Reconstructing Languages and Cultures 9783110867923, 9783110126716

Citation preview

Reconstructing Languages and Cultures

Trends in Linguistics Studies and Monographs 58

Editor

Werner Winter

Mouton de Gruyter Berlin · New York

Reconstructing Languages and Cultures

Edited by

Edgar C. Polome Werner Winter

Mouton de Gruyter Berlin · New York

1992

Mouton de Gruyter (formerly Mouton, The Hague) is a Division of Walter de Gruyter & Co., Berlin. © Printed on acid-free paper which falls within the guidelines of the ANSI to ensure permanence and durability. Library of Congress Cataloging in Publication Data Reconstructing languages and cultures / edited by Edgar C. Polome, Werner Winter. p. cm. — (Trends in linguistics. Studies and monographs : 58) Papers from a symposium on Indo-European studies sponsored by the University of Texas on the occasion of Professor Winfred P. Lehmann's retirement, held in Austin in Nov. 1986 with the IREX Conference on Linguistic Reconstruction. Includes bibliographical references and index. ISBN 3-11-012671-0 (acid-free paper) 1. Reconstruction (Linguistics) —Congresses. 2. IndoEuropean languages —Grammar —Congresses. 3. Comparative linguistics —Congresses. 4. Typology (Linguistics) — Congresses. 5. Indo-Europeans—Congresses. I. Polome, Edgar C. II. Winter, Werner, 1 9 2 3 III. IREX Conference on Linguistic Reconstruction (1986 : Austin, Tex.). IV. Series. P143.2.R4 1992 410 —dc20 91-45260 CIP

Die Deutsche Bibliothek — Cataloging in Publication Data Reconstructing languages and cultures / ed. by Edgar C. Polome ; Werner Winter. — Berlin ; New York : Mouton de Gruyter, 1992 (Trends in linguistics : Studies and monographs ; 58) ISBN 3-11-012671-0 NE: Polome, Edgar C. [Hrsg.]; Trends in linguistics / Studies and monographs

© Copyright 1992 by Walter de Gruyter & Co., D-1000 Berlin 30 All rights reserved, including those of translation into foreign languages. No part of this book may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopy, recording, or any information storage and retrieval system, without permission in writing from the publisher. Typesetting and Printing: Arthur Collignon GmbH, Berlin. Binding: Lüderitz & Bauer, Berlin. Printed in Germany

Preface

On the occasion of Professor Winfred P. Lehmann's retirement from active teaching in the Fall of 1986, a number of his friends and colleagues and former students met in Austin in early November 1986 in a symposium on Indo-European studies sponsored by the University of Texas thanks to the generous support of the Dean of the College of Liberal Arts. As this was also the time when the IREX Conference on Linguistic Reconstruction was scheduled to take place in the United States, it was arranged to combine the two meetings so that the Soviet scholars participating in the IREX-sponsored meeting on related problems would also be able to take part in the homage paid to Professor Lehmann who had been the chief American promotor of these productive scientific exchanges with the Academy of Sciences of the U.S.S.R. for at least two decades. As original theme for the Conference, the discussion of specific issues in Indo-European linguistics had originally been chosen so that a number of papers were devoted to the impact of non-Indo-European languages on Indo-European languages. There was also a particular focus on language and culture in a cross-disciplinary perspective involving archaeology, anthropology and comparative religion. The link with the IREXConference entailed the presentation of different papers on theoretical and practical problems of reconstruction involving morphology and syntax as well as etymology and the lexicon. Special emphasis was put on the role of typology in the process of reconstruction with reference to materials in Indo-European as well as in Amerindian languages. This accounts for the rich and stimulating diversity of the papers presented here. On the anecdotal side, it might be recorded that most of the Soviet participants arrived only a week later due to a snafu with regard to the obtainment of their visas. Thus, the main conference was held on November 4 — 6 with only Dr. T. Gamkrelidze as representative of the U.S.S.R., and a few days later, the other Soviet participants — Dr. V. N. Jarceva, Dr. V. P. Neroznak and Dr. Ju. S. Stepanov — were treated to a mini-symposium to which the Austin faculty and Dr. M. Mithun from Santa Barbara, California, contributed their papers again together with them. The tragic death of Dr. D. E. Francis, of the Department of Classics at the University of Texas at Austin, deprived us of the final version of

vi

Preface

his paper on which he was working when he passed away. We have retrieved all we could from his computer files and tried to present as faithful a picture of his work as humanly possible, but we are aware that what is brought here cannot reflect the whole richness and depth of his thought. We want, however, to submit his excellent paper as it is in homage to his memory. E. C. Polome

W. Winter

Contents

General papers on the problem of reconstruction Comparative linguistics Winfred P. Lehmann Comparative method, internal reconstruction, typology Henry M. Hoenigswald

3

23

Contributions from new data to the reconstruction of the proto-language Karl Horst Schmidt

35

Comparative reconstruction and typological verification: The case of Indo-European Thomas V. Gamkrelidze

63

Old and new thoughts about the configuration of the Romance language family Yakov Μ alkie I

73

Typology and deep genetic relations in North America Marianne Mithun

91

Specific papers on aspects of Indo-European reconstruction Armenian, Tocharian, and the "glottalic" theory Werner Winter

111

Reconstructing morphology: the role of o-grade in Hittite and Tocharian verb inflection Jay H. Jasanoff

129

Lexical entries in major sentence types of Proto-Indo-European Jurij S. Stepanov

157

viii

Contents

A syntactic typology of Indo-European languages (with reference to Celtic, Baltic, and Germanic languages) Viktorija N. Jarceva

185

A discourse perspective on syntactic change: Text-building strategies in Early Germanic Paul J. Hopper

217

Indian grammatical traditions and historical linguistics George Cardona

239

Philology and etymology, with focus on Anatolian Jaan Puhvel

261

Phrygian Vladimir P. Neroznak

271

Perspectives on Indo-European culture Archaeology and Indo-European comparative linguistics Homer L. Thomas

285

The traditional view of the Indo-European paleoeconomy: Contradictory evidence from anthropology and linguistics A. Richard Diebold, Jr.

317

Comparative linguistics and the reconstruction of Indo-European culture Edgar C. Polome

369

The comparison of formulaic sequences Calvert Watkins

391

Migration and linguistics as illustrated by Yiddish Robert D. King

419

Indo-European and non-Indo-European The impact of non-Indo-European languages on Anatolian Carol F. Justus The impact of non-Indo-European languages on Greek and Mycenaean E. D. Francis The impact of Dravidian on Indo-Aryan: An overview Andree F. Sjoberg Index of subjects Index of names Index of languages, dialects, and language groups

General papers on the problem of reconstruction

Comparative linguistics Winfred P. Lehmann

1. Comparison in linguistic study Comparison is the fundamental procedure in linguistics. In language, "values", to use Saussure's term, are determined by relationships. Comparison discloses those relationships and the values they establish. Procedures are the same if items like "categories" are selected as basic elements; categories too are identified, labeled and used on the basis of observed relationships. From the earliest known concern with language, linguists have accordingly examined and classified values by seeking out and comparing relationships. Comparative philology, the term used in the 19th century, accurately reflected the fundamental procedures in the science now known as linguistics. Comparison as the central procedure is not limited to the historical approach. Such an assumption simply reflects the dominant concern of the dominant linguists of the 19th century. Well-known theoretical statements illustrate the use of comparison in all linguistic concerns. Hjelmslev, for example, used Saussure's example arbre to demonstrate the differing values in the lexical sets: French arbre, bois,foret, German Baum, Holz, Wald, Danish tree, skov. Earlier, Panini began his grammar by comparing the three vrddhi vowels ä ai au with a e ο against i u r. Somewhat later, Plato has Socrates employ the same procedure in providing explanations, as for the word σώμα 'body', with which he compared σήμα 'tomb' and σήμα 'sign'. To return to another recent linguist, we note that Bloomfield's first assumption in his set of postulates reads (1926: 154): "Within certain communities successive utterances are alike or partly alike". Comparison determines the likenesses; it is the heart of his approach, as it is of anyone's involved with the analysis of language. As these statements with their examples illustrate, all of the investigators cited employ comparison regardless of the domain of their concern with language. Hjelmslev and Saussure employ it in lexical investigation; Panini in grammatical; Plato in etymological; Bloomfield in phonological and morphological. Differences in terminology, refinements in application, experience derived from previous study may well have led to ad-

4

Win/red P. Lehmann

varices beyond Panini, Plato and Saussure. For example, Hjelmslev used the term "commutation" rather than "comparison" for his determination of values. Yet it would be difficult to demonstrate that the fundamental procedure has been altered.

2. Comparative linguistics of the present Today the use of the term "comparison" and of the phrase "comparative linguistics" has been sharply restricted. The terms are not used for treatment of one language at one period; indeed they even seem inappropriate for application to such linguistic study in view of terminology now in widespread use. For study of a specific language a succession of special labels has been in vogue, from "synchronic" through "descriptive", and today most oddly "theoretical", as if all scientific concern were not theoretical. (To be fair to linguistics, chemistry has manipulated the term "theoretical" with comparable arbitrariness, applying it to computational chemistry.) In the last decades of the 20th century then we do not apply the term "comparative linguistics" to synchronic study even though it too uses comparison as its basic procedure. We apply it only to three other prominent and well-established ways of studying language, first briefly identified here, and then discussed at greater length with reference to their status and requirements. Any attention to comparative linguistics does well to start from Meillet's admirable lectures presented to the Institute for the Comparative Study of Civilizations, Oslo, 1924. The first paragraph reads (Meillet 1967: 1): "There are two different ways of practicing comparison: one can compare in order to draw from comparison either universal laws or historical information. These two types of comparison, equally legitimate, differ absolutely". After illustrating the differences between these "two types" with examples from the study of myths by the "young French scholar, Dumezil", Meillet directs his attention to the "historical and comparative method." Lingering correlation of the term "historical" with "comparative" helps maintain the notion that comparison is restricted to historical linguistics. The notion is fortified by the quotation from Saussure cited by Gamkrelidze and Ivanov at the beginning of their epoch-making work: "If the sole means of reconstruction is by comparing, conversely comparison has no purpose other than reconstruction" (1984: 2). Surely Jarceva is more accurate in saying that "the aim of comparison is to reveal the number

Comparative

linguistics

5

of markers among the objects compared" (Yartseva 1986: 21). Comparison in her view is not limited to historical ends, but it applies to any linguistic analysis. In keeping with the separate position assumed by many for synchronic study, Klimov (1986:108) identifies three "main branches" of comparative linguistics: genetic, typological, areal. Meillet's statement on the aims of the first two identifies them clearly: the historical approach is equivalent to Klimov's "genetic", the typological is a universal approach. Meillet did not discuss "areal", though he touched on it in dealing with mixed languages. Since the appearance of Sandfeld's monograph on the Balkan languages, six years after Meillet's lectures were given, areal linguistics has occupied the same plane as typological and historical linguistics. The distinction between areal and typological study is one of breadth. Both, in Jarceva's words, look for markers. The scope of typology extends to all languages; that of areal linguistics to those in a selected territory, a territory open to ready intercommunication. For various reasons that intercommunication is not extended to additional territory, whether because of the extent of a common culture, as in Northern Europe and the Asian sub-continent, or because of the extent of trade, as in the Balkans and the northwest coast of America. Both reasons probably interact in any identifiable area. In view of the distinction between areal and typological study largely by the extent of their scope, the activities and conclusions of each may be confused, as we note to some extent below. Before examining the current situation of the three identified "branches", a note on the position of "typological" with regard to synchronic linguistics is in order. To clarify our position it may be useful to cite once again Whitney's statement (1892: 6) on the procedures of linguistic study: "Every fact of every language, in the view of the linguistic student, calls for his investigation, since only in the light of all can any be completely understood. To assemble, arrange, and explain the whole body of linguistic phenomena, so as thoroughly to comprehend them in each separate part and under all aspects, is his endeavor" (1892: 6). In concentrating on features over the wide span of languages, typological linguistics sets the same aims for itself as does synchronic (descriptive / theoretical) linguistics. Typological linguistics seeks to determine the basic entities or categories of language in their relationships to other entities and categories on which language is built. The appearance of a difference between synchronic and typological linguistics may result from the restricted attention in recent synchronic study to a single language which has had such a devastating

6

Winfred P. Lehmann

effect in nullifying much of the work in the structural and generative approaches. Yet their professed aim is that of typological linguistics — the search for the fundamental features and processes of language, Meillet's universal laws. Typological linguistics understandably posits language types as "abstract models" for the comparison of languages, in Klimov's words (1986: 105) — a procedure which a monolingually based linguistic approach cannot take. Yet this procedure simply provides a ready summation of findings in relation to the fundamental features and processes of language. Both pursuits are in their infancy, as is other linguistic study. No language has been exhaustively described, so that one overall hypothesis for language, generally labeled theory, has replaced another, even of the same originator. Each branch of study also produces limited theories, such as the laryngeal theory in Indo-European linguistics or the theta theory in generative linguistics. For well-based typology, we require capable descriptions of hundreds or thousands of languages. Nonetheless, we can and do propose principles, and generalizations under the label "universal". Recognition of shortcomings may stimulate some of the humility which supposedly characterizes scientists; it may also provide incentives for students through the knowledge that they can bring somewhat greater certainty to the broad and hitherto inadequate study of language.

3. Historical linguistics Examining each of the three branches of comparative linguistics in turn, we may begin with historical linguistics, by far the most highly developed branch of the three. Comparison with historical aims was already at a high level in Grimm's correlations of 1822. His application of it was further remarkable because of his statement of the correlations by means of "rules", a far better label than the subsequent misleading term "laws". (The term "law" should be abandoned, partly because of its erroneous implications, partly because of recent excesses in applying it to a minor set of correlations which fail to account for all the examples concerned. On the other hand, devotees of laws might elaborate on their procedures, like the last followers of Ptolemy, and set out to produce epi-laws.) Grimm's procedures are also admirable because of his exhaustive listings. Moreover, he listed the examples which did not fit his correlations, in

Comparative

linguistics

7

this way preparing the ground for greater refinement, as by Grassmann and Verner. It is also remarkable that Grimm's formulations of 1822 are not made obsolete by the glottalic theory. By that theory the reconstructions which others proposed on the basis of Grimm's rules are modified, though not the rules themselves. Continuation of their validity over more than a century and a half is adequate testimony to the advanced status of historical comparison even in Grimm's day. Nonetheless, refinements in the use of comparison practiced by Grimm for historical purposes are among the important contributions of the 19th century. Included in elementary handbooks, they do not need presentation here. Of primary importance is the focusing of attention on more and more precise patterns, as with the help of increased knowledge of the several types of phonetic study. Of similar importance is the extension of application throughout all entities of language — to syllables rather than merely single sounds, to sequences of syllables or words, to suprasegmentals; to features of morphology, syntax, discourse patterns, the lexicon, to the semantic and pragmatic spheres. Clearly the application to some of these entities and features, such as syntactic, is not as welldeveloped as to others. But progress is being made. For the phonological and morphological entities we can point to the excellent statements of Hoenigswald. Citing a large number of earlier works, he examines in detail various patterns of phonological and morphological change, then illustrates the application of the comparative method to these (1960). His subsequent treatise supplements this text and its bibliography (1973), bringing comparison as applied to phonology and morphology to a high level.

4. Problems for historical linguistics Yet even this highly developed use of comparison includes two notable shortcomings: restriction to one family of languages; lack of attention to syntax and discourse. Hoenigswald's examples are taken from IndoEuropean languages. These of course have been most thoroughly explored. One might hope that they provide representative patterns and processes covering all languages. Yet recent publications suggest that reliance on examples in Indo-European alone has deprived even IndoEuropeanists of adequate insight for the solution of their problems. The glottalic theory could scarcely have been proposed without attention to

8

Winfred P. Lehmann

data in other language families; and the widespread acceptance of the laryngeal theory today may be ascribed in part to increased information about back consonants in such language families as the Afro-Asiatic and the Caucasian. Further, the presentation of the pattern of comparison of inequality by Gamkrelidze and Ivanov (1984) may finally on the basis of patterns in OV languages oust from our handbooks the reconstructions of an Indo-European comparative of inequality from one of the patterns in early Greek and Sanskrit. Broader experience with other patterns and processes may also help solve additional problems in Indo-European study, such as the development of ablaut. And apart from the contributions to Indo-European studies, such experience will provide the basis for a secure, general use of comparison for genetic purposes. A second shortcoming stemming from the limited scope of linguistics is the inadequate control of syntax and discourse. The patterns of key Indo-European languages have long been known and described. Priscian's syntax of Latin is said to be the work for which the largest number of manuscripts survive from the medieval period. But as I have pointed out before, the thorough descriptions of Latin, Greek and other Indo-European languages available to Hermann Hirt were not adequate to stifle his unhappy comment on syntax two years before the end of a long career: "One assembles a series of facts but doesn't know what to do with them" (1934: vii). Explanation, the highest goal of linguistics for Whitney and many other linguists, requires a framework, which will be adequate only if it is constructed on the basis of a thorough knowledge of the possibilities of human language. A syntactic framework is gradually being assembled, if over bodies of reluctant Indo-Europeanists; it will be followed by frameworks for discourse, semantics and pragmatics.

5. Typological linguistics Achievement of such a framework, with understanding of its functioning, is the aim of comparison for "universal laws", known as typology. Before examining the status of typology we may note that students of language have sought other means of explanation for patterns of language than a framework resulting from the study of language. We may pass over the attempts at divine explanations and note only those that seek a framework through hypotheses on the functioning of the mind. These have foundered in part because of primitive conceptions of that functioning proposed with inadequate knowledge of the functioning of the brain, in infancy

Comparative

linguistics

9

and in later stages of its development. Bloomfield's awakening to the skewed framework for linguistics achieved from the functioning of the mind as imagined by Gestalt psychology did not prevent generative transformationalists from a renewed failure. Little survives from such hypotheses but awkward terminology, such as left-branching or rightbranching, as if the mind constructed a bidimensional matrix when controlling language. The abandonment of the well-funded, massively pursued, and picturesquely presented generative-transformational approach by its founder and principal students may be adequate indication of the inevitable failure of proposing a theory of language without thorough knowledge of many languages and their use by society. As Hawkins (1989: 119) puts it: "The major challenge facing any general linguistic theory is to extract the universale from the enormous range of variation exhibited by the 5000 or so currently spoken languages". Typological linguistics sets out to assemble such knowledge, to formulate it, and to use it in providing explanations for patterns and processes of language. While there may be agreement on the aim of linguistic typology, the expectations differ considerably. As one of the most active areas of linguistic concern, the various views cannot be fully treated in a short sketch, only major positions and findings. First, it is important to recall that typological comparison has long been carried out. Typological comparison was a dominant interest of the 17th and 18th centuries, encouraged by Catherine the Great and Leibniz among others. Something of a culmination of over a century of study was published by Adelung in his Mithridates (1806 — 1816). As his readily accessible statements on Turkish illustrate (Lehmann 1978: 27 — 29), Adelung dealt with segmental and suprasegmental phonology, inflectional and derivational morphology, and with syntax. The items singled out in these spheres fall under a scheme applied in Mithridates to other languages as well, but they were not correlated under any one principle. Adelung did not propose universals.

6. Problems for typological linguistics Shortly after the publication of Mithridates Wilhelm von Humboldt formulated such principles, as the first in line of distinguished linguists of the 19th century who concentrated their efforts on typology: Steinthal, Friedrich Müller, Finck, Lewy, Pater W. Schmidt. In their attempts to formulate general principles they directed their attention at morphology.

10

Winfred P. Lehmann

Their attempts to extract through such attention an overall framework, such as Finck's eight types, or a general principle such as Schmidt's reliance on the position of the genitive, are commendable but inadequate. Other matters strike us as more significant; they are also controversial. Probably first among these is the problem whether one overall principle governs the structure of a given language. Some, for example, Klimov, make such an assumption (1986: 105 — 10); others, for example, Timberlake, reject it (1986: 103 — 104). In arriving at a position we must note that language is not a physical object, like a crystal, with clearly regulated structure. Nor is language an organism, however persistent the notion and the label, even though organisms, like languages, have more complex structures than do physical objects. The complexity of any one organism, by no means yet fully mapped or understood, stems in part from structural principles, in part from genetic regulation, in part from modifications by a given environment. As a social structure, language may well be similar to organisms in being governed by structural principles, though regulation resulting from source and environment may be expected to be far greater if only because, in Hamlet's words, language unlike flesh is not solid. Moreover, there is little doubt that language is in part modified in accordance with its function. Sorting out the characteristics governed by structural principles or universale, by historically determined configurations, by areal and functional influences is a central requirement of typology. 6.1 Typological findings in attempts to determine areal versus historical influences Rather than proceed to a statement on the role of each of these forces, I cite two examples illustrating different positions on their roles. The first is Andersen's recent book on comparative constructions (1983). Concentrating on examples from Indie, taken from the Rigveda and extending through Middle Indie, Andersen accounts for the patterns which he found as areal phenomena. By far the majority of his examples from the Rigveda have comparative constructions like those of Japanese, Turkish and other OV languages. Though Andersen accepts the fact that Proto-Indo-European was an OV language, he dissociates the comparative construction from a general framework, in this way arriving at his analysis. One cannot deny that the Indian area has comparable constructions; the major nonIndo-Aryan family, Dravidian, is consistently OV. Yet an examination of

Comparative

linguistics

11

other early Indo-European languages, Hittite, Greek, Latin, even Old English, requires us to posit an OV comparative construction for ProtoIndo-European, as I indicate in a review of Andersen as well as earlier (1974, 1987). Gamkrelidze and Ivanov agree with my position. The position holds that the comparative construction in the Rigveda is in accordance with the structure of the language, a structure historically determined. Yet the proposal of an areal explanation in Andersen's extensive study illustrates that much attention needs to be given to the role of historical versus areal forces in molding characteristic elements of language, as well as to the role of structural coherence. 6.2 Typological findings

in determining functional

effects

The second example has to do with functional influences on characteristic elements. In the study concerned, Hawkins deals with three patterns which accord with VO structure more than would be expected in OV languages if these were generally consistent: 1) relative clause positioning; 2) sentence complement positioning; 3) favoring of suffixation over prefixation. In a careful study drawing also on observations of others, Hawkins demonstrates that functional considerations, or what he calls "user-based performance", affect "system-internal" principles. In short, "different principles are often in partial conflict, or they may be gradient rather than absolute ... [complicating] the precise predictions that they make" (1989: 152). Hawkins does not reject the primacy of structural coherence for language. What he concludes on the basis of a relatively large sampling of languages, is the proposal that some characteristic patterns are favored over others for their functional appropriateness. Without extending comment unduly we may point out that the appropriateness depends on the point of view of the evaluator. For Hawkins (1989: 145) a postposed complement is functionally superior to a preposed one because it accords with a "fundamental characteristic of the human parser: the need to match arguments with their appropriate predicates quickly and readily while building a semantic interpretation of the whole sentence". Hawkins employs argumentation here that accords with the referential and conative functions of verbal communication, to use Jakobson's designations. These may well be predominant for the use of language in our culture, and even in most cultures. But when one reads Skaldic verse, or Classical Latin verse and prose, questions arise about the emotive and poetic functions of verbal communication. Disturbingly

12

Win/red P. Lehmann

preeminent for modern ears in the literatures concerned, might not the preference for these functions enjoy greater prominence in some cultures? Different motives may lead users of conative language, even in cultures whose primary concern is not emotive and poetic, to patterns which do not accord with strict functional aims as Hawkins sees them. Anyone who has dealt with scientific German and its fondness for preposed relative expressions through use of the participial modifier construction, or even with the language favored in German newspapers may wonder whether a human parser necessarily objects to "delay and interruption of the main clause" (Hawkins 1989: 145). 6.3 Background and basis of typological

generalizations

To clarify the interplay among the various forces at work we need additional scrutiny, especially of typological principles. Yet we cannot assume that they are recent generalizations, generalizations made on restricted examination of language. I have pointed out before that Pater W. Schmidt's analysis of two major language types in accordance with the position of the genitive with regard to its head is equivalent to proposing two types in accordance with the position of the object in regard to its verb. Even with this equivalence, centering on the clause rather than on a nominal phrase provides a much more satisfactory base for a language typology, in view of the role of the sentence as the minimum free form of language. Moreover, Schmidt's massive data, as well as the assemblages of data by other 19th century linguists, such as Codrington, maintains its validity in the parallel explanation of linguistic features. Accurate discrimination among the forces named above will result only from accurate analysis of reliable data, as Hawkins suggests. Yet we must also examine languages of the past, not simply those currently spoken, in part for their additional data, in part for reliable historical principles. We can no longer cite random examples in support of our historical explanations and reconstructions, as was done in the 19th century. Some time ago I pointed to the use of such examples by two eminent scholars proposing diametrically opposed accentual examples in their explanations for Holtzmann's law (Lehmann 1952: 37 — 38). It would be difficult to find stronger evidence of the need for determining typological principles in historical study. But those principles must be based on adequate data. Secure historical principles must be derived largely from three language families: Afro-Asiatic, Indo-European, Sino-Tibetan, supplemented by

Comparative

linguistics

13

less well-attested historical evidence in such families as the Dravidian and Finno-Ugric. Even in these five families, the data need scrutiny, with the Indo-European most securely described. If other approaches were not published in reputable journals, they might be disregarded. In the interests of serious scholarship I comment on them here, in contrast with earlier reluctance. Miller (1979: 310) has cited the Old Turkish pattern of teen numerals in supposed refutation of a generalization I made on the basis of a number of well-attested OV languages. The structure of the Old Turkish pattern cited is totally different from the additive pattern I discussed. Moreover, we have no information on the prehistory of the Old Turkish pattern. There is little further point in dealing with Miller's statements, other than to cite them as examples of inferences from data which themselves need explanation, lacking as they do data from an earlier period. Miller continues with an interpretation of syntax based on etymological analysis. One sympathizes with the wish to derive general principles from other families than the Indo-European, but the procedures applied must have some semblance to scientific reliability. 6.4 Foundation of linguistic standards in historical or in typological linguistics Treatment of the development of nominal inflection in Proto-Indo-European may illustrate roles of these two kinds of comparison. In 1958 I proposed that the basic endings -s, -m, -h at an earlier stage indicated respectively an individual, the result of an action, a collective, simply setting up the three endings as the core declension to which other suffixes with specific features, like -i were later added. In the earlier period, -s could also indicate a possessive relationship under appropriate intonational patterns; for this I cited support from Hittite. In their monumental treatment of Proto-Indo-European, Gamkrelidze and Ivanov posit the same endings, ascribing to -s the value of animate/ active, to -m the value of inanimate/inactive. From this contrast they derive the later paradigm of nominative and accusative, as well as genitive, supplementing my analysis by deriving the genitive plural as well as the singular from the pertinent early markers. They also link the s j m contrast in the noun with verbal contrasts. Most important, they lend support to this reconstruction of the earlier grammatical contrast by identifying Proto-Indo-European at the stage in question as a "stative-active lan-

14

Winfred P. Lehmann

guage," as discussed by Klimov (1977) and others, including Sapir. By identifying this earlier stage with a well-recognized type of language, Gamkrelidze — Ivanov (1984: 267 — 319) add considerable credence to this view of the development of Proto-Indo-European. At the conclusion of his brilliant essay: "Proto-languages as objects" (1980/1986), Ivanov ascribes greatest success in linguistic study to correct choice of a system to serve as linguistic standard, adding that the most promising standard is the proto-language of any given family. I would like to suggest that the standard should not be determined by historical, but rather by typological comparison, as the foregoing example indicates. Historical comparison, as Meillet pointed out, yields a particular standard, typological comparison, however, universal standards.

7. Areal linguistics The increased concern for areal linguistics illustrates the importance of data in proposing generalizations. Having identified use of comparison only for typological and historical study, Meillet in his 1924 lectures did little more than point to borrowing across languages in his chapter on mixed languages (Meillet 1967: 101 — 102). Then Sandfeld's impressive study of the Balkan area left no hesitation about influences across language boundaries, not only distantly related languages like Greek and Albanian, but also influences across unrelated languages like these, the Slavic languages and Turkish. Beside lexical patterns, Sandfeld found common to the Balkan languages features such as postposed articles, the patterns of future formation and complementation.

8. Problems of areal linguistics After Sandfeld's monograph there could be no doubt that neighboring languages influence one another regardless of their family relationships, if their speakers maintain communication across linguistic boundaries, often accompanied by widespread bilingualism. While areal influences are therefore without question, data from the past are essential to determine the direction of diffusion or also to determine whether historical relationship rather than diffusion is the basis of common features. A well-known difference of view on these matters was held by Sapir and Boas on the Northwest languages of America. Sapir, while placing

Comparative

linguistics

15

high value on genealogical determination, admitted diffusion of words, sounds and "even morphological elements." But he rejected "borrowing [of] fundamental features of structure, hidden away in the very core of the linguistic complex" (Sapir 1921: 219). Among tasks of areal comparison is the determination of these. That task was one of the goals of an influential article by Emeneau on the linguistic area of India (1956). In the article he concentrated on the spread of numeral classifiers, after examining features pointed out earlier by Bloch (1934/1965). Among Bloch's features which Emeneau considered significant are: distinct stems for singular and plural forms of nouns; use of the gerund in the first clauses of a multi-clause sentence, which is concluded by a finite verb; echo words of a special structure. All of these Emeneau ascribed to Dravidian origin, as he also did the areal characteristic of retroflex consonants in the Indie linguistic area. But he also pointed out that this use of the gerund is found in many other languages, among them Japanese, Korean, Altaic, Finno-Ugric, also Vietnamese and Chinese, without at the same time discussing the presence of such gerund constructions in OV languages. However one judges the features concerned, that is, whether one should agree with Emeneau in applying Sapir's adjective "profound" to them, an evaluation procedure to distinguish which features are "fundamental" or "profound" can only be determined from typological study. The extent of receptivity of speakers to borrowing on the other hand is a sociolinguistic concern, also requiring study, as elaborated by Pandharipande in his attention to the contact situation in Central India (1982). 8.1 Areal in relation to typological

linguistics

Procedures used in determining areal influences are like those used in seeking out the characteristic patterns of typological linguistics, except that the languages scrutinized are limited to those of selected territories. The results then do not concern the entire language system; they may actually be limited to minor characteristics of language, such as numeral classifiers. A collection of patterns spread under areal influence would be useful, with reference to the linguistic systems in which they have been adopted. Sherzer provides such a collection for American Indian languages north of Mexico, limiting his scrutiny to segmental phonemes and to morphemes. The problem of identifying whether patterns are areal or typo-

16

Winfred P. Lehmann

logical/universal is apparent from the title of his book. Yet with further knowledge of the languages investigated, such identification is not crucial. For even the near-universal, or "categories ... probably universal in language, in that they can be and are at times expressed in some manner in every language" (Sherzer 1976: 17) may be so clearly restricted to the languages of an area, and absent from related languages in other areas, that they may be identified as areal. A striking example among Sherzer's illustrations of probable universale is the "category" of numeral classifiers, which Emeneau used so confidently in his highly regarded paper (1956). Their absence in the other Indo-European languages gave Emeneau assurance on their areal inclusion in the Indo-Aryan languages. However widespread, like other categories listed by Sherzer, among them nominal case system, possession classes, diminutives, they can be used for purposes of areal classification if they are adequately identifiable for special presence in a given territory. Areal and typological/universal study may then make use of the same features, categories and processes. The cogency of identification under either rubric will be determined by adequacy of information and adequate rigor in discrimination by linguists. 8.2 Areal in relation to historical linguistics Identification of features resulting from areal influences as opposed to genealogical also requires adequate information and accurate discrimination. The information may be adequate, as in identifying the OV comparative of early Indie as genealogically determined. Identification is supported by the preponderance of OV comparatives in the earliest Indie texts coupled with the parallel pattern in the earliest related language, Hittite, as well as survivals in the most conservative texts of such lateattested languages as the Germanic. But the data may be far less ample, as for the American Indian languages of the northwest coast, where Boas and Sapir disagreed on the interpretation of the crucial common characteristics. Sapir's "fundamental features" had unfortunately not been identified at the time, nor are they now. Yet recent work on the Nostratic hypothesis has assumed a set of 25 lexical items which is held to be relatively stable, and hence indicative of genealogical relationship; that work has also proposed procedures for determining "the varying degrees of consonantal stability and the regular pattern of evolution for certain types of consonants"

Comparative

linguistics

17

(Shevoroshkin — Markey 1986: xvii — xviii). The assumptions are welcome, but they need thorough testing. For example, of the 25 proposed stable words, 'tongue' and 'hand' have been notoriously considered to be affected by taboo in the Indo-European family; others, such as 'blood', 'moon', 'sun' are not without problems. Yet if testing upholds the validity of the techniques and assumptions applied in work on the Nostratic hypothesis, we will have possibilities for discriminating historically determined from areal characteristics, providing welcome access towards a solution for identifying "the very core of the linguistic complex".

9. Comparative linguistics today Earlier sections have discussed comparative study carried out in accordance with specific procedures: examination of historical comparability; of comparability across languages whatever their historical background; comparability in specific areas. These procedures introduce artificial distinctions into the study of language, leading to the three varieties of comparative linguistics discussed, beside which one could name others. The common aim of all such approaches to the study of language is presumably that of Wilhelm von Humboldt, among other scholars: "in die Natur der Sprache überhaupt einzugehen" 'to concern oneself with the very nature of language'. As Humboldt (1836: cxx) stated, in pursuing that aim linguists confront an infinite quantity of details that confuse their course of action. Some seek shortcuts, in part by proposing a social or a mental overlay which makes it unnecessary to be concerned with the details, in part by assuming that one language family or even one language provides sufficient evidence for determining "the very nature of language." As noted above, attempts to buttress hypotheses based on such shortcuts with supposed mental processes only leads to further self-deception. Any attention to statements by specialists on the functioning of the brain illustrates the emptiness of such suppositions more than half a century after Bloomfield came to that conclusion (see for example Thompson 1986). Further, increasing knowledge of the changing biological architecture and the changes of the brain in early years, as by PET scanners, indicates the groundlessness of current and earlier theorizing in the abstract about language acquisition. There is no alternative to assembling and arranging linguistic data before proposing explanations.

18

Winfred P. Lehmann

Assembled data are appropriately arranged through comparison, comparison of the "details" in one language to determine its essential features, comparison of these features with the essential features of other languages, present or past, to determine general patterns, processes and principles known as universals. Linguists face hazards in each of their successive tasks. The "details" may be inadequately or erroneously noted. To forestall problems, linguists dealing with written records look to philological techniques, linguists dealing with oral records to anthropological techniques. In their further task, the determination of essential features, they may not draw on adequate data, they may draw on suspect data, they may evaluate the data faultily. Inadequacies may be introduced in moving to further abstractions, to so-called explanation. At this point, linguistics as a collaborative effort advances through the pointing out of errors and inadequacies. Without denial of their contributions, the inadequacies of the 19th century in concentrating on historical comparison illustrate previous shortcomings of our field, which commonly cites Humboldt but disregards his views. Humboldt (1836: CXX) aimed to seek out the essence of language through "individuell historische Sprachvergleichung". This phrase was unfortunately rendered by Buck and Raven, in their valiant attempt to translate Humboldt, as "individual, historical, and comparative linguistics" (1971: 69), as though Humboldt had three activities in mind. He did not. His aim was one "correct and energetic approach to penetrate" the essence of language, an aim that obviously has not been realized. Instead of an "individuell historische Sprachvergleichung" there have been many. One enjoying wide appeal has tried to shortcircuit the difficulties by positing a black box inside which the essence of language is genetically implanted in human beings. As a result of such unsupported assumptions linguistics is far behind the natural and biological sciences. The advanced status of these sciences is due to two methods, the orderanalytical and the causal-analytical. By the order-analytical, structures are compared for similarities and differences to determine an order. By the causal-analytical, experiments are performed to seek cause-effect relationships to assist in determining such an order. Introduced to the physical sciences by Galileo, the causal-analytical was eventually applied in the biological sciences, as by Gregor Mendel. Limited in biological study first by the conservatism of specialists, then by humanistic considerations, it has not been applicable in linguistics. The frequent reference to Herodotus's account of the experiment by Psammeticus illustrates its

Comparative

linguistics

19

inapplicability, and probably also the envy linguists may have of colleagues more fortunate in their methods. In short, linguistics is limited to use of the order-analytical method with its reliance on comparison. Impatience with the necessary observations, with the drudgery involved in observing and analyzing data to determine order has led to approximations of the causal-analytical method. Of these approximations, that arousing the greatest hope is participation in language acquisition, whether of the initial language or of further languages. Simple observation of acquisition provides information for comparison, but participation permits experimentation to some extent. Such activity merits applause when it is carried on patiently, without the unfounded leaps that bring undue expectation and also notoriety in a field overwhelmed with data and phenomena. But the activity has done little to lead us to Humboldt's goal. Similar hopes have been raised in attention to pidgins and Creoles. Yet shortcircuiting has not been avoided in attention to these either. In Seuren's (1985: 4) words: "creole linguistic universals ... have been proposed ... Yet, although the phenomena in question are no doubt highly intriguing, there are either too many counterexamples, or there is not enough reliable information available on the actual facts". That is, creolists need to assemble and compare data. It may be mentioned that experimentation of a severely restricted kind is possible with computers, though the grammars and lexica so far stored in them are merely fractions of those controlled by users. Yet in the interest of completeness, this possibility of applying the causal-analytical method in our field cannot be overlooked. Finally, however dismaying the restriction, linguistics of every subclassification must rely on the order-analytical method. The great contributors to our field, Whitney, Brugmann, Meillet, a m o n g others, recognized that limitation and produced works of lasting benefit by not attempting end runs around the labor necessary in applying the method. The method has yielded many achievements and great advances in the physical and biological sciences. Linguists will do well to apply it widely, refining the tools of comparison through careful studies like those presented at this conference, and testing the results against standards derived from typological study, in the continued effort to proceed towards the goal of understanding the essence of language. References Adelung, Johannes Christoph 1806—16 Mithridates oder allgemeine

Sprachkunde

I —III (Berlin: Voss).

20

Winfred P. Lehmann

Andersen, Paul Kent 1983 Word order typology and comparative constructions (Amsterdam: Benjamins). Bloch, Jules 1965 Indo-Aryan from the Vedas to modem times. Trans, by Alfred Master (Paris: Adrien-Maisonneuve). Bloomfield, Leonard 1926 "A set of postulates for the science of language", Language 2: 153 — 164. Buck, George C. — Frithjof A. Raven (transl.) 1971 Linguistic variability and intellectual development. Translation of Über die Verschiedenheit des menschlichen Sprachbaues... by Wilhelm von Humboldt (Coral Gables: University of Miami Press). Codrington, Robert 1885 The Melanesian languages (Oxford) [reprinted: Amsterdam: Philo, 1974], Ellis, Jeffrey 1966 Towards a general comparative linguistics (The Hague: Mouton). Emeneau, Murray B. 1956 "India as a linguistic area", Language 32: 3 — 16. Gamkrelidze, Tamaz V. — Vjaceslav V. Ivanov 1984 Indoevropejskij jazyk i indoevropejcy (Tbilisi: Izdatel'stvo Gosudarstvennogo Universiteta). Grimm, Jakob 1822 Deutsche Grammatik I2 (Göttingen: Dieterich) [reprinted under direction of W. Scherer, Gütersloh: Bertelsmann, 1870]. Haarmann, Harald 1976 Aspekte der Arealtypologie. Die Problematik der europäischen Sprachbünde (Tübingen: Narr). Hawkins, John A. 1989 "Competence and performance in the explanation of language universals", in: Doug Arnold et al. (ed.), Essays in grammatical theory and universal grammar (Oxford: Oxford University Press), 119 — 152. Hirt, Hermann 1931 —34 Handbuch des Urgermanischen 1—3 (Heidelberg: Winter). Hjelmslev, Louis 1970 Language. Trans, by Francis J. Whitfield (Madison: The University of Wisconsin Press). Hoenigswald, Henry M. 1960 Language change and linguistic reconstruction (Chicago: The University of Chicago Press). 1973 Studies in formal historical linguistics (Dordrecht: Reidel). Humboldt, Wilhelm von 1836 Uber die Verschiedenheit des menschlichen Sprachbaues und ihren Einfluss auf die geistige Entwickelung des Menschengeschlechts (Berlin: Dümmler). Ivanov, Vyacheslav V. 1986 "Proto-languages as objects of scientific description", in: Vitalij V. Shevoroshkin — Thomas L. Markey (eds.), 1—26. Jakobson, Roman 1960 "Linguistics and poetics", in: Thomas A. Sebeok (ed.), Style in language (Cambridge: M.I.T. Press), 350-377.

Comparative linguistics

21

Klimov, Georgij A. 1986 "On the notion of language type", in: W. P. Lehmann (ed.), 105 — 110. Lehmann, Winfred P. 1952 Proto-Indo-European phonology (Austin: University of Texas Press). 1958 "On earlier stages of the Indo-European nominal inflection", Language 34: 179-202. 1974 Proto-Indo-European syntax (Austin: University of Texas Press). 1978 "The great underlying ground-plans", in: W. P. Lehmann (ed.), 3 — 55. 1986 Language typology 1985 (Amsterdam: Benjamins). 1987 Review of Paul Kent Andersen (1983), Lingua 70.1: 2 0 0 - 2 0 5 . Lehmann, Winfred P. (ed.) 1978 Syntactic typology (Austin: University of Texas Press). Meillet, Antoine 1967 The comparative method in historical linguistics. Translated by Gordon B. Ford, Jr. (Paris: Champion). Miller, Roy Andrew 1979 "Japanese, Altaic, and Indo-European", Journal of Indo-European Studies 7: 307-313. Pandharipande, Rajeshwari 1982 "Counteracting forces in language change: Convergence vs. maintenance", Studies in the Linguistic Sciences 12: 97 — 116. Ramat, Paolo 1986 "Is a holistic typology possible?", Folia Linguistica 20: 3 — 14. Sandfeld, Kristian 1930 Linguistique balkanique. Problemes et resultats (Paris: Champion). Sapir, Edward 1921 Language (New York: Harcourt, Brace). Saussure, Ferdinand de 1959 Course in general linguistics, edited by Charles Bally and Albert Sechehaye, translated by Wade Baskin (New York: Philosophical Library). Schmidt, Pater Wilhelm 1926 Die Sprachfamilien und Sprachenkreise der Erde (Heidelberg: Winter). Seuren, Pieter Α. M. 1986 "Semantic transparency as a factor in Creole genesis", in: Pieter Muysken —Norval Smith, Substrata versus universals in creole genesis (Creole Language Library 1) (Amsterdam: Benjamins), 57 — 70. Sherzer, Joel 1976 An areal-typological study of American Indian languages north of Mexico (Amsterdam: North Holland). Shevoroshkin, Vitalij V. —Thomas L. Markey (eds. and transl.) 1986 Typology relationship and time (Ann Arbor: Karoma). Thompson, Richard F. 1986 "The neurobiology of learning and memory", Science 233: 941 —947. Whitney, William Dwight 1892 Language and the study of language5 (New York: Scribner). Yartseva, Viktoria Ν. 1986 "Basic typological units", in: W. P. Lehmann (ed.), 1 9 - 2 6 .

Comparative method, internal reconstruction, typology Henry M.

Hoenigswald

The "comparative" method (in the narrow sense) of reconstruction is intended as a formal calculus. "Internal reconstruction" makes a similar claim. Both are circumscribed by, and interpreted according to, considerations concerning language typology and, possibly, language universals. Certain premises enter into what follows; for example: (a) double articulation: every text has a phonological and also a morphemic (syntactic, etc.) structure and may be represented in some phonological and in some morphic notation; (b) for neither representation is there one true, or one best segmentation though what is perceived as different (what is "in contrast") must be written differently, a weaker requirement being that what is perceived as identical (homophonous) be written the same (as in biunique [autonomous] phonemic notation); (c) speakers can generate new discourses; (d) speakers are multilingual in the sense that they command different styles, registers, dialects, or languages; etc. In certain settings (not all of which are well understood), for example in contact with other speakers, speakers may produce a new variety of speech which differs from preexistent speech by the fact that a phonological contrast is replaced, in some phonological environments or in all, by one and the same entity; there is homophony. Such a replacement is a "sound-change". In Latin, s between vowels and r between vowels 1 merged, with the result, e. g., that serit 'strings up' (with an earlier r) and serit 'sows' (with an earlier 5) became homonyms. Since it happens to be the case that the new r was, by all accounts, "more similar" phonetically to the old r than to the old s — hence the preferred alphabetic rendering — this is conveniently expressed by saying that "intervocalic s became r" (implying that r "remained" r). Note, however, that the sound-change (merger) is a fact regardless of the phonetic description which we may or may not be in a position to supply. At each stage the phonological alphabet is self-contained, and any use of the same symbol from stage to

24

Henry Μ. Hoenigswald

stage is undefined except in terms of some conceivable permanent feature framework. As the two intervocalic consonants have "merged" it may be said that the earlier phoneme s has "split": intervocalically it is replaced by the new r, but otherwise (for example before t which itself "remains" t) by the new s (5 "remains" 5) which in this latter position contrasts with the new r (*gesit 'conducts' has gone to gerit but gestus 'conducted' is still gestus and does not rime with sertus 'strung up'). The split is only an aspect of the merger (Polivanov's law). However gradual the transition from [s] to [r] may have been — moving, as perhaps it did, through an intermediate phase [z] — the shifting spirant continued to be in complementary distribution with the other instances of s until the merger was complete. To be sure, the merger product is merely neutralized at first, and the notation ("r") chosen still reflects the judgment that it is phonetially closer to r[t than to s[t (see above) and is, therefore, open to revision and, in any case, somewhat arbitrary. This has a certain importance in the context of reconstruction. In the foregoing case the split-off phone is at the same time the merging phone (part of the occurrences of the new intervocalic r). We may call this a "primary" split and may wish to distinguish it from a "secondary" split, also subject to Polivanov's law in the sense that it is likewise a facet of merger. There are, in other words, mergers without split (namely, unconditional mergers, in which all allophones of a phoneme merge with those of another phoneme or with zero), but there is no split without a merger. There are mergers associated with a primary split (on which see above), and there are mergers associated with secondary split where the merging phone is not the split-off phone, but where it is instead the case that a contrast is merged in such a way that an allophonic difference in the environment which used to be conditioned by the contrast in question has become a contrast by default. Actually, however, the nexus between merger (in one place) and secondary split (in another) is ambiguous. Let us assume an older stage with the sequences uta (with u — [u]) and uti (with u = [Ü]), and let us also posit a sound-change merging a with i into "e". If the physical difference between the two rounded vowels persists at all — and this would be the crucial, unpredicted datum — the outcome will be ute and üte. We are then tempted to say that u has split by virtue of a physical alteration elsewhere in the stretch. But note that, in a sense, nothing has happened. The two stretches — think of them as two words with different meanings — were distinct and are still distinct; no homonymy has come into being. In fact, if we imagine, quite plausibly,

Comparative

method,

internal reconstruction,

typology

25

the new e with two allophones, [ae] after a back vowel and [i] after front vowel in the preceding syllable, the triviality of the case is even clearer: we have simply decided to phonemicize the possibly "unchanged" [utae] and [üti] first in one way and then in another. What prompts a decision in such cases is not so much a shift in the prominence of the physical differences from one pair of variables ([ae], [i]) to the other ([u], [ü]) as the circumstance that one of them may be "similar" to a third with which it is complementary and which in the environment in which it occurs undergoes "at the same time" a real merger not compensated by a secondary split: ata and ati might merge into ete, ina and ani into ene and so on. Such configurations are frequent; they, too, play a role in reconstruction (cf. Bhat 1972: 8 3 - 8 5 ) . The expression "at the same time" is, in fact, without a definition until we introduce systematic considerations of relative chronology (see below). By the same token there is no sense in which we can speak of a given merger as "one event". In instances where enough is known we tend to find that the environment for a merger may widen stepwise: a change takes place first in one, then in another, then in still another environment, until the expansion stops or until the merger ends up as an unconditional sound-change. Conversely, we may choose to describe developments as if they were combinations of more than one process (cf. Hoenigswald 1985). In Attic the proto-Greek voiceless labiovelar E m e r g e d with t into .1 before front vowels, and with ρ into ρ in most other environments, ρ ! I I

== before front vowels = in most other positions

Ρ

where the vertical alignment of "/" under "i" and of under "p" signifies what is variously called "phonetic similarity" or distinctive feature "identity" (in a universal framework). — In the history of Latin, r replaces intervocalic 5 (see above), and s replaces ss intervocalically after long vowels and diphthongs as well as after vowels at the end of a word (e. g., caussa 'cause', ess 'thou art' > causa, es), ss

s

r

r

= at word end = preceded by long vowel or diphthong, and followed by vowel

26

Henry Μ.

Hoenigswald

The two histories are usually treated under different heads, the first as a bifurcation, the second as a filling of a systemic gap (5 in inter-vowel position), and it is easy to see that the phonetic alignment gives them a different complexion. 2 But it is also clear what they share: the components are connected in the same fashion (by mergers and splits). When different sound changes (mergers and their corollaries) affect the same language in different subcommunities the two forms of speech (languages) which result from these sound-changes, and from other changes which are not sound-changes, may be subjected to the "comparative" method whereby the phonological shape of morphs and morph sequences is reconstructed. It is useful to distinguish between the procedure in the abstract and the further interpretation — including first of all the phonetic interpretation — of the results. The procedure itself is in principle based on the topological relations discussed above. We present its central features by envisaging the binary case (bifurcation only) and going on to multiple descendancy from there. When of two descendant languages one innovates by undergoing a particular merger, the result will be two partially like correspondences. If phoneme (etc.) χ merges with y in one descendant (I) but not in the other (II), there will be two correspondences a/m and a/n as the morphs of I are compared with their translations in II, where a, b, 0i and m,n, .... 0 n a r e the contrasting phonological entities (such as phonemes) of I and of II respectively. Since χ and y were in contrast with one another, the same will be true of the correspondences a/m and a/n with regard to all other correspondences. Both may, for example, occur in the environment b-/o-. If the merger in I is conditioned, e.g., occurs only after z- which itself goes to c-/p- but not after w- which itself goes to d/q-, χ and y will be reflected in three correspondences, e/r, e/s, f/s. Of these, however, while the first is in contrast with each of the other two, the third (occurring only after d-/q-) is complementary with the second (occurring only after c-/p-). This justifies the reconstruction of only two (not three) phonological entities from three correspondences. It amounts to the retrieval of x, of y, and of a primary split in language I. The principle is, then, that correspondences among descendant languages have the properties of allophones, their "phonetic similarity" being represented by complementarily distributed correspondences sharing one member (in this case, /s; note that e/r and e/s, which also share a member, e/, are not complementary and therefore do not qualify). That this characterization of the comparative method fits the formulation of "autonomous phonemics" so closely no doubt simply reflects the way in which early pho-

Comparative

method, internal reconstruction,

typology

27

nological concepts grew out of the historical and comparative tradition in linguistics. In classical Indo-European phonology the contrasting status of the Sanskrit/Latin correspondences i/i, if a, a/a could be used to reconstruct the three IE phonemes *i, *a, while s/s, sjr, rjr, with s/r found only preceded and followed by vowel/vowel and s/s found only otherwise, lead to only two Indo-European phonemes, *s and *r, complete with the primary split touched upon earlier. 3 Sometimes indeterminacies seem to turn up; frequently these only bring to light, once again, the problematics of synchronic analysis or, simply, of notation. Still, the English/German correspondence d/d occurs only after n/n and after /// (as in end and gold). After njn it is complementary with the otherwise generously distributed d/t\ after /// it is complementary with thjd. This suggests a primary split for one proto-obstruent in English (with the merger occurring after *l) and one for another proto-obstruent phoneme, actually observable in the written texts of medieval German (after n, < *ri) (cf. Hoenigswald 1960: 126 — 127). This, then, is a problem in "overlapping". In the ideal case descendant language communities are neatly separated and share particular mergers only by accident. In reality this is often not true. Besides, some mergers are more likely to take place than others if the same causal factors, about which little is known in general, are present. Ideally, however, if I shares a non-trivial merger with II to the exclusion of III it cannot share another non-trivial merger with III to the exclusion of II. Innovation, here represented by sound-change (merger), forms the graph known as the rooted tree. Such schematic trees can be constructed in various ways. For instance, if all the binary reconstructions for a set of η languages are carried out, it may be found that the immediate protolanguage for some two languages paired for reconstruction (say, A and B) is, in turn, identical with another member of the set (C). If C is now paired with A (or with B) the resulting proto-language will be identical with C itself — naturally so, since it has just been shown that A (B) has all the innovations while C has none. The "line of descent" — the relationship that makes one language an older or a later "stage" of another — is therefore not a primitive, to be utilized in the application of the comparative method, but is precisely founded on that method. Internal reconstruction, though of very great importance, is far less self-contained than the comparative method. As a matter of fact, it seems doubtful that it has a codifiable core of sufficient generality to qualify as "another method". Its functioning is best understood when it is considered in its relationship with the comparative method, and then again when

28

Henry Μ. Hoenigswald

this relationship is measured against the equally weighty relationship which ties the comparative method and its interpretation in real history to the viewpoints which have been gaining influence through the exploration of language typologies and universals. Since we have so far limited ourselves to phonological reconstruction (which is the legitimate, not merely the accidental center of the comparative method) it is best to concentrate on the way in which internal reconstruction, too, functions in that area. The central motif is that conditioned sound-change (split) can create allomorphs characterized by morphophonemic alternation. More specifically, a primary split will under favorable circumstances (depending on the nature of paradigms and morph boundaries) create alternations that are phonologically conditioned (sometimes called "regular" — a much misused term). The pair gerit : gestus stands for a whole class of paradigms with r-allomorphs and s-allomorphs distributed in such a way that r "instead of 5" is what stands between vowels. A secondary split, on the other hand (if indeed it is notationally sanctioned; see above), will produce grammatically or lexically conditioned morphophonemic alternations, i.e., alternations such that a description of their conditioning necessitates the listing of morphemes or morpheme classes. The alternation between / and ν in English goes back to the secondary split which occurred when the loss (merger with zero) of certain endings put what had been a word-interior voiced allophone in the word-final position where it contrasted with the voiceless f . The alternation can now only be stated in terms of noun and verb, singular and plural, specific vocabulary items, and the like. This is of course precisely because of the loss, by merger, of the conditioning difference which would be required to make the alternation phonologically predictable. Internal reconstruction of this kind consists, then, in starting back from the alternation: a "regular" alternation points to a primary split, under (at least partly) statable conditions; an "irregular", morphemically-conditioned alternation points to a secondary split. However, the precariousness of the inference becomes clear when we think, for instance, of languages with a fairly through-going variety of vowel harmony. In Hungarian -ban 'in' alternates with -ben {a häz-ban 'in the house' : a kert-ben 'in the garden'). Yet, to make this into an indication of primary split and thus grandly to imply that vowel harmony is never "original" would be preposterous. Alternations are not anomalies. Here we have touched upon one of the aspects of change for the understanding of which we need typological controls.

Comparative method, internal reconstruction, typology

29

At this point we must retrace our steps. Sound change is not the only change process to exist and to be the goal of reconstruction. It is useful to follow the old classification whereby change other than sound change can be subdivided (aside from mere obsolescence) into borrowing, semantic change, and analogic change — it being understood that any one of these can affect all kinds of morphemes, including constructions and transformational rules. It must further be understood that, as in the case of sound-change, the essence of change is replacement. The mere utterance of a "new" discourse, sentence, or construct in the exercise of a speaker's generative potential may be a "change" to the lexicographer. But, strictly speaking, even a borrowing consisting in the addition of a loanword to the stock of morphs qualifies as a change only insofar as it takes the place of another morph or other morphs. This is of particular importance for "analogic change". "Analogy" by itself is a very general word for the way in which new discourses, sentences, etc., are generated from other discourses, sentences, etc. Analogic change occurs when a construct so generated replaces another construct. In its simplest form the replacement is betwen allomorphs — the very characteristic of a replacement being that the replaced and the replacing item have the same meaning. As eyen is replaced by eyes on the "analogy" of other English plurals in -s, and Latin honös 'honor' (cp. gestus above) is replaced by its co-allomorph honor, we can unhesitatingly point to the (morphemic) objects — namely, the plural of the word for 'eye' and the nominative singular of the word for 'honor' — which have undergone the (allomorphic) analogic change. Now, allomorphs are either suppletive (in which case they are usually conditioned grammatically or lexically) or morphophonemically related and conditioned either grammatically and lexically or phonologically (see above). The suppletive ones differ in non-recurring ways (e. g., plural -n and -5) while the phonological difference between morphophonemically related ones recurs in other allomorphs which belong to other morphemes: wolfjwolves, sheaf I sheaves, scarfj scarves, or, for that matter, honosjhonorem (accusative singular) and gestus/gerit are examples. We have discussed the background of morphophonemically related allomorphs above. Now we must deal with the analogic intrusions, honor, roofs (replacing, in some dialects, rooves), and the rest. In the context of reconstruction — and here we are back with the comparative method — these present us with the following problem: a finding of a/m φ a/n (read as 'contrasting with') ordinarily justifies the reconstruction of a contrast in the proto-language, with a merger, into 'a', in one of the descendant lan-

30

Henry Μ.

Hoenigswald

guages. Thus, the correspondences seen in Dutch drie, dank vs. English three, thank on the one hand, and in drinken, dochter vs. drink, daughter on the other, point to two contrasting consonants at the beginning of the words, with a merger in Dutch. This, however, is true only if the words are morphemically, i.e. semantically, unrelated. As we have seen, -s in other Indo-European languages corresponds to the Latin -s of honös, and the contrast with the -rj-r of other words makes it necessary, naturally and trivially, to reconstruct two proto-phonemes in word-final position, s and r. Thanks to the analogical introduction, from non-final position, of honor there is now a word-final -rj-s in the material as well: for a fairly extended period honös and honor exist side by side. During that period the fact that they form a pair of doublets — that is, that they are semantically related as quasi-synonyms (with a stylistic nuance) and that the phonological difference between them recurs in other doublets (cf. Hoenigswald 1983) — shows that this is not a case of the non-Latin languages merging two phonemes, but rather a case of morphic change bringing about a deceptive impression of sound-change, on the Latin side. This insight further creates the presumption that other instances of word-final -rj-s represent completed analogic changes, with the doublet stage left behind. Borrowings can create the same deceptive impression, provided that the source of the borrowing is a related language or dialect; this includes the well-known "learned" borrowings from a literarily preserved ancestor. In Italian, pieve 'parish' and plebe 'populace' are doublets (cf. Hall 1964: 300 — 302); semantically they are certainly close enough to raise a suspicion, and the phonological differences recur in such pairs as pianare 'smooth out' vs. pianare 'glide', and then again in lavoro 'work' vs. laborioso 'laborious'; the fact that i: / and ν : b are associated with each other lends particular strength to the doublet pieve: plebe. The conclusion is, to put it neutrally, that two channels of transmission from an ancestor were involved here. A consideration of the quality of the stylistic difference between the components of the pair (perhaps "bookish" vs. "ordinary") and of certain phonetic factors where they can be known (/ and b are more like Latin, the "learned" source, than are i and v) will prompt us to label plebe, pianare, and laborioso as loanwords, but pieve, pianare, and lavoro as inherited by descent. To sum up, the configuration a/m=|=a/n is interpreted as follows: (1) In the absence of doublets, the proto-language was like the second descendant language, and there was a merger in the first. (2) If the material includes doublets in the second descendant, the proto-language

Comparative

method, internal reconstruction,

typology

31

was like the first descendant language, and the second descendant language has innovated. However, this innovation was not by sound-change; rather, (2a) if the phonological difference between the components of the doublet pair (as in roofs: rooves) occurs in paradigmatic alternation as well (iknife: knives) so that a split can be internally reconstructed, the innovation was an analogic change. If this is not the case, (2b) the innovation was a dialect borrowing. It should be added that the argumentation from doublets tied to alternation (2a) is closely connected with the old maxim that morphologically isolated forms — say, the Latin indeclinable eras 'tomorrow', which is not part of any paradigm — will show the phonologically regular treatment. The relative chronology of sound-changes, and of morphemic changes such that they interact with sound-changes, is subject to reconstruction, although, much as with internal reconstruction, the cogency of the inferences is often overstated. Rule orderings as such, dependent as some of them are on mere notational decisions, cannot simply be interpreted as realistic orderings of change events in time. In some cases the argument turns on whether parsimony of statement is necessarily the best guide for the attainment of historical truth. The establishment of subfamilies within language families is among other things also a device to determine relative chronologies: the innovations that are shared by the smaller sub-subfamilies are more recent than those shared by the intermediate subfamilies. West-Germanic traits postdate Common Germanic traits, and so on. The difficulties are greater when the internal reconstruction of a relative chronology is attempted. Parsimony does play a considerable role. If, for instance, two phonemic entities are split along the same lines of conditioning but also merge in each of these conditions (a^ > m I? b, > m t , a 2 > n2, b2 > n2, perhaps with an additional merger like c1>2 > n 1 2 t o ensure contrasting status for m Φ η), it is economical to let the merger of a with b in both environments (t and 2) precede the split (which is, in turn, a side effect of the merger with c) because this enables us to think of the split as one event affecting the merger product of a and b rather than as a duplication of events. In all these matters the dependence on particular alphabetizations is likely to compromise the inevitability of proposed solutions, or rather, it illustrates the importance of the tacit typological assumptions that are at the bottom of notational choices. Still, a case of the sort described has been put forward for Latin vowel weakening, for the history of the labials and the labiovelars in Celtic, for the relation between the assibilation of the

32

Henry Μ.

Hoenigswald

Indo-European palatals and the insertion of t between s and r in Slavic, and for other change complexes. 4 Behind and above the fundamental formalisms derived strictly and exclusively from the notion of contrast, typology looms large. It has long been known that the phonetic traits which we assign to phonological items reconstructed by the comparative method do not emerge from that method. They are, in fact, assigned either on the basis of something like a majority decision (if nine out of ten descendant languages show a phonetic consensus for a proto-phoneme and the tenth is different, it seems reasonable to decide that it is more likely that one descendant should have changed than that nine correlated descendants should have gone through the opposite change independently). Assignment may further be based on a belief, often strongly supported empirically, that certain phonetic alterations — such as that from stop to spirant — are more likely and "easier" than others, though it should always be remembered that sound-changes that grow out of hypercorrections must be expected to have a counter-natural direction. A sophisticated typological argument may be found in the "glottalic" interpretation of the IndoEuropean obstruent system proposed by Gamkrelidze and Ivanov, where it is argued that certain distributional properties and constraints, said to be normal in the co-occurrence of glottalized consonants are still present, though no longer in motivated form, in most of the Indo-European descendant languages even after these have shed, subphonemically but effectively, the feature of glottalization (see T. Gamkrelidze, this volume). "Ease of articulation" (sometimes combined with "ease of perception") is a traditional motif in the search for the so-called causation of soundchange. The concept is too loose and too vulnerable to the obvious charge of vicious circularity to be useful, though it is worth pointing out that, whatever its demerits, it does not violate our hard-won uniformitarian tenets. Taking the simplification of consonant clusters as a likely candidate for argument, it could be countered that the syncopation of vowels is possibly an equally widespread, quasi-natural process, and that it is likely to create new consonant clusters, but does so by a mechanism which is not at all the simple reversal of the loss of consonants from clusters (as the massive epenthesis of consonants — a rather unheard-of type of change — would be). Another universal of change that has been proposed is that of mergers favoring the unmarked partner of a phonological opposition as a target, and one might well think of converting that proposition into a guideline for reconstruction. But the uncertainties are forbidding: the premises are not settled (what kind of markedness do we

Comparative

method, internal reconstruction,

typology

33

have in mind?), and factual checks far too often bring to light contradictions and exceptions. Nor is it clear to what extent the statistical preponderance of a typological or universals-oriented state of affairs can justify an effort to base on it a procedure for reconstruction. It is impossible to suppress a query about the status of complete induction as well as the further query, close to our present concern, whether it is merely out of naivete and habit that the problem never seems to trouble us so long as we have reason to be content with the immediate results of the comparative method or the simpler forms of internal reconstruction. The historical-comparative tradition has been much reproached for an undue obsession with the sounds of morphs; why — so the complaint runs — cannot the methods leading to that kind of reconstruction simply be repeated on the higher level of "morphology", syntax, and semantics? Some of the answers have been hinted at earlier. The comparative method in the strict sense needs the semantic equivalence between cognates to establish them as cognates, the possible effects of semantic changes notwithstanding; it is obvious that there can be no appeal from the higher, morphemic level to a still higher one — double articulation being at the heart of the very design of language. To be sure, there are a few superficial resemblances, as well as a few valuable analogous principles that can serve as algorithmic steps in their own right; one which deserves mention is the rule that suppletive allomorphs being former contrastive morphs, a semantic merger of two morphemes can be inferred from such suppletion. Thus, the subjunctive mood in Latin, with its suppletion as in laudemus 1 pi. from laudäre 'praise' vs. uelimus from uelle 'will' may be seen to have had a syncretistic history — a bit of reconstruction confirmed by the semantic insight that the subjunctive combines two discrete meanings, optatival and hortative. This, however, covers very little ground. 5 The bulk of morphemic reconstruction follows qualitative, impressionistic ways of proceeding, with little evidence as yet of successful formalization. The typological and universal properties adduced are most often synchronic, and it is only when it is claimed that an anomalous condition, transitional between two states of balance, is discovered in a descendant or later stage that there is scope for the recovery of an antecedent state. Word-order typology, presumably in conformity to contingent univers a l of word-order, is sometimes made to serve in this fashion — a challenge to the old uniformitarian preconceptions. Whether it can be surmounted is the chief question which Winfred Lehmann has raised and which forms a central item on our investigative agenda.

34

Henry Μ.

Hoenigswald

Notes 1. The conditioning is formulated in simplified fashion. 2. It is one of the characteristics of gap-fillers as contrasted with bifurcations that only a "bleeding" chronology (s > r before ss > s) will fit the typological framework we prefer, namely one in which a fixed entity (s) persists or resurfaces in a 'pull-chain' sequence. 3. The procedure is easily extended to cover the occurrence of (different) mergers in I and II each; see Hoenigswald (1960: 128). — The Sanskrit/Latin data discussed here are simplified. 4. A primary split will at first create alternations which are not only phonologically statable ("regular") but also "automatic" in the sense that they are, at that time, phonologically necessary; the "gap" is there, not yet "filled", see fn. 2 above. It is honorem "because" *honösem is at that point disallowed. After the second soundchange, ss > s (known, more or less, to have occurred after a lapse of several decades) the alternation is still "regular" but no longer "automatic". 5. Semantic reconstruction is not as hopeless as some would have it (see, e. g., Cowgill 1986: 66 — 67). A rewarding formalization must certainly await the fuller development of formalisms covering the borderline area between analogic and semantic change, but it should be said that some of the more interesting comments on the reconstruction of lexical meaning have to do with the internal reconstruction of shorter constructs from longer ones. It may be of interest to historians and archaeologists — such questions are frequently raised outside linguistics proper — to know that the word for 'wheel' corresponding to Latin rota can be reconstructed internally for pre-Indo-Iranian if indeed the Indo-Iranian word for 'chariot' (Sanskrit rätha- etc.) is derived from it with a suffixal -a (Mayrhofer 1976: 3 8 - 3 9 ) .

References Bhat, D. N. Shankara 1972 Sound change (Poona: Bhasha Prakashan). Cowgill, Warren 1986 Indogermanische Grammatik 1.1: Einleitung (Heidelberg: Winter). Hall, Robert Α., Jr. 1964 Introductory linguistics (Philadelphia, New York: Chilton Books). Hoenigswald, Henry M. 1960 Language change and linguistic reconstruction (Chicago: University of Chicago Press). 1983 "Doublets", Essays in Honor of Charles Hockett, in: F. B. Agard — G. Kelley —V. B. Makkai (eds.), Essays in honor of Charles Hockett (Leiden: Brill), 1 6 7 - 1 7 1 . 1985 "Distinzioni reali e distinzioni chimeriche nella classificazione dei cambiamenti fonologici", in: L. Agostiniani — P. B. Maffei —M. Paoli (eds.), Linguistica storica e cambiamento linguistico (Pubblicazioni della Societä Linguistica Italiana, 23) (Roma: Bulzoni), 3 1 9 - 3 2 6 . Mayrhofer, Manfred 1976 Kurzgefaßtes etymologisches Wörterbuch des Altindischen 3 (Heidelberg: Winter).

Contributions from new data to the reconstruction of the proto-language* Karl Horst

Schmidt

The topic proposed to me by our hosts refers to the question to what extent the reconstruction of a proto-language can be affected or modified by new data discovered later on. New data in this sense refers first of all to the broadening of the material basis of a language or language group by the discovery of new documents or of whole languages whose genetic relationship to previously known languages can be proved by means of the identification process defined by Benveniste (1966: 102): "Le processus est lie a la consideration de la substance concrete des elements compares". I would like to call this kind of enlargement of the concrete linguistic basis "substantial enlargement" as opposed to "interpretative enlargement", the latter being a new analysis or arrangement of the material as a result of better understanding or superior interpretation of the data. Classical examples of interpretative enlargement in Indo-European morphophonemics are the laws set down by Grassmann (1863), Verner (1875), Brugmann (1876) and de Saussure (1879),1 while the laryngeal and glottalic theories have been under discussion in more recent years. 2 It must be admitted, however, that in establishing new data, substantial and interpretative enlargement are not always clearly separated from each other, but may combine to a different extent. The laryngeal theory, for instance, being confirmed by concrete evidence such as Hittite h, "fairly consistent lengthening of vowels before Indo-Iranian nr 'man' in compounds" (Cowgill 1965: 152), or certain prothetic vowels in Greek and Armenian, perhaps also in Phrygian, 3 (1)

Hitt. ha-an-za 'front' : Gk. αντί; Skt. sü-näras < *su-h2nero-s 'beautiful' : Gk. άνήρ, Classical Arm. ayr, Phryg. αναρ : Skt. nar-,

seems to be closer to substantial enlargement than the glottalic hypothesis (cf. (2)) which is based to a higher degree on typological arguments such * I would like to thank K. Hlaväcek and W. Winter for correcting my English and U. Seefloth for completing the bibliography and proofreading.

36

Karl Horst

Schmidt

as the rare occurrence of *p or ejective *p\ or the "universal" that voiced aspirated stops should imply voiceless aspirated stops: 4 II I III p[hj hm (p) t[h] f d[h} k' klh] g'hl (Gamkrelidze — Ivanov 1984: 39). Theoretically the evidence for substantial enlargement and interpretative enlargement could be taken from all the languages in the world. But since this is an Indo-European Symposium, I shall concentrate (after a general introduction) on substantial enlargement and two models of interpretative enlargement in Indo-European. My paper is arranged as follows: 1. Some principles of the reconstruction of proto-languages. 2. Evidence of substantial enlargement. 3. Two models of interpretative enlargement. 4. Summary.

1. Some principles of the reconstruction of proto-languages The reconstruction of a proto-language like Indo-European is based on the theory that genetically related languages are derived from a common ancestor called "proto-language". In proving linguistic relationship, only the "stoffliche Übereinstimmungen", as they have been called by Trubetzkoy, 5 are relevant, that is to say the substance of the linguistic sign, but not its structure or type which can change independently of the genetic relationship of a language. 6 To give just one example of the development of different linguistic types within the same language family, I refer to the agglutinating declension of Modern Armenian which in contrast to the inflecting case forms of Classical Armenian has separate morphemes for number and case: (3)

Classical Arm. nom. sg. azg 'nation, people', pi. azg-k\ gen. sg. azg-ί, pi. azg-ic > Modern Arm. nom. sg. azg, pi. azg-er, gen. sg. azg-i, pi. azg-er-i.

The Modern Armenian inflection type implies Gruppenflexion attributive syntagm: 7

of the

New data in the reconstruction

(4)

of the proto-language

37

Classical Arm. mec-i ark'ay-i του μεγάλου βασιλέως Mt. 5, 35 vs. Modern Arm. mec selan-i 'of a big table'.

The group inflection in (4) is characterized by lack of "word autonomy" 8 — the Modern Armenian attribute mec 'big' is not inflected. Moreover, Modern Armenian agglutination correlates with a reduction of Formvariation,9 i.e. allomorphic change within the same case: (5)

Classical Arm. azg, -i, -e, -iw; pi. -k\ -s, -ic, -iwk' > Modern Arm. azg, -i, -ic, -ov, -um; pi. -er, -er-i, -er-ic, -er-ov, -er-um; Classical Arm.y'i 'horse', -oy, -ov; pi. -k\ -s, -oc, -ovk' > Modern Arm. ji, -u, -uc, -ov, -um; pi. -er, er-i, -er-ic, -er-ov, -er-um.

The paradigms in (5) show that the difference between Classical Armenian i- and o-stems has been neutralized in the modern language both in the instrumental and locative singular and in the entire plural. In comparison to Modern Armenian, the French declension developed in a totally different way: (6)

Lat. capitis > French de la tete.

The separation of "Begriff, Klasse, Kasus" which Lewy (1964: 17) called Flexionsisolierung is basically equivalent to Sapir's "isolating technique" and "analytic synthesis". 10 In spite of these typologically contrasting substituents which have replaced the Indo-European declension model, nobody would doubt the genetic relationship between French and Modern Armenian. Additional aspects which must be borne in mind in connection with the reconstruction of a proto-language can only be touched upon in passing. Among these may be mentioned the difference between comparative and internal reconstruction (cf. Rix 1986: 6 — 7), the question of absolute and relative chronology, the geographic position of a language in prehistoric and historical times as well as typological considerations. Comparative reconstruction consists of a series of operations: a) Establishment of the recurrent sound and morphematic laws by which the individual linguistic continuants are connected with the reconstructed form in the proto-language, e.g.: (7)

PIE *snusos 'daughter-in-law' : Skt. snusa, Classical Arm. nu, Gk. νυός, Lat. nurus, OHG snur, SerbCS snüxä.

PIE *snusos was affected by the following laws: morphematic law: the change of feminine o-stems into ä-stems (Sanskrit, Serbian Church Slavic);

38

Karl Horst

Schmidt

phonological laws: transition of the cluster *sn- to *n- in anlaut (Classical Armenian, Greek, Latin); of s to 0 or r in intervocalic position (Classical Armenian, Greek, Latin) or before the accent (Old High German); of s to s or ch after u (Sanskrit, Serbian Church Slavic), accent shift (Latin). (Cf. Schmidt 1973: 120f.) b) Establishment of additional transformations like semantic developments or analogical processes limited to individual languages; (8) presents the case of the Old Irish genitive sethar instead of *sesar which received the spirant th on the analogy of other kinship terms: (8)

PIE *suesör 'sister', gen. *suesros : Classical Arm. k'oyr < *k'eur < *suesör, gen. dat. loc. k'er < *suesr-os, -ei, -i; Olr. siur, gen. sethar (instead of *sesar in analogy to bräthair, bräthar, mäthair, mäthar, athair, athar).

c) Attempt at a reconstruction of the phonological, morphological, syntactic and semantic systems of the proto-language. 11 Internal reconstruction, the fundamentals of which were already established by Hermann (1907), deals with morphonemic change in individual languages as a result of splitting, as we have it in the representation of Latin rhotacism by Hoenigswald: 12

The relation of Latin fer-t to fer-imus is comparable to that of Latin es-t to er-imus, the latter going back to *es-imus. The contrast between absolute and relative chronology is based on the fact that the particularly important conservative, archaic elements of a language can be differently motivated. Absolute chronology refers to the

New data in the reconstruction

of the proto-language

39

age of the oldest documents. Relative chronology, on the other hand, takes into account the different speed with which languages change, depending on geographical position (e. g. of marginal languages), language levels (e. g. basic vocabulary; grammatical subsystems) and possibly on the type as well. In (10), a plus sign indicates a high rating in the context of the respective parameter; in the table, four such parameters are combined: (1) absolute chronology, (2) relative chronology, (3) cohesion of tradition, (4) duration of tradition (cf. Schmidt 1984: 129). Hittite 1. 2. 3. 4.

absolute chronology relative chronology cohesion of tradition duration of tradition

Greek

Sanskrit Albanian

+

+

±

±

±

+

+ +

+ +

—

+ +

The combination of plus and minus signs serves to show that in terms of relative chronology, the languages with earliest attestation (Hittite, Greek, Sanskrit) are rated unevenly in the current discussion of ProtoIndo-European. The problem of geographical position is closely connected with the question of the separation of the Indo-European languages which has been under discussion since the 19th century: Lottner (1858, 1861) replaced the Greek-Latin hypothesis with his Italo-Celtic theory; Schleicher (1861) developed the idea of a family tree splitting into slawodeutsch and ariograecoitalokeltisch; J. Schmidt (1872: 27) preferred "das bild der welle ..., welche sich in concentrischen mit der entfernung vom mittelpunkte immer schwächer werdenden ringen ausbreitet"; Leskien (1876: XVII ff.) attempted to synthesize the theories of Schleicher and Schmidt; moreover he established the important principle: "Die Kriterien einer engeren Gemeinschaft können nur in positiven Uebereinstimmungen der betreffenden Sprachen, die zugleich Abweichungen von den übrigen sind, gefunden werden" (p. XIII). 13 Meillet (1922: 24) discussed the question of the unities of Indo-Iranian, Italo-Celtic and Balto-Slavic in the sense of dialect groups "qui reposent sur une periode de communaute posterieure a la periode indo-europeenne". At present, one of the most important questions of Indo-European linguistic geography concerns the position of Hittite and Anatolian within the language family 14 and the problem of the differences between eastern Indo-European, i. e., Indo-Iranian, Greek, Armenian and the proto-language. 15 In studying language typology, 16 phonological, morphological, and syntactical features must be taken into account, as has been discussed by

40

Karl Horst

Schmidt

many authors — recently by Gamkrelidze — Ivanov (1984). The persuasiveness of typological criteria is limited by the following dilemma: There are clear indications of a Proto-Indo-European type which differs from the model arrived at by direct reconstruction, but this typologically plausible reconstruct is difficult to verify because of the great time gap. To give just one example, on the levels of morphology and syntax the Indo-European nominative/accusative construction very likely was preceded by a still older syntagm consisting of the cases activus and indefinitus. This older reconstruct is indicated by a number of features among which may be mentioned: a) The old dichotomy of gender — communia (masculine and feminine) vs. neutra — which has been established for prehistoric times even though it has not yet been conclusively proved for Anatolian (Cf. Schmidt 1979 a: 793 — 794). Needless to say, however, the enlargement in i by which the nominative/accusative singular and plural of several stem-classes in Luvian is characterized — (11)

Nom./acc. sg./pl. Luvian harrani- 'eagle', Hierogl. Luv. suuani'dog'; Luv. uasu- n. 'good', uasui- c.; ura- n. 'big', uri- c.; ärraian. 'long', äraii- c. — 17

cannot be taken as proof of old feminine gender. b) The verb classes activa vs. inactivajstativa as a preliminary stage of aorist/present vs. perfect. 18 c) The difference between marked nominative/accusative case of the genus commune vs. unmarked nominative/accusative case of the genus neutrum: before the emergence of the nominative and accusative cases the system was based on the indefinite case, casus indefinitus]9 or casus absolutivus, the latter being the preliminary stage of activus and obiectivus ("Zielkasus") as a first step towards nominative and accusative. The parallels with regard to the marking of the direct object follow the principle "that separate accusative marking and verb object agreement are more likely with noun phrases that are high in animacy or definiteness" (Comrie 1981: 212). Applied to ergative or active languages, this principle produced the hypothesis presented by Silverstein (1976: 22): "If an ergative system splits simply into two two-way case-marking schemes, then minimally either the [ + ego] or the [ + tu] forms are nominativeaccusative, the rest ergative-absolutive". 20 Contrary to these indications — and that is the dilemma — the processes of language change which transformed the prehistoric type into

New data in the reconstruction

of the proto-language

41

the Proto-Indo-European reconstruct cannot be proved in detail because of the great distance in time. Nevertheless, it would be just as wrong to ignore typological considerations altogether as it would be to accept unproved hypotheses uncritically.

2. Evidence of substantial enlargement 21 According to the extent of their effect on the Indo-European reconstruction model the evidence of new data may be subdivided into three groups: (a) Languages whose discovery affects the basis of the Indo-European reconstruction model: Hittite (Anatolian) and to a lesser degree Tocharian; (b) Languages affecting mainly individual subgroups of Indo-European: e.g., Mycenaean, Mitanni, Venetian, Celtiberian; (c) Linguistic documents which affect individual languages or groups: e. g., Continental Celtic inscriptions. By far the most important evidence of group (a) is Hittite, which combines simplified grammar with very ancient tradition. The strong deviation from the traditional reconstruction system, as it is codified in Brugmann — Delbrück's Grundriß on the basis of previously known IndoEuropean languages, in particular Indo-Iranian and Greek, led to two different conclusions: on the one hand (cf. (12), the "Indo-Hittite hypothesis" established by Forrer (1921: 26 ff.) and developed by Sturtevant between 1926 and 1951: (12)

Hittite (Sturtevant 1962: 106).

Indo-Hittite

Sanskrit

Greek

Latin

Sturtevant (1964: 9) attempted "to show that Hittite and Proto-IndoEuropean are both descended from a common ancestor, which we may

42

Karl Horst

Schmidt

call Proto-Indo-Hittite". This hypothesis, which had long been rejected by most scholars, is now being discussed anew. 22 Adherents of the second conclusion believe in principle in Proto-IndoEuropean as a model very close to the old Indo-European languages Indo-Iranian and Greek, explaining the grammatical categories lacking in Hittite by later loss (so-called Schwundhypothese).23 An unprejudiced examination of the position of Hittite must distinguish between preserved archaisms and innovations in every single case. Probable archaisms are features like the reflexes of laryngeals preserved in Hittite (cf. (1). possibly the old dichotomy of genus commune and genus neutrum, the lack of a perfect tense 24 and the s-aorist, 25 the preservation of the old genitive singular of o-stems in *-os: (13)

Hitt. nom. sg. antuhs-a-s, gen. sg. antuhs-a-s \s. *-eso/-oso (Greek, Germanic, Old Prussian); *-osio (Indo-Iranian, Greek, Classical Armenian, Faliscan); -Γ (Latin, Celtic, Venetic, Messapic); *-5d (Slavic, East Baltic, possibly Dacian, Celtiberian); -eis (OscoUmbrian). 26

the distribution of the so-called dimensional cases in Old Hittite, investigated by Starke (1977: 126): (14)

Singular Sachklasse Dativ Terminativ Lokativ Ablativ Instrumentalis

Plural Personenklasse -i

-a -i -az -it

-as

or the use of the indefinite case as a locative (cf. Neu 1980). In other cases Anatolian has innovated; this has been discussed in particular with regard to the probable loss of the old optative suffix *-ieh]-j-ih]- characterized by stem gradation. 27 The lack of morphological aspect 28 may well result from another special Anatolian development. Its former existence is attested by Anatolian present stem formations. 29 The lacking differentiation of aorist and imperfect implies the loss of the archaic tempus primitivum commented on by Schwyzer (1953: 640) as follows: "Bei formeller Gleichheit ist für die Auffassung die Stellung im System entscheidend": 30

New data in the reconstruction

(15)

of the proto-language

43

Gk. aor. έγένετο : prs. γίγνεται, imperf. έγίγνετο vs. Skt. imperf. ajanata : prs. janate (cf. OLat. genit); Classical Arm. aor. eber 'he carried' < *ebheret : imperf. Skt. abharat, Gk. εφερε; Classical Arm. aor. elik' 'he left' < *elikwet : aor. Skt. aricat, Gk. ελιπε.

Although Hittite coincides with other old Indo-European languages in a set of characteristic features of declension (stem classes, cases)31 and conjugation (present stem formations, two series of personal endings, differentiation of active and medium diathesis), its archaisms point to an early separation of the Anatolian language group from Indo-European. If we disregard the innovations which took place after separation, e.g., transition from the category of state to the /»'-conjugation, loss of aspect and tempus primitivum, probably also of the optative, Anatolian corresponds to the reconstruct of an older Indo-European type characterized by the following set of features: verbs differentiated as to diathesis (state vs. action) and aspect (present/imperfect vs. aorist); classification of the nouns as to person (common gender) and non-person (neuter), only the first class occurring in the agent function; common use of the indefinite case (cf. Schmidt 1984: 133). As to Tocharian, whose declension shows such typological innovations as agglutination and special marking of the accusative singular in connection with rational beings, I refer to the statement made recently by Thomas (1985: 128): "Auf Grund der Besonderheiten, Umgestaltungen und Veränderungen in seinem sprachlichen System kann das Toch. nicht die Bedeutung für das Idg. und die Indogermanistik erlangen wie das Heth., die andere zu Beginn des 20. Jahrhunderts entdeckte idg. Sprache." That the two features just mentioned reveal recent developments of Tokharian declension is confirmed by the conservative status of certain classes: On the one hand the grammatical cases do not share the transition of inflecting to agglutinating case forms; on the other hand kinship terms, mostly /'-stems, which are generally known as archaic, lack the accusative singular mark: (16)

Sg. nom. Toch. Β päcer 'father' [Α päcar], obi. pätär; gen. pätri [Α päcri\; Β obi. petso 'husband' [A nom./obi. pats], nom./obi. soy 'son' [A se].

Group (b), i. e., material affecting mainly individual groups of the protolanguage, consists among others of particularly old Indo-Aryan and

44

Karl Horst

Schmidt

Greek documents. The Mitanni evidence from Asia Minor is confined to a few appellatives, divine and personal names; among this material we find Vedic numerals, partly in archaic and partly in innovated form: (17)

Mitanni a-i-ka- Ί ' : Vedic eka-, ti-e-ra- '3' : Ved. tri-, pa-an-za'5' : Ved. pänca-, sa-at-ta- '7' : Ved. saptä-, *na-a-ua- '9' : Ved. ηάνα- (Mayrhofer 1974: 15).

Mycenaean 32 shows a series of archaisms, e. g., the preservation of labiovelars, lack of vowel contraction and preservation of *u in nearly all positions: (18)

Mycen. qo-u-qo-roj^o\)YioXo^ = βουκόλος, a-to-ro-qo/ävθρως™ος = άνθρωπος; do-e-rojbοελος = δούλος; wa-tu/Facxu = άστυ.

The position of Venetic, which is known from inscriptions found in northeastern Italy (Este, Lägole) and shows a number of agreements with Latin (19), is disputed: (19)

Venet. ekvon : equum (with *ku); vivoi : vivö (dat., *gwiuo-), murtuvoi: mortuo (dat.; OCS mrütvü); */, r > ol, or : volti- 'will' (subst.), murtuvoi; stem formations fak- 'to make', donä- 'to present, make a present of'. 3 3

A further representative of group (b) is Celtiberian which I shall discuss in the context of Gaulish inscriptions, particularly the inscription of Chamalieres belonging to group (c). Fundamental to the identification of the Celtiberian inscriptions as Celtic by the late A. Tovar in 1946 was the analysis of the ethnical and linguistic strata of the Iberian peninsula on the basis of Gomez-Moreno's conclusive reading of the Iberian characters in the early 1920s (1925). Celtiberian syntax is characterized by a number of archaic features, first of all the basic word order subject — object — verb with the determinant preceding the determined element. The syntax of at least two inscriptions — Botorrita (20) and Penalba de Villastar — is very likely based on this principle: 34 (20)

A [3] uta : oscues : stena : uefsoniti: silabur : sleitom : conscilitom : gabiseti 'Und wer auch immer diese (Gebäude) hinzuzugewinnen sucht, soll nehmen {gabiseti) Silber zum Spalten abgeteiltes'; [4] uta : oscues : boustomue : cofuinomue macasifaJmue ailamue ambitiseti 'und wer den Rinderstall oder den Rundbau oder ein Mauerwerk oder ein anderes (sc. Gebäude) herumbauen will...';

New data in the reconstruction

of the proto-language

45

[8] ids : osias: uertatosue : temeiue : robiseti: saum : tecametinas : tatus : somei [9] enitousei 'welche arsnas (f. pi.) er entweder draußen oder zu Hause zerlegen/ töten wird, von diesen jeweilig die Zehnten sollen sie opfern diesem Tousos'; [10] tocoitei: ios : ur : antiomue : auseti : aratimue : tecametam : tatus 'für den Togets, der U. oder A. gedeihen lassen soll/wird, sollen sie den zehnten Teil opfern' (Schmidt 1986: 170-171). Among the archaisms of Celtiberian syntax unknown both in Gaulish and in Old Irish may also be mentioned: the fully inflected relative pronoun *ios, the repetition of the conjunctions *kwe, *nekwe, *ue (21)

Celtiberian Tocoitos cue : sarnicio : cue (Botorrita Al) : Gk. πατήρ άνδρών τε θεών τε (II. 1.544) vs. Lep. Latumarui Sapsutai pe vs. archaic Ir. fer δα η-έΐαί be(i)ch ro-ch lamethar forgull 'a person from whom bees escape (i.e., swarm) and who ventures (i.e., is prepared) to testify'

and the conjunction uta which is also attested in Indo-Iranian. Celtiberian has two morphological features which reflect an older Pro to-Celtic reconstruction model: (1) the genitive singular of the o-stems in -o, cf. Tocoitos cue : sarnicio : cue in (21); (2) four thematically inflected s-subjunctive/futures of the 3rd person singular which correspond to the Indo-European type as it is known from Greek, Indo-Iranian, OscoUmbrian and Middle Welsh: (22)

gabiseti 'er soll nehmen', ambitiseti 'er wird herumbauen', robiseti 'er wird zerlegen/töten', auseti 'er soll/wird gedeihen lassen (vermehren)'; hom. άμείψεται, lesb. χαλάσσομεν, ai. nesat = av. naesat 'soll führen'; osk .fust 'erit', deivast 'iurabit', umbr. ferest 'feret' : mkymr. duch 'may bring' < *deuk-s-e-t, gwares 'may succour' < : uo-ret-s-e-t, gwnech 'may do' < *ureg-s-e-t.35

Comments on the s-subjunctive/futures in (22): (a) Stem formation and ablaut degree: gabiseti is a late formation derived from the present stem: (23)

Olr. ga(i)bid Β II 'takes', Gaul, gabi budduton imon.3b

As to ablaut, 37 the full grade is not clearly apparent in ambi-tiseti, root *dheigh- or *steigh-, and robiseti, root *bhei(h)-. Full grade ei is expressed by i instead of the e one should generally expect.

46

Karl Horst

Schmidt

(b) The Celtiberian subjunctive/futures of the roots *dheigh- or *steigh~, *bhei(h)~, *aug- and of the present stem gabie-ji- prove that the Old Irish rule restricting the ^-subjunctive to "strong verbs whose root or verbal stem ends in a dental or guttural stop or spirant, or (in the present and preterite) in (Thurneysen 1946: 380) does not apply to Proto-Celtic. Like the six Old Irish verbs in (24) in which "the future stem ... falls together with that of the subjunctive", 38 the Celtiberian verbal forms in (22) do not differentiate between subjunctive and future. (24)

Olr. aingid, -anich 'protects': subj. fut. sg. 3rd pers. -ain < *anegs-t; la(i)gid'\ies' : Ιέ < *leg-s-t; sa(i)didisits' : seiss (absol.) < *sed-s-t- + vowel; reg- 'to stretch' : -re < *reg-s-t; rethid 'runs' : -re < *ret-s-t; techid 'flees' : •tess (1st pers. sg.) < *tekwsö.

We may conclude from the Celtiberian and Welsh evidence in (22) that the Old Irish differentiation between athematic inflection in the 3rd singular active and deponent as well as in the 2nd singular deponent of s-subjunctive, s-future and 5-preterite vs. thematic inflection in all other cases was the result of a mutual analogical adaptation of the subjunctive, future, and preterite categories. (c) The thematic subjunctive/futures in (22) must be looked upon in the context of two other Celtic future formations with an Indo-European background: 1. The future in *sie-/-sio- which occurs three times in the Gaulish inscription of Chamalieres: (25)

bissiet 'er wird spalten' : *bheid-, *bhid-; pissiiu mi 'ich werde sehen' : air. ad-ci 'sieht' < *-kwis-e-t; toncnaman toncsiiontio 'die den Eid schwören werden' (Schmidt 1986: 174).

When added to roots ending in a resonant, the suffix *-sie-/-sio- implies the development of a laryngeal between the root and the stem formation. This future form occurs in Indo-Iranian and probably also in Baltic and Slavic;39 recently Hollifield (1981) has tried to prove its existence in Greek: (26)

Ved. kar-i-syä-ti 'he will make', vak-syä-ti 'he will speak', Avest. vax-syä Ί shall speak', Lith. düo-siu Ί shall give', OCS bysgsteje, bysgsteje < *bhü-sio-nt- 'τό μέλλον', Gk. κείοντες· κοιμηθησόμενοι Hes. < *kei-hrsio-nt- = Ved. say-i-sya-nt-.

2. The Celtic evidence of the other future formation is restricted to the Old Irish "sigmatic and asigmatic reduplicated futures" which "originally

New data in the reconstruction

of the proto-language

47

constituted a single class" (Thurneysen 1946: 414). It corresponds in substance to the Indo-Iranian desiderative formation. In Celtic and IndoIranian this old desiderative formation is marked both by reduplication and by a thematically inflected s-suffix which, when added to roots ending in a resonant, is preceded by a laryngeal: (27)

Skt. cikirsati : kar- 'to make', krtä-, susrüsate : sru- 'to hear', srutä-; Olr. celid 'conceals' : fut. ·cela < *cechla- < *kiklä- < *kiklhse-/o-; ga(i)rid 'calls' : fut. ·gera < *gig[hse-/o-; gena 'will wound' (: Skt. jighämsatv, root han-).

Because of their non-Celtic parallels, all these future formations, the subjunctive/futures of (22) and (24) and the futures of (25) and (27), must be regarded as Proto-Celtic and old. They probably go back to a period when the Celtic «-subjunctives 40 were not yet productive and the Old Irish /-futures 4 1 did not exist at all. As to the /-future which is restricted to Old Irish, this conclusion seems self-evident. But it should also be kept in mind that the ä-subjunctive (cf. Oettinger 1984) is limited to western Indo-European languages, possibly to Italic and Celtic. 42 Rix's attempt (1977: 153; 1986: 249) at tracing the ä of the Old Irish ä-subjunctive back to a desiderative suffix *-h1se-jo-\ (28)

OLat. advenat, attulat; Olr. bera < *bher-ä-t: ind. prs. beir < *bher-e-t; Olr. -ä- < *hjse-jo-\ cf. μενουσι, Homeric μενέουσι < *men-hiSO-nti

implies the separation of the Italic and Olr. ä-subjunctives. Another weakness of this theory is its assumption that the laryngeal -h- was generalized from its original position after resonants to all positions. This is contradicted by the stem formations *-sie-jo- (25) and *-se-/o- (nos. 22, 27) where such an analogical process did not take place. As we know from the Old Irish future in (27), analogy operated at a later stage and in a different way: (29)

Olr. subj. gara-: gaba- = fut. -gera < *gigrhse-jo-: χ; χ = -giba.43.

Finally, Gaulish lubiias from La Graufesenque must be taken into account, although it is found in a mutilated context: (30)

Gaul. ]sani lubiias san[ ].llias sante[ (Lejeune — Marichal 1971).

lubiias looks like a 2nd singular subjunctive of the verb also attested in Gaulish lubitus and lubv. (31)

Gaul, lubiias : lubitus, lubi : Skt. lubhyati, Lat. lubet, libet.

48

Karl Horst

Schmidt

This analysis seems to disprove Rix's hypothesis from the phonological point of view as well, as Indo-European s in intervocalic position is preserved in Gaulish. Pointing to the lack of a semantic parallel for the transition from desiderative to subjunctive (Bammesberger 1982: 66 — 67), Bammesberger (1982: 70) explains the ä-subjunctive in Old Irish and Italic as "the shortvowel subjunctive of aoristic formations in -ä- with contraction of -ä-e-f -ο- > -ä-\ This theory seems quite plausible, but if we accept it, we must reckon with the formation of a subjunctive which because of vocalic contraction cannot be differentiated from its preterite basis. A possible alternative would be the semantic derivation from the ä-preterite already proposed by Thurneysen (1884). 44 Be that as it may, the limitation of the «-subjunctive to Italic and Celtic does not seem favourable to a derivation of this category from IndoEuropean. More likely the semantic transition of aorist to subjunctive took place as an isogloss of the neighbouring languages Italic and Celtic. Comments on the Celtiberian genitive singular of o-stems in -o (21): Although in general we should expect a different sound development, 45 the best explanation up to now seems to be the derivation of this -o from the ending of the ablative singular in *-öd. Irrespective of the source of Celtiberian -o, the different reflexes of the genitive singular of o-stems in Celtic, i.e., *-o in Celtiberian vs. *-fin the other Celtic languages, need to be looked upon in the context of two facts: 1) the development of the genitive singular of o-stems in IndoEuropean whose old ending has been preserved only in Hittite -as, while it was replaced by different morphemes in the other Indo-European languages (13); 2) Old Faliscan *-osio and Osco-Umbrian *-eis: (32)

OFalisc. eko Kaisiosio; Osc.-Umbr. -eis (gen. of /-stems) : Ose. sakarakl-eis, Umbr. popl-er; -ϊ (Latin, non-Celtiberian Celtic; younger Faliscan : marci; Messapic; Venetic).

Conclusion: Old Faliscan -osio, Osco-Umbrian -eis and Celtiberian *-o confirm the late expansion of *-F which therefore cannot be taken as a feature of the Celtic or Italic proto-language, 4 6 nor can it serve as an argument in favour of the Italo-Celtic hypothesis. Watkins (1966 a: 39) already recognized "that the community of -f in Italic and Celtic is attributable to early contact, rather than to an original unity". As to the etymological identification of the morpheme -F there is no general consensus of opinion. On the basis of Wackernagel's theory (1908) which connects the genitive in -F with the Indo-Iranian adverbial case in -Γ (as

New data in the reconstruction

of the proto-language

49

in Sanskrit grämi-bhü'to get possession of a village' ( g r ä m a - ) ; krüri-Ifr'to wound' (ikrüra- 'the wound')) it may well represent an old morpheme which has been preserved in different functions in marginal languages. 47 Concluding the Celtic section, I would like to refer to Gaulish duxtir 'daughter', recently discovered in the inscription of Larzac (Lejeune et al. 1985). The new evidence confirms O'Brien's and Hamp's presumption 48 and disproves Meillet's hypothesis referring to both Italic and Celtic: "Les anciens noms du 'fils' et de la Tille' ont disparu et ont ete remplaces par de nouveaux mots" (Meillet 1908; 1922: 38).

3. Two models of interpretative enlargement Lack of time limits this part of my paper to a short remark on a special question. Its point of departure is Gamkrelidze —Ivanov's segmentation model (1984) of the agglutinating finite verbal complex (no. 33) and of the simple sentence in Indo-European (34): (33)

ΡΑΗΓΟΒΑΛ CTPYKTYPA ΦΟΡΜ ΡΛΑΑ *-mi

ΜορφβΜΗΜΗ

ο

CErMEHTHblX

I

II

III

ΓϋΑΓΟϋΒΗΗΧ

IV

V

paHr TeMarHwe-

ΚθΗΊ>Η)ΗΚ-

JlHHHbie

Cy6i>eKT-

Hacr.

Barejib-

ocaa

THB

noKa3a-

Hax

M8

HOCTh

macHax

OnrarHB

TCJIH

BepCHX

rioBejiHT.

fiocjiejjo-

St

ΜΟρφΰΜ

Kay3arHB

( G a m k r e l i d z e - 1 v a n o v 1984: 343) Sn

(34)

HaKJi.

Bpe(acneKr)

50

Karl Horst

Schmidt

Diagram (34), which is based on Wackernagel's Law (1892) that IndoEuropean enclitics occupied the second position in the sentence and on the basic word order SOV, is illustrated by Hittite evidence: (35)

Hitt. na-an-kän ku-en-zi 'h ero (-an) yÖHBaeT'; sa-an DHal-mas[u-it-ti] DSi-i-us-mi-is pa-ra-a pa-is 'h ero {-an) XajiMacyHTy (O') Bor-Ham (S) o w n ' (Gamkrelidze - Ivanov 1984: 363).

Gamkrelidze — Ivanov evidently consider the sentence particles -kan < *k/hJom (indicating perfectivity), 49 -san < *som (indicating intraterminality) 50 and -pat < *pfhJotihJ (expressing identification or emphasis) 51 as Hittite archaisms. The arrows in their diagram (34) are meant to show that the elements under A and Β are in complementary distribution, i. e., that the existence of one of the elements {s}, {o1}, {o} on the left side of the chart implies the lack of the corresponding element S, Ο', Ο on the right side (Gamkrelidze — Ivanov 1984: 362). Diagram (33) does not correspond in all points to the oldest achievable stage of the Proto-Indo-European reconstruction model. For instance, it is disputed that the subjunctive and the optative in column II are of the same age. As already mentioned, the optative has possibly been lost in Anatolian. In spite of questionable details of this kind, diagrams like (33) and (34) are of considerable typological interest and can make important contributions to a better understanding of Indo-European structures. Typologically, (33) resembles, e. g., the Kartvelian/South Caucasian verbal complex of twelve elements as arranged by Deeters (1930: 6 — 7): (36)

1) preverb, 2) personal prefix, 3) version vowel, 4) root, 5) passive affix jenj, jd/, 6) causative affixes, 7) plural affixes referring to the object, 8) present-stem formation, 9) imperfect affixes, 10) mood vowel, 11) personal ending, 12) plural suffix.

The syntagmatic order of the morpheme chain follows the structure of the language; elements 5 and 6, 7 and 8, 7 and 9 exclude each other. (37) shows the segmentation of two examples, Old Georgian damidginnes 'he will hire us' and Svan laxzazena 'he came to meet him': (37) a) b)

1 da la

2 m X

3 i

4 dg zaz

5 en

6 in

7 η

8

9

10 e a

11 s

12

Although Proto-Indo-European was a much less incorporating and agglutinating language than Proto-Kartvelian, rank segmentation seems to

New data in the reconstruction

of the proto-language

51

be a promising method for arriving at a sharper analysis of the verbal complex in Indo-European as well. Diagram (34), on the other hand, resembles the model of a polysynthetic West Caucasian language like Adyge-Cherkess, although two differences must be noted: 1) in Adyge the syntactic relations are expressed twice, i.e., outside and inside the incorporating verbal complex; 2) in Adyge an incorporating verbal form corresponds to the simple sentence on side Β of no. 34, as may be illustrated by two examples: (38)

sn

A

Β

se we I you

wa - s - e - s'e you I lead

se we I you

sd - qd - w - e - ze I you wait

4. Summary After a discussion of reconstruction principles such as substance vs. structure, comparative vs. internal reconstruction, absolute vs. relative chronology, geographic position and type, the topic "Contributions from new data to the reconstruction of the proto-language" has been subdivided into a treatment of "substantial enlargement" and "interpretative enlargement". Among the examples of new data providing substantial enlargement, the Indo-Hittite hypothesis and the reconstruction of ProtoCeltic occupied a central position. The discussion of interpretative enlargement was restricted to a brief typological evaluation of two syntagmatic models established by Gamkrelidze —Ivanov (1984), i.e., the segmentation of the finite verbal form and of the simple sentence. Notes 1. Cf. also Schmidt (1983); Collinge (1985). 2. As to the history of the laryngeal theory, cf. Polome (1965); Lindeman (1970); as to the glottalic theory, cf. G a m k r e l i d z e - I v a n o v (1972, 1973, 1984); Hopper (1973). Both theories are discussed by Mayrhofer (1986); his report is based on the results of the laryngeal theory and gives an evaluation of the glottalic theory.

52

Karl Horst

Schmidt

3. Cf. Lindeman (1970: 32 ff., 57, 70); Rix (1976: 68 ff). 4. Additional arguments are discussed by Schmidt (1983: 343 — 344). 5. Cf. Trubetzkoy (1968 [ = 1939] 217), who, however, underestimates in this article their heuristic importance; as to Trubetzkoy's "6 strukturelle Merkmale" cf. Benveniste (1966: 109) "que le takelma possede ensemble les six traits dont la reunion constituait, aux yeux de Troubetzkoy, la marque distinctive du type indo-europeen:". 6. Cf. Hjelmslev (1968: 111). 7. As to the term Gruppenflexion cf. Finck (1910: 154), who speaks of "Gruppen von Elementen, die wie erst werdende Wörter verhältnismäßig lose miteinander verbunden erscheinen." 8. As to the term Autonomie des Wortes cf. Deeters (1957: 14): "jedes Wort des Satzes ist selbständig und trägt an sich die Formantien, die seine Beziehung zu einem anderen Wort (Konkordanz) oder seine Funktion innerhalb des Satzes (Flexion) kennzeichnen." 9. As to the term Formvariation cf. Lewy (1961: 205 — 206), who gives the definition: "Bezeichnung derselben ... inneren Form durch verschiedene äußere Formen", e.g., vir-i, cui-us, stella-rum. 10. Cf. Sapir (1921: 142) and see Skalicka (1966: 1 6 0 - 1 6 1 ) : "Das wichtigste Merkmal des isolierenden Typus ist die Verminderung der Affixe, die schließlich ganz fehlen können. So entstehen kurze, einmorphematische Wörter. In den wirklichen Sprachen gibt es wenig Affixe, zugleich aber eine Menge von 'Formalwörtern'." See also Schmidt (1975). 11. As to the reconstruction of semantic systems cf., e. g., Thieme (1964); Humbach (1967); Schmitt (1967, 1968); Meid (1978). 12. Cf. Hoenigswald (1960: 102): "es-(t) 'he is' ~ er-(imus) 'we shall be' is alternating; fer-(t) 'he carries' ~ fer-(imus) 'we carry' is non-alternating; eräs 'tomorrow', cur 'why', ära 'altar', and merus 'pure' are indeterminate; as it happens, ära has old s, merus has old r"; Schmidt (1973: 1 1 1 - 1 1 2 , 124). 13. Cf. also the modifications of this principle by Leumann (1955: 155), Schlerath (1984: 57 ff.). 14. Cf. the recent discussion in InL 9 (1984) by Campanile, Crevatin, Doria, Gusmani, Lazzeroni, Neu, Ramat, Schmidt, and Strunk; moreover Rix (1986); Oettinger (1986). 15. Cf. Porzig (1954), Meid (1975), and the criticism by Schlerath (1981, 1 9 8 2 - 1 9 8 3 [1984]: 53-69). 16. Cf. Schmidt (1966, 1972 [1977]). 17. Cf. Starke (1982: 4 0 8 - 4 0 9 ) , Strunk (1984: 149), Oettinger (1986: 25). 18. Cf., e.g., Cowgill (1979: 34): "Nominal verbs, ancestral to the Indo-European perfect and the Anatolian /»-conjugation, did not distinguish tense or voice, and were based on a third singular of nominal origin ..." (cf. also footnote 14); Schmidt (1979: 97): "Wie bei einer primär Stativen Verbalklasse nicht anders zu erwarten ist, hat das idg. Perf. dem (aus der älteren dynamischen Verbalkategorie hervorgegangenen) AoristPräsens-System gegenüber ein Defizit an Kategorien: seine Merkmale sind auf imperfektiven Aspekt und intransitive Diathese eingeschränkt." 19. As to this term cf. Böhtlingk (1851, reprint 1964: 2 1 3 - 2 1 4 ) , Hirt (1904-1905: 41 ff.); Schmidt (1982). 20. Cf. also McLendon (1978: 6: "that the system splits common nouns and personal nouns for which agent function must be marked, from pronouns, kinship terms, and proper names for which patient function must be marked"; Mallinson —Blake (1981: 52); Bossong (1980).

New data in the reconstruction

of the proto-language

53

21. Cf. Watkins (1982), who gives a general survey of the development of Indo-European linguistics between 1957 and 1982. 22. Cf. Cowgill (1974, 1979); Schmid (1979: 234): "daß sich das Hethitische vor dem Griechischen und Indoiranischen im Osten des idg. Sprachgebiets so frühzeitig abgelöst hat, daß es an einzelnen gut indogermanisch geltenden Gemeinsamkeiten keinen Anteil hat". Instead of Grundsprache, Schmid (1985: 130) now thinks of a "weiträumiges Sprachkontinuum, das durch stetige Übergänge charakterisiert ist" (cf. the criticism by Rix 1986: 22 fn. 2); see also Meid (1975: 211, 1979: 161), Strunk (1984: 1 4 2 - 1 4 3 ) . 23. Cf. Oettinger (1986); Rix (1986: 20) admits "daß das Uranatolische, die gemeinsame Vorstufe des Hethitischen und der übrigen indogermanisch-anatolischen Sprachen, sich als erste der bekannt gewordenen indogermanischen Sprachgruppen aus der gemeinsamen Grundsprache ausgegliedert hat", but on the other hand he leaves the question open "wie die Grammatik der Grundsprache zum Zeitpunkt dieser Ausgliederung ausgesehen hat". 24. Cf. Cowgill (1979: 39): "the nominal form that lies at the base of the Indo-European perfect and the Anatolian /n'-conjugation would be a pure stem without case ending, used predicatively, of a formation with o-grade of the root and -e-/-o-suffix with active meaning, of the type seen in Homeric άοιδός 'singer', πομπός 'escort'; Vedic näyä'leader', säkä- 'helper, helping'; Lat. procus, coquus; OCS proroki 'prophet'; Toch. Β plewe 'raft', boat'; Hitt. is-ha-a-as 'master' ( < *sxox-o-s or *sxoi-os)"\ cf. also footnote 22. 25. Cf. Gamkrelidze — Ivanov (1984: 389 fn. 1); Strunk (1983); Rix (1986: 2 1 - 2 2 ) : "Eine morphologische Spur hat man dann auch in der Endung -s der 3. Singular des Präteritums der /^-Konjugation (nais 'er führte'; desgleichen im -s- der Endungen -sten, -steni der 2. Plural) zu erkennen geglaubt, die man auf den Ausgang -s-t des sigmatischen Aorists zurückgeführt hat. Für jenes -s wurden aber auch schon andere Erklärungen vorgeschlagen: Vermischung der 3. mit der 2. Singular oder archaische Endung der 3. Person, auf der letztlich der sigmatische Aorist aufgebaut ist...". 26. Cf. OCS bog-a < *bhäg-öd; Lith. tev-o, Latv. tgv-a < *-äd (by analogy with the stem vowel -a- : nom. sg. -as etc.) < *-öd\ Dacian Decebalus per Scorilo; Celtiberian tocoitos : cue : sarnicio : cue. 27. Cf. Strunk (1984: 1 4 4 - 1 4 5 ) , who refers to Sommer's (1947: 63) remark "Ablaut -je- : -ϊ- im Stammsuffix" and gives additional arguments for the age of the optative; in particular: "Zu den angenommenen Grundfunktionen des prähistorischen Optativs, ... Wunsch und Vorstellung (Potentialität, Irrealität) wiedergegeben zu haben, passen nahtlos die bei diesem Modus festen Sekundärendungen, die im Gegensatz zu den Primärendungen für Zeit und Wirklichkeitsbezug merkmallose neutrale Zeichen sind" (Strunk (1984: 1 4 5 - 1 4 6 ) ; cf. also Rix (1986: 20). 28. As to the age of this category cf. on the one hand Szemerenyi (1985: 528): "daß der Aspekt nicht als gemeinsames Erbe, sondern als Ergebnis paralleler, aber unabhängiger Entwicklungen anzusehen ist" (not convincing), and on the other hand Rix (1986: 12): "daß die Annahme nicht zu umgehen ist, die Aspektopposition sei schon im Urindogermanischen entstanden". 29. Cf. Risch (1975: 249); Oettinger (1979). 30. Cf. also Schmidt (1980: 42, 45 — 46). As to the problem of simple thematic present stem formation in Old Hittite cf. Oettinger (1979: 259 ff.), Rix (1986: 20, 1 4 - 1 5 ) . 31. Cf. Neu (1979), Starke (1982).

54

Karl Horst

Schmidt

32. Cf. Rix (1976); Scherer (1959: 3 1 4 - 3 6 1 ) ; Schmitt (1977: 1 1 1 - 1 1 7 ) ; H i l l e r - P a n a g l (1986). 33. Cf. Untermann (1980) with older references, who, however, rejects the Italic (Latin) hypothesis. 34. Cf. Schmidt (1972 [1977]), Ködderitzsch (1985). 35. Cf. Schmidt (1986: 171) where, however, following Thurneysen (1946: 391), the OscoUmbrian evidence is still explained by athematic inflection; cf. moreover Rix (1976: 230); L e w i s - P e d e r s e n (1974: 288). 36. Cf. Meid (1980: 1 5 - 1 6 ) and my review in IF88 (1983): 3 3 5 - 3 3 6 ; cf. moreover Schmidt (1985: 4 0 3 - 4 0 4 ) . 37. Cf. Pokorny (1959: 84 — 85): *ueg-, *uög-, *aug-, *ug- : Lat. augere-, Goth, aukan, Lith. äugti 'to grow' etc. 38, Thurneysen (1946: § 662); cf. also Schmidt (1966 a). 39. Cf. Schulze (1966). 40. Cf. Thurneysen (1946: 380 ff.), who points to "the strong verbs agid 'drives' and ad-glädathar 'addresses'" with ä-subjunctive. 41. As to the unsolved problem of the origin of the / - f u t u r e cf. Thurneysen (1946: 396 ff.); Watkins (1966: 81): unreduplicated desiderative adjectives in root + s + u, "which were affected with the new sign -ä- and given a verbal inflexion: *dasw-a-ti ...". This theory is unconvincing, as the Vedic and Celtic desiderative formations (cf. no. 27) are in general reduplicated; Quin (1978). 42, As to Venetic tolar (hap. leg.), cf. Untermann (1980: 293 — 294); as to Messapian, cf. Haas (1962: 197, 205); as to Tocharian, cf. K.T. Schmidt (1982). 43 Moreover, Rix's hypothesis does not cover weak verbs: " F ü r den ä-Subjunktiv der schwachen Verben halte ich an der traditionellen Herleitung des ä aus dem Wurzelaorist der Wurzel b''ueh2 grundsätzlich fest"; McCone (1986: 248), however, comes to the conclusion "that all Insular Celtic subjunctives were based upon the morpheme *-sef ο-, which was sometimes preceded by a stop and sometimes by *-ä-." 44 Cf. Thurneysen (1884: 286): "Also italokeltisch *bhuäm *bhuäs *bhuät sind altererbte, grundsprachliche bildungen, die, als augmentlose formen, sowohl der Vergangenheit als dem conjunctiv (und futurum) zum ausdruck dienen konnten". 45 Cf. Untermann (1967: 288), although the shortening of the *-ö in *-öd cannot be excluded: cf. the examples given by Schmoll (1959: 78). Meillet's hypothesis "mais c'est sans doute ä -Γ que ce -eis as ete substitue" (1922: 35) 46 is disproved by the lesser text frequency of /-stems; cf. also Porzig (1954: 89) who, however, leaves the question of priority open. Cf. the discussion by Devine (1970): 1) the locative theory (85 ff.); 2) Wackernagel's 47 theory (89 ff.); 3) the -oiio-theory (93 ff.); 4) the den-theory (105 ff.), especially p. I l l : "the ifevf-theory does seem less open to objection than the others". Cf. O'Brien (1956: 1 7 8 - 1 7 9 ) : Ir. *ducht(a)ir; H a m p (1975: 3 9 - 4 0 ) . 48 Cf. Josephson (1972: 416): "-kan is a sign of manner of action as well as of aspect. The perfective action is normally seen as terminal, or maybe rather postterminal as the punctual action is seen as completed". 49 Cf. Josephson (1972: 416): "-san lacks all marked reference to a term and is the sign of intraterminality. It, however, confirms the general category of terminal action and is neutral as well as negative with regard to the term". 50 Cf. Friedrich (1960: 1 5 0 - 1 5 1 ) : "-pat 'eben, a u c h ' " . Cf. R o g a v a - K e r a s e v a (1966: 97); Schmidt (1970: 4 5 0 - 4 5 1 ) . 51 52

New data in the reconstruction of the proto-language

55

On the Celtiberian genitive singular in *-öd cf. now the Lepontic parallel suggested by J. de Hoz, "El genitivo celtico de los temas en -o-, El testimonio lepontico", in: F. Villar (ed.), Studio Indogermanica et Palaeohispanica in honorem A. Tovar et L. Michelena: 315 — 329. Concerning the Latin genitive singular in -osio (Popliosio Valesioso) on the Lapis Satricanus cf. C. de Simone, in: C. M. Stibber — G. Colonna — C. de Simone — Η. S. Versenel, Lapis Satricanus ('s Gravenhage 1980) 71 fT.

References Bammesberger, Alfred 1982 "The origin of the ά-subjunctive in Irish", Eriu 33: 65 — 72. Benveniste, Emile 1966 "La classification des langues", in: Emile Benveniste, Problemes de linguistique generale (Paris: Editions Gallimard) [first published in: Conferences de l'Institut de linguistique de l'Universite de Paris 11 (1954): 33 — 50]. Böhtlingk, Otto 1851 Über die Sprache der Jakuten, in: Dr. A.Th. v. Middendorffs Reise in den äußersten Norden und Osten Sibiriens 3 (St. Petersburg) [reprint: The Hague: Mouton, 1964]. Bossong, Georg 1980 "Syntax und Semantik der Fundamentalrelation. Das Guarani als Sprache des aktiven Typs" Lingua 50: 359 — 379. Brugmann, Karl 1876 "Nasalis sonans in der indogermanischen Grundsprache", in: Georg Curtius —Karl Brugmann (eds.), Studien zur griechischen und lateinischen Grammatik 9: 285-338. Brugmann, Karl —Delbrück, Berthold 1886 — 1900 Grundriß der vergleichenden Grammatik der indogermanischen Sprachen 1—5 (Straßburg: Trübner). Collinge, Neville Ε. 1985 The laws of Indo-European (Amsterdam — Philadelphia: Benjamins). Comrie, Bernard 1981 Language universals and linguistic typology (Oxford: Blackwell). Cowgill, Warren 1965 "Evidence in Greek", in: Winter (ed.), 142-180. 1974 "More evidence for Indo-Hittite: The tense-aspect-systems", in: Proceedings of the 11th International Congress of Linguists 2 (Bologna: II Mulino), 557-570. 1979 "Anatolian /u'-conjugation and Indo-European perfect: Instalment II", in: N e u - M e i d (eds.), 2 5 - 3 9 . Cowgill, Warren —Manfred Mayrhofer 1986 Indogermanische Grammatik 1.2 (Heidelberg: Winter). Deeters, Gerhard 1930 Das kharthwelische Verbum (Leipzig: Markert & Petters). 1957 "Die Stellung der Kharthwelsprachen unter den kaukasischen Sprachen", BK 23: 1 2 - 1 6 .

56

Karl Horst Schmidt

Devine, Andrew Mackay 1970 The Latin thematic genitive singular (Oxford: Blackwell). Finck, Franz Nikolaus 1910 Die Haupttypen des Sprachbaus (Leipzig: Teubner). Forrer, Emil 1921 "Ausbeute aus den Boghazköi-Inschriften", Mitteilungen der Deutschen Orient-Gesellschaft zu Berlin 61: 20 — 39. Friedrich, Johannes 1960 Hethitisches Elementarbuch 1: Kurzgefaßte Grammatik2 (Heidelberg: Winter). Gamkrelidze, Tamaz V.—Vjaceslav V. Ivanov 1972 Lingvisticeskaja tipologija i rekonstrukcija sistemy indoevropejskix smycnyx: Konferencija po sravnitel'no-istoriceskoj grammatike indoevropejskix jazykov. Predvaritel'nye materialy (Moskva). 1973 "Sprachtypologie und die Rekonstruktion der gemeinindogermanischen Verschlüsse. Vorläufiger Bericht", Phonetica 27: 150—156. 1984 Indoevropejskij jazyk i indoevropejcy 1—2 (Tbilisi: IzdateFstvo Tbilisskogo Universiteta). Gömez-Moreno, Manuel 1925 "El bronce de Ascoli", in: Homenaje a R. Menendez Pidal 3 (Madrid), 475 — 499 [reprinted in: M. Gömez-Moreno, Misceläneas. Historia — arte — arquelogia. 1: La antigüedad (Madrid: Aguirre, 1949), 233 — 256], Grassmann, Hermann 1863 "Über die Aspiraten und ihr gleichzeitiges Vorhandensein im An- und Auslaute der Wurzeln", KZ 12: 8 1 - 1 3 8 . Haas, Otto 1962 Messapische Studien (Heidelberg: Winter). Hamp, Eric P. 1975 *dhughHter in Irish", Μ SS 33: 3 9 - 4 3 . Hermann, Eduard 1907 "Über das Rekonstruieren", KZ 41: 1 - 6 4 . Hiller, Stefan — Oswald Panagl 1986 Die frühgriechischen Texte aus mykenischer Zeit (Darmstadt: Wissenschaftliche Buchgesellschaft). Hirt, Hermann 1904 — 1905 "Über den Ursprung der Verbalflexion im Indogermanischen", IF 17: 36-84. Hjelmslev, Louis 1968 Die Sprache. Eine Einführung. Aus dem Dänischen übersetzt, für deutsche Leser eingerichtet und mit einem Nachwort versehen von Otmar Werner (Darmstadt: Wissenschaftliche Buchgesellschaft) [original version: Sproget. En introduktion (Kobenhavn: Berlingske Forlag, 1963)]. Hoenigswald, Henry M. 1960 Language change and linguistic reconstruction (Chicago: The University of Chicago Press). Hollifield, Patrick 1981 "Homeric κείω and the Greek desideratives of the type δρασείει", IF 86: 161-189. Hopper, Paul J. 1973 "Glottalized and murmured occlusives in Indo-European", Glossa 7: 141-166.

New data in the reconstruction of the proto-language

57

Humbach, Helmut 1967 "Indogermanische Dichtersprache?", MSS 21—31 [reprinted in: Klaus von See (ed.), Europäische Heldendichtung (Darmstadt: Wissenschaftliche Buchgesellschaft, 1978), 4 5 - 5 2 ] , Josephson, Folke 1972 The function of the sentence particles in Old and Middle Hittite (Uppsala: SKRIV Service AB). Ködderitzsch, Rolf 1985 "Die große Felsinschrift von Penalba de Villastar", in: Hermann M. Ölberg —Gernot Schmidt (eds.), Sprachwissenschaftliche Forschungen. Festschrift Johann Knobloch (Innsbruck: Innsbrucker Beiträge zur Sprachwissenschaft), 221-222. Lejeune, Michel —Robert Marichal 1971 "Quelques graffites inedits de la Graufesenque' (Aveyron)", CRAI 1971: 188-212.

Lejeune, Michel — Leon Fleuriot — Pierre-Yves Lambert — Robert Marichal — Alain Vernhet 1985 Le plomb magique du Larzac et les sorcieres gauloises (Paris: Editions du Centre National de la Recherche Scientifique) [also EC 22: 88 — 177]. Leskien, August 1876 Die Declination im Slavisch-Litauischen und Germanischen (Leipzig: Hirzel) [reprint: Leipzig 1963: Zentral-Antiquariat der DDR], Leumann, Manu 1955 "Baltisch und Slawisch", in: Corolla Linguistica. Festschrift Ferdinand Sommer (Wiesbaden: Harrassowitz), 154—162 [reprinted in: M. Leumann, Kleine Schriften (Zürich/Stuttgart: Artemis, 1959), 389-398]. Lewis, Henry —Holger Pedersen 1974 A concise comparative Celtic grammar3 (Göttingen: Vandenhoeck & Ruprecht). Lewy, Ernst 1961 "Die Heimatfrage", in: E. Lewy, Kleine Schriften (Berlin: Akademie-Verlag), 2 0 0 - 2 1 3 [first published: KZ 58 (1931): 1 - 1 5 ] . 1964 Der Bau der europäischen Sprachen1 (Tübingen: Narr). Lindeman, Fredrik Otto 1970 Einführung in die Laryngaltheorie (Berlin: Göschen). Lottner, Carl 1858

" U e b e r die Stellung der Italer innerhalb des indoeuropäischen stammes", KZ

7: 1 8 - 4 9 . 1861 "Celtisch-italisch", Kuhn und Schleichers Beiträge 2: 309-321. Mallinson, Graham —Barry J. Blake 1981 Language typology. Cross-linguistic studies in syntax (Amsterdam/New York/ Oxford: North-Holland). Mayrhofer, Manfred 1974 Die Arier im Vorderen Orient — ein Mythos? (Wien: Verlag der Österreichischen Akademie der Wissenschaften). 1986 "Lautlehre (Segmentale Phonologie des Indogermanischen)", in: W. Cowgill — M. Mayrhofer, 7 3 - 1 8 5 .

58

Karl Horst Schmidt

McCone, Kim R. 1986 "From Indo-European to Old Irish: Conservation and innovation in the verbal system", in: D. Ellis Evans —John G. Griffith — Ε. M. Jope (eds.), Proceedings of the 7th International Congress of Celtic Studies held at Oxford from 10th to 15th July, 1983 (Oxford: Jesus College), 222-266. McLendon, Sally 1978 "Ergativity, case and transitivity in Eastern Porno", IJAL 44: 1—9. Meid, Wolfgang 1975 "Probleme der räumlichen und zeitlichen Gliederung des Indogermanischen", in: Rix (ed.), 204-219. 1978 Dichter und Dichtkunst in indogermanischer Zeit (Innsbruck: Innsbrucker Beiträge zur Sprachwissenschaft). 1979 "Der Archaismus des Hethitischen", in: N e u - M e i d (eds.), 159-176. 1980 Gallisch oder Lateinisch? (Innsbruck: Innsbrucker Beiträge zur Sprachwissenschaft). Meillet, Antoine 1922 Les dialectes indo-europeens2 (Paris: Champion). Neu, Erich 1979 "Einige Überlegungen zu den hethitischen Kasusendungen", in: Neu —Meid (eds.), 177-196. 1980 Studien zum endungslosen 'Lokativ' des Hethitischen (Innsbruck: Innsbrucker Beiträge zur Sprachwissenschaft). 1984 "Konstruieren und Rekonstruieren", InL 9: 101-113. Neu, Erich — Wolfgang Meid (eds.) 1979 Hethitisch und Indogermanisch (Innsbruck: Innsbrucker Beiträge zur Sprachwissenschaft). O'Brien, Μ. Α. 1956 "Etymologies and notes", Celtica 3: 168-184. Oettinger, Norbert 1979 Die Stammbildung des hethitischen Verbums (Nürnberg: Hans Carl). 1984 "Zur Diskussion um den lateinischen ä-Konjunktiv", Glotta 62: 187 — 201. 1986 'Indo-Hittite'-Hypothese und Wortbildung (Innsbruck: Innsbrucker Beiträge zur Sprachwissenschaft). Pokorny, Julius 1959 Indogermanisches etymologisches Wörterbuch (Bern —München: Francke). Polome, Edgar 1965 "The laryngeal theory so far", in: Winter (ed.), 9 - 7 8 . Porzig, Walter 1954 Die Gliederung des indogermanischen Sprachgebiets (Heidelberg: Carl Winter). Quin, E. G. 1978 "The origin of the f-future: an alternative explanation", Eriu 29: 13 — 25. Risch, Ernst 1975 "Zur Entstehung des hethitischen Verbalparadigmas", in: Rix (ed.), 247 — 258. Rix, Helmut 1976 Historische Grammatik des Griechischen, Laut- und Formenlehre (Darmstadt: Wissenschaftliche Buchgesellschaft).

New data in the reconstruction

59

of the proto-language

1977

"Das keltische Verbalsystem auf dem Hintergrund des indo-iranisch-griechischen Rekonstruktionsmodells", in: Κ . H. Schmidt —Rolf Ködderitzsch (eds.), Indogermanisch und Keltisch. Kolloquium der Indogermanischen Gesellschaft am 16. und 17. Februar 1976 in Bonn (Wiesbaden: Reichert), 132 — 158.

1986

Zur Entstehung des urindogermanischen Modussystems brucker Beiträge zur Sprachwissenschaft).

(Innsbruck: Inns-

Rix, Helmut (ed.) 1975

Flexion und Wortbildung. Akten der V. Fachtagung Gesellschaft (Wiesbaden: Reichert).

der

Indogermanischen

Rogava, Georgij V. — Zajnab I. Keraseva 1966

Grammatika adygejskogo jazyka (Krasnodar —Majkop: Krasnodarskoe Kniznoe Izdatel'stvo).

Sapir, Edward 1921

Language. An introduction to the study of speech (New York: Harcourt, Brace and Co.).

Saussure, Ferdinand de 1879

Memoire sur le systeme primitif des voyelles dans les langues indo-europeennes (Leipzig: Teubner) [reprint: Hildesheim: Olms, 1968].

Scherer, Anton 1959 Handbuch der griechischen Dialekte von Albert Thumb2 (Heidelberg: Winter). Schleicher, August 1861

Compendium der vergleichenden Grammatik der indogermanischen Sprachen 1 (Weimar: Böhlau) [41876].

Schlerath, Bernfried 1981 "Ist ein Raum/Zeit-Modell für eine rekonstruierte Sprache möglich?", KZ 95: 175-202. 1982—1983 "Sprachvergleich und Rekonstruktion: Methoden und Möglichkeiten", InL 8: 5 3 - 6 9 . Schlerath, Bernfried (ed.) 1985

Grammatische Kategorien, Funktion und Geschichte. Akten der VII. Fachtagung der Indogermanischen Gesellschaft, Berlin 1983 (Wiesbaden: Reichert).

Schmid, Wolfgang P. 1979 "Das Hethitische in einem neuen Verwandtschaftsmodell", in: Neu —Meid (eds.), 231-235. 1985 "Das Lateinische und die Alteuropa-Theorie", /F90: 129-146. Schmidt, Johannes 1872 Die Verwandtschaftsverhältnisse der indogermanischen Sprachen Böhlau).

(Weimar:

Schmidt, Karl Horst 1966

"Historische Sprachvergleichung und ihre typologische Ergänzung", 116:8-22.

1966 a 1972

"Konjunktiv und Futurum im Altirischen", SCelt 1: 1 9 - 2 6 . "Probleme der Typologie (Indogermanisch/Kaukasisch)", in: Homenaje Antonio Tovar (Madrid: Gredos), 449 — 454.

ZDMG

a

1972 [1977] "Der Beitrag der keltiberischen Inschrift von Botorrita zur Rekonstruktion der protokeltischen Syntax", Word 28: 51 —62. 1973

"Grundlagen und Methoden der historischen Sprachvergleichung", in: W. A . Koch (ed.), Perspektiven der Linguistik 1 (Stuttgart: Kröner), 105 — 133.

60

Karl Horst Schmidt 1975

"Das indogermanische Kasusmorphem und seine Substituenten", in: Rix (ed.), 268-286. 1977 "Historische Sprachvergleichung und typologisches Kriterium", in: Linguistica Generalia /. Studies in linguistic typology (AUC, Philologica 5, 1974), 83-87. 1979 "Die vorgeschichtlichen Grundlagen der Kategorie 'Perfekt' im Indogermanischen und Südkaukasischen", in: S.V. Dzidziguri (ed.), A. S. Cikobava. Sbornik, posvjascennyj 80-letiju so dnja rozdenija (Tbilisi: Mecniereba), 76 — 83. 1979 a "Zur Vorgeschichte des indogermanischen Genussystems", in: Bela Brogyanyi (ed.), Studies in diachronic, synchronic, and typological linguistics. Festschrift for Oswald Szemerenyi on the occasion of his 65th birthday 2 (Amsterdam: Benjamins), 793-800. 1980 "Armenian and Indo-European", in: John A . C . Greppin (ed.), First International Conference on Armenian Linguistics. Proceedings (Delmar, N.Y.: Caravan Books), 35 — 58. 1982 "Caucasian typology as an aid to the reconstruction of Proto-Indo-European", FoSl 5: 364-383. 1983 "Indogermanisch als Diasystem", in: Manfred Faust —Roland Harweg —Werner Lehfeldt — Götz Wienold (eds.), Allgemeine Sprachwissenschaft, Sprachtypologie und Textlinguistik. Festschrift für Peter Hartmann (Tübingen: Narr), 341-348. 1983 a Kaukasische Typologie als Hilfsmittel für die Rekonstruktion des Vorindogermanischen (Innsbruck: Innsbrucker Beiträge zur Sprachwissenschaft) [ = Schmidt 1982]. 1983 b Review of Meid 1980, IF 88: 335-337. 1984 "Rekonstruktion und Ausgliederung der indogermanischen Grundsprache", InL 9: 127-133. 1985 "Probleme keltischer Etymologie und Wortbildung", in: Hermann M. Ölberg — Gernot Schmidt, Sprachwissenschaftliche Forschungen. Festschrift Johann Knobloch (Innsbruck: Innsbrucker Beiträge zur Sprachwissenschaft), 4 0 3 - 4 0 7 . 1986 "Zur Rekonstruktion des Keltischen", ZCP 41: 159 - 1 7 9 . Schmidt, Klaus T. 1982 "Spuren tiefstufiger je;-Wurzeln im tocharischen Verbalsystem", in: Tischler (ed.), 363-372. Schmitt, Rüdiger 1967 Dichtung und Dichtersprache in indogermanischer Zeit (Wiesbaden: Harrassowitz). 1977 Einführung in die griechischen Dialekte (Darmstadt: Wissenschaftliche Buchgesellschaft). Schmitt, Rüdiger (ed.) 1968 Indogermanische Dichtersprache (Darmstadt: Wissenschaftliche Buchgesellschaft). Schmoll, Ulrich 1959 Die Sprachen der vorkeltischen Indogermanen Hispaniens und das Keltiberische (Wiesbaden: Harrassowitz).

New data in the reconstruction of the proto-language

61

Schulze, Wilhelm 1904 "Lit. kläusiu und das indogermanische Futurum", BSB 1904: 1434-1442 [reprinted in: W. Schulze, Kleine Schriften2 (Göttingen: Vandenhoeck & Ruprecht, 1966), 101-109]. Schwyzer, Eduard 1953 Griechische Grammatik l 2 (München: Beck). Silverstein, Michael 1976 "Hierarchy of features and ergativity", in: R. M. W. Dixon (ed.), Grammatical categories in Australian languages (New Haven: Humanities Press), 112 — 171. Skalicka, Vladimir 1976 "Ein typologisches Konstrukt", TLP 2: 157-163. Sommer, Ferdinand 1947 Hethiter und Hethitisch (Stuttgart: Kohlhammer). Starke, Frank 1977 Die Funktionen der dimensionalen Kasus und Adverbien im Althethitischen (Wiesbaden: Harrassowitz). 1982 "Die Kasusendungen der luwischen Sprachen", in: Tischler (ed.), 407 — 425. Strunk, Klaus 1984 "Probleme der Sprachrekonstruktion und das Fehlen zweier Modi im Hethitischen", InL 9: 135-152. 1985 "Flexionskategorien mit akrostatischem Akzent und die sigmatischen Aoriste", in: Schlerath (ed.), 4 9 0 - 5 1 4 . Sturtevant, Edgar H. 1962 "The Indo-Hittite hypothesis", Language 38: 105-110. 1964 A comparative grammar of the Hittite language. Revised edition, 2nd printing (New Haven: Yale University Press/London: Oxford University Press). Szemerenyi, Oswald 1985 "Strukturelle Probleme der indogermanischen Flexion. Prinzipien und Modellfälle", in: Schlerath (ed.), 515-533. Thieme, Paul 1964 "The comparative method for reconstruction in linguistics", in: Dell Hymes (ed.), Language in culture and society (New York —Evanston —London: Harper & Row). Thomas, Werner 1985 Die Erforschung des Tocharischen (1960-1984) (Stuttgart: Steiner). Thurneysen, Rudolf 1884 "Der italokeltische conjunctiv mit a \ Bezzenbergers Beiträge 8: 269 — 288. 1946 A grammar of Old Irish. Revised and enlarged edition with supplement. Translated from the German by D. A. Binchy and Osborn Bergin (Dublin: The Dublin Institute for Advanced Studies). Tischler, Johann (ed.) 1982 Serta Indogermanica. Festschrift Günter Neumann (Innsbruck: Innsbrucker Beiträge zur Sprachwissenschaft). Tovar, Antonio 1946 "Las inscripciones ibericas y la lengua de los celtiberos", Boletin de la Real Academia Espanola 25: 7 — 42. 1973 "Die iberischen Inschriften und die Sprache der Keltiberer", in: A. Tovar, Sprachen und Inschriften (Amsterdam: Grüner), 124 — 158 ( = Tovar 1946).

62

Karl Horst Schmidt

Trubetzkoy, N. S. 1939 "Gedanken über das Indogermanenproblem", AL 1: 81—89 [reprinted in: Anton Scherer (ed.), Die Urheimat der Indogermanen (Darmstadt: Wissenschaftliche Buchgesellschaft) 214—233]. Untermann, Jürgen 1967 "Die Endung des Genetiv singularis der o-Stämme im Keltiberischen", in: Wolfgang Meid (ed.), Beiträge zur Indogermanistik und Keltologie. Julius Pokorny zum 80. Geburtstag (Innsbruck: Innsbrucker Beiträge zur Kulturwissenschaft), 281-288. 1980 "Die venetische Sprache", Glotta 58: 281-317. Verner, Karl Adolf 1875 "Eine ausnahme der ersten lautverschiebung", KZ 23.2: 97 — 130. Wackernagel, Jacob 1892 "Über ein Gesetz der indogermanischen Wortstellung", IF 1: 333 — 436 [reprinted in: Wackernagel 1953, 1 — 104]. 1908 "Genetiv und Adjektiv", in: Melanges de linguistique offerts ä M. Ferdinand de Saussure (Paris: Champion), 125 — 152 [reprinted in: Wackernagel 1953, 1346-1373]. 1953 Kleine Schriften. Herausgegeben von der Akademie der Wissenschaften zu Göttingen (Göttingen: Vandenhoeck und Ruprecht). Watkins, Calvert 1966 "The origin of the/-future", Eriu 20: 6 7 - 8 1 . 1966 a "The interrelationships within Italic", in: Henrik Birnbaum—Jaan Puhvel (eds.), Ancient Indo-European dialects (Berkeley — Los Angeles: University of California Press) 2 9 - 5 0 . 1983 "Historical comparative linguistics and its contribution to typological studies", in: Shirö Hattori — Kazuko Inoue (eds.), Proceedings of the XHIth International Congress of Linguists, August 29—September 4, 1982, Tokyo (Tokyo: Gakushuin University), 270 — 277. Winter, Werner (ed.) 1965 Evidence for laryngeals (The Hague: Mouton).

Comparative reconstruction and typological verification: The case of Indo-European Thomas V. Gamkrelidze

An application of the principles of language typology and linguistic universale to theoretical premises of historical comparative linguistics and language reconstruction necessitates a reformulation of the aims and tasks of historical linguistics in general, and of linguistic reconstruction, in particular. The principle of typological plausibility, both synchronic and diachronic, of a postulated model for a proto-language provides a new approach to comparative linguistic studies and calls for a revision of traditional views on reconstructed proto-linguistic systems, in particular those on Proto-Indo-European and its daughter dialects. On the basis of comparative and typological evidence, the traditional and classical threefold system of Proto-Indo-European stops must be given a phonetic reinterpretation whereby the traditional "plain voiced" stops might be viewed as glottalized. The stop series in the new interpretation must be defined as: I, glottalized; II, voiced [aspirates]; III, voiceless [aspirates], where aspiration in II and III is a phonetically relevant but phonemically redundant feature. This revision of the consonantism is known in current IndoEuropean comparative studies as the "glottalic theory", which takes a new look at the Proto-Indo-European linguistic model and its diachronic transformations into the historical Indo-European languages. On the glottalic analysis these transformations prove to be totally different from those traditionally assumed. The archaic Proto-Indo-European stop inventory proves to be closer to those of languages traditionally viewed as having undergone later consonant shift or "Lautverschiebung" (Germanic, Armenian, Hittite), while languages traditionally considered phonologically conservative (especially Old Indie) prove to have undergone complex phonemic transformations in their consonantism. The typological approach to linguistic reconstruction necessitates a radical reinterpretation and reformulation of all the basic comparative work in IndoEuropean.

64

Thomas V. Gamkrelidze

I would like to recall in this connection Winfred Lehmann's words: "What had seemed one of the most solid achievements of 19th century linguistics is now modified in every section" (Lehmann 1983). Such a reinterpretation and reformulation of the basic comparative work in IndoEuropean has been undertaken by Vjaceslav Ivanov and myself in a recent two-volume monograph (Gamkrelidze — Ivanov 1984) soon to be published in English translation. The criticism levelled at the glottalic theory concerns mainly the adopted methodology of linguistic reconstruction, styled by some scholars "typological reconstruction" as opposed to traditional reconstruction viewed as "comparative reconstruction", which is considered to be the only methodologically legitimate procedure of language reconstruction, typology being a mere fallacy (cf. Dunkel 1981). I consider this to be a misunderstanding of the aims and tasks of linguistic reconstruction in general and of Indo-European reconstruction, in particular. There is no such procedure as typological reconstruction as opposed in principle to comparative or internal reconstruction. Current methodological premises of language reconstruction entail, of necessity, involving typological considerations in the process of comparative and internal reconstruction. Any linguistic reconstruction must be based on comparative evidence taking at the same time into account the typological plausibility, both synchronic and diachronic, of a linguistic system arrived at by means of comparative and internal reconstruction. To put it the other way, comparative reconstruction must go hand-inhand with typology and language universale, so as not to obtain by comparative reconstruction a system which is linguistically implausible and constituting an exception to typologically verifiable linguistic evidence. Consequently, we must speak in diachronic linguistics solely of comparative language reconstruction aided in some cases by internal reconstruction of the protolinguistic patterns, typology and language universale appearing merely as verification criteria for the proposed reconstructions. Typologically verified linguistic models arrived at by comparative and internal reconstruction must be given preference over typologically rare and implausible patterns that may also be posited theoretically on the basis of language comparison. Among diverse theoretical patterns of linguistic reconstruction arrived at with the aid of genetic comparison of related dialects typological criteria must provide help in singling out one of them which can be considered linguistically most plausible and realistic, explaining a large set of historical facts that remain unaccountable from the viewpoint of the alternative reconstructed

Comparative

reconstruction

and typological

verification

65

models. All these considerations must be involved in the procedure of comparative and internal reconstruction paying due attention to typological criteria of the linguistic plausibility of the theoretically postulated linguistic models that must reflect in the first approximation a protosystem existing in space and time. If we had a linguistic proto-system with very rare and exceptional characteristics historically attested, we would be called upon to account for its exceptional structural features setting up pre-stages to justify its peculiar and typologically exceptional traits. This would be a methodologically acceptable procedure accounting for the typological peculiarity of a historically attested linguistic system which served as a proto-system to a group of related dialects. This is precisely what is being done now by some scholars in attempts to justify the peculiar structural characteristics of the traditionally reconstructed consonantism of the Proto-Indo-European linguistic system, with three series of stops defined as plain voiced; voiced aspirates : voiceless, as if this set were not a theoretically posited linguistic construct, but a historically attested and recorded linguistic system, whose structural peculiarities should be somehow justified and accounted for. Our contention is that the Proto-Indo-European stop series should not have been posited, from the very beginning, in the shape as it had been traditionally proposed, this being the result of mere historical chance due to the influence of the then prestigious Old Indian system and to the absence at those times of a strict reconstructional methodology. As a matter of fact, in a series of phonemic correspondences d:d:d:d:t:t, etc., what sort of an entity should be posited for the protosystem, a *d, *t or a third sound, different from both the historically attested ones? Logically all three possibilities may be envisaged, since any of these entities is a priori not ruled out. The decision in such cases must rest wholly with typological considerations, with a view to obtain such a linguistic system which on the whole would be linguistically more probable and plausible, not constituting an exception to general typological evidence. That is why in this series of correspondences the preference must be given to positing, for the Proto-Indo-European system, such an entity which is phonemically unvoiced and characterized by an additional distinctive feature of glottalization. 1 Now, we are advised in some attempts to justify and rescue the traditional Proto-Indo-European consonantism, as if it were a historically attested system, and not a hypothetical construct as any other linguistic reconstruction, to view the plain voiced stops with highly marked labial

66

Thomas V. Gamkrelidze

*b2 and very common and unmarked velar *g as a result of a transformation of a previous system at a pre-Indo-European stage with voiced implosives. Coming up with new suggestions and alternative theories for the ProtoIndo-European system of stops has become very popular this last decade after the presentation, by us in 1972 and by Paul Hopper in 1973, of the glottalic theory. 3 As regards positing voiced implosives in pre-IndoEuropean instead of glottalized stops, as proposed by Haider (1985), I would like to point out that the series of voiced implosives, as shown by Greenberg (1970), is characterized by the same hierarchical relationship of markedness or dominance as the plain voiced stops (unmarked — dominant — labial: marked — recessive or totally absent — velar member), this being in contradiction with the evidence of the traditional plain voiced stops in Proto-Indo-European with highly marked labial *b and unmarked velar *g. The pre-Indo-European voiced implosives simply could not have yielded what is traditionally known in Indo-European as the series of "plain voiced stops". 4 Setting up such pre-stages for Proto-Indo-European with different sorts of phonemes to account for typological inconsistencies of the traditional system is as old as the first attempt to reinterpret the classical system undertaken by Holger Pedersen (1951) who suggested introducing such changes at the pre-Indo-European stage ("Vorindoeuropäisch") leaving intact the traditional system of Proto-Indo-European ("Gemeinindoeuropäisch"). Such internal reconstructions of different typologically consistent pre-Indo-European stages leave still unexplained the fact of the transition from such presumably stable configurations to the highly unstable system known as the traditional Proto-Indo-European one which later allegedly transformed once again into typologically stable systems of the historical Indo-European dialects (cf. Cowgill 1984 [1985]: 6). 5 I would call all such newly invented constructs fallacious attempts to rescue and salvage at all cost the traditionally received view on ProtoIndo-European, despite the fact that the contradictory character and the disadvantages of the classical Indo-European paradigm become more and more evident in current Indo-European comparative studies. I would like to adduce in conclusion the words of one of the proponents of the glottalic theory — the American linguist Paul Hopper — speculating on the strategies which those Indo-Europeanists who have not yet accepted the glottalic theory will use to confront it: Will they attempt to show that the data used in the formation of the theory are wrong (root structure constraints, distribution of *b, morphophonemics

Comparative reconstruction and typological verification

67

of mediae, point of articulation skewing)? That the data are correct but irrelevant? Or compatible with some other typological facts? Will they maintain that the theory is plausible but applicable to pre-Indo-European and can be complacently ignored by Indo-Europeanists? Will they cast aspersions on the method of external reconstruction itself, in the hope that none of the internal evidence for the theory can be sustained? Perhaps, on the other hand, those who work in the traditional paradigm will find it less troublesome to ignore the radical revisions which have been presented during the past decade in the hope that these innovations will die of neglect (which is not likely to happen) or that, being accepted by a widening circle of scholars, the changes proposed will have n o significant consequences for our view of Proto-Indo-European (an even remoter possibility). (Hopper 1982) I f o r m y p a r t f i r m l y b e l i e v e t h a t the g l o t t a l i c t h e o r y a s a n e w p a r a d i g m in I n d o - E u r o p e a n c o m p a r a t i v e l i n g u i s t i c s will g a i n w i t h t i m e a n e v e r w i d e n i n g a c c e p t a n c e a m o n g I n d o - E u r o p e a n s c h o l a r s o f all g e n e r a t i o n s , this b e i n g a s t r o n g i m p e t u s t o f u r t h e r d e v e l o p m e n t o f

Indo-European

studies, m a k i n g them m o r e theory-oriented and b r o a d e n i n g considerably their s c o p e o f r e s e a r c h . Notes 1. There has been an objection raised to positing glottalized obstruents for Proto-IndoEuropean in view of the absence in the system of lingual (strident) affricates. Such an objection presupposes an implicational universal of an existence in the system of strident affricates implied by the presence in the inventory of glottalized stops. Lingual (strident) affricates are wide-spread in the languages of the world, either with glottalized obstruents or without them (cf. Ruhlen 1977; Maddieson 1984). The only implicational universal which may be inferred on empirical evidence linking glottalized stops with affricates would be: If a language with glottalized stops is characterized at the same time by the existence of lingual (strident) affricates, one of the affricate series would also be glottalized. That is to say, in the subsystem of strident affricates we may expect the existence of the glottalized affricates of the type c', c', in view of the presence in the inventory of the glottalized stops p't' k', but not conversely: the existence of glottalized phonemes in the subsystem of stops does not imply by itself the presence of affricates. Instances of phonemic systems with glottalized stops but without affricates would be: Ge'ez (Ethiopic), Kapau (Indo-Pacific: Southwest Guinea), Maidu (Penutian: Northern California), Nez Perce (Penutian: Idaho), although systems both with glottalized stops and affricates are much more common due mainly to the widespread presence of strident affricates in the inventories of the languages of the world (cf. Ruhlen 1975). Even Proto-Indo-European, as assumed by Vennemann (1985), may have contained lingual affricates — at least allophonically. 2. In his recent attempt to refute the glottalic theory Szemerenyi tries to reject the thesis of the absence in Proto-Indo-European of voiced labial *b referring to forms with b in medial position: Lat. lübricus, libö, Goth, diups. He admits that "initially b is rare, perhaps not to be acknowledged at all; but internally it is vigorously represented" (cf.

68

Thomas V. Gamkrelidze

Szemerenyi 1985: 12). But this "vigorous representation" internally of b is restricted mainly to Western ("ancient European") dialects casting thus doubts on its Proto-IndoEuropean character. Furthermore, dealing with root-restrictions in Proto-Indo-European Szemerenyi tries to account from the traditional point of view for the MA-T and T-MA restrictions by assimilation (which is our explanation as well, in the light of the glottalic theory), leaving the M-M restriction unaccounted for, this having no satisfactory traditional explanation. Reading this work by Oswald Szemerenyi, one cannot get rid of the impression that one of the earliest proponents of the "new look" on ProtoIndo-European has abandoned his previous approach to the theoretical problems of reconstruction trying to reject all attempts of others to follow the same theoretical lines. 3. On the other hand, some of the adherents of the glottalic theory, having even contributed to some extent to its elaboration and development and basing their reconstructions on its premises, do not even feel obliged to mention by name in their papers their predecessors — the originators of the glottalic theory. The procedure is simple: In their first articles manifesting acceptance of the glottalic theory and adherence to the views of its first proponents concerning the originally glottalic character of the traditional Proto-Indo-European plain voiced stops, with all the structural consequences involved, the authors of the theory are mentioned by name, however in the ensuing works dealing with Proto-Indo-European glottalics and the history of their postulation reference is made not to the original work of the first proponents of the theory, but to their own earlier articles, creating thus a false impression and leaving the reader unaware of the genuine authors of the glottalic theory. As different from such tendencies, I would like to refer in this connection to Manfred Mayrhofer's excellent recent work on IndoEuropean comparative phonology (1986), a real state-of-the-art account, with an objective picture of the developments of the Proto-Indo-European phonological studies (cf. also his brilliant lecture of 1983 on the history of comparative Indo-European linguistics). 4. It seems to be untenable to try to account for this fact assuming change of the postulated pre-Indo-European implosive *'b to Proto-Indo-European *m (Haider 1985: 12), while *'d and *'g changed to Proto-Indo-European *d, *g respectively, leaving in the new series of Proto-Indo-European plain voiced stops a gap at the bilabial point which, by the way, is a favoured point of articulation in the series of voiced stops, as it is in the series of voiced implosives (viewing the highly dubious Proto-Indo-European root *bel'force' as an instance of voiced *b in Proto-Indo-European cannot, of course, save the situation). Apart from this, positing "voiced implosives", even for the pre-Indo-European stage, leaves unexplained the root-constraint ruling out the co-occurrence of two voiced stops, this being one of the most conspicuous typological inconsistencies of the classical Proto-Indo-European system. This constraint is well accounted for phonetically on the assumption of the rule of non-cooccurrence of two glottalized consonants, this being widely supported by typological evidence, as different from the co-occurrence of "voiced implosives". Haider would like to see insurmountable difficulties with glottalized consonants in connection with the reduplication structures of the type of Gr. didö-mi, Ο. Ind. da-dä-mi (Haider 1985: 8). Even if we consider this type of verbal reduplication as of Proto-Indo-European origin, there is no difficulty in positing a Proto-Indo-European structure t'V-foH-, with two homorganic glottalized stops in a sequence, since the typological constraint concerns the tendency of non-cooccurrence

Comparative reconstruction

and typological verification

69

of two heterorganic glottalized stops, while two homorganic glottalized consonants may combine freely in a root or a word-form. 5. Such an unstable and "less economic" traditional Proto-Indo-European system of stops is assumed also by the Polish linguist and Armenologist Andrzej Pisowicz who tries to "refute" the glottalic theory on the strength of Armenian evidence. His "refutation" of the glottalics in Proto-Indo-European Pisowicz, oddly enough, bases on the analysis of "voiced aspirates" in modern Armenian dialects, sounds which he considers to be "murmured consonants". Not arguing with him about the phonetic character of these consonants in modern Armenian dialects we still uphold the view that these sounds along with plain voiced stops are allophones of single phonemes which may be characterized as voiced stops, since they are in complementary distribution showing different patterns in diverse Armenian dialects. The principal question here is the complementarity of distribution, and not its particular patterns. We never intended to say that the distribution: aspirates initially — non-aspirates medially, is valid particularly for the dialect of Djulfa as described by Allen (1951). Allen's article is referred to in our work in connection with the general thesis of the complementarity of voiced aspirates ~ non-aspirates in modern Armenian dialects (cf. Gamkrelidze — Ivanov 1984: 42). As for the particular pattern: aspirates initially — non-aspirates medially, it is valid precisely not for Djulfa, but for a number of other modern Armenian dialects, such as the dialects from the Ararat area and others. But all this has nothing to do with the postulation of glottalics in Proto-Indo-European, and has only a bearing on the status of "voiced aspirates" in Armenian, a question on which Pisowicz shows strong adherence to traditional views. This status can be interpreted diachronically both ways as a retention of the aspiration or as its later development in Armenian dialects, an assumption that is less plausible in view of an absence in the linguistic environment of Armenian of languages with this type of sounds, be it qualified phonetically as "voiced aspirates" or as stops with '"aynized" pronunciation, as proposed by Pisowicz. One more question raised by Pisowicz that he considers to support traditional views on the Proto-Indo-European stops, in opposition to the glottalic theory, is the development of the Proto-Indo-European sequence *du- (i.e. *t'u-) in Armenian as erk-. He seems to ignore a rich literature on the subject elucidating different ways of such a development, one of the alternatives being a postulation of an unvoiced sequence -tuyielding immediately the voiceless k without positing an intermediary stage of devoicing g > k through consonant shift in Armenian that is absent in clear instances of g resulting from u (cf. gini "wine", get "water", kogi "butter", etc.) (for the latest literature on the subject cf. Szemerenyi 1985; Vennemann 1986). I consider it unnecessary to enlarge here on Pisowicz's rather outdated and superficial considerations on the phonetic character of the glottalized consonants — ejectives and injectives, on the non-voiced character of the glottalized stops, etc., a question that is dealt with in detail in our monograph. I would rather advise the author to consult in this connection Greenberg's classical paper of 1970 on glottalic stops that Pisowicz seems not to be familiar with. Another Armenologist, G. B. Dzaukjan, is also skeptical about the glottalic theory, primarily because of the wrongly understood status of the phoneme *b in Proto-IndoEuropean (cf. Dzaukjan 1986). Suffice it to say in this connection that this scholar continues to adduce in his critique OInd. pibati, Lat. bibit, etc., as illustrations of a normal distribution of the voiced *b in Proto-Indo-European (!). His recent article does not add anything new and essential to his earlier objections against positing glottalics

70

Thomas V. Gamkrelidze for Proto-Indo-European that reflect, as indicated in our monograph, a lack of understanding of the problems involved in the typological approach to the comparative reconstruction of Proto-Indo-European.

References Allen, William Sidney 1951 "Notes on the phonetics of an Eastern Armenian speaker", Transactions of the Philological Society 1950: 180-206. Cowgill, Warren 1984 (1985)Review of Yoel A. Arbeitman — Allan R. Bomhard (eds.), Bono homini donum, Kratylos 29: 1 - 1 3 . Dzaukjan, Gevork B. 1986 "O tak nazyvaemoj 'glottal'noj teorii' ν indoevropeistike" [On the so-called "glottalic theory" in Indo-European studies], Vestnik drevnej istorii 1986.3: 160-165. Dunkel, George 1981 "Typology versus reconstruction", in: Yoel B. Arbeitman —Allan R. Bomhard, Bono homini donum. Essays in historical linguistics in memory of J. Alexander Kerns (Amsterdam: Benjamins), 559 — 569. Gamkrelidze, Tamaz V. — Vjaceslav V. Ivanov 1972 "Lingvisticeskaja tipologija i rekonstrukcija sistemy indoevropejskix smycnyx [Linguistic typology and the reconstruction of Indo-European stops]", in: Konferencija po sravnitel'no-istoriceskoj grammatike indoevropejskix jazykov. Predvaritel'nye materialy (Moskva: Nauka), 15 — 18. 1984 Indoevropejskij jazyk i indoevropejcy. Rekonstrukcija i istoriko-tipologiceskij analiz prajazyka i protokul'tury [Indo-European and the Indo-Europeans. A reconstruction and historical-typological analysis of a proto-language and a proto-culture] 1—2 (Tbilisi: Izdatel'stvo Tbilisskogo Universiteta). Greenberg, Joseph H. 1970 "Some generalizations concerning glottalic stops, especially implosives", I J AL 36: 123-145. Haider, Hubert 1985 "The fallacy of typology. Remarks on the Proto-Indo-European stop system", Lingua 65: 1—27. Hopper, Paul J. 1973 "Glottalized and murmured occlusives in Indo-European", Glossa 7: 141-166. 1982 "Areal typology and the early Indo-European consonant system", in: Edgar C. Polome (ed.), The Indo-Europeans in the fourth and third millennia (Ann Arbor, Michigan: Karoma), 121 — 139. Lehmann, Winfred P. 1983 "Proto-Indo-European phonology. An overview over recent research". Maddieson, Ian 1984 Patterns of sounds (Cambridge: Cambridge University Press). Mayrhofer, Manfred 1983 "Sanskrit und die Sprachen Alteuropas. Zwei Jahrhunderte des Widerspiels von Entdeckungen und Irrtümern", Nachrichten der Göttinger Akademie der Wissenschaften 1983.5: 121-154.

Comparative reconstruction and typological verification 1986

71

Indogermanische Grammatik 1.2: Lautlehre (Segmentale Phonologie des Indogermanischen) (Heidelberg: Winter). Pedersen, Holger 1951 Die gemeinindoeuropäischen und die vorindoeuropäischen Verschlusslaute (Det Kgl. Danske Videnskabernes Selskab, Historisk-filologiske Meddelelser 32.5) (Kobenhavn). Pisowicz, Andrzej 1989 "Objections d'un armenologue contre la theorie glottale", Folia Orientalia 25 (1988): 213-225. Ruhlen, Merritt 1975 A guide to the languages of the world (Stanford, CA: University Press). 1977 "The geographical and genetic distribution of linguistic features", in: Linguistic studies offered to Joseph Greenberg on the occasion of his sixtieth birthday (Saratoga: Anma Libri), 137 — 160. Szemerenyi, Oswald 1985a "Recent developments in Indo-European linguistics", Transactions of the Philological Society 1985: 1 - 7 1 . 1985b "Armenian between Iran and Greece", in: Ursula Pieper — Gerhard Stickel (eds.), Studia linguistica diachronica et synchronica Werner Winter sexagenario (Berlin —New York —Amsterdam: Mouton de Gruyter), 783 — 799. Vennemann, Theo 1986 "Syllable-based sound changes in Early Armenian", Annual of Armenian Linguistics 7: 27—43. 1989 "Phonological and morphological consequences of the glottalic theory", in: Vennemann (ed.), 107-121. Vennemann, Theo (ed.) 1989 The new sound of Indo-European. Essays in phonological reconstruction (Berlin—New York: Mouton de Gruyter).

Old and new thoughts about the configuration of the Romance language family Yakov Μ alkie I

There exists, in the annals of Romance scholarship, no sustained tradition of heated controversy about the most cogent classification of constituent languages and dialects. To be sure, a brief, isolated attempt was made, shortly after 1800, by the Adelung-Vater team, to separate Rumanian from the other "daughters of Latin", on the theory that it qualified for the unique status of a Slavo-Romance language; but that approach was not destined to bear fruit. 1 When Friedrich Diez, in his pioneering writings, referred sporadically to "Western", "Northern", or "Southern" members of the family or, more succinctly, to the "West", etc., he had in mind a purely geographic localization of individual phenomena or processes, not the crystallization of any downright subfamilies, geographically identified. Then again, if his comparative etymological dictionary and, much later, that of his successor Wilhelm Meyer-Lübke each listed the Romance reflexes of ancestral bases (for the most part, Latin) in certain standardized sequences, the choice of such a sequence was a purely editorial decision, devoid of any profound classificatory implications; the same holds for Gustav Gröber's trail-blazing "Substrate..." 2 Graziadio Isaia Ascoli, a student of Romance also known as an accomplished Indo-Europeanist, mustered greater courage on two occasions: From the dialects of three or four scattered Alpine and Subalpine areas he extrapolated the one territorially disjointed language with which Romance scholarship has ever since operated, namely R(h)aeto-Romance; somewhat less successfully, he soon afterward attempted to subdivide France (and the adjoining sections of Belgium and Switzerland) into three rather than two language zones, adding Franco-Provengal to French proper, in the North, and to Occitan or Provencal, in the South. However, critics argued, first, that Franco-Provengal was little more than a sharply characterized dialect group scarcely comparable in rank to its reputed peers, and, second, that there sprouted in the territory of Gaul other such dialect groups, e. g., Gascon. 3 Since the time when the opening volume of Meyer-Lübke's comparative Romance grammar rolled off the press, i. e., since 1890, its author deemed

74

Yak ο ν Malkiel

it expedient to deal with Southern French and with Catalan, despite the dividing wall of the Pyrenees, as, basically, one and the same language, without, at first, making any big issue out of that assumption. 4 Things changed measurably, however, when, as an aging scholar, he published, in 1925, the single most controversial monograph of his entire career, namely Das Katalanische, whose lengthy subtitle indicated that one of his purposes was to determine the position of the chosen language vis-avis Provencal and Spanish, while the other goal was to bring to bear on the issue both selected straight historical arguments and those furnished by diachronic linguistics.5 Not surprisingly, Meyer-Lübke, by comparing feature after feature of the parent language with their respective reflexes in the records of the three medieval and modern contenders for his attention, came up with conclusions that, on the whole, favored Catalonia's closer proximity to the trans-Pyrenean North than to the peninsular Center and West; in essence, Carolingian conquests, starting with the early 8th century, were credited with having yanked loose Catalonia from its previously closer connection with Proto-Spanish. No doubt, in an effort to dramatize the contrast, the author operated with Castilian as the most characteristic representative of the Spanish (or HispanoRomance, or Ibero-Romance) alliance of dialects. Practically all critics took to task the ranking doyen of Romanists for having chosen this odd modus operandi, i.e., for having willfully neglected the evidence of the interjacent, transitional cluster of dialects between Castilian and Catalan known as Navarro-Aragonese. 6 Two reviewers, however, that same year went much farther in their sophisticated attacks, virtually re-writing the entire book. One of them, Antoni(o) Griera, saw in Catalan, his native tongue, not a fairly late, if heavy, overlay of Gallo-Romance concealing an old layer of the Mediterranean Coast variety of Hispano-Romance, but something radically different and, to him and to his friends, vastly more exciting: Already in Antiquity, he pleaded, there began to develop a pattern of provincial differentiation of colloquial Latin, with the focal point of the Peninsula — not Old Castile hugging the Bay of Biscay by any means, but Baetica or Andalusia at the southern tip — practically coinciding with AfroLatin, while the ancestor of Catalan, from time immemorial, gravitated toward Roman Gaul. 7 This way of seeing things transformed the SpanishCatalan frontier into one of the deepest cleavages to be encountered in Romance territory. Meyer-Lübke, to be sure, parried Griera's reproaches of arbitrariness (1926-1927: 116-128); yet, more effective than the senior scholar's rejoinder in weakening the impact of Griera's strictures

The configuration

of the Romance language family

75

was the eccentricity of the Catalan critic's own counterview, e. g., his eagerness to draw sweeping conclusions from our fragmentary, not to say minimal, knowledge of spoken Afro-Latin. 8 The fact that Griera, thereafter, compromised his reputation by embarking on all sorts of eccentricities cast further shadow on his 1925 role as implacable critic of Meyer-Lübke. 9 The second major attack, conducted in an entirely different key and direction, 10 came from Madrid quarters. A young, dynamic, and talented student of Ramon Menendez Pidal, namely Amado Alonso, of Navarrese background and thus predestined to become emotionally involved in the ethnoglottal controversy, came up with an exceptionally long critique, of which two installments were published immediately in the journal controlled by his teacher, while a mere draft of the third part made its appearance, anticlimactically enough, with a delay of almost twenty years and on a different continent, with a consequent loss of momentum. 11 On indirect but overabundant evidence, Menendez Pidal himself was the tacit co-author of the inordinately long critique, and he had excellent reasons indeed for allying himself with his prize pupil, for his own masterwork, the Origenes del espahol (1926), which carried an entirely different message, was already in press at that point. The strength of the latter work and of its two subsequent revisions (1929, 1950) lies not in any searching analysis of the subjacent Latinity, 12 but rather in the meticulous observation of the formative periods of the Peninsular vernacular languages, a segment of time which followed upon the Moorish invasion that started with the crossing of the Straits of Gibraltar in 711. Within Menendez Pidal's (and Alonso's) schema, Castilian and the dialects immediately adjacent to it in the West (Asturo-Leonese) and the East (NavarroAragonese) formed the powerful center of the Christians' long-drawnout counterattack against the Arabs (the reconquista), with GalicianPortuguese, to the extreme West, and Catalan, to the extreme East, constituting the flanks, as it were. The heavy coloring of Catalan with trans-Pyrenean ingredients was not denied, neither was it made the cornerstone of the entire classificatory edifice: Catalan remained solidly enclosed within the Peninsula. That same year, but above all after resuming his program of research in the wake of his return to Spain from exile (1940), Menendez Pidal began to investigate traces of Oscan, allegedly imported by legionnaires recruited in South Italy, in the dialect speech and toponymy of the Ebro Valley which, as historiographers and historians agree — began to be Romanized in the second century B.C. 13 Little by little, he also tentatively

76

Yakov

Malkiel

recognized certain connections between individual features of Spanish and their counterparts in Romanized South Italian. 14 But by then, the heated controversy set off by Meyer-Lübke's provocative book had subsided. In calm retrospect, the debate fell short of yielding any truly impressive results. The second major classificatory controversy was started, one might say, singlehandedly by the Swiss Romanist Walther von Wartburg, in the mid 'thirties. Known and admired, until his transfer from Lausanne to Leipzig, chiefly as a studious and impressively meticulous lexicographer, with a high degree of specialization in Gallo-Romance, von Wartburg launched his career at a leading German university, with a reputation for excellence in linguistics, with an inaugural lecture on the influence of Germanic languages on French lexis (1930: 309 — 325); delivered, two years later, an academy lecture (never published in full) on the reasons for the eventual division of Gaul into a northern French and a southern Provengal zone; produced, at an amazing speed, a preliminary study, ambitious by implication, of the rise of language borders within the farflung Romance domain (1934: 209 — 227); and, after the lapse of two more years, published in a tone-setting journal (whose editorship had meanwhile fallen into his lap) a sensational fifty-page article, accompanied by seven folding maps, on the crystallization of Romance language spaces, a slightly pompous metaphor for 'areas', or 'domains', or else 'territories': "Die Ausgliederung der romanischen Sprachräume" (1936: 1—48). The preliminary 1956 bibliography of von Wartburg's publications, scrupulously compiled by Kurt Baldinger and Alfred Thierbach, lists eleven immediate critical reactions, a figure conceivably not overwhelming were it not for the two circumstances that the piece at issue was not a book, but merely a lengthy journal article and that the ranks of reviewers included, for the most part, top-notch experts in several European countries. 15 It is hardly an exaggeration to remark that not since the publication, on the eve of World War I, of Jakob Jud's trail-blazing "Probleme der altromanischen Wortgeographie" (1914: 1—75) had there appeared, in my narrower discipline, a paper of article length so influential. The following year witnessed a unique alliance of von Wartburg with his colleague and next-door neighbor, the unforgotten Germanist Theodor Frings, a pooling of talents, temperaments, and competences which sought to bring about the definitive clarification of the relations of Frankish to French (and, at a later stage, of Germanic as a whole to Romance). 1 6 By 1939, von Wartburg's thinking had sufficiently matured to have allowed him to undertake a book-length presentation of his

The configuration

of the Romance

language family

77

"theory" (as he insisted on loosely calling the ensemble of his hypotheses, although it patently lacked any solid theoretical foundation), namely Die Entstehung der romanischen Völker. Note that the very title of this venture presupposes the highly hazardous equation of glottal with ethnic links. This time, 23 critical appraisals ensued, some of them published, conventionally enough, in learned journals, others — as if to complicate the lives of later new-comers to the debate — in daily newspapers. 17 By 1941, amid World War II turmoils (and despite, or perhaps on account of, that war's ideological underpinnings), a translation of the book into French, with the misleading equation of folk and speech virtually left intact (Les origines des peuples romans), became available in German-occupied Paris, an event entailing, in turn, six critical responses. There followed two fullsized monographs, slanted in part archeologically, in part toponymically, on Germanic early-medieval settlements in Northern Gaul or France (1944: 17 — 42; 1950 b); one more technical version (Die Ausgliederung der romanischen Sprachräume, 1950 a) of the entire edifice of the author's favorite ideas — note the irksome identity with the title of the 1936 article; 18 plus a revised third edition (1951) of the Entstehung.19 If one disregards subsequent translations, e. g., into Spanish, and treats as mere rear-guard actions, albeit by no means unimportant ones, the author's pleasing short monographs or articles (a) on Greek colonization in Southern Gaul (1952: 1 - 4 8 ; revised, 1956: 6 1 - 1 2 6 ) , (b) on the Romanization of Sardinia (1954: 59 — 72), and (c) on the rise of Rhaeto-Romance (1956: 23 — 44), one feels emboldened to characterize the entire 1934 — 1956 period as a single monumental effort, stretching almost uninterruptedly over more than twenty long years, to grapple with the tricky issue of the inner organization of the Romance language family. Did this protracted endeavor pay off? Thirty years after the conclusion of W. von Wartburg's massive attack on an elusive classificatory problem an exhaustive account, from a position of authority and independence, of all and any versions of his "theory" remains an unfulfilled desideratum; also, the reactions to it have been inventoried but not digested. I cannot aspire to filling this gap here and must confine myself to essentials. 20 To a distinctly higher degree than Menendez Pidal had done in 1926, von Wartburg relied on a random selection of criteria; while those that he chose were undeniably important, he refrained from stating (perhaps even failed to realize), down to the end of the road, whether any appeal to other criteria might or might not confirm his initial surmises. His principal aim was to establish a far-flung Western and a discernibly smaller Eastern wing of the cluster of Romance

78

Yakov Malkiel

languages and dialects. The dividing line that he drew, for the peak of the Middle Ages, and that subsequently became famous, was the La Spezia-Rimini line slicing off Northern Italy from the Apennine Peninsula. (Of these two localities, La Spezia overlooks the Tyrrhenian, while Rimini is on the Adriatic.) The most important pillars supporting this division were (a) the survival, at least for a while, of Latin word-final -s in the West as against its loss in the East (witness Standard Italian and Rumanian) side by side with (b) the voicing (and, subsequently, partial spirantization, even loss) of intervocalic surds in the West (cf. näta > Fr. nee, OSp. nada [nada]) as against the continuance of their original articulation in the East (cf. It. nata). The proponent failed to find it surprising that East and West, on these two scores, behaved very differently. He knew the fact, but failed to capitalize on it, that the core dialects of Sardinia reconciled the preservation of -s with the tendential survival of /p/, /t/, jk/ between vowels, and refused to pay attention to the circumstance, known since 1900, that there were spots in the Pyrenees, especially in the Upper Aragon, where /p/, /t/, /k/ in that position had not been shifted to /b/, /d/, /g/, either.21 On the positive side of the ledger let me credit the scholar here criticized with one major accomplishment: the beautiful demonstration of the fact, never before shown on that scale, that practically all of Northern Italy, en bloc, for centuries must have preserved Lat. s in word-final position. We can now afford to interrupt, for one brief moment, the dissection of W. von Wartburg's thesis and remark that both the core dialects of 20th-century Sardinia and the aforementioned modern-day Pyrenean redoubts laid bare by J. Saro'fhandy and his successors involve characteristic recession zones (Rückzugsgebiete), i.e., residues of areas once distinctly larger. In reference to Late Antiquity and the early Middle Ages it seems legitimate to invoke a single solid, if bipartite, Corso-Sardic dialectal territory, on the understanding that subsequent economic pressures, cultural infiltrations, military conquests, and waves of colonization raying out from Barcelona, Genoa, and Pisa chipped away from that unit major marginal chunks, plus Corsica almost in its entirety. 22 As regards the Iberian peninsula, with special attention to the preservation of intervocalic /p/, jij, /k/, microscopic inquiries into medieval and even post-medieval, including modern, toponymy, conducted by a scholar of the caliber of Manuel Sanchis Guarner (1960: 291 — 342), made it plain that Valencia, Murcia, and Granada, along the Mediterranean coast and farther south, as well as far-off Mallorca, in the Balearic archipelago, for a while remained almost entirely aloof from the oozing of sonantization.

The configuration

of the Romance language family

79

The fashion of voicing the surds in this position must have spread southward from across the Pyrenees, at the very least so far as the Eastern section of Spain is concerned. On this side of the Atlantic Robert A. Hall, Jr., resolutely championed this revisionist opinion on, at least, two occasions: (1974: 133) and (1975: 530-535, at 533); and I hastened to lend it additional support (1983: 192-256, at 199). If one next takes the concluding step by conjoining the evidence of the Corso-Sardic zone with the yield of toponymic explorations in Eastern Spain, including the Balearic islands, a single Tyrrhenian area of protracted stubborn resistance to sonantization begins to emerge, and along with the rise of this new bold vision the edifice of Western Romance, as erected by von Wartburg, ironically, mainly on the strength of this particular criterion, comes down in a crash, leaving no prospect whatsoever of prompt repair. 23 In addition to splitting the Romania through the arbitrarily-drawn La Spezia-Rimini line,24 von Wartburg paused to ask himself about the most plausible causes of, or forces behind, a limited number of sound shifts selected for consideration. Haunted as he was by the equation he himself had devised and relied upon of glottic and ethnic units, he hastened to make admittedly judicious allowances for all sorts of substrate influences, both Indo-European (Keltic, Oscan) and non-Indo-European (Mediterranean, Ibero-Basque); measured generously, but unevenly, superstrate influences (the felicitous term, incidentally, had been minted by himself), by making exaggerated concessions — e.g., re the genesis of rising diphthongs in Romance, especially ie — to Germanic models, in an assumed context of protracted bilingualism, while practically disregarding the impacts of Arabic and Slavic; properly took into account the respective dates of the conquest and subsequent colonisation of the individual provinces under the Republic, the principatus, and the Empire, also the sharply varying socio-educational status of the Roman settlers ('veterans'), plus the rich spectrum of communication lanes, by land and by sea. On the other hand, von Wartburg paid insufficient attention to two discrete factors comparably, if not in fact more, important: (a) The momentum of the relentless strictly internal growth of the sermo Romanus, through sheer interplay of systemic lacunae and resulting asymmetries (and the like) in the phonological structure and, conceivably, in its morphological counterpart as well, along the line of Andre Martinet's skillful dramatization of 1955, but a process fragmentarily known from before; and

80

Yakov

Malkiel

(b) The crucial point that the very centers of political power and cultural hegemony, in the Late Empire, tended to move away from the old capital toward the north: Straight history teaches us that the Imperial residence shifted from Rome to Ravenna, while well-informed practitioners of old-fashioned philology, taken in the narrowest sense of the word, report that the best-executed copies of manuscripts, from a certain cutoff point on, are traceable to the scribal schools of Lyon. Could it be true, then, that the Latinitas Lugdunensis, which could easily have extended westward all the way to Tolosa (present-day Toulouse), in the end became tone-setting for several provinces, including "Hispania citerior", thus overlaying in part of the Peninsula the result of the original wave of Latinization, which had rayed out directly from Rome, in the years of the Republic and the Principatusl Before any serious attempt can be made to replace von Wartburg's brittle theses with something less fragile, a major vacuum in the underlying theoretical foundation should by all means be filled. The principal lacuna refers to the spectrum of imaginable models of diffusion. My understanding is that front-line sociologists are far ahead of historical linguists on that score at present. For interdialectal diffusion of certain features to set in, e. g., it seems at present, as before, highly desirable, but no longer strictly necessary, for an advocate to be in a position to demonstrate the actual release of major forces through such tangible events as uneven spread of political (or ecclesiastic) power; visible economic superiority of one blessed region over its less lucky neighbor; military or naval victories and resulting conquests; gains and losses incurred through dynastic ties; pressures of taxation; the direction of pilgrimage routes leading to sanctuaries; the leveling or integrative power of vigorously conducted military campaigns, and the like. Even in the absence of such "splashy" events as are apt to be specifically reported or, at least, succinctly mentioned in medieval chronicles and/or legal documents, conditions of human climate considerably more elusive or volatile can produce an atmosphere in which a steady flow of sharply profiled features, coming from a favored, prestigious, readily accessible area X can invade with astounding speed and force of impact all or most of the adjoining area Y, with the pockets of resistance (if one detects any), as a rule, corresponding to isolated spots difficult of access and indifferent to commerce and communication. Changes so produced can spread with the speed of meteorological phenomena, such as spells of cold or of humidity. Consequently, from oblique observation of speech via texts of, say, the 9th and the 10th centuries,

The configuration

of the Romance language family

81

however skillfully interpreted, it is by no means certain that even an astute exegete can invariably draw truly cogent conclusions as to the provincial differentiation of Latin in the first few centuries of the Common Era, let alone before — unless a fraction of the total territory under scrutiny was practically cut off from any direct socio-cultural contact with the remainder (witness the exotic Latinity of the Lower Danube). Even in the absence of resounding political events a half-millenium of relative silence can have witnessed tremendous dislodgements, which need not have been of equal strength in phonology, morpho-syntax, and lexis. Without laying claim to being able, at this point, to launch any new, full-fledged hypothesis, I wish to draw your attention, in a low key, to an observation that I have of late made, again and again, in my own lexico-etymological studies: Certain characteristic words and word families solidly entrenched in Portuguese also turn up in Sardic. They may, in addition, be weakly, not to say peripherally, represented in certain constituent dialects of Spanish (Asturo-Leonese, Castilian, Navarro-Aragonese), on the one hand, and, on the other, in Southern Italy; but this is not mandatory. Preponderantly, they are absent from Occitan and French, except for secluded Gascon. As for background, they are, as a rule, of Latin ancestry. Here is a scattering of examples. From faeces, -um 'dregs, lees' there branched off, in the sermo tabernarius, a number of verbs, only two of them, including defaecäre, recorded in literature, not always in an obscene sense. To this small stock we may add *tränsfaecäre 'to decant, pour [wine] from a large container into bottles or glasses', which has left vestiges in Portuguese, Spanish (trasegar), Catalan, and later infiltrated via Genoa into Italian (and from there, into other languages) as trafficare. Next, let us examine *offaecäre 'to obstruct the gullet with dregs', strongly represented in Portuguese (ofegar 'to pant heavily, as from a spell of coughing'), and Catalan ofegar 'to stifle' beside ofec 'asthma'. Sardic boasts a whole cluster of -fegare verbs: af- and in-fegare 'to get intoxicated', izfegare 'to vomit'. 25 For 'opening in the wall' (serving the purpose of ventilation) HispanoRomance uses either a derivative from ventus 'wind', witness Sp. ventana, reminiscent of E. window, or a descendant of Lat. fenestra, cf. OFr. fenestre (mod. fenetre), It. finestra, G. Fenster, whose relevant counterparts include dialectal Sp. hiniestra and Ptg. fresta originally feestra 'aperture in a wall'. The standard word in Portuguese, however, is janela, lit. 'small door', cf. Sard, yanna 'door', from Lat. iänua 'threshold', in the Logudoro dialect; farther south, in the Campidanese flatland, one

82

Yakov

Malkiel

comes across the var. enna, plausibly traceable to the assumed Latin byform ienua. As a geographically interjacent form I can cite the toponym Jänovas, observed in the Aragonese province of Huesca. 26 The word for 'hungry' in Portuguese is uniquely difficult: Given the existence offamulentus in Latin plus hambriento in Spanish; given further the survival or new coinage of scores of local adjectives in -ento, including sedento 'thirsty' (from sede, a reflex of sitifmj), one would expect *famento, but instead encounters faminto from time immemorial. The crux begins to yield to analysis, once Sard, /famiöu/ from Logudoro comes to mind. The rise, in Sardo-Latin, of *fam-itus, reminiscent of recorded crmitus, pen-ltus (from crlnis, penis), is a sensible enough assumption, and a blend of *fam-ido with synonymous *fam-ento might indeed account for erratic and otherwise inexplicable faminto. This analysis, then, furnishes another Luso-Sardic isogloss. Slightly different are the problems raised by the verb Ptg.-Gal.-Ast. fecharjpechar/pesllar 'to lock, shut', a cluster of visibly related forms which invites joint study with the vicissitudes of (a) choir, lloir (the archaic outgrowths of clauderej-clüdere 'to close', cf. obsolescent Fr. clore, It. chiudere); (b) Fr. fermer, the echo of firmäre 'to make firm, make safe by closing firmly'; and (c) seräre 'to lock', contaminated by ferru 'iron' and transmitted in a phonosymbolic key: /s/ > /Θ/ (Sp. cerrar), conceivably so as to avoid irksome homonymy with serrar 'to saw'. 27 As Gunnar Tilander has demonstrated in an article long and thorough, if not in every respect wholly persuasive (1949 — 1950: 1 — 13), the pecharjfechar verb, which may go back to a blend of pessulu 'latch' either with ferru 'iron' or with fistula 'reed-pipe' or else with both ('creaking iron lock' > 'musical lock', said in an ironic vein), has counterparts in two other Romance speech areas, Sardinia and Gascony. The fact that in individual word histories a third territory will occasionally emerge as a, strictly, supererogatory partner detracts not in the least from the significance of the tell-tale coincidence between the Western fringe of the Iberian peninsula and Sardinia's core area. These four word histories should not be brushed off as mere anecdotes; moreover, equally weighty phonological parallels are not lacking either, though I would be abusing your patience by presenting them here in comparable detail. Everywhere one recognizes patches of what Heinrich Lausberg, not infelicitously, has called "Altromania" (1963: 27 — 76), torn apart by the intrusion of its aggressive polar opposite. Over against it, by definition, must stand something worthy to rank as "Neuromania", 28 and there actually is no better place for it, on the map of Late Antiquity,

The configuration

of the Romance language family

83

than the region, in Southern France, surrounding and including Lyon and Toulouse. I do not wish listeners or future readers to come away with the impression that, in my private estimate, the division championed by W. von Wartburg (West vs. East) is necessarily wrong, while the partial division here advocated (mid-North vs. South) is absolutely right. This is what my message decidedly was not meant to be. Rather, the idea that I have tried hard to drive home was that one should reckon with the ever-present possibility, where successive patterns of dialectalization are involved, that one territorial schema of division may have prevailed in, say, the third and fourth centuries and one almost entirely different may have emerged five or six centuries later, so long as the dialects at issue continued to be closely connected with one another. And even where an irremediable territorial separation did occur (e.g., through the wedgingin of the genetically unrelated language of some intruder, as applies to the role that Hungarian played in the life history of Rumanian), there remains the marginal — or perhaps not so marginal — possibility of — speaking with Meillet — a "developpement independant parallele". Notes 1. In an earlier paper I missed the eventual retraction, by Vater, of this bizarre classification (see 1973: 589 — 593). Five years later, Jens Lüdtke set the record straight, as I admitted in my review of his book (1979 [ - 1 9 8 0 ] : 2 1 7 - 2 2 6 ) . 2. For details see Malkiel ( 1 9 8 3 - 1 9 8 4 : 1 - 2 4 ) . 3. On Ascoli's involvement, in general, with issues in Romance linguistics and underlying points of theory see I o r d a n - O r r (1937; 1970: 1 0 - 1 2 , 2 7 - 3 0 , etc.; see Index: 588a). The two studies here singled out for mention are: (1873: 1 — 556) and (1878: 61 — 120). I refrain from taking into account in this retrospective bird's-eye view Hugo Schuchardt's celebrated pamphlet (1900), because it made its appearance, anachronistically, thirty years after the underlying probationary lecture had been delivered at Leipzig University. 4. I am referring here, principally, to Meyer-Liibke's position taken, passim, in his standard-setting comparative grammar (1890 — 1902) as well as throughout the three editions of his equally influential introduction into Romance linguistics (1901; 1909; 1920: § 18). 5. A dimension of pathos is added to Meyer-Liibke's 1925 book by the fact, inferrable from his dedication, that he wrote most of it as a house guest of Catalan friends and thus as a temporary resident of Catalonia. 6. While Griera's and Alonso's reactions overshadowed all others in sheer length and violence, there appeared a few that, in retrospect, indeed appear more balanced: those by Oscar Bloch, Josef Brüch, A. Carnoy, W[illiam] J. E[ntwistle], Pierre Fouche, F. Krüger, Antoine Meillet, K. Sneyders de Vogel, Ο. J. Tallgren [-Tuulio], W. von Wartburg, and A. Zauner. Except for Fouche's long-delayed and less than authoritative

84

7.

8.

9.

10.

11.

Yakov

Malkiel

piece, these assessments made their appearance in rapid succession, between 1925 and 1927; for details see F. Ritter (1930: 210). For our purpose W. von Wartburg's reaction is doubly and triply interesting (1928: 157 — 161): After commending the author for the wealth of criteria applied and for the novel arrangement of lexical data, and without disagreeing with Meyer-Lübke's verdict to the effect that phonology tends to bracket Catalan with Provengal, while in matters of inflection Catalan displays stronger affinity with Spanish, the reviewer concluded, almost reverentially: "Ich möchte nicht von dem Buche Abschied nehmen, ohne nochmals auf seine hohe methodische, nach gewissen Richtungen bahnbrechende Bedeutung aufmerksam zu machen". In light of subsequent events, the key statement, for today's readers, is perhaps the following incidental remark: "Wer, wie ich, geneigt ist, der Morphologie einer Sprache für ihre Charakteristik und Verwandtschaftsbeziehungen eine weit größere Bedeutung zuzumessen als etwa der Lautlehre und der Syntax..." (von Wartburg 1928: 158). Eight years later, von Wartburg reversed himself completely, sweeping morphology under the rug — without bothering to offer the mildest excuse for this strange self-contradiction. The radical position Griera took on that occasion (1925: 198 — 254) involved nothing really new: He had voiced similar views in his own journal, tacitly committed to the championship of Catalan cultural autonomy (1922: 34 — 53). Amazing, at this distance, is the ineptitude displayed by the editor of the ZRPh, Alfons Hilka (or was it a case of bias?), who published Griera's exceedingly long paper with unseemly haste; later relegated Meyer-Lübke's substantial counterview to the less conspicuous section of book reviews; and in the end published W. von Wartburg's appraisal — a genuine book review — without any editorial reference to the preceding exchange of opinions. A clear line of demarcation must be drawn between the linguists' tag "Afro-Latin" and the classical philologists' label "African Latinity". The latter refers mainly to a rather pompous, rhetorical style exhibited by several Latin writers who happened to hail from Northern Africa. "Afro-Latin", in contrast, denotes the fragmentarily reconstructed provincial variety of colloquial Latin, based on our partial knowledge of a handful of individual word histories and carried out with help from Berber evidence. Starting with the year 1941, possibly by way of response to the new political climate, Griera's interest turned abruptly from the Castilian vs. Catalan polarization and from dreams about African Latinity to Pyrenean studies, Basque and substratum problems, and the onomasticon. The transition is marked by his post-Civil War papers (1941 a) and (1941 b); the peak of the new trend is exemplified by (1957). Severe frictions between Barcelona's Institut d'Estudis Catalans and Madrid's Centro de Estudios Historicos became visible as early as Menendez Pidal's devastating review (1916: 73 — 88) of Griera's earliest major monograph, on the Catalan-Aragonese frontier. Part 1 of Alonso's composite review article ("La subagrupacion...") concentrates chiefly on Meyer-Lübke's book and takes into account only toward the end Griera's review article; Part 2 focuses on Griera's piece, but also takes cognizance, albeit superficially, of Meyer-Lübke's reply the year after (1926: 1—38, 225 — 261). The inclusion of both parts, and of the item contributed to the miscellany in honor of P. Fabra (1943: 81 — 101), in a collection of Alonso's shorter articles which he could still supervise shortly before his death (1951: 11 — 127) contains no changes, but the Preface (1951: 7 — 8) shows how he viewed the controversy in retrospect. His 1927 transfer to Buenos Aires cut short a far more ambitious project along this line. The Catalan expert in Madrid on whose expertise Alonso had leaned was Samuel Gili Gaya. The change in

The configuration

12.

13.

14.

15.

16.

17.

18.

of the Romance language family

85

climate, after the Civil War, shows in that J. Coromines' 1954 book, reprinted in 1972, produced not a ripple in Romance quarters the world over. It is usually held that Menendez Pidal's 1910 speech of welcome extended to Francisco Codera before the Academy of History embodies the nucleus of his own thinking about the dialectal structure of the Peninsula; perhaps one might add to it his, virtually simultaneous, substantial appraisal (1910b: 119 — 130) of Erik S t a a f f s L'ancien dialecte leonais. I hardly need to repeat here what I have expounded on other occasions about the history of Vol. 1 — the only one ever published — of the Origenes del espanol (1970: 371 - 4 1 1 ; 1984: 3 2 5 - 3 4 7 ; 1985: 5 - 6 8 ) . The assumption of an Oscan substratum in the Ebro Valley and in other sections of the Aragonese area goes back to the Origenes, starting with its 1926 version (cf. 1950: 296 — 306, with numerous additions), and later spilled over, not always advantageously, onto the author's manual of historical grammar (1941: § 2). For clues to the reactions by W. D. Elcock and others see A. Kuhn (1951: 340, 393, 395). More judiciously phrased references to parallel situations in select sections of Northern Spain and in South Italy, without heavy insistence on the Oscan substrate as, necessarily, the common denominator, are found in some later probings and gropings by a mellowed Menendez Pidal, e.g., in (1954a: 7 — 44), where he cautiously stated: " L o mäs verosimil es explicar las dos areas asturianas por una primitiva colonization romana procedente del sur de Italia"; see also (1954 b: 161 - 2 1 6 ) . Of the 1936 version of "Die Ausgliederung" there appeared three extended reviews by D. Alonso, F. Schürr, and E. Richter, running to 13, 10, and 5 printed pages, respectively; additionally, seven signed appraisals of standard length (G. Gougenheim, J. B. Hofmann, I. Iordan, S. Pu§cariu, G. Rohlfs, B. A. Terracini, M.Valkhoff), plus an unsigned assessment in a Zurich daily — all in all, an unusually strong repercussion. The long-lasting alliance between Th. Frings and W. von Wartburg gave the impression of being directed against E. Gamillscheg, who competed with both, and with J. Brüch for good measure, in centering attention about Romano-Germanic contacts and conflicts. The 1939 princeps of Die Entstehung marks the first peak of W. von Wartburg's impact, on the eve of World War II. Aside from once more stimulating the imagination of several Romance scholars (W. Giese, G. Rohlfs, L. Spitzer, M.Valkhoff, M. L. Wagner; D. Alonso dashed off his second 13-page review), it provoked the curiosity of Classicists, Germanists, and straight historians. Appraisals appeared in numerous, topically diversified learned journals and in five newspapers (in one of which K. Vossler acted as the invited critic); all in all there ensued 23 reactions extending over the period 1939 — 1943, plus one rejoinder by the author. The 1950 version of Die Ausgliederung... — being this time a book — provoked far more numerous responses than the first, including a 17-page review article by T. Bolelli. Almost all reviewers were Romanists, but the Norwegian generalist Alf Sommerfell, as well as Andre Martinet, joined the chorus. Among better-known critics not yet mentioned let me cite G . Bonfante, A. Henry, U.T. Holmes, L. Kukenheim, H. Lausberg, Ε. Legros, L. Mourin, J. Orr, V. Pisani, R. L. Politzer, Ε. Pulgram, S. Ullmann, V. Väänänen, R.-L. Wagner, and P. Zumthor. Two Swiss newspapers chimed in. The stronger participation of American experts may be safely ascribed to W. von Wartburg's part-time teaching experiment at Chicago, in the 1930s. A greater success was inconceivable.

86

Yakov

Malkiel

19. With the revised edition of Die Entstehung... a sobering attitude on the part of reviewing media set in: Only five critical reactions have come to my attention, all five fairly short (by M. Geizer, E. Legros, L. Mourin, J. M. Piel, V. Pisani). 20. I have previously examined W. von Wartburg's "theory" (or better, thesis), from different angles, in at least three contexts: (1972: 8 3 5 - 9 2 5 ; at 8 6 3 - 8 6 9 , 9 2 3 - 9 2 4 ) ; (1977: 2 7 - 4 7 ) ; and (1978: 4 6 7 - 5 0 0 ) . The last item is merely a distillation of a booklength manuscript temporarily left unpublished. 21. The discovery of the crucial evidence is credited to J. Saro'ihandy (1898: 85 — 94; 1901: 1 0 6 - 1 1 8 ) . The key studies are due to Alwin Kuhn (1935: 1 - 3 1 2 ) and William D. Elcock (1939); the latter also surveyed the corresponding Upper Bearnais dialects, to the north of the Pyrenean divide. 22. For an outline of the political and ecclesiastic history of the island see M. L. Wagner (1951: 7 - 3 5 ) . 23. One peculiar facet of von Wartburg's approach, starting with the pioneering 1936 article, was his admitted total reliance on Menendez Pidal's findings as presented in the latter's 1926 treatise. Conversely, the Spanish master, in his Preface to the revised third edition of that work, confessed that he had plunged into the preparation of his magnum opus without the guidance of any preliminary blueprint. The conclusions to be drawn from these two candid admissions are self-evident. 24. The nadir of his entire analysis was, one feels in retrospect, his inability to pin down the exact position of Sardic, which was described as having joined now the West, now the East; a dance routine with change of partners? 25. I have twice busied myself with the verbs that rayed out from faec- in written and colloquial Latin; for the fuller of these two treatments of the problem see (1985 b: 305 — 338). On Ptg. ofegar, its derivatives and putative cognates see my note of slightly later vintage (1986: 177 — 184). This entire subfamily has, until the recent past, been usually traced to offöcäre 'to stifle' (akin to fauces), even though the latter reappeared in Old Spanish and Old Portuguese as afogar (mod. Sp. ahogar). 26. The identification goes back to G. Rohlfs (1985: 40); as a matter of fact, the article in question appeared originally as early as (1951: 229 — 265) and was reprinted for the first time in 1956. 27. My preoccupation with this verb, the underlying noun, and fistula as a carrier of lexical contamination extends over 35 years, a stretch of time that has entailed a radical change of perspective: (1952: 299 — 338) vs. (forthcoming b). 28. True, while Lausberg is willing to operate with the concepts of "Altromania" and "Neuromania" in discussing, e. g., the fortunes of ϊ and ü (to which W. von Wartburg paid scant, if any, attention), he adopts elsewhere von Wartburg's basic concepts of Western vs. Eastern Romance and gratuitously projects them onto the plateau of the earliest period, dimly visible only through reconstruction.

References Alonso, Amado 1926 "La subagrupacion romänica del Catalan: 1. Los metodos; 2. La geografia lexica", Revista de filologia espafiola 13: 1 —38, 225 — 61. 1943 "Partition de las lenguas romänicas de Occidente", in: Miscellänia Fabra: Recull de treballs de lingüistica catalana i romänica dedicats a Pompeu Fabra... (Buenos Aires: Coni), 81 — 101.

The configuration 1951

of the Romance language family

87

Estudios lingüisticos: temas espanoles (Biblioteca romänica hispanica 2:2) (Madrid: Gredos) [2nd printing, 1954], Ascoli, Graziadio Isaia 1873 "Saggi ladini", Archivio glottologico italiano 1: 1 —556; folding map. 1878 "Schizzi franco-provenzali" Archivio glottologico italiano 3: 61—120. [Left unfinished, but cf. ibid. 2 [1876]: 3 8 5 - 3 9 5 . ] Baldinger, Kurt —Alfred Thierbach 1956 "Bibliographie der Publikationen von W. von Wartburg", in: W. v. Wartburg, Von Sprache und Mensch (Bern: Francke), 234 — 279. Coromines, Joan [1954] El que s'ha de saber de la llengua catalana (Palma de Mallorca: Moll) [2nd edition ( = Biblioteca Raixa 1), 1972]. Elcock, William D. 1939 De quelques affinitέs phonetiques entre l'aragonais et le bearnais (Paris: E. Droz). Griera, Antoni(o) 1922 " A f r o - r o m ä n i c ο ibero-romänic? Estudi sobre eis corrents histöricoculturals que han condicionat la formacio de les llengües romäniques en la peninsula iberica", Bulleti de dialectologia catalana 10: 34 — 53. 1925 "Castellä — catalä — proven^al; observacions sobre el llibre: W. MeyerLübke...", Zeitschrift für romanische Philologie 45: 198 — 254. 1941 a "Los origenes del espanol", Boletin de dialectologia espanola [= catalana] 25: 52-61. 1941 b "El origen de la lengua vasca", Boletin de dialectologia espanola [= catalana] 25: 6 4 - 7 8 . 1957 Nombres personales del Pireneo catalän de origen prerromano. Actes du 2C Congres International d'Etudes Pyreneennes 8: 7. Gröber, Gustav 1884 — 1888 "Vulgärlateinische Substrate romanischer Wörter", Archiv für lateinische Lexicographie und Grammatik 1 —6. Hall, Robert Α., Jr. 1974 External history of the Romance languages ( = Comparative Romance grammar 1) (New York: American Elsevier). 1975 "La non-lenizione nella Romania occidentale" Romance Philology 23: 530-535. lordan, Iorgu —John Orr 1937 An introduction to Romance linguistics: Its schools and scholars (Oxford: Basil Blackwell) [revised second edition, with a supplement by Rebecca Posner: 1970], Jud,Jakob 1914 [ — 1917] "Probleme der altromanischen Wortgeographie", Zeitschrift für romanische Philologie 38: 1 —75; 5 maps. Kuhn, Alwin 1935 "Der hocharagonesische Dialekt", Revue de linguistique romane 11: 1—312. 1951 Romanische Philologie 1: Die romanischen Sprachen (Wissenschaftliche Forschungsberichte 8) (Bern: Francke). Lausberg, Heinrich 1963 Die romanische Sprachwissenschaft 1: Einleitung und Vokalismus2 [revised] (Sammlung Göschen 128/128 a) (Berlin: de Gruyter).

88

Yakov Malkiel

Lüdtke, Jens 1978 Die romanischen Sprachen im 'Mithridates' von Adelung und Vater. Studie und Text (Lingua et Traditio 4) (Tübingen: Narr). Malkiel, Yakov 1952 "Studies in Hispano-Latin homonymics: pessulus, päctus, pectus, despectus, suspectus, fistula in Ibero-Romance", Language 28: 299 — 338. 1970 "... Era omme esencial' [Ramon Menendez Pidal] (1869-1968)" Romance Philology 23: 371-411. 1972 "Comparative Romance linguistics", in: Thomas E. Sebeok (ed.), Current trends in linguistics 9 (Linguistics in Western Europe) (The Hague: Mouton), 835-925. 1973 "Adelung-Vater's pioneering survey of Romance languages and dialects (1809)", Studii §i cercetäri lingvistice 15: 589-593. 1977 "Criteres pour l'etude de la fragmentation des langues romanes", in: Atti del XIV Congresso internazionale di linguistica e filologia romanza (Napoli, 15 — 20 aprile 1974) 1 (Napoli: G. Macchiaroli/Amsterdam: J. Benjamins), 27—47. 1978 "The classification of Romance languages", Romance Philology 31: 467 — 500. 1979 [1980] Review of Jens Lüdtke. Kratylos 24: 117-126. 1983 "Alternatives to the classic dichotomy family tree/wave theory? The Romance evidence", in: Gerald F. Carr (eds.), Language change (Bloomington: Indiana University Press), 192-256. 1983 — 1984 "The place assigned to Rumanian by 19th-century comparativists", Southeastern Europe 10.1: 1—24. 1984 "Ramon Menendez Pidal as etymologist", Historiographia linguistica 11: 325-347. 1985 a "La ultima fase (1939 — 1969) de la labor lingüistica de Ramon Menendez Pidal", Anuario de letras 23: 5 - 6 8 . 1985 b "La etimologia de espafiol tras-[h]egar 'transvasar', italiano trafficare 'comerciar': un nuevo balance", Medioevo romanzo 10: 305 — 338. 1986 "The etymology of Portuguese ofegar", Romance Notes 26: 177 — 184. forthcoming a "El sardo como guia por los matorrales lexicos del gallegoportugues; la derivacion de faminto 'hambriento'", to appear in Medioevo romanzo. forthcoming b "Regular sound development, phonosymbolic orchestration, disambiguation of homonyms", to appear in Transactions of the Berkeley Sound Symbolism Conference (January 1986). Martinet, Andre 1955 Economie des changements phonetiques. Traite de phonologie diachronique (Bibliotheca Romanica 1:10) (Berne: Francke). Menendez Pidal, Ramon 1910 a "Discurso de contestation a don Francisco Codera, leido ante la Real Academia de la Historia en la reception publica de..." (Madrid) [no publisher]. 1910 b Review of Erik Staaff, Etudes sur l'ancien dialecte leonais d'apres des chartes du XIII siecle, in: Revue de dialectologie romane 2: 219 — 230. 1916 Review of A. Griera i Gaja, La frontera catalano-aragonesa, estudi geogräficolingüistic. In: Revista de filologia espanola 3: 73 — 88. 1926 Origenes del espafiol. Estado lingüistico de la peninsula iberica hasta el siglo XI (Madrid: Centro de estudios historicos) [revised second edition, 1929; revised third edition ( = Obras completas 8), 1950], 1941 Manual de gramätica histörica espanola6 [revised] (Madrid: Espasa-Calpe).

The configuration of the Romance language family 1954 a

89

"Pasiegos y vaqueiros: dos cuestiones de geografia lingiiistica", Archivum (Oviedo) 4: 7 - 4 4 . 1954 b "A proposito de -II- y -I- latinas; colonization suditälica en Espana", Boletin de la Real Academia Espanola 34: 161 —216. Meyer-Lübke, Wilhelm 1890 — 1902 Grammatik der romanischen Sprachen, 1—4 (Leipzig: Fues's Verlag [R. Reisland]). 1901 Einführung in das Studium der romanischen Sprachwissenschaft (Heidelberg: Winter) [revised second edition, 1909; revised third edition, 1920]. 1925 Das Katalanische; seine Stellung zum Spanischen und Provenzalischen sprachwissenschaftlich und historisch dargestellt (Sammlung romanischer Elementarund Handbücher 5.7) (Heidelberg: Winter). 1926 —1927 "Afroromanisch und Iberoromanisch", Zeitschrift für romanische Philologie 46: 116-128. Ritter, F. 1930 Bibliographie 1925 (Supplementheft 45, Zeitschrift für romanische Philologie) (Halle: Niemeyer). Rohlfs, Gerhard 1951 "Aspectos de toponimia espanola", Boletim de fdologia (Lisböa) 12: 229 — 265. 1985 Antroponimia e toponomastica nelle lingue neolatine; aspetti e problemi (Tübingen: Narr). Sanchis Guarner, Manuel 1960 "El mozarabe peninsular", in: Enciclopedia lingüistica hispänica (dir. M. Alvar et al.) 1: Antecedentes, Onomästica (Madrid: Consejo Superior de Investigations Cientificas), 291—342. Saroihandy, J. 1898 "Mission de M.S. en Espagne", Annuaire de l'Ecole pratique des Hautes Etudes 8 5 - 9 4 [concluded in 1901: 106-118], Schuchardt, Hugo 1900 Über die Klassifikation der romanischen Mundarten. Probevorlesung gehalten zu Leipzig am 30. April 1870 (Graz: [no publisher]). Tilander, Gunnar 1949 — 50 "L'etymologie du portugais fecho, fechar elucidee par la construction des serrures primitives", Studia Neophilologica 22: 1 — 13. Wagner, Max Leopold [1951] La lingua sarda: storia; spirito e forma (Bibliotheca Romanica 1.3) (Bern: Francke). 1957 "Das Sardische im "Diccionario critico [etimologico] de la lengua castellana" von J. Corominas", Romanische Forschungen 69: 241 —272. Wartburg, Waither von 1928 Review of Meyer-Lübke, Das Katalanische..., in: Zeitschrift für romanische Philologie 48: 157-161. 1930 "Der Einfluß der germanischen Sprachen auf den französischen Wortschatz", Archiv für Kulturgeschichte 20: 309 — 325. 1932 "Die Ursache des Auseinanderfallens der Galloromania in zwei Sprachgebiete: Französisch und Provenzalisch". Summary in Forschungen und Fortschritte 8.21: 268-269.

90

Yakov 1934

Malkiel

"Die Entstehung der Sprachgrenzen im Innern der Romania", Beiträge zur Geschichte der deutschen Sprache und Literatur 58: 209 — 227. 1936 "Die Ausgliederung der romanischen Sprachräume", Zeitschrift für romanische Philologie 56: 1—48; 7 maps. 1939 Die Entstehung der romanischen Völker (Halle: Niemeyer). 1941 Les origines des peuples romans. Trs.: G. Cuenot de Maupassant (Paris: Presses Universitaires de France). 1944 " U m f a n g und Bedeutung der germanischen Siedlung in Nordgallien im 5. und 6. Jahrhundert: Sprachliches, Ortsnamen, Ausgrabungen", Schweizer Beitrüge zur allgemeinen Geschichte 2: 17—42. 1950a Die Ausgliederung der romanischen Sprachräume (Bibliotheca Romanica 1.8) (Bern: Francke). 1950 b Umfang und Bedeutung der germanischen Siedlung in Nordgallien im 5. und 6. Jahrhundert im Spiegel der Sprache und der Ortsnamen (Berlin: Deutsche Akademie der Wissenschaften) (Vorträge und Schriften, 36). 1951 Die Entstehung der romanischen Völker2 [revised] (Tübingen: Niemeyer). 1952 "Die griechische Kolonisation in Südgallien und ihre sprachlichen Zeugen im Westromanischen", Zeitschrift für romanische Philologie 68: 1—48 [separatum: Tübingen: Niemeyer, 1953; revised version: (1956): 61 — 126]). 1954 "Zum Problem der Romanisierung Sardiniens", Zeitschrift für romanische Philologie 70: 5 9 - 7 2 . 1956 Von Sprache und Mensch. Gesammelte Aufsätze (Bern: Francke). Wartburg, Walter von —Theodor Frings 1937 "Französisch und Fränkisch", Zeitschrift für romanische Philologie 57: 1 9 3 - 2 1 0 ; 58 (1938): 5 4 2 - 5 4 9 ; 67 (1951): 1 6 7 - 1 7 3 ; 70 (1954): 8 6 - 9 7 .

Typology and deep genetic relations in North America* Marianne

Mithun

Throughout most of the history of North American Indian linguistics, a tradition spanning over three centuries, scholars have devoted intensive efforts to the classification of the hundreds of languages indigenous to the continent. For the most part, this work has proceeded from the establishment of regular sound correspondences among closely related languages to the investigation of those among more distantly related groups. Once phonological correspondences are understood, grammatical systems are compared. Steady progress has been made in the grouping, subgrouping, and reconstruction of most of these languages. By 1891, Major John Wesley Powell of the Smithsonian could propose fifty-some language families north of Mexico, and these are still recognized, with a few modifications, today. Many scholars engaged in this comparative work have been impressed by the structural resemblances among these languages. An Iroquoianist encountering Algonquian languages for the first time feels a sense of instant recognition, although there are no discernible cognates between the two families. Even linguists not specializing in American Indian languages often have notions about their general character. Such structural resemblances have been noticed since the early nineteenth century. Peter Stephen Duponceau, secretary to the American Philosophical Society, had compared vocabularies of a number of languages, and divided the eastern ones into four separate families: Eskimo, Algonquian, Iroquoian, and 'Floridian' (actually a mixture of southeastern languages). Despite their profound lexical differences, on which he based his classification, he was impressed with their grammatical resemblances. In his essay Memoire sur le systeme grammatical des langues de

* I am grateful to the following speakers who have graciously shared their expertise on their native languages: Hasan Basri, of Palu, Indonesia on Selayarese, Margaret Edwards and Sonny Edwards, of Ahkwesahsne, Quebec, on Mohawk, and Frances Jack, of Hopland, California, on Central Pomo. I also appreciate comments provided by Yakov Malkiel on an earlier version of the paper.

92

Marianne

Mithun

quelques nations indiennes de l'Amerique du nord, he cited their common polysynthetic nature, the high numbers of morphemes contained in single words. Le caractere general des langues americaines consiste en ce qu'elles reunissent un grand nombre d'idees sous la forme d'un seul mot; c'est ce qui leur a fait donner par les philologues americains le nom de langues polysynthetiques. Ce nom leur convient ä toutes (au moins ä celles que nous connaissons), depuis le Groenland jusqu'au Chili, sans qu'il nous ait ete possible d'y decouvrir une seule exception, de sorte que nous nous croyons en droit de presumer qu'il n'en existe point. (Duponceau 1831: 89)

The pervasiveness of polysynthesis in America was the source of much discussion throughout the remainder of the nineteenth century and into the twentieth. Philologists, engaged in establishing relationships on the basis of demonstrable sound correspondences, began to wonder whether such structural unity might be indicative of deeper genetic ties. It was proposed on several occasions that all American languages "descended from a single parent language, for, whatever their differences of material, there is a single type or plan upon which their forms are developed and their constructions made" (Whitney 1889: 348). (A history of nineteenthcentury views on the relative value of lexical and grammatical correspondences in genetic classification can be found in Haas 1969.) Daniel Brinton, a contemporary of Powell, also noted the pervasiveness of polysynthesis, and further described the characteristics of polysynthetic languages as follows: The psychic identity of the Americans is well illustrated in their languages. There are indeed indefinite discrepancies in their lexicography and in their surface morphology; but in their logical substructure, in what Wilhelm von Humboldt called the "inner form", they are strikingly alike. The points in which this is especially apparent are in the development of pronominal forms, in the abundance of generic particles, in the overweening preference for concepts of action (verbs), rather than concepts of existence (nouns), and in the consequent subordination of the latter to the former in the proposition. This last mentioned trait is the source of that characteristic called incorporation ... I have yet to find one of which we possess ample means of analysis, in which it does not appear in one or another of its forms, thus revealing the same linguistic impulse. (Brinton 1891, cited in Haas 1969)

The criterion of structural similarity was formally applied to a classification of North American languages by Edward Sapir in 1929. Sapir was both an experienced field worker and a rigorous historical linguist. He had worked extensively with speakers of a variety of languages, and done

Typology and deep genetic relations

in North America

93

careful reconstruction based on the comparative method in several families, in particular, Athapaskan and Uto-Aztecan. It was clear to him, however, that due to the limited documentation of many American languages, and the tremendous time depths separating them, the traditional comparative method could be of little use in establishing deeper ties. After pointing out the phonological and morphological diversity to be found within North America, he proposed an overarching scheme which conflated Powell's fifty-some families into six superstocks: Eskimo-Aleut, Algonkin-Wakashan, Nadene, Penutian, Hokan-Siouan, and Aztec-Tanoan. He based these groupings purely on structural characteristics, but he was careful to specify that the results were meant to be only suggestive. It is impossible to say at present what is the irreducible number of linguistic stocks that should be recognized for America north of Mexico, as scientific comparative work on these difficult languages is still in its infancy. (Sapir 1929)

A number of Sapir's new combinations have since been confirmed by further documentation of the languages and comparative work. Others remain hypotheses, due to the great time depth involved, sufficient to obscure almost all cognates, and the antiquity of language contact, particularly in such areas as the Northwest and California. In some cases, resemblances across superstock boundaries now appear as strong as those within them. Sapir was sufficiently careful however, in specifying the degree of confidence he was willing to accord his various proposals, that this work has proven useful in pointing the way to areas inviting special investigation. The reliability of structural resemblances as indicators of deeper genetic relationship has not yet been clearly established, however. Among the grammatical features most often cited as typical of North American languages are the following: (i) (ii) (iii) (iv)

polysynthesis, pronominal forms, incorporation, and the preference for concepts of action (verbs) rather than concepts of existence (nouns) and the consequent subordination of the latter to the former in the proposition.

94

Marianne

Mithun

1. Polysynthesis The term polysynthesis refers technically to a high number of morphemes per word. The extreme polysynthesis characteristic of many North American Indian languages is familiar to most of those who have been exposed to them. Words like those below, containing large numbers of morphemes, are not at all uncommon in North America. In fact, they can be found in languages in each of Sapir's superstocks. (Hyphens indicate morpheme boundaries; periods within interlinear translations do not. The English gloss 'provide.with' in the first example thus corresponds to a single Greenlandic morpheme.) (1)

Greenlandic Eskimo (Eskimo-Aleut): aliikkus-irsu-i-llammas-sua-a-nira-ssa-gukku entertainment-provide.with-HALF.TRANSITIVIZER-one. good.at-big-be-say.that-FUTURE-1. SINGULAR. 3.SINGULAR. CONDITIONAL 'if I should say that he is a good entertainer' (Fortescue 1984: 203)

(2)

Cree (Algonquian, Algonkin-Wakashan): klsk-ik w-et-ah w-e-w sever-neck-with.handle-by.medium-3.SINGULAR-4.SINGULAR 'he severs his neck with an axe' (Wolfart 1973: 73)

(3)

Hupa (Athapaskan, Nadene): ya-na-ki-s-di-mmill-ei to.pieces-again-INDEFINITE.OBJECT-PROGRESSIVE-3d. MODAL-throw.several.things-at.last 'she smashed it' (Goddard 1911: 112)

(4)

Takelma (Penutian) bem-wa-'Ί-ίΙοχο '-xi-,:n stick-together-hand-gather-in-1 .SINGULAR Ί gather sticks together' (Sapir 1922: 65)

(5)

Mohawk (Iroquoian, Hokan-Siouan): i-ons-a-ha-hnek-ontsien-ht-e? TRANSLOCATIVE-REPETITIVE-AORIST-MASCULINE. SINGULAR. AGENT-liquid-fetch.water-CAUSATIVEPUNCTUAL 'he scooped up the water again' (Margaret Edwards: p. c.)

Typology and deep genetic relations in North America

(6)

95

Tetelcingo Nahuatl (Uto-Aztecan, Aztec-Tanoan): ti-tlahko-tla-kwj-s-h

l.PLURAL-half-UNSPECIFIED.OBJECT-eat-FUTUREPLURAL 'we will eat at noon' (Tuggy 1979: 48). Not all North American languages exhibit nearly this degree of synthesis, however. While there are no extreme isolating languages, like some of those found in Asia, there are numerous languages roughly comparable in degree of synthesis to many Indo-European counterparts. The sentence below opens a tale from Wappo, a California language. (7)

Wappo (Isolate, Hokan): lokh-noma-h

nom-khP,

khön \

goose-camp-LOCATIVE live-FACTIVE once 'Long ago at Goose Camp, chica:-met'-a:yi,

me-^ew

bear-woman-SUBJECT POSSESSIVE-husband k"esu:-has-k'a.

deer-old-with 'Bear Woman lived with her husband Deer.' (Sawyer —Somersal 1977: 106) At the same time, long words are not at all restricted to the Americas. Polysynthetic languages can be found in most areas of the world. Note the morphological structure of the verbs below from Africa, and India, respectively. (8)

Turkana (Nilo-Saharan, Nigeria): jti-k-i-

inis-ään-a-kl

NEGATIVE-TRANSITIVE-2.SINGULAR-boast-HABITUAL-DATIVE 'do not boast at me (all the time)' (Dimmendaal 1983: 449) (9)

Ngandi (Non-Pama-Nyungan, Australia): baru-ja-maria-gulk-d-i

3.PLURAL/3. MASCULINE.SINGULAR-now-neck-cutAUGMENT-PAST.PUNCTUAL 'they hanged him' (Heath 1978: 277) (10)

Sora (South Munda, India): ji-lo:-9jer/-t-am

stick-mud-leg-FUTURE-2.SINGULAR 'mud will stick to your leg' (Ramamurti 1931)

96

Marianne

(11)

Mithun

Selayarese (Austronesian, Indonesia): la-mu-paka-ta-s-suyke-aq-i FUTURE-2.SINGULAR.ERGATIVE-CAUSATIVE-BENEFACTIVE-PATIENT-INTRANSITIVE-open-BENEFACTIVE-3.ABSOLUTIVE 'you will open it for him/her/them' (Hasan Basri: p.c.)

The polysynthesis found in North America is thus neither universal nor unique. There is ample evidence that the degree of synthesis characteristic of a language can change radically over a very short period of reconstructible time. Consider Selayarese, the Austronesian language cited in (11) above. As can be seen from the verb in that example, quite a natural one in the language (Basri: p. c.), the language is polysynthetic. Yet few Austronesianists would reconstruct a highly polysynthetic parent language. Many Austronesian languages are relatively analytic, and in the more synthetic languages, the recent origins of affixes are often transparent. Furthermore, those languages with more complex morphologies do not necessarily show parallel morphological structures, a fact which suggests that their affixes are the result of independent developments. Compare the Selayarese sentence below to a similar expression in Yapese, another Austronesian language, spoken in Micronesia. The Yapese version requires several separate words: (12 a)

Selayarese (Austronesian, Indonesia): la-ku-pap-jamä-9ay-ko FUTURE-l.SING.ERGATIVE-INTRANSITIVE-work-BENEFACTIVE-2.SING.ABSOLUTIVE Ί will work for you.' (Hasan Basri: p. c.)

(12 b)

Yapese (Austronesian, Micronesia): raa gu marweel ni faan ngoom will I work for purpose to-you Ί will work for you.' (Jensen 1977: 277)

The grammatical structures are very different in the two languages, but note that both the tense and the pronominal morphemes are transparently cognate. Further examination of other grammatical markers reveals more cognates, although many are no longer precisely equivalent in function in the two languages. Similar examples of the recent development of polysynthesis can be found in many other parts of the world. In many of these cases, as in

Typology and deep genetic relations in North America

97

Selayarese and Yapese, phonological resemblances between related languages are so transparent that cognates are often easy to spot by naive inspection, while grammatical structures are radically different. Since polysynthesis is neither universal in North America, nor unique to it, and since it can develop in such a short period of time, it appears that polysynthesis alone may not be a reliable indicator of deep genetic relationship.

2. Pronominal affixes Many polysynthetic languages, like most of those cited above, exhibit bound pronominal affixes within verbs. These affixes are not simply subject 'agreement markers'. They are referential pronouns, and refer to all primary arguments of the verb (agents, patients, and/or benefactives; subjects, objects, and/or indirect objects, ergatives, absolutives, and/or datives). Pronominal affixes are typically obligatory, and appear with all verbs whether external nominal phrases are present or not. (13)

Mohawk (Iroquoian, Quebec): wa-honwa-id?t-isäk-ha FACTUAL-MASCULINE.PLURAL. AGENT/MASCULINE.SINGULAR.PATIENT-body-seek-PURPOSIVEPUNCTUAL 'They went to look for him.' (Sonny Edwards: p. c.)

(14)

Mohawk (Iroquoian, Quebec): Sok ki: ronaten:ron so this MASC.PL-RECIP-friend wahatsha.nike? FACTUAL-MASC.SG-fear-PUNCTUAL 'So his friend was afraid...' (Sonny Edwards: p.c.)

The pronominal affixes are obligatory because they are the actual arguments of their verbs. Associated nominals, like the word for 'friend' in (14), are appositives to these affixes, further identifying them, but they are not necessary for grammaticality. The verbs alone constitute complete clauses in themselves. In fact, in languages of this type, clauses very often have no separate nominals at all. Pronominal affixes appear in languages throughout North America, but they are by no means universal. Many languages have only first and

98

Marianne

Mithun

second person pronominal affixes; many others have no pronominal affixes at all, only free pronouns. One such language is Wappo, cited in (7) above. Another is Central Porno, cited below. The free pronouns are underlined. (15)

Central Porno (Pomoan, California): ? άα άα macäats'aq5 cäcPyem müutu I COP I grandfather man-old him

9

9

aa 9üdaaw I a. lot

däa?duw. Meenda müutu meen mul mihdu 9el too care, for so-PROG him so she talked NOM me hcPfial. mouth-hurt 'The man I call my grandfather, I care for him a lot, and her talking that way to him hurt me.' (Frances Jack: p. c.) Pronominal affixes are not only not universal in North America, they are also not limited to North America. Note the pronominal prefixes in the Turkana, Ngandi, and Selayarese examples cited earlier, and in the Ngandi below. (16)

Ngandi (Non-Pama-Nyungan, Australia): ma-mulupinda? -yuij MA-smelly. tortoise-ABSOLUTIVE narma-ncP-rju-ni 1. PLURAL. EXCLUSIVE/MA-still-eat-PAST. CONTIN. 'We used to eat smelly tortoises' (Heath 1978: 193)

There is also ample evidence that pronominal affixes can develop in a language that previously had none. The Austronesian languages cited earlier provide a good example. Selayarese now has a full set of ergative prefixes and absolutive suffix enclitics. (Ergatives refer to the agents of transitive clauses, and absolutives to the patients of transitives or the single arguments of intransitives.) Yapese has developed object suffixes, but no subject affixes. The rapidity with which pronominal affixes can develop is illustrated by a number of Pama-Nyungan languages of Australia. Among even closely related languages, some have only free pronouns, while others have full sets of transitive and intransitive pronominal clitics or affixes. The free form sources of the modern bound forms are still transparent in most cases. Dixon (1980: 246) notes that the development of bound pronouns is concentrated in specific areas of Australia, and "this isogloss

Typology

and deep genetic relations in North America

99

often does not coincide with strong genetic boundaries". He provides numerous examples like that below. Most of the languages in the Yolqu subgroup, from north-east Arnhem Land, lack bound pronominals; but these are found in two languages on the edges of the Yoliju area. Rithargu, the southernmost Yolqu language, has developed pronominal EN-clitics under areal pressure from its southerly neighbours Nunggubuyu and Ngandi (which have pronominal PREfixes); these are transparently reduced forms of the Yolqu independent pronouns. Djinaq, in the northwesterly corner of the Yolqu region, has moved in a different direction, and developed a system of pronominal PROclitics, that can be shown to be historically related — at slightly further remove than the Ritharrju forms — from free-form pronouns in other Yolgu languages; this was again due to areal pressure, from the prefixing languages to the west.

In some cases, the process of fusion can be seen in the earliest stages of its development. Yallop (1977: 46) notes that "the personal pronouns of Alyawarra are independent words", but "certain sequences of words may, through vowel elision and loss of primary stress, become phonetically indistinguishable from a single word." It is interesting that the pronouns may still appear in any order, either proclitic or enclitic to verbs. (17)

Alyawarra (Pama-Nyungan, Australia): Pronoun — Verb athä arika or athärika Ί saw' Verb — Pronoun arika athä or arikatha Ί saw' Pronoun — Pronoun — Verb athä nginha aräyntiya or athänginharäyntiya

'I'll be seeing you'

Verb — Pronoun — Pronoun inpika athä rinha or inpikathirinha Ί got it' (Yallop 1977: 46) Breen (1981) reports on a particularly interesting phenomenon. Margany and Gunya are closely related dialects of a (Pama-Nyungan) Mari language, spoken in Queensland, Australia. They are essentially equivalent structurally, except for one feature: "Gunya has a transparent and obviously recent system of pronominal suffixes to the verb, which Margany lacks" (Breen 1981: 275). Note the pronominal suffixes in the Gunya verb below, for example.

100

(18)

Marianne

Mithun

Gunya (Mari, Pama-Nyungan, Australia): yulbiyiijgiyandana chase-CONTINUING-FUTURE-1. SINGULAR-3. PLURALACCUSATIVE TU hunt them away' (Breen 1981: 331)

Breen points out that the bound forms are highly transparent in origin and variable in use. The singular suffixes are identical to their free counterparts minus the initial syllable (sometimes with vocalic increment).

1 2 2 3 3

Nominative Singular Nominative Singular Accusative Singular Nominative Singular Accusative Singular

Free yaya inda inana nula nuquna

Bound -ya -nda -nana -la -na

Most non-singular free and bound forms are identical. The pronouns have not fused at all in Margany. Compare the two sentences below. (19 a)

Gunya (Mari, Pama-Nyungan, Australia): wacPayiniya go-CONTINUING-PRESENT-1. SINGULAR unayujgiya

(19 b)

lie-CONTINUING = FUTURE-1. SINGULAR 'I'm going to have a sleep' (Breen 1981: 331) Margany (Mari, Pama-Nyungan, Australia): quia 3. SINGULAR

wabataijga go-CONJUNCTIVE-LOCATIVE

ijaya unangu 1. SINGULAR lie-PURPOSIVE 'As soon as he goes I'm going to have a sleep' (Breen 1981: 319) Pronominal affixes are pervasive in North America, but they are neither universal nor unique. Furthermore, they can develop over a relatively short period of time. It thus appears that the presence of pronominal affixes is not a reliable indicator of deeper genetic relationship either.

Typology

and deep genetic relations

in North America

101

3. Incorporation A number of languages throughout North America exhibit the phenomenon called noun incorporation, whereby a noun stem, usually referring to a patient, but sometimes to an instrument or location, is compounded with a verb stem to form a derived verb stem. Examples of incorporation can be seen in the Cree, Takelma, and Mohawk examples cited earlier. It is by no means universal throughout North America, however. Numerous languages show no incorporation at all. Noun incorporation is also not confined to North America. Note the examples cited earlier from Ngandi (Australia) and Sora (India), as well as those below from Hixkaryana (Brazil) and Koryak (Siberia) below. (20)

Hixkaryana (Carib, Brazil): ry-exe-mnuk-yaha 3. SUBJECT/1. OBJECT-throat-press-NONPAST 'he is choking me' (Derbyshire 1985: 226)

(21)

Koryak (Chukotko-Kamchatkan, Siberia): Mi-mitq-antak let. me-blubber-go. after 'Let me go for blubber' (Bogoras 1917)

There is also evidence that noun incorporation can develop over a relatively short period of time. The presence of incorporation is often not consistent within language families. Within the Mayan family, for example, certain languages have no incorporation at all, such as Ixil and Aguacatec. Some others, such as Kanjobal, Mam, and Chuj, have the beginnings of incorporation. A transitive verb and unmodified noun may be juxtaposed to indicate a conceptually unitary activity. Although they remain separate words, the verb and noun form a syntactic unit equivalent to an intransitive verb. The noun, no longer a syntactic argument of the clause, does not refer to a specific, countable entity; it simply narrows the scope of the verb semantically to an activity directed at a certain type of patient. Note that the Kanjobal verb below is grammatically intransitive, as shown by the use of an absolutive case pronoun for 'you'. (22)

Kanjobal (Mayan): s-at_-lo-w-i PAST-2. ABSOLUTIVE-eat-AFFIX-AFFIX 'you bread-ate'

pan bread

102

Marianne

Mithun

Compare:

s-0-a-lo-t-oq

past-3. ABSOLUTIVE-2. ERGATIVE-eat-go-OPT 1. ERG-bread 'you ate my bread' (Robertson 1980) in-pan

In Yucatec, incorporation can function in a similar way, but the incorporated noun is further integrated into the host verb, usually appearing before several suffixes. (23)

Yucatec Mayan: c'ak-ce^-n-ah-en

chop^tree-ANTIPASSIVE-PERFECTIVE-1. ABSOLUTIVE Ί wood-chopped' Compare:

t-in-cak-ah

ce?

COMP-1. ERGATIVE-chop-PERFECTIVE Ί chopped a tree' (Bricker 1978)

tree

The absence of incorporation in many Mayan languages, the transparency of the incorporated structures in the others, and the lack of relic constructions, make it clear that incorporation is a new development within this family, rather than an earlier trait that is decaying. (For more on the nature, development, and distribution of noun incorporation see Mithun 1984.) Since noun incorporation is neither universal nor unique to North America, and since it can develop relatively rapidly over time, it cannot be assumed to be a reliable indicator of deep genetic relationship either.

4. The predominance of verbs over nouns The predominance of verbs in natural discourse is a striking feature of many North American languages. The passage below, from a Mohawk legend, is quite typical. Three men went fishing, but had no luck. Finally, one went off by himself. (All verbs are underlined.) (24)

Mohawk (Iroquoian, Quebec): Khiahatkähtho?

kintsion

ih

latiiä.ti,

there he looked fish there they body be in 'He looked and saw fishes in there,

Typology and deep genetic relations

in North America

103

kentsi.io iä.kerP. fish were nice it seems they were nice fish, they say. Sok wahonnehlä.ko tsi nikentsi.io. so it unset him so fish is so nice He was surprised to see how nice the fish were. Sok wä.thlahkwe so he picked up So he picked them up, wahshakoia?l0:ko. he body removed from water he scooped them up. Sok ionsahähawe9 tsi ηόη: thön:ne>s so he took back to where they are there So then he took them back to where ki: lonnaten:ro?s tehniiähsen thonatehalä:ton. this they are friends they are two they are waiting there his two friends were waiting. A:kP wahshakohlo.li tsi na:hotela?swi:io'?ste9 Oh he told them so for his luck to become so good Oh, he told them how lucky he was wahotsiaien: ta?ne9. he fish got to catch fish.' (Sonny Edwards: p. c.) The predominance of verbs is of course not unrelated to the three other features mentioned above. Mohawk is highly polysynthetic, but the productive morphological complexity is centered within verbs. This is typical of polysynthetic languages. Mohawk nouns may be morphologically complex, but not nearly so complex as verbs. Not only are tense and aspect specified on verbs, but a variety of adverbial distinctions are made as well. Consider the morphological structure of the first word. kh-i-a-h-at-kähtho COINCIDENT-TRANSLOCATIVE-AORIST-MASCULINE. SINGULAR. AGENT-SEMI.REFLEXIVE-look-PUNCTUAL 'he looked' ('all of a sudden he set his sight there')

104

Marianne

Mithun

The coincident prefix kh- signals that the event was sudden and unexpected, the translocative i- indicates that the action was directed away from the agent, the aorist a- signals aorist tense, the semi-reflexive -atsignals middle voice, and the punctual signals punctual aspect. Verbs can carry considerable information, and the lexicon is organized to take advantage of this power. Many concepts that are rendered in IndoEuropean languages by nouns, adjectives, or adverbs are expressed by verbs in Mohawk. The pronominal prefixes present in every verb eliminate the need for separate pronouns. Each verb is a grammatical predication in itself. The incorporated nouns eliminate the need for many free noun phrases. In the first line, the noun 'fish' is free, because it represents significant new information. Thereafter, however, the noun root for 'fish', -(i)tsi-, is incorporated as in the verbs kentsi.io, 'it-fish-nice, is', nikentsi.io 'soit-fish-nice. is', and wahotsiaien:taPne9 'he-fish-got'. The notion 'to have good luck' is considered a unitary concept, so it is conveyed by a single complex verb stem: the noun -IcPsw- 'luck' is incorporated into the root -iio 'good, nice': n-a:-ho-te-lcPsw-i:io-?-st-e? PARTITIVE-OPTATIVE-MASCULINE.SINGULAR.PATIENT-SEMI.REFLEXIVE-luck-good-INCHOATIVECAUSATIVE-PUNCTUAL 'his luck would become so good'. Finally, in Mohawk, as in many polysynthetic languages, morphological verbs function syntactically not only as predicates and clauses, but also as nominals, without additional derivation or special marking. Note the word for 'friends'. ronn-at-en.ro-^s MASCULINE. PLURAL. PATIENT-SEMI. REFLEXIVE-be. friend-DISTRIBUTIVE 'they are friends to each other' > 'friends' Morphologically, this is a well-formed stative verb. It is more often used to identify than to predicate, however: it is nearly lexicalized as a syntactic nominal. The deictic particle ki: 'this/these' preceding it in the text above strengthens the expectation that the function of the phrase will be identification. Although many North American languages do show a preponderance of verbs like that above, many others do not. Note the Wappo and

Typology and deep genetic relations in North America

105

Central Pomo passages cited above, and the Nisenan below. All are California languages, from different language families, but Wappo and Central Pomo were classified by Sapir as Hokan, and Nisenan as Penutian. (The suffix glossed NOM is the nominative case, ACC the accusative. Verbs are again underlined. Nominalization serves to subordinate, and thus form appositives.) (25)

Nisenan (Maiduan, California): kapa-im k'ut'-im ktle-im bear-NOM deer-NOM woman-NOM 'Bear and Deer Woman, 9

epe:-ki-to-im sister.in.law-relate-RECIPROCAL-NOM sisters-in-law, ht:-k'oy-mukum c'i:w-i. pick-away-REMOTE.PAST clover-ACC went to pick clover long ago. ha-inhay k'ut'-im ktle-im and-say deer-NOM woman-NOM And, it is said, Deer Woman, lay-pe-im hi:ni-im mana:-im child-have-NOM small-NOM boy-NOM leaving her two small boys

pe:n-i two-ACC

hPis-ti-im house-stay-CAUS-NOM at home, ht:-k 'oy-mukum. pick-away-REMOTE.PAST went to gather clover. mi-im ha-inhay mi-im soka-di and-say the-NOM meadow-LOC the-NOM And Bear, having killed kapa-im mi-im k'ut'-im kile-i bear-NOM the-NOM Deer-NOM woman-ACC Deer Woman

106

Marianne

Mithun

w0:n-ti-im die-CAUSATIVE-NOM in the meadow, mik'i c'o:l-i wolo-na bo-mit'-in its head-ACC basket-LOC throw-into putting her head in the packing basket c'i:w-ni wa.-t'a-in clover-INSTR put-onto and covering it up with clover, to.daw-mukum. carry-reach-REMOTE.PAST brought it back.' (Smith 1977: 132-133) A predominance of verbs is pervasive in North America, but it is neither universal nor unique. It appears that this, in itself, is also not a reliable indicator of deep genetic relationship.

5. Conclusion Polysynthesis, pronominal affixes, incorporation, and a predominance of verbs are thus neither universal nor unique among American languages, although it is easy to understand why earlier scholars supposed they were. At the beginning of the nineteenth century, language families such as Algonquian and Iroquoian were already extensively documented and familiar to scholars. These languages do indeed share many of the structural features thought to characterize the continent as a whole. Western languages, by contrast, show considerably more variety. In fact, twentytwo of Powell's original fifty-eight families were in California, but significantly less was known about them at the time. Furthermore, although well-educated philologists were often familiar with a variety of languages, these were usually exclusively Indo-European, occasionally enriched by Semitic or Chinese. It is not surprising that the structures found in North America should have seemed dramatic. Even if these structures were universal, they could not be relied on as indicators of deep genetic relationship, because they can develop so rapidly. As the Selayarese, Alyawarra, Gunya, and Jacaltec examples show, they can spring into existence while phonology and lexicon remain relatively unchanged.

Typology and deep genetic relations in North America

107

Of course this does not mean that languages sharing these structural features are necessarily unrelated. More interestingly, it is not even clear that these characteristics are meaningless for language classification. It could be the case, for example, that related languages can jointly inherit other structural features that create a predisposition for the development of these. Polysynthesis, bound pronouns, incorporation, and verbiness co-occur sufficiently frequently to suggest that they should not be counted as independent features. They cannot be considered reliable indicators of deep genetic relationship, however, either singly or as a group, until more is understood about their borrowability and interrelationships. The development of pronominal affixes in Australia indicates that at least this structural feature is borrowable. It could be the case that the acquisition of a characteristic like bound pronouns facilitates the development of the others. In any case, it is clear that our understanding of genetic, areal, and typological relationships have much to gain from each other, and much to lose if carried on in isolation. References Bogoras, Waldemar 1917 Koryak texts (American Ethnological Society Publications 5) (Leiden: Brill). Breen, J. G. 1980 "Margany and Gunya", in: R. M. W. Dixon - Barry J. Blake (eds.), Handbook of Australian languages 2 (Canberra: The Australian National University Press), 2 7 5 - 3 9 4 . Bricker, Victoria 1978 "Antipassive constructions in Yucatec Maya", in: Nora C. England (ed.), Papers in Maya Linguistics (Columbia: University of Missouri), 3 — 24. Brinton, Daniel Garrison 1891 The American race (New York: N. D. C. Hodges). Derbyshire, Desmond C. 1985 Hixkaryana and linguistic typology (Summer Institute of Linguistics Publication 76) (Arlington, Texas: SIL and University of Texas). Dimmendaal, Gerrit Jan 1983 The Turkana language (Dordrecht, Holland: Foris). Dixon, R. M.W. 1980 The languages of Australia (Cambridge: Cambridge University Press). Duponceau, Peter Stephan 1838 Memoire sur le systeme grammatical des langues de quelques nations indiennes de l'Amerique du nord (Paris: A Pihan de la Forest). Fortescue, Michael 1984 West Greenlandic (London: Croom Helm). Goddard, Pliny Earle 1911 "Athapascan (Hupa)", in: Franz Boas (ed.), Handbook of American Indian languages (Bureau of American Ethnology Bulletin 40.1) (Washington: Government Printing Office), 85 — 158.

108

Marianne

Mithun

Haas, Mary R. 1969 "Grammar or lexicon? The American Indian side of the question from Duponceau to Powell", IJAL 35: 2 3 9 - 2 5 5 . Heath, Jeffrey 1978 Ngandi grammar, texts, and dictionary (Canberra: Australian Institute of Aboriginal Studies). Mithun, Marianne 1984 "The evolution of noun incorporation", Language 60: 847 — 894. Jensen, John Thayer 1977 Yapese reference grammar (Honolulu: University Press of Hawaii). Powell, John Wesley 1891 Indian linguistic families of America north of Mexico (Seventh Annual Report, Bureau of American Ethnology) (Washington, D. C.: Government Printing Office), 1 - 1 4 2 . Ramamurti, G.V. 1931 A manual of the So.ra: (or Savara) language (Madras: Government Press). Robertson, John S. 1980 The structure of pronoun incorporation in the Mayan verbal complex (New York: Garland). Sapir, Edward 1922 "The Takelma language of southwestern Oregon", in: Franz Boas (ed.), Handbook of American Indian languages (Bureau of American Ethnology Bulletin 40.1) (Washington: Government Printing Office), 1—296. 1929 "Central and North American Languages", in: Encyclopaedia Britannica14 (London and New York: Encyclopaedia Britannica Company), 5: 138 — 141 [reprinted in: David G. Mandelbaum (ed.), Selected writings of Edward Sapir, (Berkeley —Los Angeles: University of California Press, 1951), 169 — 178], Sawyer, Jesse O. —Laura Fish Somersal 1977 "Bear Woman and her children", in: Victor Golla —Shirley Silver (eds.), Northern California texts (IJAL-NATS 2.2), 1 0 5 - 1 1 3 . Smith, Richard 1977 "Bear and Deer Woman", in: Victor G o l l a - S h i r l e y Silver (eds.), Northern California texts (IJAL-NATS 2.2), 1 3 2 - 1 4 6 . Tuggy, David H. 1979 "Tetelcingo Nahuatl", in: Ronald W. Langacker (ed.), Modern Aztec grammatical sketches (Summer Institute of Linguistics Publications in Linguistics 56.2) (Arlington: SIL and University of Texas), 1 — 140. Whitney, William Dwight 1898 Language and the study of language^ (New York: Scribner). Wolfart, H. Christoph 1973 Plains Cree: A grammatical study (Transactions of the American Philosophical Society, New Series 63.5). Yallop, Colin 1977 Alyawarra: An Aboriginal language of central Australia (Canberra: Australian Institute of Aboriginal Studies).

Specific papers on aspects of Indo-European reconstruction

Armenian, Tocharian, and the "glottalic" theory Werner Winter

0. The fact that Winfred P. Lehmann has taken a very lively interest in the so-called "glottalic" theory, and the further fact that two protagonists of this theory, Tamaz Gamkrelidze and Paul J. Hopper, are among us today, leads me to deviate from my original, somewhat broader, plan and to discuss in some detail data from Armenian and Tocharian that have an immediate bearing upon aspects of the glottalic theory. 1. Before the advent of this theory, it used to be generally accepted for a fact that the Armenian consonants matching the stops of Greek owed their particular characteristics to a soundshift comparable to that presumably incurred in Germanic, with Greek and its peers retaining a more original state of affairs. Now, however, it is assumed that Armenian preserves a system certainly more archaic than that of Greek, but possibly even more conservative than that of Germanic. Gamkrelidze — Ivanov (1984: 41), who put forward this claim, cite in its support the following data: Some Modern Armenian dialects show voiceless glottalized stops in the series reflecting (in Brugmannian terms) Proto-Indo-European voiced stops; Gamkrelidze and Ivanov take this as an indication that the voiceless glottalized stops (series I) of Proto-Indo-European were preserved as such in Old Armenian. The further fact that again some Modern Armenian dialects have slightly aspirated voiced stops is viewed as pointing to a possible identical representation of Proto-Indo-European voiced aspirated stops (series II) in Armenian (Gamkrelidze — Ivanov 1984: 42); they furthermore propose that the complementary distribution found in some Armenian dialects of voiced aspirates in initial vs. voiced nonaspirates in noninitial position might have preserved an allophonic pattern proper to those Indo-European dialects of which Armenian formed part (Gamkrelidze — Ivanov 1984: 43). Finally, they assume that Armenian voiceless aspirated stops are the regular reflexes of the sounds in series III (Brugmann's voiceless unaspirated stops), except that *ph became h or zero initially, and by way of */, w intervocalically; for intervocalic *th, they posit a developmental chain *θ > *h > 0.

112

Werner

Winter

2. A number of comments are called for. I will first raise some questions concerning method. To be sure, a mere inspection of the table of changes outlined by Gamkrelidze — Ivanov (1984: 45) yields the impression of a striking near-identity of the ProtoIndo-European and Old Armenian systems posited. However, this identity is the result of a number of choices made when other decisions had been equally possible. Thus, both Arm. t (plain) and Arm. t' (globalized) are available as matches for PIE t' (Brugmannian *d), both Arm. d (plain) and Arm. dh (aspirated) are candidates for comparison with PIE d Vr an initial *y may be reconstructed. Internally, Arm. t' and k' are found in retained or former postconsonantal position (a special case will be discussed in 8.2.1); after resonants, Arm. d and g occur as variants, always with the preceding consonant preserved. Intervocalically, PIE *p *t *kw/k yield Arm. w y y. 5. Figure 1 will show the various developments of PIE *p *t *kw/k incurred: 6. Figure 1 shows quite clearly a number of characteristic properties. For one, there is a high degree of merger: all reflexes of preconsonantal PIE *p *t *kw merge, both in initial position (here evidence for *kw is lacking), where 0 is the outcome, and in the interior of forms, where Arm. w is found. Likewise, there are identical reflexes of prevocalic *p *t *kw in the sets which show consistently h or y, respectively (the latter again not clearly attested for *kw). A more limited merger occurs in intervocalic position: here *p yields Arm. w, while *t and *kw result in Arm. y. Secondly, there is considerable divergence in the reflexes of *ρ *t *kw in identical environments. Thus, a contrast Arm. ρ' φ h φ y is found in initial prevocalic position for PIE *p; for PIE *t, we have under the same conditions Arm. t' φ h φ y. The problem cannot be solved by simply disregarding one or the other manifestation of the representation of ProtoIndo-European voiceless stops; thus, the assertion by Gamkrelidze —Ivanov (1984: 43) that the development of PIE *p through */ to h and 0 had something to do with general properties of the labial and was to be interpreted in analogy to Celtic phenomena, loses all weight once the parallels in the development of PIE *t and *kw are taken into consideration. We have genuine multiple representation under identical syntagmatic conditions; as for an interpretation of this state of affairs, I see no way but to acknowledge the fact that Classical Armenian preserves more than one code, which means that more than one dialect served as the basis for the koine of Grabar (cf. Winter 1966).

Armenian,

PIE input

Environment

Tocharian, and the "glottalic"

theory

Armenian output

*P

*t

*kw/k


*d-, but rather a replacement of aspiration by voice, meaning that a feature set including + consonantal, + aspirated, ( + lenis) was changed to -(-consonantal, -I-voiced, (+lenis). The development *dh > *dj— Vgwh then did not involve reduction of the marking in a privative opposition, but a replacement of one marking in an equipollent opposition by another marking. The Tocharian dissimilation thus would be different from what

Armenian,

Tocharian, and the "glottalic"

theory

125

seem at first glance parallel developments in Greek and Indo-Iranian: here we would have a reduction- in degree of markedness. While thus a special reflex of the voiced dental can be observed in Tocharian (B A ts- in normal environment, Β A s- before palatalizing vowel, 0 before resonant), no such special treatment of the voiced stops can be seen in the labial, velar, and labiovelar sets. To take just the labials: *b is safely attested in Β tapre, Α tpär 'high' (cf. Goth, diups) and, with a different ablaut grade, in Β taupe, A top 'mine'; *p in Β päcer, Α päcar 'father'; *bh \η Β procer, A pracar 'brother'. There is no evidence that before resonant there was a loss of *b *g *gw; this leaves open the question why *d should have been so unstable in position before w, y, r, m — I find it impossible to follow the reasoning of Lehmann (1986): if glottalization had anything to do with it, one would want evidence for instability of other consonants of series I — one cannot, in my view, base an argument solely on the dental stop, the more so as here other explanations are available that would take us too far afield now. 12. The Tocharian system then has, not surprisingly, nothing to offer in favor of, or against, the glottalic theory. The theory will have to be accepted, or rejected, on other grounds. As for myself, I find the argument by Jakobson (1957: 528), which served as a trigger, rather weak: if the voiced aspirates were not distinctively voiced, the typological reasoning loses all its weight — a pattern plain Φ voiced Φ aspirated is perfectly natural. In my view it is crucial that the plainness of the tenues should not be tampered with since only by keeping them unmarked (in the Trubetzkoyan, not the Jakobsonian, sense of the term) is it possible to account for clustering behavior in a number of Indo-European languages (cf. Winter 1959). Whether voicedness, a common and, in the context sketched, most natural feature, or glottalization is to be prefered, will have to be decided on the strength of arguments internal to the reconstruction of Proto-Indo-European; thus, I find Lubotsky's proposal very intriguing indeed, while what has been said against PIE *b strikes me as exceedingly weak. The gaps in the admissible root types in Proto-IndoEuropean (cf., e. g., Hopper 1982: 128) need not be due to a single cause since plain and voiced consonants belong to one assimilation class, aspirates to another. The frequency arguments (Hopper 1982: 128 — 129) suffer on the one hand from the notion that aspirates were doubly marked as against singly-marked voiced stops, on the other from all-too-easy dismissal of what I consider clear cases of PIE *b. Still, even admitting some forms with *b would not necessarily eliminate the argument in

126

Werner Winter

favor of a glottalic rather than a voiced series, even though this may just have to be ascribed to a remote pre-Proto-Indo-European past; from my point of view it is decisive that *t should be taken to be the unmarked member of the set of stops and that both the aspirates and the voiced stops (or glottalized stops) should be singly marked. In this sense I find the Revised System in Hopper (1982: 126) more attractive than its formulation in Gamkrelidze — Ivanov (1984) since here the introduction of aspirated variants on an equal footing with unaspirated ones opens up the possibility of a system with no unmarked members. That the danger is real is indicated by the treatment of the reflexes of series III in Armenian by Gamkrelidze and Ivanov: here they really operate with a marked Proto-Indo-European source form. I hope I have shown the inherent weakness of their analysis at this point at least. References Bolognesi, G. 1954 "Ricerche sulla fonetica armena", Ricerche Linguistiche 3: 123 — 154. Gamkrelidze, Tamaz V. —Ivanov, Vjaceslav V. 1984 Indoevropejskij jazyk i indoevropejcy [Indo-European and Indo-Europeans] 1 — 2 (Tbilisi: Izdatel'stvo Tbilisskogo Universiteta). Godel, Robert 1975 An introduction to the study of Classical Armenian (Wiesbaden: Reichert). Hopper, Paul J. 1982 "Areal typology and the Early Indo-European consonant system", in: Edgar C. Polome (ed.), The Indo-Europeans in the fourth and third millennia (Ann Arbor: Karoma), 121-139. Hübschmann, Heinrich 1897 Armenische Grammatik 1: Armenische Etymologie (Leipzig) [reprint: Darmstadt: Wissenschaftliche Buchgesellschaft, 1962], Jakobson, Roman 1958 "Typological studies and their contribution to Historical Comparative Linguistics", in: Eva Sivertsen (ed.), Proceedings of the Eighth International Congress of Linguists (Oslo: Oslo University Press), 17 — 25. Krause, Wolfgang — Thomas, Werner 1960 Tocharisches Elementarbuch I: Grammatik (Heidelberg: Winter). Lehmann, Winfred P. 1986 "Reflexes of PIE d < t"\ in: Dieter Kastovsky — Aleksander Szwedek, Linguistics across historical and geographical boundaries (Berlin: Mouton de Gruyter), 4 8 3 - 4 8 9 . Meillet, Antoine 1936 Esquisse d'une grammaire comparee de l'arminien classique2 (Vienne: Mekhitaristes). Pokorny, Julius 1959 Indogermanisches etymologisches Wörterbuch (Bern/München: Francke).

Armenian,

Tocharian, and the "glottalic"

theory

127

Solta, Georg Renatus 1960 Die Stellung des Armenischen im Kreise der indogermanischen Sprachen (Wien: Mechitharisten-Buchdruckerei). Szemerenyi, Oswald 1960 Studies in the Indo-European system of numerals (Heidelberg: Winter). 1977 "Studies in the kinship terminology of the Indo-European languages with special reference to Indian, Iranian, Greek, and Latin", Acta Iranica (Textes et memoires 7: Varia), 1 —240. Van Windekens, Albert Joris 1976 Le tokharien confronte avec les autres langues indo-europeennes 1: La phonetique et le vocabulaire (Louvain: Centre International de Diabetologie Generale). Winter, Werner 1954 "Problems in Armenian phonology I", Language 30: 197 — 201. 1955 "Problems in Armenian phonology II", Language 31: 4 — 8. 1959 "Über eine Methode zum Nachweis struktureller Relevanz von Oppositionen distinktiver Merkmale", Phonetica 4 Suppl.: 28—44. 1962 a "Problems in Armenian phonology III", Language 38: 254—262. 1962 b "Die Vertretung indogermanischer Dentale im Tocharischen", Indogermanische Forschungen 67: 16 — 35. 1965 "Armenian evidence", in: Werner Winter (ed.), Evidence for laryngeals (The Hague —London —Paris: Mouton), 100—115. 1966 "Traces of early dialectal diversity in Old Armenian", in: Henrik Birnbaum—Jean Puhvel (eds.), Ancient Indo-European dialects (Berkeley — Los Angeles: University of California Press), 201—211. 1970 "Some widespread Indo-European titles", in: George Cardona —Henry M. Hoenigswald —Alfred Senn (eds.), Indo-European and Indo-Europeans (Philadelphia: University of Pennsylvania Press), 49 — 54. 1986 "Hayereni anvanakan t'ek'man mi k'ani hare'er [Nekotorye aspekty armjanskoj imennoj fleksii] [Some aspects of Armenian nominal inflection], Patmabanasirakan handes / Istoriko-filologiceskij zurnal 1986.2: 17 — 24.

Reconstructing morphology: the role of o-grade in Hittite and Tocharian verb inflection Jay H. Jasanoff

The state of our knowledge in a historical discipline depends not only on the primary evidence at our disposal but also on the manner and order in which it was acquired. Historical linguistics is no exception; it is a truism that our current views of Proto-Indo-European might be very different if the classical language encountered by the first Western travellers to India had been Hittite and the language uncovered in the Bogazköy archives had been Vedic Sanskrit. In a sense, the history of Indo-European studies in the twentieth century can be seen as an extended attempt to accommodate the Neogrammarian model of the parent language to the data that became available with the discovery of Anatolian and Tocharian. The most celebrated work of this kind has been in phonology, where the unexplained appearance of h as a segmental phoneme in Hittite was the immediate stimulus for Kurylowicz' revival of the laryngeal theory in 1927 and the elaboration of new solutions to an impressive series of problems over the following decades. The significance of Hittite and Tocharian for the study of PIE morphology has been less dramatic but no less profound. Cases come readily to mind in which the testimony of these languages has led to completely new analyses; thus, to take a well-known example, the -r of middle forms like Lat. sequitur 'follows' and Olr. sechithir 'id.', which Brugmann and Meillet regarded as a shared innovation of Italic and Celtic, is now known from forms like Hitt. artari 'stands' and Toch. Β klyaustär 'is heard' to be a genuine archaism. Not all cases, however, are so clear-cut. In many other instances, Hittite and Tocharian fail to conform to the received image of what an older Indo-European language should be, but it is not at all obvious whether this is due to innovation or conservatism. In such situations the weight of tradition can take on undue importance, for Indo-Europeanists are understandably reluctant to rethink the history of Sanskrit, Greek, or Latin for the sake of what may seem marginal simplifications in the description of Hittite and Tocharian. The danger, of course, is that by failing to undertake periodic re-examinations of our underlying assumptions we exclude a priori the possibility of recognizing

130

Jay Η. Jasanoff

important archaisms of Anatolian and Tocharian that happen to have been lost from the other Indo-European languages. A case of this kind is explored below. The Proto-Indo-European verb system, as reconstructed by Brugmann and his successors, included a number of athematic formations with inner-paradigmatic ablaut. Most of these oppose a "strong" stem with egrade to a "weak" stem with zero-grade; the former is found chiefly in the active singular, the latter in the active dual and plural and throughout the middle. The presents of *h/es- 'be' and *ieug- 'yoke, join' are typical: 3 sg. *h,es-ti (Ved. ästi) and *iu-ne-g-ti (Ved. yunäkti) contrast with 3 pi. *h,s-inti (Ved. sänti) and *iu-n-g-enti. Paradigms with a strong vocalism other than *-e- are rare. In particular, alternations involving o-grade, which will be the focus of the discussion that follows, are traditionally assumed for only a single primary category, the perfect (cf. Gk. μέμονε 'wishes', pi. μεμάασι < *memon-e, pi. *memn-er; Ved. veda 'knows', pi. vidüh < *uoid-e, pi. *uid-ir). Otherwise *-o- is found chiefly in secondary, non-ablauting formations, such as the iterative-causatives in *-eie/o- (cf. Gk. φοβέω Ί put to flight', Go. lagjan 'lay', etc.) and the iteratives in *-eh2iejo- ("*-ö/e/o-"; cf. Gk. ποτάομαι Ί fly about', Go. hwarbon 'walk', etc.). The rarity of o-vocalism in the oldest strata of the verbal system is in part responsible for the prevalence of the view, maintained, e. g., by Kurylowicz (1956: 41 ff.) and Cowgill (1979: 33 ff.), that the perfect is of nominal origin. 1 To be sure, the uniqueness of the perfect has occasionally been called into question. Athematic presents with o-grade were posited by Meillet (1916), who assumed a three-way alternation of *-o-, *-e- and *-0- to account for forms like Goth, malan 'grind', Lith. mälti 'id.' ( < *molh2-) beside Olr. melid 'grinds', OCS meljetü 'id.' ( < *melh2-) and Arm. malem Ί grind', MW malaf 'id.' (