Baixe o app para aproveitar ainda mais
Esta é uma pré-visualização de arquivo. Entre para ver o arquivo original
PERSPECTIVES ON HISTORICAL LINGUISTICS AMSTERDAM STUDIES IN THE THEORY AND HISTORY OF LINGUISTIC SCIENCE General Editor E.F. KONRAD KOERNER (University of Ottawa) Series IV - CURRENT ISSUES IN LINGUISTIC THEORY Advisory Editorial Board Henning Andersen (Copenhagen) ;Raimo Anttila (Los Angeles) Tomaz V. Gamkrelidze (Tiflis);Hans-Heinrich Lieb (Berlin) J. Peter Maher (Chicago);Ernst Pulgram (Ann Arbor, Mich.) E. Wyn Roberts (Vancouver, B.C.);Danny Steinberg (Honolulu) Volume 24 Winfred P. Lehmann & Yakov Malkiel (eds.) PERSPECTIVES ON HISTORICAL LINGUISTICS PERSPECTIVES ON HISTORICAL LINGUISTICS edited by WINFRED P. LEHMANN & YAKOV MALKIEL JOHN BENJAMINS PUBLISHING COMPANY AMSTERDAM/PHILADELPHIA 1982 Dedicated to the memory of Emile Benveniste, 1902-1976 Jerzy Kurylowicz, 1895-1978 and George Lane, 1904-1981 © Copyright 1982 - John Benjamins B.V. ISSN 0304 0763 / ISBN 90 272 3516 3 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. PREFATORY NOTE A decade and a half ago the editors of this volume arranged a conference to review aims for historical linguistics. The papers presented at that conference were thereupon published under the title: Directions for Historical Linguistics (Austin: University of Texas Press, 196 8). They may well have achieved some of their goals. In any event the considerable attention sub sequently devoted to historical linguistics suggested the usefulness of another conference. Unlike its predecessor this one required no major support; it was arranged as part of a meeting of a learned society. Two sections of the Modern Language Association program in San Francisco, December 27-29, 1979, were allotted to the Language Theory Division for presentation of papers which are now published here. While the earlier papers and their topics are by no means obsolete, the titles of the papers in this volume may indicate that historical linguistics has progressed in the meantime. Not that the historical study of language has completely overcome an unfortunate legacy from one of its most influential theorists, by which it was sharply separated from descriptive linguistics! Yet the separation between the two approaches to the study of language no longer maintains the Saussurean sharpness of even a decade ago. Linguists are coming to understand that if problems are examined in isolation from the dimension of time affecting all human activities, such isolation, as well as their views of language, may be as artificial as the context in which they are treated. Linguistics is slowly overcoming another trouble some heritage, this too bequeathed by theorists who have in part adversely influenced the historical study of language: virtually exclusive attention to the sounds and forms of language. While like other linguists, the neo- grammarians -- who set this narrow course — considered the sentence the minimal unit of language, in effect they limited their attention to its most readily manageable Vi PREFATORY NOTE segments, phonology and morphology. Such a delimitation not only led to neglect of syntax, let alone discourse; it also provided limited direction for the approaches to the topics on which it concentrated, as essays presented below illustrate. Linguistics pursued in the neogrammarian tradition, which persisted among theoreticians in this country as well as much of Europe, also for the most part treated language outside its social context. It was one of the achievements of the late Uriel Weinreich to assist in moving our field back to consideration of language as a social phenomenon. The founders of modern linguistics, notably Jacob Grimm, undertook their investigations of language as only one part of their concern with communica tion. Philology, the field in which they and their successors carried out their studies, involved the inves tigation of communication generally. In some degree then we now return to the broad approach of the founders of modern linguistics a century and a half ago. To solve a particular problem students of language may restrict their approaches. But the study of phonology, or mor phology, or syntax with no attention to language in its social setting may be most useful in refining formalism. Investigation of language in its use by social groups deepens our understanding of it, at the same time disclosing many opportunities for further study. By its title, the first essay after the introduction links this volume to its predecessor. Stating the prin ciples presented in the final essay of that volume and reviewing subsequent research, it assesses progress made in the meantime towards constructing a general theory of change. Like the following papers on phonology and morphology, it poses new questions that have arisen in the increasingly ambitious research devoted to language. Historical attention to discourse, the topic of the fifth paper, is virtually new, though it too finds predecessors among philologists who devoted themselves to texts. Finally, two essays treat etymology, one concentrating on the rigorously investigated Romance field, the other on Indo-European, especially on new insights prompted by attention to Hittite in accordance with views developed largely in study of languages totally unknown or disregarded by previous Indo-Europeanists. Etymology is the historical study of language in manageable proportions, including all sub-divisions of linguistics: phonology, morphology, syntax, discourse as well as meanings of items for individuals and society. In recalling the earlier participation of Emile Benveniste and Jerzy Kurylowicz, who included its study as well as PREFATORY NOTE Vii theoretical approaches to language in their scope of investigation, we present this collection in the tradition of students of language whose broad grasp of it guides them to illuminating all of its segments as parts of the whole. From the foregoing remarks, however, it would be wrong to infer that a much-needed partial return to certain crucial implications of earlier thinking, be it of the twentieth century, be it -- a fortiori — of the Romantic era, comes close to exhausting our goals. Quite the contrary. The advance of linguistic scholarship and science (two branches intertwined) resembles a spiral, with each generation, to be sure, trying to reject certain gratuitous exaggerations of its predecessors by moving, sometimes abruptly, in the opposite direction, but with steady progress being nevertheless achieved along another, perhaps more important, axis. The latter quali fication is a measure of the difference between a linear projection of the swings of a pendulum, where one movement virtually cancels out another and no visible over-all progress is achieved, and a genuine spiral, which unmistakably combines general progress along one line with the cancellation of extremist positions along another. Here is the acid test for this contention: Ask yourselves whether any of the papers here included would have been conceived or written at a distinctly earlier moment. In disagreement with the rigidity and schematism of previous schools of thought, we neverthe less cheerfully adopt certain minor techniques and absorb individual data developed or clarified under those regimes. It has been argued that the late Ramon Menendez Pidal was, figuratively speaking, a latter day re-incarnation of Jacob Grimm. The comparison is valid; but only on the understanding that the arsenal of tools available to Spain's intellectual giant was radically different from the modest kit with which Germany's originalgenie operated. Many changes of taste and perspective were necessary before, for example, typological analysis as envisaged here, cutting across the borders between Indo-European Romance, and Germanic — that is, domains of knowledge almost hermetically sealed off in the past — could be undertaken in an experimental vein. In planning this work we selected specialists in specific fields of historical linguistics and gave them latitude for the direction of their contribution. We have tried to maintain their individualism of outlook and style, and are grateful to them for their participa tion. We would like to express our appreciation to Viii PREFATORY NOTE Janet Johnson, Lynn Johnk and Margaret Woodruff for editorial assistance, especially with the combined bibliography. The computer facilities of the Linguistics Research Center of the University of Texas at Austin permitted assembly of the master bibliography. The Cambridge University Press authorized inclusion of the maps from Dialect Geography by J. K. Chambers and Peter Trudgill. Janet Johnson merits special thanks for preparing the camera-ready typescript. We are also grateful to the Modern Language Association for our initial forum and for a travel grant which enabled the participant from abroad to attend the meeting at which these papers were first presented. Austin, Texas & Berkeley, Calif. May 19 82 The Editors TABLE OF CONTENTS Prefatory note ν Table of contents ix Charts, figures and tables χ Abbreviations xi-xii 1. WINFRED P. LEHMANN Introduction: Diachronic linguistics 1 2. WILLIAM LABOV Building on Empirical Foundations 17 3. WOLFGANG U, DRESSLER A Semiotic Model of Diachronic Process Phonology 93 4. YAKOV MALKIEL Semantically-Marked Root Morphemes in Diachronic Morphology.. 133 5. ELIZABETH CLOSS TRAUGOTT From Propositional to Textual and Expressive Meanings; Some Semantic-Pragmatic Aspects of Grammaticalization 245 6. STEVEN N. DWORKIN Romance Etymology 273 7. CAROL F. JUSTUS Indo-European Etymology with Special Reference to Grammatical Category 291 Bibliography 329 Subject Index 373 Author Index 377 CHARTS, FIGURES AND TABLES Figure 1(a). (sj) in Brunlanes: speakers aged 70+ 43 Figure 1(b). (sj) in Brunlanes: speakers aged 25-69 44 Figure 1(c). (sj) in Brunlanes: speakers aged -24 63 Figure 2 Retrograde movement of checked (eyC) in Philadelphia 64 Table I. Studies of linguistic change and variation in urban speech communities 91-92 Chart 1. Gradual change of rules and their outputs 118 Chart 2. Diachronic deiconization of phonological processes 123 Flowchart of changes of grammatical markers 257 Table 1. Antecedent-consequent markers in the OE Boethius 260 Table 2. Antecedent-consequent markers in the ME Boethius 261 Table 3 Antecedent-consequent markers in the I.T's Boethius 262 Chart I. Active, Middle, Perfect 303 Chart II. Thematic Active and Middle 304 Chart III. Hittite -bi conjugation 305 Chart IV. Genus verbi 310 Chart V. Greek transitivizing Active 311 Chart VI. Comparative IE transitivizing Active 313 Chart VII. Transitivizing Middle, Perfect, and Active forms 315 Chart VIII. IE 'move' with transitive Active renewal *-(n)-ew 317 Chart IX. Denominative, eventive and other renewing suffixes 318 Chart X. Revised Hypothesis * s e k - : s o k - : s k - 325 ABBREVIATIONS (A) Genera l ad j . adjective 0 object adv. adverb obi. oblique arch. archaic obs (ol.) obsolete BEV See Abbreviations (B) OED Old English Dictionary Centr. Central orig. originally Class. Classical OV object verb coll. colloquial Orth. orthographic(ally) con j . conjunction pers. personally dial. dialectal pi. plural dim. diminutive poet. poetically f. feminine p.ptc. past participle F2 formant 2 PR phonological rule fig. figurative pres. present gram. grammatical prop. property gen. genitive refl. reflexive intrs. intransitive s. subject IPA International Phonetic sg. singular Alphabet sub j . subjunctive JCS Journal of Cuneiform subst. substantive Studies stand. standard lit. literally SVO subject-verb-object LCV Project LCV, see p. 51 SOV subject-object-verb LPC linear predictive coding trans. transitive LYS Labov, Yaeger & Steiner v. verb m. masculine var(s). variants M.E.D. "Modern English Dictionary VO verb object med. medieval West. Western mod. modifier WLH Weinreich, Labov and MR morphological rule Herzog n. noun nom. nominative NP noun phrase xii ABBREVIATIONS (Β) Names of Languages Am. American English ON Old Norse Andal. Andalusian OProv. Old Provencal Ar. Arabic OPtg. Old Portuguese Ast. Asturian OSp. Old Spanish Arag. Aragonese Pers. Persian BEV Black English Vernacular PIE Proto-Indo-European Brit. British English Ptg. Portuguese Cast. Castilian R. Russian Cat. Catalan Rom. Romance Du. Dutch Rum. Rumanian E. English Sc. Scottish EME Early Middle English S. It. South Italian ENE Early New English Skt. Sanskrit Fr. French Sp. Spanish Friul. Friulan(o) Sw. Swiss G. German Tusc. Tuscan Gal. Galician Vegl. Vegliote, i.e. North Gmc. Germanic Dalmatian Goth. Gothic V. Lat. Vulgar Latin Gr. Greek Gr.-Lat. Graeco-Latin Hebr. Hebrew IE Indo-European It. Italian Lang. Langobardian L(at.) Latin L Lat. Late Latin LG Low German Lith. Lithuanian LME Late Middle English ME Middle English Med. Lat. Medieval Latin MHG Middle High German NE New/Modern English NH New Hittite NHG New High German Neap. Neapolitan Norw. Norwegian OAst. Old Asturian OCSl Old Church Slavonic OE Old English OFr. Old French OFris. Old Frisian OH Old Hittite OHG Old High German INTRODUCTION: DIACHRONIC LINGUISTICS WINFRED P. LEHMANN University of Texas at Austin When historical linguistics began to develop in the 19th century, individual scholars made major contributions in distinct areas. These set directions for subsequent study. Franz Bopp produced a seminal work in morphology (1816); Jacob Grimm produced a similarly influential study in phonology (1822). Following Grimm, Rudolph von Raumer, Hermann Grassmann, Carl Verner and Ferdinand de Saussure amplified understanding of phonological change. Following Bopp, August Schleicher and Karl Brugmann made refinements in morphology; see Lehmann (1967) for pertinent examples of important contributions in phonology and morphology (see also Koerner 19 78, esp. 189-209). By the end of the century the theoretical bases of historical phonology and morphology had reached a high level, as handbooks like Eduard Sievers's Grundzuge der Phonetik (5th ed 1901), Hermann Paul's Prinzipien der Sprachgeschichte (1st ed 1880, 5th ed 1920) and the theoretical statements on morphology in Brugmann's Grundriss, Vol. II (1906-1916) illustrate. The course of development in each area followed parallel lines. First, data were assembled and generalizations were stated. Grimm's presentation of the data for his rules on the Germanic consonant shift was remarkably complete as are the paradigms in Bopp's short work. The generalizations were then refined, as in solutions of the residues — those among the Germanic consonants labeled exceptions to Grimm's law. Finally, universal principles were proposed, as in August Leskien's neogrammarian principle and in the many works on morpho logical change which are often brought together under the term analogy (Kurylowicz 1964; Anttila 1977). (1) 2 WINFRED P. LEHMANN In further areas of language, involving semantics — syntax and the lexicon — the diversity of data delayed a similar advance to universal principles. The syntactic treatments of German, Latin and Greek in the first half of the 19th century did not proceed beyond assembly of data. Grimm's highly praised volume on the syntax of the simple sentence in the Germanic languages provides a clear example (1898) . It brought together massive sets of examples, but the conclusions do not proceed beyond summaries for the data assembled. Some what later Berthold Delbrück proposed generalizations based on his study of Greek and Sanskrit syntax, tending towards principles clarifying historical syntax. But only in this century were universal principles advanced, as in Brugmann's psychological principles (1918), in Behaghel's laws (Vennemann 19 74) and in P. Wilhelm Schmidt's conclusions on the history of the genitive and adjective contructions in the Romance languages (19 26). Later works have attempted to sharpen syntactic princi ples, and to move historical syntactic study to a similar place with the study of phonology and morphology (Lehmann 1974; Lightfoot 1979). Historical study of the lexicon proved even less amenable to such advances. It attracted an indefatigible scholar, August Pott. Through volume after volume he comments on huge and amorphous compilations of data (1859-1876). Some of his observations are illuminating, in virtually poetic form, as when he compares the role of particles in sentences to that of spices, a pinch of which suffices to change the flavor of a dish. Yet rather than general principles governing semantic change, the major effort was devoted to determining and defining units which are carriers of meaning: roots and affixes, and thereupon to describe their relationships. The dictionaries in use today for historical purposes, such as the comprehensive work by Alois Walde and Julius Pokorny for the Indo-European languages, are compilations of this sort (1927-1932). The etymological dictionaries for individual languages build on these, concentrating their attention on individual words. There were attempts to provide some generalizations, as by Michel Breal (1897) and Gustav Stern (1931). But these had severely limited application, scarcely more than for the examples cited. Breal's "law of specialization" and his "law of differentiation" illustrate by their very designations that such "laws" apply only to limited lists of words; one may use them for clarifying those lists but not the lexicon in general, let alone for predicting the course of semantic change. In short, historical approaches to DIACHRONIC LINGUISTICS 3 semantics did not advance beyond the initial stages of study which William Dwight Whitney labeled "assembling" and "arranging," with virtually no success in "explaining" semantic change (Whitney 1896:6). Current attention to directions which historical linguists are exploring or to perspectives which they envision must therefore take the semantic segment as the primary focus of concern. Two earlier publications indicate preliminary steps in dealing with semantics in historical studies: a volume published as a Festschrift for Wilhelm Streitberg which was labeled 'state and tasks of linguistics' (no editor, 1924) and the prede cessor to this volume (Lehmann and Malkiel 1968). In addition, many linguists have indicated their views on historical linguistic theory in longer works, often in prefaces, as did Brugmann (1904, 1906) and Meillet (1937), or in separate articles. Some of these have been assembled, as in the widely read collections of Meillet (1921, 1934) and Benveniste (1966). Any histori cal investigation must take their views into considera tion; a survey of them here would extend this introduction unduly. The survey of 19 24 concentrated on Indo-European linguistics, after a review of "Indo-European and general linguistics" (Junker 1-64). Yet even the essays directed at Indo-European invite general studies, such as Junker's initial dictum, based on Schuchardt, that linguists must have general concerns, rather than limiting themselves to Romance or Indo-European studies: "Der Indogermanist ist Sprachforscher, oder er ist wissen- schaftlich nichts" (Junker 1924:1). It is idle to dwell on the 19 24 survey, in view of its lack of impact on general linguistics or even on Indo-European studies. Linguists were about to concentrate once again on phonology and morphology, influenced on the one hand by the axioms of Bloomfield, on the other by Prague School functionalism. In this concentrated attention there was no follow-up to the invitation to syntactic or semantic study. General linguistics embarked on its extended concern with structural phonology and thereupon generative syntax, while historical linguistic study remained in a holding pattern. In a move out of this pattern the 196 8 publication reexamined fundamentals: the artificial gulf created by Saussure between descriptive and historical study of language (3-20); impact of the morphological level on the phonological in change (21-64); further attention to the interrelationship between phonology and morphology in a clarification of the "notion of morphophoneme" 4 WINFRED P. LEHMANN which characterized the late work of Kurylowicz (65-81); reemphasis on a sociolinguistic approach in the study of language change, in the often cited essay of Weinreich, Labov and Herzog (95-198). These topics illustrate steps which diachronic linguistics needed to take for regaining its bearings. The status reached for Indo-European studies was open for reevaluation. Directions for historical linguistics were to come from attention to all languages, not merely the Indo-European. And that attention was to be directed at all segments of language, including "the area of semantic change" (1968:18). The last decade and a half has seen a flowering of historical linguistics, with historical attention to languages wherever data are available, and to all areas of language. Some contributions were prompted by the international conferences on historical linguistics, beginning with that at Edinburgh in 1973, others by special sessions of recurring meetings, such as that of the Chicago Linguistic Society (1976) , still others by special symposia, such as those held at Santa Barbara (Li 1976, 1978). Advances resulting from these have been incorporated in subsequent studies, including those here. Such advances also led to restatements, as of Indo-European syntax (Lehmann 19 74) and of diachronic principles (e.g. Lightfoot 1979, Ramat 1981). The advances are far too broad for more than brief comment here. Among the chief are statements of principles in accordance with generalizations drawn from typological study. These principles have guided restatements of long-studied phenomena. In phonology, for example, a restatement of major consequence for Indo-European studies has resulted from the application of such generalizations to the early obstruent system. When findings of the laryngeal theory eliminated the proposed voiceless aspirated stops from the Proto- Indo-European system of obstruents, maintenance of voiced aspirated stops constituted a highly suspect system, as Roman Jakobson brought to the attention of a wide group of linguists in the Oslo Congress (Jakobson 1962:531; see also Kurylowicz 1935 and Lehmann 1952: 80-81). A. Set of dentals in PIE system B. Set criticized proposed before laryngeal by Jakobson: theory: t th t d dh d dh DIACHRONIC LINGUISTICS 5 There are of course various solutions. One is to allow for Proto-Indo-European an aberrant system, on the grounds that such a system might well be temporary and on the way to restructuring, as it indeed was in all the Indo-European dialects. Another is to assume that the phonemes represented by the traditional bh dh gh were not aspirated stops but rather fricatives, as many linguists had proposed (Prokosch 19 39); others indicated their hesitation by simply proposing phonemes (Benveniste 1935) or by their transcription (Lehmann 1952:112 — dh). Another proposed solution is directed by the finding that when obstruent systems have a gap in their set of labial stops, the order in which such a gap occurs is likely to be voiceless and glottalized. Applied to the Indo-European system, with its minimal evidence for [b], this situation has been taken as grounds for recasting the Indo-European obstruent system to that represented below, proposed independently by Hopper (19 73) and by Gamkrelidze-Ivanov (1973). B'. (ambiguous value for d ) C. t t d dh t' d As may be noted from Hopperfs current designation: glottalic theory, the modified system represents a major departure in Indo-European phonology, comparable to the positing of palatals, velars and labio-velars after Schleicher, or to the laryngeal theory. These earlier departures were proposed on the basis of closer scrutiny of data or as a result of new data, as that furnished by Hittite on laryngeals. The glottalic theory on, the other hand is based solely on reasoning from typological observations. Whatever the eventual impact of the glottalic theory on the phonological system posited for Proto-Indo-European, generalizations achieved by typological scrutiny will be increasingly important as further proto-systems are posited. Such generalizations have also been applied to the syntax of Proto-Indo-European, permitting for the first time a tentative history of Indo-European syntax. It must not be forgotten that historical linguists of the past decided against providing a history of proto- languages. Brugmann states his view unambiguously, as in the preface to his second edition (1897:ix): that [a historical presentation] is the proper 6 WINFRED P. LEHMANN method of presentation, corresponding to the essence of the matter, that the facts of language history should always be presented in the context and chrono logical order, in which they actually occurred, has been indicated by us in Vol. II., p. 800f. Unfortu nately there is still a great problem in the determi nation of chronology for prehistoric periods, even of relative chronology. And I fear the time is still far distant when a presentation of Indo-European history involving the various languages will replace the usual presentation of the data -- such a presen tation will of course concentrate on the prehistoric period -- and successfully make use of the form of presentation which we consider ideal . (Translation mine: W. P. L.) Brugmann's belief as expressed here led to the presenta tion of facts synchronically rather than diachronically, even in apparently historical grammars. His view on the proper presentation directed works on proto-languages as well as the comprehensive grammars of early languages (Brockelmann 1911-1913; Schwyzer 1939- 1953; Szantyr 1965). While Schwyzer and Szantyr, for example, cite data on Homeric Greek and early Latin, they do not present the changing systems of these lan guages. By contrast, a departure from the non-historical approach to a historical sketch of the Proto-Indo- European phonological system was published in 1952 (Lehmann 1952:109-114; sketches of the declensional system and subsequently the syntactic system of Proto- Indo-European were proposed subsequently (Lehmann 1958; see also Fairbanks 1977; Lehmann 1974; Neu 19 76; Schmalstieg 1981). All of these rely on generalizations based on the comparative study of languages in typology (see Hopper 1981). Reactions to such steps towards historical treatment of proto-languages vary. Some scholars are opposed to it, holding that one must restrict one's conclusions to those based on attested data (Watkins 1976). Such restriction implies concern only for surface structures, a posture sharply limiting all linguistic study and maintained by virtually no one outside historical linguistics and by few historical linguists. Objections are also raised to the use of a framework (Watkins, op. cit.) Such objec tions must apply only to the specific framework employed, for all linguistic study is based on assumptions of frame works, whether simple structures like the vowel triangle or subject-predicate clause, or more elaborate models achieved through broader knowledge of language. All DIACHRONIC LINGUISTICS 7 conclusions on language are guided by theory, which as in all sciences, especially the social sciences, finds verification in statistical procedures. The reliability of such conclusions depends on representativeness and accuracy of observation, which will improve as further languages are incorporated in the body used for typo logical generalizations. Conclusions must not be based on individual instances of phenomena in single families or in disregard of vary ing structures in different language types. Historical linguistic theory has its best possibilities for reliable conclusions in three language families of longest attesta tion: Afro-Asiatic, Indo-European and Sino-Tibetan. The status of historical theory may be indicated by recalling that of these only the Indo-European has been rigorously investigated, and that there still are gaps in our know ledge of its development. Historical theory will be amplified as these gaps are filled and the two other families with a long period of materials are further investigated, and as language families attested for shorter periods, like Caucasian, Dravidian and Finno- Ugric are explored. In the current state of historical theory two potential dangers may be briefly noted: generalization from an instance of which we have no further information, and erroneous evaluation of data. As an example of the former we may note the Japanese expression for the reflexive by means of a "pronoun" zibun and a pronominal prefix zi-, both of which may be translated 'self'. Our current understanding of syntactic likelihood in an OV language like Japanese leads us to expect expression for reflexivization through a verbal suffix, as in Turkish and Quechua. Fortunately our scant knowledge of the history of Japanese is adequate to permit us to recognize its means for reflexivization as borrowed from Chinese, in this way avoiding the danger of proposing counter-evidence to our OV framework. This example illustrates the hazards of basing generalizations on languages of the Americas, of Africa and of other regions for which we have no early records that would provide the means to identify external influences and borrowings. Aberrancies in expression of clause order, relative clauses, reflexivization and so on may be due to borrowing like that of the Japanese expression for reflexivization from Chinese. If our information on Japanese were less, we might be tempted to modify our framework, since Japanese is in general highly consistent in OV patterning. As it is, we are faced with a particular problem which with our meager data on 8 WINFRED P. LEHMANN Old Japanese may be insoluble. Similarly we are fortunate in having adequate informa tion about the history of Sinhalese, so that we can account for irregularities in its syntactic patterns, as in the arrangement of its elements in the teen numerals (Ratanajoti 1976, Lehmann 1978:404-405). Most, have the OV pattern [ten + unit], but two do not, and for these an elegant solution was found. Regrettably, we have no such information on the early Turkic teen numerals, which have been recklessly used as counter-evidence for the generalizations on their typical arrangement (Miller 1979). Nor do we have such information on many current languages which have been cited as counter-evidence to generalizations based on well-known language families. Principles for historical linguistics cannot be proposed with data from languages without records from the past through which we might identify patterns that may have been borrowed. Investigators may be led to erroneous evaluations of data in a narrowly based science like linguistics by its terminology, which is taken almost entirely from study of a small number of Indo-European dialects. A ready example is the term "passive", which needs no extensive discussion here in view of the transparent misapplication in recent years of a particular syntactic construction — the passive in English — in attempts to account for constructions in other languages which are supposedly comparable because they have also been designated by the term passive. The passive of related languages like English and German differs, not to speak of that in older stages of Indo-European languages such as Homeric Greek and Vedic Sanskrit, and of the adver sative passive of Japanese and other Southeast Asian languages. Another term with misleading implications is "conditional clause"; still another is "relative clause." The designations for these have shifted in meaning, directed by connotations of the terms "conditional" and "relative" in English. Happily the role of conditional clauses has been identified in a penetrating study by Haiman entitled "Conditionals as Topics" (1978). Haiman's conclusions recall those of the Classical grammarians, as indicated by their designations: Gk hypothesis 'a proposed thesis or topic' and Lat. condicio 'an agreed premise on which to base a conclusion.' In their original use the force of the terms 'conditional' and 'hypotheti cal1 is not focussed on conjecture or contingency but rather on a posited situation, a protasis which is to be commented on in an apodosis. A modern interpretation of Classical grammars might interpret their terms for the DIACHRONIC LINGUISTICS 9 two segments of conditional sentences as topic and comment. Similarly, the term "relative" was applied to a clause associated with another clause; the relationship does not require a relative pronoun or other marker as implied by Lightfoot in his recent book on historical theory when he states that the lack of a relative pronoun in a lan guage deprives that language of the "recursive function ...of relativization," leaving the language "finite ... and radically different from all attested languages" (1979:30, fn. 1). Lightfoot here refers to discussions of Latin and Proto-Indo-European. But the types of relativization proposed for these has long been identi fied. Henri Weil in 1844 — translation into English 1887 — compared the early use of the Latin relative pronoun and complementation patterns to that of Turkish, a language which permits recursion without relative pro nouns as readily as do Japanese and many other OV languages. Similarly, Jacbbi provided data from many languages with such relative clauses (189 7). Unfortu nately his accurate analysis of relative clauses lacking relative pronouns was disregarded by his contemporaries and later scholars. The problem once again may be partially terminologi cal, compounded by generalization from relative clause patterns in the SVO languages of modern Europe. Like many western grammatical terms, relative clause derives from traditional grammar, which does not preserve the insights of Greek grammarians concerning such relation ships. The Greeks identified a major class of words as belonging to the class "arthron", which like the Latin translation articulus means 'joint1. The sub-variety of "articles" used to join relative clauses to principal clauses was labeled hypotaktikon; for the Greeks the relative clause was a form of hypotactic or subordinate clause. Recognizing the problem which might be caused by the subsequent designation, Jespersen preferred as a "more pertinent name" connective or conjunctive (1922: 85). Whether we retain the established name or introduce a substitute, the relative construction is a subordinate clause associated with nouns. The signal for subordina tion varies among language types. OV languages do not need characteristic "joining words" as found in SVO languages. Linguists dealing with historical syntax must be concerned with the construction itself and not be misled by the designation. If Weil's understanding of relative clauses had been incorporated in the body of historical linguistic theory, many problems might long ago have received appropriate treatment. 10 WINFRED P. LEHMANN Moreover, historical theory has long incorporated procedures which scholars may neglect only with unfortu nate results. Among these is suspicion of conclusions based on the argumentum ex silentio, that is, on the absence of evidence. Havers illustrates the dangers of such a procedure by referring to the historical present, which as has long been known is not found in Homer or in other selected texts in Greek, Germanic and Slavic. But he vigorously rejects the conclusion that for this reason it was absent in Proto-Indo-European (1931:3-4). Yet precisely this conclusion was proposed once again, with no greater credence (Kiparsky 1968). The procedures and principles of past scholarship are indeed open for modification. Yet some are so well- established that any disregard of them needs explicit justification. Like Havers and the predecessors he cites, historical linguists must know the sources, including stylistic and dialectal characteristics, and deal with them appropriately, that is, in accordance with general linguistic findings from earlier study. Havers warns explicitly, if pithily, about the need to exercise care in the treatment of translations like the Gothic (19 31: 4). Some specialists consider wulfila's translation such a close reproduction of Greek that a monolingual speaker of Gothic could not have understood it. We cannot verify this point of view, but any Gothic scholar knows that Gothic syntax is for the most part the syntax of a Greek Bible, which to complicate the problem has not been identified. Many other early texts as well are translations influenced by the original, both in Indo-European languages like Tocharian, Armenian, Old Church Slavic and so on, as well as in other families. The textual problems may be equally hazardous when syntactic patterns are incorporated that have not been influenced by direct translation but rather by syntactic borrowings, as in areal phenomena. The procedures developed by past scholars must be applied with increasing rigor. Assuming that we apply such procedures, we may ask what are perspectives that can now be proposed for historical linguistics. In broad terms these have to do with closer attention to the relationships between significant or carriers of meaning and signifie or meanings, and with language change in the complex societies of today. Concentration on carriers of meaning, on structures composed of entities identified by meaning but treated after such identification only as points in a structure, DIACHRONIC LINGUISTICS 11 has to be sure made contributions to historical theory. Some of these were corrective. Excessive concern with meaning and function had led to assigning unrealistic effects of these on form. Examples are available in the history of English phonology from 1400-1950 entitled Laut und Leben (Horn-Lehnert 1954). Its cardinal principle is stated p. 56: "Parts of the word which have l i t t l e function can be weakened/ parts that have no function can be weakened or lost. And by contrast, parts of the word with important function are maintained" (my translation). A more extensive presentation of this theoretical view was provided in Horn's earlier monograph (1921, 1923). Examples illustrating application of the view can be found throughout both works, as in the treat ment of weakly stressed vowels (1954:591-642). For example, in contrast with its widespread loss, -e- "had to be maintained [in forms like houses, edges, ended] because it was functionally important. The language as a purposeful activity insisted on its rights" (1954:595; my translation). Structuralism swept away any such attributions of language-internal influence as little more than vicious circles of reasoning and limited itself to the observation of forms in change. See also Labov's essay here, Section 3, for subsequent observation of losses in connection with the function of the concerned elements. The structural concentration on form applied through out grammar, whether dealing with phonological units or with syntactic units like NP and VP, or with lexical units like roots and suffixes. Attention was concen trated on the interrelations of the identified units. The place of the investigated units was precisely determined, as in Benveniste's thorough study of the Proto-Indo-European suffixes -ti- and -tu- (1948). However valuable such studies are, they give little attention to the role of language in its social context. A more significant critique for attention to language itself finds shortcomings in such studies because they may overlook deeper structural relationships in their major attention to form. Treatment of relative clauses as illustrated above provides an example. When identi fied as modifiers introduced by a relative pronoun, they are determined by superficial characteristics rather than by their role in language. Historical linguistics will also seek to determine more general principles regulating relationships. Still in its initial stages, such scrutiny has led to differ ences of classification. Many syntactic constructions, for example, are grouped together in Vennemann's 12 WINFRED P. LEHMANN "principle of category-constancy"; proposed some time ago, this principle is now presented succinctly in the important book published conjointly with Bartsch (1982: 34-35). By the principle, the relationship of subject to verb and object to verb are equated, as well as those of non-finite verb form and auxiliary, adjective and noun, grade modifier and adverb, in all, more than a dozen constructions. The constant factor in these con structions is taken to be endocentricity: each construc tion is "essentially of the same category as its major constituent" (34). And the entire analysis of syntax is shaped by this assumed function. Whether this basis for determining deeper relation ships leads to an accurate view of language and its change is not the chief concern here. As with all theory, the proof is in application: does the principle accord with and clarify the facts of language? In judging its accuracy, one must note that the principle of category- constancy equates two forces long distinguished in grammatical analysis: government and modification. Through such equation it faces a problem in dealing with titles, given names, etc. such as Japanese Tanaka sensei, Tanaka Michio versus English Professor Tanaka, Mickey Tanner. By Vennemann's approach titles like sensei and given names contrast in arrangement with adjectives and so on. When one on the other hand distinguishes the forces government and modification, titles are treated as standards (Lehmann 1978, 16-17, 174), whose arrangement reflects that of government constructions rather than modifiers. This approach accounts for the consistent difference in arrangement between adjectives and titles, which might well seem to belong in the same class when one applies a "principle of natural serialization" (Vennemann 1974:347). The approach recognizing this distinction, illuminates the history of Indo-European as well as other languages (Lehmann 1974, 1978). Yet both principles illustrate that the central concern in current perspectives is with functional and semantic as well as formal relationships. And the functional relationships are located at a deeper level than are those of studies like Horn-Lehnert's. The search for deeper structural relationships and their expression in language may well be the distinguish ing feature of current historical linguistics. Historical study in the past, for example, posited readily identi fiable units of expression in sorting out roots, bases or stems and affixes. Malkiel's essay here points to subtler configurations of form. The introduction, main tenance or loss of these, as well as their interplay with DIACHRONIC LINGUISTICS 13 the grosser units of past historical treatments, may well be examined in linguistic families outside the Indo- European for their improved understanding. Attention to deeper elements of language has also led to examination of such qualifiers as interrogative, nega tive, causative and others examined in modal logic. Not always identifiable by like formal expression in languages of different types or in successive stages of a language, these have eluded adequate historical inve stigations. Causative may serve as a relatively straightforward example. In its most transparent form, it is expressed by means of prefixes in VSO languages, e.g. the Hebrew inflection, by suffixes in OV languages, e.g. Japanese yomaseru 'cause to read'. The Indo-European languages illustrate one possibility of change, from the suffixed form expected of OV structure, e.g. Sanskrit sadáyati 'places < causes to sit', Gothic satjan 'set' vs. Sanskrit sadati, Gothic sitan 'sit down'. Productive in the OV stages of the Indo-European systems, the suf fix gradually was eroded as dialects moved to SVO structure. In current English the causative, like many verbal qualifiers, is expressed through auxiliaries or lexically. Even related lexical pairs: sit set, lie lay, have merged in general speech, leaving such pairs to be determined solely by semantic criteria, such as kill die. Similarly, as a long-established SVO language, spoken Chinese expresses the causative lexically (Chao 1968: 310). Even though expression for causative, like that for interrogative and negative, is relatively trans parent, investigations of expression for it in various languages and various stages of a language are valuable in improving our understanding of the essential features of language. Such investigations must'take as their fundamental criteria of analysis the semantic bases of qualifiers rather than particular surface expressions. The middle, which includes reflexive and reciprocal force, may serve as an example of even more varied formal expression. Like causative, it involves affixation: prefixation in VSO languages, e.g. Hebrew Niphal and and suffixation in OV structure, e.g. Turkish -in- or giyinmek 'dress oneself, get dressed'. In SVO languages it is expressed characteristically by pronominal elements like self, each other. Earlier works failed to associate the SVO expression with that of OV languages, and accordingly to depict adequately the history of the middle inflection in Indo-European; in the oldest texts, this expression was through a suffix, like that of Turkish. Moreover, a transitional form by expression 14 WINFRED P. LEHMANN through adjectives was generally misinterpreted. Homeric Greek phi los is often mistranslated 'dear' rather than by an appropriate equivalent of the middle, e.g. 'dear fatherland' for 'his fatherland1. This example may illustrate how syntactic analysis linked closely to form failed both in associating such comparable expressions for the middle in languages of various types and in stating the relationships of those patterns in successive stages of the language, as well as in instructing students and literary scholars in their interpretation. Identification of such varying patterns and their change is now an important task. On it will depend much of the success of efforts to understand historical developments in syntax and discourse. The treatment of modality has scarcely been examined historically, except when identified by surface forms. As one example, Lightfoot recognizes modals by the following criteria (1979:98): a. "They do not undergo Number Agreement or do Support but b. "do undergo Subject-Auxiliary Inversion and Negative Placement. C. "They cannot appear in infinitives and gerunds, d. "cannot occur adjacent to each other and cannot take normal complementation forms," If these criteria, with the exception of (d), were distinctive for modals, German would have no modals. Yet the role of modals in German syntax is much like that in English; and some dialects of English also permit (d). Rather than attempting to define syntactic features by surface criteria, we must recognize quali fiers, such as Necessitative and Voluntative, and set out to determine their manners of expression in the various types of languages. In Proto-Indo-European these logico-semantic categories of modality are expressed through suffixes, as one would expect in an OV language. In the subse quent SVO structures, auxiliaries were introduced, as in English and German. Yet in both stages of the language the same categories are prominent (cf. Calbert 1975). Study of the Indo-European languages has then disclosed fundamental verbal qualifiers. But it is unclear which we need to posit. When one examines languages with clearest surface representation, such as Quechua, the number of verbal categories expressed is considerably greater than in Indo-European. Quechua includes suffixes for Frequentative, Extemporaneous, Collaborative and Benefactive among others (Bills et al· 1969:235). DIACHRONIC LINGUISTICS 15 And logicians have identified modalities which they treat in deontic logic, alethic logic, epistemic logic, and so on. We may ask whether these modalities are to be recognized as verbal qualifiers and examined for their means of expression in different language types and differing stages of languages. Conversely morphological or lexical expressions found in specific languages or language types must not be assumed for other types without careful analysis. The passive is one example. Syntactically transparent in the Indo-European languages of SVO structure, where it serves to topicalize "objects", it has been posited also for OV structure, where the discourse effect of SVO passives is expressed by other means, such as particles. Moreover, passive forms of Classical Latin, Classical Greek and late Sanskrit have been reconstructed to include a passive in Proto-Indo-European, even though the Classical Latin and Greek forms develop largely from the older Middle, though the Vedic Sanskrit forms are not parallel with those of Greek and Latin, and though Hittite lacks parallel forms. Closer attention to the meaning of the passive might well have forestalled such reconstruction, and the assumption of a passive universal ly. As Indo-European studies move beyond the 19th century reconstructions centered on Classical Latin, Greek and Sanskrit with primary attention to morphological struc ture, and as other language families are studied for their historical development, the categories treated in . historical linguistics will be more heavily based on semantic examination and identified through study of universals and their expression in languages of various types. Paths towards such broadened understanding are laid out by Justus and Traugott. In extending its concern historical linguistics will draw on such well-charted areas as Romance etymological studies, as sketched here by Dworkin. It will also make use of advances in phonological theory, related here by Dressier to semi otics, and of advances in sociolinguistics. Recent work in sociolinguistics has begun to provide historical linguists with 'new information' through 'systematic research' which Meillet hoped for in the last of his excellent Oslo lectures of 1924 (Meillet 1967: 124-138). Such research, stimulated by the influential article of Weinreich, Labov and Herzog (196 8), and pursued with energetic guidance from Labov is in accord ance with Meillet's dictum: "What interests the linguist is not the norms but the way in which the language is 16 WINFRED P. LEHMANN used" (op.cit. 133). Earlier sociolinguistic study focussed on rural and village dialects "where every body is approximately at the same social level, where everybody has approximately the same culture." That study explicated the situation which we may assume for the communities of Proto-Indo-European, Vedic Sanskrit, Proto-Germanic and even Homeric Greek speech. But "the dialect of a city where there are people of different condition, of different educations, and of different cultures is another thing" (Meillet 1967: 133). Even the cities vary in their conditions, such as Istanbul "where peoples speaking five or six different languages live side by side" as opposed to Paris, Tiflis and so on. The quantitative techniques applied to gain insights into the different conditions of urban civiliza tion are providing historical linguistics with improved understanding of language in use, and as it changes. These investigations "will furnish precise ... informa tion on states of languages, " as Meillet hoped in the final words of his last Oslo lecture (op. cit. 138). The perspectives acquired through studies like those presented here will indeed bring about changes in principles and theory, as Meillet hoped. One may in turn hope that those perspectives will be sought with clarity of outlook comparable to Meillet's. Urges to overgeneralize, to overextend observations, to declare previous approaches obsolete must arouse hesitation. For as Meillet noted, "nothing differs more from one state of language than another state of language" (op.cit.133). Urban dialectology may well clarify the linguistic situation at the beginning of our era for Rome, as it does that today; but these differ from that of Rome of the seventh centuries B.C. and A.D. As for Meillet, requisite application of linguistic principles and of the findings of sociolinguistic and psycholinguis- tic study "depends on the tact, the judgment, and the good sense of linguists" (op.cit. 1251 We present these essays in the hope that they will suggest means for increased understanding of language, the highly flexible and complex system which in spite of its constant change provides the agency for societies to maintain themselves. Precise studies of languages in linguistic communities today as well as increasingly penetrating studies of languages in linguistic communities of central concern to historical linguists may fulfill expectations for the future of linguistics suggested by these perspectives. BUILDING ON EMPIRICAL FOUNDATIONS WILLIAM LABOV University of Pennsylvania At the first conference on Directions in Historical Linguistics in 1966, Weinreich, Labov and Herzog [WLH] proposed principles for the empirical foundations of a theory of language change.1 Since that time, a number of research efforts have been designed with these princi ples in mind, and many interpretations of research results have made reference to the principles involved. It is not too soon to attempt an assessment of the progress that has been made in building on these empirical founda tions, and to see what kind of theories of language change have begun to emerge from research of this type. 1. The fundamental principles Looking over the general statements in the first section of the 1966 article, and reviewing the points found most important by later writers, it seems that the central principles of this exposition can be stated in the form of two "empirical foundations". 1.1 Normal heterogeneity. The normal condition of the speech community is a heterogeneous one: we can expect to find a wide range of variants/styles, dialects, and languages used by members. Moreover, this hetero geneity is an integral part of the linguistic economy of the community, necessary to satisfy the linguistic demands of every-day life. For this heterogeneity to be worth remarking, it must be distinguished from free variation on the one hand, and free expression on the other. Free variation is an inevitable consequence of linguistic structure, the (17) 18 WILLIAM LABOV obverse of the fundamental notion of linguistic "same". The heterogeneous elements that create a problem for traditional linguistics description are much greater in size and extent, including variation among phonemes, morphemes, syntagms and sememes that are the invariants of such description. To include them under the heading of free variation is to reduce the notion of structure to vacuity. Free expression is an inevitable consequence of the different histories and personalities of the users of the language: they have different things to say. The heterogeneous character of the community appears in the fact that there are many alternate, semantically equivalent ways of saying "the same thing".2 The term "structured heterogeneity" emphasizes another aspect of this phenomenon that distinguishes it further from free variation. The occurrence of the variants in question is often correlated with features of the internal environment, though not exactly predict able from those features, and also with external charac teristics of the speaker and the situation: contextual style, social status and social mobility, ethnicity, sex and age. 1.2. Linguistic community. The object of linguistic description is the grammar of the speech community: the system of communication used in social interaction.3 Techniques for linguistic description must be adequate to deal with the heterogeneous character of this object, and many of the methods developed in sociolinguistic research have accordingly been devoted to the description of variation. The heterogeneous nature of the community raises the question of how this object can be circumscribed: what are the limits of the speech community to be described? The heterogeneity discussed so far is largely a feature of speech production. The community is defined on the level of interpretation; the obverse of heterogeneous speech production is homogeneity in the interpretation of the variants. Given semantic equivalence of the variants, such interpretation appears in the form of social evaluation, overt for a few social stereotypes, covert for the great majority of variables. A variety of experimental techniques have been used to detect this type of evaluation, with consistent and convergent results, but this is not the only aspect of homogeneity of interpretation in the speech community. There is also a common direction of style shifting, common directions of self-correction, and common directions of change. Those who acquire the vernacular of the speech community BUILDING ON EMPIRICAL FOUNDATIONS 19 in their formative years show this general agreement, but others do not, even when they show many of the community features of speech production. The principles put forward by WLH were supported by reference to a considerable body of evidence; empirical foundations are necessarily empirical. They were also supported by a detailed critique of opposing positions which have had a strong influence on linguistic research for more than a century, though they have no such empirical basis. Opposed to the first principle of WLH is the assumption that the speech community is normally homogeneous. A long history of investigations, begin ning with Gauchat 1905, has shown that when this idea is taken as a null hypothesis, it must be rejected. Yet it re-appears as an underlying assumption in many theories of linguistic change, and even more clearly in the apologies of field workers for the speech community that they have studied, as they explain that the normal processes of historical change cannot apply in the face of the widespread variation that they happen to have found. The vigor of the assumption of homogeneity also appears in the continued life of the second opposing principle, that the proper object of linguistic description is the idiolect. As WLH showed, the empirical refutation of the homogeneity assumption led many linguists to retreat to the margins of the arena of empirical research, or even to withdraw outside of that arena. In the tradition of structural linguistics, the idiolect became reduced to the speech of one person talking on one subject in one context for a very short time. In the generative framework, it has become the privileged introspections of the theorist, immune from contradiction or counter-evidence. But in spite of the theoretical and methodological weaknesses of this notion, it has remained useful for those who feel that the object of description should and must be homogeneous, and who have neither the inclination nor the motivation to deal with the variations found in everyday speech. So far, everything that has been said applies equally well to synchronic description or diachronic studies. Why then did WLH propose principles specifically as founda tions for the theory of language change? The first reason is a tactical one. The demand for homogeneity, and the consequent retreat to the idiolect, are not unreasonable moves as the first steps of a synchronic linguistic description.4 But it is quite otherwise for the study of linguistic change. It is possible to study completed changes as a series of discrete replacements of elements 20 WILLIAM LABOV in homogeneous systems. But no one has ever quite lost sight of the fact that change is the process of replace ment, not the outcome of that process. When we study the process directly we are immediately confronted with the heterogeneous character of linguistic systems. Change implies variation; change is variation. Since the time of Gauchat, we identify change in progress by the co existence of older and younger variants. Furthermore, that progress is rarely represented by the categorical replacement of one form by another, but normally by changes in the relative frequencies of the variants and changes in their environmental constraints. The second reason is a matter of general strategy. There is a natural alliance between dialect geographers, who study heterogeneity in space; sociolinguists, who study heterogeneity in society; and historical linguists, whose concern is heterogeneity in time. Historical linguistics may be characterized as the art of making the best use of bad data, in the sense that the fragments of the literary record that remain are the results of historical accidents beyond the control of the investigator. The data of dialect geography expand enormously the number of cases that can be considered; the data of sociolinguistic studies greatly increase the fineness of resolution of the process of change. On the other side of the ledger, the dialect geographer and sociolinguist are handicapped by the shallowness of their temporal perspective. Many of their interpretations rest on shaky ground unless they have obtained additional time depth from historical records of their own speech community: this is par ticularly true of inferences about change in progress (section 3.1). In any event, inferences drawn from the records of the present will always be more valuable if they can be supported by parallel cases drawn from completed changes extended over time. The uses of the past to explain the present, and the uses of the present to explain the past, rest upon two underlying assumptions. One is the uniformitarian doctrine: that the events that produced the historical record are the same type as those that can be observed operating around us today (Lyell 1833). The other is the traditional position of historical linguistics: that we understand some element of linguistic structure when we understand how it came to be (Jespersen 19 24). At first glance, these two supporting positions appear to involve two opposite views of the relation of past and present. The uniformitarian doctrine says that things are pretty much the same. As far as BUILDING ON EMPIRICAL FOUNDATIONS 21 language is concerned, this means that the growth of literacy, mass media, rapid communications and the exposure of more people to standard languages have not altered the basic processes of change that affect linguistic systems. On the other hand, the historical doctrine holds that things are not the same: explanations based on universal principles of human nature or the relatively constant physiology of human beings are not enough to explain historical events. This implies that the initial conditions and environmental contexts of one set of linguistic changes are significantly different from another. The contradiction is resolved when we realize that the uniformity of the uniformitarian doctrine is at a relatively high level of abstraction: the processes and relations between processes that ultimately govern historical developments. This uniformity may itself rest upon the constants of human physiology and psychol ogy that are the explanatory resources of those who search for universals of language or language change. But in the type of theory envisaged by WLH, uniformity also proceeds from certain constant relations within the speech community, and the embedding of that community in a larger spatial and temporal matrix. There may be universals of language change, independent of historical conditions. Many theories of language change are devoted entirely to the search for them. But we will see that if such ahistorical generalizations exist, they are rare. This alone may account for the limited success of recent efforts to use formal techniques to develop general theories of language change.5 The alliance that WLH proposed between dialectology, sociolinguistics and historical linguistics is oriented towards a type of theory that would redress the balance between historical and synchronic explanation, correct ing the ahistorical bias of twentieth century general linguistics. The strategy of WLH then is to contribute to the general theory of language by way of a theory of language change. The central issues confronted by that theory of change may be usefully re-stated before we pursue the findings of recent research. 2. Paradoxes and problems of linguistic change A general theory is constructed from answers to general questions about the object of interest; these questions in turn depend upon an accepted definition of the object, usually cast in structural-functional terms. Language is defined in the broadest terms as a 22 WILLIAM LABOV system for transforming information held by one person into a form that can be symmetrically decoded and interpreted by others. The most general question posed about language is "How does the system work?". The" answer we are looking for is the mechanism that carries out this transformation. I will refer to this as the primary question for synchronic analysis. Since there are at first view many different languages that work in many different ways, a reasonable strategy is to begin with the primary question for a single language, "How does language X work?". Those interested in the general theory hope that a satisfactory answer to this question will be followed, as quickly as possible, by the answer for languages in general. The search for the simplest object of description — the homogeneous speech community or idiolect — was naturally initiated by those with the strongest general interests, rather than those who were primarily interested in the description of a particular language. Inherent variation creates serious problems for those who attack the primary question with the expecta tion of finding homogeneous structure. Given many alternative ways of saying the same thing, it is difficult for them to say how the language works or if it works at all. The consequent retreat to the idiolect and the rejection of the data from the speech community have been documented in WLH and Labov 1975. The empirical principles of normal heterogeneity and linguistic community permit us to transform the primary question in a way that resolves the paradoxes implicit in the pursuit of the idiolect. Answers to "How does language X work?" will include the ability of speakers to deal with the heterogeneous elements of language structure. This implies a more specific question: "How (and to what extent) do speakers of a language control forms that are not in their productive repertoire?" This reformulation of the primary synchronic question forms an essential background to our approach to a theory of language change. The existence of language change within a speech community creates even more serious problems for those who work with the expectation of a homogeneous structure. If language X is in process of change, there is in principle no one answer to the question, "How does language X work?". It works in several different ways. The first step in the construction of a general theory of this type is then not uniquely determined. As WLH pointed out, there is an implicit contradiction between the structuralist view of language and the facts BUILDING ON EMPIRICAL FOUNDATIONS 23 of change: as linguists became more and more impressed with the systematic character of language structure, it became increasingly difficult to conceive of change within that system. For a dynamic phenomenon like language change, the primary question is one of cause rather than function, not "How does it work"? but rather "Why did it happen"? or more precisely, "Why did it happen in this particular way"? It is not merely a question of the search for initiating factors; curiously enough, as the question is made more specific the answers become more general. A satisfactory general theory of language change would give an account of the antecedent conditions that determined the initiation, rate, direction and termina tion of a given change and eventually the set of such conditions for language change in general. Certainly there are many profound and challenging problems associated with the causes of change. We do not know why change is continually renewed, why changes freeze or come to an end, or why some processes drift in the same direction for millennia. After many decades devoted to the study of the causes of change, both Saussure and Bloomfield (speaking of sound change) came to the conclusion that the causes of change were unknown.6 WLH divided the primary question into more specific questions that are more amenable to an empirical attack. These will be restated in the second half of this section, and the sections to follow will report on progress made in answering these questions. But as WLH showed, the study of causes has always been heavily influenced by preconceived ideas about the effects of change that have dominated previous discussions of the issues. The "dysfunctional" character of language change. Beyond the primary question of the causes of change, there is the equally challenging issue of its effects. Given the structural-functional concept of language, it must follow that a language in process of change will not carry out its functions as successfully as a stable language. If some members of the community have different systems for encoding and decoding than others, their outputs will not be as readily and symmetrically decoded by those others. Most theories of language change have in fact assumed that change was dysfunctional. Throughout the 19th century, systematic language change, and sound change especially, was portrayed as a negative, even destructive 24 WILLIAM LABOV force. The causes proposed to account for change reflect this point of view: discontinuities in communication networks, the principle of least effort, generalizations of slips of the tongue, random drift of target positions. Change is envisaged as the result of interference from outside of the system: the system itself readjusts at times in compensation (analogical change). Curiously enough, this systematic adjustment is less systematic than the original interference from outside. Analogical changes rarely affect all categories and classes. Overt social correction is irregular in the extreme. Borrowing is unpredictable. Yet sound change, the malignant disease that these remedies would cure, is highly systematic. The negative character of linguistic change is particularly embarrassing for those interested in the larger evolutionary view of the matter. The parallels between biological and linguistic evolution are striking and inescapable. Darwin (1871) noted eighteen similarities between the two kinds of evolution: as for example, the effects of use and disuse; the existence of vestigial elements; the irreversibility of the extinction of a species or a language. But the most important homologue for Darwin was the existence of natural selection in both biological and linguistic evolution. He proposed to adapt Max Mueller's notion that words become better as they become shorter; but here linguists have failed to follow him. The consensus among linguists is that there is evolution in language in the sense of development and diversification, but not in the sense of progressive or adaptive evolution (Greenberg 19 59).8 If one could find evidence for adaptive developments in language change, the search for the causes of change would be advanced. But no such evidence has been brought forward, and the census is the reverse: language change is dysfunctional. We are left with a massive but in complete parallel between the two kinds of evolutionary development. The empirical principles of WLH permit us to resolve the paradoxical opposition of structure and change. Language change is not viewed as exterior to the system, but rather as an integral part of its normally hetero geneous character. The negative view of change is not accepted without question. WLH pointed out that no one has yet brought forward evidence of negative effects of lessened systematicity in the course of change. The functional or dysfunctional character of language change must be viewed as an empirical issue. A primary problem of linguistic change may then be identified: BUILDING ON EMPIRICAL FOUNDATIONS 25 "How does language change without interfering with communication between members of the speech community?" Ά functional perspective on the adaptive value of change. Throughout this discussion of language change, we will have many occasions to refer to the basic communicative function of human language: the representation of states of affairs. Though "reference" and "referential" are often used in this sense, these terms are perhaps too tightly associated with the application of terms or expressions to particular objects or events. I will use representational for this basic function, to include the capacity to represent immediate, distant, imaginary and even impossible configurations of our perceptual and conceptual experience. So far, the term "communication" has been used only in this fundamental sense, the communication of representational information, and language has been accepted as a device specialized for that function. If one were to take a sceptical position on this issue, and insist for example that language was no more functional than any other fashion like clothing or cosmetics, the problem of explaining language change would take a very different form. We could explain the continual changing of language forms by the same need for novelty in expressive forms that we observe in fashion. It is only when we insist that it is important for speakers and listeners to maintain clear differences between "I can" and "I can't", "He's tired" and "He's tight", that the problem of accounting for language change takes the acute form that it does. The primacy of the representational function in the organization of human language cannot be challenged. But the recognition of other functions in this organization is an important step in the resolution of the paradoxes of change. The possibility of an adaptive value of systematic linguistic change was all but rejected in the preceding discussion, but that discussion was limited to the framework of representational communication. The acceptance of normal heterogeneity within the community forces us to look more seriously at other functions. The alternate ways of saying "the same thing" which form the heterogeneous character of the community are representational sames, equivalent in this sense of meaning. One tendency in current semantic theory is to expand the range of meaningful distinctions to include degrees of emphasis, focus and style (Chafe 197 0, Bolinger 1973, Lakoff 1970). The direction of WLH is the opposite: to restrict the term "meaning" more 26 WILLIAM LABOV narrowly and preserve the fundamental concept of referential or representational "same" (Weinreich 1980: 270-271). Thus the variants available to members of the community are free to carry out the other functions in the three-membered configuration set out by Buehler (1934): Representational Expressive : : Directive Sociolinguistic studies have given concrete form to our knowledge of the expressive function, showing how the choice of variants identifies the speaker with certain social status, age group, sex and orientation towards the addressee. Speech act theory has begun to give more substance to our ideas about the directive function of language, including the choice of means for making and refusing requests, or supporting or challenging a relevant other. We add to these uses of language the other functions discussed by Jakobson (1957) and Hymes (1974: 22-3), including the aesthetic, phatic, metalinguistic, and playful. The problem of accounting for the effects of language change remains: how can change occur without interfering with the representa tional function? Yet it is evident that there are many possible ways that language could become better or worse adapted to human needs, beyond the communica tion of representational information. Pursuing this line of thought may give us more insight into other effects of change, and ultimately perhaps to a larger view of the evolutionary mechanism involved. 2.1 Five problems for the study of language change. The primary questions identified above will serve ultimately as a guide for evaluating a general theory of language change. But the work of constructing such a theory will necessarily rest on solutions to more particular problems. WLH set out five specific problems that a theory of change must solve; it may be helpful to restate them here and relate them to the primary
Compartilhar