix trudnyx sejčas dlja sebja slov. these hard now for self words-gen. "She wanted to tell him that she had a son and that he should look for somebody else, but she could not manage these words, (so) difficult for her now." (P. Proskurin; cited in Klenin 1974: 84-5)

The reflexive pronoun sebja "self (gen.)" is coreferent with ona "she" in the main clause. This is not true in other cases, such as (11): (11)

Tolja preziral večno dovol'nogo soboj i T.-nom. despised always satisfied self-instr. and samouverennogo Davida self-confident D.-acc. "Tolja despised David, (who was) constantly satisfied with him­ self and self-confident." (lit. "Tolja despised the constantly selfsatisfied and self-confident David.")

In (11), soboj "self (instr.)" refers to the underlying subject of the reduced relative clause David and not to the main subject Tolja. Reflexivization is thus clause-bounded in (11), but not in (10). This violation of the clause-boundedness of Reflexivization may at first glance appear to be lexically motivated, since the class of adjectives it occurs with seems to be limited, including words like trudnyj "difficult", neožidannyj "unexpected", and čuždyj "alien". Closer examination, however, shows that Reflexivization into reduced relative clauses occurs, in general, when the statement encapsulated in the clause (e.g. èti slova sejčas trudny "these words are now difficult" in (10)) is made from the point of view of the higher subject, who also appears as x in the prepositional phrase dlja x "for x" in the embedded relative clause. Thus in (10), the judgment concerning the difficulty of uttering those words belonged to "her", and not to the speaker. The crucial difference, then, involves who is responsible for the judgment contained in the embedded statement. Moreover, those cases in which the main sentence has a first person subject (i.e. when P n = P s 1 ) also indicate

SHIFTERS AND NON-VERBAL CATEGORIES OF RUSSIAN

373

that the relationship between Ps1 and P s 2 is no less relevant for this type of Reflexivization than it is for other types of Reflexivization. Consider the following examples: (12)

Ja proanaliziroval połoženie i ešče raz I-nom. analyzed situation-c. and more once ubedilsja ν glavnom dlja sebja. convinced in main for self "I analyzed the situation and once again became convinced about the most important thing for me." (Proskurin, cited in Klenin 1974: 81)

(13)

S trudom perenošu ja poterju ètogo dorogogo with difficulty bear I-nom. loss-c. of-this dear mne čeloveka. to-me man "With difficulty do I bear the loss of this man, (who was) so dear to me."

In these sentences, P n = Ps1, and the judgment belongs to the speaker. In (12), however, there are actually two "selfs" of the speaker involved: the one analyzing and the one analyzed. When this sort of doubling occurs, as was the case in (2a)-(2c) and (3), Reflexivization takes place; when the speaker speaks from the point of view of a personal, undivided "I", as in (2d)-(2g), Reflexivization is blocked.6

3. Shifters and Russian word order I now examine the interaction of the category of P s with an area generally considered to belong to discourse grammar, i.e. Russian word order. As Jakobson himself pointed out (1936), the choice of the subject in the two sentences "Latvia borders on Estonia" and "Estonia borders on Latvia" is associated with different leading roles for the words Latvia and Estonia in these sentences. More recently, Kuno and Kaburaki have proposed (1977) that with "subject-centered" verbs like "marry" or "meet", the choice of subject is determined by Empathy, i.e. by the extent to which the speaker identifies with the P n . Empathy is thus, by definition, a shifter. Let me now demonstrate how Empathy and other shifters control a whole range of phenomena involving Russian word order.

374

OLGA T. YOKOYAMA

3.1 Discourse situation Any discourse situation D must be described in terms of four knowledge sets: the set of the speaker's knowledge, the set of the hearer's knowledge, and two sets of the items of the current concern of the speaker and the hearer. Let me illustrate this relationship by means of a Venn diagram, consisting of four circles A (the speaker),  (the hearer), Ca (A's matter of current concern), and C b (B's matter of current concern); the abbreviated names for the intersections in Figure 1 are given to the right of the intersec­ tion formulae in boldface. Any two normal participants of a Es share some of the content in their respective knowledge sets. They always share certain basic concepts per­ taining to the human condition, they may also share a linguistic code, as well as specific cultural and historical concepts and information; and finally, they may share an awareness of their immediate surroundings. These shared elements are represented by the intersection of the two sets A and B, i.e. by AB. The matters of current concern are subsets of both A and B, and for A and  to be engaged in informational discourse, C a and C b must intersect, which can happen only over the territory of the intersection AB. At the moment before A's utterance, the knowledge that A has pertaining to the matter of current concern, but of which  is not aware yet, appears outside Cb, i.e. in Cal and Ca2. The portion which has already been shared through communication is in the area where all four sets overlap, i.e. in Cab. There are two important facts about discourse that must be added for a proper understanding of this model. First, that this idealized model is never avail­ able per se to the actual participants of the Es. In real discourse, each speaker can at best make his/her own subjective evaluation of the content and the

Figure 1.

SHIFTERS AND NON-VERBAL CATEGORIES OF RUSSIAN

375

interrelationship of the four sets of knowledge, and on the basis of this evaluation proceed to an utterance. Because of this necessarily subjective picture of the discourse situation assumed by, say, A in preparation for an utterance, any intersection with  can only be an approximation of the real discourse situation. The less A and  know each other, the less the chances that the assumptions made by A (or B) about the discourse situation are correct, and vice versa. If A's statement reveals that A's assumptions were wrong, and if  is a cooperative interlocutor,  will correct A's mistake so that A and  may proceed to have an effective discourse. For example, if A assumes that  knows who "Mr. Smith" is and asks  where Mr. Smith's desk is,  may in fact be unable to identify "Mr. Smith" because there are two Smiths in the office; in this case  will adjust the prerequisites for effective communi­ cation by correcting A's wrong assumption and asking A if s/he meant "John Smith" or "Bob Smith". The fact that neither speaker can ever make a completely correct assessment of the true discourse situation (and that each is therefore left to create his/her own subjective picture of the situation) is one of the main reasons why the category of P s is so crucial to interpersonal communication. The other important fact, which is often taken for granted (and therefore not sufficiently appreciated), is that the relationship between A, B, C a and C b , even as it is assessed by one and the same speaker, constantly keeps changing as the communication proceeds. Every utterance alters the picture, because every utterance corrects the participants' assumptions about the discourse situation, or because the flow of information results in an increase of shared information in Cab and in AB. Thus every discourse exchange involves as many as four discrete situations, i.e. the situations as viewed by each interlocutor both before and after a given utterance. How is all of this relevant to Russian word order? In fact, the assumptions Ps1 makes about the discourse situation determine the order of elements in Russian in a very specific way, which I will now demonstrate. In order to simplify this already quite complex matter, I will exclude one powerful parameter, namely intonation, limiting the discussion to declarative state­ ments belonging to a single intonation type. 7

3.2 Discourse-initial utterances Consider a pre-discourse situation, when the speaker has made no specific assumptions about what his/her interlocutor has in mind. At this point the speaker assumes that Cab contains no more than the mutual awareness that precedes verbal communication, or "set for contact" (Jakobson 1960). It is

376

OLGA T. YOKOYAMA

usually signaled by eye contact and body language, and corresponds to something like "here am I, and here are you, and I have something to tell you". In this sort of situation, the speaker organizes his utterance according to the following ordering relations. Items with referential properties (i.e. nouns, pronouns and adverbial shifters) appear as follows: first (1) those the speaker assumes to be in the intersection Cab, then (2) those s/he assumes to be in the intersection Ca2, and finally (3) those s/he assumes to be in the intersection Cal. (Not all of these items need occur in every sentence.) The verb is placed after the items which the speaker assumes to be located in Cab. When there is more than one item from the same area, the order is determined by the degree of the speaker's identification with that item: the items the speaker identifies with more (i.e. has greater Empathy with) come earlier. Empathy may also override the order given above, provided the speaker's relationship with the hearer allows for such "egocentric" behavior. As a whole, the proposition the speaker is about to utter is placed in Cal. For example, utterance (14) is based on the speaker's assumption of the discourse situation given in Figure 2. (14)

Menja/ Moju doč' skorbü kakoj-to negodjaj. me-acc. my daughter-c. insulted some bastard-nom "Some bastard insulted me/my daughter."

The hearer B, e.g a policeman on the street, is assumed to be aware of the existence of the speaker ("me"); thus the speaker assumes this information to be in the intersection Cab. The knowledge of "some bastard" belongs in

Figure 2.

SHIFTERS AND NON-VERBAL CATEGORIES OF RUSSIAN

377

Cal, which is the lowest rung in the ordering sequence; the verb follows the item located in Cab. If the object of the insult is "my daughter", the referential knowledge of whom is presumably found in Cal along with the "bastard", the speaker's empathy plays the decisive role in choosing between the "daughter" and the "bastard" for the initial position, and the verb again precedes the last item. The order in (15) and (16) is determined by the same regularities: (15)

(16)

Vas sprašivaet kakoj-to student. you-acc. asks some student-nom. "Some student is asking for you." Naša mašina svalilas' v . our car-nom. fell in ditch "Our car fell into a ditch."

The items that precede the verbs in (15) and (16) belong, respectively, in Cab (vas "you" (acc.)) and Cal (naša mašina "our car"). The items that follow the verbs all belong in Cal. In (16), the choice between two items from Cal is again determined on the basis of the speaker's empathy: it is clearly easier for the speaker to empathize with "our" car than with some "ditch", hence the order "car" → "ditch". Note that this formulation does not need to refer either to the case or to the grammatical status of individual items, which is clearly irrelevant in all of the examples mentioned so far: in (14) and (15), the direct object precedes the subject, whereas in (16) the subject precedes the location. Consider, however, the following example, assuming that it is uttered by a policeman in a report to his superior officer: (17)

Krest'jane sela Vixrovki zarezali pomeščika Sedyx. peasants-nom. of-village V. stabbed landlord S.-acc. "The peasants of the village of Vixrovka slashed the throat of the landlord Sedyx."

In (17), the speaker (the policeman) assumes that both the peasants and the landlord are in Cal, and the objectivity of his official report to his superior precludes any expression of empathy based on his personal stand (even if there were any). In the absence of other criteria for choosing between these two claimants, the speaker chooses the agent/subject for the first position. This ordering appears to be grammatically motivated, although it is also possible to claim that the grammatical subject is nothing more than a grammaticaUzation of the universal tendency to empathize with agents over patients, and animates over inanimates. In fact, it would be no more surpris­ ing to find a grammaticaUzation of shifters (e.g. Empathy) in syntax than it is to find their grammaticaUzation in the morphology of the Russian verb, such as Jakobson demonstrated in 1957.

378

OLGA T. YOKOYAMA

Thus three major factors are responsible for Russian word order when the speaker has no reasons to assume that Cab contains anything except the set for contact: (1) the speaker's assumptions about the whereabouts of relevant items in the discourse situation, and the order their presumed locations dictate (i.e. Cab → Ca2 → Cal); (2) the speaker's Empathy; and (3) grammaticalized universal Empathy (or grammatical relations), ceteris paribus. 3.3 Non-discourse-initial utterances When the speaker has reason to assume that the interlocutor has something specific on his mind besides the set for contact, the speaker's assessment of the discourse situation is somewhat different. The main difference is that the area Cab may now contain some propositions, as well as their terms, besides the items associated with the set for contact. Let us examine a few cases of this kind. Suppose A's utterance indicates that A assumes someone has broken the window, that A thinks  knows who did it, and that A wants to find out from  who it was. A's question (18) is based on his/her assessment of the discourse situation given in Figure 3. 8 (18)

Kto razbil okno? who-nom. broke window-acc. "Who broke the window?"

Suppose further that  finds A's assessment of the discourse situation to be correct, and proceeds to satisfy A's request for information (i.e.  is going to tell A that "x" is "the kids next door"). B's assessment of the discourse situation at the onset of B's answer is as in Figure 4.

Figure 3.

SHIFTERS AND NON-VERBAL CATEGORIES OF RUSSIAN

379

Figure 4.

This is the picture of B's assessment of the discourse situation. Note that A's question has caused relocation of most knowledge items: Cab now contains the knowledge of the incompletely specified proposition, the refer­ ential knowledge of the window in question, and of the fact that A wants to know who "x" is. The specification of "x", which was not part of B's set of the current concern before A's question, has now entered C b , as has the knowledge who the kids next door are. On the basis of this assessment,  orders the elements in his/her response as follows: first comes the item  knows is in Cab (i.e. the "window"), then follows the verb, and lastly the item  knows is unknown to A in its capacity as the predicate of "x" = k, i.e. the item needed by A in order to complete the incompletely specified proposition "someone broke the window" that A possesses (namely "the kids next door"): (19)

Okno razbili sosedskie rebjata. window-c. broke neighbor kids-nom. "The kids next door broke the window."

When A cannot be assumed to have a particular predicate in mind that requires some of its terms (i.e. Ps) to be specified by  (as in Figure 4), but has only some terms in mind, then it is only these Ps's that  assumes to be in Cab. The order then is again "Cab → verb → Cb2 → CM", where Cbl and Cb2 correspond to Cal and Ca2 prior to A's utterance. Consider the exchange in (20), in which B's answer is based on B's assessment of the discourse situation in Figure 5:

380

OLGA T. YOKOYAMA

Figure 5.

(20)

B: Čto slučilos' s Andreem? what-nom. happened with A. "What happened to Andrej?" A: Andrej uexal ν Perm'. A.-nom. left to P. "Andrej went to Perm."

The speaker  need not be explicitly informed by A as to what items currently are or are not on A's mind. The model allows the speaker  to make subjective assumptions, on the basis of which  orders those items which have not appeared in the question along with those which have. Thus in (21)  places ja "T", an item  assumes to be in Cab by virtue of being implied by the set for contact, after Andrej, an item  knows to be in Cab: (21)

A: Čto slučilos' s Andreem? what-nom. happened with A. "What happened to Andrej?', : Andreja ja posiala v Perm'. A.-acc. I-nom. sent to P. "I sent Andrej to Perm."

It is not always the case, however, that the item the speaker knows to be on the hearer's mind must precede the other items in Cab. In fact, even in (21),  could have chosen to reverse the order of Andrej and ja in the answer. This would have been a more informal, more personal answer than the one in (21). The following two examples make the contrast and its mechanism especially clear:

SHIFTERS AND NON-VERBAL CATEGORIES OF RUSSIAN

(22)

A: Otkuda  tebja èto kolečko? whence at you this ring-nom. "Where did you get this ring (dim.)?" B: Mne ego podadla tëtja Ljuba. to-me it-acc. gave aunt L.-nom. "Aunt Ljuba gave it to me."

(23)

A: Otkuda  vas pojavilsja ètot pistolet? whence at you appeared this pistol-nom. "Where did you get this gun?" B: Ego mne podbrosili v mašinu. it-acc. to-me threw in car "Someone put it in my car."

381

In (22) — an intimate conversation between two girl friends —  assumes that there are two items in Cab, namely "I ( = B)" and the "ring", based on the explicit indication A gave in her question.  now orders these items according to her own Empathy, which is obviously higher for her own persona than for a ring. In (23) — a court-room interrogation —  also assumes, justifiably, that there are two items in Cab: "I ( = B)" and the "gun". In this case, however, the official relationship between Ps1 (the witness) and P s 2 (the prosecutor) does not allow Ps1 to give prominence to her Empathy, and so she places the "gun" first.9 The ordering of those items which are placed in the same area is thus determined, once again, by the degree of the speaker's identification with the P n , i.e. by Empathy, when the P s 1 /P s 2 relationship allows the speaker to display it. Russian word order is thus demonstrably conditioned by shifters involving the way the speaker views the discourse situation, and the extent to which the speaker identifies with the participants of the narrated event. Only when no distinctions mark the participants of the narrated event in either of these two categories does grammatical order come into play (as in (17)), but even this may be viewed as a grammaticalization of universal principles of Empathy.

4. Shifters and coordinating conjunctions Jakobson (1957: 499) pointed out that Russian has "evidential" particles indicating that the speaker's report is based on someone else's report. He also mentioned that Russian uses syntactic means to present direct and indirect speech. We may add that in presenting indirect speech, the speaker has morphological means to indicate the degree of his commitment to the truth value of the reported statement: while the complementizer čto reports

382

OLGA T. YOKOYAMA

someone else's speech without any commitment to its truth, other comp­ lementizers like budto and jakoby reflect varying degrees of skepticism towards the reported information. I will now examine the relevance of the speaker's position towards the relationship between two narrated events conjoined by the common conjunc­ tions a, i and no. Two of the three most common coordinating conjunctions, namely a and i, serve a dual role in Russian: they are used (1) to organize the discourse, and (2) to conjoin propositions. Leaving aside the very interesting role of a and i in organizing the discourse (for which see Yokoyama 1981), let me address myself only to the second function, i.e. that of conjoining prop­ ositions. It is in this function that a and i join no in serving to distinguish among various ways the speaker conjoins two different propositions. Con­ sider the following examples: (24)

Ja pokušalsja na ubijstvo, i menja arestovali i I-nom. attempted murder and I-acc. arrested and otdali pod sud. subjected under court "I attempted a murder, and they arrested me and tried me."

(25)

Ja pokušalsja togda na ubijstvo, no menja ne I-nom. attempted then murder but I-acc. not arestovali i ne otdali pod sud. Menja sčitali arrested and not gave under court I-acc. considered sumasšedšim. insane "I attempted a murder but they didn't arrest me or try me. They considered me insane."

(26)

Siranno. Ja pokušalsja na ubijstvo, a menja ne Strange I-nom. attempted murder yet me-acc. not sud. Značit, sčitajut arestovyvajut, ne otdajut pod arrest not give under court means consider menja sumasšedšim. me-acc. insane "It's strange. I attempted a murder but they don't arrest me or try me. That means they must think I'm crazy." (Cexov)

The function of i in (24) is clear: the speaker sees the cause-effect relationship between the two propositions, namely that he attempted to kill someone and that he was arrested and tried. This is precisely how i is used. (25) and (26) are more complex. Both a and no indicate that there is a contradiction —

SHIFTERS AND NON-VERBAL CATEGORIES OF RUSSIAN

383

that the fact that he attempted to kill someone and that they neither arrested nor tried him is contrary to the normal course events take when one attempts to murder someone. What then is the difference between the two conjunc­ tions, if both signal a contradiction? It can be shown that what distinguishes these two types of "but"-coordination is the speaker's orientation; in other words, this form of coordination in Russian is a shifter. When the speaker addresses him-/herself to the proposition s/he assumes the interlocutor will form, drawing on the information the interlocutor receives in the first of the conjoined sentences, the speaker uses no. In (25), for example, the speaker assumes that his interlocutor is likely to deduce from the first sentence "I attempted to kill someone" that the speaker was brought to justice. Address­ ing himself to this assumed deduction, the speaker proceeds to tell the hearer that this natural deduction is, however, wrong. When, on the other hand, the speaker reacts to his own deduction (made on the basis of the facts narrated by him in the first sentence), and when he finds his own deductions contrary to what is happening, he then uses a. Thus in (26), the speaker himself expected to be arrested, and yet it did not happen. He finds this incomprehensible and voices his feeling of incongruence, of betrayed expec­ tations, through the use of a (note the actual context of (26), which clearly indicates the introspective overtones of Uncle Vanja's brooding over the unexpected turn events took after he tried to kill his brother-in-law). A thus signals the conjunction of two propositions, "I attempted a mur­ der" and "They didn't arrest or try me", both of which were in Ps1's knowledge set (i.e. in Cal) prior to the utterance. The difference between (25) and (26) is that the conjoined sentence in (25) is "linked" to the first one by a proposition "For attempted murder one is arrested and tried", which the speaker assumes to be in the hearer's knowledge set, i.e. in Ca2, while in (26) the linking proposition is placed by the speaker in his own knowledge set, i.e. in Cal. This accounts for the fact that a in its function of conjoining two propositions belongs to interpersonal, subjective, egocen­ tric communication (and to internal monologues), while no is used in objec­ tive, addressee-oriented, formal discourse, e.g. scientific prose and expository writing (see Yokoyama 1986a). All of the tell-tale linguistic signs pointing to the relevance of the speaker's personal involvement and of his/her pre­ occupation with his/her own reasoning appear with a, and serve to dis­ tinguish it from the other conjunction of contradiction no (e.g. first person narration, the present tense, the involvement of the speaker in the narrated event (i.e. Ps1 = P n ), and the co-occurrence of certain "little untranslatable words" that indicate discourse relations (e.g. ved' and že)10). In other words, a occurs when the speaker not only perceives two conjoined propositions as contradictory, but when s/he also feels free to assert his/her personal feelings about this incongruence to the interlocutor.

384

OLGA T. YOKOYAMA

The relationship between the two propositions conjoined by the "simple" conjunctions a, i and no in Russian is therefore a shifter in two ways: (1) it is characterized by the speaker's assessment of the discourse situation with respect to the implications of the content of the first of the conjoined propositions, and (2) it is affected by the relationship between the inter­ locutors. I have tried to show in this account something of the fluidity of grammar, and in particular how persistently it accommodates the P s — the content of the speaker's knowledge set, his/her attitudes, and his/her assumptions about the interlocutor and about the discourse — into the surface structure of language. The relations of natural human discourse differ from truthbased semantics principally because speaker-hearer interaction is necessarily based on the speaker's subjective assessment of the real discourse situation. This sort of discourse is impossible to describe and explain adequately without reference to P s . The examples presented here, from a wide variety of linguistic components of one language, are sufficient to illustrate the power of this particular shifter category. To judge from the interest linguists have shown in pragmatics and discourse grammar over the course of the past decade, linguistics is moving from the narrower domain of formal grammar toward a broader and more inclusive conception of human linguis­ tic communication in its multifarious complexity, for which a fuller under­ standing of P s will be indispensable.

Notes * 1. 2.

3. 4.

5.

This research was supported in part by NEH Fellowship FA 2193982. See Uspenskij (1970) for an extensive discussion of the problem of point of view in verbal and other arts. See Parret's (this volume) discussion of "shifting" in and out of one's self; the situations when the " I " which is the P n and the " I " which is the P s do not coincide, resulting in Reflexivization of the first person possessive pronoun in Russian, can be characterized as "shifting out" in Parret's terms. Other Slavic languages seem to observe the same distinction; note e.g. Czech Mám rád mou/*svou maminku "I love my mommy". Klenin (1974) presents the results of field work on Reflexivization in Russian (undertaken primarily at the Academy of Sciences of the USSR), which shows regular patterns of disagreement among her group of informants; analysis along the lines discussed in this section suggests that preferences among informants' judgments reflects an increased likelihood that certain groups of informants have chosen (or imagined) particular contexts as opposed to others. Compare Kuno's suggestion (1972) that if deep structures for some embedded clauses resemble direct quotations, certain otherwise puzzling facts about English Pronominalization and Reflexivization can be explained.

SHIFTERS AND NON-VERBAL CATEGORIES OF RUSSIAN

385

6.

Cf. van Schooneveld's proposal (1982) that this kind of behavior shown by Reflexivization is marked by a general semantic feature he terms the "extension feature", which signifies a direct perceptional link between the referent and the speech situation. 7. The intonation type referred to here is discussed in detail in Yokoyama 1986b, where it is termed "Type I intonation". This is the type of Russian intonation used in utterances delivered in a controlled, planned fashion; the intonational contour of an utterance of this intonation type does not encode the speaker's assumptions about the discourse situation. 8. The utterance illustrated in Figure 3 is assumed to be made with Type I intonation (cf. note 7). 9. It can be shown that sentence-final position in questions uttered with Type I intonation is reserved for the item the questioning party expects the interlocutor to adopt as the Topic of the answer. In placing ego "it" first, then, the speaker  not only gives up her own Empathy, but also goes along with the Topic indicated by her interrogator. For further details see Yokoyama 1986b. 10. "Little words" like ved' or že, referred to as "particles" and/or "conjunction words" in traditional Russian grammar, are notoriously difficult to explain to non-native speakers. This is because such words refer to the speaker's assessment of the discourse situation, a concept that has been ill-understood and underutilized in linguistic description until now.

References Harada, S. I. 1976. "Honorifics". Syntax and Semantics, 5: Japanese Generative Grammar, ed. by M. Shibatani, 499-561. New York: Academic Press. Jakobson, Roman. 1936 [1971]. "Beitrag zur allgemeinen Kasuslehre: Gesamtbedeutungen der russischen Kasus". Selected Writings II: Word and Language, 23-71. The Hague-Paris: Mouton. . 1957 [1971]. "Shifters, Verbal Categories, and the Russian Verb". Selected Writings II: Word and Language, 130-147. The Hague-Paris: Mouton. . 1960 [1981]. "Linguistics and Poetics". Selected Writings III: Poetry of Grammar and Grammar of Poetry, 18-51. The Hague-Paris-New York: Mouton. Klenin, Emily. 1974. Russian Reflexive Pronouns and the Semantic Roles of Noun Phrases in Sentences. Ph. D. dissertation, Princeton University. Kuno, Susumu. 1972. "Functional Sentence Perspective: A case study from Japanese and English". LI 3. 269-320. Kuno, Susumu and Etsuko Kaburaki. 1977. "Empathy and Syntax". LI 8. 627-72. Peškovskij, A.M. 1914 [1956]. Russkij sintaksis v naučnom osveščenii. 7th ed. Moskva: Učpedgiz. Uspenskij, . . 1970. Poètika kompozicii. Moskva: Iskusstvo. van Schooneveld, C. H. 1982. "The Extension Feature in Russian". IJSLP 26. 445-457. Yokoyama, Olga T. 1975. "Personal or Reflexive? a Functional Analysis". Harvard Studies in Syntax and Semantics I, ed. by Susumu Kuno, 75-112. Cambridge, MA: Harvard University Linguistics Department. . 1979. "Reflexivization into Adjectival Reduced Relative Clauses in Russian". FS 2. 366-375. . 1980. "Studies in Russian Functional Syntax". Harvard Studies in Syntax and Semantics

386

OLGA T. YOKOYAMA

III, ed. by Susumu Kuno, 451-774. Cambridge, MA: Harvard University Linguistics Department. . 1981. "On Sentence Coordination in Russian: A Functional Approach" Papers from the Seventeenth Regional Meeting, 431-438. Chicago: Chicago Linguistic Society. . 1986a. "Lexical Frequency and Its Implications: the Case of Contemporary Edited Russian". SEEJ 30. 147-66. . 1986b. Discourse and Word Order. Amsterdam: John Benjamins.

English Speech Act Verbs: A Historical Perspective Elizabeth Closs Traugott Stanford University 1. Introduction Speech acts and speech act verbs, especially performative verbs such as assert, claim, command, promise have recently been much discussed in work on semantics and pragmatics not only in linguistics and philosophy (cf. Austin 1962; Searle 1979; Grice 1975; Verschueren 1980), but also in anthro­ pology (cf. Rosaldo 1982), and literature (cf. Pratt 1977).1 It is surprising, then, that very little work has been done on them from a historical perspec­ tive (cf., however, Schlieben-Lange 1983 on the development of speech act verb repertoires in French). Such a historical perspective would contribute precisely to the kind of interdisciplinary program Roman Jakobson called for, since it raises simultaneously a multiplicity of issues concerning linguistic typology and universais, language function, and cognitive as well as sociohistorical explanations for language change. The purpose of this paper is to pave the way for larger-scale diachronic cross-linguistic studies by investiga­ ting aspects of semantic change in the history of English speech act verbs, most particularly performative verbs Two main topics will be discussed. First, the regularities in the lexical sources for speech act verbs (hereafter SAVs), and the evidence these provide for how speakers of the language conceptualize SAVs, in other words, for what their linguistic belief systems about SAVs may be. Although the focus will be on performative verbs, the conclusions reached apply equally well to SAVs in general. Second, evidence will be given for the directionality of shift from one subcategory of performative verbs to another, e.g. of nonrepresentative SAVs like insist that X do, allow ( = "permit") to representative SAVs like insist that X is the case, allow that { = "acknowledge", "say"). Results from the study of regularities both in lexical sources and in shifts from one subcate­ gory to another show that the development of SAVs conforms to a principle of semantic change so far identified primarily in connection with modal auxiliary verbs: meanings shift from non-epistemic to epistemic.2

388

ELIZABETH CLOSS TRAUGOTT

Before we go on to discuss the lexical origins of SAVs, a word is in order about the subcategories of SAVs to which I will be referring. There has been considerable debate about whether a taxonomy of SAVs is possible, and if so, what the categories are (cf. Austin 1962, Fraser 1975, Searle 1979, Hancher 1979, Verschueren 1980, among others). The historical data suggest that although there do not appear to be clear discrete boundaries between subcategories, there is nevertheless evidence from constraints on semantic change that at least the following four subcategories are useful in an account of the history of English SAVs: representatives, directives, commissives, and expressives. a. Representatives: these are verbs like admit, affirm, argue, assert, claim, insist, postulate, theorize (usually that X is the case). The main force of these verbs, when used performatively, is to assert the speaker's belief in the truth (or falsehood) of what is said. In Searle's terms, they fit words to the world. However, it is not enough to suggest, as Searle does, that all that matters is the speaker's belief in or knowledge of the proposition; representatives also, as Bach and Harnish (1979) have shown, are directives, however much disguised, to the hearer to agree with or believe what the speaker has said. b. Directives: these are verbs like command, ask, pray, whereby one attempts to get people to do things, and to match the world to one's words. In performing directives, the speaker does not make an assertion about the world. Rather, the point of saying I hereby command you to X is to bring about the state of affairs in the proposition.  Commissives: these are verbs like promise, guarantee, swear, and vow, whereby one obligates oneself to do something. As in the case of directives, the focus is on making the world match one's words. d. And fourthly, there are expressives, which express, sometimes perfunc­ torily, certain feelings that the speaker has or thinks the hearer expects toward the state of affairs expressed in the proposition; they are verbs like apologize, compliment, congratulate, thank, and welcome.

2. Lexical domains from which SAVs have been derived The question of etymology has always fascinated people. It has traditionally given them insights into the culture of an earlier time, and therefore has often been regarded largely as a socio-historical enterprise. But when one discovers patterns of semantic change through doing etymological study, one can also begin to get insight into less tangible matters, including people's beliefs about conceptual domains. Thus, whatever the physicist may tell us about the nature of time and entropy, a look at the etymologies of temporal terms such as before, after, until and so forth will indicate that people's intuitions about

SPEECH ACT VERBS

389

time are based on perceptions of space. In other words, the etymologies suggest that experience of time in terms of motion and space, as provided by vision and the other senses, is highly relevant to the development of mental representations of linguistic temporal relations. Etymologies therefore can give us some insight into the relation between linguistic and other kinds of cognitive processes (cf. Traugott 1985). If one finds enough regularity of change across different semantic domains, one may even be able to postulate certain patterns of processing that may be universal. This, of course, is a difficult task that must await the historical study of many languages. It is therefore not possible here to say much about possible universais of concep­ tualization, although some initial hypotheses can be developed about how speech acting must be conceived by English-speaking people, past and present. These are discussed in 2.6., after the data have been presented. In what follows I will show that the lexical domains from which SAVs in English have been derived are quite restricted, and furthermore that the same domains act as sources whether the verbs in question are native or borrowed (largely from Latin via French, and to a far smaller extent from Greek and Scandinavian). I will start with verbs that were borrowed (not always with a speech act meaning or with the ones they now have), since these are the ones with which we are most familiar. The data are based on the two hundred and seventy-five performative verbs that Fraser (1975) lists, all of which have been checked for their etymologies;3 however, only a few examples will be given here. These performative verbs are only a subset of all possible SAVs; for example, they exclude explain, lament and other such verbs that are metalinguistic terms for linguistic events, but cannot normally have illocutionary force even in the first person present tense. The restriction to performative verbs is a matter of convenience for controlling the size of the data; it does not appear to skew the results in so far as types of lexical source are concerned.4 Early on in the study it emerged that, apart from a small number of idiosyncratic cases, the verbs in Fraser's list fell into four etymological categories; those derived from verbs of uttering such as "shout", "call"; those derived from terms for mental and psychological states such as "know", "wish", "praise"; those derived from terms for vision such as "shine", "look at", and those derived from spatial expressions such as "put under". As we shall see, those derived from terms for uttering and mental/ psychological states are essentially synecdochic since uttering and the exist­ ence of certain mental or psychological attitudes are prerequisites for SAVs. Those derived from vision and space are metaphorical in that they involve the use of terms from one domain for that of another, given certain quite rigorous schemata constraining the possible types of space etc. that can be used (cf. Lakoff and Johnson 1980, Traugott 1985).

390

ELIZABETH CLOSS TRAUGOTT

Before I exemplify these categories, a few of the verbs with idiosyncratic origins may be mentioned. All are of sociohistorical, rather than cognitive interest. Furthermore, they are metonymic rather than synecdochic or meta­ phorical. One example is provided by the root of verbs such as attest, protest. Te-, the first past of the root, comes from IE trei "three" (testimony was given by a witness who was the "third person" at a trial). -st, the second part of the root, comes from IE sta- "to stand". The whole root, as evidenced by Latin testavi, means "to witness, or to stand as the third person". These are facts pertaining to IE law, and tell us little about the mental schemata for SAVs — the SAV meaning of attest is not a direct extension of "to stand as a third person", but a metonymic meaning derived from the typical linguistic actions of a person standing in this capacity. Among other idiosyn­ cratic SAVs of similar metonymie origin are bless and toast. Bless is related to Gmc. blotham "blood" — to "bless" was in other words to "hallow with blood". Toast, as an SAV, derives from "say someone is like toasted bread" — the idea was that the name of the lady celebrated would flavor the drink like spiced toast (cf. Benveniste 1971 for other "delocutive" verbs derived from constructions of the type "say X"). 2.1 Sources in uttering/vocalization. A common domain of derivation for all types of SAVs is the concrete domain of "speaking", i.e. performing a vocal act that has acoustic corre­ lates. The main types of vocalization that occur can be ranked on a scale of vocal intensity ranging from "utter", "speak", to "call", and "shout". Some examples are: confess, profess, prophesy < IE bha- "speak" allege < Lat. particle + legare "charge" < "speak" (ultimately < the spatial meaning "collect") announce, denounce, renounce < Lat. particle + nuntiare "say/shout" advocate, invoke < Lat. ad + vocare "toward + call" call < IE gal- "call, shout" (borrowed into ME from ONorse kalla) acclaim, claim, proclaim, classify < IE kel- "shout" deplore, implore < Lat. plorare "wail" 2.2 Sources in mentali psychological states It is useful to subdivide these into three groups: cognitive, evaluative, and affective. The first involves knowledge and belief, the second evaluation according to the parameters true-false or good-bad, the third desires, pleas­ ures and sorrows.

SPEECH ACT VERBS

391

a. Cognitive states. Most SAV categories include at least some members that derive from mental state verbs having to do with cognition (having in mind, knowing, believing), assessment of truth, and being aware. Examples include: admonish, comment, mention, remind < IE men- "think" (cf. mind) acknowledge, note, recognize, diagnose < IE gno- "know" certify < Lat. certus + facere "certain + make" credit, grant < Lat. creditum "something entrusted" < credere "believe", ultimately < IE kerd-+ dhe "in heart+put" (i.e. partially locative) b. Evaluations. A small set of borrowed SAVs derive from evaluative terms, most notably: accuse, excuse < Lat. causa "cause/reason/lawsuit" adjudge < Lat. ad+judicare "to + judge" appraise, praise < Lat. pretiare "value", ultimately < pretium "price, that which is opposite/equal" evaluate < Lat. valere "be of value, be strong"  Affective states. Many speech act types are expressed by verbs originating in affective states, mainly wishes, feelings, positive and negative attitudes: volunteer < Lat. velle "wish, want" consent < Lat. sentire "feel" agree, disagree < Lat. ad+gratus "to + pleasing" plead < Lat. placitum "legal action, agreement" < piacere "to please" allow < OF al(l)ouer "permit, approve", blend of Lat. allocare "assign to a place" (i.e. a spatial) and allaudare "praise" commiserate < Lat. miserari "pity" 2.3 Sources in visual perception Most SAV categories include some members which have been derived from terms related to the field of vision, specifically terms for seeing, showing, and for color intensity (brightness, etc.). Some examples are: advise < Lat. advisare "watch, observe" < videre "see" regard < OF regarder "look at" ( < "guard, watch") speculate, specify < IE spek- "observe" (for speculate, see Lat. speculari "spy out, watch") theorize < Gk. theoros "spectator" < thea "a viewing" declare < Lat. declarare "make clear", ultimately < IE kel- (i.e. from a term for vocalization) argue < IE arg- "shine, white" (cf. argent "silver")

392

ELIZABETH CLOSS TRAUGOTT

2.4 Sources in spatial terms By far the greatest number (about 75%) of borrowed SAVs are derived from spatial terms. Spatial relations are often considered to be of five kinds: location (relation, static or dynamic, to a point of reference, cf. at-on-in, from-to), plane or dimension (cf. front-back, up-down), shape (round, square, flat), extension ( cf. large-small), and distance (cf. far-close) (cf. Lyons 1977; also Friedrich 1971 for the first three). By far the most predominant spatial relation from which SAVs are derived is location. Included under "location" are verbs for states of affairs which can only be understood in terms of spatial relations, cf. give, take, seize, carry (transfer X from Y to Z), hold, have (X be at Y) (for the relationship between locative, possessive, existential, see Clark 1978), and such verbs as lead, carry (W moves with X from Y to Z; cf. Gruber 1976, Jackendoff 1983, Talmy 1983). Many verbs involve both a spatial root and a spatial prefix; for purposes of this presentation, focus is on the roots: a.

Location. i. Dynamic: A. General motion (as attested in root meanings; more specific directions may be given by the affixes): cite, solicit < Lat. ciere "put in motion, move" concur < Lat. currere "run" interpret < Lat. inter + pres "between + goer" deduce < IE deuk- "pull, lead" offer < Lat. ferre "bring, carry" suggest < Lat. gerere "carry" B. Focus on motion toward: assent < IE sent- "go, head for" admit, commit, dismiss, permit, promise, submit < Lat. mittere "send" appeal < IE pel- "thrust, strike, drive" conjecture, object, reject < Lat. jacere "throw" command, demand, commend, countermand, recommend < Lat. manu dare "into hand give" guess < IE ghend- "seize" (borrowed in ME from Scand. gissa; cognate with get) C. Focus on motion away from: concede < Lat. concedere "go away, withdraw" relinquish < Lat. linquere "leave" refuse < Lat. fundere "pour out" exempt < Lat. eximere "take out"

SPEECH ACT VERBS

ii.

393

Static, dynamism irrelevant: insist, restore, state < IE sta- "stand" {state < N for status, rank) assess < Lat. sedere "sit (beside a judge/be assistant to a judge)": prohibit < Lat. pro+ habere "in front + have" maintain < Lat. manu tenere "in hand hold"

b. Plane. The horizontal plane is the normal plane for verbs of motion, and can be assumed to be involved in a) above; the verbs in the small sample that follows mark the special coordinates front-back, up-down, and in-out, almost always in prefixes, not roots: advance < Lat. ab + ante "from + before/in front" present < Lat. prae + esse "in front + be" propose < Lat. pro+ habere/ponere "in front + have/put" hypothesize < Gk. hypo 4- tithenai "under + place" submit < Lat. sub + mittere "under + send" suggest < Lat. sub + gerere "under + carry" grant < IE kerd- + dhe- "in heart + do/put" c. Shape. figure < Lat. figura "form, shape" < IE deigh- "knead clay" inform < Lat. informare "give shape to" tender < Lat. tendere "stretch" retort < Lat. retorquere "bend, twist back" abrogate, direct, rank, reckon, rule < IE reg- "move in straight line" One may ask whether there was any awareness among those who bor­ rowed the Latinate vocabulary that they were using spatial terms. This we can presumably never really know, but it is interesting to note that the OED very frequently gives partial evidence that there was some awareness of the more concrete meanings; for example, insist in the 1590s meant "stand on, persevere in a person's steps, dwell at length on". In some unusual cases, such as affirm, the spatial meanings ("make firm") are apparently actually derived as etymological reinterpretations from the mental and speech act meanings.

2.5 Sources of OE SAVs Since the SAVs discussed so far are borrowed from French or Latin, their origins might be considered to be simply a function of these languages rather than of more general processes of change. It is therefore important to investigate whether there is any evidence that the native speech act verbs of OE derived from domains different from those just described. As it turns out, the OE SAVs which are not idiosyncratic (cf. discussion of bless above),

394

ELIZABETH CLOSS TRAUGOTT

derive from the same four domains as the borrowed terms. However, spatial terms are significantly less productive in the roots (although they are an important factor in affixes). The most productive domains in OE are vocalization and mental verbs: a. Vocalization: cweðan "speak, say" < IE gwet- "speak" secgan "say" < IE sekw- "utter" ( > NE say) sprecan "speak" < IE spreg-/speg- "speak" ( > NE speak) spellan "say" < IE spel- "say out loud, recite" ( > NE spell, no longer an SAV) swerian "swear", andswerian "answer" < IE swer- "speak" ( > NE swear, answer) gretan "greet" < IE gher- "call out" ( > NE greet) bannan "summon, proclaim" < bha- "speak" (acc. to Buck 18.21 probably related to Skt. bha- "shine", i.e. a verb of vision) ( > NE ban, (marriage) bans) þrafian "rebuke" < IE (s)trep- "cry loudly" b. Mentali psychological states: þancian "thank" < IE tong- "think" beodan "order, announce, proclaim", abeodan "command, announce", bebeodan "command", forbeodan "forbid", bodian "announce, pro­ claim" < IE bheudh- "be aware", ( > NE bid, forbid) mynegian "remind, admonish, exhort" < IE men- "think" seðan "affirm" < OE soð "truth" acyþan "announce, confirm" < OE cuþ "known" < IE gno- "know" acsian "ask" < IE ais- "wish, desire"  Examples of OE SAVs derived from visual perception are geswutelian "declare, explain" < IE sweidh- "shimmer" and wissian "declare" < IE weidh- "see". d. As in the case of borrowed words, the majority of spatial terms from which OE SAVs are derived involve location, plane, or shape: i. Location: belecgan "charge, accuse" < IE legh- "lie, lay" tellan "tell" < IE del- "count" (cf. also Gmc til "goal") ( > NE tell) raedan "advise, explain" < IE ar- "fit together" ( > NE read, not an SAV) hatan "bid, command; promise, vow; call, name", behatan "prom­ ise", andettan "confess, acknowledge, praise" ( < and "against" + hatan) < IE kwei- "put in motion") wrecan "recite" < IE wreg- "push, drive" ( > NE wreak (ven­ geance), not an SAV)

SPEECH ACT VERBS

395

ii.

Plane: warnian "warn" < IE wer- "cover" ( > NE warn) iii. Shape: biddan "ask, order", abiddan "ask for" < IE bedh- "bend" ( > NE bid) (a)recc(e)an "tell out, express" < IE reg- "move in a straight line"

2.6 Etymological sources as evidence for linguistic belief systems about the nature of SAVs What do the four etymological sources for SAVs suggest to us about ways in which speakers conceive of this lexical domain, and about processes by which the meaning changes occur? For one, the sources in vocalization suggest that speakers were at some level aware that speaking in a special way had a particular function, in other words, that one performs certain kinds of actions in speaking; hence the selection of a verb of speaking for a specialized type of speech act. Since every speech act other than those specifically constrained to non-vocal modes such as writing has acoustic correlates, the vocalization domain can be considered to be a necessary prerequisite for most speech acting. Historically, the selection of a term for speaking to represent a speech act type must be considered synecdochic to the speech act event: it indexes a characteristic of speech acting, but does not specify any particular category of speech act. Speakers were also presumably aware that certain states of mind such as believing that X or wishing that X were conditions for saying things with performative force; hence the selection of verbs expressing mental states for some speech acts. We may note that all analyses of SAVs, whatever their terminology or exact details, point out that mental and psychological atti­ tudes figure crucially among what Searle calls "the conditions of satisfac­ tion" for SAVs. For example, asserting, claiming, declaring, and stating have among their conditions of satisfaction the fact that the speaker believes that p (where p = the proposition asserted, claimed, etc.), and that the speaker has the intention that the addressee believe that p. Imploring, requesting and so forth have among their conditions of satisfaction the fact that the speaker desires the addressee to do whatever is implored, requested, etc., and the intention that the addressee do whatever is implored, etc. partly because of the speaker's desire (cf. Searle 1979, 1983, and Bach and Harnish 1979 on various conditions of satisfaction). The semantic change undergone in these cases is therefore synecdochic in origin. What about verbs derived from terms of vision and space? Here the process of meaning change is different. Neither seeing in general nor the

396

ELIZABETH CLOSS TRAUGOTT

particular visual perception of spatial relations is a prerequisite for speech acting. What we have here are metaphors. Verbs of seeing (and also hearing) in general tend to be extended to the field of intelligence. Thus, I see="I understand" (cf. Sweetser 1984). That verbs related to visual perception are sources of SAVs suggests that there is a very wide-spread belief, however unconscious, that seeing is believing and understanding; as we have seen, believing, understanding and so forth are among the conditions of satisfaction for SAVs. This suggests that possibly verbs derived from terms of vision are not only metaphorical but also synecdochic. What of the spatials? They indicate that from very early on IndoEuropeans at least conceived of language as an object, and of propositions as things that could be transferred, put together, even stood upon. That the spatialization of propositions was and is very important conceptually is supported by evidence from the historical origins of connectives such as but < OE (on)but-an "on the outside". Like the shift from terms for vision to speech act terms, the shift from spatial to speech act terms is metaphorical. The types of spaces used are essentially the same as those used in other semantic domains, for example case relations (Givon 1975), temporals (Trau­ gott 1978), existentials (Clark 1978), clausal connectives (Traugott 1986), and mental verbs like maintain, deduce, understand (Traugott and Dasher, 1987). For the most part they assume a front-back axis through the body (a line along which things can be pulled or pushed) or an up-down axis. They are essentially spaces that are relative and not absolute, like corners and angles. As such they are the spatial relations that Talmy (1978, 1983) characterizes as cognitively fundamental. Talmy's evidence includes the availability only of the relative spatials for grammaticalization. The evidence from semantic change and shift into other domains supports his charac­ terization.

3. Evidence for shifts from one subcategory of SAV to another We have seen that there have been SAVs from OE on, and that whether they are native or borrowed, they ultimately come from the same set of lexical sources, a finding which lends strong support to the hypothesis that there are regularities in semantic change, and that they reflect ways in which language users conceptualize the domain in question. As I was investigating the etymologies of SAVs, I began to be interested in the question of whether they would provide any evidence for the existence of subcategories of SAVs. From the perspective of language change, evidence for subcategories can come from a number of factors. Weak evidence is

SPEECH ACT VERBS

397

provided by such facts as the failure of certain lexical domains to give rise to certain putative subcategories; an example, noted in 2.6, is the failure of terms for vision to give rise to directives. However, evidence of this kind is tenuous, since it could result from accidental gaps. Stronger evidence is provided by regular, unidirectional, shifts from one putative subcategory to another. This kind of historical argument resembles similar arguments in other domains; for example, the fact that aspect gives rise to tense in many languages, and that the reverse is rarely attested, suggests that the two categories, aspect and tense, should be distinguished on some level of analysis. I hypothesized that if shifts from one speech act category to another were to be predictable, there should be evidence of a shift from non-representa­ tives to representatives, not vice versa. The basis for this hypothesis is that SAVs have much in common with modal verbs and adverbs, and, as will be discussed below, epistemic modals (which concern possibility) develop later than root modals (which concern necessity, obligation, etc.). It is well known that modal auxiliaries in modern English have both root and epistemic readings. That is, they express on the one hand permission and obligation, i.e. degrees of necessity, and on the other degrees of possibil­ ity. Standard examples of the distinction include the following ambiguous sentences: She may leave. a. She has permission (whether from speaker or someone else) to leave [root] b. It is possible that she will leave (speaker assesses the likelihood of the proposition She will leave) [epistemic] She must be married. a. She is required (whether by speaker, someone else, social/moral obligation, etc.) to be married [root] b. I conclude she is married (speaker assesses the likelihood of the proposition) [epistemic] On the one hand, representative SAVs share much in common with epistemic modals: denials function like negatives in reflecting the "imposs­ ible" end of the continuum, assertions and claims function like simple indicatives in reflecting the "certain" end of the continuum, while the intermediate ranges of might, may, can etc. share with guess, hypothesize, suggest, etc. various intermediate degrees of probability. On the other hand, directive and commissive SAVs share much in common with the root modals; this is evidenced not only covertly by the semantics, but also overtly by mood morphology (most notably by the use of an imperative form instead of the full lexical phrase I command you to...; cf. also the use of subjunctive

398

ELIZABETH CLOSS TRAUGOTT

and optative inflections instead of SAVs in languages that have such mark­ ers). (For this and other similarities, see Sweetser 1984). It would seem, then, that if subcategories of SAVs truly exist, they should develop in ways similar to the modals. If modern English modal auxiliaries are ambiguous between root and epistemic readings, this was by no means always the case. Some compelling data for the development root → epistemic modal is provided by Shepherd (1982). She shows that epistemic meanings devel­ oped later than did root meanings in the history of English modal auxiliar­ ies. Thus shall (derived from "owe") indicated obligation from OE on, but did not indicate possibility until  1500. Similarly, will (derived from "desire, wish"), indicated willingness from OE times, intention from  1200, but not certainty until  1500; may (derived from "having physical power"), developed the meaning of permission  1100, and possibility  1200; and must (derived from the past tense of motan "be permitted") developed the meaning of obligation  900, and probability  1300. The only exception is can (derived from "having mental power"), which devel­ oped the meaning "be able"  1100, of possibility  1200, and of per­ mission  1500 (note, however, that even in this case a nonepistemic "ability" sense preceded the epistemic one). Shepherd's evidence is sup­ ported by Goossens (1982), who notes that in OE obligation and necessity could be expressed both lexically and grammatically (by modals and by adverbs). However, epistemic meanings could be expressed only lexically, as in phrases such as wen is paet "possibility (literally 'hope') is that/ possibly", or indirectly in "epistemically colored predictions". He goes on to claim that "There is nothing that grammaticalizes the epistemic function in Old English" (Goossens 1982: 84). The hypothesis that there should be a match between root → epistemic modal and nonrepresentative → representative SAY is well borne out by the history of the SAVs under investigation. Not all the changes are evidenced in English itself, as some SAVs were apparently borrowed from French with both nonrepresentative and representative meanings. But when we go back to French and to Latin, we find that the representative meaning developed later (with a few apparent, but not real, exceptions to be discussed below). I will start with some verbs which devel­ oped representative meaning in English:

SPEECH ACT VERBS

399

accuse: a. blame or censure (with object from EME): 1340 Hampole Pr. Cons. 5423 Many accusers þar sal be þan, To accuse þam byfor þat domesman; 1535 Coverdale 2 Macc. x. 21 Accusynge those per­ sonnes, that they had sold the brethren for money. b. disclose, reveal: 1477 Earl Rivers Dictes (Caxton) 16 Without he wold accuse them that were consenting to make werre ayenst the King. The OED suggests that the use of accuse in the sense "reveal" is in direct imitation of French. It certainly is very unusual in English. In French, however, it gave rise to the meaning "acknowledge", and is used in the standard formula for acknowledging a letter: J'accuse reception de votre lettre. admit: a. grant, permit: 1423 James I, King's Quair: If mercy sal admitten of thy service; 1682/3 Penn. Arch. L55 Desiring thee to admití, that the people may have the Nomination, b. accept as true, concede: 1532 More Confut. Tindale Wrks 1557, 668/1 That the scripture is not true, but because ye churche saith so and admyt it; 1538 Starkey, England 207 Hyt ys to be admyttyd.. .that then a nother is to be chosen.5 agree: a. assent, consent to: 1374 Chaucer Troylus III. 131. Ye wolde ... agreen that I may ben he b. concede that X; agree to be cited in OED from 1606; agree that X cited as "mod." allege: a. 1300, bring charge against (legal) b. 1400, claim that, express opinion allow: a. originally from a blend of allaudare "praise" and adlocare "place, bestow", allow meant praise, approve of from the early 14th century, permit from the 1550s b. accept as true; with NP; later = "say" 1580 Baret Alv. A 297 To Alowe, to make good or allowable, to declare to be true. 6 grant: a. 1225, allow, permit b. 1340, acknowledge that insist: a. demand (insist that X do): 1676 Guillatier's Voy. Athens (tr.) We insisted that...the Frigot should hang out either the French or the English Colours

400

ELIZABETH CLOSS TRAUGOTT

b.

maintain (insist that X is the case): 1768 Sterne, Sent. Journey, Case of Delicacy (1858) 252 I begged a thousand pardons, but insisted it was no more than an ejaculation. postulate'. a. 1533: ask ecclesiatical authority to admit a nominee (a "postulant"); demand: 1593 Hist K. Leir (1605) Dj A prince perhaps might postulate my love b. claim existence of truth: 1646 postulated inferences; 1855 H. Spen­ cer, Princ. Psychol. (1872) Lii.i. 146 That which we must postulate as the substance of Mind suggest'. a. suggest that X do: 1526 Pilg. Perf. (W de W 1531) 124b whan...he suggesteth or moueth to a man or woman to do suche thinges b. 1600 J. Pory tr. Leo's Africa 415 They suggested vnto him, that Gonsaluo was a Magician, who.. Some SAVs in English do not show significant evidence of the shift because they are used in both non-representative and representative mean­ ings from the time of borrowing on. However, a look at the prior histories of some of these verbs in Latin and French reveals the same story. For example, assert, based on the verb asserere "put, join", appears to have first been used metonymically in a speech act sense to lay claim on a person or set them free (OED "to put one's hand on the head of a slave either to set him free or to claim him"). From this derived the representative meaning claim that (e.g. claim that X is one's slave). Another example is provided by respond. In Latin spondere meant to take on a solemn, ritualistic engage­ ment, chiefly marriage (cf. spouse); it appears to have first been used with reference to a speech act in the context of reports of oracular responses to prior commitments, i.e. an oracle could "make a commitment in response" to a promise; in this sense, respondere is a commissive. From this derives the general meaning of "answer", via such meanings as "answer back" i.e. in (self-defense) (i.e. exercising authority), and ultimately, simply "answer that" (i.e. a representative). Additional evidence for the directionality of change from nonrepresentative to representative meanings comes from the differential behavior of cognate verbs in English, French, and Spanish. Certain verbs remain nonrepresentative in one language, but acquire additional representative meanings in another; we have seen that accuse did not really become a representative in English, although it did in French and Spanish. On the other hand, agréer did not become a representative in French while it did in English. There are some apparent problems with the claim that if an SAV is going

SPEECH ACT VERBS

401

to change category, the direction will be from nonrepresentative to represen­ tative and not vice versa. However, these problems turn out to be soluble. One apparent difficulty is that some cognate verbs in Latin might be thought to acquire representative meanings before the specific verb in question (usually a verb derived in post-Latin times) itself acquires a representative meaning. However, if we keep the cognates properly apart, the difficulty disappears. Consider suggest. Latin suggerere meant "put under, procure" in Classical Latin (Ernout-Meillet); by  230 a.d. it meant "announce, declare, make a report" (Souter), i.e. it had representative meaning. The English directive verb meaning "advise" would then appear to have devel­ oped later than the Latin representative meaning "report". However, we should note that suggest (with -st) is actually not derived directly from suggerere, but from the past participle of this verb via the derivative Latin noun suggestio, which meant "successive addition" (non-speech event), "advice" and "report" according to Souter. Both the directive and represen­ tative meanings therefore appear to have been available from the time that suggestio was used to name speech events, which means that suggest as a directive is not later than suggest as a representative. A different problem is raised by the fact that a number of verbs of "saying" seem to start out as representatives and to acquire non-representa­ tive force later. Among them is deny, which is etymologically Lat. de + nego "from + say not" but also means "disown, denounce" (Ernout-Meillet p.774; cf. also Souter "repudiate"). Another example may be OE sprecan "speak, say", which is also occasionally found in constructions of the type sprecan that X do Y (Goossens 1985). However, these verbs of "saying" are not true counterexamples to the hypothesis that non-representative meanings will precede representative ones. This is because the original sense of what we translate as "say" in modern English appears in Indo-European and earlier Roman and Germanic times to have been commissive rather than representa­ tive. In other words, to "say" as an SAV was to "say solemnly/vow that X", usually with a religious or legal force that we no longer associate with "say". If this is the case, then Lat. denegare and OE sprecan were really originally non-representative.7 Particularly valuable evidence for the sacral/legal aspects of "saying" in IE are provided by Benveniste (1973) in connection with two roots: deik"show, say" and weghw- "vow". He argues that the primary meaning of deik- was to show not by pointing but by saying. This "showing" is stipulative, that is, it pronounces and establishes a state of affairs, much as does saying in a game of tag The tree in the back garden is/shall be "home". For example, in Roman times only the judge could dicere ius "say the law" (dicere is derived from deik- "show, pronounce solemnly"). In doing so he announced the rules of law:

402

ELIZABETH CLOSS TRAUGOTT

to give justice is not an intellectual operation which requires meditation or discussion. Formulas have been transmitted which are appropriate to given cases, and the role of the judge is to possess and apply them (1973: 387) that is, not to simply assert them, but to stipulate and exercise authority through them. Similarly, Benveniste shows that the various descendants of weghw- (vow in English) can be explained only if "proclaim in a loud voice, utter sol­ emnly" is understood as a special sacral act, or commissive. Benveniste notes, for example, that the Iranian derivative aog- "say" is appropriately used only by the gods and Zarathustra, and means "announce with authority that binds". Latin vovere, another descendant, means to consecrate oneself in anticipation of support from the gods. In Homeric Greek eukhomai, another descendant, means both "pray" and "boast". Benveniste argues that the prayer was in fact a vow: an assertion/consecration which binds the gods; hence the meaning "pray". A "boast" in Homeric Greek was, he says, a declaration before the gods, in which the speaker's own person is figuratively offered, e.g "I consecrate myself to the gods, as being the son of so and so, or, the bravest of all" (ibid. 495). In sum, the IE verbs of "saying", when used as speech acts, had commissive properties that were later bleached into assertions. The interpretation of "saying" in IE, and of denegare and sprecan in particular, raises the larger question: to what extent are individual SAVs translatable across times and across languages? To what extent do the differences reflect differences in cultural attitude? For example, SchliebenLange (1976) has reminded us that medieval vs. modern views of "men­ tioning" are very different. In the middle ages, to mention was to invoke auctoritas, that is, the speaker's psychological belief that the source was right and furthermore that using it would attest to the speaker's ability. To us, however, a mention is often used at least to implicate that one does not fully believe that p, or to be ironic. An example of differences across languages is provided by Verschueren. He points out that the lexical origins of representatives in English reveal no overt recognition of the hearer; but such attention to the hearer is overtly lexicalized in Dutch and German. Thus we find Dutch mededelen and German mitteilen "state" (literally "with + share"), but no similar meanings in English (Verschueren 1980: 48). A full-scale historical account of the development of SAVs in any language or of one SAV across languages must always be very sensitive to the potential for considerable differences between apparently cognate or apparently translatable terms.

SPEECH ACT VERBS

403

4. Conclusion I have shown that the history of SAVs in English reveals that there is a very small set of lexical domains from which they are derived, and furthermore that there is evidence for at least two types of performative verb, depending on "direction of fit" (Searle 1979): (1) those that match words to world: representatives that focus on truth, belief, and certainty, and expressives that focus on psychological attitudes; and (2) those that match world to words (cf. command, promise). It remains to be seen whether the semantic changes outlined here are characteristic of European languages only, or whether they are universal. Preliminary investigations suggest that they may be universal. For example, it appears that the same four sources for SAVs in English are also sources for SAVs in Japanese, cf. (a) from vocalization: NJ eitan suru "exclaim, admire" < MChinese eitan "singing in a drawn-out style" + NJ suru "do"; (b) from mental states: ENJ syooti suru "consent, acknowledge" < "know, be aware of" < Chinese syoo "receive" + ti "knowledge"; (c) from visual perception: NJ mitomeru "recognize" < MJ miru "see" + tomeru "stop (something)" (the path from verb of vision to SAV is via a mental meaning); (d) from spatial relations: ukeireru "assent, agree (to), grant (a request)" < ukeru "receive" + ireru "send in, insert". Furthermore, the shift from nonrepresentative to representative is attested, cf. oosu MJ and NJ "say" (respectful) < OJ "command, order". It has been suggested that the origin of verbs of understanding (and, one may conclude, of SAVs) in verbs of seeing depends on literacy and objectifying of language in visual space (cf. Ong 1982), and that the origin of SAVs in spatial terms is similarly dependent on literacy (David Olson, p.c.). Although literacy may increase repertoires of intellectual and metalinguistic terms based on vision and space, it cannot be a prerequisite since "see" and "know" are related as far as we can tell from earliest Indo-European on, and also in contemporary "traditional" languages, i.e. languages that have not had liter­ acy until very recently. For example, Viberg (1983) shows that "see" and "know" are related in some of the languages of Australia and Papua New Guinea.8 Furthermore, Heeschen provides evidence that some spatial terms underlie metalinguistic terms in Eipo, a language of West New Guinea:9 The semantic fields that seem to be most effective in creating metalinguistic terms are those of carving and cutting, and of laying bare the edible part of a fruit. Longer bits of information, conversation, and tales are conceived as acts of enumerating names, or tying together or accumulating words. (Heeschen 1983: 172) One is reminded of the Latinate origins of such SAVs in English as rescind < Lat. scindere "cut", and assert < Lat. asserere "join oneself to" < IE

404

ELIZABETH CLOSS TRAUGOTT

ser- "line up". The belief that SAVs involve words and propositions that are objects that can be laid bare, cut, lined up, and otherwise manipulated may well prove to be a universal of language. Notes 1. This paper was inspired by lengthy conversations with David Olson. I am especially grateful to him for many hours of thought-provoking discussion, to Judith Hochberg for her help in gathering data and for her comments on an earlier draft of this paper, and to Richard Dasher for the data on Japanese. Thanks are also due to Claude Boisson, Louis Goossens, Suzanne Kemmer and Mary Pratt for their helpful remarks. I am of course responsible for all errors of fact or interpretation. Research for this paper was conducted while I was a Guggenheim Fellow and Fellow at the Center for Advanced Study in the Behavioral Sciences (supported in part by NSF Grant BNS 76-22943); I am grateful for the support in this and other endeavors. 2. As late as 1973, Lehmann could say in his revised edition of Historical Linguistics that little was known about general principles of lexical semantic change, except in small domains like kinship terms. The history of English SAVs shows that we have gone a considerable way since then. 3. This part of the study would have been impossible without the help of Judith Hochberg. The sources used are various, but chiefly The American Heritage Dictionary of the English Language (1981), Walde-Pokorny (1932), Buck (1915, 1949), Bosworth-Toller (1898), the MED and the OED. Citation-forms for IE are taken from the American Heritage Dic­ tionary. 4. There is, however, some quantitative skewing, in so far as there are more nonperformative than performative verbs that derive from types of utterance like crying, weeping, etc. 5. The MED includes meaning (b), but none of the examples are convincing. 6. MED cites a legal use "recognize as binding, admit validity of" from 1338 on; this may mean that shift needs to be sought in French. 7. Goossens (p.c.) points out that, indeed, while sprecan that X do Y is attested in his data, sprecan that X is the case is not. 8. I owe this reference to William Croft. 9. I owe this reference to Suzanne Romaine.

References Austin, J. L. 1962. How to Do Things with Words. London: Oxford at the Clarendon Press. Bach, Kent and Robert M. Harnish. 1979. Linguistic Communication and Speech Acts. Cam­ bridge, Mass.: M.I.T. Press. Benveniste, Emile. 1973. Indo-European Language and Society, tr. by Elizabeth Palmer. Coral Gables, Florida: University of Miami Press. . 1971. "Delocutive Verbs". Problems in General Linguistics, tr. by Mary Elizabeth Meek. Coral Gables, Florida: University of Miami Press. Buck, Carl Darling. 1915. "Words of Speaking and Saying in the Indo-European Languages". American Journal of Philology 36, 1-18, 125-54. Clark, Eve V. 1978. "Locative, Possessive, and Existential". Universais of Human Language

SPEECH ACT VERBS

405

IV, ed. by Joseph H. Greenberg, Charles A. Ferguson, and Edith Moravcsik. Stanford: Stanford University Press. Fraser, Bruce. 1975. "Hedged Performatives". Syntax and Semantics III: Speech Acts, ed. by Peter Cole and Jerry L. Morgan. New York: Academic Press. Friedrich, Paul. 1971. The Tarascan Suffixes of Locative Space: Meaning and morphotactics. (Indiana University Publications, Language Science Monographs 9.) Bloomington: Indiana University Press. Givón, Talmy. 1975. "Serial Verbs and Syntactic Change: Niger-Congo". Word Order and Word Order Change, ed. by Charles N. Li. Austin: University of Texas Press. Goossens, Louis. 1982. "On the Development of the Modals and of the Epistemic Function". Papers from the Fifth International Conference on Historical Linguistics, ed. by Anders Ahlqvist. Amsterdam: John Benjamins. . 1985. "Framing and the Linguistic Action Scene in Old and Present-day English: OE CWEÞAN, SECGAN, SP(R)ECAN and Present-day English SPEAK, TALK, SAY and TELL Compared".

Papers from the Sixth International Conference on Historical Linguistics, ed. by Jacek Fisiak. Amsterdam: John Benjamins. Grice, H. Paul. 1975. "The Logic of Conversation". Syntax and Semantics III: Speech Acts, ed. by Peter Cole and Jerry L. Morgan. New York: Academic Press. Gruber, Jeffrey S. 1976. Lexical Structures in Syntax and Semantics. Amsterdam: NorthHolland. Hancher, Michael. 1979. "The Classification of Cooperative Illocutionary Acts". Language in Society 8. 1-14. Heeschen, Volker. 1983. "The Metalinguistic Vocabulary of a Speech Community in the Highlands of Irian Jaya (West New Guinea)". Deutsche Forschungsgemeinschaft, Publi­ cation 15: Man, Culture, and Environment in the Central Highlands of Irian Jaya. Jackendoff, Ray. 1983. Semantics and Cognition. Cambridge, Mass.: M.I.T. Press. Lakoff, George and Mark Johnson. 1980. Metaphors we Live by. Chicago: University of Chicago Press. Lehmann, Winfred P. 1973. Historical Linguistics: An introduction. New York: Holt, Reinhart and Winston, 2nd ed. Lyons, John. 1977. Semantics. Cambridge: Cambridge University Press. Ong, Walter. 1982. Orality and Literacy: The Technologizing of the Word. New York: Methuen. Pratt, Mary L. 1977. Toward a Speech Act Theory of Literary Discourse. Bloomington: Indiana University Press. . 1981. "The Ideology of Speech Acts". Centrum 1. 5-18. Rosaldo, Michelle Z. 1982. "The Things we Do with Words: Ilongot Speech Acts and Speech Act Theory in Philosophy". Language in Society 11: 203-37. Schlieben-Lange, Brigitte. 1976. "Für eine Analyse von Sprechakten. Sprachtheorie und Prag­ matik". Akten des 10. linguistischen Kolloquiums, Tübingen 1975, I, ed. by Heinrich Weber and Harald Weydt. Tübingen: Max Niemeyer Verlag. . 1983. Traditionen des Sprechens: Elemente einer pragmatischen Sprachgeschichtsschrei­ bung. Stuttgart: W. Kohlhammer. Searle, John R. 1979. Expression and Meaning. Cambridge: Cambridge University Press. . 1983. Intentionality: An Essay in the Philosophy of Mind. Cambridge: Cambridge Univer­ sity Press. Shepherd, Susan C. 1982. "From Deontic to Epistemic: An Analysis of Modals in the History of English, Creoles, and Language Acquisition". Papers from the Fifth International Confer­ ence on Historical Linguistics, ed. by Anders Ahlqvist. Amsterdam: John Benjamins. Sweetser, Eve E. 1984. Semantic Structure and Semantic Change: A Cognitive Linguistic Study of Modality, Perception, Speech Acts, and Logical Relations. Ph.D. dissertation, University of California, Berkeley.

406

ELIZABETH CLOSS TRAUGOTT

Talmy, Leonard. 1978. "The Relation of Grammar to Cognition — a Synopsis". Proceedings of TINLAP 2, ed. by D. Waltz. Champaign, Il: Coordinated Science Laboratory, University of Illinois. . 1983. "How Language Structures Space". Spatial Orientation: Theory, research, and application, ed. by Herbert Pick and Linda Acredolo. New York: Plenum Press. Traugott, Elizabeth Closs. 1978. "On the Expression of Spatiotemporal Relations in Lan­ guage". Universais of Human Language III: Word structure, ed. by Joseph H. Greenberg, Charles A. Ferguson, and Edith Moravcsik. Stanford: Stanford University Press. . 1985. "'Conventional' and 'Dead' Metaphors Revisited". The Ubiquity of Metaphor, ed. by René Dirven and Wolfgang Paprotte. Amsterdam: John Benjamins. . 1986. "On the Origins of "and" and "but" Connectives in English". Studies in Language. Traugott, Elizabeth Closs and Richard Dasher. 1987. "On the Historical Relation between Mental and Speech Act Verbs in English". Papers from the Seventh International Conference on Historical Linguistics, Pavia, 1985, ed. by Anna Giacalone Ramat et al. Amsterdam: John Benjamins. Verschueren, Jef. 1980. On Speech Act Verbs. Amsterdam; John Benjamins, B.V. Viberg, Ake. 1983. "The Verbs of Perception: A typological study". Volume on "Explanation for Language Universals", ed. by Brian Butterworth, Bernard Comrie, and Östen Dahl. Linguistics 21. 125-62.

Primary dictionaries consulted American Heritage Dictionary of the English Language. Bosworth. Joseph, rev. by T. Northcote Toller. 1898. An Anglo-Saxon Dictionary. London: Oxford University Press. Buck, Charles Darling. 1949. A Dictionary of Selected Synonyms in the Principal Indo-European Languages. Chicago: University of Chicago Press. Ernout, Alfred and Antoine Meillet. 1959. Dictionnaire étymologique de la langune latine: histoire des mots. Paris: Klincksieck, 4th ed. Larousse. Grand Larousse de la langue française. 1971. Paris: Librairie Larousse. Middle English Dictionary. 1969-. Ed. by Sherman M. Kuhn and John Reidy. Ann Arbor: University of Michigan Press. Oxford English Dictionary. Souter, Alexander. 1949. A Glossary of Later Latin to 600 A.D. London: Oxford at the Clarendon Press. Walde, Α., rev. by J.B. Hoffman. 1954. Lateinisches etymologisches Wörterbuch. Heidelberg: Carl Winter. Walde, Alois, rev. by Julius Pokorny. 1930. Vergleichendes Wörterbuch der indogermanischen Sprachen. Berlin: Walter de Gruyter. Wartburg, Walther von. 1928-65. Französisches etymologisches Wörterbuch. Basel: R.G. Zbinden and Co.

Grammar and Pragmatics: The Two Axes of Language and Deixis Edna Andrews Duke University 1. Deixis and Shifters after Jakobson — Herman Parret The concept of deixis has become the focus of a great deal of attention in the last several decades, particularly after the appearance of Jakobson's pioneering work on "shifters" (Jakobson 1957). The so-called shifting categ­ ories of language evoke many questions which have implications in disci­ plines outside of linguistics proper, such as philosophy, discourse analysis and semiotics. Parret brilliantly demonstrates the implications a subjectively based theory of deixis can have in terms of heuristic and methodological consequences. He clearly states that the ability to analyze deixis and deictic elements must be a fundamental criterion for a useful theory of meaning from the philosopher's, as well as the linguist's, point of view. Parret's work on deixis, and in particular his treatment of the relationship between deixis and subjectivity and his definitions of pure and impure deixis, is a major contribution to both linguistics and philosophy. As a fundamental premise of his work, Parret (1985) states that "the deictic context is not external (ontological) but subjective". This definition would accurately characterize the type of deixis which is often discussed in the literature, a type of deixis named TRANSMISSIONAL by van Schooneveld (1983a, b). In transmissional deixis, the identifiers are restricted to the speaker (locutor) and addressee of the speech event. Thus, transmissional deixis exactly corre­ sponds to Parret's general definition of deixis. However, Parret further refines his definition to distinguish between PURE and IMPURE deixis. Pure deixis is given by words such as I, you, this, that, here and there, where they refer solely to the speaker and addressee and give no additional information about the interlocutors. Impure deixis is given by the third person pronouns {he, she, it) as they involve information specifically tied to the referent and have nothing to do with the present status of the speaker and addressee. Therefore, Parret measures the purity of deixis by its closeness to the "egocentric here and now". Particularly perceptive is Parret's portrayal of

408

EDNA ANDREWS

the complexity of categorizing shifters like here and distinguishing between the PURE INDEXICAL as opposed to the PURE DEMONSTRATIVE.

I would like to add to the above schema another type of deixis, which would allow for a more "objectively" given base of identifiers. Van Schooneveld calls this type of deixis PERCEPTIONAL. Perceptional deixis has in its set of potential identifiers any observer of a given narrated event. Such an expanded variety of deixis, in combination with TRANSMISSIONAL deixis, allows the linguist to rigorously discuss and categorize a fundamentally "subjective" phenomenon. Furthermore, by including perceptional deixis in linguistic analysis, the particular relevance of deixis can be demonstrated to exist not only at the CODE/MESSAGE level (C/M), but also at the C/C, / and M/M levels. These two types of deixis bring Jakobsonian linguistics more closely in tune with the Peircean notion of the internal relationship between signans and signatum and would likewise contribute to a redefinition of the Jakobsonian notion of CONTENT.

2. Praguean structure and autopoiesis: Deixis as individuation — C.H. van Schooneveld If the principle of deixis permeates the structure of language, as van Schoone­ veld has stated, then the ultimate connection between deixis and markedness, one of the fundamental principles of determining binary opposition in Jakobsonian linguistic theory, must be reconciled and made explicit. In the present work, van Schooneveld makes this final step and answers the ques­ tion concerning the primacy of semantic features. In the system of deixis developed by van Schooneveld, there are two primary types of deixis that exist in two variants: PERCEPTIONAL, TRANS­ MISSIONAL, SINGULATIVE PERCEPTIONAL and

SINGULATIVE TRANSMISSIONAL.

These four types of deixis exist in a hierarchy and are intimately related to the semantic features developed by Jakobson (1936, 1958) and expanded by van Schooneveld (1978). The distinction between perceptional and transmissional deixis was dis­ cussed above. What remains to be defined are the two types of singulative deixis and a reinterpretation of the deictic levels vis-à-vis the six semantic features as posited by van Schooneveld. Singulative perceptional deixis requires that all identification be made simultaneously in the narrated event; thus, the speaker plays a double role as identifier in both the narrated and speech events. Singulative transmissional deixis requires simultaneous identi­ fication in the speech event; thus, all identifiers must be present for the actual transmission of the utterance. Just as each FEATURE contains the information of the feature preceding it

GRAMMAR AND PRAGMATICS

409

in the hierarchy as a given, a phenomenon which is autopoietic by definition, so transmissional deixis is a more restricted type of deixis, and hence a more marked type, than perceptional deixis. Likewise, singulative perceptional deixis is less marked than singulative transmissional and the following hierarchy is formed: less marked — Perceptional deixis Transmissional deixis Singulative Perceptional deixis most marked — Singulative Transmissional deixis In order to fully integrate deixis with the six binary semantic features, which characterize properties of meaning, van Schooneveld has demonstrated that the two most basic semantic features define the different levels of deixis as well: Perceptional - 0 (unmarked) Transmissional —+ PLURALITY Singulative Perceptional —+ DIMENSIONALITY Singulative Transmissional —+ PLURALITY/ + DIMENSIONALITY By developing such a complex system of codification of different types of perception acts, a cohesive theory of meaning is the result. Van Schooneveld's work has expanded the number of semantic features from Jakobson's original three to six. Likewise, he has discovered an additional type of deixis, which exists in singulative and non-singulative subtypes. However, one of his greatest contributions to a theory of meaning is found in his discovery that both the semantic features and the four deictic levels exist in a strict hierarchy and in his reintegration of the semantic features with the deitic levels.

3. Shifters and Non-Verbal Categories of Russian — Olga Yokoyama Jakobson's "Shifters" article is an extremely stimulating piece of scholarship for its rigorous description of the factors of the speech event (Es), narrated event (En) and the various participants of these two events (Ps, Pn) in analyzing not only the Russian verbal system, but also shifters and other elements of discourse in language. Yokoyama deftly makes use of these distinctions in her analysis of the Russian diminutives, reflexive possessive pronoun, conjunctions and word order. By demonstrating the relevance of the role of the participants of the speech and narrated events, Yokoyama sets up an elegant system that provides an outstanding and lucid explanation for usage of the Russian non-verbal shifting categories in a way that most linguists and pedagogues have essentially ignored.

410

EDNA ANDREWS

Yokoyama clearly proves that choice of the reflexive possessive pronoun is not simply a matter of taste in the first and second person pronouns, but is determined by the relationship between the speaker and addressee (i.e. Ps1 → Ps2) In other words, although she notes the importance of the relation­ ship between the speaker and the referent of the process (Ps1 → P n ), Yoko­ yama points out that it is very often the case that Ps1 → Ps2 plays as important a role as Ps1 → P n , a fact that has been overlooked by most grammarians. In order to substantiate this claim, Yokoyama presents Russian data, in which the Russian personal possessive pronoun is prohibited in favor of the reflexive possessive when the statement is oriented toward the addressee (Ps1 → Ps2) or when the speaker views himself from the outside (Psla → Pslb), thus doubling himself and setting himself up as a potential addressee of the utterance (cf. Ja svojo delo znaju; Ja uvidel ν zerkale svojo krasnojo lico) (Yokoyama 1985). Furthermore, Yokoyama demonstrates that in a like manner, the reflexive possessive pronoun is rejected in favor of the personal possessive in utterances where the speaker is oriented toward himself and his inner thoughts and feelings (cf. Ja ljublju moju mat') (Yokoyama 1985). Such a principle involving addressee versus speaker orientation within the utterance is significant even in those cases where a particular grammatical form is heavily weighted in the direction of the addressee (cf. Russian imperative mood [Jakobson 1960]). Yokoyama further argues that reflexizication is not always clause-bounded, as was traditionally thought, and convincingly demonstrates by means of numerous examples (Klenin 1974, Yokoyama 1979) that once again, the relation of Ps1 → Ps2 is as important as Ps1 → n. I would like to include at this point certain other examples of Russian reflexive possessive pronominal usage which support Yokoyama's argument, as well as support the case that deixis and semantic features play an import­ ant role in the characterization of the speech and narrated events and the participants of these events. As is well known, there are several examples in Russian where the reflexive possessive pronoun can occur as the subject of the sentence, or where the reflexive refers to a referent not given in the sentence: 1 a. Svoja noša ne tjanet "A burden of one's own choice is not felt" 1 b. Každomu svojo "To each his own" 1 c. U každogo svoi interesy "Each person has his own interests" 2 a. On svoj čelovek "He's one of us" Once again, it is clear from these examples that the reflexive possessive pronoun in Russian is tied not only to the participants of the speech event, but also to the participants of the narrated event; such expressions and sayings (cf. la,b,c) refer to any P n , whereas in other examples (cf. 2a) the

GRAMMAR AND PRAGMATICS

411

participant referred to is clearly given within the speech event (Ps1). The motivation for such a duality in the reflexive has been explained by van Schooneveld in terms of deixis. Van Schooneveld (1982) has demonstrated that the reflexive pronouns are marked for both transmissional and singulative perceptional deixis. Thus, he has provided a rigorous means of determining the exact potential relationship to participants given within the speech event (transmissional deixis) or participants given within a specific narrated event (singulative perceptional deixis). In feature terms, van Schooneveld has shown that the Russian reflexive pronoun is marked, in addition to its deictic markings, for the feature EXTENSION (Jakobson's DIRECTIONALITY), while the first person pronoun is marked for DISTINCTNESS (van Schooneveld 1973, 1982, 1983, 1984). In an excellent discussion of the Russian conjunctions a and no, Yokoyama once again demonstrates the crucial role that the relationship between the participants (speaker and addressee) plays in the utterance. However, I believe there to be a very special relationship between the reflexive possessive pronoun and the conjunctions a and no that remains to be explored and elucidated. Specifically, the conjunction no, like the reflexive possessive pronoun svoj, involves the speaker addressing himself to the hearer's reaction (Ps1 → Ps2), while a, like the personal possessive pronoun moj, involves the speaker's reaction to himself (Psla → Ps1b)· Thus, one clearly sees the codified nature of the relationship between the participants of the speech and nar­ rated events in very diverse grammatical and lexical categories. The special relationship between the reflexive possessive pronoun and the contrastive conjunctions in Russian is further supported by van Schooneveld's work on the Russian conjunctions (van Schooneveld 1978, 1982), where he demon­ strates that the semantic features EXTENSION and DISTINCTNESS, stated above in connection with the reflexive and personal possessive pronouns, once again serve as the markings distinguishing a and no.

English Speech Act Verbs: A Historical Perspective — Elizabeth Traugott In the early years of the Prague School of Linguistics, one of the primary principles formulated by Jakobson, Karcevski and Trubetzkoy concerned a reformulation of the Saussurian notion of diachrony. The Prague School explicitly claimed that diachronic change is indeed systematic by definition, thereby asserting that predictability on a general level is possible. Traugott's work on English Speech Act Verbs (henceforth SAVs) proves the relevance of such a claim and demonstrates its utility in linguistic analysis. Based on a corpus of 275 verbs, Traugott focuses primarily on English performative verbs and then generalizes her conclusions to SAVs in general.

412

EDNA ANDREWS

One of the cornerstones of the analysis given by Traugott is the use of etymologies as a key to finding a pattern in semantic change. Traugott paves the way for expanding the traditional use of etymologies to other cognitive domains. In order to analyze SAVs, Traugott divides them into four etymo­ logical categories: verbs of uttering, mental and psychological states, vision, and spatial relationships. She further states that the first two groups are synechdochic in nature, the latter two metaphorical. It is only with excep­ tional examples that the principle of metonymy is involved. By way of conclusion, Traugott proposes that the material makeup of SAVs, which involves objects that can be manipulated in various ways, is possibly a language universal. Certainly there is support for such a claim from Indo-European, as well as non-Indo-European languages. Traugott's observation concerning metonymy and metaphor is extremely significant in the scheme of linguistic analysis and the quest for language universais. I would like to briefly review the significance of metonymy and metaphor within Jakobsonian linguistics and suggest an explanation for the dominant role metaphor plays in the etymologies of English SAVs. Jakobson (1956) states that the linguistic sign necessarily involves two modes of arrangement, or two poles: combination and selection. The axis of selection deals with elements related in the code which are thus given outside the context of a given message, whereas the axis of combination deals with elements related within the context of the given message or involves both code and message. Metaphor relates specifically to the axis of selection (or substitution) and metonymy to the axis of combination. By definition, the axis of selection gives rise to paradigmatic phenomena, which in turn more closely parallel the phenomenon of diachrony. Therefore, I would suggest that it is no mere coincidence that it is precisely metaphorical extensions which play a significant role in the semantic shift in the etymolog­ ies of English SAVs. It would be interesting to analyze data involving SAVs from other lan­ guages to determine whether or not metaphor plays as dominant a role as in the English SAVs. If so, we are one step closer to unraveling the complexit­ ies of diachronic language change.

Each of these papers serves as an excellent example of the rich and fruitful theoretical basis that the Jakobsonian legacy has provided linguists in their pursuit for the ultimate invariant properties that characterize human lan­ guage. Whether the discussion involves specific linguistic data, particular theories or other disciplines, the fundamental principles presented by Jakobson and the Prague School continue to be relevant and stimulating in the linguistic research of the 1980's and beyond.

GRAMMAR AND PRAGMATICS

413

References Jakobson, Roman. 1936 [1971]. "Beitrag zur Allgemeinen Kasuslehre: Gesamtbedeutungen der Russischen Kasus". Selected Writings II, 23-71. The Hague: Mouton. . 1956 [1971]. "Two Aspects of Language and Two Types of Aphasic Disturbances". Selected Writings II, 239-259. The Hague: Mouton. . 1957 [1971]. "Shifters, Verbal Categories, and the Russian Verb". Selected Writings II, 130-147. The Hague: Mouton. . 1958 [1971]. "Morfologičeskie Nabljudenija nad Slavjanskim Skloneniem". Selected Writings II, 154-183. The Hague: Mouton. . 1960 [1981]. "Linguistics and Poetics". Selected Writings III, 18-51. The Hague: Mouton. Klenin, Emily. 1974. Russian Reflexive Pronouns and the Semantic Roles of Noun Phrases in Sentences. Ph.D. dissertation, Princeton University. van Schooneveld, C.H. 1973. "The Morphemic Structure of the Slavic Word and Greenberg's Twenty-eighth Universal". Slavic Word, ed. by D.S. Worth, 443-448. The Hague: Mouton. . 1978. Semantic Transmutations. Bloomington, Indiana: Physsardt. . 1982. "The Extension Feature in Russian". IJSLP 26. 445-457. . 1983a. "Programmatic Sketch of a Theory of Lexical Meaning". Quaderni di Semantica 4:1. 158-170. . 1983b. "Comments to other Contributions of the Roundtable". Quaderni di Semantica 4:2. 117-124. . 1987. "Linguistic Structure and Autopoiesis". Language, Poetry and Poetics. Proceedings of the First Roman Jakobson Colloquium, ed. by K. Pomorska et al., 123-142. Berlin: Mouton de Gruyter. Yokoyama, Olga T. 1979. "Reflexivization into Adjectival Reduced Relative Clauses in Rus­ sian". Folia Slavica 2. 366-375.

PART FIVE TYPOLOGY AND UNIVERSALS

Two Approaches to Language Universals

Joseph H. Greenberg Stanford University Although there have been few published discussions, linguists have, since at least the mid-60s, been aware of two distinct approaches to language univer­ sais. One of these has been most closely associated with the name of Noam Chomsky and with generative grammar; the other took its start with my paper on word order (Greenberg 1963), was characteristic of most of the work of the Stanford Project on Language Universais, and is now continued by individual researchers as well as centers at Cologne and elsewhere. Among the earlier discussions of the two approaches may be mentioned Bell (1971), Bugarski (1972: 126-8), Lehmann (1972) and Ferguson (1978). These discussions are all by non-generativists, some not committed to either approach (e.g. Bugarski), or typologically oriented scholars such as Bell, Lehmann and Ferguson. They are all moderate in tone and state either that a synthesis of the two approaches is desirable (Bugarski), is already underway (Lehmann) or that there has been a degree of convergence chiefly because of changes in the generativist approach (Ferguson). From about 1978 on, however, a group of Italian generativists have been actively engaged in integrating typology into generative grammatical theory. A conference on this subject was held in Bologna in 1980 and resulted in a series of papers in Lingue e Stile the same year. The basic point on which almost all agree is that if it is the task of universal grammar to define "possible human language", then typological considerations must enter because research of this kind leads to the discovery of strong limitations on possible human languages not discoverable by any other method. Ramat (1980: 314) in the introductory paper states that "there is no real contradic­ tion between the principles of generative grammar and the typological approach." Rizzi (1980: 354) concludes that "if the work is well done, there is of necessity a fundamental convergence of results, as is the case with Greenberg's word order universais and X-bar theory". More recently, however, there have been several attacks by generativists on the typological approach to universais. The most active has been Coop-

418

JOSEPH GREENBERG

mans (1983, 1984), the former of which was an attack on Comrie (1981) and the latter on Hawkins (1983). In the summary which accompanies Coopmans (1984) he asserts: "Finally, it shows how the study of language typology might proceed within a generativist framework." This is to be done, as he says, by using Greenberg­ type language surveys to form initial hypotheses, but they will have to be followed by careful analysis within a restrictive theory of grammar. There seems to be here a basic circularity. If one has an adequate restrictive theory of grammar to begin with, why should one have to resort to typological data in order to discover restrictions to be incorporated in the grammar? Having presumably discovered a sort of preliminary pre-theoretic value for typology he then goes on to state that "the two approaches differ greatly and cannot be compared or related in any way" {ibid. 67). Another hostile critique of typology is that of Smith, once more a review of Comrie (1981). He states (1982: 255) that one of the few praiseworthy features of Comrie's book is that it is "extremely rare in even raising and discussing the issues which separate Greenberg and Chomsky". These reviews of Comrie and Hawkins elicited rejoinders (Comrie 1983, 1983), Hawkins (1985) and even one surrejoinder (Smith 1983). A further critical attack is Lightfoot (1981), devoted once more to wordorder typology. Finally a recent work of Stassen (1985), a cross-linguistic study of the comparative construction, provides in its initial chapter the most extensive exposition of the typological approach to universais since Comrie (1981). Like Comrie he asserts that the typological method can make a contribution to universal grammar, an expression which even figures in the title of his book. Two general observations may be made regarding this recent controversial literature. Topically it centers on word order typology especially in the attacks by generativists who seem practically to equate typology with word-order or more accurately, constituent order typology. This is partly understandable in the context of the discussions revolving around Hawkins (1983), a work devoted to word-order. But typology is a far broader notion and by now I and others have treated a considerable number of other topics (cf. Stassen 1985: 2 for a partial enumeration). Even if all the conclusions of word-order typology were to be refuted, and this is clearly not the case, there remains a large body of additional work which has not been subject to critical analysis by generativists. The second general observation to be made is that typologists, e.g. Hawkins and Stassen, in their desire to show that typology can make contributions to linguistic theory have uncritically accepted the equation of linguistic theory with universal grammar whereas, as I shall try to show, the notion of theory in linguistics has a far wider domain than the general principles utilizing the construction of synchronic grammars.

TWO APPROACHES TO LANGUAGE UNIVERSALS

419

My purpose in this discussion is not polemic. It is rather to clarify the issues at stake; but, given, the limitations of an article length treatment and the vastness of the topic, many important aspects will be omitted or treated only cursorily. Further, I will concentrate on setting forth my own views which differ in certain respects even from those of some individuals with whom I share much common ground (e.g. Comrie, Seiler). For generativist views I will rely mainly on Chomsky, but here an even greater variety of opinion and attitudes exist ranging from a considerable receptivity to the typological approach (e.g. Bach, Ramat) to outright rejection (Coopmans, Lightfoot). In addition there are now numerous competing varieties of generative grammatical theories. Also, there have been important changes during the period under consideration so that the treatment is, of necessity, to some degree historical. I will first discuss some characteristics of the two approaches. Secondly, since there is general agreement that the significance of universais lies in their explanatory role, I will discuss my own views regarding the structure of explanation in linguistics. Finally, I will consider very briefly some impli­ cations in these views in their relation to the present climate of theory in linguistics. Three characteristics are often mentioned as distinguishing the typological approach from that of the generativists: looking at many languages as against looking at very few, or in principle perhaps at only one, a concen­ tration on surface as against deep structure and the use versus non-use of typological methods. To a degree these three contrasts can be summarized as an emphasis on breadth as against depth. Human energy being finite, a combination of breadth and depth is beyond any single researcher's capacit­ ies. Moreover, not enough languages are known in depth to make this possible at our present state of knowledge. I believe that both strategies should be employed. Undoubtedly they appeal to different predilections of researchers. The three characteristics of the broad approach are connected in various ways. Typology, which seeks lawful relations among properties of languages necessarily requires fairly large areal and genetic samples of languages and hence inevitably emphasizes easily available and comparable data, thus tending to exclude the idiosyncratic characteristic which inevitably appear when any single language is investigated in very fine detail. These three factors, though thus interrelated, deserve separate treatment. With regard to breadth of linguistic coverage, it is clear that Chomsky himself since his first published discussion of universais (1965) has tended to believe that probably one language was sufficient to arrive at universais. During the initial period at MIT it was generally held that it was impossible to write a grammar of a language unless one had native speaker intuitions.

420

JOSEPH GREENBERG

Since the first generation of generativists were English speakers, this inevi­ tably lead to a strong concentration on English. This notion has generally been abandoned and, of course, generative grammar has spread beyond the English-speaking world. More significantly, data from a variety of languages are now frequently advanced in support or refutation of proposals in gram­ matical theory. However, Chomsky himself states as late as 1976 that "deep analysis of a single language may provide the most effective means for discovering non-trivial properties of universal grammar (1976: 56)". More recently he has somewhat relaxed his stance. While noting that "intensive studies of particular languages is likely to give a deeper insight into UG [Universal Grammar] than less far-reaching study of a wide variety of languages," he goes on to cite approvingly a 1981 dissertation by Marantz that "combines coverage of a wide variety of language types with construc­ tion of a well articulated theory of UG (1982: 92)". There remains, however, a difference between citing data from several languages in regard to a particular issue and systematically pursuing a linguistic topic by examining a broad and representative sample of languages to arrive at generalizations. The question of one versus many languages is more than just a matter of tactics. As pointed out in Comrie (1981) it is in principle impossible to arrive at implicational universais by looking at just one language. If, for example, I find a language like Sindhi with both front and back implosives, it is only by comparing it with other languages that it becomes possible to discover the universal implication that whenever a language has back implosives it has front implosives. A further difficulty is the following. In earlier discussion of universais by the MIT school a classification into substantive and formal universais was made. Both Katz and Postal (1964) and Chomsky (1965) give as examples of substantive universais the Jakobsonian set of universal phonological features. However, it is not asserted that all languages employ all of them. In fact, no single language makes use of the total set. Hence one could never discover them by investigating only one language. The second issue mentioned earlier is attention to surface structure as against deep structure. According to Chomsky: There is no reason to expect uniformity of surface and the findings of modern linguistics are thus not inconsistent with the hypotheses of universal grammarians. Insofar as attention is restricted to surface structure, the most that can be expected is the discovery of statistical tendencies such as those presented in Greenberg (1963) (Chomsky 1965: 118) Chomsky then further remarks that "in general it should be expected that only descriptions concerned with deep structure will have import for pro­ posals concerning linguistic universais". In accordance with this view,

TWO APPROACHES TO LANGUAGE UNIVERSALS

421

notions of a universal deep structure, now abandoned, were fairly widely held in the sixties That regularities in surface structure were unexpected during this period may also be gathered from the following statement This is not to say that it is unprofitable to look at a wide range of languages given our present knowledge of what languages are like...A case in point is Greenberg (1963). Greenberg took a very superficial look at thirty languages. He considered only the surface structure of certain types of sentences in these languages...of course, Greenberg could not have done anything more sophisticated along these lines since there do not exist thirty diverse languages that been studied extensively beyond the level of surface structure..., there was no reason why he should have come up with any interesting facts at all... What is remarkable is that Greenberg did come up with some results... Considering the impoverished theory of grammar in which he was working — phrase structure grammar — it is remarkable that he was able to describe any universal facts at all that are of interest. But though Greenberg was able to describe these facts he could not explain them within his theory. Indeed there exists no known theory of grammar in terms of which it can be explained. (Lakoff 1968: 6-7) It is of interest to note in regard to the foregoing statement that Lakoff believes that one must have a theory of grammar before investigating linguis­ tic phenomena. Since I was obviously not a generativist, he assumed that I must be working in the theory of phrase structure grammar that dominated American structuralism before the rise of generative grammar. In fact it was not that I was espousing either surface or deep structure theories. I was simply indifferent to the distinction between the two, often relating surface phenomena which in a generative account would at that period require a derivation in deep structure of one from the other, or both from a different structure. This whole question now appears to be moot. Around 1970 Chomsky began to develop his Extended Standard Theory as an answer to generative semantics. This involved both phonological and semantic interpretation at the level of surface structure. The very term "deep structure" was ultimately abandoned to be replaced by "initial phrase marker" which had much more modest functions than the earlier "deep structure". Thus by 1975 Chomsky observed regarding deep structure: The term "deep structure" had unfortunately proved to be misleading. It has led a number of people to suppose that it is the deep structure structures and their properties that are truly "deep" in the non-technical sense of that word while the rest is superficial, unimportant, variable across language and so on. (1975: 82) One may further observe that certain generalizations are only stateable as surface phenomena. These began to be noted as "conspiracies". A whole

422

JOSEPH GREENBERG

series of deeper rules of varied types, "conspired" to produce a surface regularity. For example in Yurok, Classical Arabic, Hausa and other lan­ guages a highly restricted set of syllable types, namely CV, CV, CVC, identical for all these languages, resulted from a diversity of rules. But universal conspiracies also exist. Thus, as far as I am aware, whatever order properties a language may possess, a first or second person pronoun always precedes its apposition as in English, "We, the people of the United States,..." Finally, in recent years, in addition to the Chomskyan model, a whole series of generative grammatical models have arisen all oriented towards surface structure. The third characteristic mentioned initially was the use of typology. In my view, typology is indispensable but it is not the sole basis for principles of linguistic explanation. Typology is inextricably bound up with universais among which the implicational type figures prominently.1 It is also related to marking theory in ways to be mentioned presently. The term "marking" in this context is used essentially in its Prague School meaning of hierarchical relations among categories and not as a set of conventions in universal grammar such that the extent to which a grammar deviates from such conventions the more marked and less highly valued it is. The interconnection of typology, universais and marking may be stated as follows. The absence of certain logically possible types is equivalent to particular kinds of universalist statements, depending on the pattern of the typology and the missing types. In practice, the most common situation is that of four logically possible types, based on the presence or absence of two properties. Of the four, one is frequently absent or statistically rare. This can be restated as a universal implication. For example, the non­ existence of languages with back implosives but without front implosives is logically equivalent to the statement that the presence of back implosives universally implies the presence of front implosives. In turn, when such an implicational relation exists, the unmarked category is the implied one. But marking involves a whole cluster of other characteristics such as non-zero expression of the marked category in morphology, its lesser text frequency in both morphology and phonology, etc. (cf. Greenberg 1966). Chomsky himself never seems to use the term "implicational universais" and until recently the term "typology" was also excluded. However, it is not difficult to see that these concepts, especially after the initial period of generative grammar, are in fact employed. Thus in the famous ninth chapter of Chomsky and Halle (1968), a work on the phonology of English, a whole series of marking conventions are set forth so that certain feature combi­ nations are designated as  and it costs a rule to change them to a marked combination, thus making the grammar less highly valued. But what is the source of the conventions themselves? They can't be derived from generative

TWO APPROACHES TO LANGUAGE UNIVERSALS

423

grammars because they are logically prior and intended to be utilized in writing the grammars themselves. The answer is clear. The unvoiced nasal, for example, is universally marked because of its relative rarity in languages and the implicational universal that the presence of unvoiced implies the presence of voiced nasals. These facts in turn are known because of earlier cross-linguistic typological surveys of investigators like Trubetzkoy and Hockett. A further value of marking hierarchies of the Prague type was proposed by Jakobson (1941) namely that the unmarked precedes the marked in first language acquisition. An important contribution of Chomsky is the immense stimulus he gave to research on child language acquisition. But we find here again an important change of position without explicit acknowledgment. In their work on the sound patterns of English, Chomsky and Halle say: We have been describing the acquisition as though it were an instantaneous process. Obviously this is not true. A more realistic model of language acquisition would consider the order in which primary linguistic data are used by the child... To us it appears that the more realistic model is too complex to be taken up in any meaningful way today and that it will be more fruitful to investigate in detail, as a first approximation, the idealized model outlined earlier, leaving refinements to a time when this idealization is better understood. (1968: 331) However, more recently Chomsky states, in what is I believe his first favor­ able reference to my work, that: Universais of the sort explored by Joseph Greenberg and others have obvious relevance to determining just which properties of the lexicon have to be learned in this manner in particular grammars — and to put it in other terms just how much has to be learned as grammar develops in the course of language acquisition. (1981: 95) If I understand correctly, the lexicon in EST contains unmarked categories that are then changed to marked categories in specific grammars, and the assumption is being made that these rules are learned by children after a period of using the unmarked form. The most obvious of the tacit use of implicational universais is, however, the X-bar conventions introduced in 1970. Originally it was posited as a variable which could be rewritten as head, modifier or head, complement and then specified as noun, verb, or prepositional phrase. However in its initial form there was no ordering of head and modifier complement in universal grammar. Later a convention was introduced by which, for any particular grammar the order of head and non-head was the same for all X-bar constructions. The obvious connection with word order typology was first pointed out in Riemsdijk (1978). There remains a deeper difference between my approach and that of

424

JOSEPH GREENBERG

others which has been much more rarely noted. I refer to the generativist assumption that linguistic theory is to be equated to the principles involved in the construction of synchronic grammars, and that, correspondingly, universais are the study of the basic properties underlying such grammars. A representative statement is Chomsky (1965: 28) that "the study of univer­ sais is the study of the properties of any generative grammar for natural language". For a considerable period, Sanders was almost alone in seeing that there is even an issue here. He noted (1974: 17) that "we can account for facts about natural languages either directly — by laws quantified over the set of all possible languages — or indirectly by laws quantified over the set of all possible grammars of such language". But then without supporting argument he asserts that it is preferable, perhaps necessary, that all facts and laws about languages be specified in terms of facts about grammars. A further limitation that these should be synchronic grammars is simply taken for granted. Later, Lieb and Seiler took note of this issue. The former (1978: I:61) states that "a language universal is a property of languages, not grammars, and does not depend on grammar". Further, regarding the approach through grammars, "the extent to which this has been accepted uncritically is stunning". In more recent discussion this issue has become more prominent. Smith (1982: 256) in his criticism of Comrie, talks about the Chomskyan paradigm which he supports as characterized by "its emphasis on grammars rather than languages". While Comrie in the very title of his paper of 1978 "Linguis­ tics is about languages" took the opposite stance. However, both Comrie and Hawkins in their replies to critics seem to accept the notion that linguistic theory is the same as universal grammar and seek to show that the typological approach can make a contribution to linguistic theory by contributing to the theory of grammar. I shall deal first with the question of generalizing over grammars rather than languages and secondly, the limitation to synchronic grammar. My objection to both is basically similar. They constitute arbitrary restrictions both on the range of relevant data to be explained and the nature of the explanations themselves. Regarding the first, there are many important properties of languages which will not appear in grammars. They may be classified into three types, grammar derived, transgrammatical and non-grammatical. By grammar derived properties will be meant ones that can be logically derived from grammars but are not explicitly stated in them. The following is an example. Nowhere in any grammar based on current models could it be stated that the French noun had two genders, although one could no doubt derive it from the rule specifying that all nouns in the lexicon must be marked as

TWO APPROACHES TO LANGUAGE UNIVERSALS

425

masculine or feminine. However, present theories give no motive for deriving such a statement. Now it turns out that if one compares two-gender languages typologically there are some interesting limits on possible types. The only features that ever appear are animacy or sex. There is, however, a third type in which feminine in grouped together with inherently sexless items so that the con­ trast is masculine versus non-masculine. However, as far as I know, there is no system with a basically ± feminine contrast. More serious in my view are what I shall call transgrammatical facts. These involve a relationship between the facts of two languages which will appear in both grammars but the relationship between which will not be found in either. A typical example is when a large number of words have been borrowed from a language A into a language  and these borrowed words show different phonological constraints or derivational formations from native words. An attempt has been made to deal with such situations by positing lexical features such as [ + French] in English or [ + Arabic] in Turkish. However what this will not show is the systematic changes which occur in the process of the transfer from A to  which may give further evidence regarding, for example, sound similarity as shown in substitution and which can be compared with evidence from other sources regarding the same subject. Finally, there are certain types of facts which will not appear in grammars at all and yet have theoretic significance. If a grammar is a model of linguistic competence, I fail to see how the fact that in languages with morphological plurals the singular is used about four times as frequently as the plural in text can or should be included in the grammar. It is well-known that frequency data play an important part in marking theory which in turn has significance for cross-linguistic generalizations. In many instances whether facts are stated directly about languages, or about grammars will make no real difference. But if a rule or other property of a grammar is wedded closely to a particular theory it will be an artifact of that theory and not related to language's capacities. An example is that Katz and Postal in the first published discussion of universais by generativists, enunciated a distinction between substantive and formal universais. In attempting to exemplify the latter they state (1964: 160): "Examples of one of the most detailed formal universais is a specification that certain rules must be found in the syntactic component of any linguistic description — for example the rule Rel → md." The rule given here means that Relative is to be rewritten as matrix dummy. But matrix dummy was a purely theoretical constraint to allow recoverability of the place of insertion of a relative clause. A year later it was no longer a universal. The other major limitation on explanatory theory is the restriction to

426

JOSEPH GREENBERG

synchronic factors. Here there has been some broadening of scope in that we have witnessed a revival of interest in historical linguistics. The seminal events here were Kiparsky (1968) and King (1969). However the generativist approach took two main forms which follow predictably from the equation of theory with synchronic grammatical theory. One of these is that historical change becomes theoretically relevant insofar as a particular complex synch­ ronic rule is strengthened as a valid generalization if it resulted from a unitary historic change (the problem of curly brackets). The other is the conception of linguistic change as the replacement of one grammar by another. This resulted in a classification of changes under such rubrics as rule addition, rule deletion, rule reordering, etc. On a broader view, it was claimed that the general trend of historical change was towards grammatical simplification. After the initial excitement, the weakness of this approach, it is fair to say, became generally evident. In Cohen (1974), one of the very few book length treatments of expla­ nation in linguistics, only Dretske, a philosopher of science, takes any note of the possible explanatory role of historical linguistics. Noting that perhaps not all linguists are doing the same things, he raises the possibility that there might be corresponding types of linguistic explanation. He asks (1974: 23): "Or should one take history as one's paradigm and think of linguistic explanations on the model of historical explanations?" It is my view that historical explanation must play an autonomous and not merely auxiliary part in the overall structure of linguistic explanation. Moveover, it is not confined to the explanation of facts of individual languages but generalizable through the concept of diachronic process as a class of similar but historically independent changes and through the notion of recurrent patterns of change of type. I shall seek to clarify these notions by a brief sketch of how I concieve of the overall structure of explanation in linguistics. There is, of course, an enormous literature on scientific explanations. In linguistics, in addition to numerous incidental but relevant discussions scattered throughout the litera­ ture, there are also several collections of papers devoted to this topic (Cohen 1984, Hornstein and Lightfoot 1981, Butterworth, Comrie and Dahl 1984). It is obviously not possible to discuss all the issues raised in this literature. In the present context, however, a few general observations are in order. One is that the concept of explanation is inextricably intertwined with that of theory. This is so much so that the expression explanatory theory, often encountered in the linguistic literature, may be considered redundant. What would a theory be that didn't explain or at least seek to explain anything? Another indication of the close connection between theory and explanation is that discussions of explanation generally center on problems of theory construction. Germane to the present discussion is that by considering a

TWO APPROACHES TO LANGUAGE UNIVERSALS

427

grammar as a theory of a language by a tacit transition it becomes the explanation of a language. Besides its connection with explanation, a further observation may be made concerning the term "theory". We expect it to have a certain degree of comprehensiveness and generality to deserve the name of theory. Often two basic levels are proposed, a lower one of empirical generalizations and a higher level from which empirical generalizations are deduced. Such a higher level statement is frequently called a law. I believe it is possible within the linguistic data to demarcate two basic levels. In each there are various sublevels. Although there seems to be an eternal regress of questioning, since it is always possible, like a pesky child, to ask for an explanation, it is nevertheless possible to say that on a particular level something has been explained when we no longer ask that question but move on to a higher level. One other preliminary observation is in order. To begin the explanatory ascent we need a lowest level as a starting point. It has been claimed that there are no facts which do not presuppose a theory; the term theory-laden is sometimes used. However, there does seem to be a lowest level of sheer facts. Without it we would never get started and without it there would be no locus for the lowest Chomskyan level of adequacy, observational adequacy. It is characteristic of this level that although it may employ theoretic terms it never explains anything else but is itself in need of explanation. Moreover, any theory which contradicted it would be rejected. An example would be the fact that in Latin the dative singular of puer "boy" is puerō. No doubt dative and singular are theoretic terms but that puerō is the dative singular of puer does not explain any other fact about Latin. 2 Moreover, any grammar which gave different results would fail at the lowest Chom­ skyan level, that of observational adequacy. At a higher level we may indeed ask questions which stem from the theory itself. Whether transformations change meaning makes sense in the framework of earlier generative grammatical theory. Are all types in a typology connected in the sense that ultimately, if not directly, any type can change into any other, what I call the hypothesis of connectivity? As a question this only makes sense in the context of diachronic typological theory. To illustrate the problem of types and levels of explanation I will start with some Turkish examples. Suppose we are told that the plural nominative of adam "person" is adamlar but that the corresponding form of el "hand" is eller, so that in one instance the plural mark is -lar and in the other -ler. If we answer that all singular nominative nominal forms with -a- in the last syllable take -lar while those with -e- take -ler we will stop asking about the plurals of adam and el. If we then ask why all nouns in -a- take -lar and all

428

JOSEPH GREENBERG

those in -e- take -ler, we can be told that there are eight vowels in Turkish and that nouns whose final vowels are , ,  or  take -lar and those with e, i, ö or ü take -ler. Once we know this we stop asking about a and e but we may ask the reason for this particular division of vowels and it can be pointed out that the first set are all back vowels and the second set all front vowels. At this point we can give all the basic rules of the Turkish vowel harmony system which works throughout the language, not only in nouns. There are also a certain number of easily specified exceptions which are either loan words (especially Arabic) or certain affixes. We can thus answer literally thousands of lower level questions by a sample set of rules. All of this is, of course, synchronic and we would certainly say that a grammar of Turkish which maximally generalized statements about vowel harmony was more explanatory than one which did not. However, we must call even this an empirical generalization rather than a theory. A theory contains one or more laws. What we have talked about is only true of Turkish. Expressed in law-like terms we would say "whenever a language is Turkish all the vowels in a word are back vowels, or all are front vowels". It is characteristic of laws that although they hold only under stated conditions, these conditions are not, in general, expressed under proper name conditions. This seems a natural point at which to separate empirical generalizations from laws. We have grown so accustomed to the view of a grammar of a language being a theory of that language and hence explana­ tory, that this point has not been generally appreciated. One of the dissenting voices is once again Sanders: Thus if a grammar is a theory about a particular language, then it is a theory which has a grossly unnatural domain — no less unnatural than a theory about a particular cat (as opposed to a theory about all cats or all animals). The domain of all languages, on the other hand, is a perfectly natural one relative to the present state of scientific knowledge and its foreseeable developments. (Sanders 1978: 16) What all this suggests is that the next natural step is not to look within Turkish itself, but rather at vowel harmony systems typologically across languages. In such investigations we generally find that there are logically possible types which do not occur and such results are stateable in universalistic, often implicational form, from which they can be deduced. In the present context this attempt is not made since it would require a separate and extensive study. A general notion, however, may thus be given of the three basic dimensions of vowel quality, namely height, the backfront opposition and lip rounding, the subordinate or marked status of the last is evident from the fact that it is never the sole principle of any system. There are, in addition, other generalizations about vowel systems that point in the same direction.

TWO APPROACHES TO LANGUAGE UNIVERSALS

429

In instances like Turkish vowel harmony, generalization within a language will to a degree be an explanation. However there are many phenomena, e.g. unique or small groups of exceptions, for which synchronic generaliz­ ation in the usual sense is ruled out. English plurals by internal vowel change is an example. The plural men is a unique example of the alternation of a and e to express the singular-plural distinction. There are, of course, other instances of vowel change plurals in English e.g. tooth/teeth; goose/geese; mouse/mice; louse/lice. Lakoff (1970), a work based on his interesting dissertation on syntactic irregularity, calls such instances minor rules even when they cover only one case, and suggests rule features in the lexicon for the entries involving these cases. It is difficult to see in what sense one can talk about a rule when there is no common property for the instances which fall under it. We must say that a of the singular is replaced by e in the plural when the lexical item is man. One is tempted to quote Romans 2:14, in which Paul talking of the pagans who are not under the Jewish law says "these not having the law, are a law unto themselves". But a historical account shows that all of these plurals fall under a generalized account since they arose from a former suffix -i which umlauted the previous vowel and was then lost or reduced. Here once more there is a kind of explanation, in that we stop asking specifically about man/men. Moreover our explanation leads to further generalization in that alternations like old/elder can be shown to result from the same sequence of changes. We can, of course, ask still higher level questions but they will refer to similar umlauting or vowel modification in other historically independent instances. Which vowels produce such changes and how are they gram­ maticized? These are diachronic typological and universalistic questions. Generativists have shown little or no interest in explaining exceptions or language specific rules. For example, Chomsky and Halle state: Counterexamples to a grammatical rule are only of interest if they lead to the construction of a new grammar of even greater generality or if they show some that underlying principle is fallacious or misformulated. Otherwise a citation of counterexamples is beside the point. (Chomsky and Halle 1968: ix) Katz and Postal seem to exclude language specific facts from the scope of explanatory theory; In short, one must distinguish between those features of a language that it has by virtue of being English, French, Chinese, etc., as opposed to one of the others from those features it has by virtue of being a natural language. A full specification of the latter is a theory of structure of natural language and the features are the universais of language. (1964: 160)

430

JOSEPH GREENBERG

This excludes an enormous mass of language specific facts from explanatory theory, yet typological and diachronic factors will explain a substantial portion of them. The efficacy of diachronic explanation is not limited to language specific irregularities. Whole clusters of synchronic cross-linguistic generalizations, of which those of marking theory are perhaps the most conspicuous, can sometimes be deduced from diachronic, often in conjunc­ tion with synchronic premises. Let us consider the example of vowel nasality. The following synchronic generalization may be noted: (1) There are languages without nasal vowels but there are no languages without oral vowels. (2) The text frequency of nasal vowels is smaller than that of oral vowels. (3) The number of nasal vowel phonemes is never greater than the number of oral vowel phonemes. (4) In conditions of external neutralization of oral and nasal vowels, only oral vowels appear. By external neutralization is meant that the conditions do not involve adjacent nasal consonants. We can deduce these generalizations from a set of diachronic principles. 1. Nasal vowels arise from oral vowels by conditioned change in nasal consonant environment. Since nasal vowels arise from oral vowels under restricted conditions, in the remaining environments the oral vowel remains. From this we deduce the first synchronic universal. We also require a further diachronic principle that vowels never completely disappear from a lan­ guage. If such a language existed — and none is known — the regularity would be reduced to the implication that whenever a language has oral vowels it has nasal vowels. A further consequence of the origin theory of nasal vowels is that since the conditioning environments are restricted and less extensive than the nonconditioning environments, the text frequency of nasal vowels is smaller (our second synchronic universal). Strictly speaking we once more need an auxiliary hypothesis, in this instance synchronic, about the relative frequency of nasal consonant environments compared to all others. Of course the conditioning consonant may be only the following or only the preceding nasal consonants or only affect certain vowels making the conditioning environments even smaller. 2. Merger of oral vowels implies the previous merger of the corresponding nasal. Since nasal vowels arise from oral vowels they are equal or even smaller in number than their oral counterparts when they arise. By the diachronic principle just mentioned they have a greater tendency to merge and thus become fewer in number. This explains the third synchronic gen­ eralization regarding the equal or greater number of oral vowel phonemes in comparison with nasal vowel phonemes. 3. In externally conditioned merger e.g. in word final, it is rather that the nasal vowel loses nasality than that the oral vowel acquires it. This explains

TWO APPROACHES TO LANGUAGE UNIVERSALS

431

the fourth synchronic universal concerning neutralization. In the environ­ ment of generalization, only the unmarked oral vowel appears since the nasal vowel has merged with it. Analogues of the three diachronic principles just enunciated hold in general for marked features in phonology. They thus can count for a large number of observed regularities which derive from synchronic marking theory. Another aspect of diachronic processual explanations is that highly lan­ guage specific, often unique rules, become understandable as part of the mechanism of change of synchronic type. An example from numeral noun order is the following. Since I first encountered it in Margi, a Chadic language of Nigeria, it may be called the Margi rule. In Margi the numeral follows the noun as does the adjective. However, there is a worldwide favoring of the order numeral preceding noun. Even Welsh, a strong VSO language with NA order has the numeral preceding the noun (hereafter QN). Whenever, contrary to this the numeral follows the noun, with very few exceptions NA order also holds so that we have the universal statistical implication NQ → NA. We may therefore hypothesize that NQ order arises from QN on the analogy of NA order in languages which have the latter. Although as we have seen, in Margi the numeral follows the noun, as does the adjective, within the numeral construction this only holds for higher units. Thus one says "thousands two", "hundreds two" but for multiples of 10 the order is "two tens", "three tens" etc, the opposite of that with nouns. As idiosyncratic as the rule may seem, an identical one occurs in Kabardian, a Northwest Caucasian language which, like Margi, has NA order. Similar phenomena, though differing in details appear in certain Nilotic languages, e.g. Nandi and Masai. In all these cases the languages are NA and NQ. Comparison within all of the genetic groups allow us to interpret QN order as a survival in the most unmarked lower numeral of the earlier order under the influence of NA order. 3 The point here is that we can often make sense out of highly idiosyncratic phenomena, sometimes found in only a single language, as part of the mechanism of changes of type when we examine synchronic typologies from a dynamic point of view. Of course all explanations call for still higher level and more comprehen­ sive explanations. In our search for these we must, I believe, go outside of linguistics proper to psychology, the physiology of sound production, the nature of the world and of the speech situation (pragmatics). This question has been raised in recent years in terms of a distinction between "strong universais" that is, those explainable on purely linguistic grounds and "weak" universais, then involving non-linguistic factors. I consider the

432

JOSEPH GREENBERG

question open but there must be, I believe, at least some weak universais. This is particularly obvious in phonology. For example, even in a work on generative phonology, the author notes (Schane 1983: 121) that "in the end, the true explanations for naturalness are likely to be extralinguistic, physio­ logical and/or psychological". An example of a universal that may be plausibly imputed to psychological processing relates once more to numeral system. There is a strong tendency in languages in higher numerals to put the higher units before the lower. For example, in English we say "three thousand four hundred and sixtyfive". Up to twenty however, lower precedes higher e.g. eighteen = 8 + 10 but after twenty the higher precedes, e.g. "twenty-one". Let us call this a switch point. There is never more than one switch point and as we ascend in the numeral system the switch is always from lower following higher to higher following lower. In many languages the higher units always precede e.g. Mandarin. I am only aware of one language with the opposite order and no switch point, Malagasy, and Keenan (p.c.) informs me that Malagasy speakers encounter processing "difficulties" with higher units. The advantage of higher units preceding lower is obvious. If I start with "three thousand" I immediately know that the number is between 3000 and 4000 and the next expression "four hundred" further determines the limits as between 3400 and 3500, etc. If smaller units are placed first and if I begin with five all I can see is that the number is 5 or more and I approximate the ultimate value much more slowly. I have sketched successive levels of explanation from single language synchronic generalizations and specific diachronic explanations to synch­ ronic typological, diachronic processual and, ultimately, extra-linguistic explanations. Of course, an enormous amount of concrete work remains to be accomplished, but at least a start has been made within the last twenty years. I have concentrated on my own approach with which I am naturally enough better acquainted with than with those of others. I have also sought to be constructive. In comparing the generative approach to the one I have just described there is a further fundamental difference which flows from what I have said. To generativists it would appear that the topic of language univer­ sais is simply a synonym for the general enterprise of constructing a formal explanatory theory of grammar. As we have seen, diachronic and typological explanatory principles seem, in principle to be excluded, although this limi­ tation is often tacitly overcome in practice. Whereas the problem of universais is central to me, for more orthodox generativists there is, in a sense, no separate area of universais research. Many generativist textbooks do not even have the term "universal" in the index. In recent years Chomsky no longer talks about linguistic universais but rather of UG (universal grammar).

TWO APPROACHES TO LANGUAGE UNIVERSALS

433

A remarkably frank statement is that of a participant in the first confer­ ence on universais organized by generativists. My conscience demands that I begin by pointing out that I am writing under false pretenses. I do not purport to have any universais of semantics to propose... Accord­ ingly, let me for the bulk of the paper forget this is a conference on language universais and simply talk about semantics in relation to English. (McCawley 1968: 125) There may be at present signs of what Kuhn has called a paradigm crisis (cf. Ramat 1979). We have numerous competing theories of grammar and with apparently no generally accepted way of choosing among them. It is sometimes the case in such situations not that one is false and the rest are true but that they all unconsciously share the same assumptions about the questions to be answered. The problems of synchronic grammatical theory are no doubt important but they are not to be identified with linguistic theory as a whole. It is my hope that the broader base of method and explanation, sketched here, can with further specific investigations and theoretical exploration make a contribution towards a more comprehensive and explanatory body of linguistic theory. From all indications it is likely, however, that the two approaches sketched here will continue for a considerable period, distinct, but influencing each other. On the one hand, the often brilliant and insightful analyses of specific grammatical phenomena in the generative literature can be helpful to the typologist. On the other, the broad cross-linguistic approach can often uncover phenomena which need to be taken account of in a theory of universal grammar. Notes 1. I am indebted to Roman Jakobson for first calling to my attention the importance of implicational universais. The present article was written in 1985. No account is taken of the considerable more recent literature on this topic. 2. This requires further elaboration (cf. Greenberg 1970). There is no absolute lowest level. The level discussed here is that of citation forms in grammars and dictionary and sentences in languages which are cited in theoretical discussion, basically the level of langue rather than parole. In langue each example is a type of which tokens occur in actual discourse (at the pragmatic level). 3. For a more detailed discussion see Greenberg (1978).

References Bell, Alan. 1971. "Acquisition- and Transmission-significant Universals. Working Papers in Language Universais 5. Committee on Linguistics, Stanford University.

434

JOSEPH GREENBERG

Bugarski, Ranko. 1972. Jezik i lingvistika. Belgrade: Nolit. Butterworth, Brian, Bernard Comrie & Östen Dahl, eds. 1984. Explanations for Linguistic Universals. Berlin: Mouton. Chomsky, Noam. 1965. Aspects of the Theory of Syntax. Cambridge: M.I.T.Research Labora­ tory of Electronics. Special Technical Report no. 11. Chomsky, Noam. 1975. Reflections on Language. New York: Pantheon Books. . 1976. "On the Nature of Language." Origins and Evolution of Speech ed. by Stephen R. Hornad, D. Steklis and Jane Lancaster. Annals of the New York Academy of Sciences, vol. 280. . 1981. Lectures on Government and Binding. Dordrecht: Foris. Chomsky, Noam & Morris Halle. 1968. Sound Patterns of English. Cambridge, Mass: M.I.T. Press. Cohen, David, ed. 1974. Explaining Linguistic Phenomena. New York: Wiley. Comrie, Bernard. 1978. "Linguistics is about Languages". SLS 8, 21-36. . 1981. Language Universais and Linguistic Typology. Oxford: Basil Blackwell. . 1983. "On the Validity of Typological Studies: A reply to Smith" Australian Journal of Linguistics 3, 93-96. . 1984. "Language Universais and Linguistic Argumentation: A Reply to Coopmans." Journal of Linguistics, 20, 155-63. Coopmans, Peter. 1983. Review of Comrie (1981). JL 19, 455-57. . 1984. "Surface Word-Order and Universal Grammar". Language. 60. 55-69. Dretske, Fred I. 1974. "Explanation in Linguistics". Cohen 1947: 21-41. Ferguson, Charles. 1978. "Historical Background of Universais Research." Greenberg 1978: I, 7-31. Greenberg, Joseph H. 1963. "Some Universals of Grammar with Special Reference to the Order of Meaningful Elements." Universals of Language, ed. by Joseph H. Greenberg. Cambridge, Mass.: M.I.T. Press. . 1966. Language Universals, with Special Reference to Feature Hierarchies. The Hague: Mouton. . 1970. "The Language of Observation in Linguistics." Working Papers in Language Universais 4. Language Universais Project, Stanford University. . 1978. "Generalizations about Numeral Systems." Universais of Human Language, ed. by J. Greenberg, vol. 3, 249-296. Stanford, CA: Stanford University Press. Hawkins, John A. 1983. Word Order Universals. New York: Academic Press. . 1985. "Complementary Methods in Universal Grammar: A reply to Coopmans". Lg. 61. 569-87. Hornstein, Norbert & David Lightfoot, eds. 1981. Explanation in Linguistics: The logical problem of language acquisition. London: Longman. Jakobson, Roman. 1941. Kindersprache, Aphasie und allgemeine Lautgesetze. Uppsala: Almqvist & Wiksell. Katz, Jerrold J. & Paul H. Postal. 1964. An Integrated Theory of Linguistic Descriptions. Cambridge: M.I.T. Press. King, Robert D. 1969. Historical Linguistics and Generative Grammar. Englewood Cliffs: Prentice-Hall. Kiparsky, Paul. 1968. "Linguistic Universais and Linguistic Change". Universals in Linguistic Theory, ed. by Emmon Bach and Robert T. Harms, LeSt 14. 271-83, 171-204. New York: Holt, Rinehart & Winston. Lakoff, George Philip. 1968. Deep and Surface Grammar. Bloomington: Indiana Linguistics Club. . 1970. Irregularity in Syntax. New York: Holt, Reinhart & Winston.

TWO APPROACHES TO LANGUAGE UNIVERSALS

435

Lehmann, Winfred. 1972. "Converging Theories in Linguistics". Language. 48. 266-75. Lieb, Hans-Heinrich. 1978. "Universals and Linguistic Explanation", Universais of Human Language, ed. by Joseph H. Greenberg I 157-202. Lightfoot, David. 1981. "Explaining Syntactic Change". Hornstein & Lightfoot 1981. 209-40. McCawley, James. 1968. "The Role of Semantics in a Grammar." Universals in Linguistic Theory ed. by Emmon Bach & Robert T. Harms, 125-170. New York: Holt, Rinehart & Winston. Ramat, Paolo. 1979. "Crisi del Formalismo? Teoria della grammatica e dati empirici". LeSt 14. 271-83. . 1980. "La nozione di "tipo" e le sue articulazioni nelle discipline del linguaggio; Introdu­ zione". LeSt 15. 311-28. Riemsdijk, H. . van. 1978. A Case Study in Semantic Markedness. Lisse: Peter de Ridder. Rizzi, Luigi. 1980. "Il programma chomskiano e la tipologia linguistica". LeSt 15. 347-70. Sanders, Gerald A. 1974. "Issues of Explanation in Linguistics". Cohen 1974. 1-20. Schane, Sanford Α. 1973. Generative Phonology. Englewood Cliffs: Prentice-Hall. Smith, N. V. 1982. Review of Comrie (1981). Australian Journal of Linguistics 2. 255-61. . 1983. "A Rejoinder to Comrie." Australian Journal of Linguistics 3. 97-98. Stassen, Leon. 1985. Comparison and Universal Grammar. Oxford: Basil Blackwell.

Invariance and Variation: The Dimensional Model of Language Universals Hansjakob Seiler University of Cologne 1. Introduction This paper represents a synthesis of two earlier stages: The organizers of the Conference had first asked me to write a paper for the panel on language typology and universais. In that paper I presented the Dimensional Model by way of exemplification, dwelling largely on the dimension of determi­ nation. In the meanwhile I was asked to serve as one of the keynote speakers. In my keynote address I used a somewhat different approach. I commented on rather than repeated the earlier paper, drew on its examples, and tried to interact with what I had learned from the papers of the other participants with regard to the relation between invariance and variation. In order to preserve the essence of what was actually presented and discussed at the Conference, I shall combine both approaches in this final version. One preliminary remark: It seems revealing to extract the pertinent fea­ tures of the critical remarks that were voiced with regard to Jakobson's invariance/variation theory and to set them against the pertinent features of his own presentation. For Jakobson the invariance/variation relationship was first of all an intuition, and a very brilliant one at that. Here we find, as distinctive properties, the mode of presentation per ostensionem, the force of his illustrations and examples, the appeal to graphic representation, to con­ cepts borrowed from mathematical topology. His interest in René Thorn's catastrophe theory belongs into this context. The logic is often implicit, and has occasionally been worked out by others. An instructive case is G. Ungeheuer's logical analysis of Jakobson's feature theory (Ungeheuer 1959: 69ff) which, according to my recollection, was applauded by Jakobson. On the critic's side, we find the endeavour to be explicit, to state what the claims are and what predictions can be derived from them; to make

438

HANSJAKOB SEILER

them vulnerable for falsification; and to show exactly how invariance is related to variation. Looking at the two clusters of characteristics, I do not see any either-or alternative in them, but rather a both-and, with variation of emphasis. Explicitness without an intuitively based content is meaningless; and illus­ trations and graphical representations also contribute to the explication process. It seems that the course of science goes from intuitive and instan­ taneous understanding in an unfolding process towards greater explicitness. This is also what happened to the UNITYP model of language universais and typology. UNITYP stands for the descriptive title of the Cologne Research Group and means Language Universais Research and Typology.1 I shall now first present this dimensional model in an overview — with particular emphasis on the variants/invariant relation which is at the heart of it. After that I shall comment on what the model does with a particular dimension — that of DETERMINATION. Then, I shall conclude with a renewed appraisal of the relation between invariants and variation.

2. Overview of the model I take it to be of primary importance that one states one's goals: Why should we engage in language universais research and typology? What do we want to explain? It is a fact that, although languages differ significantly and considerably indeed, no one would deny that they have something in common: How else could they be labelled "language"? — There is obviously unity among them, no matter how vaguely felt and for what reasons: scientific, practical, moral, etc. Neither diversity nor unity is what we want to explain. We consider both as given. What we want to explain are such facts as the comparability of languages, the translatability from one language to another, the learnability of any language, language change — all of which presuppose that speakers intuitively find their way from diversity to unity, and back again to diversity — and this is a highly salient property that deserves to be brought into our consciousness. Generally then, our basic goal is to explain the way in which language-specific facts are connected with a unitarian concept of language — "die Sprache" — "le langage". The foremost notion here is that of a process, as against the conception of language as a "formal" or "abstract object" — as some transformational­ ists would have it — in short, a thing. Let us exemplify this with translation, where the processual nature is, perhaps, most immediately evident. We know the input and the output. But what goes on in between? Here is a sequence of increasingly complex models of representing the process. Model 1: A child, five years old, in a series of experiments carried out by

THE DIMENSIONAL MODEL OF LANGUAGE UNIVERSALS

439

former co-workers of J. Piaget, where children between four and nine years of age were asked to explain the role of a translator and the associated process (Sinclair et al. 1985: 50ff): What the child said, amounts to this: A translator is someone who teaches. He has to teach each of the monolingual participants of a conversation the language of the other; after that they can understand each other. Little is spelled out here about the process of conversion from one language into another. Model 2: An old Greek peasant with whom I chatted — in Modern Greek — while travelling by train from Pyrgos to Olympia. I had told him that I was Swiss, and that I was teaching in Germany. At the end he remarked: "I am glad that for the first time in my life I have managed to speak German." — A shortcut model, so to say, comparable to the one behind the miracle of Penticost: "... for everyone heard them talking in his own tongue" (Acta Apostolorum, Chapter 2). Model 3: The layman's view. The translator selects words and constructions of the target language in such a way that their meaning will match the meaning of words and constructions of the source language. This leaves the most important part of the process in the dark: Not only are the shapes of the words in languages different, their meanings don't match either. To achieve even an approximate matching, nevertheless, we need a basis of comparison, a tertium comparationis. Model 4: Some semanticists. The tertium comparationis is that which is to be expressed in language, independently of any particular language — i.e. the conceptual-cognitive content, "das Gemeinte". The translator goes from the sound sequences and meanings of the source language first to a concep­ tual-cognitive content X of the basis of comparison and from there to the sound sequences and meanings of the target language. Here too, important aspects remain in the dark: If it is true that the meanings of the forms in different languages don't match, how do we know that they can nevertheless be correlated with a common content? The postulation of a further inter­ mediate level seems unavoidable: the level of invariants. Model 5: UNITYP. This is a model not only for translation, but for universais and typology as well (Figure 1). It visualizes the ways in which language-specific facts are related to the invariant and to the universal. L1, L 2 , L 3 ... symbolize the data of different individual languages, whereas L stands for the unitarian concept of language. The reconstruction of this relationship is carried out under two different aspects, one deductive, the other inductive. Under the deductive aspect we posit cognitive-conceptual entities as tertia comparationis. Every grammarian does that. For example, when he assembles and interprets language data pertaining to determination, i.e. a relation between a determinans and a determinatum, he must have some idea of what this relationship is about —

440

HANSJAKOB SEILER L1,L2,L3...

L

CONCEPTS

diversity

unity

tertium comparationis

language structures (solutions)

continua (interpreted as programs)

repraesentandum (problem)

variants

invariant

universal

function Figure 1.

cognitively and conceptually. This means that he presupposes a concept of determination. Inasmuch as he applies such a concept to the study of any language, it is precisely the concept that may be said to have a truly universal status. The concepts should not be confused with the meanings of particular linguistic structures. The latter, as we know, differ from one language to another, no matter how much they may have in common in particular instances. But difference and sameness must be judged on the basis of one common ground — the tertium comparationis. The inductive aspect of our research concerns the ordering of data assembled under a common concept. Here, we make generalizations regard­ ing their form and their meaning, and we try to bring them into an order according to sameness and difference. The construct of a continuum with the notions of negatively correlated gradiences is our most important tool in this task. Once the continuum is established, we can then extract a common functional denominator representing the invariant, while the pos­ itions on the continuum are the corresponding variants. Thus, the invariant has an epistemological status which is different from that of universal: The latter belongs to deduction and apriorism, the former to induction. But the invariant is the authority that avails itself to be directly compared with the universal. Roman Jakobson has reminded us — both in his writings and in his teachings — that language data must be judged and analyzed from the point of view of their function, and he has proposed the means-end model of language and a theory of language functions. The idea of goal-directedness, of teleonomy, which he shared with such biologists as François Jacob and Ernst Mayr, is also at the heart of the UNITYP model. The chart in Figure 1 symbolizes a goal-directed process. The CONCEPTS are not only the tertium comparationis but also the repraesentandum or experimendum, i.e. that which is to be represented or expressed by means

THE DIMENSIONAL MODEL OF LANGUAGE UNIVERSALS

441

of language. This representation is not a matter of course but a constant problem to be solved by the speaker and listener. The initial stage of the problem solving process is the repraesentanda, the final stage, the output, the result, are the various linguistic structures in the different languages — in the case of our example: the structures pertaining to determination. Now, as pointed out in the introduction, the notion of function as con­ ceived of in this model, is of a Janus-like character: Under the aprioristic, the deductive view it represents the purpose to be fulfilled, or the problem to be solved — while the diversified structures which we find in the diversified languages represent the corresponding solutions. Under the inductive aspect it represents the invariance/variation complex as related to a common cogni­ tive concept. Our major task then consists in showing how these variants relate to an invariant, and how such an invariant matches the presupposed conceptualization. The principal claims which we make consist in saying (1) that the data pertaining to one particular conceptual repraesentandum can be ordered in continua according to two complementary functional prin­ ciples, (2) that these constitute the invariant, (3) that the invariant matches the universal, and (4) that continua represent the programs followed by the language users in finding their way from diversity to unity. The UNITYP model is significantly distinct from most of the current language universais research in that we seek the invariant and the universal — that which all languages have in common — not in a certain selected property or set of properties present in all languages, but rather in the path­ ways that speakers and listeners follow, in the programs. In simple words: The speaker/listener of all languages use ultimately one and the same pro­ gram. If we succeed in bringing this to consciousness, it would have farreaching consequences for such disciplines as computer science, etc.

3. The dimension of

DETERMINATION

This dimension belongs to our earlier research, and some of the published statements need further elaboration. I have nevertheless chosen it because many linguists have at one time or another dealt with this same range of facts. At the beginning is the insight that there is a presupposed concept — however vague — of a relation between de terminans and determinatum. There is an intuition that this may have something to do with the purpose of narrowing down the range of intended objects. "Red apples" are certain determined apples, and "these apples" are certain determined apples too. We have a feeling both of sameness and of difference: Apparently I can achieve the narrowing down by certain characteristics, or simply by pointing.

442

HANSJAKOB SEILER

In the languages this is reflected by both sameness and differences, and in linguistic terminology this state of affairs is also reflected: Determination can be taken in a wider sense — as is mostly the case in the European tradition — where it covers any kind of modification of a nominal; or it can be taken in a narrower sense — mostly in the American tradition — where it refers to the identification of reference with predominantly gram­ matical means (by a determiner such as an article or a demonstrative). In our dimensional model we consider the maximum range in accordance with the unitary concept from which we started, and we find that the regularities of the continuum are apt to account both for samenesses and for differences. In fact, the borderline between determination in the wider and in the narrower sense coincides fairly well with the location of the turning point which we postulated for several independent reasons. In my 1978 paper on determination I concentrated on the noun phrase in Modern Standard German, with some observations on interlanguage comparison, and I have broadened the scope to practically include all possible "items" that may appear within a noun phrase (Seiler 1978: 307ff). The admittedly artificial example was (1)

alle diese meine erwähnten zehn schönen roten hölzernen Kugeln die Karls auf dem Tisch, die ich dir jetzt gebe, ein Geschenk "all these my afore-mentioned ten pretty red wooden balls of the Charles' on the table, which I am now giving to you, a gift"

Note that for determining the relative orderings it is, of course, not necessary that all "items" appear together in the same NP. Note furthermore that only nested constructions were considered: If we symbolize and number the prenominai (or the postnominal) "items" starting from the head noun (HN) Kugeln "balls" as D 1 , D 2 , . . . etc. then D 1 pertains to HN, D 2 to the complex of HN + D 1 , D 3 to the complex of HN + D 1 + D 2 etc. The normal order as it appears in (1) I have charted in Figure 2. 2 On the basis of experimenting with varying orderings and by taking markedness into account I have proposed the following two regularities (Rl having two parts): R1

R1

(i)

The range of head nouns for which a "determiner" D is potentially applicable increases with the positional distance of that "deter­ miner" from the head noun HN. (ii) The potential of a "determiner" D for singling out the object referred to by HN increases proportionally with the positional distance of D from HN.

THE DIMENSIONAL MODEL OF LANGUAGE UNIVERSALS

R2

443

"Determiners" indicate properties pertaining to the concept rep­ resented by the head noun. The degree of naturalness of such pertaining decreases proportionally with the distance of D from HN.

For a detailed justification of the gradient, continuous character of R1 and R 2 the reader must be referred to my earlier publication (Seiler 1978: 307ff). The peculiarity of this dimension that makes it appear somewhat different from the other dimensions which we studied consists in the fact that the determiner classes corresponding to the different positions may appear in praesentia, i. e. in one and the same syntagm along with one head noun or determinatum. The ordering of the data pertaining to DETERMINATION is thus iconically reflected in the sequential ordering of determiners with regard to a given head noun. As the chart indicates, R1 corresponds to the principle of extension. This principle narrows down the reference of the object denoted by HN. It does so by pointing or deixis, where the category of demonstratives is the prototypical instance. R 2 corresponds to the principle of intension. It narrows down the content denoted by HN. It does so by stating its qualities, thus by predicating. These regularities and correlated principles were first extracted from word order constraints in the pre-nominal domain. The postnominal "items" do show the same principles, although their manifestation is less by means of an ordering with regard to HN than by morphosyntactic evidence within each of these categories: It is the contrast that appears, e.g. within relative clauses, between restrictive and non-restrictive ones. This is symbolized by the two converse vertical arrows, in contradistinction to the two horizontal ones in the pre-nominal domain, which reflect word order regularities. To the extent that it can be shown that all "items" in the schema

444

HANSJAKOB SEILER

participate in the same two converse regularities, we may be entitled to subsume them under a common term: Determination. The entire domain where the two converse principles corresponding to R1 and R 2 hold, we call a dimension: the dimension of DETERMINATION. In a first approximation we may define nominal determination as follows: (Def)

Determination is a relation between a determinatum and one or more determinantia. The relation serves the purpose of narrowing down the range of possible designations of objects. The narrowing down is effected according to two converse principles: that of indicativity (extension, reference) and that of predicativity (intension, properties). Both principles are co-present in every relevant struc­ ture, but with varying degrees of dominance. The structures rep­ resent different options of narrowing down the range of objects and can be ordered in a continuum, called a dimension, according to the two inversely correlated gradients of dominance.

Note that nothing is specified about any particular means of expression. In languages like English and German, word order is a very conspicuous device, but it is neither the only available means — especially considering the post-nominal sector of the continuum — nor is it the device preferred by all languages. Note furthermore that nothing is specified about any particular categories involved. Of course, we will need categories for our descriptions; but categories are subject to variation and must be understood within the framework of the continuum, not inversely: the continuum within the framework of categorization. Categorization and continuity do not exclude, but rather condition each other in our description. Figure 2 shows a major incision point constituted by the position of the HN. Post-nominal structures differ from pre-nominal ones by the fact that they all contain an additional N and are thus more complex. This in turn may lead to the hypothesis that they may assume other functions beside that of determination. In the pre-nominal sector of the continuum I have located a turning point indicated as TP on the chart. This is a "catastrophe point" (see R. Thorn 1978), where several structural properties change. Note that this is not in contradiction with the continuity of the two gradients, which goes on before and after the TP. The differences are mainly these: 1. Determiners to the right of TP admit positional variation, whereas those to the left cannot be permuted. 2. Positions to the right of TP may be relativized, i.e. they may be transformed into a predication, whereas this is not possible with positions to the left. Intensionality and extensionality are the particular manifestations in this dimension of a still more general complementarity, viz. that between indica-

THE DIMENSIONAL MODEL OF LANGUAGE UNIVERSALS

445

tivity vs. predicativity. Thus we can say that the relation of DETERMINATION is linguistically represented either by pointing it out, deictically — or by qualifying and characterizing it — predicatively. Accordingly, deictic deter­ miners such as demonstrative and possessive pronouns are at one end of the continuum, and relative clauses and appositions at the other. Now, a principal claim is that the intermediate positions show gradience in inverse proportion of the two functional principles: An increase in extensionality/ indicativity is correlated with a decrease in intensionality/predicativity, and vice-versa. The gist of the argument is that both functional principles are co-present in all structures pertaining to DETERMINATION: This superposition is what constitutes the invariant The corresponding variation is constrained by the regularities R1 and R 2 . The regularities studied thus far are manifested by word order. For English they are almost the same as for German. For many languages we find similar, or analogous orderings, as has been shown in a number of studies that were prompted by my earlier publication on the subject. However, basically the same variation on the theme of determination with the same invariant and basically the same categorial positions involved can be represented by linguistic means which, at first sight, have nothing to do with word order. This is the lesson which we learn from the so-called bondedness hierarchy which W.A. Foley established for the Philippine languages (Foley 1980: 171ff). Here, the appearence of a ligature or linking element between a determinans and a determinatum is distributed among these lan­ guages in such a way that with relative clauses the linker is obligatory everywhere, whereas with articles and deictics it is obligatory only in Tagalog and Palauan; and the distribution with regard to the intermediate determiner classes clearly shows gradience. As Foley rightly pointed out, the underlying principle is that of bondedness, i.e. of the varying strength of the tie or cohesion between the different kinds of determiners and the head noun: The cohesion is strongest with predominantly referential/extensional determiners such as ART and DEICT; it is so to speak inherent in them. This is why they can easily do without a special linking element. Now this same argument can be used to explain why, in the word order type of representing the relation of determination, referential/extensional determiners are farthest removed from the head noun in the great majority of languages: the cohesion is strong enough. On the contrary, with the predicative/intensional determiners the cohesion is weaker. This is why they are closest to the head noun, with which they form a semantic unity. Actually this is a case of Behagel's law. A further lesson which we learn from widening our perspective to interlanguage comparison is that the two converse gradients are concerned with different operational subsystems of language behavior: Intensionality/predica­ tivity is represented by semanto-syntactic phenomena. Extensionality has to

446

HANSJAKOB SEILER

do with reference to the speaker and the speech act and related phenomena that can be subsumed under the term of pragmatics. The gradients point out directionalities and a dynamism of two forces "pulling" in opposite directions. This in turn has the far-reaching implication that the continua are the locus of language change. Thus, we observe that whenever a language develops a definite article, it nearly always originates from demonstrative pronouns (Romance, Germanic, Greek); and on the continuum of the dimension the two are in adjacent positions. This is a shift to greater predicativity and lesser indicativity. On the other hand, indefinite articles tend to develop from the numeral "one"; and again, numeral and article are found in adjacent positions on the dimension. This, then, is a shift to greater indicativity and lesser predicativity. For synchronic-systematic reasons we claimed that the article is at the turning point of the two gradients. This is supported by the diachronic fact that articles are generally an unstable category that varies a great deal both in its formation and in its use. We are still a long way from having established the interlanguage dimen­ sion of DETERMINATION that would cover all languages. Ideally this would have to be done by superposition of the intralanguage dimensions — and it has in fact been achieved for other dimensions such as APPREHENSION (the representation of things) (Seiler, Lehmann, eds. 1982; Seiler, Stachowiak, eds. 1982; Seiler 1986), POSSESSION (Seiler 1983) and PARTICIPATION (the relation between a process or state and its participants) (Seiler and Premper, forthcoming). As with all these, the dimension of DETERMINATION would ultimately assume the following graphic representation (Figure 3). Here it is important to keep in mind that the fillers of the positions are not separate categories such as "the demonstrative", "the adjective", but continua again, thus subcontinua, or rather: clusters of subcontinua. Within these, a particular morphosyntactic category acts as the prototypical instance of that subcontinuum, such as the demonstrative pronoun for the subcon-

THE DIMENSIONAL MODEL OF LANGUAGE UNIVERSALS

447

tinuum of DEIXIS. The different kinds of adjectives found in the serialisation rules of German would also have to be relegated into a subcontinuum; and likewise the variation in nominal vs. pronominal inflection of the pronominal adjectives — a subcontinuum that was claimed for Sanskrit (Kölver & Kölver 1980). As the latter case quite clearly shows, the same two comp­ lementary principles valid throughout the dimension are also valid in the subcontinuum.

4. Invariance and variation Let us now return again to this basic problem. In the quest for the essence of this relationship the focus has mainly been on a thing-like entity such as a particular word, or category with its various meanings. The UNITYP model, in contradistinction, claims to subsume linguistic structures differing widely, both in form and in meaning, as variants of one and the same invariant. This necessitates a new orientation and a shift of focus from the basically static view on the particular morpho-syntactic category or relation to the dynamic aspects of the interrelation of morpho-syntactically different structures in a model of goal-oriented language behavior. Dynamics in synchrony was one of Roman Jakobson's most cherished ideas, and so was its corollary, the convertibility of synchronic and diachronic dynamics. The price which we have to pay for this widening of our scope from categorial to operational invariance consists in greater complexity of the apparatus of criteria for making our claims accessible to testing and to eventual falsification. It will no longer suffice to use a categorial base as criteria and ask whether, given such and such initial conditions, a particular category or syntactic relation appears or does not appear in the data. Instead, we need a set of criteria that can be shown to cluster under a common denominator for each of the two complementary functional prin­ ciples. For all the dimensions which we studied, it turned out that indicativity vs. predicativity are the two basic complementary functional principles, and it can now be hypothesized that these are the constituents of invariance in language throughout. The clustering of criteria under each principle can be visualized as in Figure 4 (Seiler 1986: 3ff). What these criteria establish are two converse rankings — not taxonomic rankings based on purely quantitative considerations, but rather qualitative ones that bear resemblance to a preference theory. Instead of norms, the continua reflect options to be chosen by the language user; and instead of gradience alone, the continuum reflects implicational hierarchy; and instead of one hierarchy in a gradient leading from somewhere to nowhere, it reflects two converse hierarchies. It appears that the price of increased complexity

448

HANSJAKOB SEILER Indicativity more cohesion more grammaticalized less regular less information lexical more open to pragmatics instant recognition

Predicativity less cohesion less grammaticalized more regular more information syntactic less open to pragmatics step-by-step recognition Turning point neutralization instability catastrophe

Figure 4.

of the model is worth paying, since we get powerful constraints and an explanatory potential that is promising. The model includes a basic premise which I explicitly stated at the begin­ ning and which I should like to recall at the end of this article: Cognitiveconceptual entities represent the initial stage of the problem-solving process that is at the basis of language behavior. Knowledge of these concepts is assumed and presupposed, albeit very imperfect and vague knowledge. On the other hand, the invariants which we reconstruct give us a clearer insight into the feature composition of the concept. As for determination, we learn that it is a relation where the determinans narrows down the range of the determinatum extensionally and intensionally with varying dominance. Clearly, there is a certain amount of circularity involved. While reading some of the papers I had the impression that circles are considered to be vicious per se. I think, on the contrary, that a circle is a model of great aesthetic beauty, and, above all, a model for a kind of movement, not a rigid thing. And, when seen under the movement aspect, it appears that you can enter a circle. In literary science it is called a hermeneutic circle. I think it is valid for the study of language as well. In the course of our research we never start from absolute zero. Rather, we make premises and on this basis uncover regularities which in nuce are already contained in the prem­ ises. This has essentially to do with the fact that language is at the same time the object and the medium of our description, and that man is both the subject and the object in his quest for the essence of language.

Notes 1. The Group has its center at the Institut für Sprachwissenschaft of the University of Cologne. It is funded by the Deutsche Forschungsgemeinschaft, which is hereby gratefully

THE DIMENSIONAL MODEL OF LANGUAGE UNIVERSALS

2.

449

acknowledged. A bibliography of its published work till 1983 can be found in Seiler, Brettschneider eds. 1985. See the more up-to-date version in Seiler (1985: 435-448).

References Foley, William A. 1980. "Toward a Universal Typology of the Noun phrase". Studies in Language 4. 171-199. Kölver, Bernhard & Ulrike Kölver. 1980. "Referenz und Charakterisierung: Zur Flexion altindischer Pronominaladjektive". In: Brettschneider, G. & Ch. Lehmann (eds.) Wege zur Universalienforschung. Sprachwissenschaftliche Beiträge zum 60. Geburtstag von Hansjakob Seiler. ( = Tübinger Beiträge zur Linguistik, 145.) Tübingen: Gunter Narr Verlag. pp. 392-405. Seiler, Hansjakob. 1978. "Determination: A Functional Dimension for Interlanguage Compari­ son". In: Seiler, H. (ed.) Language Universals. Papers from the conference held at Gummersbach/Cologne, Germany, October 3-8, 1976. ( = Tübinger Beiträge zur Linguistik, 111.) Tübingen: Gunter Narr Verlag. pp. 301-328. . 1983. POSSESSION as an Operational Dimension of Language. ( = Language Universals Series, 2.) Tübingen: Gunter Narr Verlag. . 1985. "Kategorien als fokale Instanzen von Kontinua: gezeigt am Beispiel der nominalen Determination". In: Schlerath, Bernfried & Veronica Rittner (eds.) Grammatische Kategorien. Funktion und Geschichte. Akten der VII. Fachtagung der Indogermanischen Gesellschaft, Berlin, 20-25. Februar 1983. Wiesbaden: Dr. Ludwig Reichert Verlag. pp. 435-448. . 1986. APPREHENSION: Language, Object and Order. Part III.: The Universal Dimension of APPREHENSION. ( = Language Universais Series, l/III.) Tübingen: Gunter Narr Verlag. Seiler, Hansjakob & Christian Lehmann (eds) 1982. Apprehension: Das sprachliche Erfassen von Gegenständen. Teil I.: Bereich und Ordnung der Phänomene. ( = Language Universals Series, 1/I.) Tübingen: Gunter Narr Verlag. Seiler, Hansjakob & Franz-Josef Stachowiak (eds) 1982. Apprehension: Das sprachliche Erfassen von Gegenständen. Teil IL: Die Techniken und ihr Zusammenhang in Einzelsprachen. ( = Language Universals Series, 1/II.) Tübingen: Gunter Narr Verlag. Seiler, Hansjakob & Gunter Brettschneider (eds) 1985. Language Invariants and Mental Oper­ ations. Proceedings of the International Interdisciplinary Conference held at Gummersbach/ Cologne, Germany, September 18-23, 1983. ( = Language Universais Series, 5.) Tübingen: Gunter Narr Verlag. Seiler, Hansjakob & Waldfried Premper (eds). Forthcoming. Participation as a Universal Dimension of Language. ( = Language Universais Series, 6). Tübingen: Narr. Sinclair, Hermine et al. 1985. "Constructivisme et psycholinguistique génétique". Archives de Psychologie 53. 37-60. Thom, René. 1978. "La double dimension de la grammaire universelle". In: Seiler, H. (ed.) Language Universals. Papers from the Conference held at Gummersbach/Cologne, Germany, October 3-8, 1976. ( = Tübinger Beiträge zur Linguistik, 111.) Tübingen: Gunter Narr Verlag. 79-87. Ungeheuer, Gerold. 1959. "Das logistische Fundament binärer Phonemklassifikation". Studia Linguistica 13. 69-97.

Classical and Modern Universals Research Their Philosophical Background Elmar Holenstein* ΕΤΗ Zurich Abstract Universals research of the last twenty-five years has as its precursors the Cercle linguistique de Prague of the interwar period (in particular, Roman Jakobson and Nikolaj Trubetzkoy). Compared with the classical doctrines of universais from Aristotle through the grammarians of the Middle Ages and the rationalists of the 17th and 18th centuries to Husserl at the turn of this century, the renewed interest in universais in present-day linguistics may be characterized by five shifts: (1) In the traditional doctrines only categories of meaning were regarded as universal, but not categories of expression. This Platonist perspective of the tradition is no longer shared. On the contrary, recent universais research started with the exploration of interlingual constraints in the domain of the categories of expression. (2) A more empiricist attitude led in the first centuries of the modern era to the distinction between "nature" and "culture". The subsequent corre­ lation of "nature" and "universal" and of "culture" and "variable", too, is no longer maintained in recent universais research. (3) The traditional doctrines look for a logical, a priori foundation of universais. In recent research biological, psychological and functional explanations prevail. (4) In the tradition, manifest in particular in American anthropology of the first half of this century, only the essential and fundamental traits of language were regarded as universais, "the detail", however, as variable and changing. Recent universais research has uncovered invariant con­ straints in the detail of human languages as well. (5) The traditional doctrines were concerned almost exclusively with absolute elements and categories. In the renewed universais research relational (mainly implicational) universais prevail. American linguists usually begin surveys of modern universais research with a short paper by Burt W. and Ethel G. Aginsky in the journal Word (1948;

452

ELMAR HOLENSTEIN

cf. Ferguson 1978). The Aginskys' paper is a milestone as regards the history of terminology. It seems to be the first text in which the term "language universais" was used. As for the rest, it deals with universais based on crosscultural diffusion, a kind of universal that plays no part in the linguistic universais research that, starting in the sixties, has won renewed prominence — despite the fact that it is undeniably a matter of topical interest: for the first time since the origins of mankind it is again possible for cultural phenomena to spread to all human societies by means of contact. According to information from André Martinet, the editor of Word at the time, the Aginskys' paper was solicited by Roman Jakobson: The Aginskys' paper, which was the first to mention linguistic universais, had been ordered by Jakobson. It took some time for the seed to bear fruit, but, ever since, what has been produced is in the same vein. From the moment that everybody began to speak about universais I saw the shadow of Jakobson. They all yielded: even Hockett contributed to Greenberg's book (1963). Greenberg himself, who often opposed Jakobson when he and I were co-editors of Word, also swallowed the bait. (Martinet 1974: 227) In view of Jakobson's importance for the revival of linguistic universais research, as attested by a rival of some decades' standing, the question as to how he used to give a historical introduction to the topic of universais is of interest. For Jakobson, modern universais research, continuing the tradition of the medieval grammatica speculativa and the "idea, conceived by seventeenth and eighteenth century rationalism, of a universal grammar", begins with the two philosophers Edmund Husserl (1859-1938) and Anton Marty (1847-1914). The reference to 17th and 18th century rationalism is taken from Husserl's Logical Investigations (1901/1913: A 318/B1 336). The reference to the doctrines of universais of early modern philosophy is com­ mon practice. In Marty's Untersuchungen zur Grundlegung der allgemeinen Grammatik und Sprachphilosophie (1908: 69), it can be found together with the contentious specification that these were "disciples of Cartesius [Descartes]" — a claim that Noam Chomsky has since brought to high standing. Marty, being well versed in history, adduces particulars that go farther back. Before the time of the "disciples of Cartesius" he mentions the Scholastics and before them the Stoics, and above all Aristotle, whose work Perì hermēneías he cites as "the first and indeed as a very estimable contri­ bution to general grammar". A thesis formulated at the beginning of Aristotle's work On Interpretation subsequently came to be decisive for the direction of the whole tradition up to Husserl. The first to break with it without at the same time giving up the assumption of universais was Marty.

CLASSICAL AND MODERN UNIVERSALS RESEARCH

453

Aristotle (16a) states: As writing, so also is speech not the same for all men. But the mental affections themselves, of which those words are primarily signs, are the same for the whole of mankind, as are also the objects of which those affections are representations. The pattern of thought is Platonic. Plato's Kratylos (389e-390a) has it that a language maker, "whether he be here or in a foreign land, so long as he gives to each thing the proper form {eidos) of the name, in whatsoever syllables, is no worse lawgiver, whether here or anywhere else". It is as when a smith makes a tool for a certain purpose: "So long as he reproduces the same idéa, though it be in different iron, still the instrument is as it should be, whether it be made here or in foreign lands". Indeed, only the manner of expressing this "ideology" changed in the doctrines of the following centuries. According to Boethius of Dacia, writing in the thirteenth century, the substantia sermonis is universal, but not the articulatio vocis: according to Leibniz (1765: § 3.4.17) in the eighteenth, les idées but not les mots, according to Husserl (1901/1913: § 4.14) at the beginning of this century, the Bedeutungskategorien ("categories of mean­ ing") but not their Ausdruck ("categories of expression": Ausdruckskategorien). The pattern of thought is not only Platonic. It can be found in more than just one tradition, and not only in respect to language. A Japanese formula says ikei dōshitsubutsu "thing of the same quality in difference forms". The watchword una religio in rituum varietate is familiar from Catholic theology. In ethics the assumption can be found that moral principles are the same everywhere, only their "application" changes from society to society. All are in considerable agreement on the notion of virtue in general, although they diverge in the application. (Leibniz 1965: § 1.2.18, cf. 2.18.10) The application can sometimes be as extremely divergent as a case described by Herodotos in a legendary report (3.38): Reverence before dead parents is encountered everywhere. Whereas, to the horror of the Callatians of India, the Greeks express this reverence by burning their fathers, for the Callatians, to the horror of the Greeks, it ostensibly consists in eating their dead fathers. Similarly, it is conceded that emotions may be the same for all people, but that, if anything at all, the expression of such subjective matters as feelings will vary from culture to culture. In stark contrast to these traditional conceptions, modern linguistic uni­ versais research took the categories of expression as its first field of enquiry: Roman Jakobson on the level of phonology, demonstrating that there is a universal set of sense-discriminative sound features as well as of laws govern­ ing their arrangement to form the phonological system of a language; and

454

ELMAR HOLENSTEIN

Joseph H. Greenberg on the level of word order and morphology. In the two non-linguistic fields mentioned, ethics and emotions, there is also no lack of discoveries of universal forms of expression; in the field of social behaviour, for example, as regards the relationship between mother and baby (cf. Lumsden & Wilson 1981: 79ff.); and in the field of the expression of emotions in respect to the primary emotions. The facial expressions that in one culture are judged to be prototypical for primary emotional states (happiness, anger, fear, surprise, disgust and sadness) are interpreted as such by members of other cultures. The nerve stimulations upon which the most important emotions are based seem to be immediately connected to the corresponding facial muscles (Ekman 1971). But back to history again. In the course of the modern era in Europe, especially since the beginning of the nineteenth century and then on a broader basis in the first half of this century, there was a withdrawal from the classical conception of universais. The border between invariant and variable was no longer drawn between the ideas people made of things and their expression, but rather, in Kant's language (1798: 119), between "what nature makes of man" and what man makes of nature, what "he makes or can and should make of himself as a free being". The ideas one makes of the things turned out not to be independent of their linguistic expression. Thus, it was still assumed that one can in principle speak about everything in every language, but that, depending on the language, one forms different ideas of the things that are likewise universally given. "The illusion of an absolutely 'general' grammar" based on universal categories of meaning was superseded by the program of "the specific stylistics of each individual language" (Cassirer 1923: 83). The insight that cognitive categories are dependent on linguistic categories was a step forward, whereas an associated assumption was too rash, namely that what has usually been called innere Sprachform ("the inner form of language") since Humboldt and what Cassirer meant by the "style" of a language, namely word and sentence formation, can vary from language to language "in an infinitely manifold and disparate manner". It is logically conceivable that languages differ from each other arbitrarily. But it would not be natural. And human languages are natural structures that have emerged in a manner dependent {inter alia) on the cognitive capacities of language users. The linguistic relativists thought differently about the matter and pro­ ceeded on the basis of a different assumption, namely the assumption that the world of experience is a continuum that as such can be arbitrarily segmented and whose segments can be arbitrarily combined and embraced by concepts. According to this credo, we are guided by the habits and conventions that have been deposited in our respective native languages.

CLASSICAL AND MODERN UNIVERSALS RESEARCH

455

The color spectrum was used as the outstanding paradigm of the arbitrary segmentability of the world. One color merges into the next on the spectrum, continuously and seamlessly. Furthermore, it has long been known that languages differ enormously in respect to the number of words to designate color. Some languages seem to have fewer than half a dozen color terms, others have hundreds. If, however, attention is restricted to the basic color terms, those color terms that fulfil a series of linguistic, psychological and sociological criteria, it turns out that all languages have only a limited number of such basic color terms (Berlin & Kay 1969; Kay & McDaniel 1978). Eleven is said to be the highest number. If a language does not have all eleven, it is not arbitrary which it has and which are missing. If a language has only two basic color terms, then it has one for black and one for white, that is, for light and dark. If it has three, then there is also one for red; with four, then also a word either for yellow or, as it is called, for GRUE, that is, a word covering both green and blue. Many languages have only one word for green and blue. If a language has five basic color terms, then it has one for yellow as well as one for GRUE. If it has six, then it distinguishes between green and blue. With seven there is a word for brown in addition. If it has eight to eleven basic color terms, then it has words for gray, pink, orange and purple, whereby no particular order is dominant. Gray, a differentiation of the brightness or black-white dimension, can, however, appear earlier. That means that contrary to superficial impressions the linguistic division of the color spectrum is by no means arbitrary. The segmentation is almost universally the same, dependent only on the number of color terms. If one looks for a reason, a cognitive explanation presents itself. Our neurologically conditioned perception decides on the number and order of color terms, and not the other way around. The color spectrum is only physically a continuum, as regards the light waves underlying the perception of color, not however perceptually or, as can be shown, neurologically. Certain segments of the color spectrum are occupied by pure colors, red, green, blue and yellow, others by mixed colors. Although red is situated between orange and purple just as orange is between red and yellow, we only see orange as a compound color, but not red. The pure colors are also distinguished by being more conspicuous than the others. The different, neurologically con­ ditioned conspicuousness is a prominent factor that is adduced for the nonarbitrary segmentation of the color spectrum in natural languages. If a field of objects is differently segmented in different languages, then this has an effect on the meanings of the expressions. My favorite introduc­ tory example to demonstrate this change of meaning depending on the number of words for a Wortfeld ("lexical field") comes from Switzerland. It concerns a field of objects that is indubitably a continuum in the purely physical respect, just as the color spectrum is, namely the age classes of

456

ELMAR HOLENSTEIN

people. People live to an age between zero and about one hundred years. This continuum is obviously not arbitrarily segmented either. Although the difference between ten and thirty years is as large as that between thirty and fifty, the probability is significantly greater that different words will be used to designate those of ten years and those of thirty years than those of thirty and fifty. Equally, the probability is greater that an additional age group will be linguistically marked between ten and thirty years than between thirty and fifty. If this happens, it immediately affects the meanings of the words used for those of ten and those of thirty years. The evidence is as follows. Some years ago the mayor of Zürich went by foot to Bern — to demon­ strate sportsmanship and love of nature. He and his son-in-law were accompanied by two journalists (from the Züri-Leu, a local newspaper). One marched with the mayor, the other drove ahead by car to take pictures occasionally. Having reached the Canton Bern, the photographer asked at a farm there if "three men" had recently walked past. The surprising answer was: Nai — aber e Maa met zwee Puurschte ("No — but a man with two lads" — had gone past). Depending on whether one distinguishes only between boys and men, as is frequently the case in contemporary German, or between boys, lads and men {Knaben, Burschen, Männer), as is still usual among older people, a yes or no answer will have a different force and may prove misleading. When comparing the linguistic universais currently under discussion with the universais of the traditional doctrines, two differences become apparent. Both concern the absoluteness of the traditional examples. Traditionally, universais were in general given an a priori, logical foundation, based on the sense of respective categories. On the basis of their meaning, certain linguistic expressions can be connected with other expressions to form meaningful sentences, or they can be transformed into other expressions. A noun can be connected to a verb to form a meaningful sentence ("Flowers bloom" ). An adjective ("red") can be transformed into a noun ("redness"), an epithetic construction ("a white rose") into a subordinate clause ("a rose that is white"). But not every combination and not every transformation yields a meaningful utterence. "King but or similar and" is not a meaningful sentence. In conformity to the logical foundation, absolute universality was claimed for such regularities. For the kind of universais that have been uncovered by modern universais research, on the contrary, psychological and biological explanations seem more appropriate, the nature of the human brain or mind. Like other psychological and biological explanations they are correspondingly not claimed to be valid without exception, but only to have a statistical validity of high probability. They are not "strict", but only "near-universals".

CLASSICAL AND MODERN UNIVERSALS RESEARCH

457

The only explicitly logical foundation of universais that I know of in modern linguistic universais research (aside from linguistic pragmatics) is in the first manifesto of modern empirical universais research, by Trubetzkoy: Although the phonological laws valid for all languages are found by pure empirical induction, these laws can sometimes be logically deduced: Thus, for example, the connection between the "melodic correlation" and the "quantitative correlation" can be explained by the fact that the distinction between two intonations is possible only if the beginning and the end of a vowel are perceived as two distinct moments, which in turn presupposes the notion of duration. (Trubetzkoy 1933: 343) An objection often raised against the empirical explanation of universais is that peripheral organs are involved equally in the field of visual perception, from which the examples so far adduced come, and in the field of the sounds of language, from which modern universais research started. The prestructured functioning of these peripheral organs — eyes, ears, voice-producing mechanisms — is not denied. But for more abstract cognitive operations central brain processes, whose plasticity is proverbial, must be considered. In reply to this it can be pointed out that a whole series of universais from the fields of visual perception and phonology are explained with recourse to central processing operations. This applies for example to the privileged position of the first six primary color terms vis-à-vis the following five, but not necessarily to the order of the pure colors themselves, where peripheral factors seem to play a part (cf. Ratliff 1976). But even in the field of "higher" cognitive operations there is no lack of natural constraints that entail universais. Numeral categories are a clear example of abstract mental operations. The following implicational chain holds for grammatical number categories: a trial implies a dual, a dual implies a plural and a plural a singular (Greenberg 1963: 94). There is no language with a dual but no plural. The indefinite idea of a multitude is prior to the definite idea of a duality. The plural as a multitude of units seems cognitively to presuppose the concept of unity, and the dual as a definite multitude seems to presuppose the concept of an indefinite multiplicity. As in the case of color terms, these implications are an example for the fact that in a second sense many of the newly discovered universais are not absolute universais, whereby "absolute" is here to be understood not in contradistinction to "statistically probable", but to "relational". It is not assumed that certain properties are to be encountered in (almost) all langu­ ages, but only that they occur dependent on the presence of other properties that have a material connection to them. Implicational laws of the following kind are (near-)universals: "If a language has C, then it also has B; and if it has B, then it also has A; but not necessarily the other way around." The

458

ELMAR HOLENSTEIN

relational restriction is no surprise either. It results from the dynamic, historical character of language in general, with phases of elaboration and of simplification, and within individual languages from the developmental character of acquisition by their speakers. Greenberg was able to make out over fifty generalizations concerning the lexical number categories, the numerals in natural languages. The following regularities are examples of cases in which an explanation by recourse to cognitive simplicity suggests itself almost forcibly: Of the four fundamental arithmetical operations — addition and its inverse, subtrac­ tion, and multiplication and its inverse, division — the existence of either inverse operation implies the existence of both direct operations. The existence of multiplication implies the existence of addition. When a number is expressed by subtraction ..., the subtrahend is never larger than the remainder. (Greenberg 1978: 3.249 ff.) Addition is prior to and more frequent than subtraction as a means of forming numerals. This seems to be the case not only for the words, but also for the figures. In the Roman numeral system up to three primary units are generally added to the fives and tens, and only one is subtracted (XI, XII, XIII, XVI, XVII, XVIII vs. XIV, XIX). A language that forms a composite numeral by multiplication {quatre-vingts) also possesses numerals formed by addition ( vingt-et-un), but not necessarily the other way around. It is logically possible that it not be the numeral eighteen that is formed by subtracting two from twenty, as in Latin {duodeviginti), but the other way around, the numeral two by subtracting eighteen from twenty. But again, the rule applies: not everything that is logically possible is cognitively and linguistically natural. To the extent that universal regularities are concomitant with degrees of cognitive complexity and abstractness, developmental psychological expla­ nations come to mind. If, however, highly complex cognitive phenomena such as in the case of facial recognition and highly abstract distinctions such as those that take effect with certain grammatical categories apparently occur (a) immediately and (b) without the rate of errors usual in learning processes, then recourse to a specific biological predisposition to the corre­ sponding cognitive abilities seems appropriate. The distinction between specific and nonspecific articles observed in child and Creole languages, apparently independent of a corresponding model in the substrate languages, is an impressive example. For example, the indefinite article "a" is used specifically in the sentence, "A dog woke me this morning", but nonspecifically in a sentence such as, "I would like to have a dog", that is, some dog or other, it is not yet settled which. At a certain stage Creoles and children seem obstinately to use an explicit article in the first case, and

CLASSICAL AND MODERN UNIVERSALS RESEARCH

459

in the second case the so-called zero-article, that is, none. The distinction they master is not simply the concrete distinction between two individuals, this one and that one, but the abstract distinction between something that is individuated and something that is not deictically identifiable, by pointing to a certain spatiotemporal position (cf. Bickerton 1981). The naive rationalistic assumption that relatively simple cognitions (for example, the smallest whole numbers up to five or six, with practice some­ what more) can be grasped intuitively but complex cognitions only by means of linguistic constructions, is typical of traditional philosophical discussions of the relationship between thought and language. Empiristic strategies of research are now in the foreground. Empirical experience and not some conceptual reflection decides what is to be regarded as intuitively graspable. If a specific ability (e.g. the easy recognition of such complexities as faces) cannot be understood by means of conceptual analysis, the corresponding explanation is then empirical, for the most part biological. Traditional philosophical discussions of the relationship between thought and language can in still another respect be rebuked for narrowness. It can be detected in both opposed camps, the rationalist and the empiricist. On the rationalist side, the exclusive domain of research in linguistic universais was taken to be the realm of logical rationality (necessary conditions of the possibility of language as a universal means of communication as well as consistency and sensefulness, that is, the avoidance of falsehood and absurd­ ity). On the empiricist side it was conjectured that there are only contingent universais concerning objects that are accessible and important to all. Locke (1690: § 1.3.9 & 11) lists fire, sun, heat and number. The best examples of this kind from contemporary universais research are those concerning the object closest to all, one's own body and its parts (cf. Andersen 1978). Apart from that the empiricists thought not so much of cognitive as of conative and emotional predispositions as candidates for anthropological universais, the pursuit of happiness and the aversion to misery, furthermore feelings of pleasure and displeasure accompanying certain things and ideas (cf. Locke 1690: § 1.2.3; 2.7.3 f.). The economy and ecology of expediency as a possible source of anthropo­ logical universais that should be systematically examined certainly remained outside the field of research, if not the field of vision of the classical rationalists and empiricists. In this respect Anton Marty signified a turning point — in the wake of empiricist philosophy of the end of the nineteenth century, of course; but in applying this empirical perspective to language Marty was a pioneer, although with only limited impact at first and of historical impact only after more than half a century and thanks to Roman Jakobson's influence. He saw as no other theoretician of universais before him or among his contemporaries the foundation of a large and important

460

ELMAR HOLENSTEIN

realm of linguistic universais in the "universal nature of psychic and physical capacities (psychische und physische Kräfte)" (Marty 1908: 55). The assumption had both material and methodological consequences. Marty formulated it principally in his confrontation with Husserl, a Bren­ tano disciple as he himself was. Materially it means that linguistic universais are not only to be sought in the realm of the categories of meaning, founded in the sense of these categories, but also in the realm of the categories of expression, "in respect of the form that the means of expression assume and must assume everywhere". If all endeavour to communicate is dependent on the psychic and physical capacities that to a certain extent are common to all human beings, then "legitimate expectations can presumably be drawn as to how, wherever people speak to each other, the tasks set by what is to be expressed will be fulfilled". Methodologically, the assumption mentioned means that much "that is of greatest import for general grammar can, however, only be known empirically, and not a priori" and "also not from mere insight in the categories of meaning" (Marty 1908: 58, 60). The far-reaching significance of human nature, "of the psychic and physi­ cal capacities" serving human users of language, for universal regularities become clear just by juxtaposing the categories of meaning comprised by modern universais research to the categories of meaning to which the tra­ ditional doctrines of universais restricted themselves. In the traditional doctrines there was a clear tendency to restrict oneself to categories of thought that play a part in classical logic. The favorite topics were nouns, the copula, adjectives (which turned out not to be a universal word category), functional expressions such as conjunctions and modes of judgement. In the projects for a universal language the attempt was made to reduce even such a fundamental category of natural language as the verb to an adjective-like form, the participle, or even to a noun plus copula. Cicero scribit was thus transformed into Cicero est scribens or Cicero est scriptor without regard to the modification of meaning attendant on the grammatical transformation. What is most conspicuous for a modern theoretician of language about the traditional doctrines of universais, no less than about traditional logic, is perhaps the fact that the pragmatic word category of deictica is missing, the demonstratives and personal pronouns, without which our natural lan­ guages would be exceedingly difficult to use and hardly possible to learn. The deictica are among the first grammatical universais to which the univer­ sais research that was renewed in the sixties applied itself (cf. Greenberg 1963: 21, 91, 96). Another tendency of traditional universais research was to look for univer­ sais primarily or even exclusively in the realm of abstract categories. This seems to have less to do with the a priori, logical foundation of universais

CLASSICAL AND MODERN UNIVERSALS RESEARCH

461

than with the assumption that abstraction and generalization should imply each other: the more abstract something is, the more universal it is; and the more concrete, the more variable. This assumption reached its zenith and its most explicit formulation in the anthropological literature of the first half of this century. If the universais are no longer founded a priori logically, but rather according to laws of nature (particularly of biology), then a correlation of "abstract" and "universal" or "concrete" and "particular" is unlikely. What is (near-universally) common to all mammals is not only an abstract "blueprint", which is modified almost beyond recognition in the particular subgroups, cohorts, orders, families, genera down to the species and subspecies, but also such concrete details as the fact that the lower jaw is hinged directly to the skull, instead of through a separate bone (the quadrate) as in all other vertebrates, and that a chain of three tiny bones (hammer, anvil, and stirrup) transmits sound waves across the middle ear. The received opinion in anthropology up to the fifties was that various cultures only had "empty frames" or "blanket categories" in common, so to speak "chapter headings" that permit rough classification of the manners of behavior that vary from culture to culture; there was thought to be no common "behavioristic detail" that is specific as to content. Without claim­ ing completeness or attempting a systematic classification, Murdock (1945: 124) lists 73 such empty frames in alphabetical order. For the sake of illustration the first seven are listed: "age grading, athletic sports, bodily adornment, calendar, cleanliness training, community organization, cook­ ing". Other, better known ones are: education, gestures, language, medicine, music, tool making. According to Murdock, some of these universal categor­ ies can be resolved into components, which are also universal, for example language into conventional sound units (phonemes), meaningful combi­ nations of such units (words) and rules for combining words into sentences (grammar), so that "exceedingly numerous" resemblances can be found between all cultures, which, however, only "rarely, if ever" concern "specific cultural content". It is highly doubtful whether any specific element of behavior has ever attained genuinely universal distribution. The true universais of culture ... are similarities in classification, not in content. ... What cultures are found to have in common is a uniform system of classification, not a fund of identical elements. (Murdock 1945: 125) But linguistics and ethology are now producing such universais. The collapse of the anthropological tively empty frames are universal can be dated in the in the early fifties and is connected with the names Joseph H. Greenberg. In 1953 (516 f.), Kluckhohn,

evidence for precisely dogma that only rela­ literature. It happened Roman Jakobson and referring to a "lecture

462

ELMAR HOLENSTEIN

delivered by Dr. Joseph Greenberg in April 1952 to the staff of the laboratory of Social Relations, Harvard University", cites as the first and weightiest evidence for the contrary thesis the assumption that Jakobson had published a year previously, to the effect that the sounds of natural human languages are not a random mixture but rather form a system with twelve underlying binary oppositions. This system turned out to need revision in points of detail, but it was able to maintain its paradigmatic status. If universal regularities can also be ascertained on the level of expression, pertaining to the progressive elaboration of a lexical field as well as to the extension of already existing categories to categories to which cognitive access has newly been gained (cf. Williams 1976), and pertaining to sentence formation with the introduction of "synsemantica that are logically not founded", as Marty (1908: 537) puts it; and if these universal regularities depend on the "univer­ sal nature of psychic and physical capacities" (Marty 1908: 55), the cognitive and biological structure of man; that is, if there are universal regularities also on the level of the so-called "inner form of language" in addition to those on the level of the categories of meaning and of the external (phono­ logical) form of language: then it is not only appropriate, pace Cassirer, to continue to speak of a universal grammar, but even of universal stylistics. Buffon's wise dictum, Le style est l'homme même ("the style is the man himself"), receives in this perspective a new connotation, so to speak a species-specific connotation.

Note *

Translated by Donald Goodwin.

References Aginsky, Burt W., & Ethel G. Aginsky. 1948. "The Importance of Language Universals". Word 4. 168-172. Andersen, Elaine S. 1978. "Lexical Universais of Body-Part Terminology". Greenberg 1978: 3. 335-368. Aristotle. "Perì hermēneías/On Interpretation". Aristotle in Twenty-Three Volumes I (=The Loeb Classical Library, 325.) Cambridge, MA: Harvard University Press 1973. Berlin, Brent, & Paul Kay. 1969. Basic Color Terms. Berkeley: University of California Press. Bickerton, Derek. 1981. Roots of Language. Ann Arbor: Karoma. Boethius Dacus. ca. 1270. Questiones super Priscianum maiorem ed. by Jan Pinborg & H. Roos. Kopenhagen: Gad 1969. Cassirer, Ernst. 1923. Philosophie der symbolischen Formen I. Darmstadt: Wissenschaftliche Buchgesellschaft 1964. Ekman, Paul. 1971. "Universals and Cultural Differences in Facial Expressions of Emotion".

CLASSICAL AND MODERN UNIVERSALS RESEARCH

463

Nebraska Symposium on Motivation 1971 ed. by James K. Cole, 207-283. Lincoln: University of Nebraska Press. Ferguson, Charles A. 1978. "Historical Background of Universais Research". Greenberg 1978: 1. 7-32. Greenberg, Joseph H., ed. 1963. Universals of Language. Cambridge, MA: M.I.T. Press. ed. 1978. Universals of Human Language. Stanford: Stanford University Press. 4 volumes. Herodotos. Historia. Herodotus in Four Volumes II (=The Loeb Classical Library, 118.) Cam­ bridge, MA: Harvard University Press 1971. Holenstein, Elmar. 1985a. Menschliches Selbstverständnis. Frankfurt: Suhrkamp. 1985b. Sprachliche Universalien. Bochum: Brockmeyer. Husserl, Edmund. 1901/13. Logische Untersuchungen II/1 ( = Husserliana XIX/1.) Den Haag: Nijhoff 1984. Jakobson, Roman, C.G.M. Fant & Morris Halle. 1952. Preliminaries to Speech Analysis. Cambridge, MA: M.I.T. Press. Kant, Immanuel. 1798. Anthropologie in pragmatischer Hinsicht ( = Werke VII). Berlin: de Gruyter 1968. Kay, Paul, & Chad McDaniel. 1978. "The Linguistic Significance of the Meaning of Basic Color Terms". Language 54. 610-646. Kluckhohn, Clyde. 1953. "Universal Categories of Culture". Anthropology Today, ed. by Alfred L. Kroeber, 507-523. Chicago; Chicago University Press. Kuroda, S.-Y. 1973. "Edmund Husserl, grammaire générale et raisonnée, and Anton Marty". Foundation of Language 10. 169-195. Leibniz, Gottfried Wilhelm. 1765. Nouveaux essais sur l'entendement humain. Paris: GarnierFlammarion 1966. Locke, John, 1690. An Essay Concerning Human Understanding. New York: Dover 1959. Lumsden, Charles J., & Edward O. Wilson. 1981. Genes, Mind, and Culture. Cambridge, MA: Harvard University Press. Martinet, André, 1974. "Interview". Discussing Language ed. by Herman Parret, 221-247. The Hague: Mouton. Marty, Anton. 1908. Untersuchungen zur Grundlegung der allgemeinen Grammatik und Sprach­ philosophie I. Halle: Niemeyer. Murdock, George P. 1945. "The Common Denominator of Cultures". The Science of Man ed. by Ralph Linton, 123-142. New York: Columbia University Press. Platon. "Kratylos". Plato in Twelve Volumes IV (=The Loeb Classical Library, 167.) Cam­ bridge, MA: Harvard University Press 1977. Ratliff, Floyd, 1976. "On the Psychophysiological Bases of Universal Color Terms". Proceed­ ings of the American Philosophical Society 120. 311-330. Trubetzkoy, Nikolaj. 1933. "La phonologie actuelle". Journal de psychologie 30. 227-246. Williams, Joseph M. 1976. "Synaesthetic Adjectives". Language 52. 461-478.

Language Typology and Diachronic Linguistics* T.V. Gamkrelidze Georgian Academy of Sciences, Tbilisi The problems of linguistic typology and diachrony presented in the current paper occupy a central place in the rich scientific legacy of Roman Jakobson. The problematics investigated here and the solutions proposed are entirely based upon the fundamental scientific ideas advanced by the late scholar. Their dependence on these ideas is so obvious that I have not deemed it necessary to refer to his works in each instance. After the complete supremacy of the problematics of synchronic linguistics in the first half of the twentieth century, its second half was marked by a growing interest in diachronic linguistics, in problems of language change and transformation in time. This appeared to be a certain return, but on a new methodological level, to the elaboration of problems that had arisen in classical comparative-historical Indo-European linguistics of the last cen­ tury. Such a growing interest in problems of language change and in dia­ chronic linguistics was conditioned by the general development of linguistic thought in recent decades: having overcome the Saussurian antinomy between synchronic and diachronic linguistics, it strove to construct a linguistic theory that possessed a greater explanatory power than that of the purely synchronic theories advanced by descriptive, taxonomic grammar, built solely upon empirical linguistic data. One of the basic presuppositions of the reconstruction of a proto-language and of comparative-genetic linguistics as a whole is the thesis that linguistic development should be understood not as a movement of a language from a simple to a complex or more perfect state, but as a diachronic mutability or variability, a language's capability to be transformed on all levels of linguistic structure. On the sound level such diachronic transformations of a language are expressed in changes of certain phonemes into other phonemes that essen­ tially represent either a "fission" or "fusion" of two phonemes characteristic of an earlier state of the language. Such phonemic transformations are realized in conditions of redundancy of the language system, which indeed

466

THOMAS V. GAMKRELIDZE

determine the possibility of such sound changes. The redundancy of a language as an "incomplete system" indeed appears as the structural factor that makes possible a language's sound mutability. In consequence of this a language system is not a frozen structure in respect to sound changes and the diachronic "movement" of phonemes. However, the character of such diachronic "movements" of phonemes, of such sound transformations of the system, depend less upon the degree of redundancy of the system, which varies from language to language, than upon deeper characteristics of the language's structure. One of such deep structural characteristics of a language system is the hierarchic relation of "markedness" or "dominance" among the linguistic, in particular phono­ logical, units. There exist universal models of combinability, of compatibility of phonetic distinctive features in simultaneous ("vertical") succession — in the simul­ taneous bundles that make up phonemes. Certain features combine with one another on the axis of simultaneity and are more preferable than others, which manifests itself in a greater systemic and textual frequency of the phonemes into whose composition these features enter; other features are combined in the integral bundle in a more limited way, which is manifested in a lower frequency of phonemes into whose composition the given distinc­ tive features enter. To this second case relate so-called "gaps" (pustye kletki) in the paradigmatic system, which may be viewed as instances of a "difficult" combinability of features. In this sense one may speak of two basic types of combinability of features: "marked" or "recessive" and "unmarked" or "dominant". A combination of features is "recessive" when it is characterized by unusualness or rarity, which is manifested in a lower frequency of the phoneme into whose composition the given combination of features enters and in its distributive limitedness. Such a combinability of features, which is explainable by the "difficult" compatibility of these features in simultaneous sequence, may even be absent in certain linguistic systems, which is expressed by the presence of "gaps" in the paradigmatic system. A combination of features is "dominant" when it is characterized by usualness, naturalness, which manifests itself in a higher frequency and in a greater distributional freedom of the phonemic units into whose compo­ sition these features enter. Such "natural" combinations of features are explainable by their free articulatory and acoustic compatibility in simultaneous sequences, as a result of which the phoneme in question has a stronger and more functional aspect. "Functionally strong", stable bundles of distinctive features (respectively, of phonemes), defined usually as "unmarked", in opposition to "marked",

LANGUAGE TYPOLOGY AND DIACHRONIC LINGUISTICS

467

"functionally weak" and unstable bundles of features (respectively, of pho­ nemes) are renamed here as "dominant" bundles in counterbalance to "recessive" ones. Such a reformulation of the hierarchical relations of "markedness" in respect to a paradigmatic "dominance" with corresponding "dominant" and "recessive" members of the opposition is expedient in view of the polysemy of the traditional terms "marked"/"unmarked", which are still used in their original meaning to designate correlated members of a relation "possessing a feature" (merkmalhaltig)/"lacking a feature" (merkmallos). The terms "dominant/recessive" are in analogy to terms in contem­ porary molecular biology, which is characterized, as is well known, by the wide application of linguistic terminology in defining the concepts of the genetic code. The presence of hierarchical dependencies in the system among separate phonological units or "bundles" of distinctive features, making themselves felt in relations of dominance, testifies to the existence in a linguistic system of a strict stratification of phonological significations. It is indeed in correspondence with such universal signifying correlations that there occur diachronic phonemic transformations in language, A whole series of diachronic phonemic changes in the system, which seem at first glance to be uncoordinated and disconnected one from the other, may be conceived as interdependent, mutually conditioned transformations that are regulated by a detailed hierarchy of phonological significations. In particu­ lar, the manifest dominance of the front series of voiced stops and fricatives in comparison to the back series and, on the contrary, the dominance of the back series of unvoiced stops and fricatives in comparison to the front; the general dominance of consonants in relation to corresponding fricatives and so on, make it possible to define the sequence of phonemic alterations in concrete linguistic systems and to establish universally significant models for diachronic phonemic transformations. A universally significative hierarchy of phonological units presupposes, as was noted above, the presence of phonemes with a lower frequency, reaching null ("gaps" in the system). These paradigmatic regularities of the system should continually be taken into consideration both in the synchronic description of a language, as well as in linguistic diachrony, in particular in the reconstruction of linguistic systems. From the point of view of the theory of dominance the presence of a "gap" is not an anomaly, and, consequently, in the absence of data of external comparison, does not necessitate filling it in in an internal reconstruction of an ancient state of language, as is frequently practiced in diachronic linguistics (cf. instances of "filling in" of so-called "cases vides" in the works of Martinet). Another basic presupposition of comparative-genetic linguistics is the thesis of the arbitrariness of the linguistic sign. Although the "arbitrariness"

468

THOMAS V. GAMKRELIDZE

of the linguistic sign should be interpreted somewhat differently than it is presented by Saussure, — and one may assert, in the light of the "principle of complementarity", a motivated connection between signifié and signifiant on the level of "horizontal relations" — nevertheless the "vertical" relations between signifié and signifiant may be considered "arbitrary" in Saussure's sense, and on this principle is built in essence the entire system of compara­ tive-historical linguistics. In bringing to light formal-semantic similarities between two or several languages, i.e. similarities on two planes simultaneously — both the signifying as well as the signified signs of these languages — there naturally arises the question of the causes for the emergence of such a similarity in the signs of different languages. Proceeding from the thesis of a limited (in the sense indicated above) arbitrariness of the sign, such formal-semantic coincidences of signs of different languages (i.e. phonetic similarity of two or more signs in conjunction with their semantic closeness or identity) may be interpreted as a fact of accidental coincidence of two or more signs of different languages. It is entirely possible to admit that several words similar in phonetic sound and in meaning coincided in two or more languages according to purely accidental factors of combination. One can even calculate with definite approximateness the probability of an accidental coincidence in two or more languages of two or more overlaping or similar words of a given length. The probability of an hypothesis of accidental coincidence to explain such a similarity will decrease in relation to an increase in the number of languages in which such similar signs are discovered and in a still greater degree in relation to an increase in the number of signs in the languages revealing such similarities or overlappings. Another, more likely, hypothesis to explain such coincidences in corre­ sponding signs of two or more languages is historical contacts between the languages and the borrowing of words from one language in the other (or in several languages) or borrowing in both these language from a third source. But not all aspects of formal-semantic similarity of signs of two or more languages can be interpreted as the result of borrowing. There exists a type of similarity between signs of different languages, which is expressed in the presence of regular phonetic correspondences between the similar languages; this type of similarity is not explainable in the usual case by borrowing of the words of certain languages into others. A similarity of this type between signs presupposes the presence of such sound correspondences between languages in which each phoneme x of the language A corresponds in a formally and semantically similar sign of language  with the phoneme y, in such a sign of language  with the phoneme z, and so on.

LANGUAGE TYPOLOGY AND DIACHRONIC LINGUISTICS

469

Such sound correspondences between languages are found usually in groups of words and morphemes that reflect basic concepts relating to human activity and man's environment. This last type of similarity, revealing regular correspondences between sound units of the languages under examination, cannot be explained satis­ factorily either as accidental coincidences in sound and meaning of the words in the different languages or through the thesis of borrowing of words from one language into another or by their arising in these languages from a third source. The sole probable explanation of similarity of this type in corresponding signs of different languages is the assumption of a common origin of the linguistic systems under examination, i.e. the origin of these systems from some kind of common initial linguistic system, which had become trans­ formed in different directions. The explanation of phonemic correspondences between languages by a common origin from a certain initial linguistic system presupposes the necessity of reconstructing it with the aim of studying the rise and means of transformation of historically attested cognate linguistic systems. The comparison of languages, oriented toward establishing the regular phonemic correspondences, should apply logically to the reconstruction of the linguistic model, the transformation of which in various directions could give us historically attested cognate linguistic systems. The comparison of cognate languages cannot be considered a complete stage of investigation of the history of the languages under examination unless it has as its goal the reconstruction of the proto-language. The history of cognate languages having no written records can be estab­ lished only in the event that one succeeds in reducing all the various historically attested linguistic structures to common initial models. In this case it becomes possible to establish the ways these systems arose and developed, beginning with the initial state right up to the historically attested states. Such an approach to the genetic comparison of languages and the estab­ lishment of their interrelations naturally raises the question of the reconstruc­ tion of the original, proto-language system and of the linguistic methods to be used in such a reconstruction. The reconstruction of the system of the proto-language is attained by means of comparison of historically attested cognate language systems and by means of a retrospective movement from one language state to another, earlier state. The retrospective movement should be continued to the point that there will not be reached any language state from which could be deduced all the historically attested cognate language systems while admitting the number of consequent transformations that determine the "diachronic deducibility" of

470

THOMAS V. GAMKRELIDZE

the system. Such transformations move from the initial language system to its later states, which appear to be the result of its structural transformations. With the help of such diachronic transformations one can describe the rise of different cognate language systems from the theoretically postulated initial structural models. In their explanatory power the diachronic transformations enabling one to deduce the historically attested forms of the language from definite theoretical constructs, which may be considered earlier chronological stages of these forms (their "archetypes"), can be compared to the "transformations" of generative grammar, which deduce the observed elements of surface structure from theoretically postulated base constructions constituting the deep struc­ ture of a language. The description of the diachronic changes of a language through rules of transformation is in essence a sequential enumeration of discrete steps, each of which reflects one of the synchronic states in the language's development. The less the chronological distance between such "steps", the more exact and adequate will be the description of the development of the language, reflecting its sequential transitions beginning with its initial state. In reconstructing, with the help of certain methods of linguistic investi­ gation, the initial models of the language, which should in principle reflect the system of the common original language, there arises the methodological question of the reality of the proposed reconstructions and of the degree to which they correspond to the language system taken as the original one for the given cognate dialects as it existed in space and time. The conception of a thesis about the reality of the proposed reconstructions determines a whole series of methodological principles of comparative-genetic linguistic investigations and, first and foremost, their close tie with principles of linguistic typology and the establishment of linguistic universais. In this sense genetic (comparative-historical) linguistics, i.e. linguistics establishing kindred relations between groups of languages and offering the reconstruction of their original models, comprises in principle a single disciple with structuraltypological linguistics and linguistic universais. Actually, the reconstructed linguistic models of the initial language system, if they lay claim to reflecting the language as it really had existed in space and time, should be found to be in complete correspondence with the typologically deduced universal regularities of the language as established inductively or deductively on the basis of a comparison of a multiplicity of different language structures. Linguistic reconstruction contradicting language universais cannot, of course, lay claim to the reality of its representation of the language system as it existed historically. But even the correspondence of the reconstructed models to synchronic linguistic universais cannot serve as a sufficient basis

LANGUAGE TYPOLOGY AND DIACHRONIC LINGUISTICS

471

for affirming the reality of these reconstructions and the reflection in them of the concrete system of the language presupposed to be the initial one for cognate languages. An essential condition here is likewise the correspondence of the reconstructed models of diachronic typological data, the general schemes for the changes in time of certain linguistic structures, established in the investigation of concrete facts in the history of separate languages. The typological verification of the reconstructed linguistic models (syn­ chronic as well as diachronic) thus becomes one of the basic premises of the postulated initial language structures and is essential for testing their prob­ ability. Taking into account synchronic typology in linguistic reconstructions, the necessity of which Jakobson indicated, forces one to reexamine radically the traditional schemes of classic Indo-European comparative-historical grammar and to propose a new interpretation of the established linguistic correspon­ dences on the basis of the principle of structural-typological verification of the system and its diachronic deducibility. The one-sidedness and limitedness of classic comparative-historical IndoEuropean linguistics consisted in the fact that the reconstructed model for Proto-Indo-European was merely the result of an external comparison of separate cognate systems and in certain theories was supplemented by an internal reconstruction on the basis of an analysis of a certain type of relations in the limits of one system. Moreover, the explicitly linguistic probability of the model thus obtained in the sense of its typological correspondence to potentially possible linguistic structures was not taken into account. In classi­ cal Indo-European linguistics this led to the postulation of an initial language system that, contradicting synchronic typological data, could not be con­ sidered a linguistic reality. In particular, in evaluating the traditionally postulated system of ProtoIndo-European consonantism with the application of criterion of empirical reality of the language system, it transpires that the system of consonantism applied in classical Indo-European linguistics as the initial one for all the historically attested Indo-European languages is internally contradictory and does not correspond to the synchronic typology of those languages. Moreover, it contradicts universal linguistic regularities established in the study of lan­ guage universais. Consequently, such a theoretically postulated linguistic model cannot be considered to constitute a system reflecting the language as it really existed before it was transformed into the system of cognate IndoEuropean languages. Correspondingly, one cannot consider as real the pho­ netic changes and transformations that were assumed in classical IndoEuropean comparative grammar in the explanation and description of the transformation of the initial system in the historically attested Indo-European languages.

472

THOMAS V. GAMKRELIDZE

The traditional system

Reinterpreted system

I

II

III

I

(b) d g g gw

bh dh gh gh

P t

(P') t' k' k'

gwh

  kw

k

,o

II

III

b[h]

p[h]

d

[h]

g[h]

t[h] k[h]

g[h]

k[h]

g[h]o

k[h]o

Figure 1. Reinterpretation of the Proto-Indo-European system of stops.

At present one can offer a new interpretation of Indo-European consonantism and, moreover, formulate a new hypothesis as to the oldest localization of the Indo-European languages. In the new interpretation, the Proto-IndoEuropean system of stops proves to be closer to systems traditionally defined as ones with consonantal shifts (German, Armenian, Hittite and others), while systems that were considered in relation to consonantism to stand close to Proto-Indo-European (Old Indic, Greek, Italian and so on) turn out to be the result of complex phonemic transformations of the initial language system (see Figure 1). As a result of this the traditionally established transformations of ProtoIndo-European stops into the phonemic units of the separate Indo-European languages also change, acquiring in the new interpretation a direction opposite that of the traditional scheme. Correspondingly, the basic phonetic laws of classical comparativistics, such as "Grimm's law", "Grassmann's law", "Bartholomae's law", etc., also must be reconsidered, acquiring in the light of this new phonological interpretation of the Indo-European system of stops a new content. The reinterpreted system of Proto-Indo-European consonantism pro­ vides a basis from which to raise anew the question of the interrelations of dialects in the limits of a common language base and its articulation into separate historical dialects, i.e. the general problems of Indo-European areal linguistics. All these problems, which are linked with the general progress of typological and comparative-historical studies, force one to reexamine the traditional schemes of classical Indo-European comparative linguistics and to propose new comparative-historical constructions, in essence, a new system for the comparative-historical grammar of the Indo-European languages.

* Translated by Stephen Rudy.

Language Universals in Relation to Acquisition and Change: a Tribute to Roman Jakobson John A. Hawkins University of Southern California and Max-Planck-Institut Psycholinguistik

für

1. Introduction Among the many major contributions that Roman Jakobson made to linguistics and to psycholinguistics, his 1941 "Kindersprache, Aphasie und allgemeine Lautgesetze" must be considered one of the most innovative. In this work, which was translated into English in 1968 as "Child Language, Aphasia and Phonological Universais", Jakobson discussed the relationship between language universais and language acquisition. Specifically, child language data involving the acquisition of typologically different phonologi­ cal systems are considered in relation to implicational universais of pho­ nology derived from cross-language patterns of variation. His translator (A.R. Keiler) comments in the Preface that Jakobson was the first person to explicitly address the logic of this relationship, and he expresses the hope that an English translation will make the work better known and more widely understood. Jakobson's monograph has received quite a lot of attention since then. Yet I believe that its significance has still not been as widely appreciated as it should be. Nor has there been much subsequent work examining critically his formulation of the relationship between universais and acquisition and extending his method to areas of the grammar other than phonology. The present paper will accordingly build on and extend Jakobson's pioneering work in this area and is intended as a tribute to him within the context of this volume. Perhaps the most general point that Jakobson was trying to make was that the acquisition stages and error types of pre-adult grammars can be shown to obey the same implicational dependencies between linguistic properties that one observes across regular adult languages. Pre-adult gram­ mars are therefore linguistic systems just like any other and are constrained by the same general laws. And language acquisition constitutes just one of several domains (including language change, language breakdown and loss,

474

JOHN A. HAWKINS

regular synchronic variation, etc.) in which the patterns of variation permit­ ted by universal laws can be observed. Hence, it is often inappropriate to explain acquisition phenomena in terms that are unique to a theory of acquisition (e.g. in terms of the maturation of the human organism). What­ ever explains the linguistic universais will also explain the facts of acquisition. Jakobson goes to some length to provide an explanation (in acoustic and articulatory terms) for the phonological universais that he formulates. My interest in this context is not with the adequacy or otherwise of this expla­ nation, nor indeed with the correctness of the phonological universais them­ selves. Rather, I am interested in the more general issue of the logic of the relationship between universais and acquisition. Jakobson's discussion of this relationship is not completely satisfactory. He proceeds throughout his monograph in a brilliantly intuitive fashion, yet without ever making explicit what his precise claims and predictions are. For example, he nowhere provides a general formulation of just what an implicational universal can (and cannot) predict for: order of acquisition; types of errors made; directionality of substitutions in the event that either the antecedent or the consequent property of an implicational universal substitutes for the other. His book is full of suggestive and intriguing illustrations of the correspondence between universais and acquisition, yet most of these are founded on an implicit logic and on additional assumptions that are never spelled out. When he does attempt a more general statement, the formulation is partly incorrect. For example, there is a regular assumption throughout his book that the acquistion of certain structures must actually precede that of others in child language. But no implicational universal can ever predict that one structure will precede the acquisition of another. The strongest prediction we can make is that the acquistion of the consequent property of an implicational statement will either precede or occur simultaneously with that of the ante­ cedent property. This is what the cross-linguistic distribution requires, given that both antecedent and consequent properties may be present simul­ taneously in languages, or just the consequent, or neither. A more explicit formulation of the logic of implicational statements together with the precise consequences for acquisition would have clarified this matter. Jakobson also derives predictions for earliness versus lateness of acquisition from crosslinguistic frequency differences for the properties in question across regular adult languages. But the precise nature of such predictions is very delicate — quantitative differences in the distribution of linguistic properties across languages are not just of the "very frequent" versus "very rare" type he discusses, but often admit of numerous statistically significant degrees. The precise consequences for acquisition (if any) of such frequency differences must ultimately be spelled out, as must the possible confounding effects of

LANGUAGE UNIVERSALS IN ACQUISITION AND CHANGE

475

frequency of occurrence for the relevant properties in the particular lan­ guages being acquired. Some of Jakobson's critics who have attempted to find counter-examples to his predictions have also based their discussion on this same faulty logic (cf., e.g., Ferguson 1977 and Macken 1980), and it is far from clear whether their data still constitute counterevidence, once the implicational predictions are formulated more precisely, and once the universais themselves are stated in a more accurate form than they were in 1941. It is important to define the relationship between universais and acqui­ sition as clearly and correctly as possible, for at least two reasons. First, we need to know, empirically, to what extent child language and interlanguage data are just like regular adult languages, and to what extent they are not. Obviously, first and second language developmental stages exhibit numerous idiosyncratic and language-specific differences in relation to the target lan­ guages being acquired, and they are typically reduced and impoverished versions of these latter. But do they still conform to universal laws? I.e. do they still (despite their impoverishment) obey the principles underlying crosslanguage variation that have been extracted from a comparison of regular adult languages? In each of the grammatical areas for which implicational universais have been formulated, there are less expressive and more express­ ive linguistic variants, e.g. relativization on subjects only, on subjects and direct objects only, on subjects and direct objects and indirect objects, and so on (cf. Keenan & Comrie 1977). Regular adult languages will typically select a mix of more and less expressive variants from these different gram­ matical areas, thereby achieving an overall expressive capability and com­ plexity level that may be roughly equal from language to language. But a first language or an interlanguage developmental stage could still be univer­ sally consistent, even if it systematically selected less expressive options in each of these domains. Second, if acquisition data are indeed universally predictable to a greater extent than is currently realized, we can bring the whole apparatus of grammatical theory and universal grammar to bear on the acquisition task, and contribute new ideas to the explanation of language acquisition. Jakobson, of course, argued that child language data do conform to universal predictions, and that the explanations that underlie universais can be carried over wholesale to explain error types and relative orderings etc. in acquisition. For example, just as there are adult languages with rich inventories of front consonants and impoverished or non-existent palatal and velar series, so the same holds true for the acquisition stages of numerous pre-adult languages. And the underlying reasons and principles, argued Jakobson, are the same in both cases. As a result, linguistic explanations for acquisition can either supplement or replace many hitherto purely

476

JOHN A. HAWKINS

psychological or maturational explanations (cf. Jakobson 1968: 67,92; and section 7 below). The present paper will accordingly address what I shall call the "corre­ spondence question" between universais and acquisition. This question has a definitional, an empirical, and an explanatory aspect. We need to first define the predictions that can be made by universais for first and second language acquisition. Empirically, we must then test to see whether the predictions are correct. And finally we must explain the universais themselves (i.e. we must give some good reason(s) for why they define the range of permitted and non-permitted cross-linguistic variation that they do), and we must explain the fact (if it is a fact) that grammatical principles that underlie adult language variation are accessible to the learner, with the result that these same principles can structure the acquisition data as well. In the present context I shall discuss only the definitional and the empirical aspects of the correspondence question. For despite the fact that an impress­ ive number of language universais have been discovered, and despite the availability of many general explanatory ideas and hypotheses, the precise explanation for most universais is still far from clear (cf. Hawkins 1988a,b). By contrast, the actual universal regularities themselves (i.e. the crosslinguistic patterns of variation) are much clearer, and we can accordingly address in a precise and scientific manner the empirical question of whether acquisition data do or do not conform to observed universal regularities. But to proceed from current explanatory hypotheses to acquisition predic­ tions will almost certainly confuse the answer to the correspondence ques­ tion. Since the proposed explanatory principles for universais are still typically inadequate, any predictions that they might make for acquisition will generally be erroneous. On the other hand, a clear definition of the predictions made by implicational universais for acquisition, and an empiri­ cal testing of these on acquisition data, represents a research goal that can be successfully addressed right now, using available implicational universais and acquisition data. There is also a logical priority to first establishing empirically whether acquisition data do or do not conform to universal predictions. This is what Jakobson himself did. He first juxtaposed universais and acquisition data, predicting the latter on the basis of the former (chapter 2), and only later (chapter 3) advanced some explanatory speculations for the universais them­ selves (speculations that have since been superseded by more sophisticated explanations). He also investigated only a limited set of phenomena (from phonology). Since 1941 there has been dramatic progress in the discovery of implicational universais of morphology, syntax and semantics, and these can now be used to make predictions for acquisition. There have also recently been some intriguing studies examining such universais in relation

LANGUAGE UNIVERSALS IN ACQUISITION AND CHANGE

477

to first and second language acquisition, and these will be discussed in the light of our reformulated predicitons. In all cases, the predictions turn out to be remarkably correct. If this success rate continues, we will have estab­ lished a non-trivial fact (the consistency of acquisition data with implicational universais and their acquisition predictions), and the more explanatory questions that are raised by this correspondence can then be addressed. The order of presentation is as follows. In section 2 I summarize the logic of various types of universais and their predictions for historical change. Section 3 formulates and test implicational predictions for first language acquisition, and section 4 does the same for second language acquisition. Section 5 defines predictions for the directionality of errors in acquisition, section 6 discusses predictions made for acquisition by distributional or frequency universais, and section 7 summarizes our conclusions.

2. Some major types of Language Universals and their predictions for historical change Three major types of universais are commonly distinguished: absolute; implicational; and distributional. Absolute universais (VL P(L), i.e. for all languages, L, property Ρ holds of L) define properties found (or not found) in every single language. E.g. all languages have vowels; all languages have pronoun systems distinguishing at least three persons and two numbers; or (more abstractly) many of the formal and substantive universais of Chomskyan generative grammar. Implicational (or parameter-setting) universais have the form: for all lan­ guages, L, if L has property P, then L has property Q (i.e. L(P(L) ⊃ Q(L))). Such statements set limits on language variation by defining permit­ ted versus non-permitted co-occurrences of properties: i. ii. iii. iv.

P & Q (i.e. both) - P & Q (i.e. Q alone, without P) —P & —Q (i.e. neither) * P & - Q (i.e. P alone, without Q)

E.g. if a language has voiced obstruents word-finally (P), then it also has voiceless obstruents word-finally (Q). More complex implicational universais also exist defined on several properties: e.g. if a language has prepositions (P), then if it has noun before adjective word order (Q), then it will have noun before genitive word order (R) (i.e. L (P(L) ⊃ (Q(L) ⊃ R(L)))). Or implicational universais might underlie whole hierarchies such as the Keenan-Comrie (1977) Accessibility Hierarchy, e.g. if a language has a

478

JOHN A. HAWKINS

primary relativization strategy operating on oblique NPs, then that strategy will operate on indirect objects; and if it operates on indirect objects, it will operate on direct objects; and if it operates on direct objects, it will operate on subjects (that is, L (P(L) ⊃ Q(L)) & (Q(L) ⊃ R(L)) & (R(L) ⊃ S(L)))). Such a hierarchy permits all and only the following combinations of properties: i. ii. iii. iv. v.

P -P& -P& -P& -P&

& Q & R & S Q& R& S -Q& R& S -Q& -R& S -Q& -R& -S

thereby excluding large numbers of combinations such as *P & — Q & R & S, or * - P & Q & - R & - S , etc. Distributional (or frequency) universais stipulate that languages of type T are more frequent than languages of type T' (for theoretically interesting reasons). E.g. the further to the left the subject occurs in a VO language, the more such languages there are: SVOX > VSOX > VOSX > VOXS. These synchronic universais embody intrinsic predictions for language change. Absolute universais require that the properties in question will not change over time. Implicational universais predict possible and impossible orders of changes relative to one another. Historical stages of one and the same language must remain within the observed synchronic parameters. As a result, losses and gains in linguistic structures through time are quite constrained. E.g. given L (P(L) ⊃ Q(L)), and some historically recorded language with P&Q at stage 1 which loses one of these properties at stage 2, it is predicted that P will be lost first. If Q is lost first, the impossible co­ occurrence *P& —Q will result: P & Q →- P & Q / NOT P&Q→*P&-Q Conversely, given a language with — P&—Q at stage 1 which gains one of these properties at stage 2, it is predicted that Q will be gained first: - P & - Q → - P & Q / NOT - P & - Q → * P & - Q Or, given a language with — P& — Q at some historically early stage that is subsequently attested with P&Q, there are two possible pathways of change, and one impossible one: 1. 2. 3.

- P & - Q → - P & Q → P&Q (Q acquired first) — P& —Q → P&Q (P and Q acquired simultaneously) * - P & - Q → P & - Q → P&Q (P acquired first)

Distributional universais predict the relative time amounts (all things being

LANGUAGE UNIVERSALS IN ACQUISITION AND CHANGE

479

equal) during which languages remain in certain states. The more frequent some type Τ is compared to T' on current evidence, the longer languages will remain in state Τ compared to T' historically. Some relevant examples of all these predictions are given in Hawkins (1979, 1982, 1983) for universais of word order. Underlying these predictions is an important uniformitarian principle whereby the laws of the present are assumed to be those of the past as well. This assumption is formulated as the principle of Universal Consistency in History (1983: 211): at each stage in their historical evolution, languages remain consistent with implicational universais derived from current synchronic evidence.

3. Implicational predictions for first language acquisition This section defines the predictions made for first language acquisition by implicational universais. The most general of these is the principle of Univer­ sal Consistency in Acquisition (modelled on the principle of Universal Consistency in History). This principle logically implies a number of predic­ tive statements, the first of which, the basic prediction, makes predictions for relative stages in acquisition. Acquisition is then broken down into production versus comprehension, and similar predictions are formulated for developmental stages in each. For comprehension, error data are used as a test of successful comprehension. Finally a generalized error prediction statement is formulated to cover production as well. In the interests of readability, I shall first introduce some abbreviatory terminology: IL PAL AL SL TL

inter-language pre-adult language (i.e. child language) adult language source language target language

The principle of Universal Consistency in Acquisition and the basic predic­ tion for acquisition stages in pre-adult languages can be set out as follows: The Principle of Universal Consistency in Acquisition (UCA) At each stage in their evolution, PALs and ILs remain consistent with implicational universais derived from current synchronic evidence. UCA Basic Prediction for PAL Stages Given: an AL implicational U 'if P then Q' defining permitted P&Q, — P&Q and — P& — Q co-occurrences, and non-permitted *P& — Q; an adult language, L, containing both P&Q;

480

JOHN A. HAWKINS

Then:

a set of PAL data exemplifying successive developmental stages for L: PALi .... PALn, where PAL; contains both P&Q; Thru the stages PALi .... PALn, either the acquisition of Q will precede that of P, or Ρ and Q will be acquired simultaneously; but Ρ will not be acquired before Q. 1. 2. 3.

- P & - Q → - P & Q → P&Q - P & - Q → P&Q * - P & - Q → P & - Q → P&Q

In effect, the UCA principle predicts that the consequent property of an implicational universal will be acquired either prior to or simultaneously with its antecedent property, thereby avoiding a stage in which the ante­ cedent occurs without its required consequent. Let us reconsider one of Jakobson's examples: 1.

Across languages the existence of fricative consonants implies the co-occurrence of a series of stops (cf. pp. 51-2).

He observes that stops are universal, whereas many language families have no fricatives (e.g. numerous Australian, Malayo — Polynesian, African and South American languages). (For a more recent — and refined — set of universal correlations between stops and fricatives, cf. Gamkrelidze 1978). From this distribution Jakobson proposes that stops (the consequent prop­ erty) will be (a) earlier acquired than fricatives (the antecedent), (b) subject to fewer errors than fricatives, and (c) will replace rather than be replaced by fricatives if (as actually happens) one of these properties substitutes for the other in some PAL stage. We will need to modify (a) in the light of our UCA principle as follows: stops will be acquired prior to or simultaneously with fricatives. Since there are numerous languages with both stops and fricatives, as well as languages with stops only, we cannot expect that PALs will necessarily go through an initial stage corresponding to just one of these independently attested language types, the stops-only type. If they are to conform with universal regularities derived from adult languages, it is just as feasible for them to acquire stops and fricatives simultaneously, thereby corresponding to the languages with both. All we can predict is that PALs will conform to one or the other of these independently attested types. And all we can rule out is that fricatives will be acquired before stops, since no adult language has a fricative series without a co-occurring stop series. Of course, empirically it is indeed the case in the PAL data that Jakobson discusses that stops are acquired before their corresponding fricatives. But the important point is that this is not actually required by the implicational universal, though it is consistent with it. And it is essential to make this distinction since in many other examples the correctness or otherwise of the

LANGUAGE UNIVERSALS IN ACQUISITION AND CHANGE

481

universal predictions for acquisition may hinge crucially on this interpret­ ation: consequent properties are not necessarily acquired first. In their critique of Jakobson's implicational predictions for phonological acquisition, Ferguson (1977) and Macken (1980) present counterevidence to the claim that consequent properties (such as stops) are acquired before antecedent properties (such as fricatives). But in the process they carry over the same erroneous logic into the critique — no implicational universal predicts that the consequent property must be acquired first. Moreover, Jakobson's universal (1) above is a highly oversimplified statement of the cross-linguistic facts in this area, compared with Gamkrelidze (1978). And it is not clear to me that Ferguson's and Macken's data still provide counterevidence to more precise UCA predictions based on more refined universais. We will also need to modify Jakobson's prediction (b) in the same way, requiring that stops will be subject to fewer or equal numbers of errors compared with fricatives. I shall discuss below the rationale underlying his essentially correct equation between universais and errors later in this sec­ tion. His prediction (c) will remain as is. For clarity of exposition, let us now set out and test the precise predictions made by Jakobson's universal (1), as follows: 2.

Across lgs the existence of fricatives implies the co-occurrence of a series of stops. UCA Prediction: in those lgs having both stops and fricatives (e.g. /p t / alongside /f s x/) the acquisition of stops will either precede or occur simultaneously with the acquisition of fricatives. PAL data from Jakobson (1968): stops are regularly produced before fricatives across numerous different PALs.

Consider also a second phonological example taken from Jakobson: 3.

Across lgs, the existence of back (velar and palatal) oral and nasal consonants implies the existence of front (labial and dental) consonants. (E.g. the labial and dental nasals /m/ and /n/ may exist alone without the palatal and velar nasals / / and /η/, but not vice versa. And some lgs (e.g. Samoan) lack a palato-velar series of stops, but not front stops). UCA Prediction: in those lgs having both front and back nasal or oral consonants, the PAL acquisition of front consonants will either precede or occur simultaneously with the acquisition of back consonants PAL data from Jakobson (1968):

482

JOHN A. HAWKINS

Nasal Stops: the acquisition of /m/ and /n/ in PAL English, Scandinavian and German precedes the acquisition of the velar nasal /ŋ/; the acquisition of /m/ and /n/ in PAL French, Czech and Serbo-Croat precedes the acquisition of the palatal nasal / / Oral stops: the acquisition of dental /t/ precedes the acquisition of /k/ in PAL English, Swedish, German. (For more recent universais involving oral stops, cf. Gamkrelidze 1978; for nasal stops, cf. Ferguson 1966.) We can now extend this method to other grammatical areas. Example (4) defines one of many cross-language universais in the area of conditional constructions. The corresponding UCA prediction is formulated and tested against acquisition data from English supplied by Bowerman (1986). The data reveal a developmental progression which is entirely consistent with the UCA prediction. 4.

If a language has a construction expressing prototypical con­ ditional meaning (cf. Comrie 1986) by means of explicit con­ ditional markers (e.g. if S1 then S2 (P)), then that language will also express prototypical conditional meaning by means of S1 & S2 or S1,S2 constructions (Q) (cf. if you wake up late, you'll miss the bus, and wake up late and you'll miss the bus or wake up late, you'll miss the bus). UCA Prediction: conditionality will be expressed first via S1 & S2 or S1,S2, or simultaneously via these conjunctions and explicit conditional constructions; but explicit conditional constructions will not precede the conjunctions. PAL data from Bowerman (1986): PAL stages for English exhibit the following progression: a. conditionality expressed neither by explicit conditional mark­ ers nor by S1&S2 /S1,S2 constructions ( - P & - Q ) ; b. conditionality expressed by S1&S2 /S1,S2 alone ( —P&Q);  conditionality expressed by both (P&Q).

Consider next two sets of universais involving relative clauses: 5.

There are numerous universal predictions linked to the KeenanComrie (1977) Accessibility Hierarchy (AH). E.g., if a language has a primary relativization strategy (i.e. one operating on sub­ jects) which operates on some position low on the AH (e.g. oblique objects), then that same strategy will operate on all higher positions as well. UCA Prediction: in those ALs having a primary relativization strategy operating on more than subjects, the PAL acquisition

LANGUAGE UNIVERSALS IN ACQUISITION AND CHANGE

6.

483

of positions high on the AH will either precede or occur simul­ taneously with the acquisition of each position lower on the AH. E.g. SUs will either precede or be acquired simultaneously with DOs; DOs will either precede or be acquired simultaneously with IOs; etc. PAL data from Keenan & S. Hawkins (1987): acquisition data from English involving comprehension errors are consistent with this prediction (cf. the PAL prediction for comprehension below). Across lgs, if a relative clause construction can carry an apposi­ tive interpretation, it can also carry a restrictive interpretation (cf. Lehmann 1984). UCA Prediction: in those ALs containing relative clauses that can receive both types of interpretations, the PAL use of restric­ tive interpretations will either precede or occur simultaneously with the use of appositive interpretations. PAL data from Lehmann (1984): Lehmann asserts that there is supporting evidence from pre-adult English.

Another fruitful area in which universais can be tested on acquisition data is that of lexical semantics. Andersen (1978) has done this effectively for body-part terminology. She first formulates several implicational univer­ sais, tests these against order of acquisition stages for the relevant lexical items in languages for which PAL data are available, and finds that these stages are consistent with the universais. Her universais include those of (7): 7.

Across lgs the existence of a lexeme uniquely designating LEG implies the co-occurrence of a separate lexeme for ARM; across lgs the existence of a lexeme uniquely designating FOOT implies the co-occurrence of a separate lexeme for HAND. UCA Prediction: in those lgs having both LEG & ARM and FOOT & HAND, the PAL acquisition of ARM will either precede or occur simultaneously with the acquisition of LEG; similarly for HAND and FOOT.

PAL data from Andersen (1978): ARM generally precedes the acquisition of LEG; HAND generally precedes FOOT.

So far I have talked rather loosely about the "acquisition" of some property, without specifying what it means for that property to be acquired. At the very least we need to make a distinction here between production and comprehension. Most of the examples discussed above involve developmen­ tal stages in the first systematic production of one property relative to another. We can accordingly formulate a prediction for production alone, as follows:

484

JOHN A. HAWKINS

PAL Prediction for Production Given: an AL implicational U, etc, as for UCA Basic Prediction; Then: Thru the stages PALi ... PALn, either the production of Q will occur before that of P, or the production of Ρ and Q will be initiated simultaneously. All of the examples (2) through (7) provide relevant tests of this prediction, except for example (5) which involves comprehension. The intuition that we want to define for comprehension is, however, similar. Just as the consequent property of an implicational universal may be produced prior to or simul­ taneously with its antecedent property, so we want to say that the successful comprehension of the consequent property will also either precede or occur simultaneously with that of the antecedent. Exactly this intuition underlies Keenan & S. Hawkins' (1987) testing of the comprehension of relative clauses down the AH at successive PAL stages of English. They expect that ease of comprehension will decline down the AH, and this expectation is confirmed in an experimental task. More precisely, we should say that successful comprehension will be greater or equal for each higher position on the AH compared with every lower position. Unlike production, however, comprehension is more difficult to observe. One way to measure it is through errors, as in the Keenan & S. Hawkins experiment. Subjects were required to repeat back relative clauses to the experimenter after a certain interval, and it was found that the ratio of errors to correct repetitions increased down the hierarchy. The higher positions on the AH resulted in more successful comprehension as reflected in more accurate (error-free) repetition of what had been comprehended. We can therefore formulate a specific prediction for comprehension, as follows: PAL Prediction for Comprehension Given: an AL implicational U, etc, as for UCA Basic Prediction; Then: Thru the stages PAL¡... PAL n, either the successful comprehension of Q will precede that of P, or the successful comprehension of Ρ and Q will occur simultaneously. Testable Prediction: at any PAL stage, the quantity of comprehension errors for Ρ will be greater than or equal to that for Q. The types of errors discussed so far were experimentally induced. Errors may also occur quite spontaneously in production in child language, and it is within the spirit both of our UCA basic prediction and of Jakobson's examples to make the following general prediction: errors in both the production and comprehension of antecedent properties within implicational universais will be greater than or equal in number to those for consequent properties; i.e. either the consequent property will be more successfully

LANGUAGE UNIVERSALS IN ACQUISITION AND CHANGE

485

produced and comprehended (fewer errors), or both will be equally success­ fully (or unsuccessfully) produced and comprehended (equal numbers of errors). This assumption clearly underlies Jakobson's discussion of his PAL error data. He regularly expects production errors in the consonants and vowels mentioned as antecedent properties (i.e. a failure to produce the relevant sounds correctly) to be greater than (or equal to) those for conse­ quent properties. Thus, PAL errors involving fricatives are more extensive than those for stops (cf. example 2); errors involving back oral and nasal consonants are more frequent than those for front consonants (cf. example 3); and so on. Similarly for the other universal predictions: errors in the production and comprehension of LEG will either be greater than or equal to those for ARM; likewise for FOOT and HAND (cf. example 7). We can

therefore formulate a generalized error prediction to cover both comprehen­ sion and production, as follows: PAL Generalized Error Prediction Given: an AL implicational U, etc, as for UCA Basic Prediction; Then: Thru the stages PALi... PALn, errors in the production or compre­ hension of Ρ will be greater than or equal to those for Q.

4. Implicational predictions for second language acquisition Another domain in which we can test our universal predictions is that of second language acquisition. This is a field which has only recently come into its own, and it offers some interesting possibilities which are without parallel in first language acquisition. Second language acquisition proceeds through a series of stages referred to as "interlanguages" whereby the grammar and phonology of the source or native language (SL) are gradually abandoned in favour of the grammar and phonology of the target language (TL) as the TL is increasingly mastered. The question then arises as to whether such interlanguages composed of mixed properties from SL and TL (whose composition is not taught as such even when some tutoring of the second language takes place) conform to universal predictions or not. Are interlanguages consistent with implicational universais in the same way that the developmental stages of child language have been argued to be? The particular interest of second language acquisition stems from the fact that the transitional stages involved are not just incompletely mastered versions of some single language whose properties and rules are drawn from that language alone, but rather they involve a subtle synthesis of both grammars. And the kinds of combinations of grammatical properties that result provide a fertile testing ground for universal predictions, particularly

486

JOHN A. HAWKINS

when the typological distance between SL and TL is such as to permit many logically possible interlanguages, only some of which are in conformity with universal laws. Once again the strongest prediction we can make (and that we have indeed already made, since UCA as formulated above refers to both PALs and ILs) is that all ILs will be universally consistent. This in turn predicts possible and impossible interlanguage stages between SL and TL. Imagine a situation in which some SL has neither of the properties mentioned in an implicational universal ( —P& —Q), whereas the TL has both (P&Q). We predict that the acquisiton of the consequent property Q will either precede or occur simul­ taneously with that of the antecedent P, but Ρ will not be acquired before Q. Moreover, in the event that the transitional stage —P&Q is attested, this combination offers striking proof of the reality of the universais. Neither the SL nor the TL exhibits this particular combination, and hence it will have been constructed out of the properties of both model languages ( —P from SL, Q from TL) in a way which conforms with universal laws and which avoids an equally possible combination (P from TL, and — Q from SL) that would have resulted in the universally unattested *P& —Q co­ occurrence. As in section 3, I shall first formulate and test a basic prediction for IL stages derived from UCA, and will then formulate specific predictions for production and comprehension, followed by a generalized error prediction. UCA Basic Prediction for IL stages Given: an AL implicational U "if P then Q" defining permitted P&Q, — P&Q and —P&—Qco-occurrences, and non-permitted *P& —Q; a language contact situation between L1 and L2, where L1=the source language (SL) whose speakers are acquiring L2, the target language (TL); SL contains neither P nor Q ( - P & - Q ) , TL contains both (P&Q); a set of interlanguages (ILs) representing transitional stages between SL and TL, i.e. SL ... ILi ... ILj ... ILn ... TL; Then: Thru the stages ILi ... ILn, either the acquisition of Q will precede that of P, or P and Q will be acquired simultaneously; but P will not be acquired before Q Let us consider a phonological universal whose predictions for second language acquisition are discussed in Eckman (1984). 8.

If a language has voiced obstruents word-finally, then it also has voiceless obstruents word-finally.

Given a language contact situation in which native speakers of Japanese or Mandarin Chinese (with neither voiced nor voiceless obstruents word-

LANGUAGE UNIVERSALS IN ACQUISITION AND CHANGE

487

finally, i.e. — P & - Q ) are acquiring English (with both, i.e. P&Q), and a set of interlanguage stages in which some but not all of the word-final obstruents of English are being produced successfully; then the interlanguage stages should be in accordance with the predictions. Either the voiceless obstruents will be produced prior to the voiced ones, or the production of both will be acquired simultaneously. Eckman presents data which is in conformity with this prediction. Eckman also discusses an interesting syntactic universal taken from Greenberg (1966) and involving subject-verb inversion. The predictions can be set out as follows: 9.

In all languages, inversions of statement order in questions so that the verb precedes the subject occurs only in languages where the question word is normally sentence-initial; if this inversion occurs in interrogative-word questions, it also occurs in yes-no questions. UCA Prediction: given e.g. a language contact situation in which native speakers of Japanese (no initial wh-word, no inversions) are acquiring English (initial wh-word, inversion in both whquestions and yes-no questions), and a set of interlanguage stages in which inversions are made correctly for only some question structures, then the interlanguage stages will be in accordance with universal predictions. Inversions will be made in yes-no questions prior to or simultaneously with those in wh-questions. IL data from Eckman (1984): Japanese-English interlanguage data are in accordance with this prediction.

Some further illustrations of universal predictions for second language acquisition can be found in Eckman (1984) and in Luján, Minaya & Sankoff (1984). The more specific interlanguage predictions for production, comprehen­ sion and errors, corresponding to those for first language acquisition, are as follows: IL Prediction for Production Given: an AL implicational U, etc, as for UCA Basic Prediction for IL; Then: Thru the stages ILi ... ILn, either the production of Q will occur before that of P, or the production of Ρ and Q will be initiated simultaneously. IL Prediction for Comprehension Given: an AL implicational U, etc, as for UCA Basic Prediction for IL; Then: Thru the stages ILi ... ILn, either the successful comprehension of

488

JOHN A. HAWKINS

Q will precede that of P, or the successful comprehension of Ρ and Q will occur simultaneously. Testable Prediction: at any IL stage, the quantity of comprehension errors for Ρ will be greater than or equal to that for Q. IL Generalized Error Prediction Given: an AL implicational U, etc, as for UCA Basic Prediction for IL; Then: Thru the stages I L i . . . ILn, errors in the production or comprehen­ sion of Ρ will be greater than or equal to those for Q. Data relevant to the Generalized Error Prediction can be found in the papers cited in this section: the quantity of incorrect word-final obstruents in Eckman's (1984) IL data is greater for voiced than for voiceless obstruents; the quantity of verb-subject inversion errors (i.e. failure to invert) is greater for wh-questions than for yes-no questions in Eckman's syntactic IL data; etc.

5. Implicational predictions for the directionality of errors: PALs and ILs Jakobson (1968) observes that stops are not only acquired earlier than fricatives in PAL developmental stages, they actually replace fricatives in the child's rendition of adult phonemes. Thus, /p/ regularly substitutes for /f/, /t/ for /s/, /k/ for /x/ in his PAL data (pp. 51-2). If we examine these facts in conjunction with universal (1) above, it is clear that an impli­ cational consequent property is here systematically substituting for its antecedent, rather than vice versa. This is not an isolated occurrence in Jakobson's examples. Front oral and nasal stops replace back stops (cf. example 3) — /t/ substitutes for /k/ in PAL English, Swedish and German (e.g. the English child produces tut for cut), and front nasal stops substitute for back nasals in English, Scandinavian, German French, Czech and Serbo-Croat PALs (pp. 53-54). Again a consequent property substitutes for its antecedent. Jakobson does not discuss precisely when we can expect one implicationally mentioned property to substitute for another in this way: these kinds of substitutions are by no means always expected or appropriate with many other implicational universais. Nor does he actually define the principle to which he seems to be appealing here, namely that consequent properties will replace their antecedents rather than vice versa, if indeed one of these properties substitutes for the other at all. But his intuition is important and (I believe) correct, and can be formulated as follows:

LANGUAGE UNIVERSALS IN ACQUISITION AND CHANGE

489

PAL and IL Prediction for the Directionality of Errors Given: an AL implicational U "if P then Q" defining permitted P&Q, —P&Q and —P&—Q co-occurrences, and non-permitted *P& —Q; an adult language, L, containing both P&Q; or a language contact situation between SL and TL, where TL contains both P&Q; a set of PAL stages for L, or of IL stages between SL and TL; regular production or comprehension errors in PAL or IL stages involving P and Q, whereby one of these properties incorrectly substitutes for the other; Then: Q will regularly substitute for P, rather than vice versa. Further supporting evidence comes from the PAL relativization data of Keenan & S. Hawkins (1987). These authors observe that when a repetition error is made whereby one relativization position incorrectly substitutes for another (e.g. a subject for a direct object), such errors typically involve the substitution of a position higher on the AH for one that is lower. Since these higher positions represent the consequent properties within the relevant implicational universais of relativization, we have another example in which a consequent property replaces its antecedent rather than vice versa. A relevant example from second language acquisition involving phonology is discussed in Eckman (1984). He observes that word-final voiceless obstruents replace word-final voiced obstruents rather than vice versa in IL errors (cf. universal 8). The voiceless obstruents represent the consequent property in this universal. Notice finally that our prediction for the directionality of errors does not necessarily require that antecedent properties such as fricatives must be substituted for by consequent properties such as stops, when the former are subject to errors. It predicts only that if there are errors involving the substitution of one of these properties for the other, then the substitution will be in the direction predicted. It is not being claimed that a fricative must be replaced by a stop, and hence the kinds of data discussed by Ferguson (1977) and Macken (1980) whereby other sounds such as liquids may replace fricatives in e.g. the acquisition of Spanish are simply irrelevant for rather than counterexamples to this implicational prediction. Since Jakobson did not define his predictions in this area with any kind of precision, it is unclear what exactly he would predict and whether he would accept Ferguson's and Macken's data as counterevidence. The important point is, however, that this interpretation of Jakobson does not follow from the logic of an implicational statement. All that this latter can rule out is the existence of an antecedent property in some language without a corre­ sponding consequent property. If some error type could potentially produce such a situation, it is predicted not to occur. If no such violation arises,

490

JOHN A. HAWKINS

then no predictions are made, and we must look to the many other consider­ ations that Ferguson and Macken have discussed in order to make some sense of the substitutions.

6. Distributional predictions for acquisition In addition to deriving acquisition predictions from implicational universais, Jakobson also uses frequency of cross-linguistic distribution in this context. His examples suggest that rare phenomena across languages will be late acquisitions, while frequent phenomena will be early. For example, the distribution of nasal vowels across languages is relatively limited. (Of 706 languages in Ruhlen's 1978 sample, 155 have nasal vowels). Correspond­ ingly, nasal vowels in French and Polish PALs are late acquisitions which emerge only after all the remaining vowels have been successfully acquired. By contrast, nasal consonants exist in all languages and are among the earliest acquisitions in PALs. Jakobson observes also that the class of languages with a single liquid (/1/ or /r/) in lieu of the double series of English is very large. Correspondingly, a single liquid exists in numerous PALs for a long period with the second liquid (if any) being among the last speech sounds to be acquired. The Czech sibilant /r/ is one of the rarest phonemes cross-linguistically and is also the latest acquired in Czech PALs. Now, some obvious cautions are in order here. We do not necessarily want to predict that the cross-linguistically rare click phonemes of the Khoisan languages will be the last phonemes to be acquired, if (as is the case) they are very frequent in the input, i.e. in the particular language being acquired. Nonetheless, there is, once again, an important insight underlying Jakobson's examples which is worth spelling out. There are often (though not necessarily always) very good linguistic reasons why some structures should be more preferred than others across languages (cf. Hawkins 1988a,b for further discussion). And, all things being equal, we will expect the less preferred structures to be later acquired and subject to more errors than their more preferred counterparts. For given that earliness versus lateness of acquisition (and errors) are not simple reflections of the frequency of occurrence for the relevant structures in adult and target languages (as was shown convincingly in Brown's 1973 frequency-based predictions for the order of acquisition of English morphemes — the relative frequencies of these morphemes in the input were not good predictors), the factors that do determine lateness of acquisition may include the very grammatical consider­ ations that lead to the infrequency of certain structures across languages as a whole, regardless of the frequency situation in some particular language. This is not to deny that frequency in the input may have some role to play

LANGUAGE UNIVERSALS IN ACQUISITION AND CHANGE

491

in predicting order of acquisition (as in the click example). It simply asserts the plausibility of using cross-linguistic distributional evidence at all in this context. The predictions that these make are, of course, probabilistic only. But they can be considered a part of the ultimate predictive package of relevant interacting factors that will ultimately make correct acquisition predictions for typologically different languages. In addition to Jakobson's own relevant phonological examples in this context, we might add the following illustrations from the work of Dan Slobin. Slobin (1985) observes that the cross-linguistically more infrequent prenominai relative construction of Turkish is acquired significantly later in Turkish PALs than the cross-linguistically much more frequent postnominal construction in PAL stages of English. Turkish children do not employ prenominai relatives productively until their sixth year, whereas Englishspeaking children are using postnomimal relatives by age three. And Slobin (1978) points out that suffixal morphology (which is cross-linguistically much preferred to prefixing, cf. Cutler, Hawkins & Gilligan 1985) is an early acquisition in child language, even when the relevant morphological para­ digms are rather irregular and complex (as in Slavic case suffixing, for example).

7. Conclusions In this paper I have attempted to do two things: first to define some general predictions that can be made by different types of universais for (first and second) language acquisition, predictions that are largely implicit in Jakobson's own discussion of phonological universais and child language; and second to extend such predictions to morphology, syntax and semantics, areas in which there are now impressive universal generalizations and rel­ evant acquisition data that were not available to Jakobson at the time he wrote his monograph. In each case, the acquisition data are very much in accordance with the universal predictions. The significance of this correspon­ dence for the field of developmental psycholinguistics is considerable. For if the acquisition stages of PALs and ILs are in regular conformity with universal laws, this will correspondingly reduce the validity of alternative explanations for acquisition in terms that are unique to the process of acquisition per se. The explanation for the universais will automatically be the explanation for the relative order and error types etc. in acquisition as well. Jakobson (1968) foresaw the significance of such a correspondence between implicational universais and acquisition, but this aspect of his work has not had the impact upon developmental psycholinguistics that it

492

JOHN A. HAWKINS

deserves. This is not to deny, of course, that numerous other relevant considerations must also be incorporated into an eventual theory of language acquisition, having to do with the maturation of the human organism, or (for second language acquisition) involving factors unique to language con­ tact situations, or (for first and second language acquisition) involving frequency of occurrence for the relevant structures in the input. But the important point is that if there are profound grammatical principles underly­ ing all languages and constraining cross-language variation (as current research on universais suggest), and if pre-adult languages and interlanguages are similarly constrained by these principles, it follows that many acquisition phenomena may be the way they are simply because of the properties of the grammars that are being acquired, rather than because of some independent and more psychological principles and processes unique to acquisition. Hence, many proposed explanations for acquisition phenom­ ena may actually have little to do with acquisition per se, but rather may reflect the fact that what is being acquired is a grammar obeying certain principles and permitting certain types of variation rather than others, among which are the observed developmental and interlanguage stages in language acquisition. The explanations for universais may be various (sem­ antic, functional, innateness, etc, cf. Hawkins 1985, 1988a,b) and the search for such explanations is currently being extensively pursued. But within an acquisition context, it is important to know to what extent developmental stages are simply reflecting grammatical and explanatory principles that are equally productive in regular adult languages, in previous historical stages of these, etc., and to what extent they are not. And at the very least, a systematic investigation of grammatical universais in an acquisition context will enrich the stock of principles that one can draw on in predicting and explaining language acquisition.

References Andersen, E.S. 1978. "Lexical Univerals of Body-part Terminology". Universais of Human Language Vol. 3, ed. by J. H. Greenberg. Stanford: Stanford University Press. Bowerman, M. 1986. "First Steps in Acquiring Conditionals". On Conditionals, ed. by E.C. Traugott, A. ter Meulen, J.S. Reilly, and C.A. Ferguson. Cambridge: Cambridge University Press. Brown, R. 1973. A First Language: The Early stages. Cambridge, Mass.: Harvard University Press. Comrie, B. 1986. "Universals of Conditional Sentences". On Conditionals, ed. by E.C. Traugott, A. ter Meulen, J.S. Reilly, and C.A. Ferguson. Cambridge: Cambridge University Press. Cutler, Α., J.A. Hawkins & G. Gilligan 1985. "The Suffixing Preference: a Processing Expla­ nation", Linguistics. 23.5, 723-58.

LANGUAGE UNIVERSALS IN ACQUISITION AND CHANGE

493

Eckman, F.R. 1984. "Universais, Typologies and Interlanguage". Language Universais and Second Language Acquisition, ed. by W. Rutherford. Amsterdam: John Benjamins. Ferguson, C.A. 1966. "Assumptions About Nasals: A sample study in phonological universais". Universals of Language, 2nd ed., ed. by J.H. Greenberg. Cambridge, Mass.: MIT Press. Mass. . 1977. "New Directions in Phonological Theory: Language Acquisition and Universais Research". Current Issues in Linguistic Theory, ed. by R.W. Cole, 247-99. Bloomington, Ind.: Indiana University Press. Gamkrelidze, T.V. 1978. "On the Correlation of Stops and Fricatives in a Phonological System". Universals of Human Language Vol. 2, ed. by J. H. Greenberg. Stanford: Stanford University Press. Greenberg, J.H., 1966. "Some Universais of Grammar with Particular Reference to the Order of Meaningful Elements". Universals of Language, 2nd ed., ed. by J. H. Greenberg. Cambridge, Mass.: MIT Press. Hawkins, J.A. 1979. "Implicational Universais as Predictors of Word Order Change", Language 55, 518-48. . 1982. "Language Universais and the Logic of Historical Reconstruction", Linguistics 20, 367-390. . 1983. Word Order Universals. New York: Academic Press. . 1985. "Complementary Methods in Universal Grammar: A reply to Coopmans", Lan­ guage 61, 569-87. . 1988a. "On Explaining some Left-Right Asymmetries in Syntactic and Morphological Universais". Studies in Syntactic Typology, ed. by M. Hammond, E. Moravcsik and J. Wirth. Amsterdam: John Benjamins. , ed. 1988b. Explaining Language Universais. Oxford: Basil Blackwell. Jakobson, R. 1968. Child Language, Aphasia and Phonological Universals. The Hague: Mouton. Keenan, E.L. & B. Comrie 1977. "Noun Phrase Accessibility and Universal Grammar". Linguistic Inquiry 8, 63-99. Keenan, E.L. & S. Hawkins. 1987. "Experimental Results on the Psychological Validity of the Accessibility Hierarchy". Papers in Universal Grammar, ed. by E. L. Keenan. London: Croom Helm. Lehmann, . 1984. Der Relativsatz. Tubingen: Gunter Narr Verlag. Luján, M., L. Minaya and D. Sankoff 1984. "The Universal Consistency Hypothesis and the Prediction of Word Order Acquisition Stages in the Speech of Bilingual Children'. Language 60.2, 343-71. Macken, M.A. 1980. "Aspects of the Acquisition of Stop Systems: A Cross-Linguistic Perspec­ tive'. Child Phonology: Perception and Production, ed. by G. Yeni-Komshian, J.F. Kavanaugh and C.A. Ferguson. New York: Academic Press. Ruhlen, M. 1978. "Nasal Vowels". Universais of Human Language Vol. 2, ed. by J.H. Greenberg. Stanford: Stanford University Press. Slobin, D.I. 1978. "Universal and Particular in the Acquisition of Language". Proceedings of the Conference on Language Acquisition: State of the Art, ed. by L.R. Gleitman and E. Wanner. . 1985. "The Acquisition and Use of Relative Clauses in Turkic and Indo-European Languages". Studies in Turkish Linguistics, ed. by K. Zimmer and D. Slobin.

Paralinguistic Universals and Preconceptual Thinking in Language Ivan Fónagy C.N.R.S. Paris This contribution is intended as an attempt at prelinguistic reconstruction, based on the assumption that different forms of pre-verbal communication are still contained in some way in speech acts which are performed today. I shall try to examine three verbal patterns that may be considered as verbal remnants of three different stages of pre-verbal communication. They may also be studied as three forms of preconceptual ideation. The three verbal forms to be examined are: (a) vocal style; (b) intonation; and (c) metaphor. My presentation — written in my Hungarian-French Proto-English — is essentially based on previous studies in the domain of "phono-stylistics" (see Léon 1971, Fónagy 1977), prosody, and poetic language. To facilitate the discussion I shall occasionally restate some of the results which have already been published.

1. Vocal style 1.1 The structure of vocal style The radiocinematographic analysis of Hungarian and French emotive speech (Fónagy 1983; Fónagy, Han and Simon, 1983) clearly shows that the same emotive attitude induces similar distortions in the articulation of vowels and consonants. For instance, anger is expressed in French as well as in Hung­ arian by: 1. 2. 3.

fast, spasmodic tongue movements (i.e. short transitions, the tongue is stiffened for brief periods in extreme positions); highly increased muscular tension (increase of tongue and palate contact in plosives, strong contraction of the uvula); increased maxillary angle and labial distance (for /a/ from 15.0 to 25.6 mm);

496 4. 5.

IVAN FONAGY

the tongue is withdrawn during the articulation of both vowels and consonants; the mandible is withdrawn, the lower incisors are retracted to a position considerably behind the upper incisors; the upper inci­ sors bite on the lower lip in articulating /w/ or /v/.

Oral mimicry is particularly spectacular in the case of rolled /r/ pronounced in anger. The tongue is strongly elevated towards the upper ridge of the teeth, and resists "heroically" the pressure of outflowing air: it regains its elevated position four or five times, whereas it vibrates twice during neutral speech. The expression of tenderness differs significantly from that of anger: 1. 2. 3. 4. 5.

articulation is smooth and continuous; tongue and lip muscles are more relaxed; the uvula less con­ tracted; the tongue position is more advanced for the same vowels; labialization of /i/ and /e/ frequently occurs; alveolar plosives are more or less palatalized.

Data obtained by means of myographic and tomographic analysis are consistent with radiocinematographic findings. During tender phonation the laryngeal ventricle widens out. In angry speech the ventricle is constricted into a narrow passage (Fόnagy 1983: 43-50). As a consequence, despite increased effort of the expiratory muscles, the acoustic intensity is relatively low in anger, and still lower in hatred (Fónagy 1966). The ratio acoustic energy/physiological energy is considerably smaller in aggressive emotions than it is for tender feelings. Analogously, the relative duration of conson­ ants, especially that of plosives, increases, that of vowels decreases in speech dominated by aggressive emotions (Table 1). How to interpret such shifts in articulation and voice production? Some years ago I suggested considering concrete speech sounds in live speech as the result of a double encoding procedure (Fónagy 1971a). The phoneme /e/, for instance, in Én ' T ' was repeatedly pronounced as [i:n], much closer than /i/ in Igen "Yes" pronounced with ostensible indifference (Figure 1). Actresses playing the role of a young mother speaking to her child trans­ formed /i/ in [y] and /e/ in [œ]: [moe sy vwajõ] mais si, voyons. The Hungarian hearer perceives the /i:/ in En as a very tense /e/; the French hearer interprets /ce/ in Mais as an /ε/ spoken with tenderness: [i:] → /e:/ + tension (hatred) [œ] → /ε/ + tenderness The concrete speech sounds can be easily broken down into two com­ ponents, since they are, in fact, composed of two widely differing elements:

3.98